i

TEXTBOOK OF CANCER
EPIDEMIOLOGY
ii
iii

THIRD EDITION

TEXTBOOK OF
CANCER
EPIDEMIOLOGY

EDITED BY

H A N S -​O L O V   A D A M I
D AV I D J .   H U N T E R
PA G O N A   L A G I O U
LORELEI MUCCI

With an introductory c­ hapter by

BRIAN MACMAHON

1
iv

1
Oxford University Press is a department of the University of Oxford. It furthers
the University’s objective of excellence in research, scholarship, and education
by publishing worldwide. Oxford is a registered trade mark of Oxford University
Press in the UK and certain other countries.

Published in the United States of America by Oxford University Press

198 Madison Avenue, New York, NY 10016, United States of America.

© Oxford University Press 2018

First Edition published in 2002
Second Edition published in 2008
Third Edition published in 2018

All rights reserved. No part of this publication may be reproduced, stored in

a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by license, or under terms agreed with the appropriate reproduction
rights organization. Inquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above.

You must not circulate this work in any other form

and you must impose this same condition on any acquirer.

Library of Congress Cataloging-in-Publication Data

Names: Adami, Hans-Olov, editor. | Hunter, David J. (David John), editor. |
Lagiou, Pagona, editor. | Mucci, Lorelei A., editor.
Title: Textbook of cancer epidemiology / edited by Hans-Olov Adami, David J. Hunter,
Pagona Lagiou, Lorelei Mucci ; with an introductory chapter by Brian MacMahon.
Description: Third edition. | Oxford ; New York : Oxford University Press, [2018] |
Includes bibliographical references and index.
Identifiers: LCCN 2017029789 | ISBN 9780190676827 (hardcover : alk. paper)
Subjects: | MESH: Neoplasms—epidemiology
Classification: LCC RA645.C3 | NLM QZ 220.1 | DDC 614.5/9994—dc23
LC record available at https://lccn.loc.gov/2017029789

1 3 5 7 9 8 6 4 2
Printed by Sheridan Books, Inc., United States of America
v

We dedicate this 3rd edition to

Dimitrios Trichopoulos, MD
our close friend, long-​term collaborator, and brilliant role model
vi
vii

CONTENTS

Preface to the Third Edition ix 9. Esophageal Cancer 183

Preface to the Second Edition xi Christian C. Abnet, Olof Nyrén,
Preface to the First Edition xiii and Hans-​Olov Adami
About the Editors xv
10. Stomach Cancer 213
Contributors xvii
Weimin Ye, Olof Nyrén,
and Hans-​Olov Adami
PART I  11. Colorectal Cancer 243
1. Accomplishments in Cancer Epidemiology 3 Nana Keum, Kana Wu,
Edward Giovannucci,
Brian MacMahon
and David J. Hunter
2. Measures and Estimates of Cancer Burden 31
12. Cancer of the Liver and Biliary Tract 277
Paolo Boffetta, Dana Hashim,
Christina Bamia, Sherri Stuver,
and Pagona Lagiou
and Lorelei Mucci
3. The Origins of Cancer 57
13. Pancreatic Cancer 309
Jeffrey Ecsedy and David J. Hunter
Vassiliki Benetou, Anders Ekbom,
4. Genetic Epidemiology of Cancer 81 and Lorelei Mucci
Christopher Haiman and 14. Cancer of the Lung, Larynx,
David J. Hunter and Pleura 327
5. Biomarkers in Cancer Epidemiology 101 Jyoti Malhotra, Paolo Boffetta,
Paolo Boffetta and Lorelei Mucci and Lorelei Mucci
6. Concepts in Cancer Epidemiology 15. Skin Cancer 355
and Etiology 111 Adèle C. Green, Catherine M. Olsen,
Pagona Lagiou, and David J. Hunter
Dimitrios Trichopoulos, 16. Breast Cancer 381
and Hans-​Olov Adami
Rulla Tamimi, Susan Hankinson,
and Pagona Lagiou
PART II  17. Cervical Cancer 421
Karin Sundström, Joakim Dillner,
7. Oral and Pharyngeal Cancer 137 and Hans-​Olov Adami
Jennifer R. Rider, Paul Brennan, 18. Endometrial Cancer 441
and Pagona Lagiou
Marta Crous-​Bou,
8. Nasopharyngeal Cancer 159 Immaculata De Vivo,
Ellen T. Chang and Hans-​Olov Adami and Pagona Lagiou
vii

viii Contents

19. Ovarian Cancer 459 25. Thyroid Cancer 609

Penelope M. Webb, Susan J. Jordan, Carlo La Vecchia,
and David J. Hunter Cristina Bosetti,
20. Prostate Cancer 481 and Hans-​Olov Adami
Kathryn M. Wilson 26. Hodgkin Lymphoma 627
and Lorelei Mucci Henrik Hjalgrim, Mads Melbye,
21. Testicular Cancer 525 and Pagona Lagiou
Lorenzo Richiardi, 27. Non-​Hodgkin Lymphoma 649
Giovenale Moirano, Karin Ekström Smedby,
and Pagona Lagiou Mads  Melbye,
22. Urinary Bladder Cancer 543 and Hans-​Olov Adami
Manolis Kogevinas, Jonine Figueroa, 28. Leukemias 673
Montserrat Garcia-​Closas, Eve Roman, Alexandra Smith,
and Lorelei Mucci and Lorelei Mucci
23. Kidney Cancer 571
Xuehong Zhang, Eunyoung Cho, Epilogue 691
and Hans-​Olov Adami Index 695
24. Brain Cancer 587
Dominique Michaud, David Savitz,
and Lorelei Mucci
ix

P R E FA C E T O
THIRD EDITION

When we began planning this third edition of We are in debt to all colleagues who joined
our Textbook of Cancer Epidemiology, Dimitrios us in the effort to produce this Textbook, as well
Trichopoulos, coeditor of the first two editions, as coauthors of chapters in prior editions who
was still among us. As always, his knowledge, did not participate this time. But we are partic-
wisdom, and engagement were priceless, and ularly grateful to Kristina Glimsjö. Her unsur-
our planning for the book was indeed advanced passable competence, exemplary coordination of
when he suddenly left us in December 2014. the work, and meticulous technical editing of all
Subsequently, we slowly recovered from the enor- chapters, made the timely finishing of this third
mous loss, formed a new team of editors, and edition possible. We also acknowledge expert
decided to dedicate this edition to our admired assistance from Jill McDonald.
friend. Since our first edition was sent to print some
Similar to the second edition, this edition 15 years ago, the knowledge base in cancer epide-
has been extensively revised and updated thanks miology has grown enormously. The continued
to dedicated work by outstanding lead authors, steady, but slow, progress in reducing population
many joining us in this endeavor for the first time. cancer mortality rates through treatments only
Also similar to the second edition, we have main- emphasizes the importance of improvement in
tained the original structure, we have kept Brian primary prevention through discovery and elim-
MacMahon’s historical perspective on our disci- ination of the causes of cancer. To achieve this—​
pline and the “Concepts in Cancer Epidemiology when the global burden of this malady keeps
and Etiology” chapter written with the seminal growing—​requires a struggle on two fronts. One
contribution of Dimitrios Trichopoulos, and we is to effectively translate what we already know
have maintained the highly standardized format into effective, population-​based prevention pro-
of the site-​specific chapters (Part II). grams. The other is to embrace the many remain-
Because identification of genomic determi- ing enigmas and unknowns in the causation of
nants of cancer risk is such a rapidly expanding cancer in humans. We hope this book will stimu-
field—​ with continuously updated information late its readers to accept the challenges and con-
available on the Web—​we have restricted this tinue the endless search for new discoveries.
information in the site-​specific chapters with a
focus on genomic variants that appear mechanis- Stockholm and Boston H.O.A.
tically intriguing. Instead, we have expanded the Oxford and Boston D.H.
description of principles of genotyping as well as Athens and Boston P.L.
of analyses and interpretation of data emanating Boston L.M.
from the rapidly growing and increasingly com-
plex genetic field. This is dealt with in the general
chapter on genetic epidemiology.
x
xi

P R E FA C E T O
SECOND EDITION

Little more than 5  years have elapsed since our This contribution provides a historical perspec-
Textbook of Cancer Epidemiology was first sent tive on the evolution of cancer epidemiology.
to the print. As we hoped, and indeed predicted, We hope this text will be particularly useful to
the progress in our discipline has been substan- younger colleagues who entered our field of sci-
tial during these years. We therefore felt a grow- entific inquiry recently. We have also added a
ing need to update the text now, when the first new chapter on the nasopharyngeal carcinoma.
edition has been printed three times. Without Although largely unknown in Western countries,
changing the overall disposition of the book or this malignancy is endemic and has considerable
the uniform structure of site-​specific chapters, public health consequences in some parts of Asia.
this second edition has been thoroughly revised. We are in debt to all colleagues who joined us
We achieved this goal in many instances by invit- in the effort to produce this second edition. But
ing a new colleague to join the authors of the first we are particularly grateful to Kristina Glimsjö.
edition. As a corollary, many chapters are now Without her unsurpassable competence, exem-
written by three rather than two authors, always plary coordination of the work and meticulous
with deep involvement of the editors. technical editing of all chapters, it is hard to
Beside a thorough update—​often with con- imagine how a timely finishing of this second
siderable expansion of genetic and molecular edition would have been possible.
epidemiology—​ the second edition has some
novel features too. Most importantly, Brian Boston and Stockholm H.O.A
MacMahon, a towering Nestor in epidemiology, Boston D.H.
has written an extensive introductory chapter. Boston and Athens D.T.
xii
xii

P R E FA C E T O
FIRST EDITION

We believe there are several reasons that famil- and causal reasoning or any expertise in medi-
iarity with cancer epidemiology will become cine, oncology, or biology.
increasingly important for many health profes- Part 1 gives an introduction to basic concepts
sionals besides those working in the traditional in epidemiology, a description of the global bur-
realm of epidemiology and public health. First, in den of cancer, definitions and characterizations
the history of medicine, primary prevention has of the various measures used, and approaches
often, sooner or later, turned out to be the most used to reveal genetic determinants of cancer risk
successful and most cost-​ effective—​ sometimes and integrate biologic markers in the epidemio-
the only possible—​approach to disease control. logic research process. Part 2 is devoted to spe-
Water sanitation and vaccination programs have cific cancer sites or types, each chapter starting
saved more lives than antibiotics and intensive with a brief synopsis that introduces the clinical
care. Second, now that the human genome has characteristics of the specific cancer. The struc-
been sequenced, a gigantic challenge will be to ture of these chapters is uniform. They all have
understand how genes, environment, and life- three main sections:  Descriptive Epidemiology,
style interact in the causation of human cancer. Genetic and Molecular Epidemiology, and Risk
Epidemiologic approaches are needed to achieve Factors. In order to facilitate reading, risk factors
this understanding. Third, therapeutic break- are discussed in the same order throughout all
throughs in oncology have certainly been impor- these chapters.
tant but few. If real progress ultimately occurs, as Admittedly, this short textbook does not
we all hope, it is not obvious that novel treatments cover all sites and types of cancer. Rare malignan-
will be available, or affordable, to the majority of cies such as sarcomas have been omitted, as have
the world’s population. some endocrine tumors, childhood cancers—​
Hence, primary prevention of cancer will accounting for less than 1% of the total—​and cer-
likely be an increasingly important and integral tain anogenital cancers. It is worth emphasizing,
part of any health professional’s duties. The first though, that little is known about the etiology of
step in preventing any disease is to understand the forms of cancer that we excluded.
its causes and the proportion of cases due to each Other volumes on cancer epidemiology are
cause. This is the background against which this available—​some larger and more detailed than
book was written. It is intended for students of ours, such as the important book by Schottenfeld
epidemiology, medicine, public health, biology, and Fraumeni (1996), and others written by
and the behavioral sciences, and for practitio- distinguished colleagues who have chosen
ners of medicine, public health, and other health approaches different from those we adopted.
professions. We assume no deep familiarity with A range of book choices is beneficial to prospec-
epidemiologic theory, study design, biostatistics, tive readers, who will be able to choose on the
xvi

xiv Preface to First Edition

basis of their particular objectives, orientation, in linking complex human characteristics and
or style. behavior, often during many decades of life, to
Most of what we know about the causes of the occurrence of cancer. We hope that the evi-
cancer in humans has been generated by epide- dence put together in this book will leave the
miologic research. Nevertheless, readers of this reader better equipped to follow the expansion
textbook may find the gaps in knowledge, the of cancer epidemiology during the years to come.
ambiguities, and the abundant contradictory
findings frustrating. But a tidier picture would Stockholm H.O.A.
be unrealistic, given the enormous challenges Boston D.H.
for the young discipline of cancer epidemiology Boston and Athens D.T.
xv

ABOUT THE EDITORS

Hans-​Olov Adami, MD, is Emeritus Professor Pagona Lagiou, MD, is Professor of Hygiene
of Cancer Epidemiology and former Chairman and Epidemiology at the School of Medicine,
of the Department of Medical Epidemiology and National and Kapodistrian University of Athens,
Biostatistics at Karolinska Institutet, Stockholm, Greece, and Adjunct Professor of Epidemiology
Sweden, as well as Professor of Epidemiology at the Harvard T.H. Chan School of Public
and former Chairman of the Department of Health, Boston, United States.
Epidemiology at the Harvard T.H. Chan School
of Public Health, Boston, United States. He is Lorelei Mucci, ScD, is Associate Professor
Professor at the Clinical Effectiveness Research of Epidemiology at the Harvard T.H. Chan
Group, University of Oslo, Oslo, Norway. School of Public Health, and Leader of Cancer
Epidemiology, Dana-Farber/Harvard Cancer
David J.  Hunter, MBBS, is the Richard Doll Center, Boston, United States.
Professor of Epidemiology and Medicine at the
University of Oxford, and Vincent L.  Gregory
Professor of Cancer Prevention Emeritus at the
Harvard T.H. Chan School of Public Health,
Boston, United States.
xvi
xvi

CONTRIBUTORS

Christian Abnet, PhD, MPH Ellen T. Chang, ScD

Metabolic Epidemiology Branch Center for Health Sciences
Division of Cancer Epidemiology and Genetics Exponent Inc.
National Cancer Institute Menlo Park, California
Bethesda, Maryland United States
United States
Eunyoung Cho, ScD
Christina Bamia, PhD Department of Dermatology
Department of Hygiene, Epidemiology, and The Warren Alpert Medical School
Medical Statistics Brown University
School of Medicine Providence, Rhode Island
National and Kapodistrian University of Athens United States
Athens, Greece
Marta Crous-​Bou, PhD
Vassiliki Benetou, MD Department of Medicine
Department of Hygiene, Epidemiology, and Harvard Medical School
Medical Statistics Boston, Massachusetts
School of Medicine United States
National and Kapodistrian University of Athens
Immaculata De Vivo, PhD
Athens, Greece
Department of Medicine
Paolo Boffetta, MD Harvard Medical School
Tisch Cancer Institute Boston, Massachusetts
Icahn School of Medicine at Mount Sinai United States
New York, New York
Joakim Dillner, MD
United States
Department of Laboratory Medicine
Cristina Bosetti, PhD Karolinska Institutet
Department of Epidemiology Stockholm, Sweden
IRCCS
Jeffrey Ecsedy, PhD
Instituto di Ricerche Farmacologiche
Translational Medicine
“Mario Negri”
Millenium Pharmaceuticals Inc.
Milan, Italy
Cambridge, Massachusetts
Paul Brennan, PhD United States
Genetic Epidemiology Group
Anders Ekbom, MD
International Agency for Research on Cancer
Clinical Epidemiology Unit
Lyon, France
Department of Medicine
Karolinksa University Hospital
Stockholm, Sweden
xvi

xviii Contributors

Jonine Figueroa, MPH, PhD Manolis Kogevinas, MD

Usher Institute of Population Health Sciences Centre for Research in Environmental
and Informatics Epidemiology (CREAL)
Edinburgh Cancer Research Centre Municipal Institute of Medical Research (IMIM)
University of Edinburgh Barcelona, Spain
United Kingdom
Carlo La Vecchia, MD
Montserrat Garcia-​Closas, MD, PhD, MPH Department of Clinical Sciences and
Division of Cancer Epidemiology and Genetics Community Health
National Cancer Institute Università degli Studi di Milano
Rockville, Maryland Milan, Italy
United States
Jyoti Malhotra
Edward Giovannucci, ScD Rutgers Cancer Institute of New Jersey
Department of Nutrition and Epidemiology New Brunswick, New Jersey
Harvard T.H. Chan School of Public Health United States
Boston, Massachusetts
Mads Melbye, MD
United States
Department of Epidemiology Research
Adèle C. Green, MB, BS Statens Serum Institut
QIMR Berghofer Medical Research Institute Copenhagen, Denmark
Brisbane, Queensland
Dominique S. Michaud, ScD
Australia
Department of Public Health and Community
Christopher Haiman, ScD Medicine
Department of Preventive Medicine Tufts University Medical School
University of Southern California Boston, Massachusetts
Los Angeles, California United States
United States
Giovenale Moirano
Susan Hankinson, ScD Unit of Cancer Epidemiology
Department of Biostatistics and Epidemiology Department of Medical Sciences
University of Massachusetts, Amherst University of Turin
Amherst, Massachusetts Turin, Italy
United States
Olof Nyrén, MD
Dana Hashim, MD Department of Medical Epidemiology and
Department of Preventive Medicine Biostatistics
Icahn School of Medicine at Mount Sinai Karolinksa Institutet
New York, New York Stockholm, Sweden
United States
Catherine M. Olsen, PhD
Henrik Hjalgrim, MD QIMR Berghofer Medical Research Institute
Department of Epidemiology Research Brisbane, Queensland
Statens Serum Institut Australia
Copenhagen, Denmark
Lorenzo Richiardi, MD, PhD
Susan J. Jordan, MB, BS, PhD Unit of Cancer Epidemiology
QIMR Berghofer Medical Research Institute Department of Medical Sciences
School of Public Health University of Turin
University of Queensland Turin, Italy
Brisbane, Queensland
Jennifer R. Rider, ScD, MPH
Australia
Department of Epidemiology
Nana Keum, ScD Boston University School of Public Health
Department of Nutrition Boston, Massachusetts
Harvard T.H. Chan School of Public Health United States
Boston, Massachusetts
United States
xxi

Contributors xix

Eve Roman, PhD Rulla Tamimi, ScD
Department of Health Sciences Channing Division of Network Medicine
Epidemiology & Cancer Statistics Group Brigham and Women’s Hospital
University of York Harvard Medical School
York, United Kingdom Boston, Massachusetts
United States
David Savitz, PhD
Department of Epidemiology Penelope M. Webb, PhD
School of Public Health QIMR Berghofer Medical Research Institute &
Brown University School of Public Health
Providence, Rhode Island University of Queensland
United States Brisbane, Queensland, Australia
Karin Ekström Smedby, MD Kathryn M. Wilson, SCD
Department of Medicine Department of Epidemiology
Solna, Clinical Epidemiology Unit Harvard T.H Chan School of Public Health
Karolinska Institutet and Hematology Center Boston, Massachusetts, United States
Stockholm, Sweden
Kana Wu, MD, PhD
Alexandra Smith, PhD Department of Nutrition
Department of Health Sciences Harvard T.H. Chan School of Public Health
Epidemiology & Cancer Statistics Group Boston, Massachusetts
University of York United States
York, United Kingdom
Weimin Ye, PhD
Sherri Stuver, ScD Department of Medical Epidemiology and
Department of Epidemiology Biostatistics
Boston University School of Public Health Karolinksa Institutet
Boston, Massachusetts Stockholm, Sweden
United States
Xuehong Zhang, MD, ScD
Karin Sundström, MD Department of Medicine
Department of Laboratory Medicine Harvard Medical School
Karolinska Institutet Brigham and Women’s Hospital
Stockholm, Sweden Boston, Massachusetts
United States
xx
1

PA R T I
2
 1
Accomplishments in Cancer Epidemiology
BRIAN MACMAHON t

The proper study of Mankind is Man

(ALEXANDER POPE, An Essay on Man)

K nowledge of the environmental causes of

human cancer has come principally from
study of the human experience. The roles of
Necessarily, the subject matter of this chapter
cuts across that of the or:ganJ and site-specific
chapters that follow, but a review of accomplish­
epidemiology have been to describe the dis- ments necessitates an approach that is more his­
tribution of cancer in populations and to seek torical than state of he art. From a historical
deviations from randomness that offer clues perspective, tbe classification based on exposures
as to their explanations, to assess the validity followed i this chapter seems more functional
and strength of associations that are suspected than one based on organs and sites as followed
causes, and to evaluate the effectiveness of pre- in later chap,ters, although, as with all classifica­
ventive measures by continued monitoring of tions, one classification is in fact no more "natu­
cancer incidence and mortality. Experiments- i.- al" than the other.
on laboratory animals, inferences from t e In this chapter, risks reported in the original
natural world outside man, biochemistry, ana as "odds ratios;' "standardized mortality ratios;'
microbiology have all played importamt roles, or "hazard ratios" are identified as relative risks
but observation of human beings has often sug- or risk ratios (R) when they are approximately
gested lines of investigation and is essential in equivalent.
evaluating the efficacy of preve tive measures.
In this chapter, we review er:f demiologic evi­ THE CIGARETTE EPIDEMIC
dence on cigarette smoking, ionizi g radiation, In the early years of the 20th century cancer of
occupation, the physical and microbiologic the lung was a rare disease: one hundred years
environments, th reproductive experience of later it is the most frequent fatal cancer in most
women, alcohol, and iatrogenic exposures as of the western world, accounting for 174,000
they bear on cancer risk in humans. In addi­ deaths annually in the United States alone (Jemal
tion to accomplishments that identified specific et al, 2006). As shown in Chapter 2, in the United
cancer causes, we note some situations in which States the epidemic struck early in the 20th cen­
epidemiologic studies provided reassurance tury, peaked during the 1990s, and then began
that an agent suspected of being carcinogenic slowly to decline. Between 1930 and 1990, the epi­
has been demonstrated not to be so under the demic was responsible for about 4 million deaths
circumstance of human exposures: such find­ from lung cancer in the United States alone. In
ings are also accomplishments. many parts of the world the epidemic has not yet
Had this chapter been written in 1930, barely peaked and promises to extend well into the 21st
75 years ago, the list of accomplishments would century. Epidemiology's place in cancer research
have been modest. To summarize the progress was established by its role in investigating this
that has been made in these 75 years requires epidemic: the discipline provided descriptive
economy of language and information-not to statistics to show that the epidemic existed, sug­
mention certain arrogance. We can present only gested its cause, proved the identity of that cause,
in outline the main epidemiologic discoveries. and evaluated progress toward its control.
4

4 Textbook of Cancer Epidemiology

By 1930 several hypotheses were being offered were reported from the study of the ACS cohort
to explain the increasing lung cancer rate: these (Thun et al, 1995).
included the introduction of cigarettes, the use of Cigarette smoking has been associated with
tar to pave roads and air pollution from industry several malignancies other than lung cancer
and coal fires in homes. In 1940 Muller reported (Doll et al, 2005). These are of less concern, either
a case-​control study implicating cigarette smok- because they are less common or because it has
ing, and after a hiatus occupied by a world war not been established that their associations with
this was followed by a handful of studies from tobacco are causal. They include cancers of the
Britain and the United States, of which the most mouth, oral pharynx, larynx, esophagus, pan-
influential was that of Doll and Hill (1950). They creas, liver and bladder, and possibly kidney and
established the extraordinarily high risk of lung myeloid leukemia (Boyle, 2005). Also relevant in
cancer associated with cigarette smoking: smok- considering the impact of the cigarette epidemic
ing of pipes and cigars were also shown to be is the synergistic role that cigarette smoking plays
associated with lung cancer but with risks much with other known carcinogens such as asbestos,
lower than that for cigarettes. Clearly, cigarette nickel, and radon.
smoking was the cause of the epidemic:  it had In an illustration of the use of epidemiology
increased rapidly since the beginning of the cen- in the evaluation of preventive measures, Rodu
tury and its risks were much higher than those and Cole (2007), noting that the proportion of
associated with any other suspected factor. In Americans who smoke had declined by 50% since
1964, the Surgeon General of the United States 1965, estimated the number of deaths attributable
recognized cigarette smoking as a cause of lung to smoking in 1987 and 2002:  between the two
cancer. years smoking-​attributable deaths had declined
While cigarettes were clearly shown to be by 41% in males and 30% in females. It is likely
causes of lung cancer in case-​control studies, that about half of this decline was attributable to
further evidence was needed for its general reduction of deaths from malignant disease.
acceptance. Cohort (follow-​up) studies are not
undertaken lightly:  they require that data on I O N I Z I N G R A D I AT I O N
exposure be obtained on a large number of indi- Ionizing radiation (IR) is unique among carcino-
viduals (the cohort) and that a mechanism be gens in the variety of tissues with which it comes
available for ascertaining the occurrence of dis- in contact and in which it can induce neoplasia.
ease among the cohort members. Evidence from Almost all our knowledge of its carcinogenic risks
case-​control studies was sufficient to stimulate in humans comes from epidemiologic studies.
two cohort studies: one based on British physi-
cians (Doll & Hill, 1954) and one on the exten- Skin Cancer
sive network of volunteers for the American The earliest observations of cancer following IR
Cancer Society (ACS) (Hammond, 1966). The were of skin cancers among pioneers in the dis-
characteristics of the tobacco–​lung cancer asso- covery and investigation of radium and x-​rays.
ciation were confirmed and refined in terms The exposures were very large and irrelevant to
of amount and duration of smoking, effects current practices in the handling of radioactive
of smoking-​cessation, age, gender, and other material, but skin cancers are still being reported
variables. in survivors of the A-​bombs, after radiation ther-
The cohort studies also identified that tobacco apy for benign skin conditions, and in a few occu-
causes deaths from many other diseases than lung pations. Clinical observations rather than formal
cancer. The cohort of physicians, defined in 1951, epidemiologic studies are responsible for most of
was followed to its 50th anniversary in 2001, by these reports, and estimates of risk are lacking.
which time 74% of the enrolled men had died
(Doll et  al, 2004). The age-​adjusted death rate Leukemia
for smokers was almost double that of lifelong Increased risk of leukemia associated with IR
nonsmokers, but only part of the excess could be is the best documented of its neoplastic effects.
attributed to lung cancer: 14% were attributed to Two bodies of data contributed importantly to
lung cancer, 24% to ischemic heart disease, 21% the characterization of the relationship—​patients
to other cardiovascular disease, and 16% to non- with ankylosing spondylitis treated with x-​rays
malignant respiratory disease, mostly chronic between 1935 and 1954 (Smith & Doll, 1982) and
obstructive lung disease. Similar associations the survivors of the atomic bombings in Nagasaki
5

Accomplishments in Cancer Epidemiology 5

and Hiroshima (Little et al, 2004). They showed a Cancer of the Lung

linear relationship between level of exposure and The earliest known epidemic attributable to IR—​
risk of leukemia from the lowest levels for which though not recognized as such until late in its
reliable estimates can be made to the very high course—​was the “mountain sickness” that killed
levels that cause cell death. thousands of workers in the mineral mines of
The role of age at exposure to IR came into Saxony and Czechoslovakia between the 17th
prominence in 1958 with a study describing a and 20th centuries. Lorenz (1944) recounts the
doubling of risk of leukemia and solid tumors history of these mines and the unique illness
in childhood among children exposed to diag- associated with them. Härting and Hesse recog-
nostic x-​rays in utero (Stewart et al, 1958): many nized the disease as lung cancer in 1879 but it was
studies followed. Levels of IR exposure involved not attributed to the presence of radon gas until
in pelvimetry (the most common reason for pel- the 20th century, when it developed among ura-
vic x-​ray) are uncertain, but they were consider- nium miners in other areas.
ably lower than the levels at which carcinogenesis Particularly well documented is the experi-
had been observed or predicted from studies of ence of uranium miners of the Colorado Plateau
adults. A  consensus emerged that the associa- of the western United States, but large studies
tion probably existed, though at lower levels of have also been undertaken and are continuing in
risk than suggested in the 1958 study: the possi- several other countries. An analysis of the pooled
bility of it being attributable to social or medical data from 11 studies showed a linear relation-
confounding factors calls for caution in its inter- ship between lung cancer risk and accumulated
pretation. Because x-​rays have been replaced by underground mining experience (Lubin et  al,
other methods of pelvic and fetal imaging, this 1995). The latent period between first exposure
controversy will probably never be resolved, but and death from cancer averaged 25 years (Samet,
it played an important role in discouraging use of 1989): this fact and the disregard of the European
x-​rays in pregnancy and in developing alternative experience were major factors in the failure to
methods of accomplishing its objectives. recognize this epidemic in the United States at an
earlier stage (Archer et al, 2004).
Cancer of the Breast Several authors state that uranium miners
Increased rates of breast cancer have been are heavy smokers, but among nonsmoking ura-
observed among women receiving multiple nium miners the risk of lung cancer associated
fluoroscopies in the treatment of tuberculo- with the highest level of cumulative exposure was
sis; radiation treatment for benign skin lesions, 30 times as high as for those exposed at the low-
enlargement of the thymus, or cancer of the est levels (Gilliland et  al, 2000). However, there
cervix; and in women employed as radiology is clearly an interaction of some form between
technicians in the early years or as painters of the two variables: differences between studies in
watch dials using radium for its fluorescence. age at first exposure and length of follow-​up may
With regard to cancer of the breast IR has a dual explain differences between studies in sugges-
effect: exposure of the ovary to levels sufficient to tions as to whether the interaction is additive or
ablate ovarian function decreases breast cancer more or less so (Archer, 1986).
risk, but direct exposure of the breast increases There is no evidence of increased risk of any
it. Patients with cervix cancer whose IR exposure neoplasm other than lung cancer among ura-
was sufficient to ablate the ovaries had reduced nium miners (Darby et al, 1995).
breast cancer risk, but women receiving 3.1 Gy to
the breast had cancer risk three times as high as Cancer of the Thyroid
those receiving no radiation (Boice et al, 1989). The thyroid gland is particularly sensitive to IR in
Among women with Hodgkin’s disease, those childhood. In A-​bomb survivors, the risk of thy-
receiving the highest radiation exposures to the roid cancer is limited to persons under 20 years of
breast had an eightfold increase in breast cancer age at the time of exposure (see ­chapter 25): most
incidence (Travis et  al, 2003). Importantly, for of the other exposures that have been associated
young women treated with substantial radiation with increased risk have also involved young per-
to the breast, risk increased with age:  cumula- sons, including patients treated with IR for hem-
tive breast cancer frequencies at ages 35, 45, and angiomas, enlarged thymus glands, or Hodgkin’s
55, were 1%, 7%, and 29%, respectively (Travis disease. Concern for risk of thyroid cancer arose
et al, 2005). in the use of iodine-​ 131 in the treatment of
6

6 Textbook of Cancer Epidemiology

thyrotoxicosis, but follow-​up studies have been (Cologne et  al, 1999), and bladder (in A-​bomb
null. The epidemiologic studies of this topic are survivors); cancers of the stomach and kidney (in
reviewed by UNSCEAR (2000). patients with ankylosing spondylitis and radium
dial painters); and cancers of the brain and ovary
Thorotrast (in patients treated with IR for benign condi-
Thorotrast, an x-​ray contrast medium introduced tions). The potential of IR to induce neoplasia in
in 1928, was used extensively in Germany, where most human tissues is well documented.
it was developed, in some other European coun-
tries and in Japan. An account of the develop- C A N C E R I N   O C C U PAT I O N A L
ment of thorotrast, its uses, and the early stages SETTINGS
of understanding its complications is given by Many cancer risks in occupations were first noted
Abbatt (1979). The medium consisted of tho- in reports from physicians or others familiar with
rium dioxide and thorium-​232 and was injected working conditions, the seminal example being
intravenously or intra-​arterially:  its slow elimi- Percival Pott’s 1775 description of scrotal can-
nation from the body and its radioactive half-​life cer in chimney sweeps:  associations of cancer
of 400  years ensured that recipients would be with arsenic, asbestos, chemical dyes, chromium,
exposed to alpha radiation for the remainder of nickel, and vinyl chloride also came under sus-
their lives. It had extensive use in visualizing vas- picion in this way. Epidemiologic studies aimed
cular systems and exploring traumatic injuries. to show that such observations represented more
Abbatt estimates that sufficient thorotrast was than coincidence, estimate the size of the risks,
produced for use in at least 2.5 million, and prob- and locate the risks to particular sections of the
ably 10 million, examinations. workplace or work processes. We note here only
The first cases of angiosarcoma of the liver the numerically most important occupational
(ASL) in patients receiving thorotrast were carcinogens:  a few other suspected carcinogens
reported in the 1940s. There have been follow-​ that have received some attention in occupational
up studies totaling over 5,000 patients (van Kaick settings, such as dioxins, DDT, cadmium, and
et al, 1999; Mori et al, 1999; Nyberg et al, 2002; beryllium, are not included since the evidence of
dos Santos Silva, 2003). They showed exceedingly their occupational carcinogenicity is either null
high incidence and mortality rates for cancer of or inconclusive.
the liver (around 40 times expected) and myelog-
enous leukemia (5–​10 times expected). All types Acrylonitrile
of liver tumor were involved, but because of the Acrylonitrile (AN) is used in the manufacture of
rarity of ASL, the relative risk (R) for this tumor acrylic fibers and as a fumigant:  exposure is by
was larger than for other types. Risks of both liver inhalation of the volatile liquid. Beginning in the
cancer and leukemia increased with the volume early 1980s, reports of cases of cancer in exposed
of thorotrast injected and with the time elapsed workers multiplied. A  1998 meta-​ analysis
after exposure: excess risks remained high more included 25 epidemiologic studies and suggested
than 40  years after injection (dos Santos Silva no increased risk associated with AN for cancer
et al, 2003). as a whole or for any specific neoplasm (Collins
Epidemiology documented the extent of the & Acquavella, 1998). A subsequent follow-​up of
thorotrast tragedy, but it was an exercise in lock- workers potentially exposed to AN in eight US
ing the barn door after the horses had fled. The industries showed no increase in mortality for
potential of internalized radium and its daugh- any neoplasm except cancer of the lung, which
ters to produce fatal hematologic disorders was somewhat elevated (1.4) only in the high-
and bone cancer were well known by the 1940s est quintile of cumulative exposure (Blair et  al,
(Martland et  al, 1925). The rapid accumulation 1998). A small case-​control study gave R for lung
of case reports (IAEA, 1965) gave ample warning cancer of 2.9:  the association is weaker than in
of the toxic properties of thorotrast well before that in the larger US study and contrary to the
epidemiologic studies were under way. The tail of bulk of evidence, and chance variation cannot be
this epidemic has not yet passed. excluded (Scélo et al, 2004).
Epidemiologic studies are reassuring with
Other Cancers regard to the carcinogenicity of AN, limiting
Other cancers seen in persons heavily exposed concern to lung cancer, and then to the highest
to IR include those of the brain, stomach, liver exposures experienced historically.
7

Accomplishments in Cancer Epidemiology 7

Arsenic numbers were reported on another cohort from

Occupational exposure to arsenic is primarily by the same source (Selikoff et al, 1965).
inhalation of dust. Many metal ores are contami- Subsequently, in studies of asbestos work-
nated with arsenic, and excess lung cancer has ers in many countries and different occupations
been found among miners or smelters of gold, the association with lung cancer was confirmed
tin, and copper. A  50-​year follow-​up of work- unequivocally, an exposure–​ risk relationship
ers in a Montana smelter found a 50% excess of described, the induction period estimated, and
deaths from lung cancer that was highest after the risks associated with specific types of asbes-
time spent in areas with medium or high arsenic tos clarified:  among the three major types of
exposure:  among workers with 10 or more asbestos, crocidolite conveys the highest cancer
years work in such areas, observed deaths were risk and chrysotile the lowest. The association
three and four times, respectively, the number with stomach cancer has been found in other
expected (Lubin et al, 2000). Other cancers were studies, although not all, and usually at levels
not in excess. of risk lower than that noted by Selikoff, and
Exposure to arsenic compounds also occurs in small excesses of cancers of the esophagus and
the manufacture and use of wood preservatives, kidney have also been reported (Enterline et al,
insecticides, and herbicides. The many reports 1987; Armstrong et  al, 1988; Raffn et  al, 1989;
of illness associated with these activities include McDonald et al, 1993).
cancers of the lung and other sites (International Mesothelioma, a highly malignant cancer of
Agency for Research on Cancer (IARC), 1980), the pleura and peritoneum, was extremely rare
but the effects of arsenic have not been clearly prior to 1900, but during the next century rap-
differentiated from those of other known and idly became a signature for asbestos exposure.
suspected carcinogens in these compounds. The first epidemiologic study was of cases from
asbestos fields in South Africa (Wagner et  al,
Asbestos 1960). Reports from North America and Europe
The name “asbestos” refers to a group of fibrous followed. For six countries in Western Europe,
chemicals that can be spun into thread and annual deaths from mesothelioma increased
woven into fabrics that are heat-​resistant, light- from virtually zero at the beginning of the cen-
weight, and low cost. These “miracle fibers” tury to around 2,500 in 1998 (Peto et al, 1999). In
were used in a wide variety of products during the United States during 1973 to 2001, rates for
the 20th century. Discovery of the dark sides of males increased sharply until the mid-​1990s but
asbestos was a major epidemiologic accomplish- thereafter began to decline (Price & Ware, 2004).
ment of the second half of the century: it led to As with lung cancer, crocidolite carries the
a total ban on the use of asbestos in Britain and highest risk of mesothelioma, and from a review
other European countries and heavy regulation of data from 71 studies Yarborough (2006) con-
and the prospect of a total ban in the United cluded there is little evidence that chrysotile,
States (Bartrip, 2004). unless contaminated with other fibers, causes
Three major health problems are associated mesothelioma. As with lung cancer, latent peri-
with asbestos:  pulmonary fibrosis and emphy- ods between first exposure and appearance of
sema (asbestosis), lung cancer, and mesotheli- disease are long—​ often many decades (Peto
oma. While our concern here is only with the last et al, 1982).
two, asbestosis was an important cause of mor-
bidity and mortality and a major force in the call Benzene
for control of the material. Benzene is a solvent and starting material for the
The first epidemiologic study of asbestos with manufacture of other chemicals. It is also added
cancer was of workers in a small British asbes- to gasoline, of which it constitutes about 2%. The
tos plant (Doll, 1955). Over a period of 35 years, hemotoxicity of benzene has been known since
11 died of lung cancer whereas 0.8 would have the 19th century, but demonstration of its role in
been expected. A  larger cohort was assembled leukemia is credited to Aksoy et al (1974), who
from members of a New York Asbestos Workers reported 26 cases of leukemia among 28,000
Union. Observed and expected deaths from lung leather workers in Istanbul and estimated the
cancer were 45 and 6.6, from stomach cancer “incidence” at 14/​ 105, which they believed to
12 and 4.3, and from cancers of the colon and be a “significant” increase over that in the gen-
rectum 17 and 5.4 (Selikoff et  al, 1964). Larger eral population. Shnatter et  al (2005) reviewed
8

8 Textbook of Cancer Epidemiology

22 published studies and concluded that there occupational cancers. The authors estimated
was an increase in risk of acute myeloid leuke- that by the time of the study bladder cancer had
mia (AML) with benzene exposure and a positive occurred in 10% of the exposed population, and
relationship between level of exposure and risk. that, even assuming no further exposure after
The risks were evident in the rubber, leather, and 1951, given the estimated incubation periods, an
paint industries. Data on chronic myeloid leu- additional 10% could be expected.
kemia and lymphoid leukemia were found to be
sparse and inconclusive. Relative risk for work- Chromium
ers at the highest levels of cumulative exposure The most common forms of chromium are tri-
have been around 10 (Yin et al, 1987; Glass et al, valent (Cr-​3) and hexavalent (Cr-​6), of which
2003), with earlier studies showing even higher Cr-​3 is the more stable. Cr-​6 has powerful oxi-
risks (Rinsky et al, 1981). The exposure–​risk rela- dative properties and is thought to be most likely
tionship was evaluated by Hayes et  al (2001) in responsible for the metal’s carcinogenic potential.
a combination of information from two large In many studies the exposure is simply referred
cohorts, one from the United States and one to as “chromium.”
from China:  there was progressive increase in Occupational exposure to chromium is pri-
risk of myeloid leukemia and its precursors with marily through inhalation of contaminated dust.
increasing cumulative benzene exposures over a Its association with lung cancer has been sus-
10-​fold range of exposure. pected for more than a century. Meta-​analyses
based on 49 epidemiologic studies published
Chemical Dyes since 1950 showed an excess mortality (R 1.4)
In the late 19th and early 20th centuries there from lung cancer among chromium-​ exposed
were many clinical reports of what came to be workers that became lower (1.1) when the anal-
called “aniline tumors of the bladder” in work- ysis was limited to studies of better quality and
ers engaged in the manufacture and use of chem- adjusted for smoking (Cole & Rodu, 2005). For
ical dyes. The progenitor of many retrospective stomach cancer in these analyses R was also
cohort studies in occupations was the “field sur- slightly elevated (1.1) but was lower (0.8) on
vey” of tumors of the urinary bladder conducted adjustment for socioeconomic status. No other
by Case et  al (1954) in the British chemical cancer showed higher-​than-​expected frequency.
industry. Among workers employed since 1920 Increase in risk of lung cancer with increase in
for more than 6  months in the manufacture of exposure has been noted in several studies, with
aniline, benzidene, alpha-​naphthylamine (ANA), risks in the highest categories of exposure being
or beta-​naphthylamine (BNA), 341 cases of blad- 2–​3 times those in the lowest exposure groups
der cancer occurred, 298 (87%) of which were (Park et al, 2004).
in workers who had had contact with one of the Clearly, risk of lung cancer is increased with
chemicals:  127 of the latter died of the disease, exposure to occupational chromium, but even at
whereas four would have been expected based the highest levels of exposure, chromium is only
on national mortality data (R 32). There was no a weak carcinogen.
external source of data from which the expected
number of incidence cases could be computed. Formaldehyde
Risk was highest (R 86)  after exposure to Formaldehyde has been used as a disinfectant and
BNA—​ among the highest risks ever reported preservative in industry, embalming, and medi-
for an occupational exposure; intermediate risks cine. Human exposure is primarily by inhalation
were associated with ANA and benzidene (45 and of the gas, though skin absorption is possible with
62, respectively): ironically, there was no increase the use of the liquid formalin. Formaldehyde was
in risk among workers exposed only to aniline, reviewed by a working group of IARC (2004b)
although numbers were small. Risk increased and judged “sufficient” to place the chemical in
with duration of exposure, but some increase its Group  1 (Carcinogenic to humans). But with
was seen after exposures of less than 1 year. Some the possible exceptions of nasopharyngeal can-
tumors appeared within 2 years of first exposure, cers (NPCs) and leukemia, the evidence of carci-
but others after more than 45  years:  the aver- nogenicity in humans is weak.
age interval between first exposure and death Elevated risk of NPC has been found in many
was 18  years, presaging the long latent periods studies: in the largest cohort study, R in the most
that came to be recognized as characteristic of highly exposed subgroup was 1.7, based on
9

Accomplishments in Cancer Epidemiology 9

six cases (Hauptmann et  al, 2004). Collins et  al evaluation of nickel exposures, and it is clear that
(1997), in a summary of 47 epidemiologic stud- the risk associated with nickel is much greater
ies of formaldehyde-​ exposed workers, found in smokers than in nonsmokers (Andersen et al,
essentially no increased risk for any type of can- 1996; Sorahan & Williams, 2005).
cer other than leukemia. Leukemia mortality was
found somewhat elevated in six of seven cohorts Polychlorinated Biphenyls
of persons exposed to formaldehyde in embalm- Because of their persistence in the environment
ing or medical occupations (IARC, 2004b). Two and biological fluids, polychlorinated byphenyls
recent US industrial cohort studies showed (PCBs) have been suspected of producing a vari-
small excesses of leukemia, particularly myeloid ety of health effects, including cancer. Reports
leukemia, in the most heavily exposed groups in the last two decades have raised suspicions
(Hauptmann et al, 2003; Pinkerton et al, 2004). about association with specific neoplasms, but
no clear pattern emerged. Bosetti et  al (2003b)
Nickel reviewed six cohort studies of workers occupa-
The main uses of nickel are as alloys in stain- tionally exposed to PCBs, finding no excess can-
less steel and in the manufacture of batteries, cer overall (0.9) and no excess mortality for any
but the principal studies of its health effects are specific cancer. In 2006 the US National Institute
of workers in mining and processing. Bradford for Occupational Health and Safety updated the
Hill (1939) reported the first epidemiologic study follow-​up of three cohorts of occupationally
of workers in a Welsh nickel refinery:  among exposed workers assembled earlier (Prince et al,
workers during the years 1929–​38, there were 2006; Ruder et al, 2006). (For analysis, data from
16 deaths from lung cancer when 1 would have two of the cohorts were combined.) Deaths from
been expected and 19 from cancer of the nose cancer overall were not elevated (0.8 and 1.0).
when the expected was less than 1.  There was Modest increases were seen for a few sites, but
no excess in deaths from other cancers. In a later numbers were very small and there was no con-
study in the same refinery, there were 48 deaths sistency, either between the cohorts or with sites
from lung cancer (R 5)  and 13 from nasal can- suspected on the basis of earlier studies, and
cer (R 150)  (Doll, 1958). Comparable results there was no convincing association with esti-
were obtained from early studies in refineries mated levels of exposure.
in Canada (Mastromatteo, 1967)  and Norway Studies of cancer following PCB expo-
(Pedersen et  al, 1973). In later studies from the sure are reassuring as to the chemicals’ lack of
Welsh and Norwegian refineries smaller excess carcinogenicity.
risks were found (Doll et al, 1977; Grimsrud et al,
2003). Grimsrud and Peto (2006) summarized Vinyl Chloride
the experience of the Welsh refinery, compar- Vinyl chloride (VC) is a gas that is polymerized to
ing the experience of workers first employed in produce polyvinyl chloride—​a widely used plas-
three periods:  1902–​19, 1920–​29, and 1930–​69. tic. Cleaning polymerization chambers exposes
Relative risk for lung cancer in the three periods workers to residual VC monomer. Suspicion of
was 6.2, 0.7, and 1.4, and for nasal cancer 380, 73, the carcinogenicity of VC was raised in 1974 by
and 9, respectively—​the last figure (for nasal can- a report of ASL—​a very rare tumor—​in exposed
cer) being based on only two deaths. The analysis workers. Within two years 40 such cases had
indicates that nickel is a potent carcinogen, and been reported (Schlatter, 1976). In a literature
also illustrates the role of epidemiologic studies review in 1999, McLaughlin and Lipworth (1999)
in evaluating the efficacy of control measures. concluded that, with the exception of ASL, no
Exposures in nickel refineries include other adverse outcomes (including cancers) could be
agents besides nickel, some of which are also car- conclusively linked to VC exposure. Bosetti et al
cinogens: they include arsenic, sulfuric acid mists, (2003a) undertook an analysis of data from two
cobalt, and sometimes asbestos. In an analysis of meta-​analyses of epidemiologic studies in Europe
data from the Norwegian refinery in which indi- (Mundt et al, 2000) and North America (Ward
vidual assessments of these exposures were made, et al, 2001): in total, there were data on more
estimates of R for nickel exposure changed only than 22,000 workers from 56 plants. There were
slightly when adjusted for these variables alone 71 deaths from ASL in the cohorts: there were
or in combination (Grimsrud et al, 2005). Several no external rates of this malignancy for compar-
authors have adjusted for smoking habits in the ison, but the tumor is rare. Deaths from cancer
10

10 Textbook of Cancer Epidemiology

of the lung and larynx were lower than expected, basal and squamous cell carcinoma and mela-
and mortality from soft tissue sarcoma, brain, noma of the skin cancer. Some evidence relat-
and hematopoietic system were not higher than ing to the role of sun exposure in skin cancers,
expected. A modest excess of non-​ASL tumors of such as their greater frequency in fair-​skinned
the liver (R 1.4) was considered likely to be due people (they are about 10 times more frequent
to misdiagnosis of ASL. A second analysis of the in whites than blacks) and their occurrence in
same material, with the addition of six smaller areas of the skin most exposed to sunlight comes
studies, found similar results for tumors of the from clinical observation, but other evidence of
liver, but a significant excess of soft tissue tumorsthe relationship, such as their greater frequency
(R 2.5) that was also suspected of being due to in persons in outdoor occupations and in sunny
misdiagnosis of ASL (Boffetta et al, 2003). Two geographic locations or areas in proximity to the
later, small studies gave conflicting evidence on equator required epidemiologic studies (Oliveria
whether lung cancer risk is increased in VC work- et al, 2006).
ers (Mastrangelo et al, 2003; Scélo et al, 2004). The relationship of skin cancer risk to dura-
The evidence suggests that ASL is the only tion and intensity of sun exposure, continuous
cancer to be found in excess among VC workers. or repeated exposures and skin-​burning have all
been explored: but there is no consensus on the
Other Occupational Settings influence of these parameters, except that they
With few exceptions, exposure to poison gas appear to differ between the pathologic types
has largely been a problem of an occupation—​ of cancer. The interrelationships of these fea-
soldiering. In the context of cancer, the most tures and the roles of genetic markers are com-
relevant is mustard gas that was used exten- plex:  they are reviewed by Kricker et  al (1994)
sively during the last years of the war of 1914–​18. and English et  al (1997). There is evidence of
Another seminal follow-​up study was conducted association of risk with age, exposure in child-
by Case and Lea (1955). They identified deaths hood being particularly harmful. In addition,
among veterans receiving disability pensions as lifetime alcohol consumption has been shown to
a result of gas poisoning: 29 died of lung cancer, be positively related to risk in high-​risk areas (Le
whereas 14 would have been expected. There was Marchand et al, 2006), but it seems possible that
no excess of deaths from any cancer site other this is confounded by association with level or
than lung. Not unexpectedly given the date of type of sun-​exposure.
this, cigarette smoking was not considered as a
possible explanation of the excess. Arsenic
Exposure to wood dust has been reported in Some chemicals that have been shown to be car-
several studies as a cause of NPC (Demers et al, cinogenic in specific settings may have similar
1998), but a multicenter case-​control US study roles in the general environment. One of these
found only a modest association with wood dust is arsenic. Unlike its role in occupational cancer,
that was explained by confounding by a stronger the predominant general exposure to arsenic is
association with formaldehyde (Vaughan by ingestion—​in drinking water—​and the prin-
et al, 2000). cipal tumors involved are not of the lung but of
There is substantial literature on cancer risks the bladder. The significance of this problem is
on occupations in which chemical exposures are that it may not be economically or politically fea-
multiple and individual causes not identifiable—​ sible to reduce human exposure to a level below
for example, agricultural workers, rubber that at which carcinogenicity has been demon-
workers, and metal welders. Such studies have strated. While the capacity of arsenic to induce
relevance to the specific industries or plants iden- cancer in humans is beyond question, no animal
tified and may have utility in suggesting avenues model has been found. This leaves epidemiology
for more detailed investigation. as the principal tool for investigating its carcino-
genic potential at low exposure levels, although
THE PHYSICAL it is doubtful that epidemiological studies on a
ENVIRONMENT scale necessary to influence the debate will ever
be conducted.
Ultraviolet Radiation Drinking water was first associated with blad-
Exposure to ultraviolet radiation (UVR) from the der cancer in areas of Taiwan in which “Blackfoot”
sun is the most frequent cause of the common disease—​ a serious constrictive disease of the
1

Accomplishments in Cancer Epidemiology 11

peripheral arteries—​was endemic (Tseng et  al, terms of lung cancer, mesothelioma, and other
1968; Chen & Wang, 1990): the disease is due to asbestos diseases are documented by Bonn
high arsenic content of water from artesian wells. (1999), Mzileni et al (1999), Abratt et al (2004),
Correlations of cancer with arsenic content of and many others.
drinking water have also been found in Chile, In the less dramatic conditions of North
Argentina, Finland, and China:  these correla- America, no excess of cancer was found in one
tions were most consistent and strongest for can- Canadian town with asbestos deposits or another
cer of the bladder, but elevated risks for cancers with asbestos processing facilities (Neuberger
of the stomach, lung, and skin were also reported. et al, 1984). A review of 13 epidemiologic studies
This evidence is ecological, so that while the asso- of the roles of ingested asbestos showed no evi-
ciations are strong they leave unclear whether dence of increased risk of cancer (Marsh, 1983).
arsenic is the sole or only responsible factor. In Detection of health effects of low-​level asbestos
a cohort study in north Taiwan, in which each exposure may become easier as the “background”
family had its own well, it was possible to obtain of the occupational disease diminishes, but envi-
a more precise definition of exposure: there was ronmental exposures will also diminish, and the
a biologic gradient of risk of urinary tract cancer full consequences of the asbestos epidemic will
with level of exposure, with R for arsenic concen- never be known.
trations of less than 10, 10–​49, 50–​99, and 100
or more ug/​liter of 1, 2, 8, and 15, respectively Benzene
(Chiou et al, 2001). Sources of benzene in the environment include
The level of 50 ug/​liter, suggested by Chen gasoline, automobile exhaust, emissions from
et al, 2004 as a level at which arsenic in drinking industrial processes, and cigarette smoke.
water may cause observable increases in tumors, However, because of the high volatility of the
is the maximum currently permitted in US drink- chemical, exposures to the general population—​
ing water. In 2001 the Environmental Protection except for smokers—​are likely to be transient and
Agency proposed to reduce this standard to 10 very low compared to those found in enclosed
ug/​liter, but enforcement was delayed “for fur- work spaces. Nevertheless, the issue has received
ther study.” Such study will require judicious bal- some attention. In one study of gas station atten-
ancing of the costs of compliance (which will be dants a small excess of deaths from AML was
considerable, perhaps prohibitively so) with esti- reported (Jacobsson et  al, 1993), but in larger
mates of the reduction in the number of cancers studies leukemia cases were below expectation
prevented, the latter being determined prima- (Lagorio et al, 1994; Lynge et al, 1997).
rily by the statistical model used to extrapolate Most studies of this problem are ecological
the effect of exposures below the levels for which in nature. In a small area of Pennsylvania, gas-
data are available (Frumkin & Thun, 2001). oline leaked from underground storage tanks,
exposing the population to chronic low levels of
Asbestos gasoline in the air and drinking water for several
Spread of asbestos dust in close proximity to years. A  study of affected households showed
mining, processing, and distributing facilities incidence rates of cancer to be in line with rates
and in the removal or disintegration of asbes- for the state as whole, but there were four cases
tos insulation, brake linings, and other prod- of leukemia where only one was expected:  the
ucts raises concern for the health of populations residences of two cases of AML were said to have
beyond production workers. These concerns are “bordered the projected gasoline plume” (Patel
serious in the North Cape Provinces of South et al, 2004).
Africa, where asbestos, including the highly Rates of lymphatic and hemopoietic cancers
carcinogenic crocidolite, was discovered in the in an area in Italy that contained a major source
1880s and mined throughout the 20th century. of environmental benzene were compared with
Although the mining was done by males, females local areas with no such sources (Parodi et  al,
and children were frequently involved in proc- 2003). For leukemia, rates in males were about
essing the ore (Wagner, 1991). Exploitive labor twice as high in the exposed as in the control
practices and living conditions, leaking distri- areas, but among females, the reverse was the
bution facilities, and careless disposal of incom- case. For non-​Hodgkin’s lymphoma rates were
pletely extracted rock are described by Braun and also higher in the exposed than in the control
Kisting (2006). Consequences to populations in areas in males, but not in females. There is little
12

12 Textbook of Cancer Epidemiology

reason to think that environmental exposures recognition was important for the stimulation
would affect males but not females, and if these and acceptance of broad measures for cigarette
differences are real, it would seem reasonable to control.
seek their explanation elsewhere—​perhaps in an
occupational exposure. Comparisons have also THE MICROBIOLOGIC
been made of residents of areas believed to have ENVIRONMENT
had high benzene exposure because of proxim- The last 50  years have seen increasing recogni-
ity to traffic congestion (Crosignani et  al, 2004; tion of roles for microorganisms in cancer etiol-
Raaschou-​Nielsen et  al, 2001). The findings are ogy. In addition to the schistosome infestations
not striking and are inconsistent. In addition to that cause bladder cancer, the flukes related to
the usual problems of interpreting ecologic asso- liver cancer, and the human T-​cell lymphotro-
ciations, benzene is not the only possibly carcino- phic viruses (HTLVs) related to leukemia in
genic constituent of pollution from automobiles. certain parts of the world, there are more wide-
Overall, there is little evidence for a cancer spread human cancers, including those of the
effect from environmental contamination with liver, cervix, and stomach, in which infectious
benzene. agents are involved. In addition, epidemiologic
studies have supported the prediction of Burnet
Chromium (1965) that persons with compromised immune
As with arsenic, exposure to chromium in the response to infection are at increased risk of
general environment is not by inhalation but by cancer. Parkin (2006) estimated that about 18%
ingestion and possibly from contact with water (1.9 million cases annually) of all world cancers
contaminated by discharges from industrial can be attributed to infection and that preven-
plants. After a review of four studies of envi- tion of these would lower cancer incidence by
ronmentally exposed populations, Proctor et  al 8% in developed countries and 26% in devel-
(2002) concluded that Cr-​6 is not carcinogenic oping areas. While credit for the identification
via the oral route of exposure from currently of microbiologic agents as causes of cancer
permitted concentrations of the compound in rests securely with the microbiologic sciences,
drinking water. epidemiologic studies have assisted in suggest-
Cigarette smoke is inhaled not only by smok- ing directions in which these sciences might be
ers but by others around them—​in homes, work applied and in evaluating the effectiveness of
places, restaurants, and elsewhere. The names vaccination programs.
“passive” and “involuntary” smoking has been
given to such exposures. In 1981, two studies Hepatitis Viruses
of lung cancer in married women showed that Hepatocellular carcinoma (HCC) is among the
women who did not smoke but were married to world’s most common neoplasms, and the dis-
men who did had approximately twice the risk covery of hepatitis viruses as its principal cause
of women married to nonsmokers (Trichopoulos rivals in public health importance the discovery
et al, 1981; Hirayama, 1981). An analysis of data of the carcinogenic effects of cigarette smok-
from 37 studies of the issue suggested that life- ing. Microbiologic discoveries depended on the
time nonsmokers who lived with a smoker had development of increasingly sensitive laboratory
a risk of lung cancer 24% higher than those not techniques:  epidemiologic evidence contributed
living with a smoker (Hackshaw et  al, 1997). by demonstrating a link to a known infectious
There were positive exposure–​risk relationships disease, identifying the means of transmission
both with the amount smoked by the companion of the responsible viruses and factors associated
and with the duration of residence with a smoker. with their persistence in the host, evaluating the
A  later analysis based on 76 studies gave an efficacy of vaccination programs and exploring
excess risk of 29% (Taylor et al, 2001). The reality the roles of alcohol, tobacco, and aflatoxin.
of the relationship between involuntary smoking Hepatocellular carcinoma was known clin-
and lung cancer was recognized by the IARC in ically to be a complication of chronic hepatitis
2004 (IARC, 2004a) and the Surgeon General of and cirrhosis. During the 1960s it became evident
the United States in 2006. The numerical impact that there are (at least) two forms of virus-​related
of involuntary smoking is small relative to that of hepatitis—​one of relatively short duration, trans-
active smoking and it is likely to become smaller mitted by contagion with typical incubation
as the prevalence of smoking declines, but its periods of 20–​30 days and often called infectious
13

Accomplishments in Cancer Epidemiology 13

hepatitis and a second clinically more severe dis- of this virus have followed closely the course of
ease, with longer incubation periods and dura- those for HBV, sometimes both viruses being
tions, prone to recurrence, often transmitted by evaluated in the same study. Risks associated
blood products, and referred to as serum hepatitis with infection with HCV alone are of about the
(Krugman et al, 1967). Infectious hepatitis came same order as those with HBV alone, and the
to be known as hepatitis A, caused by the homon- risks associated with infection by the two viruses
ymous virus hepatitis (HAV), whereas serum together appear to be additive (Sun et  al, 2003;
hepatitis was distinguished as hepatitis B and C, Yu et al, 2005).
again caused by homonymous viruses HBV and High alcohol consumption, liver cirrhosis,
HCV, respectively. A  previously unknown anti- and tobacco use are established causes of HCC,
gen found in an Australian Aborigine in 1966, and in countries with low prevalence of HCV and
and initially called Australia antigen (Blumberg HBV, such as in western Europe and North
et  al, 1965), was subsequently identified as the America, may be responsible for the majority of
surface antigen of HBV (HBsAg) and became a liver cancers (Trichopoulos et al, 1987).
marker of infection.
The association of HBV with HCC was first Human Papilloma Viruses
observed ecologically:  the high prevalence of Epidemiologic features of cancer of the cervix
chronic infections with HBV, as evidenced by have long suggested that this disease is associ-
high prevalence rates of serum HBsAg, in Asia ated with early onset and variety of sexual activ-
and Central Africa and high prevalence in males ity (Jones et al, 1958; Beral, 1974). They include
corresponded to the epidemiologic features of associations of high rates of the disease with early
HCC (Maupas & Melnick, 1981). Following age at marriage or onset of sexual activity, fre-
Prince et  al, 1975, high prevalence of serum quency of intercourse, number of sexual partners
HBsAg-​positivity was found in cases of HCC in including the number of sexual partners of the
many case-​control studies. Follow-​up studies fol- spouse as well as of the patient herself, and poor
lowed, with R varying between more than 100 genital hygiene. Low rates were also seen among
in Taiwan (Beasley et  al, 1981)  and 15 in Spain members of religious or social groups per-
(Ribes et al, 2006). Risk of HCC was also found ceived as having low levels of sexual promiscu-
to be associated with the amount of HBV DNA in ity (Brinton, 1992). Early hypotheses focused on
plasma: in a cohort of men known to be HBsAg possible chemical carcinogens transmitted from
positive, risks for men with estimated HBV DNA male partners in sperm or smegma or infection
serum loads in the highest quintile were seven with herpes simplex virus type 2.
times those of men in the lowest quintile (Yu The evidence associating human papilloma
et al, 2005). virus (HPV) infection with cancer of the cervix
A vaccine against HBV was developed and (as well as other genital cancers both in males
widely used in high-​prevalence countries. In a and females) is reviewed by Bosch et  al (2002).
disease with such devastating consequences, it is It is the result of the convergence of two distinct
not surprising that evidence of the vaccine’s effi- paths of inquiry:  (1) clinical, pathologic, and
cacy came from observations in nonrandomized epidemiologic studies demonstrating a seamless
populations. In a prospective study of Korean progression in cervical epithelial tissue from mild
males, of whom 13% had been vaccinated in the to severe dysplasias to carcinoma-​in-​situ (CIS) to
year prior to the observation period, the inci- infiltrating carcinoma, a progression recognized
dence of HCC was more than 50 times as high in the more recent designation of the premalig-
in unvaccinated as in vaccinated men. (Lee et al, nant stages as cervical intraepithelial neoplasia
1998) A nationwide vaccination in Taiwan began (CIN), and (2)  development after the 1970s of
in 1984 with vaccination of newborns of HBsAg-​ sensitive microbiologic techniques for the identi-
positive mothers and was extended in stages fication of HPV DNA and its subtypes.
to all newborns and then to all schoolchildren The largest cross-​sectional study of the asso-
(Chien et  al, 2006):  over a 10-​year period the ciation is one sponsored by IARC in which more
incidence of HCC in children declined by 50% than 1,000 specimens from patients with invasive
(Chang et al, 1997). cervical cancer in 22 countries were evaluated
The HCV was also identified as an important (Bosch et  al, 1995). HPV DNA was detected in
cause of HCC, both as a cofactor with HBV and 93% of tumors. HPV types 16 and 18 predomi-
as an independent agent. Epidemiologic studies nated:  other HPV subtypes were also detected
14

14 Textbook of Cancer Epidemiology

with high though lower frequencies. Relative risk strongly suspected as a cause of gastritis and duo-
associated with any HPV type was 83 for squa- denal ulceration (Marshall & Warren, 1984) and,
mous cell carcinomas and 69 for adenocarcino- less definitively, as a factor in cancer of the stom-
mas (Bosch et al, 2002). ach. A  review of five meta-​analyses of the rela-
In a study of cases of CIS with evaluation of tionship of the organism to gastric cancer led to
HPV16 in specimens taken up to 26 years previ- the conclusion that, while the data are conflict-
ously, women with high virus loads before age 25 ing, the consensus is that the frequency of gas-
were 30 times more likely to develop CIS, often tric cancer is about doubled in the presence of
when the original smears were cytologically H. pylori (Eslick, 2006). The main uncertainty is
normal (see ­chapter  17) (Ylitalo et  al, 2000). In whether this represents a causal association or an
the same study, there was a strong relationship opportunistic infection by the bacillus, either of
between the level of virus load and subsequent the tumor itself or of one or more of its progeni-
risk of CIS. Schiffman et al (2002) found that the tors, such as chronic gastritis.
natural history of HPV infection often includes In the only randomized experiment reported
periods before and after observable cytologic to date, patients with H.  pylori infection were
abnormality during which HPV DNA is a more assigned equally to receive either chemotherapeu-
sensitive indicator of subsequent malignant tic eradication therapy or placebo: after 7.5 years,
change than is cytology. The use of HPV assays 7 of the treatment and 11 of the placebo group
as an adjunct to, or possibly replacement for, developed gastric cancer (Wong et al, 2004). The
cytologic screening, is a matter of current debate significance of this finding has been criticized on
(Josefsson et al, 2000). the basis of the relatively short follow-​up period
A vaccine—​in lay terms a “cervical cancer (Parsonnet & Forman, 2004). Prospects for clar-
vaccine”—​was developed against HPV: its effec- ification of the role of this organism in gastric
tiveness in terms of reduction of risk of cervical cancer depend on large-​scale prospective studies,
cancer remains to be demonstrated. The ethics both observational and experimental. If a role
and economics of use the vaccine will depend in for the agent is established, it is evident from the
considerable part on the knowledge of the epide- high prevalence of the infection and the relatively
miology of cervical cancer itself. low frequency of the disease that there are impor-
tant modifiers of that role.
Helicobacter pylori
The descriptive epidemiology of cancer of the Immunodeficiency
stomach is challenging. For example, incidence Three recent developments have provided
rates vary over a fivefold range, with highest rates opportunities for investigation of the role of the
in Japan, Finland, and South America (Parkin immune system in protecting against cancer: the
et  al, 2005). One contribution of epidemiology increased survival of infants with congenital
has been to recognize that the disease includes immune deficiency, the epidemic of the acquired
two entities that are pathologically and epide- immunodeficiency syndrome (AIDS), and the
miologically distinct. One of these occurs at the use of artificial means of depressing patients’
proximal (cardiac) end of the stomach, the other immunity prior to organ transplantation.
at the distal (pyloric) end. Both are more common Congenital immunodeficiency is a feature of
in males than females, but the proximal by a fac- a few rare clinical syndromes, such as Wiskott-​
tor of about 5, the distal by about 2. Distal tumors Aldrich syndrome, ataxia telangiectasia, and a
are more common in blacks than whites in the few others. Estimates of risk of cancer associated
United States, while proximal tumors are twice with these syndromes are derived from clinical
as common in whites as blacks. In most western series rather than formal epidemiologic studies.
countries, rates for the disease as a whole have They vary between syndromes and are heav-
been declining (to about one-​third of those seen ily influenced by the duration of follow-​up, the
around 1930), but proximal cancers have been more important component of which is the dura-
increasing in recent years in Europe, Australia, tion of survival of immunocompromised infants.
and New Zealand (Crew & Neuget, 2006). Overall risks of malignancy are low—​generally
Despite these striking features of the disease, less than 5%—​but very much higher than in the
immediate prospects for prevention of stomach general population of infants—​in some instances
cancer come from the incidental discovery of a by a factor of 1,000 or more. A  striking feature
microorganism—​Helicobacter pylori—​that is is the predominance of non-​Hodgkin lymphoma
15

Accomplishments in Cancer Epidemiology 15

among the malignancies in these patients, gen- (MacMahon et  al, 1970b; Collaborative Group,
erally constituting more than half of the tumors 2002b), but the role of childbearing itself has
(Muller & Pizzo, 1995). been more fully characterized. It is evident that
The association of the human immunodefi- the protective effect of childbirth is not prima-
ciency virus (HIV), responsible for AIDS, with rily a function of the number of children borne
cancer has been demonstrated in clinical reports but of the woman’s age at the time she bears her
and in clinicoepidemiologic studies such as the first child (MacMahon et al, 1970a): if births after
cohorts of AIDS patients enrolled in Switzerland the first have an additional protective effect, it is
(Clifford et  al, 2005), Italy (Nasti et  al, 2003), small relative to that of the first (Trichopoulos
England (Newnham et  al, 2005), across Europe et al, 1983).
(Mocroft et  al, 2000), and in the United States The relationship has been further clari-
(Kest et al, 2005). The dominant cancer seen in fied in two respects: (1) while the risk of cancer
AIDS patients is Karposi’s sarcoma—​a previously increases with age at first birth up to about age
uncommon tumor of lymphoid origin affecting 30, first births after that age are associated with
principally the skin but also internal organs, seen increased risk relative to that for nulliparous
mainly in tropical areas and formerly believed, women, although, since most first births occur to
often correctly as it developed, to be of venereal mothers under 30 years of age, the overall effect
origin. The disease is estimated to be 10,000 times is protection, and (2) the risk of cancer is actually
more common in HIV-​infected than noninfected increased in the first 10 years after a first birth—​
persons and is associated with infection by her- even a birth at a young age—​but after 10 years it
pes virus, type 8.  Other tumors associated with is decreased (Lambe et  al, 1994):  Since most of
AIDS, but at lower levels of frequency (R 5–​10) woman’s life years occur more than 10 years after
are non-​Hodgkin lymphoma and Hodgkin’s dis- birth of her first child, the overall effect is pro-
ease, both associated with Epstein Barr virus tection. These observations are consistent with
infections (Beral & Newton, 1998). the idea that pregnancy stimulates replication
Patients undergoing organ transplant have a and differentiation of breast cells, thus increasing
three-​to fourfold higher risk of cancer than the in the short term the probability of a malignant
general population (Taioli et al, 2006), a phenom- mutation or expansion of an initiated clone but
enon attributed to the immune suppression arti- leading later to a population of cells at low risk
ficially induced prior to surgery. The pattern of of transformation—​in line with a model sug-
post-​transplant (predominantly renal transplant) gested by Russo et al (1990) to account for similar
cancers is similar to that seen in AIDS patients, observations in rats.
with Kaposi’s sarcoma predominating, tumors of In a meta-​analysis of 12 case-​control studies,
lymphatic origin sharply increased, and a scat- risk of cancer of the ovary was also found to be
tering of other cancers showing small increases inversely associated with parity, women who had
(Penn, 2000; Pedotti et  al, 2003; Serraino et  al, 3–​4 full-​term pregnancies had about half the risk
2005). The risks for post-​ transplant tumors of those who had none (Whittemore et al, 1992).
appear to be lower than those for AIDS patients, La Vecchia et  al (1984) suggested that, as with
but it is not clear whether this is a function of cancer of the breast, the association with high
differences in survival of the two groups or of the parity could largely be accounted for by an asso-
nature of the immunosuppressant. ciation with late age at first birth, but this obser-
vation has not been as fully investigated as has
that with cancer of the breast.
REPRODUCTION As with cancers of the breast and ovary,
The utility of studying variation in reproductive risk of cancer of the uterine endometrium is
experience as a clue to cancer etiology became also reduced by childbearing (Henderson et  al,
clear when early reports suggested that cancer 1983; Brinton et  al, 1992), but specific patterns
of the breast is unusually frequent among nuns are different. From a large Swedish study it was
and single women. The first epidemiologic study estimated that, among parous women, each addi-
was that of Lane-​ Claypon (1926), indicating, tional birth was associated with a 20% decrease
among other things, that breast-​ feeding pro- in endometrial cancer risk (Lambe et  al, 1999).
tects a woman against breast cancer. The role of There was no consistent association with age at
breast-​feeding remains controversial, though it is first birth, but the older a woman is at the time
not as significant as was thought for many years of her last birth the lower her risk. It has been
16

16 Textbook of Cancer Epidemiology

estimated that risk declines by 15% for each (R 2.5). The increase in larynx cancer risk associ-
5-​year delay in a woman’s age at last birth (Lambe ated with drinking among nonsmokers seems to
et al, 1999). occur only in parts of the larynx that are in direct
The different patterns of reproductive expe- contact with alcohol—​an idea consistent with the
rience preceding cancers of the breast, ovary, early study of Tuyns et al (1988).
and endometrium, while suggesting agreement The effects of alcohol and tobacco in esopha-
on the idea that childbearing promotes anticar- geal cancer were evaluated in a case-​control study
cinogenesis, pose challenges to interpretation in in high-​risk areas of South America (Castellsagué
terms of mechanisms. et al, 1999). Supporting conclusions from earlier
studies (Tuyns, 1983), this study showed inde-
ALCOHOL pendent effects of both alcohol and tobacco with
Associations of some types of cancer with heavy multiplicative effects of the two exposures in
and regular use of alcoholic beverages were first combination:  heavy alcohol consumption com-
noted clinically:  epidemiologic studies estimate bined with high use of so-​called black tobacco
the size of the risks involved. In addition, since was associated with a risk more than 100 times
alcohol use is associated with other variables, of that of nondrinkers-​ nonsmokers:  the associa-
which the most relevant is cigarette smoking, tions were seen both in men and women. For can-
epidemiologic studies are needed to separate the cer of the mouth and pharynx the combined and
contribution of alcohol to cancer rates follow- independent effects of alcohol and tobacco were
ing joint exposures. A  recent and comprehen- clearly shown by Rothman and Keller in 1972.
sive review of the epidemiologic literature and Analyses specifying the effects of different
of mechanisms that might underlie associations types of alcoholic beverages generally show high-
between alcohol use and cancer risk is that of est risk in the beverage most typical of the area
Boffetta and Hashibe (2006). It is hardly neces- in which the study was carried out, as for exam-
sary to note that assessments of alcohol intake, ple calvados became the assigned culprit of the
particularly intakes in the past are approxima- esophageal cancer epidemic in districts of north-
tions and quantitative estimates of risk are crude. ern France. There is still much to be learned about
With these limitations, Boffetta et al (2006) esti- the roles of alcohol concentration and adulter-
mated that alcohol-​drinking accounts for 5.2% of ants in the carcinogenicity of alcoholic beverages
world cancer in males and 1.7% in females. to the upper respiratory-​digestive tracts.
Associations with high cancer risk have been In light of the role of the liver in the metab-
noted for all organs with which unmetabolized olism of alcohol and the role of chronic alcohol
alcohol comes in contact, with the exception of abuse in hepatic cirrhosis, the association of
the stomach—​that is, the mouth and pharynx, alcohol use with HCC is hardly surprising. The
larynx, and esophagus. Corrao et al (2004) in a meta-​analysis of Corrao et  al (2004) suggests
meta-​analysis of 156 studies, estimated R associ- R  of 1.8 associated with the daily consumption
ated with daily use of 25 g of alcohol (estimated of 100 g of alcohol—​perhaps smaller than clini-
as two “drinks”) for these three cancer sites were cal impressions and some studies would lead one
1.9, 1.4, and 1.4: for daily use of 100 g they were to suppose. As noted earlier, lifestyles associated
6.5, 3.9, and 3.6, respectively. with heavy alcohol consumption—​ including
The most direct evidence of an effect of alco- opportunities for multiple infections with hepati-
hol use independent of that due to smoking is tis viruses—​must be considered as at least partial
by examination of its use among nonsmokers. explanations of the increased risk of HCC.
With respect to oral and pharyngeal cancer, the Small excesses of cancers of the colon and
results of such a study by Fioretti et al (1999) are rectum (1.2 and 1.4, respectively, in Corrao et al,
generally consistent with those of earlier inves- 2004)  are more difficult to explain. Even more
tigators:  among lifetime nonsmokers, reported enigmatic is the association of alcohol use with
consumption of three of more drinks per day cancer of the breast (2.4), possibly related to
was associated with an R of 3.  In a similar association of alcohol intake with estrogen levels
study of cancer of the larynx, while R for non- (Reichman et  al, 1993). It is difficult to identify
drinkers increased regularly with the number other known epidemiologic features of breast
of cigarettes smoked daily to 14 for smokers of cancer that explain the association that has been
25 daily: among nonsmokers increased risk was found consistently over the last two decades
seen only for drinkers of 8 or more drinks per day (Longnecker, 1994; Hamajima et al, 2002).
17

Accomplishments in Cancer Epidemiology 17

I AT R O G E N I C E X P O S U R E evidence of increasing risk with duration of use

Complications of therapy that appear imme- (Collaborative Group, 1996). Another meta-​
diately or within a short time are often recog- analysis, based on 34 studies of premenopau-
nized clinically, but those that develop after long sal women, found a similar increase in breast
intervals or require accumulation of exposure cancer risk with OC use overall (R 1.2) and an
over time can usually be ascertained only by indication that, while there was little difference
long-​term follow-​up studies in which epidemio- in risk between parous and nonparous women,
logic principles are central. Such complications risk was somewhat higher among parous women
include neoplastic disease. The iatrogenic causes who used OCs prior to their first full-​term preg-
of cancer associated with IR and have already nancy: among those with 4 or more years of use
been referred to. prior to their first birth R was 1.5 (Kahlenborn
et  al, 2006). The data do not permit an evalua-
Diethylstilbestrol tion of whether the small increased risk among
Between 1940 and 1970 diethylstilbestrol (DES) current and recently discontinued OC users rep-
was prescribed for about 2  million pregnant resents an increased frequency of diagnosis of
women with a history of miscarriage or other disease or a real increase, but an indication in the
complications of pregnancy. In 1970, Herbst and Collaborative Group study that the additional
Scully reported seven cases of a very rare adeno- cancers were limited to tumors localized to the
carcinoma of the vagina in young women seen breast points in the direction of more frequent
at one hospital, and in the next year Herbst et al diagnosis.
(1971) published a study of 8 cases and 32 nor- Long-​term use of OC is associated with
mal infants delivered in the same institution: the decreased risk of ovarian cancer, a conclusion
mothers of 7 cases but none of the controls had reached in a meta-​analysis of 12 case-​control
received DES during pregnancy. A  registry for studies (Whittemore et al, 1992). A later analysis
vaginal adenocarcinoma was established, and in of six studies showed reduction of risk (0.7) asso-
virtually all cases there were histories of maternal ciated with any use of OC and a further reduc-
DES therapy (Ulfelder, 1973). The therapy was tion (to 0.5) with use for 5 years of more (Bosetti
discontinued. et al, 2002).
Whether prenatal exposures in general are Use of OCs is also associated with reduced
associated with increased risk of cancer of the risk of cancer of the endometrium. In a large
breast is currently an important area of investi- Swedish case-​control study, R for endometrial
gation (Lagiou et  al, 2006), and the experience cancer risk was reduced to 0.5 after 3+ years of
of women exposed to DES prenatally, who are use and to 0.2 after 10 years of use (Weiderpass
now in their middle and later years, is relevant. et al, 1999b). The decrease remained for at least
A  large cohort of women exposed prenatally to 20 years after cessation of use—​a matter of bio-
DES and comparable unexposed women was logic as well as practical interest.
identified from early studies and followed pro- Evidence suggests a positive association of
spectively (Palmer et al, 2006). There was a small OC use with risk of cervical cancer, but while
increase in risk (1.4) and a suggestion that the attempts were been made in some studies to con-
risk increased with age, although numbers were trol for sexual activity, it is not clear that this var-
small in the older age-​groups. Continued follow-​ iable can be assessed with sufficient accuracy for
up of this group is important. its elimination as a source of confounding.
From a practical viewpoint, Schlesselman
Oral Contraceptives (1995) used available estimates of risk to com-
Introduction of oral contraceptives (OCs) pute additional or reduced annual tumors among
in the early 1960s raised fears that their use 100,000 US women aged 20–​54 that would be
would increase risks of breast cancer and other expected with use of OC for 8 years. They were
hormone-​ dependent tumors. Combined anal- for breast +151, cervix +125, liver + 41, endome-
ysis of 54 studies suggested a small increase in trium –​197, and ovary –​195. The net effect (+317,
risk (1.2) of breast cancer for current users and –​392) is small though slightly favorable.
an even smaller increase for women who had
discontinued use within 10  years:  women who Exogenous Estrogens/​Progestins
had discontinued use 10 or more years pre- Quite different conclusions come from studies
viously had no increase in risk:  there was no of hormones prescribed for relief of symptoms
18

18 Textbook of Cancer Epidemiology

in the menopausal years—​ initially estrogens of the kidney was confirmed (McCredie et  al,
and later progestins or a combination of the 1995; Rosenberg et al, 1998).
two. Increased risk of breast cancer was first Other findings on the use of analgesics is
demonstrated in a follow-​up of patients tak- less depressing—​even positive. Thun et al (1991)
ing premarin—​a mixture of estrogens derived found a reduction of close to 50% in risk of death
the urine of pregnant mares. There was a 30% from colon cancer among men and women
increase in risk of breast cancer in the group, and reporting regular use of aspirin over a period of
among those taking the medication for 30 years at least 1 year. This was confirmed for colorectal
or more the risk was doubled (Hoover et  al, cancer both for men and women (Giovannucci
1976). An extensive literature—​both epidemio- et  al, 1994, 1995). The extensive literature that
logic (Collaborative Group, 1997; Weiderpass followed addressed primarily two questions: (1)
et al, 1999a) and experimental (Roussouw et al, is the apparently protective effect of aspirin
2002)—​ generally supports these conclusions. shared with nonaspirin NSAIDs (NA-​NSAIDs)?
Formulations combining progestin with estro- (2) does it extend to cancers in sites other than
gens are associated with higher risk than those the large bowel?
containing with estrogen alone (Newcomb et al, On the first question, Garcia Rodriguez
2002). In a randomized trial, women receiving et al (2001) reviewed 18 studies published since
estrogen alone (Anderson et al, 2004; Stefanick 1988:  there were six in which both aspirin and
et  al, 2006)  had lower breast cancer risk (0.8) NA-​ NSAIDs were evaluated:  the combined R
than those receiving no hormones, but partici- from these was 0.6 both for aspirin and for NA-​
pants had had prior hysterectomies (and prob- NSAIDs. In Garcia Rodriguez’s own data nine
ably oophorectomies) and consequently both individual NSAIDs were evaluated:  although
trial groups were initially at low risk of breast numbers were small, they did not show heteroge-
cancer. neity of the association between drugs. There is at
The association of endometrial cancer with present no reason to think that aspirin is different
exogenous estrogen is similar to that of breast from other NSAIDS in their inverse association
cancer in that risk is elevated and increases with with colorectal cancer.
duration of use: a 1995 meta-​analysis of 30 pub- The second question was the subject of a
lished studies estimated R as 2.3 for estrogen review and meta-​analysis of 47 published papers
users and 10 after 10 or more years of use (Grady by Gonzales-​Perez et al (2003). Aspirin use was
et al, 1995). However, the association differs from found to be associated with reduced risk of can-
that of breast cancer in that addition of proges- cer of the esophagus (0.5), stomach (0.5), and
tins reduces the increase associated with the use breast (0.8). Data for other NSAIDs gave similar
of estrogens alone (Persson et  al, 1989; Beral results.
et  al, 2005). In a randomized trial, women and It is premature to suggest that NSAIDs be
given estrogen/​ progestagen combinations had used for cancer prevention, but this has been
slightly lower risk (0.8) than those given placebo considered a future possibility (Baron, 2003).
(Anderson et al, 2003).
DIET
Analgesic Medications The contribution of diet to human cancer risk
Known for many years as a cause of renal nephrop- may be considerable. That this subject does not
athy, phenacetin was banned or controlled in have a more prominent place in this chapter is
many jurisdictions. In a follow-​up of analgesic because specific dietary constituents have not
abusers among males with chronic pyelonephri- been established as being or not being causally
tis, 11 developed urinary-​tract tumors:  all were related to cancer risk. However, epidemiologic
using phenacetin-​containing drugs (Bengtsson & studies have made considerable progress toward
Angervall, 1970). The induction times in these that end. The recent reviews of Willett (2001) and
cases were long (6–​35 years) and similar to those Lagiou et al (2002) are informative.
noted earlier for occupational bladder tumors. Suggestions of a role for some environmen-
A larger study confirmed increased risk (R 12) of tal factor in cancer etiology, among which diet
renal cancer following phenacetin use, but no seemed a likely candidate, came from studies of
increase with acetaminophen (McCredie et  al, international variations in rates of specific neo-
1993). That acetaminophen is not associated with plasms. These first became evident in mortality
cancers, either of the renal pelvis or of the body statistics compiled and published by the World
19

Accomplishments in Cancer Epidemiology 19

Health Organization and brought to attention study protocol. However, the secondary analyses
in a series of monographs by Mitsuo Segi (1962, do not carry the teleological advantages usually
1964) and, with the development of cancer regis- associated with randomization:  initial risk was
tries throughout the world, were much enhanced not considered in the randomization and “adher-
by incidence data compiled by IARC (Parkin et al, ence” carried different implications for the trial
2005). Studies of migrant populations showing and comparison groups since women in the
that cancer rates of migrants and their descen- comparison group were not asked to make die-
dents almost invariably move toward those of the tary modifications. For colorectal cancer, rates
host country, though at different rates, confirmed were slightly higher in the intervention than the
the environmental origin of the international dif- comparison group (R 1.08) and there was no dif-
ferences: studies of migrants to Hawaii (Haenszel ference relating to assessed adherence to study
& Kurihara, 1968) and Australia (McCredie et al, protocol (Beresford et al, 2006).
1990) were early and influential in this regard. In The results of this major undertaking are dis-
the context of diet, hypotheses focused particu- appointing. When viewed against the three-​to
larly on cancers of the breast and lower intesti- fivefold differences in disease rates that epide-
nal tract—​sites in which international variation miologic studies indicate need to be explained,
seemed less likely to be explained by difference the differences observed in the trial are small.
in diagnostic or reporting practices. The high-​ However, they do not throw doubt on the pos-
caloric, high-​fat, and low-​fiber diets of the devel- sible significance of diet as an explanation of the
oped countries came under particular suspicion. epidemiologic observations:  the populations in
Case-​control studies showed moderate which major differences in cancer rates are noted
increases in risk for cancers of the breast and experienced their diets for their entire lives,
lower digestive tract associated with diets high whereas for the participants in this trial changes
in animal fat and low risk for colorectal cancer occurred over a relatively short period and, pos-
associated with high intake of fruits and veg- sibly more significantly, at a fairly late age. The
etables, but these associations were generally most important lesson of this study may be that
not confirmed in prospective studies (Willett, randomized trials are not the most effective way
2001). In a combined analysis of 13 prospec- to investigate the role of dietary factors in cancer
tive studies, incorporating close to 726,000 men causation.
and women followed for between 6 months and It is clear that obesity—​ a characteristic
20  years, there was essentially no difference in related at least in part to diet—​is associated with
risks of colon or rectal cancer between persons increased cancer risk. In the American Cancer
in the highest and the lowest quintiles of fiber Society prospective study of over 900,000 US
intake, whether measured as reported intake of adults, death rates from cancer in the heavi-
fruits and vegetables or all sources of dietary fiber est members of the cohort were, for men, 52%
(Park et al, 2005). In another analysis of the data higher and, for women, 62% higher than for
from two large cohorts of health professionals, men and women of “normal” weight (Calle et al,
no relationship of cancer as a whole was found 2003). The authors estimated that current pat-
either for fruits and vegetable as a whole or for terns of overweight and obesity in the United
any of several subgroups (Hung et al, 2004). States account for 14% of cancer deaths in men
Much was expected of the recent Women’s and 20% of those in women.
Health Initiative trial in which 49,000 US women While several cancers contribute to the excess
aged 50–​79 were randomly assigned to one of rates associated with obesity, including cancers of
two groups, to one of which dietary interven- the colon, rectum, gallbladder, pancreas, and kid-
tion, aimed toward reduction of total fat and ney, cancer of the breast is the one site for which
increasing fruits and vegetables, was offered and the observations are most consistent and that
encouraged: both groups were followed, at latest offers the greatest potential for fitting into the
report for an average of about 8 years (Prentice overall pattern of knowledge of the disease, stem-
et al, 2006). Breast cancer incidence was slightly ming particularly from the positive correlation
reduced in the intervention group relative to between endogenous sex hormones and breast
the comparison group (R 0.91), and secondary cancer risk, at least in postmenopausal women
analyses suggested that the reduction was some- (Key et al, 2002; Collaborative Group, 2002a) and
what greater in women of initially higher risk the increased levels of plasma estrogens asso-
and in those who had been most adherent to the ciated with body mass index (Hankinson et  al,
20

20 Textbook of Cancer Epidemiology

1995). In a prospective study of 87,000 women material in the field of cancer prevention? This
aged 30–​55  years followed for 26  years, women chapter makes the case that epidemiology has
who gained 25 kg or more of weight after age 18 made an effective contribution to the knowledge
and who did not take postmenopausal hormones necessary for cancer prevention, both as an inde-
had twice the breast cancer incidence of compa- pendent discipline and in partnership with other
rable women who maintained weight: those who sciences.
lost 10 kg or more since menopause had less than Can specific areas of epidemiologic research
half the risk of those who maintained postmeno- be identified that on the basis of past contribu-
pausal weight (Eliassen et al, 2006). tions should be continued and strengthened?
An important area of investigation of the eti- With increasing specificity, predictions become
ology of cancer of the breast was opened with less reliable. But, again relying on the historical
accumulating evidence that the origins of the record, the following are suggested:
disease may be found in factors affecting breast
growth and development during the prenatal and 1. National and local statistics on causes of
early postnatal period: among these must be con- death and cancer registration and efforts
sidered both maternal and infant childhood diet to spread these resources to areas of the
(Trichopoulos, 1990; Lagiou et al, 2006). One of world that are not currently covered will
the first indications of prenatal influence was the continue to be productive. As in the past,
association of breast cancer risk with the patient’s the customs and diets of populations
birthweight. Unlikely as it seemed on first obser- not now covered by death certification
vation; this relationship is now well established or cancer registration may reveal clues
(Ahlgren et  al, 2003). In a classic illustration of that have significance worldwide.
the value of record linkage, Ahlgren et al. (2004) Such statistics are also essential for the
combined information from birth certificates, evaluation of the efficacy of preventive
school health records, and a national cancer reg- measures.
istry to show positive associations between adult 2. Facilities for maintenance and
breast cancer risk and birthweight, height at ages computerization of original medical
8 and 14, and body mass index and growth rates and demographic records and statistical
around the time of puberty. The possibility that techniques for record linkage need
associations of birthweight and childhood growth continued development. The creation of
rates with breast cancer risk may be influential in the National Death Index in the United
increasing rates of the disease among migrants States and existing comparable facilities in
and yet be compatible with the weak or null asso- Britain and the Nordic countries provide
ciations between adult diets and breast cancer abundant examples of the value of such
risk is explored by Lagiou et al (2003, 2006). resources.
Additives to and contaminants of diet have 3. Multicenter studies, such as the
frequently been suspected of carcinogenic poten- collaborative case-​control programs
tial and provoked much discussion and regula- of the US National Cancer Institute
tory action. But most such concerns derive from and the EPIC prospective program
extrapolations from experimental work rather in Europe, are useful in expediting
than epidemiologic observations:  they have meaningful results and narrowing
rarely been evaluated epidemiologically, and lie confidence intervals of observed risks.
outside the realm of this chapter. Another, and distinct, role for multicenter
studies is the identification of unknown
CONCLUSION etiologic factors in diseases that show
That the past is prologue to the future is the marked geographic variation or diseases
modus operandi of academic admissions com- associated with diets or customs that are
mittees and other groups selecting candidates for limited in their geographic distribution.
advancement to new responsibilities. Exceptions In this context centers would be selected
to the success of such selections abound and for their heterogeneity, of either disease
are often highlighted, principally because they rate or environment, rather than their
are exceptions but the exceptions come about homogeneity.
in divergent ways that are difficult to predict. 4. Populations in which specific
What does the past offer in the way of predictive environmental and social characteristics
21

Accomplishments in Cancer Epidemiology 21

are ascertained for linkage to later health Ahlgren M, Sorensen T, Wohlfart J, Hafldottir A,
outcomes need to be expanded. Groups Holst C, Melbye M. Birth weight and risk of breast
of health practitioners established in cancer in a cohort of 106,540 women. Int J Cancer
Britain and the United States and the 2003;107:997–​1000.
population of American Cancer Society Aksoy M, Erdem S, DinCol G. Leukemia in shoe-​
volunteers illustrate the productivity of workers exposed chronically to benzene. Blood
such populations in the identification 1974;44:837–​41.
of possible cancer causes that are not Andersen A, Berge SR, Engeland A, Norseth T.
Exposure to nickel compounds and smoking in
recorded routinely.
relation to incidence of lung and nasal cancer
5. The study of cancer in occupational
among nickel refinery workers. Occup Environ
settings has been highly productive.
Med 1996;53:708–​13.
While such studies appear to be near Anderson GL, Judd HL, Kaunitz AM, Barad DH,
the end of their potential for identifying Beresford SA, Pettinger M et al. Effects of estrogen
unknown carcinogens, industry continues plus progestin on gynecologic cancers and associated
to develop new chemicals and processes diagnostic procedures: the Women’s Health Initiative
of which the effects of long-​term exposure randomized trial. JAMA 2003;290:1739–​48.
have to be monitored. In addition, Anderson GL, Limacher M, Assaf AR, Bassford T,
occupational studies will remain relevant Beresford SA, Black H. Effects of conjugated
for a long time to monitor the efficacy of estrogen in postmenopausal women with hyster-
preventive measures. ectomy; the Women’s Health Initiative random-
6. There is need for much stronger resources ized controlled trial. JAMA 2004;291:1701–​12.
than now exist for detecting associations Archer VE. Underground mining, smoking, and lung
between use of medications and cancer (Letter to the Editor). J Natl Cancer Inst
unexpected outcomes, including cancer—​ 1986;76:553–​4.
particularly for medications taken over Archer VE, Coons T, Saccomanno G, Hong DY.
long periods of time, and for diseases Latency and the lung cancer epidemic among
such as cancer that have long incubation United States uranium miners. Health Phys
periods. 2004:87:480–​9.
Armstrong BK, de Klerk NH, Musk AW, Hobbs MS.
Mortality in miners and millers of crocidolite in
It would be unfortunate if the recitation of
Western Australia. Br J Ind Med 1988;45:5–​13.
accomplishments in this review were taken as an
Baron JA. Epidemiology of non-​ steroidal anti-​
invitation to complacency. In the unlikely event
inflammatory drugs and cancer. Prog Exp Tumor
that all of the advances described here were to Res 2003;37:1–​24.
come to full fruition—​that is, if morbidity and Bartrip PWS. History of asbestos related disease.
mortality from cancers of the lung, ionizing radi- Postgrad Med J 2004;80:72–​6.
ation, known occupational carcinogens, and can- Beasley RP, Hwang LY, Lin CC, Chien CS.
cers of the oral cavity, pharynx, stomach, cervix, Hepatocellular carcinoma and hepatitis B virus: a
and liver were to be eliminated—​the incidence of prospective study of 22,707 men in Taiwan. Lancet
cancer would be reduced by about 20% and mor- 1981;ii:1129–​33.
tality by about 35% (Jemal et al, 2006). Cancers of Bengtsson U, Angervall L. Analgesic abuse and tumours
the breast, endometrium, ovary, prostate, bowel, of the renal pelvis (Letter). Lancet 1970;i:305.
pancreas, and brain as well as the leukemias and Beral V. Cancer of the cervix: a sexually transmitted
lymphomas would remain as major causes of disease? Lancet 1974;1037–​40.
morbidity and mortality. Beral V, Bull D, Reeves G; Million Women Study
Collaborators. Endometrial cancer and hormone-​
REFERENCES replacement therapy in the Million Women Study.
Abbatt JD. History of the use and toxicity of thorotrast. Lancet 2005;365:1543–​51.
Environ Res 1979;18:6–​12. Beral V, Bull D, Reeves G, Million Women Study
Abratt RP, Vorobiof DA, White N. Asbestos and mes- Collaborators. Endometrial cancer and hormone-​
othelioma in South Africa. Lung Cancer 2004;45 replacement therapy in the Million Women Study.
Suppl 1:53–​6. Lancet 2005;365:1543–​51.
Ahlgren M, Melbye M, Wohlfart J, Sorensen TI. Beral V, Newton R. Overview of the epidemiology
Growth patterns and the risk of breast cancer in of immunodeficiency-​associated disorders. Natl
women, N Engl J Med 2004;351:1619–​26. Cancer Inst Monogr 1998;23:1–​6.
2

22 Textbook of Cancer Epidemiology

Beresford SAA. Johnson KC, Ritenbaugh C, Lasser Brinton LA, Berman ML, Mortel R, Twiggs LB,
NL, Snetselaar LG, Black HR et al. Low-​fat dietary Barrett RJ, Wilbanks GD et al. Reproductive, men-
pattern and risk of colorectal cancer. The Women’s strual, and medical factors for endometrial can-
Health Initiative randomized controlled dietary cer: results from a case-​control study. Am J Obstet
modification trial. JAMA 2006;295:643–​54. Gynec 1992;167:1317–​25.
Blair A, Stewart PA, Zaebst DD, Pottern L, Zey JN, Burnet M. Somatic mutation and chronic disease.
Bloom TF et  al. Mortality of industrial workers BMJ 1965;i:338–​42.
exposed to acrylonitrile. Scand J Work Environ Calle EE, Rodriguez C, Walker-​Thurmond K, Thun
Heath 1998;24 Suppl 2:24–​41. MJ. Overweight, obesity, and mortality from can-
Blumberg BS, Alter HJ, Visnich S. A “new” antigen in cer in a prospectively studied cohort of U.S. adults.
leukemia sera. JAMA 1965;191:541–​6. New Engl J Med 2003;348:1625–​38.
Boffetta P, Hashibe M. Alcohol and cancer. Lancet Case RAM, Hosker ME, McDonald DB, Pearson JT.
Oncology 2006;7:149–​56. Tumours of the urinary bladder in workmen
Boffetta P, Hashibe M, La Vecchia C, Zatonski W, engaged in the manufacture and use of certain
Rehm J. The burden of cancer attributable to alco- dyestuff intermediates in the British chemical
hol drinking. Int J Cancer 2006;11:884–​7. industry. Br J Indust Med 1954;11:75–​104.
Boffetta P, Matisane L, Mundt KA, Dell LD. Meta-​ Case RAM, Lea AJ. Mustard gas poisoning, chronic
analysis of studies of occupational exposure to bronchitis and lung cancer. Br. J Prev Soc Med
vinyl chloride in relation to cancer mortality. 1955;9:62–​72.
Scand J Work Environ Health 2003;29:220–​9. Castellsagué X, Muñoz N, De Stefani E, Victora CG,
Boice JD Jr, Blettner M, Kleinerman RA, Engholm G, Castelletto R, Rolón PA et  al. Independent and
Stovall M, Lisco H et al. Radiation dose and breast joint effects of tobacco smoking and alcohol
cancer risk for patients treated for cancer of the drinking on the risk of esophageal cancer in men
cervix. Int J Cancer 1989;44:7–​16. and women. Int J Cancer 1999;82:657–​64.
Bonn D. Asbestos—​ the legacy lives on. Lancet Chang MH, Chen CJ, Lai MS, Hsu HM, Wu TC, Kong
1999;353:1336. MS et  al. Universal hepatitis B vaccination in
Bosch FX, Lorinez A, Muñoz N, Meijer CJLM, Shah KV. Taiwan and the incidence of hepatocellular carci-
The causal relation between human papillomavirus noma in children. Taiwan Childhood Hepatoma
and cervical cancer. J Clin Path 2002;55:244–​65. Study Group. N Engl J Med 1997;336:1855–​9.
Bosch FX, Manos M, Muños N, Sherman M, Jansen Chen CJ, Wang CJ. Ecological correlation between
AM, Peto J et al. Prevalence of human papilloma- arsenic level and well water and age-​ adjusted
virus cervical cancer:  a worldwide perspective. J mortality from malignant neoplasms. Cancer Res
Natl Cancer Inst 1995;87:796–​802. 1990;50:5470–​4.
Bosetti C, La Vecchia C, Lipworth L, McLaughlin JK. Chen CL, Hsu LI, Chiou HY, Hsueh YM, Chen SY,
Occupational exposure to vinyl chloride and can- Wu MM et  al. Ingested arsenic, cigarette smok-
cer risk: a review of the epidemiologic literature. ing, and lung cancer risk:  a follow-​ up study
Eur J Cancer 2003a;12:427–​30. in arseniasis-​ endemic areas in Taiwan, JAMA
Bosetti C, Negri E, Fattore E, La Vecclia C. Occupational 2004;292:3026–​9.
exposure to polychlorinated biphenyls and cancer Chien YC, Jan CF, Kuo HS, Chen CJ. Nationwide
risk. Eur J Cancer Prev 2003b;12:251–​5. hepatitis B vaccination program in Taiwan: effec-
Bosetti C, Negri E, Trichopoulos D, Franceschi S, tiveness in the 20  years after it was launched.
Beral V, Tzonou A et al. Long-​term effects of oral Epidemiol Rev 2006;28:126–​35.
contraceptives on ovarian cancer risk. Int J Cancer Chiou HY, Chiou ST, Hsu YH, Chou YL, Tseng CH,
2002;102:262–​5. Wei ML et  al. Incidence of transitional cell car-
Boyle P. Tobacco smoking and the British doctors’ cinoma and arsenic in drinking water:  a study
cohort (Editorial). Br J Cancer 2005;92:419–​20. of 8,102 residents in an areniasis-​ endemic
Braun L, Kisting S. Asbestos-​related disease in South area in Northeastern Taiwan. Am J Epidemiol
Africa. The social production of an invisible epi- 2001;153:411–​8.
demic. Am J Pub Health 2006;96:1386–​96. Clifford GM, Polesel J, Rickenbach M, Maso LD,
Bradford Hill A. Report to the Mond Nickel Company Keiser O, Kofler A et al. Cancer risks in the Swiss
(1939). HIV cohort study: associations with immunodefi-
Brinton LA. Epidemiology of cervical cancer—​ ciency, smoking, and highly active antiviral ther-
overview. In:  Muñoz N, Bosch FX, Shah KV, apy. J Natl Cancer Inst 2005;97:425–​32.
Meheus A. (eds.) The Epidemiology of Cervical Cole P, Rodu B. Epidemiologic studies of chrome and
Cancer and Human Papilloma Viruses. IARC cancer mortality:  a series of meta-​analyses. Reg
Publ 119. Lyon: IARC, 1992: 3–​23. Toxicol Pharmacol 2005; 43:225–​31.
23

Accomplishments in Cancer Epidemiology 23

Collaborative Group. (The Endogenous Hormones Doll R. Cancer of the lung and nose in nickel workers.
and Breast Cancer Collaborative Group.) Br J Indust Med 1958;15:217–​23.
Endogenous sex hormones and breast can- Doll R, Hill AB. Smoking and cancer of the lung: pre-
cer in postmenopausal women:  reanalysis of liminary report. BMJ 1950;2:739–​48.
nine prospective studies. J Natl Cancer Inst Doll R, Hill AB. The mortality of doctors in relation to
2002a;94:606–​16. their smoking habits:  a preliminary report. BMJ
Collaborative Group on Hormonal Factors in Breast 1954;228:1451–​55.
Cancer. Breast cancer and hormonal contracep- Doll R, Mathews JD, Morgan LG. Cancers of the lung
tives:  collaborative reanalysis of individual data and nasal sinuses in nickel workers:  a reassess-
on 53,297 women with breast cancer and 100 239 ment of the period of risk. Br J Ind Med 1977;
women without breast cancer from 54 epidemio- 34:102–​5.
logic studies. Lancet 1996;347:1713–​27. Doll R, Peto R, Boreham J, Sutherland I. Mortality
Collaborative Group on Hormonal Factors in Breast in relation to smoking: 50 years’ observations on
Cancer. Breast cancer and hormone replacement male British doctors. BMJ 2004;438:1519–​28.
therapy:  collaborative reanalysis of data from 51 Doll R, Peto R, Boreham J, Sutherland I. Mortality
epidemiological studies of 52,705 women with from cancer in relation to smoking:  50  years
breast cancer and 108,411 women without breast observations on British doctors. Br J Cancer
cancer. Lancet 1997;350:1047–​59. 2005;92:426–​9.
Collaborative Group on Hormonal Factors in Breast dos Santos Silva I, Malveiro F, Jones ME, Swerdlow
Cancer and Valerie Beral. Breast cancer and breast AJ. Mortality after radiological investigation with
feeding:  collaborative reanalysis of individual radioactive thorotrast: a follow-​up study of up to
data for 47 epidemiological studies in 30 coun- 50 years in Portugal. Radiat Res 2003;159:521–​34.
tries, including 50,302 women with breast cancer Eliassen AH, Colditz GA, Rosner B, Willett WC,
and 93,973 women without the disease. Lancet Hankinson SE. Adult weight change and
2002b;360:187–​95. risk of postmenopausal breast cancer. JAMA
Collins JJ, Acquavella JF. Review and meta-​analysis 2006;296:193–​201.
of studies of acrylonitrile. Scand J Work Environ English DR, Armstrong BK, Kricker A. Sunlight and
Health 1998; 24 Suppl 2:671–​80. cancer. Cancer Causes Control 1997;8:271–​83.
Collins JJ, Acquavella JF, Esmen NA. An updated Enterline PE, Harley J, Henderson V. Asbestos and
meta-​ analysis of formaldehyde exposure and cancer:  a cohort followed up to death. Br J Ind
upper respiratory tract cancers. J Occup Environ Med 1987;44:396–​401.
Med 1997;39:639–​51. Eslick GD. Helicobacter pylori infection causes gas-
Cologne JB, Tokuoka S, Beebe GW, Fukuhara T, tric cancer? A review of the epidemiologic, meta-​
Mabuchi K. Effects of radiation on incidence of analytic, and experimental evidence. World J
primary liver cancer among atomic bomb survi- Gastroenterol 2006;12:2991–​9.
vors. Radiat Res 1999;152:364–​73. Fioretti F, Bosetti C, Tavani A, Franceschi S, La Vecchia
Corrao, G, Bagnardi V, Zambon A, La Vecchia C. A C. Risk factors for oral and pharyngeal cancer in
meta-​analysis of alcohol consumption and the risk never smokers. Oral Oncol 1999;35:375–​8.
of 15 diseases. Prev Med 2004;38:613–​9. Frumkin H, Thun MJ. Arsenic. CA Cancer J Clinic
Crew KD, Neugut AI. Epidemiology of gastric cancer. 2001;51;254–​62.
World J Gastroenterol 2006:12:354–​62. Garcia Rodriguez LA, Huerta-​Alvarez. Reduced risk
Crosignani P, Tittarelli A, Borgini A, Codazzi T, of colorectal cancer among long-​term users of
Rovelli A, Porro E, et  al. Childhood leukemia aspirin and nonaspirin and nonsteroidal antiin-
and road traffic: a population-​based case-​control flammatory drugs. Epidemiology 2001;12:88–​93.
study. Int J Cancer 2004;108:596–​9. Gilliland FD, Hunt WC, Archer VE, Saccomano G.
Darby SC, Whitley E, Howe GR, Hutchings SJ, Kusiak Radon progeny exposure and lung cancer risk
RA, Lubin JH et al. Radon and cancers other than among non-​ smoking uranium miners. Health
lung cancer in underground miners:  a collab- Phys 2000;79:365–​72.
orative analysis of 11 studies. J Natl Cancer Inst Giovannucci E, Egan KM, Hunter DJ, Stampfer MJ,
1995;87:378–​84. Colditz GA, Willett WC et  al. Aspirin and the
Demers PA, Boffetta P. Cancer Risk from Occupational risk of colorectal cancer in women. N Engl J Med
Exposure to Wood Dust:  a Pooled Analysis of 1995;333:609–​14.
Epidemiological Studies. IARC Technical Report Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA,
No. 30 Lyon: IARC, 1998. Asherio A, Willett WC. Aspirin use and the risk
Doll R. Mortality from lung cancer in asbestos work- for colorectal cancer and adenoma in male health
ers. Br J Indust Med 1955;12:81–​6. professionals. Ann Intern Med 1994;121:241–​6.
24

24 Textbook of Cancer Epidemiology

Glass DC, Gray CN, Jolley DJ, Gibbons C, Sim MR, Henderson BE, Casagrande JT, Pike MC, Mack T,
Fritschi L et  al. Leukemia risk associated with Rosario I, Duke A. The epidemiology of endo-
low-​level benzene exposure. Epidemiology metrial cancer in young women. Br J Cancer
2003;14:569–​77. 1983;47:749–​56.
Gonzales-​ Perez A, Garcia Rodriguez LA, Lopez-​ Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma
Ridaura R. Effects of non-​ steroidal anti-​ of the vagina:  association of maternal stilbestrol
inflammatory drugs on cancer sites other than the therapy with tumor appearance in young women.
colon and rectum: a meta-​analysis. BMC Cancer N Engl J Med 1971;284:878–​81.
2003;3:28. Hirayama T. Non-​smoking wives of heavy smokers
Grady D, Gebretsadik T, Kerlikowske K, Ernster V, have a higher risk of lung cancer:  a study from
Petitti D. Hormone replacement therapy and Japan. BMJ 1981;282:183–​5.
endometrial cancer risk:  a meta-​analysis. Obstet Hoover R, Gray LA, Cole P, MacMahon B. Menopausal
Gynecol 1995;85:304–​13. estrogens and breast cancer. N Engl J Med 1976;
Grimsrud TK, Berge SR, Haldorsen T, Andersen A. 295:401–​5.
Can lung cancer risk among nickel refinery work- Hung H-​C, Joshipura KJ, Jiang R, Hu FB, Hunter D,
ers be explained by occupational exposures other Smith-​Warner SA et al. Fruit and vegetable intake
than nickel? Epidemiology 2005;16:146–​54. and risk of major chronic disease. J Natl Cancer
Grimsrud TK, Berge SR, Martinsen JI, Andersen A. Inst 2004;96:1577–​84,
Lung cancer incidence among Norwegian nickel-​ IAEA (International Atomic Energy Agency).
refinery workers 1953-​ 2000. J Environ Monit Thorotrast:  A Bibliography of its Diagnostic Use
2003;5:109–​7. and Biological Effects. Vienna: IAEA, 1965.
Grimsrud TK, Peto J. Persisting risk of nickel-​related IARC (International Agency for Research on Cancer).
lung cancer and nasal cancer among Clydach Arsenic and Arsenic Compounds. Monographs on
refiners. Occupat Environ Med 2006;63:365–​6. the Evaluation of Carcinogenic Risk of Chemicals
Hackshaw AK, Law MR, Wald NJ. The accumulated to Humans, Vol. 23. Lyon: IARC, 1980.
evidence on lung cancer and environmental IARC. Tobacco Smoking and Involuntary Smoking.
tobacco smoke. BMJ 1997:315:980–​8. Monographs on the Evaluation of Carcinogenic
Haenszel W, Kurihara M. Studies of Japanese migrants. Risks to Humans, Vol. 83. Lyon: IARC, 2004a.
I. Mortality from cancer and other diseases among IARC. Formaldehyde, 2-​ Butoxyethanol and I-​
Japanese in the United States. J Nat Cancer Inst tert-​Butoxy-​
2-​propanol. Monographs on the
1968;40:43–​68. Evaluation of Carcinogenic Risks to Humans, Vol.
Hamajima N, Tajima HK, Rohan T, Calle EE, Heatth 88. Lyon: IARC, 2004b.
CW Jr, Coates RJ et al. Alcohol, tobacco and breast Jacobsson R, Ahlbom A, Bellander T, Lundberg I.
cancer—​collaborative analysis of individual data Acute myeloid leukemia among petrol station
from 53 epidemiological studies, including 58,515 attendants. Arch Environ Health 1993;48:255–​9.
women with breast cancer and 95,067 without the Jemal A, Siegel R, Ward E, Murray T, Xu J, Smigal C,
disease. Br J Cancer 2002;87:1234–​45. Thun MJ. Cancer Statistics, 2006. CA:  Cancer J
Hammond EC. Smoking in relation to the death rates Clin 2006;56:106–​30.
of one million men and women. Monogr Natl Jones EG, MacDonald I, Breslow L. A study of epide-
Cancer Inst 1966;19:127–​204. miologic factors in carcinoma of the uterine cer-
Hankinson SE, Willett WC, Manson JE, Hunter DJ, vix. Am J Obstet Gynecol 1958;76:1–​10.
Colditz GA, Stampfer MJ et  al. Alcohol, height, Josefsson AM, Magnusson PK, Ylitalo N, Sorensen P,
and adiposity in relation to estrogen and prolactin Quarforth-​Tubbin P, Andersen PK et al. Viral load
levels in postmenopausal women. J Natl Cancer of human papilloma virus 16 as a determinant for
Inst 1995;87:1297–​303. development of cervical carcinoma in situ: a nested
Hauptmann M, Lubin JH, Stewart PA, Hayes RB, Blair case-​control study. Lancet 2000;355:2189–​93.
A. Mortality from lymphohematopoietic malig- Kahlenborn C, Modugno F, Potter DM, Severs WB.
nancies among workers in formaldehyde indus- Oral contraceptive use as a risk factor for premen-
tries. J Natl Cancer Inst 2003;95:1615–​23. opausal breast cancer: a meta-​analysis. Mayo Clin
Hauptmann M, Lubin JH, Stewart PA, Hayes RB, Blair Proc 2006;81:1290–​302.
A. Mortality from solid cancers among work- Kest H, Brogly S, McSherry G, Dashefsky B, Oleske J,
ers in formaldehyde industries. Am J Epidemiol Seage GR 3rd. Malignancy in perinatally human
2004;159:1117–​30. immunodeficiency virus–​infected children in the
Hayes RB, Songnian Y, Dosemeci M, Linet M. United Sates. Pediat Infect Dis J 2005;24:237–​42.
Benzene and lymphohematopoietic malignancies Key T, Appleby P, Barnes I, Reeves G, Endogenous
in humans. Am J Ind Med 2001;40:117–​26. Hormones and Breast Cancer Collaborative
25

Accomplishments in Cancer Epidemiology 25

Group. Endogenous sex hormones and breast Lorenz E. Radioactivity and lung cancer:  a critical
cancer in postmenopausal women:  reanalysis review of lung cancer in the miners of Schneeberg
of nine prospective studies. J Natl Cancer Inst and Joachimsthal. J Natl Cancer Inst 1944;5:1–​15.
2002;94:606–​16. Lubin JH, Boice JD Jr, Edling C, Hornung RW, Howe
Kricker A, Armstrong BK, English DR. Sun exposure GR, Kunz E et al. Lung cancer in radon-​exposed
and non-​melanotic skin cancer. Cancer Causes miners and estimates of risk from indoor expo-
Control 1994;5:367–​92. sures. J Natl Cancer Inst 1995;87: 817–​27.
Krugman S, Giles JP, Hammond J. Infectious hepati- Lubin JH, Pottern LM, Stone BJ, Fraumeni JF Jr.
tis: evidence for two distinctive clinical. epidemi- Respiratory cancer in a cohort of copper smelter
ological, and immunological types of infection. workers: results from more than 50 year of follow-​
JAMA 1967;200:95–​103. up. Am J Epidemiol 2000;151:554–​65.
La Vecchia C, Decarli A, Franceschi S, Regalio M, Lynge E, Andersen A, Nilsson R, Barlow L, Pukkala E,
Tognoni H. Age at first birth and the risk of Nordlinder R et al. Risk of cancer and exposure to
epithelial ovarian cancer. Int J Cancer 1984; gasoline vapors. Am J Epidemiol 1997;145:449–​58.
73:663–​6. MacMahon B, Cole P, Lin TM, Lowe CR, Mirra AP,
Lagiou P, Adami HO, Trichopoulos D. Early life diet Ravnihar B et al. Age at first birth and breast cancer
and the risk for adult breast cancer. Nutr Cancer risk. Bull World Health Organ 1970a;43:209–​21.
2006;56:158–​61. MacMahon B, Lin TM, Lowe CR, Mirra AP, Ravnihar
Lagiou P, Hsieh CC, Trichopoulos D, Xu B, Wuu J, B, Salber EJ et  al. Lactation and cancer of the
Mucci L et  al. Birthweight differences between breast: a summary of an international study. Bull
USA and China and their relevance to breast can- World Health Organ 1970b; 42:185–​94.
cer aetiology. Int J Epidemiol 2003;32:193–​8. Marsh GM. Critical review of epidemiologic stud-
Lagiou P, Trichopoulou A, Trichoroulos D. Nutritional ies related to ingested asbestos. Environ Health
epidemiology of cancer:  accomplishments and Perspect 1983;53:49–​56.
prospects. Proc Nutr Soc 2002;61:217–​22. Marshall BJ, Warren RJ. Unidentified curved bacilli in
Lagorio S, Forastiere F, Iavarone I, Rapiti E, Vanacore the stomach of patients with gastritis and peptic
N, Perucci CA et  al. Mortality of filling sta- ulceration. Lancet 1984;i:1311–​5.
tion attendants. Scand J Work Environ Health Martland HS, Conlon P, Knef LF. Some unrecognized
1994;20:331–​8. dangers in the use and handling of radioactive
Lambe M, Hsieh C-​C, Trichopoulos D, Ekbom A, substances. JAMA 1925;85:1769.
Pavia M, Adami HO. Transient increase in the risk Mastrangelo G, Fedeli U, Fadda E, Milan G, Turato A,
of breast cancer after giving birth. N Engl J Med Pavanello S. Lung cancer risk in workers exposed
1994;331:5–​9. to poly(vinyl chloride) dust: a nested case-​referent
Lambe M, Wuu J, Weiderpass E, Hsieh CC. Childbearing study. Occupat Environ Med 2003;60:423–​8.
at older age and endometrial cancer risk (Sweden). Mastromatteo E. Nickel: a review of its occupational
Cancer Causes Control 1999;10:43–​9. health aspects. J Occup Med 1967;9:127–​36.
Lane-​Claypon JE. A Further Report on Cancer of the Maupas P, Melnick JL. Hepatitis B infection and pri-
Breast with Special Reference to Its Associated mary liver cancer. Prog Med Virol 1981;27:1–​5.
Antecedent Conditions. Report on Public Health McCredie M, Coated MS, Ford JM. Cancer incidence
and Medical Subjects. No. 32. London: Ministry of among migrants to New South Wales. Int J Cancer
Health, 1926. 1990;46:228–​32.
Le Marchand L, Saltzman BS, Hankin JH, Wilkens LR, McCredie M, Pommer W, McLaughlin JK, Stewart
Franke AA, Morris SJ et al. Sun exposure, diet, and JH, Lindblad P, Mandel JS et  al. International
melanoma in Hawaii Caucasians. Am J Epidemiol renal-​cell cancer study. II. Analgesics. Int J Cancer
2006;164:232–​45. 1995:60:345–​9.
Lee MS, Kim DM, Kim H, Lee HS, Kim CY, Park TS McCredie M, Stewart JH, Day NE. Different roles for
et al. Hepatitis B vaccination and reduced risk of phenacetin and paracetamol in cancer of the kid-
primary liver cancer among male adults: a cohort ney and renal pelvis. Int J Cancer 1993;53:245–​9.
study in Korea. Int J Epidemiol 1988;27:316–​9. McDonald JC, Liddell FD, Dufresne A, McDonald
Little MP, Blettner M, Boice RD Jr, Bridges BA, Cardis AD. The 1891–​1920 birth cohort of Quebec chrys-
E, Charles MW et  al. Potential funding crises otile miners and millers: mortality 1976–​88. Br J
at the Radiation Effects Research Foundation. Ind Med 1993;50:1073–​81.
Lancet 2004;364:557–​8. McLaughlin JK, Lipworth L. A critical review of the
Longnecker MP. Alcoholic beverage consumption in epidemiologic literature on health effects of occu-
relation to risk of breast cancer: meta-​analysis and pational exposure to vinyl chloride. J Epidemiol
review. Cancer Causes Control 1994;5:73–​82. Biostat 1999;4:253–​75.
26

26 Textbook of Cancer Epidemiology

Mocroft A, Katlama C, Johnson AM, Pradier chromate industry: a quantitative risk assessment.
C, Antunes F, Mulcahy F et  al. AIDS across Risk Anal 2004;24:1099–​108.
Europe, 1994–​98:  the EUROSIDA study. Lancet Park Y, Hunter DJ, Spiegelman D, Bergkvist L, Berrino
2000;356:291–​6. F, van den Brandt PA et al. Dietary fiber intake and
Mori MT, Kido C, Fukutomi K, Kato Y, Hatakeyama risk of colorectal cancer: a pooled analysis of pro-
S, Machinami R et  al. Summary of the entire spective cohort studies. JAMA 2005;294:2904–​6.
Japanese thorotrast follow-​ up study:  updated Parkin DM. The global burden of infection-​
1998. Radiat Res 1999;152:S84–​7. associated cancers in the year 2002. Int J Cancer
Muller BU, Pizzo PA. Cancer in children with pri- 2006;118:3030–​44.
mary or secondary immunodeficiencies. J Pediat Parkin DM, Whelan SL, Ferlay J, Storm H. Cancer
1995;126:1–​10. Incidence in Five Continents. Vols. 1–​ 8.
Muller FH. Tabakmissbrauch und lungencarcinoma. CancerBase No. 7. Lyon: IARC, 2005.
Z. Krebsforsch 1940;49:57–​85. Parodi S, Vercelli, M, Stella A, Stagnaro E, Valerio F.
Mundt KA, Dell LD, Austin RP, Luippold RS, Noess R, Lymphohaematopoietic system cancer incidence
Bigelow C. Historical cohort study to 31 December in an urban area near a coke oven plant; an eco-
1995 of 10,109 men in the North American vinyl logic investigation. Occupat Environ Med 2003;
chloride industry, 1842–​72: update of cancer mor- 60:187–​93.
tality. Occup Environ Med 2000;57:774–​81. Parsonnet J, Forman D. Helicobacter pylori infection
Mzileni O, Sitas F, Steyn K, Carrara H, Bekker P. Lung and gastric cancer—​for want of more outcomes.
cancer, tobacco, and environmental factors in the JAMA 2004;291:244–​5.
African population of the Northern Province, Patel AS, Talbott EO, Zborowski JV, Rychek JA, Dell
South Africa. Tobacco Control 1999;8:398–​401. D, Xu X et  al. Risk of cancer as a result of com-
Nasti G, Talamini R, Antinori A, Martellotta F, munity exposure to gasoline vapors. Arch Environ
Jacchetti G, Chiodo G. AIDS-​ related Kaposi’s Health 2004;59:497–​503.
sarcoma:  evaluation of potential new prognostic Pedersen E, Høgetveit AC, Andersen A. Cancer of res-
factors and assessment of the AIDS Clinical Trial piratory organs among workers at a nickel refinery
Group Staging System in the HAART era—​the in Norway Int J Cancer 1973;12:32–​41.
Italian Cooperative Group on AIDS and Tumors Pedotti P, Cardillo M, Rossini G, Arcuri V, Boschiero
and the Italian cohort of patients naive from anti- L, Caldara R et al. Incidence of cancer after kidney
retrovirals. J Clin Oncol 2003;21:2876–​82. transplant: results from the North Italy transplant
Neuberger M, Kundi M, Friedl HP. Environmental program. Transplantation 2003;76:1448–​51.
asbestos exposure and cancer mortality. Arch Penn I. Cancer in renal transplant recipients. Adv Ren
Environ Health 1984;39:261–​5. Replace Ther 2000;7:147–​56.
Newcomb PA, Titus-​Ernstoff L, Egan KM, Trentham-​ Persson I, Adami HO, Bergkvist L, Lindgren A,
Dietz A, Baron JA, Storer BE et al. Postmenopausal Pettersson B, Hoover R et al. Risk of endometrial
estrogen and progestin use in relation to breast cancer after treatment with oestrogens alone or in
cancer risk. Cancer Epidemiol Biomarkers Prev conjunction with progestagens:  results of a pro-
2002;11:593–​600. spective study. BMJ 1989;298:147–​51.
Newnham A, Harris J, Evans HS, Evans BG, Moller Peto J, Decarli A, La Vecchia C, Levi, Jones JR. The
H. The risk of cancer in HIV-​infected people in European mesothelioma epidemic. Br J Cancer
southeast England:  a cohort study. Br J Cancer 1999;79:666–​72.
2005;92:194–​200. Peto J, Seidman H, Selikoff IJ. Mesothelioma mortal-
Nyberg U, Nilsson G, Travis LB, Holm LE. Hall ity in asbestos workers: implications for models of
P. Cancer incidence among Swedish patients carcinogenesis and risk assessment. Br J Cancer
exposed to radioactive thorotrast:  a forty-​ year 1982;45:124–​35.
follow-​up survey. Radiat Res 2002;157:419–​25. Pinkerton LE, Hein MJ, Stayner LT. Mortality
Oliveria SA, Saraiya M, Geller AC, Heneghan MK, among a cohort of garment workers exposed to
Jorgensen C. Sun exposure and risk of mesotheli- formaldehyde:  an update. Occup Environ Med
oma. Arch Dis Child 2006 2006;91:131–​8. 2004;61:193–​200.
Palmer JR, Wise, LA, Hatch EE, Troisi R, Titus-​ Pott P. Chirurgical Observations. London:  Hawes,
Ernstoph L, Strohsnitter W et  al. Prenatal Clarke & Collings, 1775.
diethylstilbestrol exposure and risk of breast Prentice RL, Chlebowski RT, Patterson R, Kuller LH,
cancer. Cancer Epidemiol Biomarkers Prev Ockene JK, Margolis KL et al. Low-​fat dietary pat-
2006;15:1509–​14. tern and risk of invasive breast cancer: the Women’s
Park RM, Bena JF, Stayner LT, Smith RJ, Gibb HJ, Lees Health Initiative randomized controlled dietary
PS. Hexavalent chromium and lung cancer in the modification trial. JAMA 2006;295:629–​42.
27

Accomplishments in Cancer Epidemiology 27

Price B, Ware A. Mesothelioma trends in the women:  principal results from the Women’s
United States:  an update based on Surveillance, Health Initiative randomized controlled trial.
Epidemiology, and End Results Program JAMA 2002;288:321–​3.
data for 1973 through 2003. Am J Epidemiol Ruder AM, Hein MJ, Nilsen N, Waters MA, Laber
2004;159:107–​12. P, Davis-​King K et  al. Mortality among work-
Prince MM, Hein MJ, Ruder AV, Waters MA, Laber ers exposed to polychlorinated biphenyls (PCBs)
PA, Whelan EA. Update:  cohort mortality study in an electrical capacitor manufacturing plant
of workers highly exposed to polychlorinated bio- in Indiana:  an update. Environ Health Perspect
phenyls (PCBs) during manufacture of electrical 2006;114:18–​23.
capacitors, 1940–​ 1998. Environmental Health Russo J, Gusterson BA, Rogers AE, Russo IH, Welling
Global Access 2006;114:13–​24. SR, van Zweiten MJ. Comparative study of human
Prince MM, Szmuness W, Micon J, Demaille J, and rat mammary carcinogenesis. Lab Invest
Diebolt G. Linhard, J et  al. A case-​control study 1990;62:224–​78.
of the association between primary liver cancer Samet JM. Radon and lung cancer. J Natl Cancer Inst
and hepatitis B infection in Senegal. Int J Cancer 1989;81:745–​57.
1975;16:376–​83. Scélo G, Constantinescu V, Csiki I, Zaridze D,
Proctor DM, Otani JM, Finley BL, Paustenbach DJ, Szeszenia-​Dabrowska N, Rudnai P et  al.
Bland JA, Speizer N et  al. Is hexavalent chro- Occupational exposure to vinyl chloride, acrylo-
mium carcinogenic via ingestion? A  weight-​ nitrile and styrene and lung cancer risk (Europe).
of-​evidence review. J Toxicol Environ Health A Cancer Causes Control 2004;15:445–​52.
2002;65:701–​46. Schiffman M, Wheeler CM, Castle PE. Human pap-
Raaschou-​ Nielsen O, Hertel O, Thomsen BL, illomavirus DNA remains detectable longer
Olsen JH. Air pollution from traffic at the resi- than related cervical abnormalities. J Infect Dis
dences of children with cancer. Am J Epidemiol 2002;186:1169–​72.
2001;153:433–​43. Schlatter C. Risks to consumers and PVC workers
Raffn E, Lynge E, Juel K, Korsgaard B. Incidence of from vinyl chloride. Schweiz Med Wochenschr
cancer and mortality among employees in the 1976;106:647–​50.
asbestos cement industry in Denmark. Br J Ind Schlesselman JJ. Net effect of oral contraceptive use on
Med 1989;46:90–​6. the risk of cancer in women in the United States.
Reichman ME, Judd JT, Longcope C, Schatzkin A, Obstet Gynec 1995;85:793–​801.
Clevidence BA, Nair PP et  al. Effects of alcohol Schnatter AR, Rosamilia K, Wojcik NC. Review of lit-
consumption on plasma and urinary hormone erature on benzene exposure and leukemia sub-
concentrations in premenopausal women. J Natl types. Chem Biol Interact 2005;153–​154:9–​21.
Cancer Inst 1993;85:722–​7. Segi M. Cancer mortality for selected sites in 24
Ribes J, Clèries R, Rubió A, Hernández JM, Mazzara countries, 1950–​7. With Kurihara M.  1958–​59,
R, Madoz P et  al. Cofactors associated with 1960–​61. Department of Public Health, Tohoku
liver disease mortality in an HBsAg-​ positive University School of Medicine. Sendai, Japan,
Mediterranean cohort:  20  years of follow-​up. Int 1962, 1964.
J Cancer 2006;119:687–​94. Selikoff IJ, Hammond EC, Churg J. Asbestos exposure
Rinsky RA, Young RJ, Smith AB. Leukemia in benzene and neoplasia. JAMA 1964;188:22–​6.
workers. Am J Ind Med 1981;2:217–​45. Selikoff IJ, Hammond EC, Churg J. Relation between
Rodu B, Cole P. Declining mortality from smok- exposure to asbestos and mesothelioma. N Engl J
ing in the United States. J Nicotine Tobacco Res Med 1965;272:560–​5.
2007;9:781–​4. Serraino D, Piselli P, Angeletti C, Minetti E, Pozzetto
Rosenberg L, Rao RS, Palmer JR, Strom BL, Zauber A, Civati G et al. Risk of Kaposi’s sarcoma and of
A, Warshauer ME et al. Transitional cell cancer of other cancers in Italian renal transplant patients.
the urinary tract and renal cell cancer in relation Br J Cancer 2005;92:572–​5.
to acetaminophen use (United States). Cancer Smith PG, Doll R. Mortality among patients with
Causes Control 1998;9:83–​8. ankylosing spondylitis after a single treatment
Rothman KJ, Keller A. The effect of joint expo- with x-​rays. BMJ 1982;284:449–​60.
sure to alcohol and tobacco on risk of can- Sorahan T, Williams SP. Mortality of workers at a
cer of the mouth and pharynx. J Chronic Dis nickel carbonyl refinery, 1958–​ 2000. Occup
1972;25:711–​6. Environ Med 2005;62:80–​5.
Rossouw JE, Anderson GL, Prentice RL, LaCroix AZ, Stefanick ML, Anderson GL, Margolis KL, Hendrix
Kooperberg C et  al. Risks and benefits of estro- SL, Rodabough RJ, Paskett ED et  al. Effects
gen plus progestins in healthy postmenopausal of conjugated equine estrogens on breast
28

28 Textbook of Cancer Epidemiology

cancer and mammography screening in post- tobacco smoking and ethanol consumption in the
menopausal women with hysterectomy. JAMA etiology of hepatocellular carcinoma. Int J Cancer
2006;295;1647–​57. 1987;39:45–​9.
Stewart A, Webb J, Hewitt D. A survey of childhood Tseng WP, Chu HM, How SW. Prevalence of skin can-
malignancies. BMJ 1958;i:1495–​508. cer in an endemic area of chronic arsenicism in
Sun CA, Wu DM, Lin CC, Lu SN, You SL, Wang LY Taiwan. J Natl Cancer Inst 1968;40:453–​63.
et al. Incidence and cofactors of hepatitis C virus–​ Tuyns AJ. Oesophageal cancer in non-​smoking drink-
related hepatocellular carcinoma:  a prospective ers and in non-​drinking smokers. Int J Cancer
study of 12,008 men in Taiwan. Am J Epidemiol 1983;32:443–​4.
2003:57:674–​82. Tuyns AJ, Esteve J, Raymond L, Berrino F, Benhamou
Surgeon-​ General. Smoking and Health. Report of E, Blanchet F et  al. Cancer of the larynx/​hypo-
the Advisory Committee to the Surgeon-​General pharynx, tobacco and alcohol:  IARC interna-
of the United States. Department of Health, tional case-​control study in Turin and Varese
Education and Welfare. Public Health Service (Italy), Zaragoza and Navarra (Spain), Geneva
Publication No. 1103. Washington, DC:  US (Switzerland) and Calvados (France). Int J Cancer
Government Printing Office 0-​714-​422, 1964. 1988;41:483–​91.
Surgeon General. The Health Consequences of Ulfelder H. Stilbestrol, adenosis, and adenocarci-
Involuntary Exposure to Tobacco Smoke. noma. Am J Obstet Gynecol 1973;117:794–​8.
A Report of the Surgeon General. US Department UNSCEAR, 2000. United Nations Scientific
of Health and Human Services. National Center Committee on the Effects of Ionizing Radiation.
for Chronic Disease and Health Promotion, Sources and Effects of Ionizing Radiation, Report
Atlanta, GA, 2006. to the General Assembly, with Scientific Annexes.
Taioli E, Piselli P, Arbustini E. Boschiero L, Burra P, New York: United Nations, 2000.
Busnach G et  al. Incidence of second primary van Kaick G, Dalheimer A, Hornik S, Kaul A,
cancer in transplanted patients. Transplantation Liebermann D, Luhrs H et  al. The German
2006;81:982–​5. thorotrast study: recent results and assessment of
Taylor R, Cumming R, Woodward A, Black M. Passive risks. Radiat Res 1999;152(suppl.6):S64–​71.
smoking and lung cancer:  a cumulative meta-​ Vaughan TL, Stewart OA, Teschke K, Lynch CF, Swanson
analysis. Aust NZ J Public Health 2001;25:203–​11. GM, Lyon JL et al. Occupational exposure to form-
Thun MJ, Day-​ Lally CA, Callee EE, Flander WD, aldehyde and wood dust and nasopharyngeal carci-
Health CW Jr. Excess mortality among cigarette noma. Occup Environ Med 2000;57:376–​84.
smokers:  changes in a 20-​ year interval. Am J Wagner JC. The discovery of the association between
Public Health 1995;85:1223–​30. blue asbestos and mesotheliomas and the after-
Thun MJ, Namboodiri MM, Heath CW. Aspirin use math. Br J Indust Med 1991;48:399–​403.
and reduced risk of fatal colon cancer. N Engl J Wagner JC, Sleggs CA, Marchand P. Diffuse pleu-
Med 1991;325:1593–​6. ral mesothelioma and asbestos exposure in the
Travis LB, Hill DA, Dores GM, Gospodarowicz M, North Western Cape province. Br J Indust Med
van Leeuwen FE, Holowaty E et al. Breast cancer 1960;17:260–​71.
following radiotherapy and chemotherapy among Ward E, Boffetta P, Andersen A, Colin D, Comba P,
young women with Hodgkin’s Disease. JAMA Deddens JA et al. Update of mortality and cancer
2003;290:465–​75. incidence among European workers employed
Travis LB, Hill D, Dores GM, Gospodarowicz M, van in the vinyl chloride industry. Epidemiology
Leewen FE, Holowaty E et  al. Cumulative abso- 2001;12:710–​8.
lute breast cancer risk for young women treated Weiderpass E, Adami HO, Baron JA, Magnusson
for Hodgkin’s lymphoma. J Natl Cancer Inst C, Bergstrom R, Lindgren A et al. Risk of endo-
2005;97:1428–​37. metrial cancer following estrogen replacement
Trichopoulos D. Hypothesis: does breast cancer origi- with and without progestins. J Natl Cancer Inst
nate in utero? Lancet 1990;335:939–​40. 1999a;91:1131–​7.
Trichopoulos D, Hsieh C-​C, MacMahon B, Lin T-​M, Weiderpass E, Adami HO, Baron JA, Magnusson C,
Lowe CR, Ravnihar B et al. Age at any birth and Lindgren A, Persson I. Use of oral contraceptives
breast cancer risk. Int J Cancer 1983;31:701–​4. and endometrial cancer risk (Sweden). Cancer
Trichopoulos D, Kalandidi A, Sparros L, MacMahon Causes Control 1999b;10:277–​8.
B. Lung cancer and passive smoking. Int J Cancer Whittemore AS, Harris R, Itnyre J. Characteristics
1981;27:1–​4. relating to ovarian cancer risk: collaborative anal-
Trichopoulos D, Day N, Kaklamani E, Tzonou A, ysis of 12 case-​control studies. II. Invasive epithe-
Muñoz N, Zavitsanos X et  al. Hepatitis B virus, lial ovarian cancer in white women. Collaborative
29

Accomplishments in Cancer Epidemiology 29

Ovarian Cancer Group. Am J Epidemiol Yin SN, Li GL, Tain FD, Fu ZI, Jin C, Chen YJ et al.
1992;136:1184–​203. Leukemia in benzene workers:  a retrospective
Willett WC. Diet and cancer: one view at the start of cohort study. Br J Ind Med 1987;44:124–​8.
the millennium. Cancer Epidemiol Biomarkers Ylitalo N, Sorensen P, Josefsson AM, Magnusson PK,
Prev 2001:10:3–​8. Andersen PK, Ponten J et al. Consistent high viral
Wong BC, Lam SK, Wong WM, Chen JS, Zheng TT, load of human papillomavirus 16 and risk of car-
Feng RE et al. Helicobacter pylori eradication to cinoma in situ: a nested case-​control study. Lancet
prevent gastric cancer in a high-​risk region of 2000;355:2194–​8.
China: a controlled trial. JAMA 2004;291;187–​94. Yu MW, Yeh SH, Chen PJ, Liaw YF, Lin CL, Liu CJ
Yarborough, CM. Chrysotile as a cause of mesothe- et  al. Hepatitis B virus-​genotype and DNA level
lioma:  an assessment based on epidemiology. and hepatocellular carcinoma: a prospective study
Critical Rev Toxicology 2006;36:165–​87. in men. J Natl Cancer Inst 2005;97:265–​72.
30
31

2
Measures and Estimates of Cancer Burden
P A O L O B O F F E T TA , D A N A H A S H I M , A N D P A G O N A   L A G I O U

T he social, economic and political climate

among countries and world regions is
dynamic and rapidly shifting. Global invest-
population. Nationwide cancer registries, which
typically contain both incidence and mortality
data, exist in a small number of countries. On
ments and resources must be allocated prudently the global scale, 290 cancer registries provide
to effectively reduce cancer burdens and improve coverage for 68 countries worldwide and rep-
health outcomes in moderately short time peri- resent cancer for 1.33 billion persons (Forman
ods. Averting premature mortality and high et al, 2013).
healthcare expenditure depends on a continuous Mortality data provided to the World Health
and consistently updated assessment of burden Organization (WHO) account for about 18.6 mil-
of disease, injuries, and risks. Comprehensive lion deaths annually, representing one-​third of
and cohesive assessments of global cancer bur- all worldwide death estimates. Civil registration
den are therefore pivotal in cancer control. coverage of cause-​of-​death data varies from 4% in
Amid the backdrop of urbanization, cultural China to 100% in the Netherlands (WHO, 2015).
globalization, and growing global prosperity, the For countries reporting cause-​of-​death data, the
burden of injuries, communicable diseases, and percentage of ill-​defined causes of death also var-
maternal/​childhood diseases are decreasing in ies widely, for example, from 8% in Israel to 45%
many countries while the burden of noncom- in Thailand. In a study by Mathers and colleagues
municable diseases is increasing. Among non- (2015), data from death registrations supplied to
communicable diseases, deaths from site-​specific the WHO were complete for 64 out of 115 mem-
cancers ascended the causes of death list in both ber states reporting such data. To estimate global
low-​and high-​ income countries; collectively, cancer mortality, extrapolations are done based
neoplasms are the second-​largest cause of death on cancer incidence data (from population-​or
worldwide (Forouzanfar et  al, 2015). To obtain hospital-​based registries) or rates in neighbor-
an accurate quantification of cancer burden, ing countries with better-​quality data (Murray &
manifold measurements that capture the num- Lopez, 1996a, 1996b; Ferlay et  al, 2002; Parkin
ber of deaths, incidence/​mortality rates, and time et al, 2005b).
trends with respect to variations between coun-
tries, regions, and risk factors must be consid- Incidence
ered. This chapter addresses the various methods Cancer incidence rates determine the probabil-
for measuring cancer burden and the complexi- ity of an individual being diagnosed with cancer
ties resulting from practical applications of these over a particular period of time (Table 2-​1). In
measurements, and provides an overview of cancer epidemiology, incidence rates represent
global cancer patterns and trends. the net effect of internal (genetic) and exter-
nal (environmental) causal and preventive fac-
MEASURES tors, weighted by their frequency and intensity.
OF CANCER BURDEN Synergistic and antagonistic effects between
Measures of disease burden can be used to deter- these exposures, whether genetic, environmen-
mine the relative contribution of cancer to over- tal, or both, are integrated into the measure.
all morbidity and mortality in any population Incidence rates allow a direct comparison of the
throughout the world. Incidence data are avail- contribution of risk factors in different popula-
able only for a small proportion of the global tions, ethnic groups, countries, and time periods.
32

32 Textbook of Cancer Epidemiology

TABLE 2-​1   MEASURES OF CANCER BURDEN, THEIR DETERMINANTS

AND LIMITATIONS

Measure Definition Determinants Limitations

Incidence Number of new cases, Burden of exposure to causes Often supplied by population-​
often per 105 person-​ of cancer, weighted by the risk based cancer registration limited
years or absolute imparted by each cause. to a small proportion of global
­number of new cases population.
per year. Influenced by diagnostic intensity,
screening, and primary prevention
programs.

Cumulative Proportion of people Incidence. Requires no loss to follow-​up, no

incidence who develop cancer competing risks, same period of
before a defined age. follow-​up time for all study subjects
and unchanged exposure status
throughout follow-​up.

Prevalence Proportion of Incidence, prognosis, and Requires population-​based

­population with cancer mortality from other causes. ­registration and follow-​up.
at a certain point in Cured patients cannot be readily
time (or over a period identified.
of time).
Does not inform on cancer
survivability.

Survival Proportion of cancer Natural history of disease. Requires long-​term follow-​up of

patients surviving for Stage at diagnosis. large number of patients.
a specified time after Spurious patterns may arise due to
Therapeutic efficacy.
diagnosis. lead-​time bias.
Influenced by screening—​spurious
patterns may arise due to lead-​time
and length bias.
Sensitive to misclassification of
cause of death.
Definition of disease may change
over time.
Background life expectancy may
differ between populations and may
be unrelated to cancer.
Provides no information on quality
of life lived.

Mortality Number of cancer Incidence. Influenced by adequacy of

deaths, often per Prognosis. death certification, including
105 person-​years, or autopsy rates.
absolute number of Influenced by changes in treat-
deaths per year. ment effectiveness and prevention
programs.
Provides no information on burden
caused by disability.
3

Measures and Estimates of Cancer Burden 33

TABLE 2-​1  CONTINUED

Measure Definition Determinants Limitations

Life-​years Number of years lost Incidence. Requires reliable population life

lost between age at death Age at diagnosis. tables.
and expected (in the
Prognosis.
absence of this disease)
age at death.
Disability-​ Combines impact of Incidence. Requires reliable population life
adjusted cancer on both quality Age at diagnosis. tables.
life years of life and survival. Difficult to quantify disability
Prognosis.
(DALYs) adequately.
Expected longevity.
Not able to demonstrate synergistic
Residual disability.
or multiplicative relationships to
cancer burden.

Other determinants, however, affect incidence standards, and must be recorded using predefined
rates, such as diagnostic patterns (e.g., use of a coding and reporting standards.
screening technique that detects asymptomatic Second, aside from logistical concerns of
lesions). Compared to incidence rates, mortal- registration efficiency and reliability, incidence
ity rates reflect the combined effects of incidence data may also suffer from methodological limi-
and treatment. Cancer survival rates depend on tations. These limitations may entail both under-​
stage at presentation, a function of public and and overestimation of the “true” incidence, and
professional awareness, as well as access to can- may distort comparisons between population
cer screening and treatment. Hence, including groups and across time periods. For example,
less advanced cancers, which may not be clini- increased diagnostic intensity (Helgesen et  al,
cally relevant, would result in a spurious increase 1996)  or introduction of new diagnostic tech-
in case survival rates (Bailar & Gornik, 1997). nologies, such as computerized tomography or
The practical use of incidence rates to meas- magnetic resonance imaging, may advance the
ure the cancer burden has limitations (Enstrom time of diagnosis, and may also entail detection
& Austin, 1977; Doll et  al, 1994; Bailar & of slow-​growing nonlethal cancers that would
Gornik, 1997). First, reliable incidence data have escaped attention without these technolo-
are available only for a small fraction of the gies. Use of new screening tools—​whether in
global population (Parkin et  al, 1999, 2005b) organized or opportunistic settings—​would have
and disproportionally in high-​ income coun- a similar effect on incidence. The introduction of
tries. Data from high-​quality population-​based prostate specific antigen (PSA) testing to detect
cancer registries have been compiled since 1966 early prostate cancer is the preeminent exam-
in a series of volumes named Cancer Incidence ple. Following its widespread use in the United
in Five Continents. The 10th edition updates States, the reported incidence of prostate cancer
information on incidence rates until about increased sharply (Hankey et al, 1999; McDavid
2007 (Forman et al, 2013). et al, 2004). This increase is interpreted as an arti-
Key prerequisites for reliable rates include fact, due to the screening introduction (Barry,
census of the entire population by age, sex, and 2001; McDavid et al, 2004). In contrast, declining
other key demographic variables over time; pop- autopsy rates may spuriously cause decreasing
ulation access to adequate diagnostic facilities; cancer incidence trends because more cancers
histologic confirmation for a high proportion of escape diagnosis (Burton et al, 1998).
cancer cases; and complete, as well as timely, noti- Nonmelanoma skin cancer represents a
fication of all newly diagnosed (incident) cases to special case, because it is often diagnosed and
the registry. Cases must be diagnosed fastidiously, treated by dermatologists or general practitio-
according to internationally accepted diagnostic ners, without histologic verification. Under these
34

34 Textbook of Cancer Epidemiology

circumstances, it is difficult to measure the inci- in rates cannot be attributed to confounding due
dence of this common cancer without special, to age.
resource-​ intense provisions. For this reason, Standardization allows a comparison of rates
nonmelanoma skin cancer is often excluded from of cancer between different countries, but the
overall cancer incidence statistics. results obtained depend on the standard popu-
These potential limitations should not lation used. Using one particular standard may
undermine the usefulness of incidence rates. occasionally rank country rates differently than
Incidence rates convey essential information on using an alternative population as the standard
population cancer risk that must be interpreted (Smith, 1992). The main reason for this is that
within the context of data reliability, risk fac- cancer incidence varies strongly with age. In
tor prevalence, and healthcare access. In addi- general, higher-​income countries have a higher
tion, incidence rates provide a benchmark from proportion of older people compared to lower-​
which effectiveness in cancer prevention can be income countries.
measured. Issues of validity always need consid-
eration, but caution should be exercised when Cumulative Incidence
generalizing across registries, cancer sites, or Cumulative incidence, like incidence rate, pro-
time periods. The strengths and potential lim- vides a measure of the burden of new disease
itations should be carefully assessed for each (Table 2-​1). Unlike incidence rates, cumulative
specific issue and cancer. Biologic understand- incidence is a more easily interpretable and intu-
ing and clinical experience are often needed to itively appealing measure of disease frequency.
achieve this. The increase of noncommunicable Cumulative incidence is defined as the propor-
diseases in low-​to middle-​ income countries tion of people who develop a disease among
underscores the need for strengthening vital those at risk over a specified period of time and
statistics and registration systems (Jong-​wook, is interpreted as the probability, or risk, that an
2003) to better understand cancer rates in these individual will develop the disease during that
populations and achieve progress in cancer time period (Day, 1992). Tables 2-​3A and 2-​3B
prevention. report cumulative incidence data for different
cancer sites and populations. The cumulative
Age Standardization incidence of an Australian woman developing
Comparison of crude incidence rates between skin melanoma by age 74 is 28 per 1,000, which
two or more populations may be misleading due can be interpreted as a 2.8% probability or risk
to differences in their age distributions. The aim of developing skin melanoma by the time she is
of age standardization is to remove the effect of aged 74 years (Table 2-​3A).
these differences and allow a direct comparison A number of assumptions are necessary in
of cancer burden among populations. order for the cumulative incidence to be a valid
Incidence rates (or mortality rates) are measure (Table 2-​ 1). Cumulative incidence
adjusted to a standard population, usually the assumes:  (1) there is no loss to follow-​up and
world population or country population (Smith, (2)  the entire population at risk is followed for
1992). Individuals in each age or category of the the same time period. The reality remains that
standard population contain a fixed number of this is difficult to accomplish over long follow-​up
people at a given time; the age distribution of this periods. Cumulative incidence measures assume
standard population is used for calculating the that there are no competing risks (i.e., individuals
age-​standardized rate for the population being are not dying from other causes). Assumptions
studied. required for cumulative incidence are often not
In Tables 2-​2A and 2-​2B, the reported rates are met, and using incidence rates is a more appro-
standardized to the world population. Therefore priate measure of the disease (Parkin et al, 2003).
they are the rates that one would expect if each
country had the same age distribution as the Prevalence
world standard. The age-​standardized incidence Cancer prevalence reflects the extent to which
rate of stomach cancer for white US women is the cancer exists in a population at a particular
2.8 per 100,000 person-​years, while it is 14.1 per point in time. It is defined as the proportion of
100,000 person-​years for women in China (Table the population, often by sex and age group, with
2-​2A). Because both rates have been standard- a past or current diagnosis of cancer (Table 2-​1).
ized using the same population, the difference It does not take into account individuals who
 35
TABLE 2-​2 A   AGE-STANDARDIZED, TO THE WORLD POPULATION, INCIDENCE RATES OF CANCER OF VARIOUS SITES AMONG
WOMEN, PER 100,000 PERSON-YEARS

  Registry US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14
Cancer

Lip 0.2 0 0.1 0 0.1 0.5 0.5 0.2 0.1 0.2 0 1.1 0.2 0.2
Tongue 1.2 0.9 1 0.8 0.5 0.8 0.9 0.4 0.3 0.7 0.6 1 1.2 0.6
Mouth 1 1.1 1 0.7 0.5 0.9 1.3 1.1 0.8 1 1 1.1 1.9 0.9
Other oropharynx 0.1 0.2 0.1 0.2 0 0.6 0.3 0.1 0.2 0.3 0.3 0.1 0.5 0.1
Hypopharynx 0.2 0.3 0.2 0.2 0 0.2 0.3 0.1 0.4 0.1 0.3 0.1 0.3 0.1
Pharynx unspecified 0.1 0.1 0.1 0 0 0 0.1 0 0.1 0.1 0.1 0.1 0 0.1
Esophagus 1.1 2.1 3.8 1.9 1.9 1 2.2 0.6 1.2 0.8 11.5 1.7 2.6 1.4
Stomach 2.8 5.1 3.8 17.3 14.1 3.2 4.7 5.5 5.5 8.5 5.9 3.4 11.6 13.9
Small intestine 1 1.8 0.7 0.4 0.8 1 0.9 0.8 0.6 0.6 0.2 0.8 0.9 0.6
Colon 17.6 25.3 15.3 14.4 14.8 16 22.3 15.8 13.4 19.8 4.8 21.9 15.7 9.6
Rectum 6.8 8 7.4 6.5 8.3 7.8 11.2 8.3 6.2 9 4.6 10 10.6 4.7
Liver 2 2.9 1.6 8.7 7.1 1.7 0.7 2.2 2.4 2.8 8.7 1.9 3 3.2
Gallbladder and 1.6 2 1.3 3.8 4.2 2.2 1.5 2.9 4.3 1.9 0.4 1.5 3.6 4.9
biliary tract
Pancreas 6 8.6 5.2 5.8 5.3 4.8 4.5 5.4 5 5.5 2.1 5.3 4.3 3.4
Larynx 1 1.4 0.6 0.2 0.2 0.3 0.9 0.6 0.7 0.8 0.2 0.4 1.3 0.6
Lung (incl. trachea 35.3 35.5 21.5 13.9 18.1 17.2 24.4 8.9 17.9 14.8 5.1 19.2 14.4 8.9
and bronchus)
Bone 0.9 0.7 0.6 0.4 1.1 0.7 1 0.9 0.7 0.8 1.7 0.7 1.3 1.4
Melanoma of skin 14.4 0.7 9.9 0.3 0.4 13.9 15.2 8.6 5.8 10.7 2 28.4 5 3.1
Breast 88.8 83.9 85.9 37.3 39.2 81.1 96 69.6 60.3 88.1 32.9 82.5 69.3 49.2
Cervix uteri 6.3 7.9 7.5 7.1 4.2 6.9 5.3 3.8 13.9 5.8 54.3 5.3 28.5 20.5
Corpus uteri 17.4 14.2 13.1 5.4 5.8 14 13.5 13.5 16.7 16.5 5.6 10.4 7.4 5.5
Ovary 9.7 7.1 13.7 6 7.6 9.3 8.1 7.5 12.3 7.9 6.9 7.8 9.6 8.7
(continued)
 36
TABLE 2-​2 A  CONTINUED

  Registry US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14
Cancer

Kidney 6.7 7.2 4 1.7 3.3 4.4 5.1 4.2 5.9 6.7 1.7 5 5 2.9
Bladder 5.7 4.1 3.3 1.9 2.3 4.9 6.7 6.8 3.8 6.3 1.8 2.6 5.8 2.2
Brain, nervous 5 3.1 3.8 2 6.2 11 4 5.7 8.9 5.8 0.9 4.6 6 4.4
system
Thyroid 13.1 7.1 3.6 3.8 9.1 3.7 2.5 9.2 7.8 11.7 3.2 10.7 14.8 11.3
Hodgkin lymphoma 2.4 2.1 2.1 0.3 0.2 1.6 1.7 2.8 1.5 2.9 1.9 2 1.6 1.1
Non-​Hodgkin 11.3 8.6 7.6 4.1 3.9 7 8.2 7.7 5 7.9 7.2 10.1 8.9 8.4
lymphoma
Multiple myeloma 2.6 6.4 2.5 0.9 1 2.4 2.4 2.2 1.7 2.9 2.1 3 3.2 2.3
Lymphoid leukemia 4 2.3 3.2 1.9 1.3 3.6 3.4 2.1 2.3 2.8 0.5 3.7 3 3.4
Myeloid leukemia 3.2 3.1 2.7 2.1 1.9 2.3 2.6 2.6 2.4 2.6 0.3 3.3 3.8 2.8
Leukemia 0.2 0.2 0.1 0.2 0.7 0.3 0.1 0.2 0.1 0.5 0.6 0.2 0.9 0.7
unspecified
All sites 290.2 275.8 303.6 158.8 179.1 254.6 281.8 292.6 246.5 313.6 214.9 273.1 474.6 202.2

1
USA SEER: white, 2003–​2007. 8
Navarra, 2003–​2007.
2
USA SEER: black, 2003–​2007. 9
Cracow City, 2003–​2007.
3
West Midlands Region, 2003–​2007. 10
Florence and Prato, 2003–​2007.
4
Osaka Prefecture, 2003–​2007. 11
Kyadondo County, 2003–​2007.
5
Shanghai, 2003–​2007. 12
New South Wales, 2003–​2007.
6
Sweden, 2003–​2007. 13
Goiania, 2003–​2007.
7
Eindhoven, 2003–​2007. 14
Cali, 2003–​2007.
 37
TABLE 2-​2 B   AGE-STANDARDIZED, TO THE WORLD POPULATION, INCIDENCE RATES OF CANCER OF VARIOUS SITES
AMONG MEN, PER 100,000 PERSON-YEARS

  Registry US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14
Cancer

Lip 0.9 0.1 0.3 0 0.1 0.8 1.1 3.3 0.9 0.6 0.1 3.6 1.4 0.3
Tongue 3.4 2.7 2.1 1.7 0.7 1.3 1.5 1.8 1.7 1.7 0.8 2.6 4.8 1.3
Mouth 1.9 2.2 1.9 1.4 0.8 1.2 2.4 2.2 1.8 2 1.3 2.1 6 1.4
Other 0.4 0.9 0.4 0.8 0.2 1.6 0.8 1.6 0.8 0.4 0.8 0.3 2.7 0.2
oropharynx
Hypopharynx 0.8 1.8 0.7 1.7 0.2 0.5 0.8 1.6 0.3 0.3 1.3 0.8 3.4 0.2
Pharynx 0.2 0.4 0.3 0.2 0.1 0.1 0.1 0.4 0.1 0.2 0.1 0.4 0 0.1
unspecified
Esophagus 5.2 6.7 9.5 11.3 6.6 3.5 7.8 5.5 3.2 1.9 15.6 4.9 10.7 2.8
Stomach 5.8 10.4 10.2 45.3 26.6 6.3 9.3 14.3 13.9 17.6 8 7.7 24.4 26
Small intestine 1.5 2.4 1 0.7 1.1 1.6 1.1 0.8 0.6 0.9 0.1 1.2 1.9 0.7
Colon 22.6 31.9 21.4 20.5 16 18.2 27.5 26.8 22 28 4.1 27.4 18.9 9.5
Rectum 11.1 11.7 16.4 13.4 11.6 12.6 19 17.1 13.5 14.9 3.8 18.4 11.9 6.5
Liver 6.4 10.1 3.6 25.6 21.7 3.4 1.6 7.8 4.8 8.8 11.4 5.4 7 4.2
Gallbladder and 1.6 1.7 1.1 5.2 2.9 1.4 2 2.4 2.8 3.2 0.3 1.5 3.1 2.6
biliary tract
Pancreas 8.1 10.9 6.5 9.4 7 5.4 6.3 8 7.8 7.9 2.7 7.2 5.4 3.9
Larynx 4.3 7.6 3.5 2.6 2.4 1.9 4.7 11 7.5 7.3 2.5 3.6 7.3 4.2
Lung (incl. 48 66.8 39.4 41.6 40.9 20.7 52.2 50.5 58.7 48.7 5.2 34.7 30.3 19
­trachea and
bronchus)
Bone 1.1 0.8 0.8 0.6 1.1 0.7 1.2 1.1 0.9 0.8 1.9 1.2 2.1 1.9
Melanoma of skin 19.6 0.7 8.8 0.3 0.5 13 11.4 6.7 6.3 11 1.1 40.8 8 4
Prostate 101.9 168.3 67.3 16.3 10.5 107.6 68.8 64.2 35.8 52.9 42.5 111.1 157.4 66.9
Testis 5.9 1.1 6.6 1.3 0.7 6.2 6.2 4.1 5.1 5 0.3 6.1 1.7 2.9
(continued)
 38
TABLE 2-​2 B  CONTINUED

  Registry US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14
Cancer

Kidney 12.7 14.8 7.4 4.5 6.3 7.1 8.7 8.8 9.4 15.6 1.1 10.2 7.9 4.4
Bladder 22.6 12.1 13.2 8.2 8 17.5 28.7 39.8 20 33.6 2.6 9.1 15.4 6.8
Brain, nervous 6.7 3.8 6 2.4 5.4 9.5 6.1 7.8 8.2 6.7 1.9 6.5 10.5 5.8
system
Thyroid 4.2 2.2 1.2 1.4 2.6 1.3 1.3 3 1.8 4.7 1.7 3.3 3.1 2.3
Hodgkin 2.9 2.6 2.4 0.4 0.2 1.8 2.7 3.3 1.8 2.7 1.6 2.4 2.6 1.7
lymphoma
Non-​Hodgkin 16.1 12.9 10.8 6.4 5.4 9.9 11.3 10.5 8 13.1 9.6 14.5 11.4 11.7
lymphoma
Multiple 4.1 9.3 3.6 1.3 1.3 3.5 3.6 2 1.8 4.4 2.1 4.7 4.8 2.7
myeloma
Lymphoid 6.7 4.5 5.3 2.2 1.8 5.8 4.9 4.1 3.3 3.7 0.9 6.3 4.4 3.7
leukemia
Myeloid 4.6 4.1 4 3.2 2.5 2.7 3.2 2.4 3.5 3.5 0.6 5.2 4.5 3.1
leukemia
Leukemia 0.3 0.5 0.2 0.3 0.7 0.2 0 0.3 0.3 0.7 0.8 0.3 0.9 0.9
unspecified
All sites 359.5 435 362 244.3 204.8 301.8 342.6 423 299 401.6 180.8 375.9 631.9 226.6
1
USA SEER: white, 2003–​2007. 8
Navarra, 2003–​2007.
2
USA SEER: black, 2003–​2007. 9
Cracow City, 2003–​2007.
3
West Midlands Region, 2003–​2007. 10
Florence and Prato, 2003–​2007.
4
Osaka Prefecture, 2003–​2007. 11
Kyadondo County, 2003–​2007.
5
Shanghai, 2003–​2007. 12
New South Wales, 2003–​2007.
6
Sweden, 2003–​2007. 13
Goiania, 2003–​2007.
7
Eindhoven, 2003–​2007. 14
Cali, 2003–​2007.
 39
TABLE 2-​3 A   CUMULATIVE INCIDENCE, PER 1,000 PERSONS, OF CANCER OF VARIOUS SITES AMONG WOMEN BY AGE 74*

Cancer US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14

Lip 0.2 0 0.1 0 0.1 0.6 0.5 0.4 0.1 0.2 0 1.1 0.2 0.2
Tongue 1.4 1 1.1 0.8 0.6 0.9 1.1 0.5 0.4 0.8 0.8 1.2 1.3 0.7
Mouth 1.2 1.2 1.2 0.7 0.6 1 1.7 1.1 0.7 1.1 1.5 1.4 2.2 1
Other 0.1 0.2 0.1 0.2 0 0.6 0.4 0.1 0.2 0.3 0.3 0.1 0.6 0.1
oropharynx
Hypopharynx 0.2 0.3 0.2 0.2 0 0.2 0.4 0.1 0.5 0.2 0.6 0.1 0.2 0.1
Pharynx 0.1 0.2 0.1 0 0 0 0.1 0 0.1 0.1 0.1 0.1 0.1 0
unspecified
Esophagus 1.4 2.5 4.2 2.2 2.1 1.2 2.6 0.7 1.3 1 15.8 1.8 3.2 1.7
Stomach 3.1 5.6 4.1 18.9 15.6 3.6 5.3 5.6 6.4 9.4 7.7 3.8 13.7 15.7
Small intestine 1.2 2.2 0.9 0.5 1 1.2 1 1 0.7 0.8 0.3 0.9 1 0.8
Colon 19.8 29.8 17.4 16.2 17.9 19.1 26.7 17.5 16 21.9 5.8 26 17.8 10.9
Rectum 7.9 9.4 8.5 7.6 9.9 9.4 13.6 9.5 7.6 10.3 5 11.8 12.7 5.6
Liver 2.4 3.5 1.8 10.6 8 2 0.8 2.3 2.5 3.3 10.7 2.2 3.7 3.5
Gallbladder and 1.9 2.3 1.6 3.9 4.4 2.8 1.6 2.9 5.6 2 0.7 1.7 4.4 6
biliary tract
Pancreas 7 10 6.1 6.6 6.2 6.2 5.9 5.9 5.8 6 2.9 6 5.4 3.7
Larynx 1.3 1.7 0.7 0.2 0.2 0.4 1.1 0.7 0.9 1.1 0.2 0.5 1.6 0.7
Lung (incl. 45.6 44.7 26.3 15.7 21.2 22.7 31.2 9.7 22 18.4 7.8 23.8 17.6 10.8
trachea and
bronchus)
Bone 0.7 0.6 0.4 0.3 1 0.6 0.8 0.7 0.5 0.8 1.8 0.6 1.1 1.3
Melanoma of 14.7 0.7 10 0.3 0.5 14.5 15 8.8 6.5 10.8 2.9 30.1 5.1 3.5
skin
Breast 103.2 95.8 96.4 39.8 43.1 93.7 108 74.7 68.6 97.7 38.8 94.1 76.7 55.7
Cervix uteri 6.2 8.1 6.8 7 3.9 6.6 5 3.9 14.4 5.6 61.5 5.2 30.9 21.9
(continued)
 40
TABLE 2-​3 A  CONTINUED

Cancer US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14

Corpus uteri 21.3 18.3 16.7 6 6.8 18.3 17.4 16.7 20.7 18.3 7.5 13 9.2 6.7
Ovary 11 8.1 15.6 6.2 8 10.9 9.8 8.6 14 8.7 7.7 8.8 10.9 9.8
Kidney 7.9 8.5 4.7 1.9 3.7 5.3 6.3 4.3 7.3 7.8 1.9 5.7 5.8 3.5
Bladder 6.7 4.7 3.6 2 2.6 5.8 7.9 8 4.8 7.3 2.3 2.7 6.6 2.3
Brain, nervous 4.7 2.8 4 1.6 6.7 11.3 3.7 5.9 8.2 6 1 4.6 6.6 4.5
system
Thyroid 12.6 7.4 3.4 4.1 8.7 3.5 2.5 8.9 8.2 11.6 3.5 10.4 15.1 11.8
Hodgkin 2 1.8 1.7 0.2 0.1 1.3 1.3 2 1.2 2.2 2.2 1.6 1.4 1
lymphoma
Non-​Hodgkin 12.9 9.5 8.6 4.4 4.2 7.8 9 8.7 5.7 9 7.7 11.3 10 10
lymphoma
Multiple 3 7.9 3 1.1 1.2 3 2.9 2.8 2 3.4 2.7 3.5 4.3 3.1
myeloma
Lymphoid 3.6 2.3 2.9 1.4 0.8 3.4 2.9 1.6 2 2.2 0.4 3.2 2.3 2.5
leukemia
Myeloid 3.2 3.1 2.7 1.8 1.9 2.3 2.8 2.2 2.1 2.5 0.2 3.3 4 2.9
leukemia
Leukemia 0.2 0.2 0.1 0.1 0.7 0.3 0.1 0.1 0.1 0.5 0.6 0.2 0.7 0.6
unspecified
All sites 330.9 317.6 338.3 171.4 197.2 292.4 320.3 318.4 280.8 342.7 248.8 305.2 528.4 226.3

*Interpretable as probability
1
USA SEER: white, 2003–​2007. Navarra, 2003–​2007.
8
2
USA SEER: black, 2003–​2007. Cracow City, 2003–​2007.
9
3
West Midlands Region, 2003–​2007. 10
Florence and Prato, 2003–​2007.
4
Osaka Prefecture, 2003–​2007. 11
Kyadondo County, 2003–​2007.
5
Shanghai, 2003–​2007. 12
New South Wales, 2003–​2007.
6
Sweden, 2003–​2007. 13
Goiania, 2003–​2007.
7
Eindhoven, 2003–​2007. 14
Cali, 2003–​2007.
 41
TABLE 2-​3 B   CUMULATIVE INCIDENCE, PER 1,000 PERSONS, OF CANCER OF VARIOUS SITES AMONG MEN BY AGE 74*

Cancer US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14

Lip 1 0.1 0.4 0 0.1 0.9 1.3 4 0.9 0.7 0 3.9 1.7 0.3
Tongue 4.2 3.6 2.5 2.1 0.8 1.6 1.8 2 1.9 2.1 0.8 3.1 5.6 1.4
Mouth 2.3 2.7 2.2 1.7 0.9 1.5 2.7 2.4 2.2 2.2 1.2 2.6 7.2 1.8
Other 0.5 1.1 0.4 1 0.2 1.9 0.9 1.9 0.9 0.5 0.8 0.4 3 0.3
oropharynx
Hypopharynx 1 2.3 0.9 2.2 0.2 0.7 0.9 1.9 0.4 0.3 1.6 1.2 4.4 0.2
Pharynx 0.3 0.5 0.3 0.2 0.1 0.1 0.1 0.5 0 0.2 0.1 0.5 0 0.1
unspecified
Esophagus 6.4 8.5 11.3 14.5 7.1 4.4 9.4 6.9 4.6 2.4 19.8 5.8 13.3 3.2
Stomach 6.7 12 11.8 53.6 30.6 7.1 11.1 16.1 16.7 20.3 11 9.1 28.3 31.2
Small intestine 1.8 3.1 1.2 0.8 1.3 2 1.2 0.9 0.7 1.2 0.1 1.4 2.6 1
Colon 26.4 38.3 24.9 23.8 18.9 21.4 32 32 27.7 32.7 4.9 33.6 22.5 11
Rectum 13.5 14.1 20.4 16.9 13.8 15.6 23.5 21.7 17.1 17.9 5.1 22.9 13.8 7.8
Liver 7.6 12.2 4.2 32.6 24.1 4 1.8 9.5 6.3 10.3 13.6 6.2 8.4 4.9
Gallbladder 1.8 2.1 1.3 5.6 3.2 1.7 2.3 2.5 3.5 3.6 0.4 1.8 3.2 2.9
and biliary
tract
Pancreas 9.7 13.3 7.5 11 7.8 7 7.8 9.9 9.3 9.8 3.9 8.5 6.2 4.4
Larynx 5.4 10 4.3 3.3 2.8 2.3 5.8 13.2 9.4 9.1 3.1 4.7 9.3 5.3
Lung (incl. 60.5 84.4 47.1 46.4 46.7 27.2 65.6 61.7 76.9 61 6.5 42.6 40.7 22.2
trachea and
bronchus)
Bone 0.9 0.7 0.7 0.5 1 0.6 1 0.8 0.9 0.7 2.6 0.9 1.6 1.6
Melanoma of 22 0.9 9.7 0.3 0.6 14.5 13 7.1 6.5 11.8 1.5 45 8.7 4.3
skin
Prostate 138.5 226 85.7 19.3 11.5 143.8 91.3 90 46.1 68.4 53.6 146.4 202.3 85.8
Testis 4.5 0.8 5.1 1 0.5 4.7 4.7 2.9 3.8 3.9 0.2 4.7 1.3 2.1
(continued)
 42
TABLE 2-​3 B  CONTINUED

Cancer US-​W1 US-​B2 UK3 Japan4 China5 Sweden6 Netherlands7 Spain8 Poland9 Italy10 Uganda11 Australia12 Brazil13 Colombia14

Kidney 15.2 18.1 8.9 5.2 7.1 8.7 10.6 10.4 11.9 18.6 0.7 12 9.2 5.1
Bladder 26.3 14.1 15.1 9.1 8.6 20.5 34.9 49.4 24.5 40.8 3 9.8 17 7.9
Brain, nervous 6.6 3.7 6.3 2.1 5.5 9.8 6.4 7.8 8.4 6.9 2 6.9 11.5 6
system
Thyroid 4.6 2.4 1.2 1.5 2.6 1.4 1.2 2.8 1.7 4.7 2.5 3.4 3.2 2.2
Hodgkin 2.5 2.3 2.1 0.4 0.2 1.5 2.1 2.7 1.4 2.3 1.7 2 2.2 1.7
lymphoma
Non-​Hodgkin 17.8 14.1 12 6.7 5.6 11 13.1 10.8 8.9 13.5 8.8 16.5 11.9 12.7
lymphoma
Multiple 4.9 11.4 4.1 1.4 1.6 4.2 4.2 2.3 2.6 5.3 2.8 5.6 6 3.4
myeloma
Lymphoid 6.4 4.5 4.9 1.7 1.3 5.7 5.2 3.8 3.2 3.1 1.2 5.9 3.2 3
leukemia
Myeloid 4.7 4.4 4.1 3.3 2.3 2.8 3.5 2.5 2.6 3.1 0.6 5.1 4.6 3.2
leukemia
Leukemia 0.3 0.4 0.2 0.2 0.8 0.3 0 0.1 0.2 0.7 0.8 0.3 0.8 0.9
unspecified
All sites 434.2 543.3 418.1 282.9 227.3 364.7 408.2 505.5 361.4 467.3 210.3 446.3 739.2 265

*Interpretable as probability
1
USA SEER: white, 2003–​2007. 8
Navarra, 2003–​2007.
2
USA SEER: black, 2003–​2007. 9
Cracow City, 2003–​2007.
3
West Midlands Region, 2003–​2007. 10
Florence and Prato, 2003–​2007.
4
Osaka Prefecture, 2003–​2007. 11
Kyadondo County, 2003–​2007.
5
Shanghai, 2003–​2007. 12
New South Wales, 2003–​2007.
6
Sweden, 2003–​2007. 13
Goiania, 2003–​2007.
7
Eindhoven, 2003–​2007. 14
Cali, 2003–​2007.
43

Measures and Estimates of Cancer Burden 43

have been cured of malignancy or may have infections, or accidents. Those who remain alive
tumor recurrence. Therefore, knowledge of the for the study period or are lost to follow-​up are
incidence rate, the course of cancer treatment “censored.” Although the malignancy is often the
and fatality rates are necessary to interpret prev- predominant cause of death, particularly during
alence. For example, prevalence may be relatively the early years of follow-​up of a cancer patient,
high for a rare malignancy such as testicular can- other causes of death play an important role as
cer that strikes young individuals and carries a the observation period continues. These other
favorable outlook. To avoid this problem, some- causes of death contribute to a “baseline” death
times patients are no longer counted when they rate to which the cancer-​specific rate is added.
have exceeded 5-​year survival. When two groups with a similar baseline rate
Cancer prevalence does provide useful are compared, the absolute difference in survival
information for allocating healthcare resources. between the two groups may be relatively stable
The need for patient care, regular checkups, treat- over time. As the baseline rate grows, the rela-
ment of long-​term complications (whether due tive difference will be lower if all causes of death
to the disease or the treatment), and resources are considered than if only deaths caused by the
for terminal care depends on the number of sub- malignant disease of interest are considered.
jects in the population with a cancer diagnosis. The cause-​specific, or corrected survival, pro-
Prevalence may increase as a consequence of vides a more valid estimate of the excess death
improved prognosis, even if incidence rates are rate attributable to the particular cancer under
stable. Advancement of the time of diagnosis study. It does so by excluding deaths from other
(lead-​time) following introduction of screening causes. Such patients are “censored” at their date
may also increase cancer prevalence. Reliable of death from another cause or when they are
estimates of cancer prevalence require linkage lost to follow-​up. The assumption when calcu-
between long-​term cancer registration, survival lating corrected survival rates is that the causes
databases, and updated population registers. of death are correctly classified. This cause may
be obvious in a cancer patient who dies with dis-
Survival seminated advanced disease following a period
Survival data are calculated chiefly for the pur- with, for instance, progressive fatigue, cachexia,
pose of providing an expected cancer outcome jaundice, or organ failure. Sometimes, however,
without necessarily including a death outcome. attributing death to one specific cause may be
Despite their intuitive appeal as a measure of complex or impossible, even if an autopsy is car-
prognosis and therapeutic efficacy, survival data ried out. First, pathologists may offer conflict-
also have numerous determinants, limitations, ing opinions on tumor malignancy based on
and pitfalls, as indicated in Table 2-​1, but other gross examination or the histological specimens
measures in Table 2-​ 1, which could be used obtained. Second, even once a tumor is agreed to
instead of survival data, have their own limita- be malignant, the malignancy may not have been
tions. Survival is still the best available measure the cause of death, or could simply be a contrib-
for evaluating the effectiveness of cancer care; uting cause of death. Residual disease, side effects
the different survival measures serve different of treatment, or other comorbidities may act
purposes, each with its own limitations. These solely or synergistically to an underlying cause
measures, though, have one prerequisite in com- of death. Fortunately, this gray zone is small in
mon:  the need for complete follow-​up. If this many instances and therefore corrected survival
cannot be achieved, we need to know the date provides a reliable estimate of prognosis, unin-
when follow-​up ended for each patient. The cal- fluenced by other causes. Corrected survival also
culated survival probabilities would be correct allows more meaningful comparisons between
only if losses to follow-​up are unrelated to the patient groups with different age distributions.
outcome, that is, if we do not lose track preferen- Calculation of the relative survival is based on
tially of those with a good or bad prognosis. a statistical methodology that allows the investi-
The observed survival conveys the probability gator to be blinded to information about causes
of surviving a specific time period, typically start- of death of patients. It does so by relating the
ing at the date of cancer diagnosis. All causes of observed survival to the expected survival in a
death are taken into consideration—​both those group from the general population with the same
linked to the cancer of interest and those due to age and sex distribution. Two basic assumptions
other diseases such as coronary heart disease, must be met to assure valid estimates of relative
4

44 Textbook of Cancer Epidemiology

survival. First, we need information about sur- cancer incidence rates. Additionally, no follow-​
vival prospects in the absence of cancer. Reliable up is required, death is an unequivocal event and,
life tables from the general population must be in many countries, a death certificate is required
available. Second, the group of cancer patients before a funeral can take place. The first attempts
under study must have the same “baseline” risk to collect death statistics began in the 17th cen-
of dying as the general population. For exam- tury. Modern death registries were established in
ple, residents of particular neighborhoods or most high-​income countries by the middle of the
smokers have a higher death rate than the gen- 20th century. In 1990, the coverage of mortality
eral population at large. If such individuals are statistics ranged from only about 1% in the sub-​
overrepresented in the group of cancer patients Saharan Africa to around 40% in Latin America
studied, then their baseline risk of dying will be and close to 100% in economically developed
underestimated. Hence, the relative survival will countries (Murray & Lopez, 1996a).
overestimate the excess death risk attributable to Mortality rates have limitations too (Table
their cancer. In most situations, however, cor- 2-​1), and these have sometimes been underes-
rected and relative survival indices agree because timated. The reliability of this measure depends
the underlying measurement goal is the same. critically on how accurately causes of death are
It has been noted that available estimates of classified. As already discussed, even in a patient
survival are based on patients diagnosed several with a known cancer diagnosis, it is not always
years before. To estimate 10-​year survival, tradi- easy to tell whether this disease was the underly-
tional methods require a cohort of patients where ing or a contributing cause of death, or whether
at least some individuals are diagnosed over it was irrelevant. In settings without public
10  years ago and all patients contribute to the awareness of early cancer symptoms, poor access
survival proportion for the first year, including to healthcare, limited facilities for diagnostic
those diagnosed 10 or more years ago. It might be workup, or low autopsy rates, opportunities for
more appropriate to estimate each of the succes- misclassification of causes of death are abun-
sive conditional survival proportions using only dant. It must also be noted that individuals who
the most recently diagnosed patients (Brenner are cured do not appear in the mortality statis-
et al, 2004; Dickman & Adami, 2006). For exam- tics, thereby reducing the utility of mortality as
ple, the conditional survival proportion for the a measure of the overall cancer burden. In addi-
first year would be based on patients diagnosed tion, an individual’s quality of life and the bur-
during the previous year; the conditional survival den to healthcare resources are not accounted
proportion for the second year would be based for. Finally, death certificates classify cancers
on patients diagnosed the year before that, and according to the site of origin, without informa-
so on. Patients diagnosed many years in the past tion on other important characteristics, such as
would only contribute to the estimates of con- histology.
ditional survival proportions for later intervals. Little information exists regarding the valid-
This approached has been termed period analy- ity of mortality statistics and the comparability of
sis and allows for more accurate prediction of the rates across different populations and time peri-
survival of newly diagnosed groups of patients ods (Murray & Lopez, 1996a, 1996b). Moreover,
(Dickman & Adami, 2006). mortality rates do not meaningfully reflect the
burden of certain cancers with a favorable prog-
Mortality nosis, such as nonmelanoma skin cancer or
Because cancer mortality rates measure the rate endometrial cancer, or of others with a high cure
of dying from specific or from all cancers at the rate, for example cancer of the thyroid and tes-
population level, they are widely considered the tis. At the other end of the spectrum, mortality
single most important set of indicators of the rates approximate incidence rates for cancers that
burden of cancer (Table 2-​1). Due to the fact that remain almost inevitably fatal, for example those
reduction in mortality is the accepted target for of the lung, liver, and pancreas.
improvements in cancer control, mortality rate
comparisons are also the preferred measurement Temporal Trends
for evaluating prevention programs, including Changes in trends of age-​standardized cancer
screening. incidence or mortality rates are estimated using
Mortality rates present several advantages. annual percent change (APC) over several years.
Foremost is their much wider availability than Epidemiologically, APC can be equivalently
45

Measures and Estimates of Cancer Burden 45

defined as transformations of either (1) the slope entailed by cancer (and other diseases and con-
of the line that runs through the log of each stan- ditions) is calculated as “disability weight.” The
dardized rate, (2) the ratio of the last rate to the advantage of DALYs is that the impact of time of
first rate in the series, or (3) the geometric mean lost life and disability are integrated into one sin-
of the proportional changes in the rates over the gle measure. The DALYs for cancer are calculated
series (Fay et al, 2006). For short time intervals as the sum of the YLL due to premature mortality
with small change in rates, the second transfor- in the population plus the years lost due to disa-
mation can be used. For longer time intervals, bility (YLD) for people living with the cancer or
trends can be estimated by fitting multiple inci- its sequelae: DALY = YLL + YLD.
dence or mortality rates to segmented linear The YLL basically corresponds to the num-
regression models. This is accomplished using ber of deaths due to the cancer multiplied by
the free joinpoint software (http://​surveillance. the standard life expectancy at the age of death
cancer.gov/​joinpoint/​) used by many registries due to the cancer. The DALYs measure uses the
throughout the world for characterizing trends lowest observed death rate for any age group in
in cancer rates. Joinpoint uses an algorithm that countries of more than 5  million in population
tests whether a multisegmented line is a statisti- (Murray et al, 2010). The reference life tables for
cally significantly better fit then a straight or less-​
the 2010 Global Burden of Disease project have a
segmented line (Clegg et al, 2009). life expectancy of 86 years for males and females
at birth. The YLL is thus inversely related to the
Years of Life Lost age of death, with individuals of older age having
Though introduced relatively recently and not less YLL.
widely used, the years of life lost (YLL) measure The YLD corresponds to the number of prev-
addresses the fundamental difference between alent cases multiplied by the disability weight.
dying from cancer in childhood versus later in By using prevalence, rather than incidence, indi-
life (Table 2-​1). Typically, YLL can be obtained at viduals living with disability who need health
a group level from computer programs that cal- and other social services are reflected in burden
culate relative survival rates. In practical usage, estimates even when incidence of the disease has
YLLs are often weighted for some indicator of declined.
quality of the years of life with the disease, com- In general, disability weights represent the
pared to healthy life. Ranking cancers by YLL impact of the disease on an individual’s life
rather than by incidence or mortality rates can using a scale from 0 (perfect health) to 1 (death).
dramatically influence the apparent impact of a Undergoing a mastectomy, which is postopera-
specific form of cancer. Prostate cancer is by far tively disabling and psychologically upsetting
the most common cancer in men in many west- yet potentially curative, has a disability weight
ern populations, sometimes accounting for 25% of 0.036. In contrast, metastatic cancer, which is
or more of all diagnosed cancers. However, age permanently disabling and carries a poor prog-
at diagnosis is typically older, often over 75 years. nosis, has a disability weight of 0.451. The disa-
Therefore, if cancers are ranked by YLL at the bility weights are empirically estimated based on
population level, then, in the United States, stratified random sampling household surveys
prostate cancer is ranked lower than other life-​ administered to multiple countries, incorporat-
shortening malignancies. If cancers are ranked ing responses of linguistic, cultural, and socioec-
by YLL per person dying of cancer, then the aver- onomic diversity (Salomon et al, 2012; Salomon
age for all cancers is 15.3 years, while for prostate et al, 2015).
cancer it is 9.2 years (Ries et al, 1997). The DALYs are advantageous for providing
multidimensional planes of comparative analy-
Disability-​Adjusted Life Years sis. Use of DALYs enables comparisons of cancer
The disability-​adjusted life years (DALYs) meas- burden with other diseases, across world regions,
ure was developed with the rationale that differ- time periods, age groups, and sexes and accord-
ences in the burden of cancer should also take ing to exposure to risk factors (e.g., dietary,
into account the reduced quality of life of can- comorbidities, air pollution). The DALYs have
cer patients. To calculate DALYs, the same ideal two important limitations. First, they are not yet
life expectancy is used for all population groups. able to measure the complex synergistic (mul-
One DALY represents one lost year of healthy tiplicative or additive) contributions to cancer.
life (Table 2-​1), and the reduced quality of life Second, the same principles that grant the use of
46

46 Textbook of Cancer Epidemiology

the DALY metric across multiple planes of com- sexes, incidence and mortality rates of nearly all
parison also limit its interpretability of cultural, comparable cancer sites are greater among men
geographical, and socioeconomical underpin- than women (Figs. 2-​1A, 2-​1B), with the excep-
nings in cancer disparities within populations. tion of thyroid and gallbladder cancer.
The DALYs also provide limited genetic, mech-
anistic, or biological insight on cancer develop- Global Cancer Trends
ment and progression and are interpretable only and Patterns
as an estimator of burden. In both high-​and low/​ middle-​income coun-
Notwithstanding these limitations, DALYs tries with good-​quality data, mortality rates for
serve as a valuable tool for comparing different all cancers have decreased globally by 1% dur-
cancers, risk factors, and interventions in differ- ing the period 2000–​2010 (Hashim et al, 2016).
ent populations, and they may provide important An approximately 1% decrease has also been
guidance in developing public health policies. reported in regional studies from the United
States and Europe (La Vecchia et al, 2015;
CANCER IN THE WORLD Seigel et al, 2015), and findings have been sim-
Cancer represents a major cause of death in most ilar worldwide using GLOBOCAN data (Parkin
populations of the world. The Cancer and Other et al, 2005; Ferlay et al, 2015; Hashim et al,
Causes of Death Table 2-​4 ranks top 10 causes 2016). Declines are observed in most countries
of death in low/​middle-​and high-​income coun- for the most common cancers worldwide: lung
tries in 2013 as reported by the Global Burden (in men only), stomach, colorectal, breast, uter-
of Disease (GBD) 2013 Mortality and Causes of ine, and prostate cancer. However, liver cancer
Death Collaborators (Forouzanfar et al, 2015). mortality rates have increased in a large pro-
Cancer mortality accounts for a greater propor- portion of countries, particularly in Latin and
tion of deaths in high-​income countries; the most North America. Lung cancer mortality rates in
common cancer deaths occur from lung, colorec- women have also increased in most countries
tal, and stomach cancer. In contrast, cancer does (Figure 2-​2).
not account for such a substantial fraction of Cancer mortality trends must be interpreted
mortality in low-​and middle-​income countries. in view of country and regional comparisons. In
In low/​middle-​income countries, only lung can- addition to population demographic character-
cer makes it to the top 10 causes of death. Across istics, trends in mortality rates across countries

TABLE 2-​4   LEADING CAUSES OF DEATH IN LOW/MIDDLE- AND HIGH-INCOME

COUNTRIES IN 2013
Low-​and middle-​income countries High-​income countries

Cause % of total deaths Cause % of total deaths

1  Ischemic heart disease 12.7% Ischemic heart disease 21.5%

2  Hemorrhagic stroke 9.5% Ischemic stroke 8.4%
3 Chronic obstructive 5.7% Alzheimer disease 7.0%
­pulmonary disease
4  Ischemic stroke 5.2% Lung cancer 7.4%
5 Lower respiratory infections 5.1% Chronic Obstructive 4.8%
Pulmonary Disease
6 Tuberculosis 3.0% Hemorrhagic stroke 3.8%
7  Diarrheal diseases 3.0% Lower respiratory infections 3.8%
8  HIV/​AIDS 2.9% Colorectal cancer 3.1%
9 Diabetes 2.5% Diabetes 2.5%
10  Lung cancer 2.3% Stomach cancer 1.8%
Total number of deaths 28.1 million Total number of deaths 10.2 million
47

Measures and Estimates of Cancer Burden 47

A Breast
Colon and rectum
Cervix
Lung
Uterine
Stomach
Ovary
Thyroid
Liver
Non-Hodgkin's lymphoma
Leukemia
Pancreas
Oesophagus
Kidney
Brain, nervous system
Lip, oral cavity
Gallbladder
Bladder
Multiple Myeloma
Nasopharynx
Hodgkin lymphoma
Other Pharynx
Larynx
0 10 20 30 40 50
Age-standardized rate per 100 000 women-years
Incidence Mortality

FIGURE 2-​1A  Age-​standardized, to the world population, incidence and mortality rates, per 100,000 women-​years, by
cancer site, excluding nonmelanoma skin cancer—​Estimated world total, around 2012. 
Source: Globocan, 2012. Available at: http://​globocan.iarc.fr.

and regions are shaped by three main determi- Five-​year survival of lung cancer in the United
nants: (1) primary prevention (i.e., prevalence of States is poor at 17.4%, and most lung cancers
risk factors); (2) early diagnosis, including screen- remain incurable (SEER, 2012). Lung cancer
ing; and (3)  access or availability of treatment mortality is most effectively averted by the avoid-
(Table 2-​5). ance of modifiable risk factors, primarily tobacco
smoke (Table 2-​6). The converse of this statement
Global Cancer Trends and Patterns is also true: a higher prevalence of exposure to risk
by Site factors increases the mortality from lung cancer.
Mortality rates attributed almost exclusively to In the past, tobacco smoking had a higher prev-
modifiable risk factors include those of cancers alence within most countries among males; over
of the lung, stomach, and liver. Trends in mor- the past 30 years, this has changed (Malvezzi et al,
tality rates of these cancers are correlated with 2015). During the mid-​to-​late 20th century, the
trends in their incidence rates, are similar across sharp increase in the mortality from lung can-
most countries with less regional variation than cer among US women and men reflected the
other cancers, and are less influenced by eco- tobacco smoking epidemic; decline in mortality
nomic inequalities. Most importantly, mortality during the last few years reflects the earlier abate-
and incidence rates have similar values. ment of that epidemic among men, but less so
48

48 Textbook of Cancer Epidemiology

B Lung

Prostate

Colon and rectum

Stomach

Liver

Oesophagus

Bladder

Kidney

Non-Hodgkin lymphoma

Leukaemia

Lip, oral cavity

Pancreas

Brain, nervous system

Larynx

Other Pharynx

Gallbladder

Thyroid

Multiple Myeloma

Nasopharynx

Testis

Hodgkin lymphoma

0 5 10 15 20 25 30 35 40
Age-standardized rate per 100 000 men-years
Incidence Mortality

FIGURE 2-​1B  Age-​standardized,
to the world population, incidence and mortality rates, per 100,000 men-​years, by
cancer site, excluding nonmelanoma skin cancer—​Estimated world total, around 2012. 
Source: Globocan, 2012. Available at: http://​globocan.iarc.fr.

among women (Figs. 2-​3A, 2-​3B). Between 2000 prevalence rates of female smoking. Although
and 2010, lung cancer mortality rates decreased tobacco smoking is declining, lung cancer mor-
among males worldwide. An increased mortality tality is approaching equal rates between men
rate was observed among females in most coun- and women in several European countries and is
tries, except Ireland, Lithuania, most Asian, and expected to continue increasing for females world-
some Latin American countries. The convergence wide (Malvezzi et al, 2015).
of smoking patterns between men and women For lung cancer, as well as other rapidly
is reflected in the increase of female lung can- advancing or fatal cancers, such as liver or pan-
cer incidence and, sequentially, female lung can- creatic cancer, mortality rates provide good
cer mortality rates worldwide (Fig.  2-​3). Spain approximation of incidence rates; for cancers
(+6.2%) and Belgium (+5.2%) have experienced with variable curative treatments, better prog-
the largest APC increases, whereas Denmark noses (e.g., testicular cancer), or slower and
has the highest female lung cancer standardized nondebilitating growth, incidence rates cannot
rates overall (28.1/​100,000) (Hashim et al, 2016). approximate mortality. Figures  2.1A (women)
These European countries had coinciding high and 2.1B (men) demonstrate the variability in
49

Males Females

Colorectum

Colorectum
Stomach

Stomach
Prostate

Uterus
Breast
Lung

Lung
Liver

Liver
Argentina
Brazil
Canadaa
Chilea
Colombiaa
Costa Rica
Cuba
Ecuador
Mexico
Puerto Rico
USA
Uruguaya
Venezuelaa
Australia
New Zealanda
South Africaa
Hong Kong
Israel
Japan
Kuwait
Malaysiab
Philippinesb
Korea
Singapore
Belarusa
Kazakhstan
Kyrgyzstan
Norway
Moldova
Russian Federation
Serbia
Switzerland
Ukraine
Austria
Belgium
Bulgaria
Croatia
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Latvia
Lithuania > –2.5
Luxembourg
Malta –1.1 to –2.5
Netherlands –0.3 to –1.0
Poland
+0.2 to –0.2
Portugal
Romania +0.3 to +1.0
Slovakia +1.1 to +2.5
Slovenia
Spain > + 2.5
Sweden Missing
United Kingdom

FIGURE 2-​2  Annual percent change in cancer-​specific mortality rates for males and females by country, 2000–​2010. 
50

50 Textbook of Cancer Epidemiology

TABLE 2-​5   KEY DETERMINANTS OF GLOBAL CANCER MORTALIT Y

TRENDS BY CANCER SITE

Cancer site Primary prevention Early diagnosis, screening Treatment

Lung + –​ –​
Stomach + –​ –​
Breast + + +
Colorectum + + +
Prostate –​ –​ +
Uterus (mostly cervix) + + +
Liver + –​ –​

mortality and incidence rates by each cancer site. (Hashim et al, 2016). Rising transcontinental liver
The comparison between mortality and incidence cancer mortality rates can be explained by life-
rates, which may be calculated as a mortality-​ style exposure changes coupled with lack of liver
to-​incidence ratio, allows a crude assessment of cancer screening protocol, vaccination, or effec-
the fatality of each type of cancer. By contrast, tive treatment. Because the liver is a common site
breast cancer is both the most common primary of metastasis, a concern remains that some meta-
malignancy and causes the greatest number of static liver cancers may be misclassified as primary
female cancer deaths, although the mortality-​to-​ lesions, leading to an overestimation of the inci-
incidence ratio is the lowest. dence and mortality; this is due to the uncertain
The largest contribution to the decrease in over- validity of death certification in many countries
all cancer mortality came from stomach cancer (Bosetti et  al, 2008). Although misclassification
(–​2.7% in men and –​2.8% in women) (Hashim et al, may slightly skew these observed trends, the con-
2016). The greatest declines in the past 10  years sistent increase in liver cancer mortality in all age
occurred in continental Asia, Latin America, and groups, the constant sex ratios, and the correla-
the former Soviet Union (Fig. 2-​3). The decrease tion between hepatitis virus infection prevalence
in stomach cancer incidence in the past several and liver cancer mortality rates, weigh against a
decades has been attributed to Helicobacter pylori gross misrepresentation of increasing liver cancer
infection control (Peleteiro et al, 2012). Additional mortality worldwide.
potential contributors include the availability of Mortality from colorectal, breast, and uterine
fresh produce, reduction in dietary salt intake, (mostly cervical) cancer is most likely decreas-
gastric cancer screening, and sanitation in high-​ ing in high-​income countries due to screening
risk countries (Parkin, 2006; Choi et al, 2014). and improved treatment. Nevertheless, among
Primary liver cancer has a chiefly infectious countries with good-​quality mortality data, mor-
etiology and showed an increase in mortal- tality rates remain higher in high-​income than
ity rates in many countries. Approximately 90% low/​middle-​ income countries due to the per-
of liver cancers are hepatocellular carcinomas, sistence of underlying risk factors. Breast and
attributed to hepatitis B and C virus infection, in colorectal cancer mortality rate increases noted
addition to nonalcoholic steatohepatitis, tobacco in low/​middle-​income and transitional-​income
smoking, alcohol drinking, and overweight/​obe- countries are attributed to increases in obesity,
sity (Chuang et al, 2010). Liver cancer mortality physical inactivity, and changes in reproductive
has increased 2.5% or more per year in North factors (McCormack et al, 2011; Althuis et al,
American, most Latin American, and some Asian 2005). These mortality rate increases from 2000
countries (Hashim et  al, 2016). A  decrease in to 2010 were observed for breast cancer in con-
liver cancer mortality has been observed in Korea tinental Asia and South Africa and for colorectal
(>–​2.0% in both males and females) and Japan cancer in Latin America and Asia (Hashim et al,
(>–​3.3%), although incidence and mortality rates 2016). Uterine cancer mortality was observed to
in these countries and neighboring Asian coun- have the largest declines in Latin America, Asia,
tries remain among the highest in the world and the countries of the former Soviet Union.
51

Measures and Estimates of Cancer Burden 51

TABLE 2-​6   INDIVIDUAL AND JOINT CONTRIBUTIONS OF CAUSAL FACTORS TO

MORTALIT Y FROM SITE-SPECIFIC CANCERS IN THE WORLD, AROUND 2013

Cancer site Total number PAF (%) for individual Due to joint

of deaths risk factors hazards of
(in thousands risk factors
per year)

Mouth and 135 Alcohol use (28%), smoking (31%) 52%

oropharynx
Esophageal 440 Alcohol use (20%), smoking (25%), low fruit and 63%
vegetable intake (27%)
Stomach 841 Smoking (8%), diets high in sodium (45%) 50%
Colon and rectum 771 Alcohol use (5%), overweight and obesity (9%), physical 60%
inactivity (15%), low fiber intake (12%), high red meat
intake (7%), suboptimal calcium intake (15%)
Liver 818 Alcohol use (14%), Smoking (9%), drug use (14%) 41%
Pancreatic 352 Smoking (18%), overweight and obesity (7%) 23%
Lung, trachea, and 1,640 Smoking (62%), low fruit and vegetable intake (13%), 84%
bronchus ambient particular matter pollution (25%), indoor
smoke from household use of solid fuels (8%)
Breast 471 Physical inactivity (10%), overweight and obesity (8%), 23%
alcohol use (7%)
Cervix uteri 236 Unsafe sex (100%), smoking (4%), 100%
Corpus uteri 68 Overweight and obesity (39%) 39%
Bladder 174 Smoking (25%) 25%
Leukemia 265 Smoking (7%), overweight and obesity (6%), occupation 9%
exposure to benzene and formaldehyde (1%)
All cancers 8,236 Smoking (18%), overweight and obesity (5%), ambient 35%
particular matter pollution (5%), diet high in sodium (5%),
alcohol use (5%), low fruit and vegetable intake (4%),
unsafe sex (3%), occupational exposure to asbestos (3%),
low physical activity (2%), suboptimal calcium intake (2%),
indoor smoke from household use of solid fuels (2%)

This decline can be attributed to increased cer- risk factors and higher prostate cancer inci-
vical cancer screening in these regions, entailing dence (Lee et al, 2007). Despite its being the
detection and removal of preneoplastic lesions. most common male-​ specific cancer, little is
Mortality from cervical cancer is expected to known about prostate cancer etiology beyond
decline further with widespread screening and a predominance in populations of African ori-
vaccination (Pimenta et al, 2013). gin and a role of several low-​penetrance genes.
There is a substantial gap between prostate The increased mortality rates observed in
cancer mortality and incidence rates. Lower Latin America, low-​income Asian countries,
prostate specific antigen (PSA) screening rates and the former Soviet Union are likely to be
in Asian and African countries have been asso- explained, at least in part, by increased diag-
ciated with lower prostate cancer incidence noses due to screening coupled with a lack of
(Marugame et al, 2006), whereas other studies access to treatment (Hashim et al, 2016; Zhou
demonstrate an association between certain et al, 2016).
52

52 Textbook of Cancer Epidemiology

100
A. Females, by site
Uterine corpus*
7
6
80 5
Deaths per 100,000 females

4
3
2
60 1
0
1970 1990 2000

Lung & bronchus

40
Breast

Liver & Stomach Uterus (corpus and Colorectum

intrahepatic bile
20 duct
cervix combined)
Pancreas

0
1930 1940 1950 1960 1970 1980 1990 2000 2012
Year of death

FIGURE 2-​3A  Age-​standardized,
to the 2000 US standard population, cancer death rates (per 100,000 person-​years)
among women for selected cancers, United States, 1930 to 2012. 
Source: Siegel et al, 2016.
*Rates are age adjusted to the 2000 US standard population. Due to changes in International Classification of Diseases (ICD) coding,
numerator information has changed over time. Rates for cancers of the lung and bronchus, colorectum, liver, and uterus are affected by these
changes. Mortality rates for liver, pancreatic, and uterine corpus cancers are increasing.

Global Differences in Cancer Survival relative 5-​year survival of patients diagnosed

The success of early detection and cancer treat- with invasive cancer during 2005 to 2009.
ment may be measured by improvements in These countries have variable levels of eco-
survival from cancer (Sankaranarayanan et al, nomic development, gross national income
2011). Table 2-​7 displays the estimated overall per capita, and development of health services.

100 B. Males, by site

Lung & bronchus

80
Deaths per 100,000 males

60

Stomach
Prostate
40
Colorectum

Liver &
20 intrahepatic bile Leukemia Pancreas
duct

0
1930 1940 1950 1960 1970 1980 1990 2000 2012
Year of death

FIGURE 2-​3B  Age-​standardized,
to the 2000 US standard population, cancer death rates (per 100,000 person-​years)
among men for selected cancers, US, 1930 to 2012. 
Source: Siegel et al, 2016.
53

Measures and Estimates of Cancer Burden 53

TABLE 2-​7   FIVE-YEAR NET SURVIVAL RATES (%) FOR MAJOR CANCERS

IN SELECTED COUNTRIES, 2005–2009

Five-​year relative survival proportions (%)

Cancer site Algeria Brazil US China Denmark

Stomach 10.3 24.9 29.1 31.3 17.9

Colon 57.2 58.2 64.7 54.6 55.9
Rectum 45.5 55.9 64.0 53.2 58.4
Liver 17.5 11.6 15.2 12.5  6.1
Lung 14.8 18.0 18.7 17.5 11.3
Breast 59.8 87.4 88.6 80.9 82.0
Cervix 55.1 61.1 62.8 59.9 64.8
Ovary 41.8 31.8 40.9 38.9 37.3
Prostate 58.5 96.1 97.2 63.8 77.2
Leukemia (adults) 13.6 20.3 51.8 21.2 56.8
Acute lymphocytic 54.1 65.8 87.7 61.1 87.2
­leukemia (children)

The differences in survival between countries long latency periods, such as breast or prostate
and populations are caused by aberrations cancer.
along the cancer care continuum, extending Comparative data on cancer survival among
from primary prevention to access to treatment countries, particularly high-​ income countries
to health promotion for survivors. Specifically, with greater survival rates, serve as a benchmark
wide differences persist among different for reducing cancer burden. Future improve-
countries, including public awareness, soci- ments can be made through investments in
oeconomic development, human resources, improving awareness, health services infrastruc-
availability of and accessibility to early diag- ture, and diagnostic and/​or treatment accessibil-
nosis and treatment due to the different levels ity (Sankaranarayanan et al, 2011).
of development of health services, and ongoing
health service investments (Sankaranarayanan CONCLUSION
et al, 2011). For example, the survival rate of Epidemiological observations support the
stomach cancer is higher in China than in hypothesis that cancer development and pro-
non-​Asian countries, in part due to localized gression is due to an interaction of a constella-
screening programs and public awareness of tion of environmental exposures with the genetic
high stomach cancer rates within the country makeup of the populations. This underscores the
(Zhao et al, 2009). importance of using various and complemen-
Data quality and reliability issues also tary epidemiological measurements to obtain a
remain pertinent, although they affect survival cohesive and comprehensive panorama of cancer
rate disparities to a lesser extent. For low/​mid- burden. Efforts to quantify the impact of can-
dle-​income countries, differences in overall cer are limited primarily by the fact that only a
survival can arise if the cancers distributions small proportion of the global population is cov-
differ, even if therapeutic success is identical ered by cancer registries. The two most populous
across settings for specific cancer sites. This is countries, China and India, have only partial
particularly true if cancers with a poor prog- death certification coverage (4% and 9%, respec-
nosis, such as lung, esophagus, or liver, account tively) and this coverage is not representative of
for a larger proportion of the total. In addition, the entire population. In 75 countries, includ-
an excess death rate may continue for a longer ing more than 90% of African countries, no
period of time for solid malignancies with information on cause of death was available after
54

54 Textbook of Cancer Epidemiology

1990 (GBD 2015 Mortality and Causes of Death empirical evaluation, computational realisation
Collaborators, 2016). Hence, any global estimate and applications. Eur J Cancer 2004;40:326–​35.
relying on extrapolation and statistical mod- Burton EC, Troxclair DA, Newman WP 3rd.
els is subject to inaccuracies at the expense of Autopsy diagnoses of malignant neoplasms: how
individuals in the lowest-​income and politically often are clinical diagnoses incorrect? JAMA
isolated countries, whose cancer experiences 1998;280:1245–​1248.
remain largely unknown. Cancer registration is Choi KS, Suh M. Screening for gastric cancer: the use-
an essential component of any cancer control fulness of endoscopy. Clin Endosc 2014;47:490–​6.
Chuang SC, Lee YC, Hashibe M, Dai M, Zheng T,
program (WHO, 2002); data on cancer occur-
Boffetta P. Interaction between cigarette smoking
rence and prevalence of risk factors should be
and hepatitis B and C virus infection on the risk
collected according to the basic demographic
of liver cancer: a meta-​analysis. Cancer Epidemiol
characteristics (e.g., age, sex, urban versus rural Biomarkers Prev 2010;19:1261–​8.
areas, ethnicity) to provide the basis for alloca- Clegg LX, Hankey BF, Tiwari R, Feuer EJ Edwards
tion of resources and priority setting. BK. Estimating average annual per cent change in
The occurrence of cancer and its course fol- trend analysis. Stat Med 2009;28:3670–​82.
lowing diagnosis and treatment show many Dickman PW, Adami HO. Interpreting trends in can-
salient features. Foremost, the incidence and cer patient survival. J Int Med 2006;260:103–​117.
mortality of many types of cancer vary between Doll R, Fraumeni JF, Muir CS, eds. Trends in Cancer
different geographic settings, suggesting that Incidence and Mortality. Vol. 19/​20. New York: Cold
a large proportion of the disease burden is not Spring Harbor Laboratory Press, 1994.
explained by genetic differences in populations Enstrom JE, Austin DF. Interpreting cancer survival
and can be reduced by intervening on modifia- rates. Science 1977;195:847–​51.
ble risk factors. In low/​middle-​income countries, Fay MP, Tiwari RC, Feuer EJ, Zou Z. Estimating
this burden is expected to increase if the increas- average annual percent change for disease rates
ing cultural globalization combined with lack of without assuming constant change. Biometrics
prevention programs continue. While the former 2006;62:847–​54.
is inevitable, the latter is avoidable. Third, these Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers
trends highlight the need for identifying preven- C, Rebelo M et al. Cancer incidence and mortality
tive measures and treatments aimed at reducing worldwide: sources, methods and major patterns in
the burden of cancer, in particular those cancers GLOBOCAN 2012. Int J Cancer 2015;136:E359–​86.
Forman D, Bray F, Brewster DH, Gombe Mbalawa C,
showing increasing incidence, such as liver cancer
Kohler B, Piñeros M et al, eds. Cancer Incidence
in men and women and lung cancer in women.
in Five Continents. Vol. 10. Lyon:  International
Agency for Research on Cancer, 2013. Available
REFERENCES
at: http://​ci5.iarc.fr (accessed February 17, 2016).
Allemani C, Weir HK, Carreira H, Harewood R, Spika
Forouzanfar MH, Alexander L, Anderson HR,
D, Wang XS et  al. Global surveillance of cancer
Bachman VF, Biryukov S, Brauer M et al. Global,
survival 1995–​2009: analysis of individual data for
regional, and national comparative risk assess-
25,676,887 patients from 279 population-​based
ment of 79 behavioural, environmental and occu-
registries in 67 countries (CONCORD-​2). Lancet
pational, and metabolic risks or clusters of risks
2015;385:977–​1010.
in 188 countries, 1990–​2013:  a systematic analy-
Althuis MD, Dozier JM, Anderson WF, Devesa SS,
sis for the Global Burden of Disease Study 2013.
Brinton LA. Global trends in breast cancer inci-
Lancet 2015;386:2287–​323.
dence and mortality 1973-​1997. Int J Epidemiol
GBD 2015 Mortality and Causes of Death Collaborators.
2005;34:405–​12.
Global, regional, and national life expectancy, all-​
Bailar JC 3rd, Gornik HL. Cancer undefeated. N Engl
cause mortality, and cause-​specific mortality for
J Med 1997;336:1569–​74.
249 causes of death, 1980–​2015: a systematic anal-
Barry MJ. Prostate-​specific antigen testing for early
ysis for the Global Burden of Disease Study 2015.
diagnosis of prostate cancer. N Engl J Med
Lancet 2016 Oct 8;388(10053):1459–​1544.
2001;344:1373–​7.
Hankey BF, Feuer EJ, Clegg LX, Hayes RB, Legler JM,
Bosetti C, Levi F, Boffetta P, Lucchini F, Negri E, La
Prorok PC et al. Cancer surveillance series: inter-
Vecchia C. Trends in mortality from hepatocel-
preting trends in prostate cancer—​part I: Evidence
lular carcinoma in Europe, 1980–​2004. Hepatol
of the effects of screening in recent prostate can-
2008;48:137–​45).
cer incidence, mortality, and survival rates. J Natl
Brenner H, Gefeller O, Hakulinen T. Period analy-
Cancer Inst 1999;91:1017–​24.
sis for “up-​to-​date” cancer survival data:  theory,
5

Measures and Estimates of Cancer Burden 55

Hashim D, Boffetta P, La Vecchia C, Rota M, Bertuccio Parkin DM, Pisani P, Ferlay J. Estimates of the world-
P, Malvezzi M, Negri E. The global decrease in wide incidence of 25 major cancers in 1990. Int J
cancer mortality:  trends and disparities. Ann Cancer 1999;80:827–​41.
Oncol 2016 May;27(5):926–​33. Parkin DM, Whelan SL, Ferlay J, Storm H. Cancer
Helgesen F, Holmberg L, Johansson JE, Bergstrom R, Incidence in Five Continents, Vols. 1–​8. IARC
Adami HO. Trends in prostate cancer survival in CancerBase No. 7.  Lyon, France:  International
Sweden, 1960 through 1988: evidence of increas- Agency for Research on Cancer, 2005b.
ing diagnosis of nonlethal tumors. J Natl Cancer Parkin DM, Whelan SL, Ferlay J, Teppo L, eds.
Inst 1996;88:1216–​21. Cancer Incidence in Five Continents, Vol.
Jong-​wook L. Global health improvement and 8.  IARC Scientific Publication No. 155. Lyon,
WHO: shaping the future. Lancet 2003;362:2083–​8. France:  International Agency for Research on
La Vecchia C, Rota M, Malvezzi M, Negri E. Potential Cancer, 2002.
for improvement in cancer management:  reduc- Peleteiro B, La Vecchia C, Lunet N. The role of
ing mortality in the European Union. Oncologist Helicobacter pylori infection in the web of gastric
2015;20:495–​8. cancer causation. Eur J Cancer 2012;21:118–​25.
Lee J, Demissie K, Lu SE, Rhoads GG. Cancer inci- Pimenta JM, Galindo C, Jenkins D, Taylor SM.
dence among Korean-​ American immigrants in Estimate of the global burden of cervical adeno-
the United States and native Koreans in South carcinoma and potential impact of prophylactic
Korea. Cancer Control 2007;14:78–​85. human papillomavirus vaccination. BMC Cancer
Malvezzi M, Bertuccio P, Rosso T, Rota M, Levi F, La 2013;13:553.
Vecchia C et  al. European cancer mortality pre- Ries LAG, Kosary CL, Hankey BF, Miller BA, Harras
dictions for the year 2015: does lung cancer have A, Edwards BK, eds. SEER Cancer Statistics
the highest death rate in EU women? Ann Oncol review, 1973–​ 1994. NIH Pub. No. 97-​ 2789.
2015;26:779–​86. Bethesda, MD: National Cancer Institute, 1997.
Marugame T, Katanoda K. International comparisons Salomon JA, Haagsma JA, Davis A, de Noordhout
of cumulative risk of breast and prostate cancer, CM, Polinder S, Havelaar AH et  al. Disability
from cancer incidence in five continents Vol. VIII. weights for the Global Burden of Disease 2013
Jpn J Clin Oncol 2006;36:399–​400. study. Lancet Glob Health 2015;3:e712–​23.
McCormack VA, Boffetta P. Today’s lifestyles, tomor- Salomon JA, Vos T, Hogan DR, Gagnon M, Naghavi
row’s cancers:  trends in lifestyle risk factors for M, Mokdad A et al. Common values in assessing
cancer in low-​and middle-​income countries. Ann health outcomes from disease and injury:  dis-
Oncol 2011;22:2349–​57. ability weights measurement study for the
McDavid K, Lee J, Fulton JP, Tonita J, Thompson TD. Global Burden of Disease Study 2010. Lancet
Prostate cancer incidence and mortality rates and 2012;380:2129–​43.
trends in the United States and Canada. Public Sankaranarayanan R, Swaminathan R, Lucas E.
Health Rep 2004;119:174–​86. Cancer survival in Africa, Asia, the Caribbean
Murray CJL, Lopez AD, eds. The Global Burden of and Central America (SurvCan). IARC Sci Publ
Disease: A Comprehensive Assessment of Mortality 2011;162:1–​5.
and Disability from Diseases, Injuries and Risk Siegel RL, Miller KD, Jemal A. Cancer statistics, 2015.
Factors in 1990 and Projected to 2020. Global Burden CA Cancer J Clin 2015 Jan-​Feb;65(1):5–​29.
of Disease and Injury Series, Vol. 1.  Cambridge, Smith PG. Comparison between registries:  age-​
MA: Harvard School of Public Health, 1996a. standardized rates. In:  Cancer Incidence in Five
Murray CJL, Lopez AD, eds. Global Health Statistics. Continents, Vol. 6. Lyon, France:  International
Global Burden of Disease and Injury Series, Vol. 2. Agency for Research Cancer, 1992: 862–​4.
Cambridge, MA, Harvard School of Public Surveillance, Epidemiology, and End Results (SEER)
Health, 1996b. Program. SEER*Stat Database:  Incidence—​SEER
Murray CJ, Ezzati M, Flaxman AD, Lim S, Lozano 9 Regs Research Data, Nov 2014 Sub (1973–​2012)
R, Michaud C et  al. Global Burden of Disease Katrina/​Rita Population Adjustment—​Linked
2010:  design, definition, and metrics. Lancet to County Attributes—​ Total U.S., 1969–​ 2013
2012;380:2063–​6. Counties, National Cancer Institute, DCCPS,
Parkin DM. The global health burden of infection-​ Surveillance Research Program, Surveillance
associated cancers in the year 2002. Int J Cancer Systems Branch, April 2015. Available at:  www.
2006;118:3030–​44. seer.cancer.gov (accessed February 18, 2016).
Parkin DM, Bray F, Ferlay J, Pisani P. Global can- World Health Organization (WHO). National
cer statistics, 2002. CA Cancer J Clin 2005 Cancer Control Programmes:  Policies and
Mar-​Apr;55(2):74–​108. Managerial Guidelines. 2 ed. 2002:  119–​ 123.
56

56 Textbook of Cancer Epidemiology

Available at:  http://​www.who.int/​cancer/​pub- Zhao P, Dai M, Chen Wanqing, Li N. Cancer trends in

lications/​nccp2002/​en/​ (accessed February China. Jpn J of Clin Oncol 2009;40:281–​5.
18, 2016). Zhou CK, Check DP, Lortet-​Tieulent J, Laversanne
World Health Organization Global Health Observatory M, Jemal A, Ferlay J et  al. Prostate cancer inci-
Data Repository. Census and Civil Registration dence in 43 populations worldwide: An analysis of
Coverage Data by Country, 2015. Available at: time trends overall and by age group. Int J Cancer
http://​ apps.who.int/​ g ho/​ d ata/​ n ode.main.121 2016;138:1388–​400.
(accessed February 12, 2016).
57

3
The Origins of Cancer
J E F F R E Y E C S E D Y A N D D AV I D J .   H U N T E R

A n understanding of the origins of cancer

is fundamental to cancer epidemiology.
In order to evaluate risk factors and preventive
including cell-​cycle control, limitations on repli-
cative potential, absence of appropriate growth
stimuli, metabolic restrictions, programmed
behaviors in cancer, one must consider the com- cell death, and immune-​mediated eradication.
plex and interwoven mechanisms that lead to In addition, transformed cells must alter their
the development of cancer. By understanding extracellular environment to support continuous
such mechanisms, an epidemiologist should be cellular propagation and tumor expansion. The
able to integrate biological plausibility into his factors that facilitate cancer initiation and pro-
or her research. This chapter covers more bio- gression are often the same factors that regulate
logical ground than is customary in epidemiol- normal cell and tissue behavior. It is through pro-
ogy texts, but we deemed this desirable for three gressive genetic and epigenetic changes that these
reasons: First, the rapid and continuous develop- regulatory mechanisms go awry, causing cells to
ments in cancer biology can create knowledge gradually transform from a normal into a malig-
gaps for many midcareer epidemiologists as well nant state. Understanding how cancer arises is
as for epidemiologists with predominantly statis- essential to all oncology fields, including cancer
tical rather than biological backgrounds. Second, epidemiology. The genetic predispositions and
the natural history of cancer should be accounted external risk factors associated with increased
for in epidemiologic research and often dictates cancer occurrence manifest themselves as molec-
conceptual and statistical approaches in differ- ular and cellular events, and it is these events that
ent situations—​for example, in studying factors are ultimately responsible for the development of
that initiate rather than promote carcinogene- the cancerous state.
sis. Third, this is an area where new and relevant
knowledge is continuously generated, and epide- DEFINING CANCER
miologists should have sufficient foundational Cancer is a diverse family of diseases that
understanding of cancer biology to follow these spawn from almost every cell type in the body.
pertinent developments. The intent of this chap- Each cell type gives rise to distinct forms of can-
ter is to convey basic molecular and cellular con- cer, and the diversity is greatly increased by the
cepts essential to understanding cancer biology; fact that multiple forms of cancer can develop
it is not intended to give a complete overview from each cell type, depending on both the loca-
of cancer biology. For a more comprehensive tion of the cell and the underlying genetic aber-
exposition on the origin and development of rations. Moreover, most tumors are genetically
cancer, read The Biology of Cancer, 2nd Edition unstable and therefore are composed of genet-
(Weinberg, 2013). ically diverse subclones within a single tumor
The development of cancer is an evolution- mass (Burrell & Swanton, 2014). Despite broad
ary process beginning with an accumulation of diversity, several features are common to all can-
genetic alterations within a single cell, leading to cers (Hanahan & Weinberg, 2011). These include:
the continuous proliferation and expansion of
that cell’s descendants, and culminating in one • unrestricted cellular proliferation
or more invasive masses that ultimately can kill • circumvention of cell-​cycle control
the organism. For cancer to evolve along this • growth without appropriate signals
path, a multitude of barriers must be overcome, • modified metabolism
58

58 Textbook of Cancer Epidemiology

TABLE 3-​1   COMPARISON OF BENIGN AND MALIGNANT TUMORS

Benign tumor Malignant tumor

Invasiveness Noninvasive, often encapsulated Invasive

Rate of Growth Slow, often static Rapid
Differentiation Well differentiated; often resembles Undifferentiated
tissue of origin
Metastasis Never Often metastatic

• escape from programmed cell death growth-​ restrictive mechanisms, thereby con-

• altered interactions between cells and the ferring a selective advantage to abnormal cells,
surrounding environment allowing them to outgrow and displace their
• evasion of immune-​mediated eradication neighbors (Cairns, 1975; Weinberg, 1996). In
• invasiveness into normal tissue addition, the accumulation of genetic aberra-
tions detected in cancer may result from a muta-
The term “cancer” derives from the Greek tor phenotype. Mutator phenotypes are caused
word for crab, “karkinoma,” which the physi- by mutagenic events early in cancer development
cian Hippocrates used to describe the radiating that subsequently jeopardize genomic stabil-
appendage-​like projections extending from some ity and the fidelity of DNA replication, thereby
breast tumors. Cancer manifests itself as either a predisposing tumor cells to the accumulation
solid mass or a nonsolid leukemia in the circu- of further mutations (Fox et al, 2013; Jackson &
latory system. The term “tumor” is nonspecific Loeb, 1998; Loeb, 1991). Given the known muta-
for a lump or swelling, and tumors are charac- tion rate of normal cells, it is unlikely that cells
terized as either benign or malignant (Table 3-​1). would acquire sufficient mutations to become
The hallmark of malignancy is the invasiveness cancerous in a human lifetime. In other words,
of tumor cells into the surrounding and/​or dis- the large number of genetic alterations present
tant normal tissue. Malignant tumor cells invade in many cancers cannot be accounted for by ran-
the surrounding tissue, whereas benign tumor dom mutagenic events (Loeb, 1991). Therefore,
cells remain clustered in a single mass and are the mutator phenotype of premalignant cells
often encapsulated. Benign tumors are usually causes them to be more susceptible to mutations
not fatal, although at certain anatomic sites, such than normal cells.
as in the skull, they can cause pressure on other Most solid tumors originate from areas of
organs and lead to substantial morbidity and, if hyperplasia, characterized by increased local tis-
untreated, death. Secondary growths at a distance sue size due to abnormal cellular proliferation.
from the primary cancer are called metastases. Cells from hyperplastic tissue preserve their
Several characteristics are used to determine the normal architecture and orientation. For exam-
degree of tumor malignancy, including rate of ple, hyperplastic glandular epithelial cells retain
growth, degree of differentiation, extent of inva- a normal polarized columnar shape. When nor-
siveness, and metastatic potential. In most cases, mal cellular architecture and orientation are
metastasis characterizes the highest degree of lost, the abnormal tissue is termed dysplasia.
tumor malignancy, as it is usually the cause of Eventually the cytologic features of the abnormal
death in cancer patients. cells progress to resemble fully developed cancer.
However, if no abnormal cells invade the normal
Cancer Progression underlying tissue (the hallmark feature of malig-
Cancer progression can occur over long periods nancy), the abnormal tissue is defined as carci-
of time, whereby a prospective proliferating can- noma in situ. Once cells invade the underlying
cer cell clone progressively accumulates genetic tissue, the abnormal tissue is defined as cancer.
aberrations resulting in multiple descendent Genetic aberrations can continue to accumulate
subclones, some of which may be bestowed in cancerous tissue and may alter the growth
with increased tumor malignancy. The early rate, the extent of invasiveness, and the meta-
mutagenic events of cancer initiation overcome static potential. Progressive genetic changes may
59

The Origins of Cancer 59

also influence the efficacy of various anticancer (ECM substructures) that normally act as migra-
therapies, including surgical resection, radiation tion barriers. The tumor cells then adhere to
therapy, and chemotherapy. and enter the endothelial-​lined walls of blood
It is not unreasonable to expect that genetic or lymphatic vessels through a process known
lesions that lead to the initial transformation of as intravasation. While tumor cells are migrat-
cells also increase proclivity toward cancer phe- ing to distant sites through these vessels they are
notypes characteristic of more advanced malig- referred to as circulating tumor cells. The thin
nancies, including invasiveness and metastasis. endothelial walls of capillaries and some lym-
However, within the Darwinian view of cancer phatic channels provide little resistance to tumor
progression, transformed cells with a predispo- cell penetration. Tumor cells need to survive
sition for increased malignancy would still rely the mechanical stress and shearing forces of the
on further genetic alterations for these genetic circulatory flow; in fact most circulating tumor
lesions to manifest as advanced cancer. In fact, cells are not alive. Those that do survive can then
when one views cancer progression in the context adhere to and exit vessel walls through a proc-
of natural selection, there is a relatively straight- ess known as extravasation. Finally, for the met-
forward comparison between how mutations astatic tumor cell to proliferate and develop into
drive cancer evolution and how genomic altera- a secondary tumor mass, it must be able to adapt
tions drive the evolution of new species. Similar to a foreign environment, overcome the loss of
to emergence of new species, the survival and survival and growth signals that were present in
subsequent proliferation of any one cancer cell is the primary tumor, and eventually develop its
dependent on its ability to respond to the existing own vascular network. Although dissemination
environmental pressures, and this dependency is of cancer cells to distant sites is a frequent event
a reflection of the cell’s underlying genetics. in many tumors, few of these cells will evolve
to form detectable tumors (macrometastases).
Metastasis Formation of macrometastases, a process known
Metastasis is defined as the migration of tumor as metastatic colonization, is a rare event due to
cells from a primary mass to distant sites in the absence of survival signals unique to the pri-
the body, where the migrating cells take resi- mary site that normally cultivate tumor growth
dence and eventually develop into secondary and inhibitory factors in the secondary site that
tumor masses. Tumors can metastasize to mul- suppress tumor growth.
tiple regions of the same organ or to different
organs, making complete eradication of meta- Cellular Origin of Cancer
static cancers extremely difficult. In almost 50% Two models address the cellular origins of can-
of patients, surgical excision of the primary cer (Fig. 3-​1). In the first model, cancer originates
tumor does not cure the disease, as metastasis from a single cell, which undergoes continuous
has already occurred, sometimes as undetectable heritable alterations initiating proliferation and
micrometastases that may not become apparent expansion, ultimately resulting in the billions of
for years or even decades (Fidler & Balch, 1987; cells often necessary for tumor detection. As the
Fidler & Ellis, 1994). Metastatic cancers have the genetically heterogeneous cells of the cancer orig-
potential to disrupt many organ systems. For inated from a single ancestor clone, the cancer is
these reasons, metastasis is considered the most said to have a monoclonal lineage. An alternative
severe and common life-​threatening complica- model suggests that multiple cells within a tissue
tion of cancer. simultaneously switch to a cancerous state, possi-
Many complex interdependent events need to bly as a result of an exogenous agent. As the cells
occur in the tumor cells and in the surrounding of the cancer are descendants of many clones, the
stroma for metastasis. The tumor cells must first cancer is said to have a polyclonal lineage.
detach from neighboring cells and invade the Experimental evidence supports a monoclo-
surrounding stroma. This is mediated by altered nal origin of cancer. For example, a normal female
cell–​
cell and cell–​extracellular matrix (ECM) contains two classes of cells: those with either the
interactions. Some tumor cells decrease or alter maternal or paternal X-​chromosome inactivated,
the expression of cell–​cell cohesive molecules, an event that occurs in each cell early in devel-
allowing them to detach from adjacent cells. In opment. Therefore, the descendants of each cell
addition, tumor cells utilize specialized proteases possess the same inactivated X-​ chromosome.
to destroy the surrounding basement membranes The state of X-​chromosome inactivation, either
60

60 Textbook of Cancer Epidemiology

Monoclonal Origin of Cancer Polyclonal Origin of Cancer

Transformation Transformation
event events

Normal cells Normal cells

Cancer cell
Cancer cells

Homogeneous
tumor mass

Accumulation of Heterogeneous
genetic aberrations tumor mass

Heterogeneous
tumor mass

FIGURE 3-​1  The cellular origin of cancer. Two models address the cellular origin of cancer, a monoclonal model and
a polyclonal model. In the monoclonal model, a single normal cell is transformed into a cancerous cell through vari-
ous mutagenic events. The cancer cell is then able to proliferate uncontrollably to form a homogeneous tumor mass.
Subsequent genetic aberrations sustained by some of these cells result in the heterogeneous population of cancer cells
characteristic of many tumors. Most experimental evidence favors a monoclonal origin of cancer. In the polyclonal
model, multiple normal cells are simultaneously transformed into cancer cells. These cancer cells are then able to
proliferate uncontrollably to form a heterogeneous tumor mass. 

maternal or paternal, and the genes expressed by results in the formation of an aberrant oncogene,
either chromosome can then be used as inherited which has a role in CML pathogenesis. The site of
markers to trace cell lineage. The X-​chromosome the breakpoint and rejoining has been sequenced
encodes two isoforms of the glucose-​6-​phosphate and can vary when different patients are com-
dehydrogenase (G6PD) gene. Hematopoietic pared; however, the breakpoint site is the same
cancers arising from females heterozygous for for all the CML cells in a given patient (Zalcberg
G6PD express only one isoform of the gene, sug- et al, 1986). It is unlikely that for a single patient
gesting that all of the cancer cells from individ- several cells undergo identical translocations,
ual patients derive from a single cell (Fialkow, and more likely that all the leukemia cells derive
1979). In the majority of cases examined for from a single ancestor clone and possess a mon-
X-​chromosome inactivation in many forms of oclonal lineage.
cancer, all of the cells from an individual tumor Despite the monoclonal origin of cancer,
were found to have either the maternal or pater- tumors and leukemias comprise heterogene-
nal copy of G6PD expressed, but not both. This ous populations of malignant cells (Fig. 3-​ 1).
evidence suggests that all the transformed cells The recent application of next-​generation DNA-​
from a tumor derive from a single clone and sup- sequencing technologies has enabled a renewed
ports the theory of a monoclonal origin of cancer. recognition and deepened the understanding on
Another classic example demonstrating the extent of genetic heterogeneity that occurs
the monoclonal lineage of cancer comes from between different tumors and within individual
studying chronic myeloid leukemia (CML), a tumors (Burrell & Swanton, 2014). Tumor het-
disease that typically occurs in adults. Chronic erogeneity is driven by the fact that tumor cells
myelogenous leukemia is associated with a spe- continue to acquire genetic mutations after the
cific chromosomal abnormality resulting from a initial entry into a transformed state and evolve
translocation between the long arms of chromo- through natural selection into more aggressive
somes 9 and 22 (Khouri, 1997). This translocation forms, resulting in a worse prognosis for the
61

The Origins of Cancer 61

patient. For example, the majority of glioblasto- cancer stem cells have undergone mutations
mas (the most malignant grade of brain tumors) rendering these regulatory mechanisms inef-
originate from less-​malignant astrocytomas and fective. Despite their malignant state, these cells
anaplastic astrocytomas. After surgical resection maintain the capacity to differentiate into their
and radiation of brain tumors, secondary tumors programmed fates. An example of this is mul-
that arise from incomplete resection can have tiple myeloma, a lymphoid stem cell leukemia.
a higher grade of malignancy than the original This disease is characterized by the presence of
tumor. Another example is melanoma. Most mel- extremely high levels of B lymphocytes in all
anoma patients with a specific mutation in a gene stages of development, ranging from the precur-
known as BRAF initially respond to a targeted sor stem cells to the fully differentiated antibody-​
therapy that specifically binds to and inhibits the secreting plasma cells. However, outside of
mutated BRAF protein, but eventually relapse hematological malignancies, the cancer stem cell
occurs. This is due to either acquired or innate hypothesis has been challenged, in large part due
resistance. In acquired resistance, new mutations to the inability to identify this cell population in
occur in some tumor cells that confer resistance most tumors. Nevertheless, the cancer stem cell
mechanisms to the targeted therapy. In innate niche has garnered significant attention, as these
resistance, tumor subclones that harbor intrin- cells have been hypothesized to be resistant to the
sic mechanisms that allow them to be insensitive majority of therapeutic interventions and there-
to the targeted therapy exist as a minority pop- fore drivers of refractory and relapsed disease.
ulation at the time of initiation of the therapy.
For both acquired and innate resistance, as the Clinical Behavior of Cancer
sensitive tumor subclones die, the insensitive Several factors provide important clues in pre-
subclones continue to proliferate and eventually dicting the biological and clinical behavior of
become the dominant subclone. cancer, which may be necessary for directing
Tumors can continue to accumulate genetic appropriate therapy. Histogenesis (tissue origin)
mutations simply as a result of their rapid rate of is the predictor of tumor behavior most com-
cell division. It is a general rule that cells under- monly used by cancer pathologists. In simple
going rapid turnover have a much greater chance terms, histogenesis is based on the concept that
of acquiring mutations. This is due primarily to tumors derived from the same tissue gener-
the increase in the likelihood of acquiring genetic ally behave in a similar way and tumors derived
mutations during DNA replication. Mutations from different tissues behave differently. Another
are also amplified in tumors when there are important prognosticator of tumor behavior is
lesions in DNA mutation detection or repair the histologic pattern of tissue differentiation.
systems. These systems, which include cell cycle In general, the more tumors resemble the tissue
checkpoint and DNA repair mechanisms, are from which they derived, the less malignant they
discussed in greater detail later on in this chapter. are and the better the prognosis.
Tumors are often graded as well, moderately,
Cancer Stem Cells or poorly differentiated. Poorly differentiated
There has been major debate regarding the tumors do not resemble the tissue of origin and
hypothesis that cancer originates from the trans- are generally highly malignant. To return to the
formation of normal stem cells (Pattabiraman example of brain tumors, astrocytomas can be
& Weinberg, 2014; Shackleton et al, 2009; Singh subdivided into fibrillary, gemistocytic, and pro-
et al, 2015). Normal stem cells are undifferenti- toplasmic based on their morphologic similarity
ated, and undergo self-​renewing cell division, to normal and reactive astrocytes. In contrast,
whereby one daughter cell can proliferate into glioblastoma (a grade four astrocytoma) cells
differentiated progeny, and the other daughter retain no resemblance to astrocytes and confer
cell retains the original stem cell characteristics a grim prognosis. The proliferative status, indi-
of being undifferentiated, and maintaining the cated by the number of cycling cells, and various
capacity for unlimited replication. The cancer molecular markers are also used to predict the
stem cells hypothesis postulates that a subset of behavior of cancers.
cancer cells exists in tumors that have the ability The stage of tumor development is another
for unlimited self-​renewing cell division, similar commonly used prognostic indicator. Tumor
to normal stem cells. However, whereas prolif- stage refers to the extent of tumor development
eration of normal stem cells is strictly regulated, at the time of presentation; the higher the stage,
62

62 Textbook of Cancer Epidemiology

the worse the prognosis. Several factors contrib- Proto-​oncogenes are nonmutated alleles of
ute to tumor stage, including tumor size, extent genes that contribute to cellular replication dur-
of invasiveness in the surrounding normal tissue, ing normal growth and maintenance of tissues.
presence of lymph nodes involved with meta- When proto-​oncogenes sustain mutations that
static spread, and presence of distant metastases. impede their normal regulation, they become
Molecular characteristics of tumors are more hyperactivated and uncontrollably stimulate cel-
frequently being applied as prognostic and diag- lular replication. Hyperactivated mutant alleles
nostic tools as well. For example, breast cancer of proto-​oncogenes are termed oncogenes.
can be divided into several subtypes based on the Tumor suppressor genes prevent cell growth
presence or absence of specific molecular markers. in the absence of proper mitogenic (growth-​
These breast cancer subtypes include luminal A, stimulatory) signals. Although the name implies
luminal B, HER2 type and triple negative/​basal-​ that tumor suppressor genes function in sup-
like. Both luminal A and luminal B breast cancers pressing cancer initiation, these genes also par-
are characterized by the expression of the estrogen ticipate in regulating growth during normal
receptors (ERs) and/​or progesterone receptors development. Functional mutations in tumor
(PRs). Luminal B breast cancers are HER2 posi- suppressor genes result in the loss of growth-​
tive or are highly proliferative. HER2 is a recep- inhibitory mechanisms. Restoration of tumor
tor tyrosine kinase, and HER2 positive cancers are suppressor gene function to a cancer cell should
characterized by HER2 protein overexpression or suppress the cell’s growth.
gene amplification. Luminal A breast cancers are Unlike proto-​oncogenes and tumor suppres-
HER2 negative and are typically not highly pro- sor genes, DNA repair genes do not stimulate or
liferative. Triple negative breast cancers are neg- inhibit cellular replication. As their name implies,
ative for ER, PR, and HER2. Diagnosing breast DNA repair genes function in sensing and repair-
cancers by molecular subtyping has both prognos- ing DNA mutations and therefore act as caretak-
tic and therapeutic value. For example, ER-​and ers of the genome (Kinzler & Vogelstein, 1997).
PR-​ positive breast cancers can be treated with Functional mutations of DNA repair genes result
hormonal therapies, and HER2-​ postive breast in a mutator phenotype that accelerates accumu-
cancers can be treated with therapeutic agents that lation of mutated proto-​oncogenes and tumor
have been designed to selectively bind and inhibit suppressor genes, thereby facilitating cancer ini-
HER2 signaling. Triple negative breast cancer has tiation and progression (Fox et al, 2013; Jackson
the worse prognosis and is typically treated with & Loeb, 1998; Loeb, 1991). Restoration of DNA
an assortment of chemotherapies. More and more, repair gene function to a repair-​deficient cell will
molecular markers are being identified in multiple not affect its growth but will reduce the likeli-
cancers that allow for customized diagnosis and hood of progressive accumulation of mutations.
treatment of an individual’s disease. This practice, Much of what is known about cancer genes
applied to oncology, is known as personalized or stems from studies of viruses in human tumors.
precision medicine. Viral infection has served as a model of tumor-
igenesis in experimental systems and contrib-
A MOLECULAR utes directly to several human cancers (Moore &
APPROACH TO Chang, 2010). In fact, about 15% of all cancers
U N D E R S TA N D I N G   C A N C E R have an infectious etiology (Plummer et al, 2016).
Viruses associated with human cancers include
Cancer Genes Epstein Barr virus, hepatitis B and C viruses,
Some genes involved in the development and human T-​cell leukemia virus type 1, some human
maintenance of normal cells and tissues may, papilloma viruses, and several others. Tumor
in certain circumstances, be involved in cancer viruses contribute to cancers by expressing genes
initiation and progression. These genes fall into that mimic cellular oncogenes or interfere with
three categories:  proto-​oncogenes, tumor sup- the expression and activity of various proto-​
pressor genes, and DNA repair genes. Mutations oncogenes and tumor suppressor genes.
in proto-​oncogenes, tumor suppressor genes, and
DNA repair genes cause cancer. These mutations Cancer Genomics
can be acquired through germline transmission New insights into cancer biology and genetics
(inheritance from a parent) or by exogenous have been made by sequencing the genomes of
somatic damage during life. multiple cancers. Several consortia have been
63

The Origins of Cancer 63

created to assimilate comprehensive genomic is known as the two-​hit theory of carcinogene-

information on multiple cancers, including The sis (Fig. 3-​2). Armitage and Doll first proposed
Cancer Genome Atlas (TCGA) Project and its this theory, suggesting that age-​specific incidence
international version the International Cancer rates of most cancer could be explained by events
Genome Consortium. Over about 8  years until that occur in two stages, one or both of which
2014, TCGA assimilated data on over 10,000 could be somatic mutations in the same growth-​
individual cancers derived from 25 different controlling region of homologous chromosomes
tumor types (Editorial, 2015). Moreover, for (Armitage & Doll, 1957). This concept was elabo-
many tumors, mRNA gene expression and DNA rated by Knudson, who predicted that develop-
methylation patterns were also determined. ment of childhood retinoblastoma depended on
Some of the primary observations made from two genetic hits, subsequently shown to be muta-
these efforts include new ways of classifying can- tions in both alleles of the tumor suppressor gene
cer subtypes based on genetic signatures (i.e., Rb (Retinoblastoma gene) (Knudson, 1971).
shared mutations) rather than histological fea- Mutations resulting in the hyperactivation of
tures, an appreciation of the great complexity of a single proto-​oncogene or the functional inac-
mutational landscapes in most cancers, and the tivation of a single tumor suppressor gene are
correlation of mutational patterns with some car- typically insufficient for cancer initiation. Cells
cinogenic exposures (e.g., ultraviolet light in mel- utilize numerous checks and balances that pro-
anomas and cigarette smoking in non-​small-​cell hibit uncontrolled growth in response to a sin-
lung cancers). One hope is that by sequencing a gle activating signal. Multiple genetic lesions in
patient’s tumor, specific “driver” mutations will various growth regulatory pathways are neces-
be found that may respond to available drugs or sary for cancer to arise from a single cell clone.
lead to the development of new drugs. Examples The continuous accumulation of genetic lesions
of this approach include the development of necessary for cancer genesis is termed the multi-
vemurafenib for advanced melanomas that har- step theory of carcinogenesis. Multiple mutagenic
bor a specific mutation (V600) in the BRAF hits are necessary for cancer development (Hahn
gene. However, hopes that the same mutation et al, 1999; Vogelstein & Kinzler, 1993). The num-
in other tumor types would be equally respon- ber of mutagenic hits necessary for cancer forma-
sive led to mixed results (Hyman et  al, 2015). tion varies, depending on the growth regulatory
For many individual cancers, specific “driver” pathways affected and the level of regulation. The
mutations cannot be found with today’s bioin- accumulation of mutagenic hits results not only
formatic knowledge so the tumor sequence does in the transformation of a cell from a normal into
not provide actionable information. Others have a cancerous state but also in the increase of tumor
said that the main lesson to be drawn from the malignancy after the initial transforming events.
cancer-​sequencing data is that cancers are com-
plex and subject to Darwinian evolution such Exogenous Mutagenesis
that drug resistance arises quickly, and thus in Many exogenous carcinogens cause genetic
most cases new drugs targeting a specific muta- lesions. Several examples exist of exogenous car-
tion will only delay, and not prevent, tumor pro- cinogens leading to somatic mutations, including
gression (Tannock & Hickman, 2016). chemicals such as polycyclic aromatic hydrocar-
bons, toxins, and physical agents such as ultravi-
Genetic Carcinogenesis olet radiation. Although exogenous carcinogens
Oncogenic mutations are dominant at the cellu- induce mutations by various mechanisms, one
lar level, as only one of the two proto-​oncogene unifying outcome is the interference with
alleles needs to be mutated in a cell for the DNA replication fidelity, resulting in a mutated
mutant gene product to influence downstream genetic sequence. The risk of cancer occurrence
events that lead to cellular proliferation. Tumor is increased when these mutations occur in
suppressor mutations are recessive at the cellu- cancer genes.
lar level, as both allele copies must be inactivated Many chemicals that normally do not induce
for cells to overcome the inhibitory mechanisms DNA damage are altered by metabolic path-
of the gene product. DNA repair mutations are ways in the body, thereby inducing carcinogenic
also recessive. The successive mutagenic inactiva- activity and resulting in genetic lesions. Various
tion in both alleles of tumor suppressor or DNA metabolic pathways are responsible for detoxi-
repair genes that facilitate cancer development fying or excreting chemical carcinogens. These
64

64 Textbook of Cancer Epidemiology

(a) Paternal Maternal (b) Paternal Maternal

genes genes genes genes

Inherited germline
mutation = First No inherited
Offspring
mutagenic hit in germline mutation
all cells

Somatic mutation = Somatic mutation =

Second mutagenic First mutagenic hit in
hit in a single cell a single cell

Cancer Somatic mutation =

Second mutagenic hit
in a single cell

Cancer

FIGURE 3-​2  The two-​hit theory of carcinogenesis. A minimum of two mutagenic events, one in each allele of a tumor
suppressor or DNA repair gene, is necessary to facilitate cancer development. A. The offspring inherits the first muta-
genic hit in all cells from a germline mutation in the paternal genes. Therefore, only one somatic mutation in any
cell is necessary for the second mutagenic hit, resulting in the inactivation of both alleles, thereby facilitating cancer
development. B. The offspring inherits no mutagenic hits from the parents. Therefore, two successive somatic muta-
tions in the same cell are necessary for the first and second mutagenic hits, resulting in the inactivation of both alleles. 

processes engender the idea that differences 2010). For instance, hypermethylation of a tumor
in the sequence or abundance of activating or suppressor gene causes the gene product to be
detoxifying enzymes may partially determine absent. Therefore, a combination of an inherited
individual susceptibility to exogenous carcino- or acquired somatic mutation in one copy of the
gens. Therefore, much attention has been focused gene, and a hypermethylated promoter in the
on endogenous molecules participating in these other allele, would result in complete loss of the
metabolic pathways and the role they play in gene product in the cell. This is an example of the
DNA damage (Hussain & Harris, 2000; Irigaray two-​hit theory of carcinogenesis.
& Belpomme, 2010). Hypermethylation is an example of an epige-
netic phenomenon. Epigenetic changes influence
Epigenetic Mechanisms gene expression without directly affecting the
of Gene Silencing gene sequence (Jones & Laird, 1999). As a result,
In addition to mutations in genes or loss of genetic epigenetic changes are not inherited. For some
material from chromosomes, recent work has genes in certain organs, promoter hypermeth-
indicated that genes may be silenced or turned ylation increases with age, perhaps providing
off by nonmutational mechanisms. A  silenced part of the explanation of why many cancers are
gene is still present in the genome, but is not strongly age dependent. Continued understand-
expressed, and is thus functionally equivalent ing the environmental and endogenous causes
to a gene that expresses a nonfunctional mutant of gene silencing will almost certainly prove
product or has been lost altogether. All genes have valuable in elucidating the way epidemiologi-
promoter sequences that determine the state of cally determined risk factors cause cancer cells
expression, and a common mechanism for silenc- to arise.
ing is the addition of methyl groups to key sites Another endogenous epigenetic mechan-
in these promoter sequences (Kulis & Esteller, ism for gene silencing involves microRNAs.
65

The Origins of Cancer 65

MicroRNAs, also known as miRNAs, are single-​ The cell cycle is overseen by various check-
stranded RNA molecules of approximately 22 points, quality controllers that monitor the con-
nucleotides in length. These molecules consti- dition of DNA throughout the cell cycle and
tute a unique category of nonprotein-​ coding ensure proper orchestration of cycle events,
regulators of gene expression (Calin & Croce, thus protecting genomic integrity (Fig. 3-​3). In
2006; Esquela-​Kerscher & Slack, 2006; Jonas & response to DNA damage or replicative stress,
Izaurralde, 2015). The miRNAs interfere with checkpoints perform diverse functions includ-
gene expression at the level of mRNA transcripts ing activation of appropriate effectors neces-
by two mechanisms: (1) they induce degradation sary to repair the damage or alleviate the stress.
of mRNAs to which they have complete or par- Checkpoints arrest the cell cycle until the damage
tial sequence complementarity through a mech- is mended and activate DNA repair machinery.
anism known as RNA interference (RNAi), or In extreme cases, they cause cells to commit sui-
(2) they suppress protein translation by binding cide by initiating programmed cell death (known
to the 3’ untranslated regions of target messen- as apoptosis).
ger RNAs. The number of miRNAs suspected to Checkpoints comprise a complex network of
be present in the human genome ranges up to proteins that act as sensors, transducers, or effec-
1500 (Ameres & Zamore, 2013). These molecules tors (Zhou & Elledge, 2000). The sensors identify
effect a diverse array of biological processes, genomic insults and initiate network activity. An
including cell proliferation, differentiation, and example of a sensor is the ataxia-​telangiectasia
death, and individual miRNAs have been shown mutated protein (ATM), which participates in
to be important regulators of specific tumor sup- identifying DNA double-​ strand breaks. Once
pressor genes and oncogenes. activated, sensor checkpoint proteins signal to

The Cell Cycle and Checkpoints

The division of a cell into two genetically equiv- S
alent daughter cells proceeds in an orderly fash-
ion and is known as the cell cycle. The cell cycle
is divided into four main phases:  the mitotic
phase (M), the postmitotic/​ presynthesis gap
phase (G1), the DNA synthesis phase (S), and Cyclin/CDKs
G1 G2
the postsynthesis/​ premitotic gap phase (G2) Checkpoints
(Fig. 3-​3). Cells that are not actively dividing may
either be temporarily removed from the cycle by
entering a resting state defined as G0, or be per-
manently removed from the cycle by terminal
differentiation. G0 M
Proper orchestration of the cell cycle depends
on the completion of all the steps in one phase FIGURE  3-​3  The successive phases of the cell cycle. The
before the next phase begins. Oscillating lev- cell cycle is characterized by four main phases: a mitotic
els of cyclin/​ cyclin-​dependent kinase (CDK) phase (M), a postmitotic/​ presynthetic phase (G1), a
complexes regulate cell-​cycle progression (Fig. DNA synthetic phase (S), and a postsynthetic/​premitotic
3-​3). These complexes phosphorylate substrates phase (G2). The G1 phase marks the beginning of the
required for DNA synthesis (S phase) or mito- cell cycle. This is followed by the S phase, which is char-
sis (M phase). Several different cyclin/​ CDK acterized by the duplication of each chromosome into
complexes participate in cell-​cycle progression. two sister chromotids. The G2 phase follows the S phase
Exogenous agents and events, including growth and precedes the M phase. The M phase is characterized
factors, mitogens/​ antimitogens, differentiation by nuclear and cytoplasmic division of a parent cell into
inducers, cell–​ cell and cell–​ ECM interactions, two genetically identical daughter cells. Cells that are not
oxygen pressure, and nutrients, ultimately con- dividing may be temporarily removed from the cycle by
trol cyclin/​CDK activity. These agents/​events uti- entering a resting state, defined as G0, or permanently
lize various signal transduction mechanisms to removed from the cycle by terminal differentiation.
control cyclin/​CDK activity by affecting expres- Cyclin/​CDK complexes drive cell-​ cycle progression.
sion levels, activity, and stability and by regulat- Checkpoints monitor the quality of DNA throughout
ing specific inhibitors (CDK inhibitors). the cell cycle and ensure proper orchestration of events. 
6

66 Textbook of Cancer Epidemiology

transducer checkpoint proteins that in turn acti- diverse developmental abnormalities in these
vate a variety of effector checkpoint proteins. As patients, and a higher cancer incidence is shared
is discussed in what follows, ATM (a sensor) acti- by all. Diseases characterized by these defects
vates p53 (a transducer), which stimulates p21 include ataxia-​telangiectasia, Bloom’s syndrome,
(an effector). The p21 protein is associated with Werner’s syndrome, and Fanconi anemia, among
cell-​cycle arrest at the G1/​S transition. several others. These diseases, which are inher-
The checkpoint network regulates phase tran- ited in a recessive manner, are rare.
sition at different points throughout the cell cycle, Ataxia-​telangiectasia is characterized by cer-
although the proteins involved and their activ- ebellar degeneration, immunodeficiency, radi-
ity vary, depending on the phase in which they ation sensitivity, and cancer predisposition.
are activated. The G1/​S checkpoint ensures that Ataxia-​telangiectasia results from mutations in
damaged DNA is repaired prior to replication in the ATM gene that encodes a protein involved in
the S phase. The S phase checkpoint is activated cell-​cycle progression, and possibly in detection
when DNA damage occurs in the early events and repair of double-​strand DNA breaks (Lavin,
of DNA synthesis. Any DNA damage sustained 2008; Savitsky et  al, 1995). Left unrepaired,
at the end of DNA synthesis or prior to mitosis double-​strand DNA breaks lead to high rates of
activates the G2/​M checkpoint. The mitotic spin- chromosomal rearrangements.
dle checkpoint ensures that all chromosomes Bloom’s syndrome is characterized by prena-
are properly attached to mitotic spindles and tal and postnatal growth retardation, immunode-
aligned to the metaphase plate before anaphase ficiency, premature aging, and a high incidence
begins. Defects in the mitotic spindle checkpoint of cancer. Werner’s syndrome is characterized
result in aneuploidy, which can lead to cell-​cycle by premature aging and early onset of osteopo-
arrest, cell death, or continued cell proliferation, rosis, artheosclerosis, type II diabetes mellitus,
increasing the risk for cancer initiation. and cancer. Bloom’s syndrome and Werner’s syn-
The cell-​cycle machinery is commonly tar- drome arise from mutations in genes that encode
geted in tumorigenesis. Regulators of the cell DNA helicases (Ellis et al, 1995; Gray et al, 1997).
cycle qualify as proto-​ oncogenes, tumor sup- Loss of Bloom’s or Werner’s helicase activity can
pressor genes, or DNA repair genes, according lead to entangled DNA strands, causing muta-
to their function in cell-​cycle progression. For tions to arise during DNA replication and chro-
example, hyperactivating mutants of cyclin/​CDK mosomal segregation.
complexes, which facilitate cell-​cycle progression, Fanconi anemia is characterized by growth
are classified as oncogenes. Functional muta- retardation, skeletal malformations, and can-
tions in p53 prevent its cell-​cycle arrest activ- cer predisposition. Unlike the other genomic
ity (mediated through p21 and other proteins); instability disorders, Fanconi anemia arises
therefore, p53 is classified as a tumor suppressor from mutations in any one of at least 16 genes
gene. Defects in some checkpoint mechanisms (Kottemann & Smogorzewska, 2013). Patients
prevent proper orchestration of DNA replication with defects in these genes demonstrate hyper-
or separation of chromosomes during mitosis. sensitivity to DNA cross-​linking agents, as the
Defects in others result in impaired detection Fanconi anemia proteins play a key function
and repair of DNA damage. These checkpoints DNA damage recognition and repair.
are therefore classified as DNA repair (caretaker)
genes. Nevertheless, as DNA repair checkpoints The p53 and Retinoblastoma
ensure DNA integrity, functional defects in Pathways
these checkpoints result in accelerated accumu- The Rb and p53 tumor pathways serve as para-
lation of mutated proto-​oncogenes and tumor digms for understanding cell-​cycle regulation.
suppressor genes. Rb and p53 are components of two signal trans-
duction pathways that negatively regulate cel-
Genomic Instability Disorders lular replication. Description of these pathways
Several human genetic disorders result from here is intended to be illustrative for how cell-​
defects in DNA damage sensing and repair genes cycle progression is negatively regulated; multi-
or in DNA recombination genes. These disorders ple other pathways provide similar functions in
result in gross chromosomal aberrations and cells. In nonreplicating cells, the Rb protein is
are therefore characterized as genomic instabil- normally “turned on,” thus suppressing cell-​cycle
ity disorders. Genomic instability leads to many progression. In response to mitogenic signals,
67

The Origins of Cancer 67

Rb is “turned off,” allowing cells to proceed DNA Damage Oncogenic Singal

through the cell cycle. The p53 protein is nor-
mally “turned off ” in cells. In response to DNA
damage or oncogenic signals, p53 is “turned on,”
thus inducing cell-​cycle arrest or programmed ATM mdm2
cell death.
Both Rb and p53 are tumor suppressors, as p53
their functional inactivation results in the loss of
cell-​cycle control, leading to cellular transforma-
tion. Indeed, the Rb and p53 genes are commonly p21 Bax mdm2
mutated in human cancers. Other components
of the Rb and p53 signal transduction pathways
are also mutated in certain human cancers. These G1/M Programmed p53
comprise various oncogenes and tumor suppres- Cell cycle arrest cell death Downregulation
sor genes, depending on whether they stimulate
FIGURE 3-​4  Signaling through the p53 pathway. The p53
or suppress Rb and p53 activity.
protein is normally turned off in cells. In response to
DNA damage or oncogenic signals, p53 undergoes both
p53
qualitative and quantitative changes and is subsequently
The p53 protein was first described in 1979
turned on. ATM and other protein kinases phosphorylate
through its association with the transforming
p53 in response to DNA damage, altering p53 protein–​
simian virus 40 large T protein (Kress et al, 1979;
protein interactions and increasing p53 DNA binding
Lane & Crawford, 1979; Linzer & Levine, 1979).
affinity. Oncogenic signals free p53 of mdm2, thereby
Subsequently, p53 was identified as a tumor sup-
stabilizing the protein and resulting in increased p53
pressor and remains the most extensively studied
levels. Activated p53 serves as a sequence-​specific tran-
gene in the field of cancer research. The attention
scription factor and promotes expression of target genes,
focused on p53 is due in large part to its loss of
including p21, Bax, and mdm2. Activation of these genes
function in many human cancers. Approaching
results in cell-​cycle arrest, programmed cell death, and
one-​half of all human cancers possess mutations
decreased p53 activity and stability, respectively. 
in at least one copy of the p53 gene that result
in direct inactivation of the p53 protein; the spe-
cific frequency of inactivating mutations dif- genomic perturbations or oncogenic signals pre-
fers between cancer types (Hollstein et al, 1991; vents potentially cancerous cells from attaining
Greenblatt et al, 1994). In addition to mutations uncontrolled growth status and eventually devel-
in the p53 gene itself, many other cancers pos- oping into full-​blown cancer.
sess mutations in genes encoding p53 regulatory Activation of the p53 protein requires both
proteins that result in indirect inactivation of the quantitative and qualitative changes. Quantitative
p53 protein. Germline mutations in p53, though changes result from a dramatic increase in the
rare, can also occur. Li-​Fraumeni syndrome is a steady-​state levels of cellular p53. This is due pri-
genomic disability disorder defined by individu- marily to the removal of a p53-​bound inhibitor,
als who inherit a mutant p53 allele and are pre- mdm2 (Fig. 3-​4) (Bieging et  al, 2014; Bottger
disposed to sarcomas and cancers of the breast, et al, 1997). The mdm2 protein binds to and pro-
brain and adrenal glands. motes rapid degradation of p53. When mdm2 is
Under normal cellular conditions, wild-​type removed, p53 is not targeted for rapid degrada-
p53 protein exists in an inactive state and is pres- tion and p53 protein levels increase. However, the
ent in low levels. In response to various intra- increase in protein levels alone is not adequate
cellular or extracellular stresses, including DNA for p53 to become a transcriptional activator;
damage or oncogenic signals, p53 switches to an qualitative changes that alter p53 conformation
active state and stabilizes (Fig. 3-​4). Activated are also necessary (Bieging et al, 2014; Prives &
p53 serves as a sequence-​specific transcription Hall, 1999).
factor and promotes expression of target genes. The qualitative changes in p53 include post-
The proteins expressed from the p53 target translational modifications, including phospho-
genes then initiate a variety of cellular events, rylation and dephosphorylation events, among
most significantly cell-​cycle arrest and apopto- others. These changes alter the protein–​protein
sis. Therefore, activation of p53 in response to interactions of p53 and increase the DNA
68

68 Textbook of Cancer Epidemiology

binding affinity of p53, allowing it to act as a tran- childhood retinoblastoma, characterized by mul-
scriptional activator. The protein kinase ATM tiple tumor foci in both eyes (Varmus & Weinberg,
and several other protein kinases phosphoryl- 1993). These individuals also are predisposed to
ate p53 in response to DNA damage, leading to other forms of malignancy later in life. The Rb
p53 activation (Fig. 3-​4). The p53 protein is also protein functions as a tumor suppressor prima-
activated in response to aberrant growth signals, rily by inhibiting cell-​cycle progression in the G1
such as overexpressing Ras oncogenes (Groth phase. This activity is mediated by Rb interaction
et al, 2000). with E2F transcription factors, thereby inhibit-
Once activated, p53 acts as a sequence-​ ing the expression of E2F target genes (Dick &
specific transcription factor, upregulating expres- Rubin, 2013; Weintraub et  al, 1992). E2F target
sion of genes whose products induce cell-​cycle genes facilitate progression of the cell cycle. In
arrest (either at the G1 or G2 phase), cellular response to mitogenic signals, the Rb protein
senescence or programmed cell death (Fig. 3-​4). is phosphorylated by CDKs (Fig. 3-​ 5). CDK-​
Additional noncanonical roles for activated p53 mediated phosphorylation of Rb disrupts Rb/​
also include metabolism, invasion, and metasta- E2F transcription factor complexes, resulting in
sis (Bieging et al, 2014). One such target gene is expression of E2F target genes (Adams & Kaelin,
p21, which is partly responsible for arresting the 1998). In addition to inhibiting cell-​cycle pro-
cell cycle in the G1 phase (Abbas & Dutta, 2009). gression, Rb can regulate genomic stability and
When p21 expression is elevated in response to apoptosis (Dick & Rubin, 2013).
p53 activation, it impedes cell-​cycle progression
at the G1 phase by interfering with the CDKs. Growth-​Regulatory Signaling
The p53 target genes leading to programmed Pathways
cell death include Bax and Fas, among others The mediators of growth-​regulatory signal trans-
(Bieging et  al, 2014). The function of these two duction pathways are targets of deregulation in
genes is discussed in greater detail in the follow- cancer. These mediators constitute an extremely
ing section. Interestingly, the mdm2 gene is also a diverse assortment of membrane-​ spanning
target for p53 transcriptional activation, thereby receptors, intracellular transducers, and effec-
allowing p53 to negatively autoregulate its own tors. Membrane-​spanning receptors are normally
activity. activated in response to a multitude of growth-​
stimulatory signals, including growth factor poly-
Retinoblastoma peptides, hormones, and cell–​cell and cell–​ECM
Individuals who inherit a mutant copy of the contacts. Activated receptors then relay signals to
Rb gene are predisposed to the development of intracellular transducers, which elicit a diverse

Mitogenic signal

Cyclins/CDKs

“Rb on” Rb E2F

cell cycle
not expressed
progression genes

P
Rb “off” Rb
E2F cell cycle
expressed
progression genes

FIGURE 3-​5 Signaling through the Retinoblastoma (Rb) pathway. In nonreplicating cells, Rb is turned on and interacts
with E2F transcription factors, thereby repressing expression of genes involved in cell-​cycle progression. Rb is turned
off by phosphorylation through cyclin/​CDK complexes in response to mitogenic signals. Subsequently, E2F tran-
scription factors are liberated from Rb and activate the expression of cell-​cycle progression genes. 
69

The Origins of Cancer 69

range of cellular responses and require complex Ligand

effectors. Depending on whether they positively
or negatively regulate growth signal transmis-
Ras
sion, mediators of these pathways are classified
as proto-​oncogenes or tumor suppressor genes.
Unlike normal cells, cancer cells are able ErbB-2
to divide in the absence of appropriate growth
stimuli. Cancer cells circumvent dependency on
these stimuli by generating their own growth sig-
Other signaling MAPK
nals (via oncogenic mutations) and by escaping pathways cascade
growth-​suppressive systems (via tumor suppres-
sor mutations). In addition, some cancers secrete
polypeptide growth factors that in an autocrine
fashion stimulate their own growth. Importantly,
specific mutations in growth regulatory genes C-Myc
can result in susceptibility to therapies that spe- Gene expression
cifically target those mutations while to a varying
extent limiting unwelcome adverse events to less
targeted therapies. However, further mutations in
FIGURE  3-​6  Growth-​ regulatory signaling pathways.
these growth regulatory genes also lead to thera-
ErbB-​2 is a member of the membrane-​spanning epider-
peutic resistance, making cancers refractory to
mal growth factor receptor family. In response to ligand-​
multiple therapeutic modalities (Holohan et  al.,
mediated activation, ErbB-​2 initiates various signaling
2013). ErbB-​2, Ras, and c-​Myc serve as examples
pathways, among them the mitogen-​activated protein
for cancer-​deregulated membrane-​spanning
kinase (MAPK) cascade. Ras functions as an intracellu-
receptors, intracellular transducers, and effec-
lar transducer of various signaling pathways. For exam-
tors, respectively (Fig. 3-​6). Description of these
ple, Ras transduces ErbB-​2 signals to the MAPK cascade.
genes here is intended to be illustrative for how
c-​Myc functions as an effector by regulating expression
growth regulatory signaling pathways function;
of genes involved in a variety of cellular functions. The
many other growth regulatory pathways exist in
MAPK cascade and several other signaling pathways
cells as well.
activate c-​Myc. ErbB-​2, Ras, and c-​Myc expression are
often deregulated in cancer. 
ErbB-​2
In many cancers, growth factor receptors are
overexpressed or contain hyperactivating muta- activity of ErbB-​2, is used in the treatment of the
tions resulting in their constitutive activation. majority of breast carcinomas that overexpress
One such receptor, ErbB-​2 (also known as HER2), ErbB-​2 (Pegram & Slamon, 2000). Trastuzumab-​
is a member of a transmembrane tyrosine kinase DM1, an ErbB-​2-​specific antibody bound to a
receptor family known as the epidermal growth cellular toxin, is also a commonly used therapy
factor receptors. Deregulation of ErbB-​2 activity as it binds selectively to cancers overexpressing
in cancers is primarily due to gene amplification ErbB-​2 and kills tumor cells upon internalization
and overexpression or to structural mutations of the toxic molecule (Hudis, 2007).
that result in hyperactivation of the tyrosine
kinase domain. In either case, ErbB-​2 signaling Ras
is activated independently of appropriate growth One downstream intracellular transducer of
stimuli. ErbB-​2 signaling is the Ras proto-​oncogene,
Several human cancers, including breast, gas- which is also a target for deregulation in tumori-
tric, and ovarian cancers among others, are asso- genesis. The Ras gene encodes a G protein (gua-
ciated with ErbB-​2 gene amplification or protein nine nucleotide binding protein) that localizes
overexpression (Yan et al, 2015). ErbB-​2 serves as to the inner leaflet of the plasma membrane. Ras
an ideal target for anti–​breast cancer therapy, as it belongs to a large group of related proteins known
is expressed at high levels in many breast cancers as the Ras family of GTPases. After delivery of
and is correlated with a poor prognosis in breast the appropriate signal, Ras binds to the guanine
cancer patients (Lupu et al, 1996). Trastuzumab, nucleotide GTP and undergoes conformational
a monoclonal antibody that blocks the kinase changes that activate the protein. Activated
70

70 Textbook of Cancer Epidemiology

Ras then stimulates many diverse downstream result in increased expression of genes required
pathways, the key signaling pathway being the for glucose uptake and metabolism in cancer
mitogen-​activated protein kinase (MAPK) cas- cells (Dang et al, 2009).
cade. Oncogenic Ras mutants fire continuously ATP production in cancer cells has an
even without input from growth factor receptor increased dependence on aerobic glycolosis and
signals. a decreased dependence on oxydative phospho-
Ras mutations have been implicated in up to rylation relative to normal cells. This phenom-
30% of all human cancers (Hernandez-​Alcoceba enon was first described by Otto Warburg and
et  al, 2000). Ras and its family members also is commonly referred to as the Warburg effect
play a significant role in the intracellular changes (Warburg, 1956). Given that glycolysis is much
associated with cellular migration in tumor inva- less efficient than oxidative phosphorylation in
sion and metastasis. Though Ras itself has proven ATP production per molecule of glucose, cancer
to be intractable for developing targeted thera- cells rely on a high rate of glycolysis for energy
pies, therapeutic agents have been developed for production and for generation of biosynthetic
proteins that signal downstream of Ras including intermediates.
BRAF and MEK (Holderfield et al, 2014). There are several explanations for the glyco-
lytic dependence of tumors (Cairns et al, 2011).
c-​Myc In the presence of sufficient glucose, the rate of
One downstream effector of many growth-​ ATP production occurs faster through glycoly-
regulatory signal transduction pathways is the sis than oxidative phosphorylation. In addition,
proto-​oncogene c-​Myc. c-​ Myc is a transcrip- many tumors often provide an unhospitable
tion factor that regulates expression of many environment for cellular growth, including low
genes associated with a variety of cellular func- nutrient supplies, oxygen pressure, and pH levels.
tions, including replication, growth, metabo- Under low oxygen conditions (also referred to as
lism, differentiation, and apoptosis (Dang, 2012). hypoxia), glycolysis allows for the production of
Regulation of these genes by c-​Myc differs among ATP, unlike oxidative phosphorylation, which
cell types, suggesting that tissue-​specific factors requires oxygen as a key reactant. In addition,
exist that modulate c-​Myc expression. Many of glycolysis may provide an advantage to cellular
these c-​Myc mutants lose the ability to be shut growth and proliferation, as the intermediate of
down by normal mechanisms and thus remain glycolysis pyruvate can be converted to lactate
constitutively active (Nesbit et  al, 1999). c-​Myc and used as a carbon source for biosynthesis of
belongs to a family of transcription factors that the major metabolic building blocks including
also includes L-​Myc and N-​Myc, which are also nucleic acids, amino acids, lipids, and carbohy-
disregulated in cancer. drates. Beyond the increased dependence on
Mutations that disrupt c-​ Myc regulation glycolysis for ATP production, many other met-
are among the most common mutations found abolic pathways are altered in cancer cells, pro-
in human cancer. In fact, most human cancers viding multiple opportunities for exploiting these
display c-​Myc amplification and/​or overexpres- changes for therapeutic intervention.
sion (Nesbit et  al, 1999). Almost all cases of
Burkitt’s lymphoma arise from chromosomal Programmed Cell Death
rearrangements that juxtapose the c-​Myc gene (Apoptosis)
located on chromosome 8 with regulatory ele- Programmed cell death, or apoptosis, is a genet-
ments of immunoglobulin genes located on other ically regulated form of cell death characterized
chromosomes. by a series of intracellular molecular events.
Several changes in cellular morphology distin-
Modified Metabolism guish apoptosis from other forms of cell death
The genetic alterations that impact growth reg- (i.e., necrosis or autophagy), including cell
ulatory signals in cancer cells including both shrinkage, nuclear fragmentation, chromoso-
oncogenic and tumor suppressive mutations mal condensation, and membrane blebbing. The
can also modify or reprogram cellular metabo- removal of cells by apoptosis is an essential proc-
lism pathways leading to the growth advantages ess in embryonic development and in homeosta-
required to support the rapid proliferation of sis of adult tissues (Suzanne & Steller, 2013; Vaux
many cancers (Boroughs & DeBerardinis, 2015). et al, 1988). In developing embryos, for example,
For example, activating mutations in c-​ Myc superfluous cells are removed during digit and
71

The Origins of Cancer 71

brain formation by apoptosis. In adults, apop- contacts the Fas ligand. Fas then activates spe-
tosis participates in the postpartum involution cific members of the caspase family, which ini-
of mammary glands and in homeostasis of cell tiate the proteolytic degradation characteristic of
number by balancing cell birth with cell death. In apoptosis.
the context of cancer, apoptosis provides a means Killer T lymphocytes utilize the death receptor
of eliminating cells that have sustained genetic pathway to destroy target cells. Killer T lymphocytes
aberrations from cytogenetic insults, thereby trigger the death receptor (extrinsic) pathways in
preventing damaged cells from attaining uncon- target cells by expressing cell surface Fas ligand. In
trolled proliferation. addition to Fas, several other death receptors exist
Genetic aberrations, such as DNA damage or that activate apoptosis by similar mechanisms. The
oncogenic signals, can lead to apoptosis by p53 other pathway for activating apoptosis (intrinsic) is
activation, for example, which induces expression triggered when cytogenetic insults, by an unknown
of several proapoptotic genes. In addition, cer- process, alter specific proapoptotic proteins, result-
tain types of DNA damage also trigger apoptotic ing in leakage of cytochrome c from mitochondria
cell death by mechanisms that are independent of into the cytosol (Fig. 3-​7). One such proapoptotic
p53. CD8+ (killer) T lymphocytes can also acti- protein is Bax, whose expression is upregulated in
vate apoptosis in cells that present foreign anti- response to p53 activation. Cytosolic cytochrome
gens as a result of genetic aberrations. Therefore, C induces conformational changes in the interme-
the apoptotic process must fail or be inhibited for diary protein Apaf1, which subsequently activates
a cell to transform to a cancerous state. the caspase cascade.
The changes in cell morphology that distin- Elucidation of the cell death pathways has
guish the apoptotic phenotype reflect the activ- provided insight into the antiapoptotic changes
ity of a family of proteases known as caspases. that occur in tumor cells (Holohan et al, 2013).
Caspases degrade specific intracellular polypep- Many tumor cells express low levels of Fas or
tides that ultimately result in cell death. Caspase none at all, thereby evading killer T-​lymphocyte
activity is activated by distinct pathways:  one is recognition and subsequent death (O’Connell
triggered by receptor-​mediated signaling at the et al, 2000; Sikora et al, 1998). Tumor cells also
cell surface (extrinsic pathway), and the oth- secrete soluble polypeptides that act as decoy
ers are triggered by cytogenetic insults (intrin- Fas receptors by binding surface Fas ligands
sic pathway) (Fig. 3-​7) (McIlwain et  al, 2013). (Pitti et  al, 1998). Saturation of Fas ligands
The receptor-​mediated signaling pathway, often with decoy receptors prevents activation of Fas
referred to as the death receptor pathway, is receptors. One gene encoding a decoy poly-
triggered when the extracellular domain of the peptide was amplified in approximately 50% of
membrane-​ spanning signaling molecule Fas lung and colon carcinomas examined (Pitti et al,

Death receptor pathway Cytogenetic insult pathway

p53 activation
Killer T-lymphocyte

CD8 Fas ligand Bax activation

MHC class I Fas Mitochondrial

cytochrome C leakage
Cancer cell

Caspase activation

CELL DEATH

FIGURE 3-​7 Apoptotic signaling pathways. Apoptosis is induced in cells through death receptor pathways or by cyto-
genetic insult. Killer T lymphocytes recognize foreign antigens presented on the cell surface through CD8–​major his-
tocompatibility complex (MHC) class I interactions. The death receptor pathway is then activated by Fas/​Fas ligand
contact and is mediated by intracellular caspase activity. Cytogenetic insult results in p53 activation, which in turn
increases Bax expression. Bax stimulates leakage of mitochondrial cytochrome C, which eventually results in caspase
activation and ultimately cell death. 
72

72 Textbook of Cancer Epidemiology

1998), suggesting that this may be a common a quiescent state, but not from a senescent state,
mechanism of avoiding apoptosis by some can- upon treatment with appropriate growth stimuli.
cers. Fas receptor signaling can also be blocked Despite an inability to divide, senescent cells can
by reduced expression of downstream caspases. be maintained in culture for long periods of time.
Tumor cells also avoid apoptosis by overex- Senescence can be induced by exceeding rep-
pressing antiapoptotic proteins such as Bcl-​ 2. licative limits as well as through other cellular
Bcl-​2 inhibits cell death by interfering with the stresses, including DNA damage.
activity of mitochondria-​ associated proapop- Cells can evade senescence and attain a state
totic molecules such as Bax. Overexpression of of replicative immortality by undergoing spe-
Bcl-​2 provided the first example of an oncogene cific genetic alterations. As an example, mouse
that indirectly contributed to cell proliferation embryonic fibroblasts containing homozygous
by inhibiting cell death (Vaux et al, 1988). Bcl-​2 deletions for some tumor suppressor genes,
overexpression is detected in a variety of tumor including p53, divide continuously in culture
types and is associated with a poor prognosis (Jacks, 1996). Rodent cells can also spontane-
(Kaufmann & Gores, 2000; Scarfo & Ghia, 2013; ously avoid senescence at a very low frequency
Strand & Galle, 1998). when grown in culture, presumably through
In addition to evading killer T lymphocytes, sporadically acquired genetic mutations.
suppression of death receptor pathways in cer- Interestingly, normal human cells grown in cul-
tain circumstances may be important in tumor ture rarely give rise spontaneously to immor-
initiation and survival. Loss of essential growth talized derivatives. The underlying mechanism
factors, alterations in cell–​ECM interactions, and for this species-​ specific difference is unclear.
other suboptimal growth conditions can induce However, human cells treated with mutagens or
Fas-​mediated cell death (Le-​ Niculescu et  al, infected by certain viruses can attain an immor-
1999; Rytomaa et al, 1999). Therefore, evasion of talized state; although in certain circumstances,
the apoptotic process may allow tumors to sur- activation of oncogenic signaling in normal cells
vive and grow in environments that are restric- may also lead directly to senescence (Mooi &
tive to normal cells (Kaufmann & Gores, 2000). Peeper, 2006). The evasion of cellular senescence
Such suboptimal growth conditions are common and entry into an immortalized state is one of
in tumor masses. the distinguishing features of cancer.
Many anticancer therapeutic agents act by The mechanism(s) that forces cells to become
inducing cytogenetic insults that subsequently senescent is often referred to as a molecular
lead to apoptotic death. Therefore, defects in clock, which keeps track of cell divisions; at a
apoptotic mechanisms circumvent the efficacy of critical point, an alarm signals and prevents fur-
many anticancer treatments. As a result, signifi- ther proliferation (Sedivy, 1998). The molecular
cant efforts have attempted to identify new meth- clock is thought to exist in the form of telomeres,
ods that counteract antiapoptotic mechanisms. simple repeating DNA sequences located at chro-
These methods may help to overcome drug mosome ends that are associated with protein
resistance in some tumors, thereby increasing the complexes (Kong et  al, 2013). Telomeres effec-
efficacy of some standard anticancer therapies. tively form a cap that prevents chromosomal
erosion, fragmentation, and aberrant rearrange-
Unlimited Replicative Potential ment. The DNA-​ protein complexes of telo-
of Cancer meres oscillate stochastically between two states,
Normal somatic mammalian cells have a lim- capped and uncapped (Fig. 3-​ 8) (Blackburn,
ited replicative potential, where, after continu- 2000). Uncapping of telomeres occurs normally
ous rounds of division, they lose their capacity in dividing cells. With each successive round of
to proliferate. This phenomenon is demonstrated cellular division, telomeres shorten due to the
in continuous in vitro cultures, where cells cease nature of the lagging strand during DNA replica-
to replicate after a certain number of divisions tion of linear chromosomes (Watson, 1972).
(Hayflick & Moorhead, 1961). Viable cells no As telomere length decreases, the stochastic
longer able to divide are said to be senescent. oscillation of protein capping shifts toward the
Senescence is an irreversible process, which dis- uncapped state. At a certain point, telomeres are
tinguishes it from quiescence, the normal with- uncapped for periods long enough to activate
drawal from the cell cycle observed in many cell-​cycle checkpoint mechanisms that initiate
adult cells. Cells can reenter the cell cycle from cell-​cycle arrest, thereby inducing senescence
73

The Origins of Cancer 73

Cell division
Cap
(Telomere shortening)
DNA
Telomerase
(Telomere elongation)

Prevention of chromosomal Activation of cell

erosion, fragmentation, cycle checkpoints
and aberrent rearrangement

FIGURE 3-​8  Telomeres regulate cell survival. Telomeres oscillate stochastically between capped and uncapped states.
Telomere shortening favors the uncapped state, and telomere elongation favors the capped state. Capped telomeres
prevent chromosomal erosion, fragmentation, and aberrant rearrangement. With each successive round of cellular
division, telomeres shorten, shifting the balance toward the uncapped state. If telomeres are left in the uncapped
state for long periods, cell-​cycle checkpoints are activated, resulting in cell-​cycle arrest and subsequent senescence.
Telomerases extend telomeres, thereby shifting the balance toward the capped state and conferring on cells unlimited
replicative potential. 

(Blackburn, 2000). This limitation on the rep- remodeling during normal development, tissue
licative potential of cells can be overcome by a maintenance, and injury repair. In tumors, how-
specialized reverse transcriptase known as telom- ever, the activity of these extracellular factors
erase. Telomerase extends the repeating telomeric is almost always irregular and facilitates tumor
sequence lost during each round of replication, initiation, survival, growth, invasiveness, and
thus preventing chromosomal shortening, stabi- metastasis.
lizing telomere protein caps, and conferring on
cells unlimited replicative potential. The Extracellular Matrix
In humans, telomerase is not expressed in The ECM constitutes a noncellular organized
normal somatic cells but is expressed in germ meshwork of proteins and polysaccharides
cells and stem cells. Telomerases are also widely secreted locally into the extracellular space.
expressed in human cancers. For cancerous cells These molecules communicate with membrane-​
to avoid senescence and attain unlimited repli- spanning glycoproteins and glycolipids, and
cative potential, the progressive loss of telomericparticipate in normal cell and tissue dynam-
sequence must be prevented, thus explaining ics including cell growth, shape, motility, and
the activation of telomerases in many tumors. survival. Integrins and growth factor receptors
Considering the fact that telomerases are acti- are two groups of membrane-​spanning glyco-
vated in many tumors and not in normal somatic proteins that direct extracellular/​ intracellular
cells, they provide an ideal target for therapeuticcommunication. The membrane-​ spanning and
intervention against cancer. ECM molecules expressed by tumors often dif-
fer quantitatively and qualitatively from the cell
EXTRINSIC FACTORS or tissue from which they are derived. These dif-
INFLUENCE CANCER ferences can significantly affect tumor behavior.
PROGRESSION For example, the integrin α2β1, which decreases
Although much cancer research focuses on cel- metastasis risk, is expressed at lower levels in
lular intrinsic mechanisms of tumorigenesis, the advanced breast cancer (Ramirez et  al, 2011).
extracellular environment plays a critical role in Moreover, some breast tumor epithelial cells
cancer initiation and progression. Bidirectional express aberrant integrins that interact inap-
communication between extracellular and intra- propriately with the ECM (Boudreau & Bissell,
cellular systems is requisite for tumorigene- 1998; Glukhova & Streuli, 2013). Restoration of
sis. Extracellular factors that influence tumor the appropriate integrins returns the malignant
growth include ECM (extracellular matrix) epithelial cells to a nonproliferating and differen-
structural proteins, proteinases, growth factor/​ tiated state, regardless of the preexisting genetic
signaling molecules, and tumor-​infiltrating host lesions. In addition, differential expression of
cells. These factors function in tissue growth and multiple extracellular molecules can create an
74

74 Textbook of Cancer Epidemiology

environment favorable to the survival and expan- non-​small-​cell lung cancer are both associated
sion of cancer cells (Bonnans et al, 2014). These with a more aggressive tumor phenotype and a
findings highlight the critical role that extra- worsened prognosis for patients (Luo et al, 2014;
cellular/​intracellular communication plays in Yang et al, 2014). Altered cell-​extracellular inter-
tumorigenesis. actions lead to activation of intracellular signal-
In addition to expressing different ECM and ing cascades that affect cell shape and motility by
membrane-​spanning proteins, tumors remodel causing cytoskeletal changes necessary for tumor
preexisting ECM proteins using extracellular pro- invasion and metastasis (Li et  al, 2014; Mack
teinases. Several classes of proteolytic enzymes et al, 2011).
degrade the ECM, including a class of metal-​
dependent endopeptidases known as matrix Angiogenesis
metalloproteinases (MMPs). Within tumors, both Tumor expansion relies on adequate blood
neoplastic tumor cells and infiltrating-​host cells supplies for delivery of oxygen and nutrients.
express ECM-​degrading proteinases (Coussens Therefore, proliferating tumor cells (and normal
et  al, 2000). These proteinases normally partic- cells) must be located within 100 to 200 µm of
ipate in tissue growth, tissue maintenance, and blood vessels, a distance equal to the intratis-
injury repair. In tumors, however, MMPs are sue diffusion limit of oxygen (Carmeliet & Jain,
important regulators of ECM remodeling asso- 2000). Tumor cells become necrotic or apop-
ciated with growth, invasion, metastasis, and totic when not in close proximity to blood sup-
formation of new blood vessels. For example, plies. Tumors access blood supplies by forming
MMPs and other extracellular proteinases partic- extensive vascular networks through the proc-
ipate in the regulation of polypeptide growth fac- ess of angiogenesis—​the sprouting and expan-
tor activity (Bergers & Coussens, 2000; Lu et al, sion of blood vessels from preexisting vessels.
2012). Some growth factors are directed to the Angiogenesis differs from vasculogenesis—​ the
plasma membrane as latent precursors, and upon assembly of blood vessels from endothelial
proteolytic cleavage, these factors are activated. precursors—​which is how blood vessels originate
Other growth factors are bound to ECM proteins in developing embryos for example. In addition
that suppress their activity. Upon proteinase-​ to enabling tumor expansion, angiogenesis may
mediated degradation of those ECM proteins, be an important step in tumor initiation, since
the growth factors are released and subsequently this process can be visualized in transformed
activated. tissues prior to the appearance of solid tumors
A positive correlation exists between increased (Hanahan & Folkman, 1996). The sprouting
MMP expression and increased tumor invasive- of vascular networks in tumors also provides
ness and metastasis in many human cancers. For migrating tumor cells easy access to the circu-
example, in breast cancer, increased expression of latory system. Thus, angiogenesis also facilitates
the specific MMP isoforms (MMP-​2 and MMP-​9) tumor metastasis.
in serum are associated with a worsened prognosis Many endogenous pro-​and antiangiogenic
(Ren et  al, 2015). MMP expression is induced in molecules exist that are derived from cancer cells,
response to cytokines, growth factors, and altered endothelial cells, stromal cells, blood, and the
cell–​cell or cell–​ECM contacts, and is negatively ECM (Carmeliet & Jain, 2000; Wang et al, 2015).
regulated by a group of nonspecific inhibitors, Proangiogenic factors include vascular endo-
including tissue inhibitors of metalloproteinases thelial growth factor (VEGF), fibroblast growth
(TIMPs) (Bourboulia & Stetler-​Stevenson, 2010). factor (FGF) family, platelet derived growth fac-
The balance between MMP and TIMP activity is tor (PDGF) among many others. Proangiogenic
crucial in cancer progression, as tumor invasion growth factors stimulate endothelial cell prolif-
and metastasis are inhibited by specific TIMPs. eration and migration by binding to transmem-
Tumor invasion and metastasis also require brane tyrosine kinase receptors, initiating signal
changes in cell surface adhesion molecules. cascades, and activating the cell proliferative
Structural or expression changes of many cell and mobility machinery. Antiangiogenic fac-
surface molecules are associated with decreased tors include angiostatin and endostatin among
cell–​cell contact and cellular/​tissue differentia- others. The relative contribution of each of
tion and with increased tumor invasiveness and these factors to the vascular density in tumors
metastasis. For example, decreased expression of likely depends on tumor type, location, size, and
E-​cadherin or increased expression of CD44 in stage of progression (Carmeliet & Jain, 2000).
75

The Origins of Cancer 75

Anti–angiogenic Pro–angiogenic
molecules molecules
VEGF
Angiostation Angiopoietin
Endostatin FGFs

Tumor expansion Tumor expansion

inhibition promotion

The balance of pro-​and antiangiogenic molecules controls angiogenesis. Antiangiogenic molecules,

FIGURE  3-​9 
including angiostatin and endostatin, impede vascular formation and subsequently inhibit tumor expansion.
Proangiogenic molecules, including vascular endothelial growth factor (VEGF), angiopoietin, and fibroblast growth
factors (FGFs), facilitate vascular formation and subsequently promote tumor expansion. 

Variations in the microenvironment of tumors, major histocompatibility complex (MHC)

such as the presence or absence of hypoxia (dep- class  I  molecules that present an array of cel-
rivation of adequate oxygen), may also influence lular peptides on the cell surface. All cells in
vascular density (Carmeliet & Jain, 2011; Wang the body express MHC class  I  complexes that
et al, 2015). present antigens to killer T lymphocytes during
As angiogenesis is necessary for tumor immunosurveillance (Fig. 3-​10). Cells that pres-
growth and metastasis, inhibiting angiogenesis ent foreign antigens to killer T lymphocytes are
may be an effective approach to cancer therapy destroyed by multiple methods including cyto-
(Wang et  al, 2015). The balance of proangio- lytic perforin-​granzyme mechanisms or by Fas
genic and antiantigenic factors controls angio- receptor–​induced apoptotic death. The recogni-
genesis (Fig.  3-​9) (Hanahan & Folkman, 1996). tion and subsequent removal of foreign antigen
Tumors shift the balance in favor of the proan- by the immune system is known as immuno-
giogenic factors, thus allowing for the formation surveillance. In certain circumstances, immu-
of extensive vascular networks and guaranteeing nosurveillance may eradicate tumors after early
their survival. On the other hand, it is possible transforming events, thereby preventing devel-
through pharmaceutical intervention to shift the opment of full-​blown cancer.
balance in favor of the antiangiogenic factors, Tumors elicit immune responses that result
thus depleting the tumor vasculature, resulting in the infiltration of multiple inflammatory cells,
in decreased tumor growth and metastasis. One including macrophages, lymphocytes, and poly-
clinically successful approach for the treatment morphonuclear granulocytes. However, despite
of cancer has been to interfere with the proan- the considerable immune infiltrates and pres-
giogenic growth factor VEGF with the mono- entation of foreign antigens, tumors are able to
clonal antibody bevacizumab. Bevacizumab has evade immune recognition and immunologi-
been approved for the treatment of multiple solid cal clearance by a diverse array of mechanisms
tumor indications including metastatic colorec- (Vinay et  al, 2015). In healthy individuals, the
tal cancer, metastatic non-​small-​cell lung cancer, immune response is negatively regulated by a
metastatic breast cancer, and glioblastoma. variety of pathways collectively referred to as
immune checkpoints. These checkpoints are
Evasion of Immune-​Mediated crucial for tolerance against self antigens and
Eradication in managing the extent and magnitude of an
The genetic alterations that result in tumor immune reaction to damaged cells or foreign
transformation have the potential to be accom- antigens in order to minimize damage to healthy
panied by the generation of tumor-​specific anti- cells. Cancer cells can develop multiple means to
gens (also referred to as neoantigens) (Klein & highjack these checkpoints to suppress immune-​
Klein, 2005; Schumacher & Schreiber, 2015). mediated eradication despite the presence of
Theoretically, these antigens can elicit an neoantigens. For example, many tumors have
immune response that ultimately can kill the increased cell surface expression of the immune
tumor. Killer T lymphocytes are able to mon- checkpoint protein PD-​L1 (Lipson et  al, 2015).
itor the contents of cells by interacting with PD-​L1 binds to its cognate receptor PD1, which
76

76 Textbook of Cancer Epidemiology

Fas
ligand
Killer CD8
T lymphocyte

Foreign antigen Perforins–

Granzymes
Cancer cell
MHC class I Fas

Apoptosis Cytolysis

CELL DEATH

FIGURE  3-​10  Killer T lymphocyte–​mediated cell death. Killer T lymphocytes recognize foreign antigen through
CD8—​major histocompatibility complex (MHC) class I interactions. Killer T lymphocytes then destroy target (can-
cer) cells by inducing apoptosis through Fas/​Fas ligand contacts or by inducing cytolysis through perforin-​granzyme
activity. 

is expressed on cytotoxic killer T cells and sup- 1991; Vinay et  al, 2015). Along this same line,
presses their activation by inhibiting intracellu- tumor cells may fail to process or present anti-
lar signaling pathways that would normally lead gen that normally would be recognized by T lym-
to T cell activation (Wherry & Kurachi, 2015) phocytes. Tumors also utilize various methods
(Figure 3-​11). The cytotoxic T-​lymphocyte anti- to avoid T lymphocyte–​induced apoptotic death.
gen 4 (CTLA4) is another molecule expressed on Some tumors may evade immune-​ mediated
exhausted T cells and inhibits activation when eradication by secreting cytokines that inhibit
bound to its cognate receptors CD80 or CD86 on immune responses, for example prostaglandin
antigen presenting cells (Dijkstra et al, 2016). In E2, transforming growth factor β, or interleu-
addition, some tumor cells do not express MHC kin 10 among others (Vinay et al, 2015). Lastly,
class  I  complexes or express them at greatly many tumors comprise high levels of regulatory
reduced levels (Garrido et al, 1993; Ruiter et al, T cells (Tregs), which function in suppressing the

Tumor
APC T Cell cell

CD80/86 CTLA4 PD1 PD-L1

Inactivation ∫
exhaustion

Immune checkpoint inhibitors

FIGURE 3-​11  The PD1 /​ PD-​L1 interaction between T cells and tumor cells as well as the CTLA4 /​ CD80 or CD86
interation between T cells and antigen presenting cells (APCs) are known as T cell checkpoints and lead to intracel-
lular signals that induce T cell inactivation or “exhaustion.” Inhibition of these immune checkpoints with targeted
therapies that disrupt the PD1 /​ PD-​L1 and CTLA4 /​ CD80(86) interactions can overcome these checkpoints leading
to T cell activation and in some cases immune-​mediated killing of tumor cells. 
7

The Origins of Cancer 77

activity of effector lymphocytes including killer to evaluate basic research and therapeutic find-
T cells. ings. The future of cancer research therefore lies
Inhibiting immune checkpoints has emerged in the collaborative efforts of epidemiologists,
as one of the most promising therapeutic strat- biologists, and clinicians. It is hoped that the col-
egies for treating cancer patients. For example, lective efforts of these groups will translate into
antibody therapies directed against multiple improved cancer prevention programs and anti-
immune checkpoints including PD-​L1 expressed cancer therapies.
on tumor cells and PD1 and CTLA4 expressed on
cytotoxic T cells has shown remarkable clinical REFERENCES
benefit in multiple cancer types by significantly Abbas T, Dutta A. p21 in cancer:  intricate net-
prolonging overall survival relative to other works and multiple activities. Nat Rev Cancer
available treatments, and providing a minority 2009;9:400–​14.
of patients with durable long-​term remissions Adams PD, Kaelin WG Jr. Negative control elements
(Postow et  al, 2015). Initial evidence suggests of the cell cycle in human tumors. Curr Opin Cell
that a higher load of mutations in a cancer cell Biol 1998;10;791–​97.
may predict a greater probability of response Ameres SL, Zamore PD. Diversifying microRNA
sequence and function. Nat Rev Mol Cell Biol
to the immune checkpoint inhibitors, perhaps
2013;14:475–​88.
because mutations alter protein conformation
Armitage P, Doll R. A two-​stage theory of carcinogen-
and present the immune system with abnormal
esis in relation to the age distribution of human
proteins (i.e., neoantigens) that the immune sys-
cancer. Br J Cancer 1957;11:161–​69.
tem is more likely to recognize as foreign and Bergers G, Coussens LM. Extrinsic regulators of epi-
thus to target these cells for destruction (Dijkstra thelial tumor progression:  metalloproteinases.
et al, 2016). Curr Opin Genet Dev 2000;10:120–​7.
Bieging KT, Mello SS, Attardi LD. Unravelling mecha-
CONCLUSION nisms of p53-​mediated tumour suppression. Nat
One of the major goals in cancer research is to Rev Cancer 2014;14:359–​70.
elucidate the molecular and cellular events driv- Blackburn EH. Telomere states and cell fates. Nature
ing the progression of cells from a normal to a 2000;408:53–​6.
cancerous state. This transformation relies on Bonnans C, Chou J, Werb Z. Remodelling the extra-
the accumulation of mutations in various proto-​ cellular matrix in development and disease.
oncogenes, tumor suppressor genes, and DNA Nature Rev Mol Cell Biol 2014;15:786–​801.
repair (caretaker) genes. Under normal circum- Boroughs LK, DeBerardinis RJ. Metabolic pathways
stances, these genes participate in regulating promoting cancer cell survival and growth. Nat
cell and tissue behavior. However, progressive Cell Biol 2015;17:351–​59.
accumulation of mutations in these genes causes Bottger A, Bottger V, Sparks A, Liu WL, Howard SF,
cells to transform to a malignant state. These Lane DP. Design of a synthetic Mdm2-​binding
mutations collaborate to overcome the numer- mini protein that activates the p53 response in
ous barriers and obstacles cells erect to prevent vivo. Curr Biol 1997;7:860–​9.
Boudreau N, Bissell MJ. Extracellular matrix signal-
uncontrolled growth, including cell-​cycle con-
ing:  integration of form and function in nor-
trol, absence of appropriate growth stimuli,
mal and malignant cells. Curr Opin Cell Biol
modified metabolic pathways, programmed cell
1998;10:640–​6.
death, limited replicative potential, and immune-​
Bourboulia D, Stetler-​Stevenson WG. Matrix metal-
mediated eradication. In addition, these muta- loproteinases (MMPs) and tissue inhibitors of
tions ultimately result in altered interactions with metalloproteinases (TIMPs):  Positive and nega-
the extracellular environment, facilitating tumor tive regulators in tumor cell adhesion. Sem Cancer
initiation and progression. Biolog 2010;20:161–​8.
An understanding of the molecular and cel- Burrell RA, Swanton C. The evolution of the unstable
lular events associated with the origin of can- cancer genome. Curr Opin Gen Dev 2014;24:61–​7.
cer is fundamental to cancer epidemiology. Cairns J. Mutation selection and the natural history of
Cancer-​related risk factors and genetic predis- cancer. Nature 1975;255:197–​200.
positions identified by epidemiologic methods Cairns RA, Harris IS, Mak TW. Regulation of cancer
manifest themselves as molecular and cellu- cell metabolism. Nat Rev Cancer 2011;11:85–​95.
lar events. Moreover, the complexity of cancer Calin GA, Croce CM. MicroRNA signatures in human
biology necessitates epidemiologic approaches cancers. Nat Rev Cancer 2006;6:857–​66.
78

78 Textbook of Cancer Epidemiology

Carmeliet P, Jain RK. Angiogenesis in cancer and and growth arrest in mouse embryo fibroblasts
other diseases. Nature 2000;407; 249–​57. lacking p21Cip1 and p27Kip1 without activat-
Carmeliet P, Jain RK. Molecular mechanisms and ing cyclin D-​ dependent kinases. J Biol Chem
clinical applications of angiogenesis. Nature 2000;275:27473–​80.
2011;473:298–​307. Hahn WC, Counter CM, Lundberg AS, Beijersbergen
Coussens LM, Tinkle CL, Hanahan D, Werb, Z. MMP-​ RL, Brooks MW, Weinberg RA. Creation of
9 supplied by bone marrow-​derived cells contrib- human tumour cells with defined genetic ele-
utes to skin carcinogenesis. Cell 2000;103:481–​90. ments. Nature 1999;400:464–​8.
Dang CV. MYC on the path to cancer. Cell Hanahan D, Folkman J. Patterns and emerging mech-
2012;149:22–​35. anisms of the angiogenic switch during tumori-
Dang CV, Le A, Gao P. MYC-​induced cancer cell genesis. Cell 1996;86:353–​64.
energy metabolism and therapeutic opportunities. Hanahan D, Weinberg RA. Hallmarks of cancer: the
Clin Cancer Res 2009;5:6479–​83. next generation. Cell 2011;144:646–​74.
Dick FA, Rubin SM. Molecular mechanisms under- Hayflick L, Moorhead PS. The serial cultivation
lying RB protein function. Nat Rev Mol Cell Biol of human diploid cell strains. Exp Cell Res
2013;14:297–​306. 1961;25:585–​621.
Dijkstra KK, Voabil P, Schumacher TN, Voest EE. Hernandez-​Alcoceba R, del Peso L, Lacal JC. The Ras
Genomics-​and transcriptomics-​ based patient family of GTPases in cancer cell invasion. Cell
selection for treatment with immune checkpoint Mol Life Sci 2000;57:65–​76.
inhibitors, JAMA Oncol 2016; 2:1490–​95. Holderfield M, Deuker MM, McCormick F, McMahon
Editorial. The future of cancer genomics. Nat Med M. Targeting RAF kinases for cancer ther-
2015;21:99. apy:  BRAF-​mutated melanoma and beyond. Nat
Ellis NA, Groden J, Ye TZ, Straughen J, Lennon Rev Cancer 2014;14:455–​67.
DJ, Ciocci S et el. The Bloom’s syndrome gene Hollstein M, Sidransky D, Vogelstein B, Harris CH.
product is homologous to RecQ helicases. Cell p53 Mutations in Human Cancers. Science
1995;83:655–​6. 1991;253:49–​53.
Esquela-​ Kerscher A, Slack FJ. Oncomirs—​Holohan C, Van Schaeybroeck S, Longley DB,
microRNAs with a role in cancer. Nat Rev Cancer Johnston PG. Cancer drug resistance: an evolving
2006;6:259–​69. paradigm. Nat Rev Cancer 2013;13:714–​26.
Fialkow PJ. Clonal origin of human tumors. Ann Rev Hudis CA. Trastuzumab: mechanism of action and use
Med 1979;30:135–​43. in clinical practice. N Engl J Med 2007;357:39–​51.
Fidler IJ, Balch CM. The biology of cancer metasta- Hussain SP, Harris CC. Molecular epidemiology and
sis and implications for therapy. Curr Probl Surg carcinogenesis:  endogenous and exogenous car-
1987;2:129–​209. cinogens. Mut Res 2000;462:311–​22.
Fidler IJ, Ellis LM. The implications of angiogenesis Hyman DM, Puzanov I, Subbiah V, Faris, JE, Chau I,
for the biology and therapy of cancer metastasis. Blay JY, et al. Vemurafenib in multiple nonmela-
Cell 1994;9:185–​88. noma cancers with BRAF V600 mutations. N Engl
Fox EJ, Prindle MJ, Loeb, LA. Do mutator muta- J Med 2015;373:726–​36.
tions fuel tumorigenesis? Cancer Met Rev Irigaray P, Belpomme D. Basic properties and molec-
2013;32:353–​61. ular mechanisms of exogenous chemical carcino-
Garrido F, Cabrera T, Concha A, Glew S, Ruiz-​Cabello gens. Carcinogenesis 2010;31:135–​48.
F, Stern PL. Natural history of HLA expression Jacks T. Lessons from the p53 mutant mouse. J Cancer
during tumour development. Immunol Today Res Clin Oncol 1996;122:319–​27.
1993;4:491–​99. Jackson AL, Loeb LA. On the origin of multiple
Glukhova MA, Streuli CH. How integrins control mutations in human cancers. Sem Cancer Biol
breast biology. Curr Opin Cell Biol 2013:2:633–​41. 1998;8:421–​29.
Gray MD, Shen JC, Kamath-​ Loeb AS, Blank, A, Jonas S, Izaurralde E. Towards a molecular under-
Sopher BL, Martin GM, et  al. The Werner syn- standing of microRNA-​mediated gene silencing.
drome protein is a DNA helicase. Nat Genet Nat Rev Genet 2015;16:421–​33.
1997;17:100–​03. Jones PA, Laird PW. Cancer epigenetics comes of age.
Greenblatt MS, Bennett WP, Hollstein M, Harris C. Nat Genet 1999;21:163–​67.
Mutations in the p53 tumor suppressor gene: clues Kaufmann SH, Gores GJ. Apoptosis in cancer: cause
to cancer etiology and molecular pathogenesis. and cure. BioEssays 2000;22:1007–​17.
Can Res 1994;54:4855–​78. Khouri I. Molecular biology of leukemias.
Groth A, Weber JD, Willumsen BM, Sherr CJ, In:  Devita VRJr, Hellman S, and Rosenberg
Roussel MF. Oncogenic Ras induces p19ARF SA (eds.) Cancer:  Principles and Practice of
79

The Origins of Cancer 79

Oncology. Philadelphia, PA:  Lippincott-​ Raven, Mack NA, Whalley HJ, Castillo-​Lluva S, Malliri A.
1997: 2285–​2292. The diverse roles of Rac signaling in tumorigene-
Kinzler KW, Vogelstein B. Cancer-​ susceptibility sis. Cell Cycle 2011;10:1571–​81.
genes. Gatekeepers and caretakers. Nature McIlwain DR, Berger T, Mak TW. Caspase functions
1997;24;386:761–​63. in cell death and disease. Cold Spring Harbor
Klein G, Klein E. Surveillance against tumors—​ is Persp Biol 2013;5:a008656.
it mainly immunological? Immunol Letters Mooi WJ, Peeper DS. Oncogene-​ induced cell
2005;100:29–​33. senescence—​halting on the road to cancer. N Engl
Knudson AG Jr. Mutation and cancer: statistical study J Med 2006;355:1037–​46.
of retinoblastoma. Proc Natl Acad Sci U S A Moore PS, Chang Y. Why do viruses cause cancer?
1971;68:820–​23. Highlights of the first century of human tumour
Kong CM, Lee XW, Wang, X. Telomere shortening in virology. Nat Rev Cancer 2010;10:878–​9.
human diseases. FEBS J 2013;280:3180–​93. Nesbit CE, Tersak JM, Prochownik EV. MYC onco-
Kottemann MC, Smogorzewska A. Fanconi anaemia genes and human neoplastic disease. Oncogene
and the repair of Watson and Crick DNA cross- 1999;18:3004–​16.
links. Nature 2013;493:356–​63. O’Connell J, Bennett MW, Nally K, Houston A,
Kress M, May E, Cassingena R, May P. Simian virus O’Sullivan GC, Shanahan, F. Altered mechanisms
40-​transformed cells express new species of pro- of apoptosis in colon cancer:  Fas resistance and
teins precipitable by anti-​simian virus 40 tumor counterattack in the tumor-​immune conflict. Ann
serum. J Virol 1979;31:472–​83. N Y Acad Sci 2000;910:178–​92;discussion 193–​5.
Kulis M, Esteller M. DNA methylation and cancer. Pattabiraman DR, Weinberg RA. Tackling the cancer
Adv Genet 2010;70:27–​56. stem cells—​what challenges do they pose? Nat
Lane DP, Crawford LV. T antigen is bound to a Rev Drug Disc 2014;13:497–​512.
host protein in SV40-​transformed cells. Nature Pegram M, Slamon D. Biological rationale for HER2/​
1979;278:261–​63. neu (c-​erbB2) as a target for monoclonal antibody
Lavin MF. Ataxia-​telangiectasia: from a rare disorder therapy. Sem Oncol 2000;27:13–​9.
to a paradigm for cell signalling and cancer. Nat Pitti RM, Marsters SA, Lawrence DA, Roy M, Kischkel
Rev Mol Cell Biol 2008;9:759–​69. FC, Dowd P et  al. Genomic amplification of a
Le-​Niculescu H, Bonfoco E, Kasuya Y, Claret FX, decoy receptor for Fas ligand in lung and colon
Green DR, Karin M. Withdrawal of survival fac- cancer. Nature 1998;396:699–​703.
tors results in activation of the JNK pathway in Plummer M, de Martel C, Vignat J, Ferlay J, Bray F,
neuronal cells leading to Fas ligand induction and Franceschi S. Global burden of cancers attribut-
cell death. Mol Cell Biol 1999;9:751–​63. able to infections in 2012:  a synthetic analysis.
Li H, Peyrollier K, Kilic G, Brakebusch C. Rho Lancet Glob Health 2016;4:e609–​16.
GTPases and cancer. BioFactors 2014;40:226–​35. Postow MA, Callahan MK, Wolchok JD. Immune
Linzer DI, Levine AJ. Characterization of a 54K dal- checkpoint blockade in cancer therapy. J Clin
ton cellular SV40 tumor antigen present in SV40-​ Oncol 2015;33:1974–​82.
transformed cells and uninfected embryonal Prives C, Hall PA. The p53 pathway. J Pathol
carcinoma cells. Cell 1979;17:43–​52. 1999;187:112–​26.
Lipson EJ, Forde PM, Hammers HJ, Emens, LA, Taube Ramirez NE, Zhang Z, Madamanchi A, Boyd
JM, Topalian SL. Antagonists of PD-​1 and PD-​L1 KL, O’Rear LD, Nashabi A et  al. The alpha(2)
in cancer treatment. Sem Oncol 2015;42:587–​600. beta(1) integrin is a metastasis suppressor in
Loeb LA. Mutator phenotype may be required for multi- mouse models and human cancer. J Clin Invest
stage carcinogenesis. Cancer Res 1991;51:3075–​79. 2011;121:226–​37.
Lu P, Weaver VM, Werb, Z. The extracellular matrix: a Ren F, Tang R, Zhang X, Madushi WM, Luo D, Dang
dynamic niche in cancer progression. J Cell Biol Y et  al. Overexpression of MMP Family mem-
2012;196:395–​406. bers functions as prognostic biomarker for breast
Luo Z, Wu RR, Lv L, Li P, Zhang LY, Hao QL, Li W. cancer patients:  a systematic review and meta-​
Prognostic value of CD44 expression in non-​small analysis. PLoS One 2015;10:e0135544.
cell lung cancer: a systematic review. Int J Clin Exp Ruiter DJ, Mattijssen V, Broecker EB, Ferrone, S. MHC
Pathol 2014;7:3632–​46. antigens in human melanomas. Sem Cancer Biol
Lupu R, Cardillo M, Cho C, Harris L, Hijazi M, Perez 1991;2:35–​45.
C et al. The significance of heregulin in breast can- Rytomaa M, Martins LM, Downward, J. Involvement
cer tumor progression and drug resistance. Breast of FADD and caspase-​8 signalling in detachment-​
Cancer Res Treat 1996;38:57–​66. induced apoptosis. Curr Biol 1999;9:1043–​46.
80

80 Textbook of Cancer Epidemiology

Savitsky K, Bar-​Shira A, Gilad S, Rotman G, Ziv Y, Vinay DS, Ryan EP, Pawelec G, Talib WH, Stagg J,
Vanagaite L et  al. A single ataxia telangiectasia Elkord E et al. Immune evasion in cancer: mecha-
gene with a product similar to PI-​3 kinase. Science nistic basis and therapeutic strategies. Sem Cancer
1995;68:1749–​53. Biol 2015;35 Suppl:S185–​98.
Scarfo L, Ghia, P. Reprogramming cell death:  BCL2 Vogelstein B, Kinzler KW. The multistep nature of
family inhibition in hematological malignancies. cancer. Trends Genet 1993;9:138–​41.
Immunol Letters 20013;155:36–​9. Wang Z, Dabrosin C, Yin X, Fuster MM, Arreola A,
Schumacher TN, Schreiber RD. Neoantigens in can- Rathmell WK et al. Broad targeting of angiogene-
cer immunotherapy. Science 2015;348:69–​74. sis for cancer prevention and therapy. Sem Cancer
Sedivy JM. Can ends justify the means? telomeres Biol 2015;35: Suppl:S224–​43.
and the mechanisms of replicative senescence and Warburg O. On the origin of cancer cells. Science
immortalization in mammalian cells. Proc Natl 1956;123:309–​14.
Acad Sci U S A 1998;95:9078–​81. Watson JD. Origin of concatemeric T7 DNA. Nature
Shackleton M, Quintana E, Fearon ER, Morrison SJ. 1972;239:197–​201.
Heterogeneity in cancer: cancer stem cells versus Weinberg R. The Biology of Cancer. 2 ed. Hamden,
clonal evolution. Cell 2009;138:822–​29. CT: Garland Science, 2013.
Sikora J, Dworacki G, Zeromski J. Expression of Fas Weinberg RA. How cancer arises. Sci Am
and Fas ligand and apoptosis in tumor-​associated 1996;75:62–​70.
lymphocytes and in tumor cells from malignant Weintraub SJ, Prater CA, Dean DC. Retinoblastoma
pleural effusions. Nat Immun 1998;16:244–​55. protein switches the E2F site from positive to neg-
Singh AK, Arya RK, Maheshwari S, Singh A, Meena ative element. Nature 1992;58:259–​61.
S, Pandey P et al. Tumor heterogeneity and can- Wherry EJ, Kurachi M. Molecular and cellular insights
cer stem cell paradigm: updates in concept, con- into T cell exhaustion. Nat Can Rev Immunol
troversies and clinical relevance. Int J Cancer 2015;15:486–​99.
2015;136:1991–​2000. Yan M, Schwaederle M, Arguello D, Millis SZ, Gatalica
Strand S, Galle PR. Immune evasion by Z, Kurzrock R. HER2 expression status in diverse
tumours:  involvement of the CD95 (APO-​1/​Fas) cancers:  review of results from 37,992 patients.
system and its clinical implications. Mol Med Cancer Me Rev 2015;34:157–​64.
Today 1998;4:63–​8. Yang YL, Chen MW, Xian, L. Prognostic and clini-
Suzanne M, Steller, H. Shaping organisms with apop- copathological significance of downregulated E-​
tosis. Cell Death Diff 2013;20:669–​75. cadherin expression in patients with non-​small
Tannock IF, Hickman JA. Limits to personalized can- cell lung cancer (NSCLC): a meta-​analysis. PLoS
cer medicine. N Engl J Med 2016;375:1289–​94. One 2014;9:e99763.
Varmus H, Weinberg RA. Genes and the Biology Zalcberg JR, Friedlander ML, Minden MD. Molecular
of Cancer. New  York, NY:  Scientific American evidence for the clonal origin of blast cri-
Library, 1993. sis in chronic myeloid leukaemia. Br J Cancer
Vaux DL, Cory S, Adams JM. Bcl-​2 gene promotes 1986;53:459–​64.
haemopoietic cell survival and cooperates Zhou BB, Elledge SJ. The DNA damage response:
with c-​myc to immortalize pre-​B cells. Nature putting checkpoints in perspective. Nature
1988;335:440–​2. 2000;408:433–​39.
81

4
Genetic Epidemiology of Cancer
C H R I S T O P H E R H A I M A N A N D D AV I D J .   H U N T E R

I n this chapter we explore the genetic epidemi-

ology of cancer:  the identification and quan-
tification of inherited genetic factors, and their
CANCER GENES AND
MECHANISMS OF GENE
DISRUPTION
potential interaction with the environment, in Discoveries in the field of molecular genetics
the etiology of cancer in human populations. have provided substantial insight into the genetic
The availability of the complete DNA sequence of events required for the initiation and progres-
our genome, generated by the Human Genome sion of cancer in humans. Genes involved in
Project, was followed by the creation of genome-​ cellular growth, differentiation, apoptosis, and
wide databases of common genetic variations. DNA repair have been directly implicated in
This information, in addition to recent inno- cancer development. Some of these genes were
vations in high-​ throughout genotyping and described in the previous chapter. Alterations
sequencing technology, has provided researchers that disrupt their normal activity have been
with the necessary tools to more comprehensively identified, and include mutation of the DNA
examine inherited genetic variation in relation to sequence, loss of heterozygosity (LOH), gene
cancer risk. These advances have revolutionized amplification, chromosomal translocation, and
how we study inherited susceptibility to cancer silencing of transcription though DNA methyl-
and provide the framework for a more complete ation. At the molecular level, cancer is complex;
understanding of the role our genes play in the cell transformation and tumor growth result
development of human cancer. from the accumulation of multiple genetic and
We also describe the techniques used to iden- epigenetic alterations.
tify genetic elements that contribute to cancer A distinction is made between DNA muta-
susceptibility. To facilitate the discussion, we tions originating in germline cells and those
have included a short glossary of genetic terms occurring in somatic cells. The former are inher-
at the end of the chapter. We describe the older ited sequence variants passed on to progeny and
research methods for identifying the chromoso- found in the DNA of all cells of the body. Cancers
mal localization of high-​risk predisposing genes, resulting from inheritance of germline muta-
such as linkage analysis within pedigrees and tions are considered hereditary in origin, and
allele-​sharing methods, as we feel it is impor- approaches to identify these are the focus of this
tant to understand the foundations of the field. chapter. In contrast, mutations in somatic cells
We also review the epidemiologic study designs that arise after birth during the process of carcin-
that can be helpful in identifying low-​risk alleles ogenesis are limited to the cancerous tissue and
in candidate gene and genome-​wide association are called somatic mutations. Sporadic cancers
studies (GWAS), as well as gene–​environment are generally attributed to somatic mutations.
interactions. These concepts are further explored
in ­chapter 6. We also describe some of the geno- Genetic Variation
typing and sequencing platforms commonly A human gene is composed of coding (exons),
employed for high-​ throughout genome analy- noncoding (introns), and regulatory noncoding
sis, and the concept of Mendelian randomization DNA sequences. Once the human gene sequence
and how it may be useful in the study of biomark- was available, the number of protein-​coding genes
ers and environmental causes of cancer. in the human genome was initially estimated to
82

82 Textbook of Cancer Epidemiology

be approximately 30,000–​40,000 (International frameshift mutations, that is, they shift the three-​
Human Genome Sequencing Consortium, 2001), base reading frame (the codon) in RNA that is
but later estimates are as low as 19,000 (Ezkurdia translated into normal protein. Frameshift muta-
et  al, 2014). An individual’s complete genome, tions either yield an incorrect string of amino
composed of the linear sequence of the four acids or introduce a stop codon, which results in
nucleotide bases (adenine, guanine, cytosine, and a shortened protein.
thymine), is unique (except for identical twins). In coding regions, base pair substitutions
This uniqueness contributes to the observed phe- may appear as either silent mutations that do not
notypic diversity among individuals. Individual alter the encoded amino acid, missense muta-
variation in one or many genes may explain dif- tions that result in an amino acid change, or non-
ferences in susceptibility to human diseases such sense mutations that give rise to a stop codon and
as cancer. the premature truncation of the protein during
translation. Deletions, insertions, and noncoding
Mutations and Polymorphisms single-​base replacements may also influence the
Aberrations in the DNA sequence of genes may regulation of transcription and RNA splicing.
alter the function of the encoded proteins. They Some mutations are lethal to the embryo,
may be present as germline mutations in indi- while others may be compatible with survival but
viduals with a familial predisposition to cancer, are associated with a significant elevation in dis-
may occur de novo in a sperm or egg cell or in the ease risk. Only a fraction of genetic alterations
fertilized egg, or may occur during life as somatic result in a change in phenotype. When mutations
mutations in individual cells. These aberrations are identified within tumors or in the germline,
include nucleotide deletions, insertions, and particularly if they do not truncate the protein,
substitutions (Fig. 4-​ 1). Nucleotide insertions laboratory studies may be necessary to estimate
and deletions of fewer than three bases result in their biological relevance.

FIGURE 4-​1  Examples of mutations in cancer-​predisposition genes.

83

Genetic Epidemiology of Cancer 83

Mutations existing at a population allele fre- aggregation may be due to shared environ-
quency greater than 1% (i.e., in 2% of individuals) mental exposures or lifestyle practices, such as
are called polymorphisms (Khoury et al, 1993) and smoking or reproductive histories among fam-
convey a range of biological phenotypes. Most ily members, or may be due to the inheritance
polymorphisms have no functional consequence, of one or more predisposing gene mutations.
because they occur in a noncoding sequence and Mutations in single genes have been identified
do not influence a regulatory element, they do as the causal factors in some inherited cancers,
not alter the amino acid sequence of the protein, and data suggest that inherited genetic fac-
or the alteration does not affect the function of tors also make a substantial contribution to
the protein. However, genetic polymorphisms are the causation of sporadic cancers (Mucci et al,
known to contribute to the phenotypic differences 2016; Pharoah et al, 2002). The degree of can-
observed both within and between ethnic groups. cer risk associated with inheriting a mutation
As discussed later in this chapter, intense searches, depends on gene penetrance, that is, the prob-
both within candidate genes and genome-​wide, are ability that an individual carrying a mutation
currently underway to identify polymorphic vari- will develop disease, which can be as great as
ants that are associated with complex multifactorial 100% and may be modified by a combination
diseases including cancer. of genetic and environmental factors.
The biggest change that has occurred in Familial clustering of cancer may also occur
our understanding of the genome in the last by chance. The common characteristics of truly
10 years has been the appreciation of the role inherited cancer that differentiate it from the
of the variation in introns within protein-​ sporadic form include a substantially increased
coding DNA regions and variation in inter- risk of cancer (usually at a single or small num-
genic regions, previously labeled as “silent” ber of specific anatomic sites) in relatives from
or “junk” DNA. Most of the genetic variants several generations, a Mendelian pattern of can-
associated with cancer in GWAS are in inter- cer inheritance in family pedigrees, an earlier age
genic regions and introns, not in the exons of cancer onset, and multiple or bilateral cancers
that code for the amino acid sequence of the among affected individuals.
proteins, and account for less than 3% of the Classic genetic epidemiologic methods devel-
genome. In parallel, we have learned over 80% oped to identify the chromosomal location of
of the genome is transcribed, a phenomenon inherited cancer susceptibility genes in fami-
sometimes called “pervasive transcription” lies include segregation analysis—​to determine
(Khurana et  al, 2016). Some of the earliest whether there is an underlying Mendelian pattern
evidence for this phenomenon came through of genetic inheritance, linkage analysis—​to local-
epidemiologic GWAS studies in which most of ize the chromosomal region coinherited with
the genetic variants discovered were in inter- cancer that contains the cancer-​ causing gene,
genic regions; subsequently, analyses of RNA and allele-​sharing methods—​to identify cancer
transcripts showed this definitively. The prob- susceptibility loci. The latter two approaches use
able explanation of this phenomenon is that marker loci in genome-​wide searches for DNA
many of the loci associated with cancer and regions that cosegregate with a specific disease
other diseases are coding for RNAs that reg- phenotype.
ulate expression of protein-​ coding genes or
modulate their function. This complex archi- Segregation Analysis
tecture of the human genome also suggests an Segregation analysis is a statistical technique
explanation for why humans have about the initially carried out to identify and describe the
same number of protein-​coding genes as lower underlying genetic mode of disease inheritance
organisms; the greater complexity of regula- in families (Khoury et al, 1993). The goal of seg-
tion of these genes allows them to be deployed regation analysis is to establish whether a genetic
or modified in more complex ways. model, such as Mendelian dominant or recessive
inheritance, is compatible with the observed
S T R AT E G I E S O F G E N E pattern of disease occurrence in pedigrees.
I D E N T I F I C AT I O N Segregation analysis also provides gene pene-
Clustering of cancer within some families trance and allele frequency parameter estimates
has been well documented for most cancer required for classic linkage analysis methods. In
sites (Borch-​Johnsen et  al, 1994). Familial addition to monogenic single-​gene Mendelian
84

84 Textbook of Cancer Epidemiology

models of inheritance, more complex models Positional Cloning Supplanted

can be evaluated, thanks to advances in statisti- by Regional Sequencing
cal techniques. These include polygenic models Once a chromosomal region has been observed
(the combination of many genes), the contribu- to occur more frequently among cancer-​affected
tion of environmental factors, and the consid- family members, the molecular genetic tech-
eration of variable age at onset among family nique known as positional cloning could then be
members (Khoury et al, 1993). Since the advent applied to isolate and identify the disease gene
of techniques to directly analyze DNA in fam- (Watson et  al, 1992). This strategy did not rely
ily members, segregation analysis is no longer a on the biological function of the genetic defect,
mainstay of genetic epidemiology, as it has been but rather on “chromosome walking,” carried
substantially supplanted by these direct tech- out by using overlapping cloned DNA fragments
niques. Another study design to gain an initial that scanned the DNA in the neighborhood of
estimate of the genetic contribution to a disease the marker locus. The availability of the human
is the twin study—​usually comparing the joint gene sequence database now eliminates the need
occurrence of the disease on monozygotic (iden- for constructing the DNA sequence, and a list of
tical) twins who have identical genomes with the known and predicted genes in a chromosomal
joint occurrence in dizygotic twins who share region can be rapidly downloaded.
only half their genome. The genes identified as potentially disease
causing are then sequenced, seeking the muta-
tion that is almost always carried by family
Linkage Analysis to Identify members with cancer—​and rarely or never in
Chromosomal Location those without it. If the mutation has not led to an
Once a pattern is established through segregation obvious alteration in protein function, such as a
analysis, genetic linkage analysis in cancer-​prone stop codon leading to protein truncation, in vitro
family pedigrees (Ott, 1991)  will localize the analyses might make it possible to characterize
chromosomal region coinherited with cancer that how the genetic mutation disturbs the protein’s
contains the cancer-​causing gene. This is based normal biological function.
on the predicted model of disease transmission,
and the analysis is conducted as a genome-​wide Examples of High Penetrance
search for chromosomal loci that occur more fre- Genes for Cancer Syndromes
quently in family members with cancer than in A number of cancer susceptibility genes have
those without it, by using markers spaced across been discovered by linkage analysis in family
the genome to identify regions that are coinher- pedigrees. Examples of genes and the cancers
ited in cases of a cancer within a family and not in to which they predispose (Table 4-​ 1) include
noncases. Multiple family pedigrees with availa- CDKN2A (p16) on chromosome 9q21 and famil-
ble DNA specimens are normally required to ial melanoma (Cannon-​Albright et al, 1992), the
conduct a genome-​wide scan, in which typically RET oncogene on chromosome 10 and multiple
300 or more markers evenly spaced across the endocrine neoplasia (types 2A and 2B) (Eng et al,
chromosomes are typed for each family mem- 1996), Rb on chromosome 13q and retinoblas-
ber. The measure used to evaluate genetic linkage toma (Friend et al, 1986), BRCA1 on chromosome
between genetic loci across families is the LOD 17q21 and early-​ onset breast-​ovarian cancer
score, which is calculated as a ratio of likelihoods (Hall et al, 1990; Easton et al, 1993), BRCA2 on
on the log10 scale. The probability or likelihood chromosome 13q and early-​onset breast cancer
of observing the inheritance pattern of two loci at (Stratton et al, 1994), APC on chromosome 5q and
a specified level of linkage or genetic recombina- familial adenomatous polyposis (FAP) (Leppert
tion is divided by the likelihood of observing the et  al, 1987), and hMSH2, hMLH1, hPMS1, and
data assuming no linkage between the two loci hPMS2 on chromosomes 2p15-​16, 3p21, 2q21,
(Morton, 1955). A  LOD score greater than 3 is and 7q22, respectively, and hereditary nonpol-
conventionally used to indicate that the observed yposis colorectal cancer (HNPCC) (Leppert et al,
data are more likely to occur under the hypothe- 1987; Peltomaki et al, 1993; Bronner et al, 1994;
sis of linkage between the marker and the causa- Nicolaides et al, 1994; Papadopoulos et al, 1994).
tive gene and thus provides strong support for a The pace of discovery of these high-​penetrance
further search for the causative gene in the neigh- genes has slowed in recent years; intriguingly,
borhood of the marker. despite segregation evidence for families at high
85

Genetic Epidemiology of Cancer 85

TABLE 4-​1   EX AMPLES OF INHERITED risk of prostate cancer no causative gene has been
CANCER-PREDISPOSITION GENES found. Similarly, there are families that appear to
carry a gene mutation for breast cancer but do
Gene Syndrome not link to BRCA1 or BRCA2; BRCA3 has been
hypothesized but not discovered.
Oncogenes
The study of families with early-​ onset
RET Multiple endocrine neoplasia 2, breast cancer serves as a paradigm for the use
familial medullary thyroid cancer of genetic epidemiologic methods to identify
MET Familial papillary renal carcinoma genetic determinants underlying familial inher-
syndrome itance of cancer due to mutations in a single
gene. Familial clusters of breast cancer are well
Tumor suppressors (some of these are DNA established—​ the classic observations of Paul
damage response genes) Broca in his wife’s family being a well-​known
APC Familial adenomatous polyposis example (Broca, 1886). Women with one or
more first-​degree relatives with breast cancer
VHL von Hippel-​Lindau syndrome
have 1.5 to 3.5 times the risk of developing breast
WT1 Wilms tumor syndrome cancer compared with women without a family
Rb Hereditary retinoblastoma history (Pharoah et al, 1997). Segregation analy-
ses in family pedigrees with multiple affected
NF1 Neurofibromatosis 1 members with breast cancer alone or breast and
NF2 Neurofibromatosis 2 ovarian cancers suggested a genetic component
with an autosomal dominant mode of inherit-
p53 Li-​Fraumeni syndrome
ance (Newman et al, 1988).
p16/​CDKN2A Hereditary melanoma syndrome Linkage analysis was conducted as a genome-​
PTCH Nevoid basal cell carcinoma wide search among 23 families with multiple
syndrome individuals affected with early-​onset breast can-
cer to identify genetic loci that may be associated
MEN1 Multiple endocrine syndrome with the disease (Hall et al, 1990). The initial dis-
(MEN1) covery, later confirmed by others (Narod et  al,
BRCA1 Hereditary breast-​ovarian 1991), linked a region on chromosome 17q21
syndrome with familial inheritance of early-​onset breast
BRCA2 Hereditary breast-​ovarian syn-
and ovarian cancers. Positional cloning was
drome, pancreas and prostate
used to localize the BRCA1 gene in this region
cancer
(Miki et  al, 1994). BRCA1 is a large gene con-
sisting of 22 coding exons, in which many thou-
PTEN Cowden syndrome sands of separate germline variants have been
PALB2 Breast cancer, pancreas cancer identified (see Breast Cancer Information Core
Database:  https://​research.nhgri.nih.gov/​bic/​).
DNA damage response genes (some are tumor Among women with early-​onset breast/​ovarian
suppressors) cancer, approximately 95% of the variants were
hMSH2, hMLH1 Hereditary nonpolyposis colon frameshift, nonsense, or splice site mutations that
cancer (HNPCC) aka Lynch led to truncation of the encoded protein. Linkage
Syndrome to 17q21 was observed in only 45% of high-​risk
breast cancer families, suggesting that other
hPMS1, hPMS2 includes endometrial, other gas- high-​penetrant susceptibility genes may explain
trointestinal and other cancers the remaining portion of breast cancer within
MYH Colorectal cancer these families (Hall et al, 1990).
ATM Ataxia-​telangiectasia
A second gene responsible for early-​ onset
breast cancer, designated BRCA2, was next
XP(A-​G) Xeroderma pigmentosum identified on chromosome 13q12-​13 in families
BLM Bloom syndrome in which linkage to the BRCA1 locus was not
observed (Wooster et  al, 1994). From high-​risk
pedigrees, the cumulative probability of devel-
oping breast cancer for carriers of germline
86

86 Textbook of Cancer Epidemiology

mutations in BRCA1 or BRCA2 was initially esti- penetrance). Misspecification of gene penetrance
mated at over 80% (Ford et al, 1994, 1998). and variable penetrance of different mutations
Gene penetrance estimates from linkage stud- may obscure the results of linkage studies (Ott,
ies are generally overestimates because, by defini- 1991). Penetrance is also age-​dependent, with
tion, they only include families with an unusually older individuals being more likely to develop
large number of affected members. In other disease. For instance, for a dominant gene with
words, families that do not show a striking excess incomplete penetrance, it is difficult to deter-
of breast cancer, but may carry a mutated gene, mine whether a young, high-​risk, nondiseased
are excluded from this ascertainment scheme. individual who carries the linked marker loci has
Results from a population-​ based kin-​ cohort in fact inherited the mutated gene, because dis-
study, a design employed to estimate the pene- ease has not yet occurred and may not occur. An
trance of an autosomal dominant gene, have set age-​dependent penetrance function can be used
the estimate for BRCA1 mutations below 60% in to assign a probability of having a susceptibility
Ashkenazi Jews (Struewing et al, 1997), although locus for those who are unaffected (Ott, 1991),
a later study in New York City suggested a higher but this will obviously misclassify some family
estimate (King et al, 2003). Subsequent estimates members.
suggest a penetrance by age 70 of 55%–​ 65% A large number of family pedigrees may
among BRCA1 carriers and 45% for BRCA2 car- also be required to detect linkage if gene pene-
riers (Mazzola et  al, 2016). BRCA1 and BRCA2 trance is low. An accurate case definition is also
are responsible for about 70%–​80% of early-​onset important for determining the genetic etiology
breast cancer in high-​risk families, however they of a disease. Labeling a family member who does
are estimated to only account for less than 20% not truly have the disease as a case can dramati-
of the excess risk of breast cancer associated with cally reduce the power of the pedigree approach.
having a family history (Anglian Breast Cancer However, defining the case status may be diffi-
Study Group, 2000), suggesting the existence of cult, as mutations in specific genes may result in
other familial breast cancer susceptibility loci variable degrees of the disease phenotype (var-
(Ford & Easton, 1995). iable expressivity). In addition, the inclusion of
Epidemiologic studies have also identified family members with disease not caused by the
geographic differences in the mutation frequen- genetic factors (phenocopies) can decrease the
cies of BRCA1 and BRCA2. These differences power to detect linkage.
are due to “founder” mutations in these genes Genetic heterogeneity, that is, when a single
originating throughout human evolution in sub- disease is caused by mutations in independent
groups of the population. For example, Ashkenazi genes at different loci, may also add complexity
Jews have a greater prevalence of specific muta- to linkage studies. Thus, when families are pooled
tions in these genes than other subgroups. They to study linkage, a mixing of genetic effects
have a combined prevalence of greater than 2% from various chromosomal regions, such as the
due to three common mutations: a two-​base-​pair involvement of BRCA1 and BRCA2 with early-​
deletion (185delAG) and a one-​base-​pair inser- onset breast cancer, may decrease the power
tion (5382insC) in BRCA1 and a one-​base-​pair to detect linkage with a specific genetic region.
deletion in BRCA2 (6174delT) (Struewing et al, A  larger number and closer spacing of highly
1997). Similarly, a five-​ base-​
pair deletion in informative (polymorphic) markers allow a more
BRCA2 (999del5) among Icelanders is relatively widespread evaluation of chromosomal regions
common (0.6%) (Thorlacius et  al, 1996). This and greater efficiency in locating disease loci.
deletion explains a large percentage of high-​risk
familial breast cancer in Iceland.
Alternative Methods
Limitations of Segregation Although the majority of familial cancer genes
and Linkage Analysis were initially identified by linkage analysis and
The sample size required for sufficient power to positional cloning, alternative strategies have also
provide a LOD score greater than 3 in a genome-​ been employed. For instance, the strategy used
wide linkage analysis depends on a number of to distinguish p53 as the gene responsible for
factors. For many mutations in cancer suscepti- Li-​Fraumeni syndrome was based on informa-
bility genes, the probability of a carrier develop- tion about the known biological function of this
ing cancer is much less than 100% (incomplete candidate gene. Li-​Fraumeni syndrome is a rare
87

Genetic Epidemiology of Cancer 87

autosomal dominantly inherited disorder char- linkage analysis studies when several genes with
acterized by a significantly elevated risk of devel- small to modest effects predispose an individ-
oping early-​onset cancers, including soft tissue ual to the disease. Allele-​sharing methods test
sarcomas, osteosarcomas, leukemias, and brain for linkage using affected sib or relative pairs.
and breast tumors (Malkin et al, 1990). Fundamental to these methods is the hypothesis
Somatic mutations in the p53 gene were that the two relatives who have the same type of
observed in many tumors of individuals with cancer share more genetic loci than those who
Li-​Fraumeni syndrome (Hollstein et  al, 1991). are discordant for this cancer. As in classic link-
Previously, p53 had been located on the short age analysis, marker loci are utilized and geno-
arm of chromosome 17 (McBride et al, 1986), and typed in family members. A higher frequency of
experimental data showed that p53 was impor- marker transmission on specific chromosomes
tant in regulating cell growth (Hinds et al, 1989). in diseased relatives suggests that predisposing
Based on these observations, p53 was sequenced genes may exist in the neighborhood of these
in the germline of Li-​Fraumeni syndrome fami- markers.
lies to search for mutations that might predispose Two types of allele-​ sharing measures are
to the cancer syndrome. Mutations were found used: the number of alleles shared that are iden-
that were more common in cases in the families tical by state (IBS: alleles that are the same) and
than in noncases (Malkin et al, 1990). the number of alleles shared that are identical by
An additional gene discovery strategy is descent (IBD:  alleles inherited from a common
based on the knowledge of protein homology ancestor). In general, incorporating alleles shared
with lower organisms and is exemplified in stud- IBD in a study is more powerful than incorpo-
ies of HNPCC (Table 4-​1), the autosomal dom- rating alleles shared IBS (Bishop & Williamson,
inantly inherited disease characterized by an 1990). However, IBD measures require parental
elevated risk of colon and rectal cancer (Aaltonen marker genotypes to be known—​which may not
et al, 1993). Tumors in patients with HNPCC dis- always be accessible, particularly if parents are
play insertions and deletions at repetitive DNA deceased. Under normal Mendelian inheritance
sequences, a condition described as “microsatel- for a biallelic gene, the probability that two sib-
lite instability” (MSI) or “replication error phe- lings share alleles is 25% for sharing none, 50%
notype” (RER) (Aaltonen et al, 1993). Mutations for sharing one, and 25% for sharing two. Any
in yeast mismatch repair genes (PMS1, MLH, deviation from these expected distributions indi-
MSH2) lead to MIN in yeast (Strand et al, 1993). cates potential linkage of a marker locus and a
Linkage studies mapped HNPCC loci to chromo- disease gene.
somes 2p and 3p (Peltomaki et al, 1993; Bronner Issues similar to those encountered in family-​
et  al, 1994). Human mismatch repair homologs pedigree linkage analyses such as case definition,
(hMSH1, hMLH) were discovered on the basis of reduced gene penetrance, phenocopies, genetic
similarity in sequence (homology) between the heterogeneity, and the large number of highly
yeast and human genes, and the human genes variable markers necessary complicate the eval-
were shown to map to chromosomes 2p and 3p. uation of genetic linkage using allele-​ sharing
Germline mutations in these genes were later methods. In addition, the requirement for fami-
identified in the majority of individuals with lies with at least two affected individuals from
HNPCC (Leach et al, 1993; Bronner et al, 1994; which a DNA sample can be obtained is a poten-
Nicolaides et al, 1994; Papadopoulos et al, 1994). tial limitation compared with association studies.

Allele-​Sharing Methods Sequencing to Identify

An obvious Mendelian mode of disease inherit- Rare Alleles
ance will not usually be apparent for a common The ability to sequence panels of candidate genes,
disease with a multigenic etiology or a strong whole exomes, and whole genomes, has led to the
environmental component, such as the majority discovery of almost two dozen additional genes,
of cancers that do not occur in high-​risk families. that when mutated, substantially increase risk of
Alternative study designs for evaluating genetic breast and/​or ovarian cancer (Nielsen et al, 2016).
linkage that do not require the mode of disease Most of these are in genes associated with genome
inheritance to be specified (model-​ free) have maintenance, DNA repair, or cell-​ cycle check-
been developed (Fishman et al, 1978). In general, points, and many of them form protein complexes
these methods are more informative than classic with BRCA1 and BRCA2. Because most of these
8

88 Textbook of Cancer Epidemiology

mutations are even less common than mutations set of three nucleotides may code for the same
in BRCA1 and BRCA2, and have lower penetrance, amino acid) are more frequent than base changes
establishing causality through statistical means is at nonsynonymous sites (Halushka et  al, 1999).
often difficult and requires biological evidence of This further illustrates that genetic variation is
alteration of function. constrained by the influence of each variant on
protein function.
APPROACHES TO IDENTIFY Nucleotide sequence variations occur and
C O M M O N , L O W -​R I S K are maintained in different geographic popula-
DISEASE ALLELES tions over the course of human history. These
Generally, around 5% of most cancers can be variations include single nucleotide polymor-
explained by highly penetrant, rare mutations phisms (SNPs), sequence insertions and dele-
in cancer susceptibility genes transmitted in tions, and highly variable repeating nucleotide
families with an established mode of inherit- segments. SNPs are the most common form
ance. However, for many cancers, a more com- of variation and exist every ~300 bp on aver-
plex underlying etiology has been suggested, due age across the genome. SNPs with frequencies
to highly prevalent polymorphisms in multiple equal to or above 5% are estimated to account
genes potentially in combination with expo- for ~90% of human heterozygosity, meaning that
sures to nongenetic factors such as smoking, most genetic variation is shared between any two
diet, and other lifestyle variables. In the 1990s individuals (International Hapmap Consortium,
the predominant approach was to compare the 2005). A central hypothesis regarding the under-
prevalence of single gene variants in cases and lying nature of genetic variation contributing to
controls, the so-​called candidate gene strategy. complex diseases such as cancer is the common-​
Candidate cancer genes are usually chosen on disease common-​ variant (CD/​ CV) hypothesis
the basis of their function in biological pathways (Lander, 1996; Chakravarti, 1999). The under-
thought to be important in the pathogenesis of lying premise of the CD/​CV hypothesis is that
cancer. Biologically functional polymorphisms alleles with appreciable frequencies in the popu-
in candidate genes may contribute to our under- lation contribute to important variation in risk of
standing of the genetic etiology of cancer, serving disease. These causal variants are most likely to be
as biomarkers of individual cancer susceptibil- old and predate the divergence of human popula-
ity. These low-​penetrance polymorphisms have tions and thus, most will be shared today across
been observed to convey only small excess risks. ethnically diverse populations. An increasing
However, if they are common, then they may number of common variants have been repro-
result in substantial population-​ attributable ducibly associated with risk of cancers through
risks. If, in addition to categorizing individuals GWAS, an unbiased and agnostic view to uncover
on the basis of genotype, we incorporate environ- the genes and biological pathways that contribute
mental and lifestyle exposure histories into epi- (unambiguously) to cancer development.
demiologic studies, we will be better equipped to
define population subgroups that are genetically LINKAGE DISEQUILIBRIUM
susceptible or resistant to disease. This will con- A N D TA G G I N G   S N P S
sequently improve our ability to associate specific There are two basic strategies for the discov-
exposures with specific cancers. ery of variants that alter cancer risk, by using
Although the human species is 99% identical marker SNPs spaced across the genome, or by
at the DNA sequence level, there is a remarkable fully sequencing genes, regions, the exome, or
spectrum of phenotypic diversity in the popula- the whole genome. Initial advances in high-​
tion. Most of this observed variation, including throughput genotyping technology (discussed
susceptibility to disease, is thought to stem from later) made it possible to survey the vast major-
the less than 1% differences in our DNA. Genetic ity of human genetic variation due to common
diversity is greater in noncoding than in cod- alleles. Genotyping large numbers of SNPs in
ing regions, as noncoding variants are less likely different racial and ethnic population samples
to result in alterations in protein function and revealed the genome to be composed of discrete
thus to be selected against. Nucleotide variation regions, or blocks, of linkage disequilibrium
at codons that do not result in a change in the (LD) where nearby SNPs show strong correla-
amino acid (synonymous sites) due to the redun- tion with one another (Gabriel et al, 2002). The
dancy of the genetic code (i.e., more than one size of LD block regions have been shown to be
89

Genetic Epidemiology of Cancer 89

highly variable across the human genome, yet be selected from, or be representative of, the pop-
their locations appear to be moderately con- ulation giving rise to the cases.
served across populations. These observations A critical potential bias is the selection of
led to the development of LD-​based approaches controls from populations with different ethnic
to survey common genetic variation in a candi- backgrounds than the cases. Selection of cases
date gene or region, or genome-​wide, for associ- from an ethnic population with a greater fre-
ation with disease risk (Daly et al, 2001). Due to quency of an allele than the controls will result
the high degree of correlation between SNPs in in a positive association between the allele and
LD block regions, only a subset of SNPs, called disease even if there is no true underlying asso-
tagging SNPs, are needed to comprehensively test ciation in the population. This problem is often
human genetic variation (Johnson et  al, 2001). called “population stratification” by geneticists,
Tagging SNPs serve as proxies for known and and there are several studies in which it has led to
unknown common genetic variation and allow spurious results (Knowler et al, 1988). In conven-
an indirect assessment of the association with a tional epidemiologic parlance, population strat-
disease-​causal allele. The utility of this approach ification can be seen as a form of selection bias,
has prompted efforts to characterize and catalog in that the controls do not represent a random
polymorphisms across the genome, resulting in sample of the population that gave rise to the
rich SNP databases to facilitate the exploration of cases. Population stratification may be defended
genotype-​phenotype associations. against in the design of the study by matching on
ethnicity or restricting the study population to a
EPIDEMIOLOGIC STUDIES defined ethnic group. It can also be controlled for
In contrast to linkage analysis, used to local- in the statistical analysis if information on this
ize disease loci, association studies quantify the variable is available. However, simulation studies
effect of known genetic variants on disease in suggest that large differences in the prevalence
terms of relative risk. These methods have been of a genotype combined with substantial differ-
shown to be more powerful than linkage anal- ences in ethnicity between cases and controls are
ysis in identifying disease genes with modest required before serious bias occurs due to pop-
effects (Risch & Merikangas, 1996). The term ulation stratification (Wacholder et  al, 2000).
“association studies” is often used by geneticists A  variety of methods have been proposed to
to describe conventional epidemiologic stud- control for this source of bias, mostly relying on
ies designed to evaluate associations between stratification or multivariate techniques that con-
genotypes and cancer, as well as their joint effects trol for a series of SNPs that identify the ancestral
with environmental exposures. These include subpopulations that may give rise to differential
traditional case-​control studies with unrelated ancestry between the cases and controls (Price
controls, cohort studies, and alternative designs, et  al, 2006). These methods now use thousands
including family-​based case-​control, kin-​cohort, of SNPs and are routine in the analysis of genetic
and case-​only designs. case-​control studies.
The use of incident rather than prevalent cases
Case-​Control Studies is also preferable unless it can be shown that gen-
The most common study design for assessing otype does not influence the duration of survival.
relationships between genotypes and cancer risk A “survivor bias” may occur if the prevalence of
is the case-​control study using unrelated controls. the gene variant is different in the prevalent cases
Comparisons of allele frequencies are initially compared with the original case group.
made between individuals with cancer (cases) Genotype misclassification can also lead to
and those without it (controls). The measure of erroneous conclusions regarding the relation-
association computed between genotype and dis- ship between genetic polymorphisms and dis-
ease is the familiar odds ratio, which serves as a ease. Often the biological function of the genetic
measure of relative risk. The usual procedures polymorphism being studied is not known. In
for proper design, conduct, and analysis of case-​ these studies, the polymorphic site may serve as a
control studies also apply to case-​control stud- genetic marker of a nearby linked functional var-
ies in which a gene variant is the exposure. They iant, either within or in proximity to the candi-
include the selection of an appropriate control date gene. As a result, gene–​disease associations
group (Wacholder et  al, 1992)  and reduction/​ may be underestimated if a high degree of LD is
elimination of potential biases. Controls should not present between the nonfunctional marker
90

90 Textbook of Cancer Epidemiology

and the true functional variant. On the other among Ashkenazi Jews to estimate the cumula-
hand, a positive result between a polymorphism tive lifetime risk of breast cancer and other can-
in a candidate gene and a specific cancer is far cers associated with mutations in BRCA1 and
from conclusive evidence that the gene is caus- BRCA2 (Struewing et  al, 1997). The kin-​cohort
ally associated with cancer, as LD with variants at design is sensitive to biases, including selection
a considerable genetic distance may account for bias, which may result if probands with a family
the association. history of disease are more likely to participate.
Misclassification of disease status in relatives may
Case-​Control Studies also occur if the results are based solely on the
Nested case-​control studies within well-​defined probands’ recollection.
cohorts are the conceptually preferred method
to evaluate associations between genes and dis- Case-​Only Design
ease if enough cases are available to provide ade- The case-​only design is an efficient and valid
quate statistical power. However, the sample sizes approach designed specifically to evaluate gene–​
needed to provide adequate power to detect the environment interaction when there is independ-
RRs in the range of 1.2 or less typically associated ence between genotype and the environmental
with low-​ penetrance alleles mean that results exposure. This design avoids the potential prob-
from case-​control studies and case series have lems of control selection as well as concern
become the main source of data for large-​scale regarding systematic differences in the meas-
GWAS, sometimes using “controls” from large urement of past exposures in cases and controls
population databases established for the study of in case-​control studies by employing cases only
other diseases or phenotypes. The advent of sta- (Clayton & McKeigue, 2001). A comparison with
tistical control for population stratification (Price regard to exposure prevalence is made between
et al, 2006) enabled studies in which the possibil- cases with and without the susceptibility gen-
ity of confounding by different population struc- otype to provide an interaction parameter that
ture in cases versus controls could be minimized could measure departure from a multiplicative
statistically, rather than by trying to enroll con- genotype–​exposure association. The joint effects
trols who approximated the exact population that of exposure and genotype are divided by the
gave rise to the cases. product of the individual genotype and exposure
effects (Andrieu & Goldstein, 1998). Without
Family-​Based Association Studies controls, however, the prevalence of exposure
Other case-​control study designs using nontra- and the genotype among nondiseased individu-
ditional controls have also been introduced to als is unknown. Thus, the independent effects of
evaluate associations between genotype and dis- genotype and environmental exposure cannot be
ease (Gauderman et  al, 1999). One such hybrid directly assessed in this design.
method is the case-​ parental-​control design
known as the “transmission disequilibrium test” Gene–​Environment Interaction
(TDT) (Spielman et al, 1993). Alleles transmitted and Polygenic Models
to affected offspring are compared with those of of Disease Risk
a fictitious control composed of nontransmitted Establishing gene–​environment interactions will
parental alleles. One obvious limitation is the always be more challenging than establishing the
necessity of measuring the parental genotype, main effects of a genetic variant. To the issues
which may not be readily available. Similar meth- discussed above must be added the problems of
odologic issues, such as LD, must also be consid- measuring the relevant exposures, the fact that
ered in evaluating the results of nontraditional multiple statistical models of gene–​environment
case-​control study designs. interaction are possible, and the problem that
A novel kin-​ cohort design was developed gene–​environment interactions may vary from
to estimate mutation penetrance of autosomal study to study as a function of ethnic and geo-
dominant genes (Wacholder et al, 1998). A pro- graphic differences in both genotype and expo-
band, with or without disease, is genotyped, and sure prevalence, making the assessment of the
the history of disease in first-​degree relatives is reproducibility of interactions problematic
assessed from the participant. This protocol, (Hunter, 2005). Probably the largest single issue
which allows estimation of the cumulative prob- is that of sample size—​typically studies of even
ability of disease in mutation carriers, was used a simple multiplicative interaction between two
91

Genetic Epidemiology of Cancer 91

dichotomous variables need to be four or more it is possible to efficiently examine whether

times the size needed to detect the main effects. there is a departure from multiplicative model of
Relatively few studies have both exposures meas- interaction for the established genetic and envi-
ured and DNA available from sufficient cases ronmental factors. Almost no departures from
and controls to have substantial power to detect multiplicative interaction have been observed
interactions, meaning that much of the literature and/​or replicated in these studies of breast and
to date is composed of underpowered studies prostate cancer, suggesting that most genetic
in which the chances are high than any statis- and lifestyle factors are statistically independent
tically significant interaction is a false positive. and their influence on disease risk can be mod-
Until multiple individual studies of adequate size eled by multiplying together their individual rel-
are available, then pooling data across consor- ative risks (Barrdahl et al, 2014).
tia of studies may be the only way of obtaining
adequate sample sizes (Kraft & Hunter, 2005), GENOTYPING METHODS
although differences in measurement of environ- USED IN EPIDEMIOLOGIC
mental exposures across studies may limit this STUDIES
approach. Many strategies have been used to identify muta-
Most studies prior to the GWAS era took a tions and polymorphisms in cancer susceptibil-
candidate-​ gene approach, concentrating par- ity genes and in GWAS. The polymerase chain
ticularly on variation in genes that metabolize reaction (PCR), by which millions of copies of
known carcinogens. For instance, glutathione-​S-​ a specific area of the genome can be made from
transferase M1 (GSTM1) is an enzyme involved DNA present in a sample as small as a single cell,
in detoxification of activated carcinogens (Strange has been the most important single development
& Fryer, 1999). The GSTM1 null genotype has (Mullis & Faloona, 1987). The methods fall into
been reported to interact with smoking in lung two groups—​those that genotype for a set of spe-
and other smoking-​related cancers. In addition, cific known polymorphisms or mutations, and
gene–​gene and gene–​gene–​environment interac- those that sequence the genome.
tions have been detected between these and other Epidemiologic studies require large samples
carcinogen-​metabolizing genes and environmen- and automated, high-​ throughput genotyping
tal exposures (Olshan et  al, 2000). However, in techniques to detect the modest associations
a large meta-​analysis of 43 studies with >18,000 between common alleles and cancer. The abun-
subjects, the relative risk for the GSTM1 deletion dance of SNPs discovered in the human genome
was reported to be 1.17 (95% CI 1.07–​1.27), and and the ease in which they can be detected, has
in a subset of 21 studies with more than 9,500 resulted in the development of a number of high-​
subjects for whom the authors had original pri- throughout SNP genotyping platforms. In stud-
mary data the relative risk for the main effect ies that require the analysis of a limited number
of genotype was nonsignificant, and no inter- of SNPs, such as missense or tagging SNPs in a
action with smoking was observed (Benhamou handful of genes, but in a large numbers of sub-
et  al, 2002). Studies such as these demonstrate jects, the TaqMan method has been demonstrated
the sample sizes necessary to document the to be most cost-​efficient. TaqMan is a single-​plex
main effects of genetic low-​penetrance alleles, fluorogenic 5’ nuclease assay that allows for the
and illustrate the dangers of relying on a small analysis of one SNP at a time. The Sequenom
number of studies to document even biologi- MassARRAY is a midlevel multiplexing platform
cally plausible gene–​ environment interactions. that allows for the processing of up to 40 SNPs
Simonds et  al, (Simonds et  al, 2016)  provide a simultaneously. High-​ volume SNP genotyping
useful review of the candidate-​gene approach to platforms such as the Illumina BeadArray and
gene–​environment interaction studies for cancer. the Affymetrix GeneChip array are capable of
Large-​scale collaborations, such as the NCI analyzing up to 4  million SNPs simultaneously
Breast and Prostate Cohort Consortium (Hunter and are most cost-​efficient in large studies. SNPs
et al, 2005), have examined variation in genes in on these chips can “tag” a suitable amount of the
these and others pathways as well as gene–​gene common genetic variation in the human genome
and gene–​environment interaction in breast and and used in GWAS (discussed later); in addition,
prostate cancer. By comparing genetic variants tens of thousands of specific SNPs can be incor-
established through GWAS with the known life- porated into the design of these chips to follow-​
style risk factors for breast cancer, for instance, up on known associated variants or to saturate a
92

92 Textbook of Cancer Epidemiology

region in order to try and establish the variants are more cumbersome to exchange and ana-
that have the strongest association with disease lyze and for which privacy and confidentiality
and are thus more likely to be causal. concerns make sharing data more burdensome
(Pasaniuc & Price, 2016).
W H O L E -​G E N O M E S C A N S , To date, GWAS have identified more than
C O L L A B O R AT I O N , A N D 700 common variants for cancer (https://​www.
CONSORTIA ebi.ac.uk/​gwas/​), which provide strong support
Because of the vast number of SNPs that can for a polygenic inheritance model. Common risk
now be evaluated efficiently in association stud- alleles are estimated to explain a sizable fraction
ies, concerns about multiple comparisons cannot of familial cancer risk (i.e., the risk associated
be ignored. The more SNPs studied, the greater with a positive family history). For example, for
the chance of observing a false-​positive associa- breast cancer, common risk loci have estimated to
tion based on chance alone. In a study of 500,000 explain 20% of the twofold familial risk of breast
SNPs per sample, 25,000 would be expected to cancer in women of European descendance
be statistically significant at the p  =  0.05 level. (Michailidou et al, 2017), and 33% of the famil-
By convention, “genome-​wide significance” has ial component of prostate cancer (Al Olama et al,
been set at p < 5 × 10-​8, corresponding to one 2014). While these estimates are promising, for
million associations tested at the 0.05 level. most cancers, a large fraction of cancer heritabil-
Demanding this stringent level of significance ity has yet to be explained (Manolio et al, 2009).
partially solves the false-​positive problem, but at This missing heritability is likely due to large
the price of creating false negatives and rejecting numbers of common and rare alleles, the iden-
true associations. The high cost of whole-​genome tification of which will require large consortia
scans per sample meant that initial studies had studies and the incorporation of next-​generation
modest numbers of cases and controls and rep- sequencing for comprehensive and well-​powered
lication of initial associations in multiple other investigations of genetic variation in the human
studies was crucial to rule out false-​positive find- genome.
ings. With this in mind, consortia of studies have Genome-​ wide association studies have
been formed for some of the most common can- also provided insight into shared biology path-
cers, for example the National Cancer Institute ways for a number of cancers (Fehringer et  al,
Breast & Prostate Cancer Cohort Consortium 2016). Pleiotropy, defined as a single variant in
(http://​epi.grants.cancer.gov/​BPC3), the Breast the same region of the genome being associ-
Cancer Association Consortium (http://​bcac. ated with risk of different phenotypes, has been
ccge.medschl.cam.ac.uk/​), and the GAME-​ON revealed through GWAS for a number of can-
studies (https://​epi.grants.cancer.gov/​gameon/​). cer sites. Examples include chromosome 8q24,
The falling cost of SNP chips (in 2016 a cost of where genetic variants have been associated with
approximately $25 for the chips and reagents) risk of prostate, colorectum, breast, bladder, and
means that some consortia are now able to sur- other cancers (Easton et al, 2007; Rothman et al,
vey (but not sequence) the whole genome in 2010; Thomas et  al, 2008). Another example is
all samples. By the end of 2016 more than 180 in and near the telomerase reverse transcriptase
genetic variants had been associated with breast (TERT) gene at 5p15, which encodes for telom-
cancer, and about the same number with pros- erase activity, where genetic variants have been
tate cancer, with over 100 variants associated associated with risk of a number of cancer sites,
with colorectal cancer and many dozens of vari- including glioma, lung, prostate, colorectal, and
ants associated with each of the other common breast cancer (Peters et  al, 2012; Haiman et  al,
cancers. A  catalog of genome-​wide associated 2011; Turnbull et al, 2010). Genome-​wide asso-
variants is curated by the European Molecular ciation studies have also provided insight into
Biology Laboratory at http://​www.ebi.ac.uk/​ genetic heterogeneity for many cancers. As an
gwas/​. example, in breast cancer, risk alleles have been
A methodologic development that has greatly discovered that are specifically associated with
facilitated the conduct of these consortia has estrogen receptor positive and negative disease,
been statistical tools to pool summary results which provides support for the hypothesis that
from individual studies without the need to breast cancer subtypes have different genetic eti-
exchange the primary individual-​level data that ologies (Garcia-​Closas et al, 2013).
93

Genetic Epidemiology of Cancer 93

P O LY G E N I C R I S K   S C O R E S height was associated with the risk of colorectal,

Although the relative risks associated with com- lung, or prostate cancers. They observed that the
mon genetic variants tend to be small—​as small genetic prediction equation for height was asso-
as 1.02 per risk allele, when combined into a ciated with risk of colorectal and lung cancers
polygenic risk score the variants can be used but not prostate cancer, providing evidence that
to generate substantial relative risks between the association observed in observational studies
those who have inherited a higher number of between height and colorectal and lung cancers
risk alleles versus those who have inherited a was not due to taller people having different life-
smaller number. Polygenic risk scores are usu- style risk factors for these cancers.
ally constructed simply by assigning the value
of 1 for a heterozygote carrier of a specific risk ASSIGNMENT OF
allele, and 2 for a homozygote. Across dozens BIOLOGICAL FUNCTION IN
or hundred of risk alleles an approximately nor- RISKS REGIONS
mal distribution of risk alleles is present for The majority of genetic variants associated with
cases, and a normal distribution with a lower cancer risk that have been discovered through
number of median risk alleles for controls. For GWAS are located in noncoding sequence, mak-
breast cancer for instance, based on the 172 risk ing their biological interpretation difficult. Recent
alleles, the relative risk for women in the top 1% studies have highlighted the functional nature of
of risk had a 3.5-​fold higher risk of breast cancer noncoding DNA sequence, suggesting that more
than the population average (Michailidou et al, than 80% of the human genome may have func-
2017). These estimated risks could, for instance, tional activity (Bernstein et  al, 2012; Farnham
be used to make recommendations about age at 2012; Gerstein et al, 2012). The NIH Epigenomics
onset of mammographic screening that are tai- Roadmap and the ENCODE (Encyclopedia of
lored to a woman’s individual risk and may thus DNA Elements) project have developed an ency-
increase the benefit-​risk of screening (Chatterjee clopedia of functional elements in noncoding
et al, 2016). A continuing issue in genetics stud- sequence that are being used to isolate the biolog-
ies is the underrepresentation of diverse ances- ically functional alleles at GWAS-​identified risk
tries; most of the data derive from European region (Schaub et  al, 2012). These efforts have
ancestry or Asian ancestry populations, with defined functional classes of noncoding “open”
little data from people of African ancestry and chromatin that include promoters, enhancers,
Hispanic ethnicity (Popejoy et  al, 2016). Some and insulators bound by particular transcription
cancer-​associated variants are “private” to spe- factors and other proteins, such as modified his-
cific continental ancestries, and although many tones. Highlighting their important relationship
variants are associated with a specific cancer in to disease etiology, one of these chromatin classes
multiple ancestral populations, the strength of (“strong enhancers,” making up about 2% of the
association may vary, suggesting that risk pre- genome) was found to be strongly enriched in
diction algorithms need to be customized to the a set of phenotype-​associated SNPs taken from
major continental ancestry groups. 151 GWAS studies (Ernst et al, 2011). ENCODE
has also reported a significant enrichment for
MENDELIAN functional SNPs among the currently identified
R A N D O M I Z AT I O N GWAS-​associations in noncoding sequence, with
Mendelian randomization utilizes individ- the variants concentrated in regulatory DNA
ual genetic variants, or polygenic risk scores as marked by DNase hypersensitive sites (DHSs);
instrumental variables to assess the causality these SNPs were found to perturb transcription
between phenotypes or behaviors and risk of spe- factor recognition sequences, frequently alter
cific cancer. The technique assumes that genetic allelic chromatin states, and cluster in transcrip-
predictors of the phenotype exist, that they influ- tional regulatory pathways. Intersecting maps of
ence risk of cancer only through the phenotype, regulatory elements defined by histone modifica-
and that they are independent of other potential tions, regions of open chromatin and transcrip-
confounders of the phenotype–​cancer relation- tion factor binding sites, and GWAS risk loci
ship. For instance, utilizing over 400 variants that have already facilitated the localization of func-
have been associated with adult height, Khankari tional variant(s) at cancer risk regions (Meyer
et  al (2016) examined whether a risk score for et al, 2013; Jia et al, 2009).
94

94 Textbook of Cancer Epidemiology

In addition to localizing the functional var- epidemiologic application is that the types and
iant, an additional challenge is deciphering the patterns of mutations in tumors may give clues
actual gene in a risk region that is influenced by to the etiologic agents as has been found for
genetic variation. Genes in the vicinity of a risk smoking-​ related cancers, aflatoxin and some
allele are likely candidates, however gene reg- liver cancers, aristolochic acid and urothelial
ulation can span over many megabases. One cancers, and UV-​induced skin cancers (Hollstein
approach used to prioritize potential gene can- et al, 2017).
didates is expression quantitative trait locus
(eQTL) analysis, where genetic variants are cor- CONCLUSION
related with transcript levels. An eQTL-​based Clarifying the complex etiologic role of environ-
analysis is a powerful method to connect genes mental and genetic factors in carcinogenesis will
and GWAS risk variants and has been applied be challenging but necessary to fully understand
to implicate candidate genes at a number of the causes of cancer and how it can be prevented.
cancer risk loci (Couch et  al, 2016; Thibodeau One concern is the potential risk to study subjects
et al, 2015; Lawrenson et al, 2015). Tissue-​based in epidemiologic studies. Guidelines must con-
resources have been developed to aid eQTL anal- tinue to be established to minimize the potential
ysis, including the Genotype–​Tissue Expression risks to research participants while maximizing
(GTEx) project (GTEx Consortium 2013), how- the benefits to society. Confidentiality regarding
ever, even larger and more racial and ethnically the results of genetic analyses, the potential for
diverse resources will be needed for eQTL-​based health insurance or employment discrimination,
analyses of less common alleles and in non-​ and the feedback, if any, to study participants
European ancestry populations. are other concerns, as is, of course, the need for
very large studies to evaluate gene–​environment
S O M AT I C M U TAT I O N S interaction.
IN CANCER The future of this discipline as an integrated
Unlike other diseases, cancer in an individual part of cancer epidemiology will also rely on
can be said to have two genomes—​the inherited continued development of innovative molecular
germline genome that is the subject of this chap- and statistical techniques as well as novel study
ter, and the tumor genome of the cancer itself. designs. The object of these studies is to identify
Sequencing of the exomes or whole genomes of the genetic markers that will aid in assessing can-
hundreds of tumors of each anatomic type of cer risk and elucidate cancer biology. Such stud-
cancer has revealed that no two tumors are the ies may also potentially provide clues for cancer
same, and that tumors may have thousands of therapy, and facilitate early detection and screen-
“somatic” mutations that accumulate during the ing efforts in genetically susceptible subgroups of
development of the cancer. There is hope that, by the population.
analyzing these mutations in a tumor, so-​called
driver mutations can be ascertained that may G L O S S A RY *
be targeted by available drugs or the subjects of
Allele One of the variant forms of a gene at a
drug development, in order to slow or eliminate
particular locus, or location, on a chromosome.
the growth of cancers and approach so-​called
Different alleles produce variation in inherited
precision oncology. While there are some nota-
characteristics such as hair color or blood type.
ble examples of the success of this approach, for
An individual inherits one allele from each
example, imatinib for chronic myelogenous leu-
parent.
kemia, and vemurafinib for metastatic melano-
mas with a BRAF V600 mutation, the success of Allele-​sharing methods An approach used for
the approach is limited by the heterogeneity of studying complex traits. It examines marker
many tumors, the ability of tumor cells to mutate allele-​sharing between pairs of affected relatives.
to escape control by specific drugs, and the com- If relative pairs (or sib pairs) share marker alleles
plexity of the mechanisms of cancer develop- more often than would be expected by chance,
ment and growth that mean that we do not have this suggests that a susceptibility locus may be
drugs for most mutations (Tannock & Hickman, linked to the marker.
2016). This area is not strictly the subject of can- Apoptosis Programmed cell death, the body’s
cer epidemiology, as opposed to cancer therapy, normal method of disposing of damaged,
and is not described here in depth. A  potential unwanted, or unneeded cells.
95

Genetic Epidemiology of Cancer 95

Autosomal Pertaining to the 22 pairs of chromo- Gene The functional and physical unit of heredity
somes that are not sex chromosomes. passed from parent to offspring. Genes are pieces
Chromosome One of the thread-​like “packages” of DNA, and most genes contain the information
of genes and other DNA in the nucleus of a cell. for making a specific protein.
Different kinds of organisms have different num- Gene amplification An increase in the number
bers of chromosomes. Humans have 23 pairs of of copies of a gene sequence. A tumor cell some-
chromosomes, 46 in all: 44 autosomes and 2 sex times amplifies, or copies, short or long DNA
chromosomes. Each parent contributes one chro- segments, resulting in increased signal from
mosome to each pair, so children get half of their specific genes.
chromosomes from their mothers and half from Genome All the DNA contained in an organism
their fathers. or a cell, which includes both the chromosomes
Codon Three bases in a DNA or RNA sequence within the nucleus and the DNA in mitochondria.
that specify a single amino acid. Genome-​Wide Association Study A method to
Deoxyribonucleic acid (DNA) The chemical scan the genome of hundreds or thousands of
name for the molecule inside the nucleus of a cell cases of a disease and controls in order to identify
that carries the genetic instructions for making genetic markers that are more, or less, common
living organisms. among the cases.
DNA chip technology A DNA chip (also called Genomic imprinting An epigenetic process
gene chip or DNA microarray) is a small, flat resulting in the silencing of an inherited allele
surface on which DNA strands are immobilized of a gene. The functional genomic changes are
in distinct spots; each strand contains a unique not caused by inherited mistakes in the DNA
DNA sequence. DNA chips use the same proc- sequence, but are believed to occur through the
ess of hybridization that is used in conventional methylation of DNA in the regulatory regions of
southern and northern blots; the hybridization these genes.
signals are detected and analyzed. DNA chips Genotype An individual’s collection of genes.
are more efficient than conventional methods Often refers to the two alleles of a specific gene.
and can be used for a wide array of biological
Genotyping The process of determining an indi-
experiments.
vidual’s genotype, or specific allelic composition
DNA repair A process that repairs DNA damage. for a certain gene or set of genes.
Dominant allele The allele that is phenotypically Germline The cell lines from which gametes are
expressed despite the presence of other alleles of derived, that is, eggs and sperm.
the same gene.
Heterozygosity The presence of different alleles
Enzyme A protein that encourages a biochemical at a locus.
reaction, usually speeding it up.
Homozygosity The presence of the same alleles
Epigenetic The term that refers to any factor that at a locus.
can affect the phenotype without a change in the
Intron A segment of a gene that is initially tran-
genotype, usually by activating or deactivating
scribed into RNA but is then removed from the
one or more genes.
primary transcript by splicing together the exon
Exon The region of a gene that contains the code sequences on either side of it. Intronic sequences
for producing the gene’s protein. Each exon codes are not found in mature mRNA. Originally
for a specific portion of the complete protein. In thought to be of little biological consequence, it is
some species (including humans), a gene’s exons now clear that the intronic sequences can influence
are separated by long regions of DNA (called gene expression and alternative protein sequences.
introns).
Linkage Coinheritance of two or more genes or
Frameshift mutation Nucleotide insertions and DNA sequences because their loci are in close
deletions that cause a shift in the three-​base-​ proximity on the same chromosome, such that
reading frame in RNA that is translated into after meiosis they remain associated more often
normal protein. These mutations either yield than the 50% expected for unlinked genes.
an incorrect string of amino acids or introduce
Loss of heterozygozity (LOH) Loss of the wild-​
a stop codon, which will result in a shortened
type allele of a tumor suppressor gene.
protein.
96

96 Textbook of Cancer Epidemiology

Meiosis A special type of cell division occur- Phenotype The observable traits or character-
ring in the germ cells of sexually reproducing istics of an organism, for example hair color,
organisms during which gametes containing the weight, or the presence or absence or a disease.
haploid chromosome number (a single set of Phenotypic traits are not necessarily genetic.
chromosomes) are produced from diploid (con- Polymerase chain reaction (PCR) A technique
taining two sets of chromosomes) cells. for amplifying a short stretch of DNA. The
Mendelian inheritance The manner in which method depends on the use of two flanking oli-
genes and traits are passed from parents to child- gonucleotide DNA primers and repeated cycles
ren. Examples of Mendelian inheritance include of primer extension using DNA polymerase.
autosomal dominant, autosomal recessive, and Polymorphism A gene that exists in more than
sex-​linked genes. one version (allele) and where the rare allele can
Microsatellite instability (MIN) Microsatellites be found in more than 2% of the population.
are segments of DNA (usually located in noncod- Positional cloning A process that, through
ing regions) consisting of a segment of mono-​, di-​, gene-​mapping techniques, is able to locate a
or trinucleotide repeats (e.g., CACACACACACA). gene responsible for a disease when little or no
These microsatellites are highly susceptible to information is known about the biochemical
expansions or contractions when the machinery basis of the disease.
for DNA repair is somehow compromised.
Recessive allele An allele that has no obvious
Missense mutation A single DNA base substi- phenotypic effect in a heterozygote; it produces
tution resulting in a codon specifying a different a phenotypic effect only in the homozygous
amino acid. This may alter or abrogate the pro- condition.
tein function, or have no effect.
Restriction fragment length polymorphism
Mutation A permanent structural alteration in (RFLP) A variation in DNA sequence that alters
DNA. In most cases, such DNA changes either the length of a restriction fragment. These varia-
have no effect or cause harm, but occasionally a tions may be simple point mutations, which cre-
mutation can improve an organism’s chance of ate or destroy a restriction site, or variable-​length
surviving and passing on the beneficial change regions (so-​called VNTRs, variable number of
to its descendants. A  germline mutation occur- tandem repeats). RFLPs provide convenient
ing in an egg or sperm is passed on to offspring, a markers for linkage analysis.
somatic mutation occuring in a cell in the body is
Ribonucleic acid (RNA) A molecule similar to a
not passed on to offspring.
single strand of DNA. In RNA, the letter U, which
Nonsense mutation A single DNA base substitu- stands for uracil, is substituted for T in the genetic
tion resulting in a stop (termination) codon that code. RNA delivers DNA’s genetic message to the
truncates the protein. cytoplasm of a cell, where proteins are made.
Nucleotide One of the structural components, or Silent mutation A mutation that has no apparent
building blocks, of DNA and RNA. A nucleotide effect on phenotype expression.
consists of a base (one of four chemicals: adenine,
Single nucleotide polymorphism (SNP) A
thymine, guanine, and cytosine) plus a molecule
polymorphism involving variation of a single
of sugar and one of phosphoric acid. In RNA ura-
base pair.
cil take the place of thymine.
Somatic cell Any cell of an organism not involved
Oncogene A gene that when mutated contributes
in the germline, that is, not an egg or sperm.
to the development of cancer.
Somatic mutation A mutation occurring in
Pedigree A simplified diagram of a family’s gene-
a somatic cell during life rather than in the
alogy that shows family members’ relationships
germline.
to each other and how a particular trait or disease
has been inherited. Stop codon Any of three codons—​UAA, UAG,
or UGA—​that signal the termination of the syn-
Penetrance A phenomenon that refers to the
thesis of a protein.
observable expression, or lack of it, of the mutant
gene. For a dominant disease, it is defined quan- Transcription The synthesis of a single-​stranded
titatively by determining the proportion of RNA molecule from a double-​ stranded DNA
obligate gene carriers (heterozygotes) with the template in the cell nucleus, catalyzed by RNA
phenotype. polymerase.
97

Genetic Epidemiology of Cancer 97

Translation The process of synthesizing a Broca P. Traite des tumeurs. Vol. 1. Des tumeurs en
polypeptide directed by the sequence of a general. Paris, Asselin, 1866.
specific mRNA. Bronner CE, Baker SM, Morrison PT, Warren G,
Smith LG, Lescoe MK et  al. Mutation in the
Translocation Breakage and removal of a large
DNA mismatch repair gene homologue in
segment of DNA from one chromosome, fol-
hereditary non-​ polyposis colon cancer. Nature
lowed by the segment’s attachment to a different 1994;368:258–​61.
chromosome. Cannon-​Albright LA, Goldgar DE, Meyer LJ, Lewis
Tumor suppressor gene A gene that codes for a CM, Anderson DE, Fountain JW et al. Assignment
protein that keeps cell division in check. When of a locus for familial melanoma, MLM, to chro-
mutated, uncontrolled cell growth may occur. mosome 9p13-​p22. Science 1992;258:1148–​52.
X-​linkage Genes on the X-​chromosome, or traits Chakravarti A. Population genetics: making sense out
determined by such genes, are X-​linked. of sequence. Nat Genet 1999;21:56–​60.
Chatterjee N, Shi J, Garcia-​ Closas M. Developing
*Adapted from the Glossary at the National Human and evaluating polygenic risk prediction models
Genome Research Institute website (https://​www.
for stratified disease prevention. Nat Rev Genet
nhgri.gov/​glossary).
2016;17:392–​406.
Clayton D, McKeigue M. Epidemiological methods
REFERENCES for studying genes and environmental factors in
Aaltonen LA, Peltomaki P, Leach FS, Sistonen P,
complex diseases. Lancet 2001;358;1356–​60.
Pylkkanen L, Mecklin JP et al. Clues to the path-
Couch FJ, Kuchenbaecker KB, Michailidou K,
ogenesis of familial colorectal cancer. Science
Mendoza-​Fandino GA, Nord S, Lilyquist J et  al.
1993;260:812–​6.
Identification of four novel susceptibility loci for
Al Olama AA, Kote-​Jarai Z, Berndt SI, Conti DV,
oestrogen receptor negative breast cancer. Nat
Schumacher F, Han Y et  al. A meta-​analysis of
Commun 2016;7:11375.
87,040 individuals identifies 23 new susceptibility
Daly MJ, Rioux JD, Schaffner SF, Hudson TJ, Lander
loci for prostate cancer. Nat Genet 2014;46:1103–​9.
ES. High-​resolution haplotype structure in the
Andrieu N, Goldstein AM. Epidemiologic and genetic
human genome. Nat Genet 2001;29:229–​32.
approaches in the study of gene–​ environment
Easton DF, Bishop DT, Ford D, Crockford GP, and the
interaction:  an overview of available methods.
Breast Cancer Linkage Consortium. Genetic link-
Epidemiol Rev 1998;20:137–​47.
age analysis in familial breast and ovarian can-
Anglian Breast Cancer Study Group. Prevalence and
cer. Results from 214 families. Am J Hum Genet
penetrance of BRCA1 and BRCA2 mutations in a
1993;52:678–​701.
population-​based series of breast cancer cases. Br
Easton DF, Pooley KA, Dunning AM, Pharoah PD,
J Cancer 2000;83:1301–​8.
Thompson D, Ballinger DG et  al. Genome-​wide
Barrdahl M, Canzian F, Joshi AD, Travis RC, Chang-​
association study identifies novel breast cancer
Claude J, Auer PL et  al. Post-​ GWAS gene-​
susceptibility loci. Nature 2007;447:1087–​93.
environment interplay in breast cancer:  results
Eng C, Clayton D, Schuffenecker I, Lenoir G, Cote
from the Breast and Prostate Cancer Cohort
G, Gagel RF et al. The relationship between spe-
Consortium and a meta-​analysis on 79,000 women.
cific net protooncogene mutations and disease
Hum Mol Genet 2014 Oct 1;23(19):5260–​70.
phenotype in multiple endocrine neoplasia type
Benhamou S, Lee WJ, Alexandrie AK, Boffetta P,
2:  International RET Mutation Consortium.
Bouchardy C, Butkiewicz D et  al. Meta-​and
JAMA 1996;276:1575–​9.
pooled analyses of the effects of glutathione S-​
Ernst J, Kheradpour P, Mikkelsen TS, Shoresh N,
transferase M1 polymorphisms and smoking on
Ward LD, Epstein CB et al. Mapping and analysis
lung cancer risk. Carcinogenesis 2002;23:1343–​50.
of chromatin state dynamics in nine human cell
Bernstein BE, Birne E, Dunham I, Green ED, Gunter
types. Nature 2011;473:43–​9.
C, Snyder M. An integrated encyclopedia of
Ezkurdia I, Juan D, Rodriguez JM, Frankish A,
DNA elements in the human genome. Nature
Diekhans M, Harrow J et  al. Multiple evidence
2012;489:57–​74.
strands suggest that there may be as few as 19,000
Bishop DT, Williamson JA. The power of identity-​
human protein–​coding genes. Hum Mol Genet
by-​state methods for linkage analysis. Am J Hum
2014;23:5866–​78.
Genet 1990;46:254–​65.
Farnham PJ. Thematic minireview series on results
Borch-​Johnsen K, Olsen JH, Sorensen TI. Genes and
from the ENCODE project:  integrative global
family environment in familial clustering of can-
analyses of regulatory regions in the human
cer. Theor Med 1994;15:377–​86.
genome. J Biol Chem 2012;287:30885–​7.
98

98 Textbook of Cancer Epidemiology

Fehringer G, Kraft P, Pharoah PD, Eeles RA, Hinds PW, Finlay CA, Levine AJ. The p53 proto-​
Chatterjee N, Schumacher FR et  al. Cross-​ oncogene can act as a suppressor of transforma-
cancer genome-​ wide analysis of lung, ovary, tion. Cell 1989;57:1083–​93.
breast, prostate, and colorectal cancer reveals Hollstein M, Sidransky D, Vogelstein B, Harris CC.
novel pleiotropic associations. Cancer Res P53 mutations and human cancers. Science
2016;76:5103–​14. 1991;253:49–​53.
Fishman PM, Suarez B, Hodge SE, Reich T. A robust Hollstein M, Alexandrov LB, Wild CP, Ardin M,
method for the detection of linkage in familial dis- Zavadil J. Base changes in tumour DNA have the
ease. Am J Hum Genet 1978;30:308–​21. power to reveal the causes and evolution of cancer.
Ford D, Easton DF. The genetics of breast and ovarian Oncogene 2017;36:158–​67.
cancer. Br J Cancer 1995;72:805–​12. Hunter DJ. Gene-​environment interactions in human
Ford D, Easton DF, Bishop DT, Narod SA, Goldgar diseases. Nat Rev Genet 2005;6:287–​98.
DE et al. Risks of cancer in BRCA1-​mutation car- Hunter DJ, Riboli E, Haiman CA, Albanes D, Altshuler
riers. Breast Cancer Linkage Consortium. Lancet D, Chanock SJ et  al; National Cancer Institute
1994;343:692–​5. Breast and Prostate Cancer Cohort Consortium.
Ford D, Easton DF, Stratton M, Narod S, Goldgar D A candidate gene approach to searching for low-​
et al. Genetic heterogeneity and penetrance anal- penetrance breast and prostate cancer genes. Nat
ysis of BRCA1 and BRCA2 genes in breast cancer Rev Cancer 2005;5:977–​85.
families: the Breast Cancer Linkage Consortium. International Hapmap Consortium. A haplotype map
Am J Hum Genet 1998;62:676–​89. of the human genome. Nature 2005;437:1299–​320.
Friend SH, Bernards R, Rogelj S, Weinberg RA, International Human Genome Sequencing
Rapaport JM, Albert DM et  al. A human DNA Consortium. Initial sequencing and analysis of
segment with properties of the gene that pre- the human genome. Nat Genet 2001;409;860–​921.
disposes to retinoblastoma and osteosarcoma. Jia L, Landan G, Pomerantz M, Jaschek R, Herman
Nature 1986;323:643–​6. P, Reich D et  al. Functional enhancers at the
Gabriel SB, Schaffner SF, Nguyen H, Moore JM, gene-​poor 8q24 cancer-​linked locus. PLoS Genet
Roy J, Blumenstiel B et  al. The structure of hap- 2009;5:e1000597.
lotype blocks in the human genome. Science Johnson GC, Esposito L, Barratt BJ, Smith AN,
2002;296:2225–​9. Heward J, Di Genova G et al. Haplotype tagging
Garcia-​Closas M, Couch FJ, Lindstrom S, Michailidou for the identification of common disease genes.
K, Schmidt MK, Brook MN et  al. Genome-​wide Nat Genet 2001;29:233–​7.
association studies identify four ER negative-​ Khankari NK, Shu XO, Wen W, Kraft P, Lindstroem
specific breast cancer risk loci. Nat Genet S, Peters U et  al. Association between Adult
2013;45:392–​8. Height and Risk of Colorectal, Lung, and Prostate
Gauderman WJ, Witte JS, Thomas DC. Family-​based Cancer: Results from Meta-​analyses of Prospective
association studies. Monogr Natl Cancer Inst Studies and Mendelian Randomization Analyses.
1999;26:31–​7. PLoS Med 2016;13:e1002118.
Gerstein MB, Kundaje A, Hariharan M, Landt SG, Khoury MJ, Beaty TH, Cohen BH. Fundamentals
Yan KK, Cheng C et al. Architecture of the human of Genetic Epidemiology. New  York:  Oxford
regulatory network derived from ENCODE data. University Press, 1993.
Nature 2012;489:91–​100. Khurana E, Fu Y, Chakravarty D, Demichelis F, Rubin
GTEx Consortium. The Genotype-​Tissue Expression MA, Gerstein M. Role of non-​coding sequence
(GTEx) project. Nat Genet 2013;45:580–​5. variants in cancer. Nat Rev Genet 2016;17:93–​108.
Haiman CA, Chen GK, Vachon CM, Canzian F, King, M-​C, Marks JH, Mandell, JB. Breast and ovarian
Dunning A, Millikan RC et  al. A common vari- cancer risks due to inherited mutations in BRCA1
ant at the TERT-​CLPTM1L locus is associated and BRCA2. Science 2003;302:643–​6.
with estrogen receptor-​negative breast cancer. Nat Knowler WC, Williams RC, Pettitt DJ, Steinberg AG.
Genet 2011;43:1210–​4. Gm and type 2 diabetes mellitus: an association in
Hall JM, Lee MK, Newman B, Morrow JE, Anderson American Indians with genetic admixture. Am J
LA, Huey B et  al. Linkage of early onset Hum Genet 1988;43:5205–​26.
breast cancer to chromosome 17q21. Science Kraft P, Hunter D. Integrating epidemiology and
1990;250:1684–​9. genetic association:  the challenge of gene-​
Halushka MK, Fan J-​B, Bentley K, Hsie L, Shen N, environment interaction. Philos Trans R Soc Lond
Weder A et  al. Patterns of single-​nucleotide pol- B Biol Sci 2005;360:1609–​16.
ymorphism in candidate genes for blood pressure Lander ES. The new genomics: global views of biology.
homeostasis. Nat Genet 1999;22:239–​47. Science 1996;274:536–​9.
9

Genetic Epidemiology of Cancer 99

Lawrenson K, Li Q, Kar S, Seo JH, Tyrer J, Spindler TJ dominant transmission in high risk families. Proc
et al. Cis-​eQTL analysis and functional validation Natl Acad Sci USA 1988;85:3044–​8.
of candidate susceptibility genes for high-​grade Nicolaides NC, Papadopoulos N, Liu B, Wei YF,
serous ovarian cancer. Nat Commun 2015;6:8234. Carter KC, et  al. Mutations of two PMS homo-
Leach FS, Nicolaides NC, Papadopoulos N, Liu B, Jen logues in hereditary nonpolyposis colon cancer.
J, Parsons R et al. Mutations of a mutS homologue Nature 1994;371:75–​80.
in hereditary nonpolyposis colorectal cancer. Cell Nielsen FC, van Overeem Hansen T, Sorensen CS.
1993;75:1215–​25. Hereditary breast and ovarian cancer:  new
Leppert M, Dobbs M, Scambler P, O’Connell P, genes in confirmed pathways. Nat Rev Cancer
Nakamura Y, Stauffer D et al. The gene for familial 2016;16:599–​612.
polyposis coli maps to long arm of chromosome 5. Olshan AF, Weissler MC, Watson MA, Bell DA.
Science 1987;238:1411–​3. GSTM1, GSTT1, GSTP1, CYP1A1, and NAT1
Malkin D, Li FP, Strong LC, Fraumeni JF Jr, Nelson polymorphisms, tobacco use, and the risk of head
CE, Kim DH et al. Germline p53 mutations in a and neck cancer. Cancer Epidemiol Biomarkers
familial syndrome of breast cancer, sarcomas, and Prev 2000;9:185–​91.
other neoplasms. Science 1990;250:1233–​8. Ott J. Analysis of Human Genetic Linkage. Rev. ed.
Manolio TA, Collins FS, Cox NJ, Goldstein DB, Baltimore: Johns Hopkins University Press, 1991.
Hindorff LA, Hunter DJ et  al. Finding the miss- Papadopoulos N, Nicolaides NC, Wei YF, Ruben SM,
ing heritability of complex diseases. Nature Carter KC, Rosen CA et al. Mutation of the mutL
2009;461:747–​53. homologues in hereditary colon cancer. Science
Mazzola E, Chipman J, Cheng SC, Parmigiani G. 1994;263:1625–​9.
Recent BRCAPRO upgrades significantly improve Pasaniuc B, Price AL. Dissecting the genetics of com-
calibration. Cancer Epidemiol Biomarkers Prev plex traits using summary association statistics. Nat
2014;23(8):1689–​95. Rev Genet 2017 Feb;18(2):117–​ 27. doi:10.1038/​
McBride OW, Merry D, Givol D. The gene for human nrg.2016.142. Epub 2016 Nov 14. Review.
p53 cellular tumor antigen is located on chromo- Peltomaki P, Aalonen LA, Sistonen P, Pylkkanen L,
some 17 short arm (17q13). Proc Natl Acad Sci Mecklin JR, Jarvinen H et al. Genetic mapping of
USA 1986;83:130–​4. a locus predisposing to human colorectal cancer.
Meyer KB, O’Reilly M, Michailidou K, Carlebur S, Science 1993;260:810–​2.
Edwards SL, French JD et al. Fine-​scale mapping Peters U, Hutter CM, Hsu L, Schumacher FR, Conti
of the FGFR2 breast cancer risk locus:  putative DV, Carlson CS et  al. Meta-​ analysis of new
functional variants differentially bind FOXA1 and genome-​ wide association studies of colorectal
E2F1. Am J Hum Genet 2013;93:1046–​60. cancer risk. Hum Genet 2012;131:217–​34.
Michailidou K, Lindstrom S, Dennis J, Beasley J, Hui Pharoah PD, Day NE, Duffy S, Easton DF, Ponder
S, Kar S et al. Large-​scale genetic association anal- BA. Family history and the risk of breast cancer: a
ysis identifies 65 new breast cancer risk loci and systematic review and meta-​analysis. Int J Cancer
predicts target genes. Nature 2017; Oct 23. doi: 1997;71:800–​9.
0.1038/nature24284. [Epub ahead of print]. Pharoah PD, Antoniou A, Bobrow M, Zimmern RL,
Miki Y, Swenson J, Shattuck-​Eidens D, Futreal PA, Easton DF, Ponder BA. Polygenic susceptibility to
Harshman K, Tavtigian S et al. A strong candidate breast cancer and implications for prevention. Nat
for the breast and ovarian cancer susceptibility Genet 2002;31:33–​6.
gene BRCA1. Science 1994;266:66–​71. Popejoy AB, Fullerton SM. Genomics is failing on
Morton NE. Sequential tests for the detection of link- diversity. Nature 2016;538:161–​4.
age. Am J Hum Genet 1955;7:277–​318. Price AL, Patterson NJ, Plenge RM, Weinblatt ME,
Mucci LA, Hjelmborg J, Harris J, Czene K, Havelick Shadick NA, Reich D. Principal components anal-
D, Scheike T et  al. Familial risk and heritabil- ysis corrects for stratification in genome-​ wide
ity of cancer in the Nordic Twin Cohorts. JAMA association studies. Nat Genet 2006;38:904–​9.
2016;315:68–​76. Risch N, Merikangas K. The future of genetic studies
Mullis KB, Faloona Fa. Section II. In:  Wu R (ed.) of complex diseases. Science 1996;273:1516–​7.
Methods of Enzymology. Vol 155, Part F.  San Rothman N, Garcia-​Closas M, Chatterjee N, Malats
Diego, CA: Academic Press, 1987: 335–​50. N, Wu X, Figueroa JD et  al. A multi-​ stage
Narod SA, Feunteum J, Lynch HT, Watson P, Conway T, genome-​wide association study of bladder cancer
Lynch J et al. Familial breast-​ovarian cancer locus identifies multiple susceptibility loci. Nat Genet
on chromosome 17q21-​q23. Lancet 1991;338:82–​3. 2010;42:978–​84.
Newman B, Austin M, Lee M, King M-​C. Inheritance Schaub MA, Boyle AP, Kundaje A, Batzoglou S, Snyder
of human breast cancer:  evidence for autosomal M. Linking disease associations with regulatory
10

100 Textbook of Cancer Epidemiology

information in the human genome. Genome Res candidate genes for prostate cancer-​risk SNPs uti-
2012; 22:1748–​59. lizing a normal prostate tissue eQTL data set. Nat
Simonds NI, Ghazarian AA, Pimentel CB, Schully SD, Commun 2015;6:8653.
Ellison GL, Gillanders EM, Mechanic LE. Review Thomas G, Jacobs KB, Yeager M, Kraft P, Wacholder S,
of the Gene-​Environment Interaction Literature Orr N et al. Multiple loci identified in a genome-​
in Cancer: What Do We Know? Genet Epidemiol wide association study of prostate cancer. Nat
2016 Jul;40(5):356–​65. doi:10.1002/​gepi.21967. Genet 2008;40:310–​5.
Epub 2016 Apr 7. Thorlacius S, Olafsdottir G, Tryggvadottir L,
Spielman RS, McGinnis RE, Ewens WJ. Transmission Neuhausen S, Jonasson JG, Tavtigian SV et al. A
test for linkage disequilibrium:  the insulin gene single BRCA2 mutation in male and female breast
region and insulin-​dependent diabetes mellitus cancer families from Iceland with varied cancer
(IDDM). Am J Hum Genet 1993;52:506–​16. phenotypes. Nat Genet 1996;13:117–​9.
Strand M, Prolla TA, Liskay RM, Petes TD. Turnbull C, Rapley EA, Seal S, Pernet D, Renwick A,
Destabilization of tracts of simple repetitive DNA Hughes D et al. Variants near DMRT1, TERT and
in yeast by mutations affecting DNA mismatch ATF7IP are associated with testicular germ cell
repair. Nature 1993;365:274–​6. cancer. Nat Genet 2010;42:604–​7.
Strange RC, Fryer AA. Glutathione S-​transferase: influ- Wacholder S, Hartge P, Struewing JP, Pee D,
ence of polymorphism on cancer susceptibil- McAdams M, Brody L et  al. The kin-​ cohort
ity. In:  Vineis P, Malats N, Lang M, d’Errico study for estimating penetrance. Am J Epidemiol
A, Caporaso N, Cuzick J et  al (eds.) Metabolic 1998;148:623–​30.
Polymorphisms and Susceptibility to Cancer. Wacholder S, McLaughlin JK, Silverman DT, Mandel
IARC Sci. Pub. No. 148. Lyon:  International JS. Selection of controls in case-​control studies.
Agency for Research on Cancer, 1999: 251–​70. Am J Epidemiol 1992;135:1019–​28.
Stratton MR, Ford D, Neihausen S, Seal S, Wooster R, Wacholder S, Rothman N, Caporaso N. Population
Friedman LS et al. Familial male breast cancer is stratification in epidemiologic studies of common
not linked to the BRCA1 locus on chromosome genetic variants and cancer: quantification of bias.
17q. Nat Genet 1994;7:103–​7. J Natl Cancer Inst 2000;92:1151–​8.
Struewing JP, Hartge P, Wacholder S, Baker SM, Watson JD, Gilman M, Witkowski J, Zoller M. Mapping
Berlin M, McAdams M et  al. The risk of cancer and cloning human disease genes. In:  Watson
associated with specific mutations of BRCA1 and JD, Gilman M, Witkowski J, Zoller M (eds.)
BRCA2 among Ashkenazi Jews. N Engl J Med Recombinant DNA. 2 ed. New  York:  Scientific
1997;336:1401–​8. American Books, 1992: 511–​37.
Tannock IF, Hickman JA. Limits to personalized can- Wooster R, Neuhausen SL, Mangion J, Quirk Y, Ford
cer medicine. New Engl J Med 2016;375:1289–​94. D, Collins N et  al. Localization of a breast can-
Thibodeau SN, French AJ, McDonnell SK, Cheville cer susceptibility gene, BRCA2, to chromosome
J, Middha S, Tillmans L et  al. Identification of 13q12-​13. Science 1994;265:2088–​90.
10

5
Biomarkers in Cancer Epidemiology
P A O L O B O F F E T TA A N D L O R E L E I   M U C C I

INTRODUCTION risk factors, characterized by complex exposure

M olecular methods are now well integrated

into epidemiological studies of cancer. While
approaches to study the genetic epidemiology of
circumstances and weak effects, has become
increasingly difficult to study with traditional
epidemiological approaches. In parallel, an
cancer are described in the preceding ­chapter  4, increasing desire to understand the mechanisms
this chapter deals with the use of other molecular of carcinogenesis has led to the integration of a
methods and biomarkers that are used to measure range of biological markers and the develop-
exposures, to assess early evidence of malignant ment of models. These models, which need to be
transformation, to measure individual susceptibil- tested in human studies, represent an adequate
ity, and to assess cancer subtypes. Such molecular theoretical framework for molecular epidemio-
epidemiology studies have a number of impor- logical research. Furthermore, developments in
tant strengths, but also are subject to problems of molecular biology, genetics, and pathology, such
design and analysis similar to those encountered as the increasingly high-throughput of automatic
in traditional epidemiologic investigations. analytical equipment, allow the large-​scale appli-
The use of biological approaches to meas- cation of assays into large epidemiology studies
ure variables of interest in epidemiology stud- that would otherwise be very resource intensive.
ies is not new. In the areas of infectious and In the context of epidemiology, a biomarker
cardiovascular diseases, research that would be has been defined as a substance, structure or
included under the rubric of molecular epide- process that (1) can be measured in the human
miology has been conducted for decades. As an body or its products and (2)  may influence the
example, in the 1960s, elevated serum choles- incidence or outcome of disease in human popu-
terol level (a biomarker of biologically effective lations (Workshop Report, 1997). A  distinction
dose) was linked to the risk of ischemic heart has been made between biomarkers of exposure,
disease in the Framingham Heart Study (Truett intermediate events, disease, and susceptibility
et  al, 1967). The field of molecular cancer epi- (Table 5-​1; Fig. 5-​1). This distinction, however,
demiology was established more formally in the is somewhat arbitrary. For example, chromo-
early 1980s, with the applications of biomarkers somal aberrations in a range of cell types have
of exposure to carcinogens to population-​based been used for decades to monitor exposure to
studies (Perera & Weinstein, 1982). Since then, environmental carcinogens (Tucker et al, 1997).
the use of increasingly sophisticated biologically From this point of view, they can be classified
based markers of exposure, intermediate effect, as biomarkers of exposure. However, there is
outcome, and susceptibility has become standard evidence pointing toward a role of chromoso-
practice in cancer epidemiology. The maturity of mal aberrations in prediction of cancer risk,
the field is reflected by the increasing number of irrespective of exposure (Norppa et  al, 2006).
studies within textbooks and journals as well as In this respect, they can be seen as intermedi-
presentations in academic programs and profes- ate biomarkers. Furthermore, it is important to
sional societies (Ambrosone & Harris, 2010; Lin note that any scheme, such as that represented in
et al, 2006; Porta et al, 2002; Ugolini et al, 2007). Fig. 5-​1, reflects the current understanding of a
Several reasons may explain this expan- complex biological phenomenon such as carcin-
sion. The identification of new carcinogens or ogenesis and our ability to measure events that
102

102 Textbook of Cancer Epidemiology

methodological issues encompassing the use bio-

TABLE 5-​1   CLASSES OF BIOMARKERS
markers in cancer epidemiology and review the
USED IN MOLECULAR CANCER
contribution of biomarker-​based research to the
EPIDEMIOLOGY identification of human carcinogens.
Class Examples
METHODOLOGICAL ISSUES
Exposure DNA or protein adduct I N B I O M A R K E R -​B A S E D
Viral DNA EPIDEMIOLOGY STUDIES
Urinary metabolite of carcinogen
In the following section, we outline several of the
key methodological issues in biomarker cancer
Disease Gene mutation epidemiology studies. Some of these are issues to
Chromosomal aberration consider more generally in cancer epidemiology,
Preneoplastic lesion while others are specific to biomarker research.
Susceptibility Activation of metabolic enzyme
Bias
Reduced DNA repair activity
All three of the main types of bias that are rec-
ognized in epidemiology may also operate in
are considered relevant to it. The increase in the biomarker-​based studies. Selection bias arises
understanding of the different events in carcin- from lack of comparability of groups included
ogenesis, and the development of relevant and in the study (e.g., cases and controls). For exam-
valid biomarkers, represents the main challenge ple, exposed cases might be more (or less) likely
to molecular cancer epidemiology. to participate in a study than exposed controls.
Any set of biomarkers used in molecular epi- Information bias involves misclassification of
demiology faces the risk of becoming soon obso- participants with respect to disease or exposure
lete, as shown by the rapid development and use status. In biomarker-​based studies, information
of new classes of assays and new approaches to bias encompasses the issues of inherent validity,
measure established markers. Moreover, the util- reproducibility, and stability of markers. Finally,
ity of biomarkers in molecular epidemiology confounding is a special form of bias, generated
research relies on the ability to accurately meas- by coexposure to causal factors other than those
ure the marker, which can be influenced by: (1) under study (discussed later).
the collection, processing and storage of biolog- Selection bias can be avoided by properly
ical specimen; (2) intra-​and interassay variabil- identifying the study population and by optimiz-
ity; and (3)  within-​person variability over time ing the response rate. Furthermore, it can be con-
(Tworoger & Hankinson, 2006). In the follow- trolled in the analysis by identifying factors that
ing sections of this chapter, we address some are related to selection and by controlling them

PRIMARY PREVENTION AND CLINICAL INTERVENTIONS

EXPOSURE
ASSESSMENT MARKERS OF DOSE EARLY DETECTION & CLINICAL
TO HUMANS PROGNOSTIC MARKERS DISEASE
POPULATION
STATISTICS
Absorption Further
distribution Repair Cell proliferation genetic
metabolism replication clonal expansion change

Mutations in Genomic
Environmental Metabolites DNA adducts, instability
or in body fluid reporter genes, Malignant
protein adducts Aberrant gene
endogenous & excreta oncogenes, tumor
suppressor genes expression,
agents Altered
cell structure

SUSCEPTIBILITY FACTORS
genetic/environmental

FIGURE 5-​1  Schematic representation of application of biomarkers in molecular cancer epidemiology. 

Source: National Resource Council, 1987.
103

Biomarkers in Cancer Epidemiology 103

as confounders. Unfortunately, many molecular including through careful work in piloting bio-
epidemiological studies pay too little attention to markers, to minimize the sources of variation
the definition of source population and selection other than the intergroup variability. Such sources
of participants. This is particularly common in are numerous and can include understanding
studies of genetic factors, since it is considered the effect of the circumstances in which biologi-
that any selection of participants is unlikely to be cal samples are collected, processed, stored, and
associated with the genetic factors under study. analyzed on the biomarker values; the technical
In general, prospective studies are less prone to aspects of the assays; the inclusion of quality con-
selection bias than retrospective studies. trol samples in each assay in order to assess var-
iability; and how cases and controls are arranged
Biomarker Variation within batches. It is important to ensure that, if
Sources of variation in biomarker-​based mea- all sources of variation cannot be controlled (as
surements can arise from several sources. First, is often the case), they should apply equally to the
variation might arise from true intergroup dif- groups being compared. Therefore, if long-​term
ferences such as between cancer cases and con- storage of samples might affect the measurement,
trols. Indeed, this is the very phenomenon that it is important to match the groups under study
molecular epidemiological studies usually aim by duration of sample storage. In such cases, mis-
to address. However, other sources of variation classification is said to be “nondifferential,” that
exist that can generate misclassification or meas- is acting equally on the groups being compared.
urement error of the biomarker. This variation In most instances, nondifferential misclassifica-
can be inter-​or intra-​assay or can be inter-​or tion produces bias toward the null value, that is,
intraindividual. Interassay variability exists when it reduces the strength of an existing association.
different assays are used to measure the same On the other hand, a misclassification that
biologic marker and the assays generate different is “differential” with respect to case/​control (or
results. Interindividual variability might be due exposed/​unexposed) status generates a bias in an
to genetic or environmental factors that influ- unpredictable direction. For example, if there is
ence how well the biomarker measures what one substantial interbatch (or interreader) variability
thinks it does. Intraindividual variability refers in the biomarker measurement, the inclusion of
to components of variation within an individ- samples of cases and controls in different batches
ual such as the circadian rhythm and melatonin would generate differential misclassification,
levels measured across a day. Finally, measure- while a proper mix of case and control samples
ment error might arise from sampling and labo- in each batch would at most result in nondiffer-
ratory variation, that is, assaying replicates from ential misclassification. If more of the biological
the same sample and getting different results. samples from cases were collected in the summer
Table 5-​2 provides some examples of sources of months, and samples from controls in the win-
variation for selected biomarkers used in molec- ter months, and seasonality affects the biomarker,
ular cancer epidemiology. this could lead to differential misclassification if
Proper precautions should be taken during the not considered in the design phase. Careful plan-
design phase of molecular epidemiology studies, ning of biomarker studies as well as detailed

TABLE 5-​2   SOURCES OF VARIATION FOR SELECTED BIOMARKERS

USED IN MOLECULAR CANCER EPIDEMIOLOGY

Biomarker Interindividual Intraindividual Laboratory

Hormones + + –​
Environmental agents + +/​–​ +
DNA adducts in white blood cells + + +
Chromosomal aberrations + –​ +
Genetic polymorphism + –​ -​

+: more important variability

–​: less important variability
+/​–​: variability depends on biomarker
104

104 Textbook of Cancer Epidemiology

testing of biomarkers through pilot studies can (Thorgeirsson, 2008). There was no association
help to address issues of misclassification. between rs1051730 genotype and lung cancer
risk in never smokers, whereas there was a sug-
Confounding gestion of a stronger association among light ver-
The use of biomarkers does not prevent con- sus heavy smokers (Spitz et al, 2008).
founding. For example, an association between Another example of interaction, between
tobacco smoking and cancer of the uterine cer- two biomarkers, is aflatoxin, hepatitis B virus
vix has been observed in many populations, but (HBV), and primary liver cancer. Aflatoxin,
it is likely to be confounded by infection with the a family of toxins produced by types of fungi
human papilloma virus (HPV), a necessary cause that occur on some agricultural crops grown in
of cervical cancer. In many populations, smok- warm and humid climates, is an established risk
ers are more frequently HPV positive than non- factor for primary liver cancer. Investigators
smokers. Hence, HPV would be a confounder examined the possible interaction of aflatoxin
no matter how smoking, infection, and cervical with HBV (Ross et  al, 1992). When compared
cancer are assessed (via questionnaires, medi- to HBV-​negative subjects exposed to aflatoxin,
cal records, biochemical methods, or molecular the relative risk in HBV-​positive subjects also
techniques). Furthermore, to use biomarkers positive for aflatoxin markers was 60, which
might introduce confounding. If, for example, was greater than the product of the relative risks
workers occupationally exposed to polycyclic for the two factors separately (4.8 for HBV and
aromatic hydrocarbons (PAHs) have a higher 1.9 for aflatoxin). Thus, a synergism between
consumption of tobacco (an important source of aflatoxin and HBV in primary liver cancer was
PAHs) than other workers, then the assessment of suggested. The wide confidence interval in the
occupational exposure with a biomarker of PAH group with both exposures (6.4–​560) precludes,
is confounded by tobacco smoking (Fig. 5-​2) however, rejections of the null hypothesis of no
(Bonassi, 1989). This would have not been the interaction according to a multiplicative model
case if occupational exposure and tobacco smok- (4.8 × 1.9 = 9.1). This wide interval also stresses
ing were assessed via traditional (e.g., industrial another methodological concern in molecular
hygiene) methods. Issues of confounding should epidemiological studies, namely, the need for a
be considered during the design phase of a bio- large sample-​size.
marker study as well as in the analysis phase.
Random Error
Interaction or Effect Modification An important problem in biomarker-​based epi-
Biomarkers have been widely applied to studies demiological research can be an insufficient num-
of gene–​ environment and gene–​ gene interac- ber of subjects included in a study, due to either
tions in the pathogenesis of cancer and other logistical or financial constraints. The strengths
chronic diseases. A notable example of this is the of biomarker studies for some exposure–​disease
interaction between tobacco dependence and associations do not supersede the potential for
genetic variants in CHRNA5 with lung cancer unstable and conflicting results across studies
risk. Genome-​wide association studies of lung due to sample size and statistical power. Indeed,
cancer risk identified variants in CHRNA5, in the any biomarker-​based measure introduced in epi-
nicotinic acetylcholine receptor family (Amos, demiology should be compared with alternative
2008). Moreover, these variants also independ- measures, and the possible gain in sensitivity and
ently were associated with nicotine dependency specificity of the biomarker measure should be

Occupational exposure to PAHs Traditional approach

Molecular approach

Biomarker of PAHs exposure Lung cancer

Other sources of PAHs

(tobacco smoking)

FIGURE 5-​2  Example of confounding in biomarker-​based studies. 

105

Biomarkers in Cancer Epidemiology 105

considered in the light of the possible decrease humans but strong evidence that an experimen-
in the number of study subjects. A  problem tal carcinogen operates via the same mechanism
related to the generation of false positive results in humans (IARC, 2010). Mechanistic evidence
is the tendency to selectively report significant may include data on preneoplastic lesions, tumor
results, in particular when they show an effect in pathology, genetic and related effects, structure-​
the expected direction. The net result is a biased activity relationships, metabolism and toxicoki-
reporting of positive over negative or null results. netics, and physicochemical parameters.
It may be argued that an initial report of In the first 115 volumes of the Monographs,
false positive results should be considered no a total of 118 agents were classified as human
major scientific problem, since subsequent stud- carcinogens. These agents include single com-
ies, aimed to replicate the early positive results, pounds and infectious agents (e.g., vinyl chlor-
will eventually establish the truth. However, this ide and hepatitis B virus), mixtures (e.g., wood
approach is inefficient and represents an impor- dust), and exposures in certain occupations, such
tant waste of resources, more so when it comes as coke production. For 92 of these agents (78%),
to expensive molecular epidemiological studies. the classification was driven by positive evi-
A  preferable approach is to critically evaluate dence from “traditional” epidemiologic studies,
and report results on the basis of criteria other which were mainly prospective cohort studies in
than—​or including but not limited to—​statistical which exposures, outcomes, and covariates were
significance (Ioannidis, 2008). Biological plausi- assessed using nonmolecular tools such as envi-
bility, possible sources of bias and confounding, ronmental measures, self-​reports, and medical
and numbers of tested associations are among records. Tobacco smoking is a notable example
such criteria. Statistical approaches have been of this group of agents.
proposed to take into account the possibility For an additional 12 agents (10%), the eval-
that significant results are generated by chance uation of carcinogenicity was driven by the
when many comparisons are made. In addition, results of epidemiologic studies that used molec-
authors should be encouraged to systemati- ular markers to measure the relevant exposures
cally report their results, even those “negative” (and, in some cases, potential confounders).
or “null.” Provisions have been developed in Eleven of these carcinogens are biological agents
the studies leveraging ’omic based approaches, (Table 5-​3), whose infection status was measured
and it has become customary to test hundreds using laboratory methods. The only nonbiologic
of thousands of markers in one analysis, to take agent for which biomarker-​based studies gener-
into consideration multiple comparisons, and ated the evidence leading to the classification in
to provide access to the whole set of results in IARC Group 1 is aflatoxin. In this case, the bio-
dedicated databases. In other areas of molecular marker was a urinary marker of exposure includ-
cancer epidemiology, however, testing specific ing adducts produced by the covalent binding of
hypotheses remains the prevalent approach, and aflatoxin B1 at N7 of guanine. Several prospec-
selective reporting of results remains a potential tive studies have examined aflatoxin biomarkers
concern. in prospectively collected samples, resulting in
strong and consistent associations with risk of
CONTRIBUTION OF primary liver cancer (Ross et al, 1992).
B I O M A R K E R -​B A S E D For the remaining 14 agents classified in
EPIDEMIOLOGY TO THE IARC Group  1 (12%), the evidence of carci-
I D E N T I F I C AT I O N O F nogenicity in humans was less than sufficient
HUMAN CARCINOGENS (Table 5-​4), and, for 11 of them, their classifi-
The classification of an agent as a Group 1 carcin- cation as human carcinogens was supported by
ogen in the International Agency for Research on mechanistic evidence in humans, generated via
Cancer (IARC) Monograph program can be used biomarker-​based studies (for the dioxin-​like com-
as benchmark for the identification of human car- pounds, 3,4,5,3’,4’-​Pentachlorobiphenyl, 2,3,4,7,8-​
cinogens. An agent is classified in IARC Group 1 Pentachlorodibenzofuran, and “dioxin-​like PCBs,”
based on either (1) sufficient evidence of carcino- the classification was based on evidence for sim-
genicity in humans, which indicates that a causal ilarity of carcinogenic action with that of 2,3,7,8-​
association is credible, and chance, bias, and tetrachlorodibenzo-​para-​dioxin, an established
confounding could be ruled out with reasonable carcinogen, provided by mechanistic studies in
confidence; or (2) less than sufficient evidence in animals).
106

106 Textbook of Cancer Epidemiology

TABLE 5-​3   BIOLOGICAL AGENTS CLASSIFIED AS HUMAN CARCINOGENS (GROUP 1)

Biological agent Cancers associated with infection Underlying mechanistic events

Epstein–​Barr virus (EBV) Burkitt lymphoma; Hodgkin Cell proliferation, inhibition of

lymphoma; lymphoma (extranodal apoptosis, genomic instability, cell
NK/​T-​cell, nasal type); nasopharynx; migration
non-​Hodgkin lymphoma (immune
suppression related)
Hepatitis B virus (HBV) Liver (hepatocellular carcinoma) Inflammation, liver cirrhosis,
chronic hepatitis
Hepatitis C virus (HCV) Liver (hepatocellular carcinoma); Inflammation, liver cirrhosis, liver
non-​Hodgkin lymphoma fibrosis
Kaposi’s sarcoma herpes virus Kaposi sarcoma; lymphoma Cell proliferation, inhibition of
(KSHV) ­(primary effusion) apoptosis, genomic instability, cell
migration
Human immunodeficiency Anus; cervix; eye (conjunctiva); Immunosuppression (indirect
virus, type 1 (HIV-​1) Hodgkin lymphoma; Kaposi action)
­sarcoma; non-​Hodgkin lymphoma
Human papilloma virus (HPV) Anus*; cervix; oral cavity*; Immortalization, genomic i­ nstability,
types 16, 18, 31, 33, 35, 39, 45, ­oropharynx*; penis*; tonsil*; inhibition of DNA ­damage response,
51, 52, 56, 58, 59 vagina*; vulva* antiapoptotic activity
Human T-​cell lymphotrophic Leukemia and/​or lymphoma (adult Immortalization and
virus, type-​1 (HTLV-​1) T cell) ­transformation of T cells
Helicobacter pylori Lymphoma (low-​grade B-​cell Inflammation, oxidative stress,
mucosa-​associated lymphoid-​tissue altered cellular turnover and gene
gastric lymphoma); stomach expression, methylation, mutation
(noncardia carcinoma)
Clonorchis sinensis Liver (cholangiocarcinoma) N/​A
Opisthorchis viverrini Liver (cholangiocarcinoma) Inflammation, oxidative stress, cell
proliferation
Schistosoma haematobium Urinary bladder Inflammation, oxidative stress

* Association established only for HPV-​16

N/​A, not available

The data used for the classification of the may interact with the aryl hydrocarbon receptor
agents listed in Table 5-​4 refer to three main (Dietrich & Kaina, 2010).
types of mechanism. The first line of evidence The second type of mechanistic study con-
involves formation of DNA and protein adducts cerns the detection of specific genetic alterations
and interaction with cellular receptors. This in tumors of exposed animals and humans. This
includes studies based on DNA and proteins has been the case of agents causing mutations in
adducts. Adducts have been interpreted as TP53, including ultraviolet radiation and aris-
markers of biologically effective dose of carcino- tolochic acid, and in KRAS, including polycyclic
gens as well as cancer risk. For example, DNA aromatic hydrocarbons such as benzo(a)pyrene
adducts and A:T→T:A transversions in the TP53 (B[a]‌P), and aristolochic acid.
gene similar to those in experimental animals Finally, structural alterations in the genome
exposed to aristolochic acid have been observed have been used to identify human carcinogens.
in patients with severe nephropathy and uro- Some of these alterations, including sister chro-
thelial tumors exposed to this agent (Chen et al, matid exchange (SCE) and micronuclei (MN) for-
2016). Receptor-​ mediated carcinogenesis has mation, are unspecific, whereas others, including
been evoked in the case of dioxin-​like agents that mixed lineage leukemia (MLL) translocations,
107

Biomarkers in Cancer Epidemiology 107

TABLE 5-​4   GROUP 1 AGENTS WITH LESS THAN SUFFICIENT EVIDENCE IN HUMANS

BUT WITH STRONG MECHANISTIC EVIDENCE

Agent Evidence from Evidence from Mechanistic evidence

cancer studies cancer studies in in humans*
in humans animals

Acetaldehyde associated Inadequate Sufficient Very high risk of certain upper

with consumption of ­aerodigestive tract cancers in aldehyde-​
­alcoholic beverages dehydrogenase-​deficient populations in
genetic epidemiology studies
Areca nut Inadequate Sufficient Genotoxicity, including sister chromatid
exchange (SCE) and micronuclei (MN)
formation
Aristolochic acid Limited Sufficient Genotoxicity, including DNA adducts and
TP53 mutations
Benzo[a]‌pyrene Inadequate Sufficient Genotoxicity, including SCE, MN
­formation, DNA damage and 8-​oxo-​
deoxyguanosine, formation and specific
diolepoxide-​induced DNA adducts that
cause TP53 and K-​RAS mutations
Benzidine, dyes Inadequate Sufficient Strong mechanistic evidence indicating that
­metabolized to these dyes are converted to benzidine, a
carcinogen to humans, and produce DNA
adducts and other genotoxic effects similar
to those of benzidine
Ethylene oxide Limited Sufficient Genotoxicity, including SCE, ­chromosomal
aberrations, MN formation, mutagenic
and clastogenic activity, and DNA and
­hemoglobin adducts
Etoposide Limited Inadequate Genotoxicity, including SCE, aneuploidy,
mixed lineage leukemia (MLL) gene
translocations
4,4’-​Methlenbis Inadequate Sufficient Genotoxicity similar to other aromatic
(2-​chloroaniline) (MOCA) amines that are known to cause can-
cer of the urinary bladder in humans;
DNA adducts; protein adducts; SCE; MN
formation
4-​(Nnitrosomethylamino)-​ Inadequate Sufficient Genotoxicity, including DNA and hemo-
1-​(3-​pyridyl)-​1-​butanone globin adducts
(NNK), N’-​
Nitrosonornicotine (NNN)
Neutron radiation Inadequate Sufficient Structural and numerical chromosomal
aberrations (including rings, dicentrics, and
acentric fragments)
3,4,5,3’,4’-​ Inadequate Sufficient Strong evidence of action through the
Pentachlorobiphenyl same Ah receptor–​mediated mechanism
(PCB-​126**) (based on animal studies) as 2,3,7,8-​
tetrachlorodibenzo-​para-​dioxin, a
carcinogen in humans
(continued)
108

108 Textbook of Cancer Epidemiology

TABLE 5-​4  CONTINUED

Agent Evidence from Evidence from Mechanistic evidence

cancer studies cancer studies in in humans*
in humans animals

2,3,4,7,8-​ Inadequate Sufficient Strong evidence of action through the

Pentachlorodibenzofuran same Ah receptor–​mediated mechanism
(2,3,4,7,8-​PeCDF) (based on animal studies) as 2,3,7,8-​
tetrachlorodibenzo-​para-​dioxin, a
carcinogen in humans
“Dioxin-​like” polychlo- Inadequate Sufficient for Strong evidence of action through the
rinated biphenyls, with a PCB-​118 and same Ah receptor mediated mechanism
toxicity equivalency factor PCB-​126 (based on animal studies) as 2,3,7,8-​
according to WHO (PCB-​ tetrachlorodibenzo-​para-​dioxin, a
77, PCB-​81, PCB-​105, PCB-​ carcinogen in humans
114, PCB-​118, PCB-​123,
PCB-​126**, PCB-​169, PCB-​
156, PCB-​157, PCB-​167,
PCB-​189)
Ultraviolet radiation Inadequate Sufficient Specific C>T transition in TP53 in
­squamous cell carcinoma and solar keratosis

* Mechanistic evidence from animal studies is not included in the table, except for PCBs and 2,3,4,7,8-​PeCDF
** PCB-​126 is classified both as individual agent and as a member of “dioxin-​like” PCBs

are specific to certain cancer types. SCE and MN of exposure to and mechanisms of actions of car-
formation have been reported with exposure cinogens need to be identified, keeping in mind
to many carcinogenic agents; studies reporting that, in most cases, cancer is associated with long-​
these alterations were critical in the evaluation term, low-​dose exposures to carcinogens and is
of agents such as areca nut, the primary ingredi- based on biological events that take place during
ent in betel quid, which is widely used in South several decades. Due to temporal variability of
Asia, and the aromatic amine, 4,4’-​Methlenbis(2-​ exposures, assessment of biomarkers in a sin-
chloroaniline). Translocations in MLL gene are gle biological sample may not provide accurate
found in approximately 10% of human leukemia information about lifelong exposure to respective
cases (Winters & Bernt, 2017). The chemother- carcinogens. Even in case of cancers associated
apy drug etoposide has been classified as human with high exposures during a short time, samples
carcinogen because it has been shown to induce may not be collected in the most appropriate time
chromosomal translocations affecting MLL, a for assessment of exposure. Therefore, attempts
common finding in cases of acute myeloid leuke- should be made to assess exposures in repeated
mia related to chemotherapy regimens including samples in different time periods to have a more
this agent. comprehensive picture about patterns of expo-
sure. The concept of “exposome,” assessment of
CONCLUSIONS exposure from the perinatal period until adult-
Molecular epidemiology can provide evidence hood (Rappaport, 2011), has been introduced to
for exposure and biological effects of carcino- address the variability of exposure in epidemio-
gens. This can be particularly helpful in cases of logical studies, which also applies to molecular
mixed compounds, as identification of the effect epidemiology, but studies based on an exposomic
of individual exposures in traditional epidemio- approach have so far delivered limited results.
logic studies could be challenging. However, As discussed earlier, the contribution of
except for biological agents, its contribution to molecular cancer epidemiology has mainly been
the identification of human carcinogens has been limited to the identification of carcinogens acting
relatively modest. Additional biological markers through a geno-​or cytotoxic mechanism, that is,
109

Biomarkers in Cancer Epidemiology 109

when carcinogens induce damage to the DNA. Kant P, Hull MA. Excess body weight and obesity—​the
There is increasing understanding on the impor- link with gastrointestinal and hepatobiliary can-
tance of nongenotoxic mechanisms in human cer. Nat Rev Gastroenterol Hepatol 2011;8:224–​38.
cancer. Important examples are certain hor- Lin BK, Clyne M, Walsh M, Gomez O, Yu W, Gwinn M
mones (such as 17β-​estradiol and androstene- et al. Tracking the epidemiology of human genes
dione) (National Toxicology Program Technical in the literature:  the HuGE Published Literature
Report 2010; Spink et  al, 2009), immunosup- database. Am J Epidemiol 2006;164:1–​4.
pressants (such as cyclosporine) (Bugelski et  al, Moon HS, Chamberland JP, Aronis K, Tseleni-​
Balafouta S, Mantzoros CS. Direct role of adipo-
2010; van Kesteren et al, 2009), and overweight/​
nectin and adiponectin receptors in endometrial
obesity (Kant & Hull, 2011; Moon et  al, 2011).
cancer:  in vitro and ex vivo studies in humans.
These biomarkers act through receptor modifica-
Mol Cancer Ther 2011;10:2234–​43.
tion, effects on immune system, and epigenetic National Toxicology Program Technical Report.
changes, but these mechanisms are not clear in Toxicology and carcinogenesis studies of andro-
the majority of cases. Overall, as nongenotoxic stenedione (CAS No. 63-​ 05-​
8) in F344/​ N rats
mechanisms of carcinogenicity are not well iden- and B6C3F1 mice (gavage studies). Natl Toxicol
tified and their biomarkers are sparse, the contri- Program Tech Rep 2010;Ser 1:7–​1171.
bution of these mechanisms to identification of Norppa H, Bonassi S, Hansteen IL, Hagmar L,
human carcinogens has been limited. This area Stromberg U, Rossner P et al. Chromosomal aber-
of molecular epidemiology deserves priority rations and SCEs as biomarkers of cancer risk.
through the development of biological markers Mutat Res 2006;600:37–​45.
of nongenotoxic mechanisms and alterations that Perera FP, Weinstein IB. Molecular epidemiology
are specific to this class of carcinogens. and carcinogen-​ DNA adduct detection:  new
approaches to studies of human cancer causation.
J Chronic Dis 1982;35:581–​600.
REFERENCES Porta M, Malats N, Vioque J, Carrato A, Soler M, Ruiz
Ambrosone CB, Harris CC. The development of L et al. Incomplete overlapping of biological, clin-
molecular epidemiology to elucidate cancer risk ical, and environmental information in molecular
and prognosis: a historical perspective. Int J Mol epidemiological studies: a variety of causes and a
Epidemiol Genet 2010;1:84–​91. cascade of consequences. J Epidemiol Commun
Amos CI, Wu X, Broderick P, Gorlov IP, Gu J, Eisen T Health 2002;56:734–​8.
et al. Genome-​wide association scan of tag SNPs Rappaport SM. Implications of the exposome for
identifies a susceptibility locus for lung cancer at exposure science. J Expo Sci Environ Epidemiol
15q25.1. Nat Genet 2008;40:616–​22. 2011;21:5–​9.
Bonassi S, Merlo F, Pearce N, Puntoni R. Bladder can- Ross RK, Yuan JM, Yu MC, Wogan GN, Qian GS, Tu
cer and occupational exposure to polycyclic aro- JT et al. Urinary aflatoxin biomarkers and risk of
matic hydrocarbons. Int J Cancer 1989;44:648–​51. hepatocellular carcinoma. Lancet 1992;339:943–​6.
Bugelski PJ, Volk A, Walker MR, Krayer JH, Martin Spink BC, Bennett JA, Pentecost BT, Lostritto N,
P, Descotes J. Critical review of preclinical Englert NA, Benn GK et  al. Long-​term estrogen
approaches to evaluate the potential of immuno- exposure promotes carcinogen bioactivation,
suppressive drugs to influence human neoplasia. induces persistent changes in gene expression, and
Int J Toxicol 2010;29:435–​66. enhances the tumorigenicity of MCF-​7 human
Chen CH, Dickman KG, Huang CY, Shun CT, Tai HC, breast cancer cells. Toxicol Appl Pharmacol
Huang KH et  al. Recurrence pattern and TP53 2009;240:355–​66.
mutation in upper urinary tract urothelial carci- Spitz MR, Amos CI, Dong Q, Lin J, Wu X. The
noma. Oncotarget 2016;7:45225–​36. CHRNA5-​ A3 region on chromosome 15q24-​
Dietrich C, Kaina B. The aryl hydrocarbon receptor 25.1 is a risk factor both for nicotine depend-
(AhR) in the regulation of cell-​cell contact and ence and for lung cancer. J Natl Cancer Inst
tumor growth. Carcinogenesis 2010;31:1319–​28. 2008;100:1552–​6.
IARC. Preamble. In: Some Aromatic Amines, Organic Thorgeirsson TE, Geller F, Sulem P, Rafnar T, Wiste
Dyes, and Related Exposures. IARC Monographs A, Magnusson KP et al. A variant associated with
on the Evaluation of Carcinogenic Risks to nicotine dependence, lung cancer and peripheral
Humans, Vol. 99. Lyon: IARC Press, 2010: 9–​38. arterial disease. Nature 2008;452:638–​42.
Ioannidis JP. Interpretation of tests of heterogene- Truett J, Cornfield J, Kannel W. A multivariate anal-
ity and bias in meta-​analysis. J Eval Clin Pract ysis of the risk of coronary heart disease in
2008;14:951–​7. Framingham. J Chronic Dis 1967;20:511–​24.
10

110 Textbook of Cancer Epidemiology

Tworoger SS, Hankinson SE. Use of biomarkers in epi- van Kesteren PC, Beems RB, Luijten M, Robinson J,
demiologic studies:  minimizing the influence of de Vries A, van Steeg H. DNA repair-​deficient
measurement error in the study design and analy- Xpa/​p53 knockout mice are sensitive to the non-​
sis. Cancer Causes Control 2006;17:889–​99. genotoxic carcinogen cyclosporine A:  escape
Tucker JD, Eastmond DA, Littlefield LG. Cytogenetic of initiated cells from immunosurveillance?
end-​points as biological dosimeters and predic- Carcinogenesis 2009;30:538–​43.
tors of risk in epidemiological studies. IARC Sci Winters AC, Bernt KM. MLL-​rearranged leukemias-​
Publ 1997;142:185–​200. Review. an update on science and clinical approaches.
Ugolini D, Puntoni R, Perera FP, Schulte PA, Bonassi Front Pediatr 2017;5:4.
S. A bibliometric analysis of scientific production Workshop Report Application of biomarkers in can-
in cancer molecular epidemiology. Carcinogenesis cer epidemiology:  workshop report. IARC Sci
2007;28:1774–​9. Publ 1997;142:1–​18.
1

6
Concepts in Cancer Epidemiology and Etiology
P A G O N A L A G I O U , D I M I T R I O S T R I C H O P O U L O S † , A N D H A N S - ​O L O V   A D A M I

E pidemiology has been a powerful tool in the

identification of causes of infectious diseases
and the elucidation of the conditions underlying
exposure, such as many infectious and occupa-
tional diseases, or documented by a constella-
tion of clinical and/​or laboratory findings, for
epidemic outbreaks, which are frequently but not example, malignant tumors, connective tissue
always of infectious etiology. Around the middle disorders, or psychoses. In terms of a particu-
of the 20th century, first in the United Kingdom lar individual, exposure to a cause of a disease
(Doll & Hill, 1950) and later in the United States implies that the individual is now more likely
and the rest of the world (Wynder & Graham, to develop the disease, although there is no cer-
1950; Clemmesen & Nielsen, 1957; MacMahon, tainty that this will happen. The complexity of
1957), epidemiology expanded in scope by focus- biological phenomena and our ignorance or lim-
ing also on the etiology of chronic diseases, irre- ited understanding of many of the underlying
spective of the nature of the causal agents. Since processes hinder a deterministic, logically unas-
then, epidemiology has developed and matured sailable, explanation of disease causation. Hence,
to become a rich and powerful toolbox for the causation of disease can only be conceptualized
study of biologic phenomena in humans. With a in a probabilistic (stochastic) sense that involves
number of fine textbooks nowadays available to statistical terms and procedures. For instance,
students of epidemiology (for instance Miettinen, while heavy smokers are much more likely to
1985; Hennekens & Buring, 1987; Walker, 1991; develop lung cancer than nonsmokers, most
MacMahon & Trichopoulos, 1996; Rothman & smokers never develop lung cancer and some
Greenland, 1998; Rothman, 2002; and several nonsmokers do.
others), this chapter is not intended to expand In epidemiology, there are several models of
on methods or quantitative considerations. For causality that have been applied to help clarify
the purpose of better understanding the logic the role of various exposures in the etiology of
underlying cancer epidemiology, however, cen- disease. The causal pies presented by Rothman
tral concepts in epidemiology—​the study of dis- (1976) provide perhaps the most coherent
ease etiology—​are reviewed. We examine cohort approach to conceptualizing causality in a vari-
and case-​control studies, with special reference ety of epidemiologic settings (Rothman, 1986).
to studies of genetic epidemiology, we consider Each of these pies describes a set of exposures
the impact of chance and systematic errors (con- that work together on the same pathway to cause
founding and bias), and we trace the process of disease (Fig. 6-​1). Different exposures may occur
causal reasoning. Familiarity with these concepts within a short time span, or may happen decades
is essential for critical reading and understand- apart. Once every exposure in a causal pie has
ing of the chapters on specific cancers. A glossary occurred, that is, the pie is complete, disease is, in
found at the end of the chapter provides a sum- a deterministic context, inevitable. Table 6-​1 pro-
mary of definitions for words in italics. vides a summary of the attributes of the causal
pie model.
ETIOLOGY Causality is rarely, if ever, characterized by
a simple one-​to-​one correspondence between a
Causality particular exposure and a specific disease. If so,
The definition of a cause should apply to all dis- the presence of the exposure would be both nec-
eases, whether defined on the basis of a particular essary and sufficient for the occurrence of the
12

112 Textbook of Cancer Epidemiology

age
tobacco alcohol genes tobacco alcohol
HPV

no low no low
dental age dental
fruit fruit genes
visit visit

FIGURE 6-​1 The causal pie model describes a set of exposures that work together in the same pathway to cause disease.
These are hypothesized ways in which a series of exposures could interact biologically over time to cause disease. This
figure provides an example of suggested sufficient causes of oral cancer. Tobacco is an established component cause in
many cases of oral cancer. However, tobacco use by itself is not enough for the disease to occur; in addition, oral can-
cer can occur among people who have never used tobacco. In a given causal pie, the complementary exposures can
occur simultaneously, or many years apart. If even one of the component causes did not occur, disease would be pre-
vented by this pathway, although a person could develop the disease by another mechanism (a different causal pie). 

disease. By necessary we mean that the disease single factors that are in and by themselves suffi-
cannot occur without the presence of that expo- cient (although not necessary) for the causation
sure (although other exposures may be required of a certain disease. Even powerful exogenous
for the occurrence of the disease). By sufficient factors, such as life-​ long heavy smoking, and
we mean a set of exposures that inevitably pro- strong genetic influences, like those conveyed
duce disease. There may of course be different by dominant breast cancer genes, do not always
ways by which one could get disease, and thus cause disease in an individual.
sufficient causes may not be necessary. Certain exposures are by definition neces-
In cancer epidemiology, the only known sary (although not sufficient) for the occurrence
examples of exposures that are sufficient to cause of a particular disease. For example, chronic
disease refer to the genetic origin of some rare lead disease cannot occur in the absence of lead
cancers due to dominant genes with complete exposure, and a motor vehicle injury requires
penetrance. In this instance, the causal pie would the involvement of a motor vehicle (MacMahon
require only one factor for the pie to be complete et  al, 1960; Hill, 1965; Rothman, 1976; Susser,
and this would be the way that carriers would get 1991; MacMahon & Trichopoulos, 1996).
the specific cancer. Also rare is the existence of Again, while these represent necessary causes,

TABLE 6-​1   AT TRIBUTES OF THE CAUSAL PIE

Attribute Description

Inevitability Completion of a sufficient cause (causal pie) is synonymous with eventual

­occurrence (though not necessarily diagnosis) of the disease.

Causality A component cause (piece of a causal pie) can involve presence of a detrimental
exposure or absence of a preventive exposure.

Burden of disease The amount of disease caused by a sufficient cause depends on the prevalence of all
complementary component causes.

Temporality Component causes can act far apart in time

Interaction Component causes in the same pie interact biologically to cause disease

Attributable fraction Different component causes are responsible for more than 100% of disease cases.
Disease prevention Blocking the action of any component cause prevents completion of the respective
sufficient cause and therefore prevents disease by that pathway.

Source: Rothman, 1976.

13

Concepts in Cancer Epidemiology and Etiology 113

there are additional cofactors that must work We could study two groups of people who are
in concert before disease is inevitable. Most comparable on every characteristic, except that
human cancers can occur via several path- one group had exposure and one did not. The
ways, so it is hard to define any single necessary randomized controlled trial closely approximates
cause. Asbestos, in relation to mesothelioma this goal. By randomly allocating who receives
(cancer of the pleura), and human papilloma an exposure, for example treatment, and who
virus infection, in relation to cervical squa- does not, the exposure occurs only because the
mous cell cancer, are close to being necessary. investigator has assigned it. For example, an
However, cases of these cancers do arise with- investigator randomly assigns one group of peo-
out the exposure being documentable, either ple to receive vitamin E supplements (exposed),
because the exposure occurred but could not while the other group receives a placebo (unex-
be identified or because these exposures are not posed). Study participants are then followed for-
necessary for all cases. ward in time to see whether they develop cancer.
For most diseases, there is no one necessary Whether someone receives vitamin E then does
cause. Indeed there may be numerous causal pies not depend on whether or not the subject for
by which disease can occur. Such an example is example smokes, drinks, eats a high fat diet, or
illustrated in Fig. 6-​1, with suggested sufficient has a certain genetic susceptibility.
causes of oral cancer. In the first example, expo- In this way, the randomization in a trial
sure to tobacco and alcohol over time are contrib- makes the two (or more) groups, those exposed
uting factors (component causes) in the etiology and those unexposed, comparable on other study
of oral cancer. However, the oral cancer would factors that might cause the disease. Hence, the
not have occurred in the presence of a dental visit unexposed group is a proxy of what would have
that could have treated precancerous lesions and happened to the exposed group if they had been
might have prevented the disease. While smok- unexposed, that is, if we could have sent them
ing is a component cause in many causal pies for back in the time machine. Comparability is essen-
oral cancer, people can get oral cancer without tial in order to ascribe any changes in the fre-
smoking, as shown by the second causal pie in quency of disease to alterations in the exposure.
this figure. While some researchers describe the ran-
domized controlled trial as the gold standard
Interventional Epidemiology of scientific studies, this design is impractical in
How do we design a scientific study to evalu- the majority of epidemiologic situations. For one
ate whether a particular exposure (for example, thing, most exposures we study are detrimen-
asbestos) is a cause of a specific disease (for exam- tal. If we want to study the impact of asbestos on
ple, lung cancer)? To understand the most appro- lung cancer, we cannot ethically randomize peo-
priate design in practice, it is useful to begin by ple to live in a house with asbestos. But even for
describing the ideal scientific study. Imagine for exposures that are not necessarily detrimental,
a moment that we have access to a time machine. randomization may be difficult or impractical.
In an imaginary study, we follow a group of For instance, it is very difficult and expensive to
individuals from birth to death, where everyone randomize a large group to eat a low-​fat versus a
is exposed to asbestos, and we observe whether normal diet, and have everyone comply with this
they develop lung cancer. We then send everyone allocation over the course of many years. Most
back in a time machine, to live the exact same trials are thus only conducted for no more than
lives they lived, except that we completely remove a few years, an unrealistically short period to test
asbestos from the environment so that no one is the effect of most exposures because of the long
exposed. We then compare whether there are latency between exposure and diagnosis of can-
changes in the frequency of occurrence of lung cer. Furthermore, in many randomized trials,
cancer before and after use of the time machine. subjects become noncompliant over time, that
Since the same people lived identical lives but for is, people allocated to the intervention arm stop
the presence/​absence of asbestos, any difference taking the intervention, and those in the original
in the frequency may be attributed to alterations placebo or usual care arm may adopt the inter-
in the exposure to asbestos, which leads to the vention (a phenomenon called cross-​over). This
definition of cause. diminishes the contrast between the original ran-
How then can we develop the time-​machine domized groups, reducing the power to detect a
analogy into a realistic epidemiologic approach? difference in disease rates between the groups.
14

114 Textbook of Cancer Epidemiology

Because of the limitations of the randomized or hepatitis B virus and liver cancer are striking
controlled trial, the observational cohort and examples. It becomes more difficult when the
case-​control designs are extensively used in epi- association is biologically compelling but the
demiology. As is discussed later in the chapter, epidemiologic experience weak, for example, in
attention to both the design and analysis of these studies of low-​level ionizing radiation and leuke-
studies may allow us to approximate the stan- mia or passive smoking and lung cancer. Causal
dards of comparability, necessary to validly eval- interpretation also becomes problematic when
uate the effect of an exposure on the frequency the epidemiologic association is fairly convinc-
of a disease. ing but the biological rationale is uncertain, as it
is with respect to red meat and colorectal cancer
Observational Epidemiology or alcohol and breast cancer. When an epidemio-
The essence of observational epidemiology is the logic association is weak, is derived from a study
noninterventional investigation of disease causa- with questionable quality and floats in a biologi-
tion in human population groups. The argument cal vacuum, inferring causation is perilous.
is that only by studying humans is it possible
to draw confident conclusions about normal STUDY DESIGN
or pathological processes concerning humans
(MacMahon, 1979; MacMahon & Trichopoulos, Descriptive Studies
1996). In vitro studies, such as those involv- It is possible to distinguish observational epide-
ing cell cultures, and studies in laboratory ani- miological studies into descriptive and analytic.
mals are valuable. They are indeed indispensable In descriptive studies the frequency of occur-
when toxic exposures or invasive procedures rence of a disease (incidence)—​or of death from a
like repeated biopsies are needed for the study disease (mortality)—​is estimated in a population,
of physiologic or pathologic processes, such as by routinely available time, place and/​or group
carcinogenesis. However, in vitro systems are fre- characteristics. Descriptive studies are essen-
quently artificial, and there are physiological and tially exploratory and hypothesis-​generating. For
metabolic differences between humans and lab- instance, descriptive studies that documented the
oratory animals that hinder interspecies analo- increasing trend of lung cancer incidence among
gies. These analogies are further complicated by men, but not among women, in the early part of
the unavoidably limited number of animals used the 20th century pointed to tobacco smoking as a
in laboratory studies and the relatively short life likely cause of this disease. In contrast, the objec-
span of these animals, both of which impose the tive of analytic studies is to document causation
administration of high doses of suspected agents from the pattern of association in individuals
in order to generate a sufficient number of out- between one or more exposures on the one hand
comes. Consequently, questionable quantitative and a particular disease on the other.
extrapolations to humans have to be undertaken.
Even when experimental studies, such as ran- Ecologic Studies
domized controlled trials, in humans are ethical, Ecologic studies in epidemiology occupy an
they are, with few exceptions, impractical because intermediate position between descriptive and
most diseases are rare and their latent period, analytic investigations, in that they share many
that is, the time between exposure to a cause and characteristics with descriptive studies, but
the appearance of a clinical disease, is long. This serve etiologic objectives. In ecologic studies,
makes it necessary to enroll unrealistically large the exposure and the disease under investiga-
numbers of compliant volunteers for a very long tion are ascertained not for individuals but for
period (Hennekens & Buring, 1987; MacMahon, groups or even whole populations (Morgenstern,
1979; MacMahon & Trichopoulos, 1996). 1982). Thus the prevalence of hepatitis B virus
Observational studies, that is nonexperimental (HBV) in several populations could be corre-
studies, represent the mainstream of modern epi- lated with the incidence of liver cancer in these
demiology. Such studies seek to document causal populations, even though no information could
relations on the basis of associations between par- be obtained as to whether any particular indi-
ticular exposures and cancer or other diseases. vidual in these populations was or was not an
Inference of causation on the basis of association HBV carrier and has or has not developed liver
is easy when the association is both strong and cancer. Associations from ecologic studies are
biologically credible—​smoking and lung cancer, viewed with skepticism, because these studies are
15

Concepts in Cancer Epidemiology and Etiology 115

susceptible to unidentifiable and intractable con- thus both should be taken into account. Person-​
founding as well as to several other forms of bias time is the sum of all the time contributed in a
(Morgenstern, 1982; Greenland & Robins, 1994). study by subjects at risk of a disease.
When an exposure is fairly common, for exam- Theoretically, an ambitious investigator might
ple, smoking, or even prevalence of HBV carriers, wish to include the entire world population in
ecologic studies can provide useful evidence on the an epidemiologic study during many decades.
possible effects of these exposures. For instance, Needless to say, such a study would provide mar-
following the increase in tobacco consumption, velous opportunities to evaluate many different
the incidence of lung cancer increased sharply over exposures in relation to many diseases. Millions
time, and the incidence of primary liver cancer is of person-​years would be generated even within
higher in populations with higher prevalence of a few weeks. In real life, however, any investiga-
HBV. As a corollary, lack of an association in eco- tor has to restrict the person-​time from which
logic studies between a widespread exposure that information is harvested. This specified person-​
has rapidly increased over time and the incidence time is called the study base. Defining the person-​
of a disease allegedly caused by this exposure, does time to be included in the study base may include
not support a strong causal relation. geographic restrictions, defined time periods, and
certain age limits. Personal characteristics such
Analytic Studies as gender, ethnicity, and occupation may further
Analytic epidemiologic investigations ascertain specify the study base. For example, the study
exposure and disease outcome in individuals and base may be composed of all British doctors who
are usually distinguished into cohort and case-​ answered a questionnaire in 1951 (Doll and Hill,
control studies, although there are also several 1956), or by all Swedish women who were aged 50
variants of these prototype designs (MacMahon to 74 between 1994 and 1995 (Weiderpass et al,
& Trichopoulos, 1996; Rothman & Greenland, 1999), and who generated person-​time until they
1998). The objective of analytic epidemiologic died or until the follow-​up was completed.
studies is to ascertain whether a particular expo- Thus, the study base is simply the person-​
sure, such as a physical, chemical, or biological time of a population of individuals at risk of a
agent, and a specific cancer or other disease are disease under study. Defining the study base
unrelated (independent) or associated. An asso- is a crucial step in the design and conduct of
ciation does not necessarily indicate causation. an epidemiologic study. There are three central
Chance, bias, and confounding (discussed later) considerations. One is to accommodate realistic
can also generate associations, and they fre- goals with regard to feasibility and resources, as
quently do. Causation is unlikely when there is certainly no investigator is independent of time
no association observed. Even if a causal relation and money. A second goal is to make the study
does exist, however, it may sometimes be difficult efficient. For example, it would make little sense
to document it, particularly when the association to study the association between smoking and
is weak, the study has limited statistical power, or cancer in a population where very few are smok-
the exposure is seriously misclassified. ers. Likewise, a study of diet and prostate cancer
would be inefficient among men younger than
Person-​Time and Study Base 40, since virtually no cases arise among such
The concepts of person-​time and study base are young people. The final challenge is to identify a
fundamental to the design and analysis of epide- study base that allows valid inferences concern-
miologic studies. As the name implies, there are ing associations between exposure(s) and a par-
two key components in our description of the ticular disease, that is, a study base that does not
person-​time, namely the number of people and impose intractable confounding or raise insur-
the time they are followed. To illustrate this, we mountable obstacles of other biases.
could ask how many brain cancer cases we would Person-​time is the source of any event we
expect if we followed one million people exposed want to investigate, for example the occurrence of
to x-​rays for zero seconds. Conversely, how many cancer. To help set a foundation for better under-
cases would we expect if we followed zero peo- standing person-​time in a study base, we will
ple for one million years? The answer in both use the example of x-​rays and risk of brain can-
instances is, of course, zero. Hence, neither peo- cer (Fig. 6-​2). In this population, five people have
ple nor time alone provides adequate information been exposed to x-​rays, and another five have not
about the disease experience of a population, and been exposed and remain unexposed during the
16

116 Textbook of Cancer Epidemiology

Person 1 Brain cancer

Person 2 Brain cancer

Exposure to
x-rays Person 3

Person 4 Brain cancer

Person 5

Person 6

Person 7
No exposure Person 8
to x-rays
Person 9 Brain cancer

Person 10

0.0 1.0 2.0 3.0 4.0 5.0

Time

FIGURE 6-​2  Experience of a theoretical study population over time. Five individuals who were exposed to x-​rays and
five individuals who are unexposed are followed over time to see whether they develop brain cancer. Among those
exposed to x-​rays, persons 3 and 5 are followed for the duration of the study period, which in this case was 5 years.
Person 1 develops brain cancer after 1 year, person 2 develops brain cancer after 4 years, while person 4 develops can-
cer after 2 years. Persons 1, 2, and 4 stop contributing person-​time after they develop brain cancer, since they are no
longer at risk for the disease. The total person-​time in the exposed group is 17.0 person-​years. Similarly, we can look
at the population of people unexposed to x-​rays over time. Of the five people who are unexposed, only one develops
brain cancer during the study period (Person 9). The remaining people are followed completely for 5 years. The
experience of these 10 individuals over time, that is the person-​time that the subjects contributed, is the study-​base. 

study period. While in real life the study popu- When we discuss analysis of epidemiologic
lations are much larger, we use this elementary studies in what follows, we see how the person-​
example to illustrate the principles. time data help us to compare disease incidence
Among the people exposed to x-​rays, persons between exposed and unexposed people.
3 and 5 were followed from the time they were
exposed to x-​rays till the end of the study period, Cohort Studies
a total of 5 years each. Persons 1, 2, and 4, how- The word “cohort” derives from the similar
ever, developed brain cancer at the end of years Latin word, which identified one of the ten
1, 4, and 2, respectively. Once these individuals divisions in a Roman legion. In epidemiology,
develop brain cancer, they are no longer at risk cohorts are groups of individuals, which can
of the disease, and thus no longer contribute be followed over time. In cohort studies, indi-
information to the study base. The person-​time viduals are classified according to their expo-
among those exposed to x-​rays is estimated by sure and are observed for ascertainment of
summing up the person-​time of all the individu- the frequency of disease occurrence or death
als while at risk for the disease, that is: in the various exposure-​ defined categories
(Fig.  6-​
3A). In each category the frequency
(2 persons × 5 years) + (1 person × 1 year) of occurrence is calculated either as risk or as
+ (1 person × 4  years) + (1 person × incidence rate. Risk describes the proportion of
2 years) = 17 person-​years those who developed the disease under study
We can similarly sum up the person-​time among among all individuals in this category. Rate
the group of five individuals who were not describes the number of those who developed
exposed to x-​rays: the disease divided by the person-​time dur-
ing which the individuals in this category have
(4 persons × 5  years) + (1 person × been under observation. Cohort studies have
2 years) = 22 person-​years the following defining characteristics:
17

Concepts in Cancer Epidemiology and Etiology 117

A
x1 Exposed cases

Exposed person-time (P1)

Non-exposed person-time (P0)

Incidence rate ratio = x0 Unexposed cases

(x1/P1) (x0/P0)

FIGURE 6-​3A  A cohort study comprises individuals who are either exposed or unexposed to the factor(s) of inter-
est. When these people are followed over time, they generate person-​time. Newly diagnosed cases of a particular
disease that occur while person-​time is accumulated are recorded. The exposure status of a person can change. A
person could be smoking high ​tar cigarettes for 5 years, then switch to light cigarettes for 15 years, and then quit.
Consequently, each person can contribute to person-​time in different exposure groups. A case is considered exposed
if the disease occurred when the person who developed the disease was accumulating exposed person-​time. A case is
non-exposed if it occurred while the person was accumulating non-exposed person-​time. The example assumes, for
simplicity, zero latency. The total amount of exposed and non-exposed person-​time and the number of exposed and
non-exposed cases can be calculated. After that, one can determine whether more cases occurred in the exposed or
non-exposed group per unit of person-​time, that is, one can calculate the incidence rate ratio. This ratio will indicate
whether there is a relationship between the exposure and the disease of interest. 

Cohort Studies Are Exposure-​Based a study factor, include the British Doctors Study
The groups to be studied are selected on the and the Nurses Health Study. The British Doctors
basis of exposure. In special exposure cohorts, cohort, established in 1951, consisted of more
the groups are chosen on the basis of a partic- than 30,000 doctors from Great Britain. In this
ular exposure. In general population cohorts, landmark study, Doll and colleagues prospec-
groups offering logistical advantages for follow-​ tively followed the cohort and collected updated
up are initially chosen and the individuals are information on multiple exposures, particularly
classified according to their exposure status. smoking, over several decades. Indeed, prospec-
Special exposure cohorts may be necessary when tive data from the British doctors were among
rare exposures need to be studied, such as those the first to demonstrate convincingly the role
encountered in the occupational setting. For of tobacco in the etiology of lung cancer (Doll
example, to study efficiently the effect of vinyl and Hill, 1956). More than four decades later,
chloride on liver angiosarcoma, or aromatic data from the British Doctors have continued to
amines on bladder cancer, epidemiologic studies provide insight into the etiology of cancer (Doll
have been conducted in cohorts of workers in the et al, 2005).
plastic and dyestuff manufacturing industries, Another notable cohort is the Nurses Health
respectively. Study, which began in 1976 with over 120,000
The general population cohort is appropriate US registered nurses. This cohort was assem-
when the exposure under consideration is fairly bled initially to evaluate prospectively the effect
common. Classical examples of general popula- of oral contraceptives on the risk of breast can-
tion cohorts, in which the profession facilitated cer (Hennekens et  al, 1984). Subsequently,
accessibility of cohort members rather than being diet and many other exposures have been
18

118 Textbook of Cancer Epidemiology

studied in relation to the risk of cancer as well Open cohorts are composed of individuals
as other chronic conditions (Zhang et al, 2005). who contribute person-​time to the cohort only as
Information on these diverse exposures has been long as they meet the criteria for a membership-​
collected biennially through questionnaires. defining state (Fig. 6-​3A). Examples of such cri-
Moreover, blood samples have allowed research- teria include place of residence, age, and health
ers to explore biomarkers and genetic factors. status. Once individuals can no longer be char-
For example, prospective data from the Nurses acterized by the defining state(s), they cease to
Health Study has provided insight into the role contribute person-​time to the open cohort and
of both exogenous and endogenous estrogens in are no longer members. Open cohorts are used,
breast cancer etiology. A particular characteristic for example, in cancer epidemiology studies
of these types of cohorts is that the individuals based on registry data (Hansson et  al, 1996).
can be followed almost completely over time, due A person could be a member of the cohort, for
to their membership in groups with a high inter- example, only as long as he or she was a resident
est in health studies and registration require- of Sweden and was not diagnosed with the can-
ments that facilitate initial contact and long-​term cer under study. If the person emigrated from
follow-​up. Sweden to another country, he or she stopped
contributing person time to the cohort at that
Cohort Studies Are Patently or time. Similarly, if someone born outside of
Conceptually Longitudinal Sweden immigrates there later in life, he or she
The study groups are observed over a per- would begin contributing person-​ time to the
iod of time to determine the frequency of dis- cohort at that time. In studies based on open
ease occurrence among them. The distinction cohorts it is not possible to directly measure risk,
between retrospective and prospective cohort otherwise referred to as “cumulative incidence.”
studies depends on whether the cases of dis- Analyses are based on person-​time using inci-
ease occurred in the cohort at the time the study dence rate measures.
began. In a retrospective cohort study, exposures As an example, assume that in a closed cohort
and health outcomes occurred before the inves- among 5,000 nonsmoking men followed for an
tigation started. These are typically assembled average period of 10 years (P0 = 50,000 person-​
from preexisting records of a population over years), x0 = 25 were diagnosed with lung cancer,
time, for example, the employment histories of a and among 10,000 smoking men followed for an
factory can be linked to recorded health outcome average period of 8  years (P1  =  80,000 person-​
information of the workers. In a prospective years), x1 = 600 were diagnosed with lung cancer.
cohort study, the relevant causes may or may not In this example the incidence rate among non-
have acted and the cases of disease certainly have exposed would then be 50 per 105 person-​years
not yet occurred. Hence, following identification and among exposed 750 per 105 person-​years.
of the study cohort, the investigator must wait for The relative risk (incidence rate ratio) would be
the disease to appear among cohort members. 600 / 80,000
Methodologically, there are two types of , or 15. The conclusion is that there
25 / 50,000
cohort studies: closed or fixed cohorts, and open
is a 15-​fold increase in lung cancer occurrence
or dynamic cohorts. Closed cohorts are frequent
from smoking.
in occupational epidemiology and the study of
outbreaks, whereas open cohorts dominate can-
cer epidemiology and form the conceptual basis Case-​Control Studies
for most case-​control studies. The key distinc- In case-​control studies, patients diagnosed with
tion between open and closed cohorts is how the disease under consideration form the case
membership in the cohort is determined. In a series. As in cohort studies, their exposure to
closed cohort, it is determined by a membership-​ the factor under investigation is ascertained, for
defining event that occurs at a point in time. For example, through questionnaires, interviews,
example, people who were living in Hiroshima examination of records, undertaking of labora-
and Nagasaki when the atomic bombs were tory tests in biological samples, and other means
dropped in 1945 are part of a cohort whose (Fig. 6-​3B). Using the same methods, the pattern
membership began on the date of the bomb- of exposure to the study factor(s) is then esti-
ing. These subjects remain in the cohort until mated in the population, or more strictly in the
they die. person-​time from which the case series arose.
19

Concepts in Cancer Epidemiology and Etiology 119

B
x1 Exposed
cases

y1 Exposed
Exposed person-time (P1) controls

Non-exposed person-time (P0)

y0 Unexposed
controls

Odds ratio =
x1y0/x0y1 x0 Unexposed
cases

FIGURE 6-​3B  It is not always practical or economical to evaluate the entire study person-​time. The case-​control study
is a more efficient design. Instead of enumerating the total amount of exposed and non-​exposed person-​time that
makes up the study person-​time, the ratio of exposed to non-​exposed person-​time is estimated. This is achieved
by randomly selecting people (controls) without the disease of interest from the underlying study person-​time and
determining their exposure status. If a sufficient number of controls are selected, without regard for their exposure
status, then the exposure distribution in the controls will estimate that in the total study person-​time. The exposure
distribution among the controls is then compared to that among the cases of the disease of interest that have arisen
in the study person-​time. An estimate can then be made of the odds of exposure among the cases compared to that
among controls, or the odds ratio, which is an unbiased estimator of the incidence rate ratio and so indicates whether
there is an association between the disease and the exposure of interest. 

This is done among control subjects selected as to the study of the etiology of cancer and other
a sample of the study base from which the cases conditions.
arose. If only two categories of exposure are rel-
evant (exposed and unexposed), the relative risk Nested Case-​Control Studies
can be estimated by dividing the odds of expo- Some case-​ control designs are methodologi-
sure among cases with the corresponding odds cally superior to others. The best example is the
among the controls, the odds ratio. nested case-​ control design. The definition of
Thus, if among 200 male patients diagnosed this study design is still somewhat ambiguous
with lung cancer (cases), a = 150 were smokers (Walker, 1991; Rothman & Greenland, 1998).
and b  =  50 nonsmokers, whereas among 300 A  definite requirement, however, is that con-
men similar in age to the cases but without lung trols are chosen from the clearly defined person-​
cancer (controls), c  =  50 were smokers and time from which all cases have arisen. In other
d = 250 were nonsmokers, the odds ratio would words, if one of the controls had developed the
disease under study, he or she would have defi-
a / b ad 150 × 250
be = = or15 . This measure is a nitely been included among the cases. Defining
c / d bc 50 × 50 the underlying person-​time from which a series
good approximation to the relative risk (or risk of cases—​for example, lung cancer cases present-
ratio, or rate ratio). Hence, similar to the cohort ing at a referral hospital—​arose can be difficult.
study, these data from a case-​ control study Sampling controls from a cohort different from
show a 15-​fold excess of lung cancer among the one that gave rise to the cases often results in
smokers. selection bias.
There are some features of case-​control stud- According to a more strict definition, the
ies that make this design susceptible to bias (see term nested case-​control study is used only when
following). A  well-​designed case-​control study, the underlying cohort and the corresponding
however, is a valid and cost-​efficient approach person-​time have been previously enumerated
120

120 Textbook of Cancer Epidemiology

and the exposure information was collected prior and, thus, the ability to substantiate a true asso-
to the diagnosis. In other words, the controls are ciation (Rothman & Greenland, 1998). What is
selected from exactly the same person-​time that necessary, however, is that, whenever matching
gave rise to the cases, the study base. Unlike the has been used in the enrollment of cases and con-
traditional case-​control design, which is liable to trols, the statistical analysis should accommodate
bias due to selective participation and differences the matching process. This can be done through
in recall, this nested design preserves the valid- either a matched analysis (for example, condi-
ity of a prospective cohort study. Case-​control tional modeling) or unmatched analysis with
studies nested in an existing cohort are being explicit control for the matching factors (proper
used increasingly for cost-​efficiency when anal- application of unconditional modeling).
ysis of all cohort members requires substantial
resources. Studies of the Genetic
Nested case-​control studies have been fre- Epidemiology of Cancer
quently used in occupational epidemiology Genetic epidemiology of cancer is considered
(Rothman & Greenland, 1998). The occupational in more detail in a distinct chapter (­chapter  4).
cohort can often be readily defined, whereas Here, we refer briefly to such studies, to provide
abstraction of detailed exposure information an integrated picture of epidemiologic designs
from existing records requires substantial work. available for the study of cancer etiology. Two
Hence, it is more efficient to investigate only the main types of epidemiologic studies are used for
cases of interest and a sample from the cohort the identification of genes predisposing to can-
that is the controls. Nowadays, nested case-​ cer: genetic linkage studies and genetic associa-
control studies are used routinely when exposure tion studies.
information is derived, often through expensive Genetic linkage studies are generally under-
laboratory procedures, from biologic samples taken in families with a high cancer burden and
such as blood or blood products, tissue, urine, rely on the principle that two genetic loci, or a
or nails. cancer and a particular locus, are linked when
One such example is a study of selenium sta- they are transmitted together from parent to
tus and breast cancer risk in the Nurses Health offspring more often than expected by chance.
Study (Hunter et al, 1990). On the basis of prior Linkage extends over large regions of the genome
evidence that selenium intake may influence and refers to a locus, rather than specific alleles
breast cancer risk and since selenium levels in in that locus, which can vary from study to study
toenails are a reliable source of selenium exposure and from family to family (Teare & Barrett,
over several months, the participating women 2005). Such studies have led to the identifica-
were asked to provide toenail clippings in 1982. tion of genes that have substantial impact on the
After 4 years of follow-​up, there were 434 cases of occurrence of breast cancer and colorectal can-
breast cancer. It would have been very expensive cer, but these genes are generally rare, presuma-
and inefficient to get exposure information for bly because of natural selection pressure.
all the 62,000 nurses who had returned toenail Genetic association studies can be of either
samples at the start of follow-​up. Hence, 434 con- cohort or, more frequently, case-​control design.
trols without breast cancer were sampled from They are frequently undertaken in the general
the cohort. Using this design meant that only population, rather than in families, and are con-
868 rather than 62,000 samples had to be sent ceptually similar to traditional epidemiological
to the laboratory for selenium analyses (Hunter investigations. The difference is, however, that
et al, 1990). instead of focusing on environmental factors, like
smoking or diet, genetic association studies evalu-
Matching in Case-​Control Studies ate as “exposures” specific alleles (rather than loci)
Occasionally, case-​control studies are matched. of genetic polymorphisms, usually single nucleo-
This means that controls are chosen so as to match tide polymorphisms (SNPs). The specific alleles
particular cases with respect to gender, age, race, may be etiologically related to cancer or, much
or any other factor that is likely related to the dis- more frequently, very closely linked to the truly
ease under investigation but not intended to be etiological allele, which may not be known. The
analyzed in the particular study. Matching is not actually investigated allele and the true etiological
strictly necessary, nor does it increase the valid- allele are said to be in linkage disequilibrium, that
ity of results. But it improves statistical efficiency is, they are so closely linked that they tend to be
12

Concepts in Cancer Epidemiology and Etiology 121

inherited together. Two loci in linkage disequilib- Let us take, as an example, tossing a fair
rium are obviously linked, but two linked loci may (unbiased) coin that has a 50% or 0.5 probabil-
not be in linkage disequilibrium if they are suffi- ity of turning up heads and an identical proba-
ciently apart in the chromosome to be separated, bility of turning up tails. Tossing the coin three
sooner or later, by the frequent cross-​over process times and getting three heads in a row is some-
in the meiosis phenomenon during the gener- what unusual but it can hardly be taken as indi-
ation of gametes. In other words, linkage covers cation that the coin is systematically influenced
longer genetic regions than linkage disequilib- (biased) toward tails. The p-​value in this instance
rium (Cordell & Clayton, 2005; Teare & Barrett, is 0.25 and is calculated by multiplying 0.5 × 0.5 ×
2005). The specific allele may be chosen for study 0.5 = 0.125, and then doubling 0.125, because the
because the corresponding locus is thought to be symmetrically opposite outcome, three tails in a
involved in the etiology of the cancer under inves- row, is as extreme as three heads in a row. Getting
tigation (e.g., a candidate gene). Many SNPs over five heads or tails in a row generates some sus-
large parts of the genome, or even over the whole picion (P = (1 / 2)5 × 2 = 0.0625) . But if 100 peo-
genome may also be evaluated, with little or no ple have tossed a fair (unbiased) coin five times
prior evidence that most of them are etiologically each, it should be expected that about six (100 ×
relevant or are in linkage disequilibrium with eti- 0.0625) among them would have obtained either
ologically relevant genes. In the latter situation, five heads or five tails in a row.
most statistically significant findings are likely It must be realized that stochastic (prob-
to be “false positives,” and special procedures are abilistic), in contrast to deterministic, pro-
recommended to delineate which ones among cesses always have built-​ in uncertainty. In
the apparent associations are probably genuine their research, all investigators want to reduce
(Wacholder et al, 2004). Genetic association stud- chance-​related uncertainty as much as possi-
ies have not been very successful to date in identi- ble in order to allow more reliable conclusions.
fying genes or polymorphisms involved in cancer This can be achieved mainly by enrolling pro-
etiology, possibly because the respective relative gressively larger numbers of individuals in a
risks deviate very little from the null value, but study. The remaining uncertainty can always
also because the tools to examine alleles across be assessed by utilizing statistical procedures
the genome simultaneously—​rather than at a lim- that generate a number of summary statistics,
ited number usually selected on the basis of weak including the p-​value.
prior probabilities of being truly associated—​are The true meaning of the p-​value, however,
only just becoming available. is poorly understood, and the concept itself is
widely misused. Surprisingly, this misunderstand-
THE ROLE OF CHANCE ing and misuse is quite common even in scientific
Before an epidemiologic association could be research. P-​values are traditionally expressed as
considered true and therefore deserve interpre- numerical fractions of 1. For example, a p-​value
tation in causal terms, the role of chance and sys- of 0.1 for a particular positive association (or dif-
tematic errors should be considered. ference) indicates that there is a 10% chance that
such an association or a more extreme one (or a
The P-​Value symmetrically opposite one, that is, an inverse
Our daily lives are full of highly unlikely events association) would appear by chance, even if there
and coincidences. At the extremes, thousands were in reality no association at all.
of people have become wealthy from lotter- In essence, the p-​value is interpretable as such
ies. Many more have died in strange accidents; when only one comparison or one test is per-
even though the probabilities for the respective formed. When multiple comparisons or multiple
events are extremely small, say one in 100,000 tests are carried out the set of the respective p-​
or smaller, the lesson is simple:  highly unlikely values loses its collective interpretability. Various
events happen by chance all the time. Chance procedures for adjusting p-​values according to
does not operate differently in scientific research the number of comparisons undertaken or tests
and everyday life. In science, however, proper performed have been proposed (Wacholder
quantification and judgment, relying on sound et al, 2004).
substantive knowledge, are necessary before con- A p-​value of 0.05 or smaller is traditionally—​
sidering chance as an unlikely explanation for a and indeed arbitrarily—​ treated in medi-
phenomenon. cal research as evidence that an observed
12

122 Textbook of Cancer Epidemiology

association may not have arisen by chance. For Confidence Intervals

example, the proportion of long-​term smok- In order to integrate information about the
ers is found to be larger among lung cancer strength of an association (as reflected in the rela-
patients than among individuals without the tive risk effect measure, described later on) and its
disease and the p-​value for this difference is, statistical significance, the concept of confidence
say, 0.05. This implies that the probability of interval has been developed. Most common are
finding a difference of this magnitude or larger 95% confidence intervals. With a 95% confidence
(in absolute terms) is 5% if smoking were unre- interval, one can be 95% confident that the inter-
lated to lung cancer. In this situation, chance is val covers the true measure of association (for
considered unlikely to explain the association. example the relative risk). But in five times out
However, small p-​values, including values con- of a hundred, the true measure is not included.
siderably smaller than 0.05, do not guarantee The confidence interval is closely linked to the
that an association (or difference) is genuine—​ p-​value. The width of the confidence interval is
let alone causal. determined primarily by the desired level of con-
Even when the p-​value is very small and was fidence and the sample size. Hence, the interval is
generated from a carefully conducted study, it wider if it includes the true value with 95% con-
could still be dismissed when the relevant result fidence than with, for example, 80% confidence.
makes no sense (Miettinen, 1985). Hence, a sta- Likewise, smaller studies create wider confidence
tistically significant association, linked by con- intervals—​that is, greater uncertainty about the
vention to a p-​value of 0.05 or less, does not true value—​than larger studies.
necessarily imply causation. Systematic errors,
generated by confounding or bias (discussed S Y S T E M AT I C   E R R O R S
later), cannot always be confidently discounted
in observational epidemiology. Moreover, as The Experimental Study
indicated at the end of this chapter, the existence The chance-​related issues apply to all types of
of a genuine association that can be confidently studies, observational as well as experimental.
attributed to causation does not necessarily As discussed earlier, experimental studies under-
imply that someone who developed the disease taken under optimal conditions are methodo-
following the exposure did so because of that logically superior to observational studies. With
exposure. randomization of exposure, complete follow-​up
A common misconception (Miettinen, of study subjects, and double-​blind assessment of
1985) is that if a p-​value (for example, p = 0.03) outcome, they are not as liable to the pitfalls of
has been properly derived, then its comple- typical observational studies, that is, confound-
ment, 0.97 in our example, can be interpreted ing and bias (Miettinen, 1985; Hennekens &
as the likelihood that the respective association Buring, 1987; MacMahon & Trichopoulos, 1996;
is indeed causal. This misconception is rarely Rothman & Greenland, 1998; Rothman, 2002).
stated explicitly in the scientific literature, but Proper evaluation of the association between a
it underlies the conclusions of many epidemi- particular exposure and a specific disease pre-
ologic reports that are not securely anchored supposes that every other factor that could influ-
in methodological principles and biomedical ence disease occurrence is either constant among
substance. subjects studied or distributed equally between
Lastly, it must be recognized that the p-​value exposed and unexposed subjects.
itself does not convey any information about the In other words, an experimental study uses
strength of the respective association. A  weak random allocation of study subjects into those
association may be statistically highly signifi- who will be exposed and those who will not.
cant (very small p) when the study is large, and Thus, the two or more groups will tend to be sim-
a strong association may be statistically non- ilar in distribution to known as well as unknown
significant (large p) when the study is small. factors that may influence the results. In some
Hence, all p-​ values are inherently dependent studies, blinding of researchers and study sub-
on the study size, because statistical power—​the jects through the use of appropriate procedures
ability to detect an association (or a difference) and devices (for example, indistinguishable
when it exists—​increases when a study is larger inert pills, the so-​called placebos) may further
(Rothman & Greenland, 1998). assure that every factor that can affect disease
123

Concepts in Cancer Epidemiology and Etiology 123

occurrence, other than the exposure under study, epidemiologic studies have also generated con-
is kept at about the same level between the flicting results, unwarranted concern about eve-
exposed and unexposed groups. ryday exposures, and considerable confusion
Experimental studies aim to fulfill the Latin over the rational ranking of public health priori-
dictum ceteris paribus (other things being equal). ties (Taubes, 1995). The problem arises because
However, in humans, the optimal conditions that epidemiologic studies must confront not only the
completely eliminate confounding and bias are vagaries of chance but also the problems of sys-
difficult to create even in randomized controlled tematic errors that undermine their validity.
trials. Moreover, as already indicated, there is no
way to fully control the inherently unpredicta- Confounding
ble role of chance, except by the use of very large Confounding is the systematic error generated
numbers of study subjects—​an unrealistic objec- when another factor that causes the disease
tive in many studies. under study, or is otherwise related to it, is also
The randomized controlled trial, with its independently related to the exposure under
methodological advantages, dominates exper- investigation (Fig. 6-​4A). Thus, if one wishes
imental research in laboratory animals. In to examine whether hepatitis C virus (HCV)
humans, however, the undertaking of experi- causes liver cancer, hepatitis B virus (HBV)
ments faces serious obstacles, the most important would be a likely confounder. Confounding
of which are ethical. It is obviously not acceptable arises because HBV causes liver cancer and car-
to expose humans intentionally to a potentially riers of HBV are more likely to also be carriers
carcinogenic agent in order to ascertain cancer of HCV (because these two viruses are largely
causation. For this reason, most randomized transmitted by the same routes). Hence, if the
controlled trials in humans have been performed confounding influence of HBV is not accounted
to evaluate treatment effectiveness and occasion- for in the design (by limiting the study to HBV-​
ally to determine the preventive potential of vac- negative subjects) or in analyses of the data,
cines, vitamins, or other supplements. In most then the strength of the association between
instances, research on disease etiology has to rely HBC and liver cancer would be overestimated
either on animal models—​with inherently dubi- (Fig. 6-​4A).
ous assumptions about interspecies similarities A more trivial example is the strong associ-
and exposure dose extrapolations—​or on epide- ation between carrying matches or a cigarette
miologic studies with an observational design. lighter and developing lung cancer. Obviously,
Epidemiologic studies have indeed gener- neither matches nor lighters cause lung cancer
ated most of what is currently known about the and their association to the disease is due entirely
etiology of human diseases in general, and can- to confounding by cigarette smoking. The con-
cer in particular. At the same time, however, founding factor, cigarette smoking, is the true

A
Exposure increases risk Outcome
(e.g., HCV) (e.g., liver cancer)

+ increases risk

Confounder
(e.g., HBV)

FIGURE 6-​4A 
Infection with hepatitis C virus (HCV), a cause of liver cancer, is (positively) confounded by hepatitis B
virus (HBV) infection, another cause of liver cancer. If this confounding is disregarded, the strength of the association
between HCV and liver cancer will be overestimated. 
124

124 Textbook of Cancer Epidemiology

Exposure no causal association Outcome

(e.g., carrying matches) (e.g., lung cancer)

+ increases risk

Confounder
(e.g., smoking)

FIGURE 6-​4B  An
association between carrying matches and lung cancer would arise spuriously due to confounding
by smoking—​the major cause of lung cancer—​unless this confounding is accounted for in the design or the analysis. 

cause of lung cancer and the dependence of ciga- In case-​control studies, the ascertainment of
rette lighting on matches or lighters generates the exposure occurs after the occurrence of disease.
confounded, entirely spurious association of the Therefore, this study design is particularly sub-
latter two factors with the disease (Fig. 6-​4B). ject to information bias. In particular, cases may
There are several ways to deal with confound- be likely to remember their exposures differently
ing, some simple, others more complicated. They than controls, a form of information bias called
all assume that two conditions are satisfied: one, recall bias. For example, a reasonable concern is
that all the confounders have been identified or that cases, or their relatives, are inclined to rumi-
at least suspected, and two, that the identified nate about the disease and identify a particular
or suspected confounders can be adequately exposure as the causative agent, either for con-
conceptualized and accurately measured. When scious or subconscious reasons. Cases may also
the study is fairly large, it is always possible to try harder than controls to recall relatives with
evaluate all suspected confounders in the anal- the disease of interest, leading to a biased esti-
ysis. However, the ability to conceptualize and mate of the effect of the family history (Chang
accurately measure all of them is frequently et al, 2006).
beyond the control of any investigator. The result A well thought-​out protocol, standardized
is what has been termed “residual confound- procedures, and built-​in quality control meas-
ing,” confounding that is left unaccounted for ures can reduce bias and allow some quantifica-
(MacMahon & Trichopoulos, 1996; Rothman & tion of its potential impact. However, complete
Greenland, 1998). assurance that bias has been eliminated can
never be achieved. In addition, the reliance of
Bias case-​ control studies on a control series that
Compounding the problems of epidemiologic simultaneously has to meet criteria of compli-
studies is that the data are almost never of opti- ance, comparability to the case series, statisti-
mal quality. Data collection relies on the recol- cal efficiency, and general practicality makes
lection of exposures and their accurate reporting them susceptible to selection bias of unpredict-
by study participants, laboratory procedures, or able direction and magnitude. Such bias arises
existing records. These sources are rarely perfect. when eligible controls are not representative
For example, studies on diet rely on individuals’ of the population, or more strictly the person-​
imperfect recall on how frequently they eat spe- time, that gave rise to the cases (Wacholder et al,
cific foods, or on serum markers of nutrients that 1992a; Wacholder et al, 1992b; Wacholder et al,
are far from perfect indicators of long-​term con- 1992c).
sumption. Such misclassification, or information Assume, in the same example above that con-
bias, can influence the relative risk in any direc- trols refuse to participate more often if they are
tion and, thus, entails exaggeration, underesti- smokers than if they are nonsmokers. We would
mation, or even reversal of the true associations. then underestimate smoking in the control
125

Concepts in Cancer Epidemiology and Etiology 125

group and thereby overestimate both the differ- Although the absolute measures are easily
ence between cases and controls and the excess interpreted, more common are effect measures
risk. Hospital controls, neighborhood controls, that are taken as ratios and collectively known as
and controls enrolled through searches of tele- the relative risk. This term includes the risk ratio,
phone lists have their own problems, and these rate ratio, odds ratio, standardized mortality
have been extensively discussed (MacMahon & ratio, and standardized incidence ratio. The risk
Trichopoulos, 1996). ratio is simply the cumulative incidence of dis-
In contrast to selection and information ease among the exposed, divided by the cumu-
biases, issues of chance and confounding are lative incidence among the unexposed. The rate
equally relevant to cohort and case-​ control ratio is a ratio of the rates of disease among the
investigations (Hennekens & Buring, 1987; exposed and unexposed. The odds ratio is the
MacMahon & Trichopoulos, 1996; Rothman & odds of disease among the exposed divided by
Greenland, 1998). the odds of disease among the unexposed. Lastly,
the standardized mortality ratio or standardized
A N A LY S I S O F incidence ratio is a ratio of the observed num-
EPIDEMIOLOGIC STUDIES ber of deaths or cases in a cohort, divided by the
expected number of deaths or cases in the general
Effect Measures population, usually stratified by age and gender.
The underlying goal of epidemiology is to deter- A relative risk value of 1.0 implies that the
mine the magnitude of change in disease fre- exposure under study does not affect the inci-
quency caused by an exposure. How do we dence of the disease under consideration. Values
accomplish this? We could measure the cumu- below and above 1.0 indicate a negative (inverse)
lative incidence or incidence rate among those and a positive association, respectively. For exam-
exposed to a factor. For example, we could ple, a relative risk of 0.5 implies that the disease
observe that the incidence rate of breast cancer occurs only half as frequently among exposed as
in a population of alcoholic women is 60/​10,000 among unexposed individuals; the studied factor
person-​years. This information provides an esti- appears to be protective. In contrast, if the rela-
mate of the overall disease burden in this study tive risk is 1.5, then the occurrence (usually the
base. However, we do not know how many cases incidence) is 50% higher among exposed than
would have arisen in the study base if all the among unexposed individuals.
women in this population had not been alcohol- Studies based on follow-​up of closed cohorts
ics. In epidemiology, the unexposed group stands may be analyzed by using either cumulative inci-
in for the person-​time experience of the exposed dence (risk) measures or by counting person-​
group had it not been exposed. Thus, we need time and calculating incidence rate measures.
to harvest information from both exposed and Analyses based on cumulative incidence meas-
unexposed person-​time. ures are only useful under certain conditions,
There are several ways through which an such as no loss to follow-​up, no competing risks,
association, or lack thereof, is assessed. Consider and unchanged exposure status throughout
a population of women exposed to a high satu- follow-​up. In addition, study subjects should be
rated fat diet and a group exposed to low satu- followed for the same period of time. Whether
rated fat diets who are followed for 5 years to see or not these conditions are met, it is always valid
if they develop breast cancer. The absolute effect to conduct analyses based on person-​time, using
of the high-​fat diet would be the difference in the incidence rate measures.
cumulative incidence between the two groups,
or the difference in the incidence rates. Since the Interaction
experience of the low saturated fat group should The term “interaction” has been used to describe
represent what would have happened to the high different biological and statistical concepts.
saturated fat group if they had not eaten the high Indeed, even in the epidemiologic literature,
saturated fat and if the two groups are equiva- statements about interaction are often ambigu-
lent with respect to other breast cancer risk fac- ous and inadequately specified. From a biolog-
tors, the difference in risks or rates represents the ical point of view, component causes within the
excess risk or rate. These absolute effect measures same sufficient cause may be thought of as inter-
are called the “risk difference” and “rate differ- acting (Fig. 6-​1). In other words, the exposures
ence,” respectively. act synergistically to produce disease, since in
126

126 Textbook of Cancer Epidemiology

TABLE 6-​2   DEFINITIONS OF INTERACTION. RELATIVE RISKS OF DEVELOPING A

CERTAIN DISEASE AMONG SUBJECTS EXPOSED (+) OR NOT EXPOSED (–) TO ONE
OR B OTH FACTORS DENOTED A AND B. SUBJECTS EXPOSED TO NEITHER OF THESE
FACTORS COMPRISE THE REFERENCE CATEGORY AND THEIR RELATIVE RISK IS BY
DEFINITION 1.0

TABLE 6-​2 A  STATISTICAL INTERACTION ON THE ADDITIVE SCALE WITH EX AMPLES

OF SUBADDITIVE AND SUPERADDITIVE FACTORS
Factor B

Factor A –​ +

–​ 1.0 3.0
(reference)
+ 2.0 Subadditive
4.0 6.0 Expected under additive effects assumption
8.0 Superadditive

TABLE 6-​2 B   STATISTICAL INTERACTION ON THE MULTIPLICATIVE SCALE WITH

EX AMPLES OF SUBMULTIPLICATIVE AND SUPERMULTIPLICATIVE FACTORS

Factor B

Factor A –​ +

–​ 1.0  3.0
(reference)
+  8.0 Submultiplicative
4.0 12.0 Expected under multiplicative effects assumption
16.0 Supermultiplicative

TABLE 6-​2 C   EFFECTS ON LUNG CANCER RISK OF SMOKING, ASBESTOS

AND BOTH FACTORS

Asbestos

Smoking –​ +

–​ 1.0 5.2
(reference)
+ 10.9 53.2

Source: Hammond et al, 1979.

the absence of one factor, disease will not occur expectation of the joint effect of factors A  and
by that mechanism. From an epidemiologic B can be assessed in either an additive or a
point of view, interaction is frequently charac- multiplicative way.
terized as effect-​modification:  that is, a factor We can use the example in Table  6-​2A–​C to
A and factor B alone have a certain relationship illustrate how interaction is assessed. When a
with a disease, but together the factors have an multiplicative scale is assumed, there is statis-
effect different than that expected on the basis tical interaction if the relative risk among those
of the magnitude of their individual effects. The exposed to both factors A  and B (that is, RRAB)
127

Concepts in Cancer Epidemiology and Etiology 127

is different than the product of the two individ- There are not any clear cut guidelines on
ual relative risks (that is, RRA * RRB). When an whether to assess interaction in the additive or
additive scale is assumed, there is interaction if multiplicative setting for the various disease out-
the RRAB is different than [RRA + RRB −​ 1]. In this comes examined in epidemiology, although both
example, the expected relative risk for someone approaches are used (Brennan, 1999).
with both exposures is 6.0 [6.0 = (4.0 + 3.0 −​ 1)]
under the additive effect assumption (Table   6-​ Meta-​Analysis
2A), whereas it is 12.0 [12.0 = 4.0 * 3.0] under the Random variation per se in epidemiologic stud-
multiplicative effect assumption (Table 6-​2B). ies is not an insurmountable problem. Larger
Hence, interaction between two exposures is studies and eventually quantitative summary
present when the relative risk is significantly dif- analyses are increasingly used. Such systematic
ferent from what is expected according to a speci- statistical evaluations of results of several inde-
fied scale. Thus, for those with both exposures, pendent investigations can effectively address
we would have interaction on the additive scale genuine chance-​related concerns. Quantitative
if the relative risk is significantly different from summary analyses have been termed meta-​
6.0 (Table 6-​2A), and on the multiplicative scale if analyses and pooled analyses. There is no com-
the relative risk is significantly different from 12.0 pletely accepted distinction between the two
(Table 6-​2B). If the relative risk following expo- terms, although meta-​analysis is used more fre-
sure to both factors compared to having neither quently when published results are combined. By
is greater than the sum (minus the reference risk contrast, in pooled analysis primary individual-​
of 1, which should not be counted twice) or prod- level data from different studies may be made
uct of the individual risks, we call this interaction available to an investigator who undertakes the
superadditive or supermultiplicative, respectively. task of combining them. This facilitates the use
If the relative risk is significantly lower, we refer of uniform exposure categories and statistical
to this as either subadditive or submultiplicative. analyses across studies and may permit analy-
We can illustrate the concept of interac- ses that were not in the original publications.
tion using data from an epidemiological study For instance, analyses of effect modification for
of asbestos, smoking, and lung cancer risk. The which each initial study may have been too small
source population for the data shown in Table to be informative.
6-​2C is a cohort of insulation workers from the Meta-​ analyses and pooled analyses have
United States and Canada (Hammond et  al, been widely and effectively used for random-
1979). The exposed person-​time was the expe- ized controlled trials and intervention studies,
rience of over 12,000 male workers with at least because in properly undertaken investigations
20  years of asbestos exposure. The comparison of this nature confounding and bias are non-
person-​time came externally from the experience issues (Sacks et  al, 1987). For observational
of more than 73,000 men of similar social class. epidemiologic studies, however, the role of
Compared to men who had neither exposure, meta-​ analysis is not universally accepted
those who were smokers but were not exposed to (Shapiro, 1994; Feinstein, 1995). Some inves-
asbestos occupationally had a relative risk of 10.9; tigators are concerned that no statistical sum-
the relative risk of those exposed to asbestos but marization can effectively address problems
who were not smokers was 5.2. For those exposed generated by residual confounding, unidenti-
to both asbestos and smoking, the relative risk fied bias, and the way investigators choose to
of lung cancer was 53.2 compared to those with present their results (legitimately, but occa-
neither factor. In this example, there appears to sionally selectively or arbitrarily). Nevertheless,
be interaction on the additive scale, since the meta-​ analyses have provided important,
RRsmoker and asbestos  =  53.2 is substantially higher widely accepted, data, even when derived from
than the expected relative risk of 15.1 under the observational data.
additive model [RRsmoker + RRasbestos –​1 = 10.9 +
5.2 –​ 1]. We do not, however, observe interaction CAUSAL INFERENCE
on the multiplicative scale, since the relative risk IN EPIDEMIOLOGY
for both smoking and asbestos (53.2) does not
represent a significant departure from what is General Principles
expected under the multiplicative effect assump- Regulatory agencies and policy makers may rec-
tion (56.7 = RRsmoker * RRasbestos= 10.9 * 5.2). ommend standards, set limits, or authorize action
128

128 Textbook of Cancer Epidemiology

even when the scientific evidence is weak. These establish this. Nor do meta-​ analyses establish
decisions serve public health objectives by intro- causality. These techniques essentially address
ducing a wide safety margin, but they should not the issue of chance and provide no guarantee that
be confused with the establishment of causation a particular bias, unrecognized confounding, or
based on scientific considerations alone. selective reporting have not operated in the con-
When results of an observational epide- stituent studies. It is at this stage that both bio-
miologic study designed to address a specific logic and epidemiologic considerations should
hypothesis are striking, the study is large, and be taken into account in interpreting the results
there is no evidence of overt confounding or of empirical studies.
major biases, it is legitimate to attempt etiologic Criteria for inferring causation from epi-
inferences. In contrast, interpretation becomes demiologic investigations have been proposed,
problematic when a weak association turns out over the years, by several authors, including
to be statistically significant, for example, in MacMahon, Pugh, and Ipsen (1960), the USA
a large but imperfect data set. Although that Surgeon General (US Department of Health,
association could reflect a weak—​but genuine—​ 1964), Sir Austin Bradford Hill (Hill, 1965), the
causal association, it might also be the result of IARC (1987), and others. In spite of differences
residual confounding, subtle unidentifiable bias, in emphasis, a similar set of principles has been
or chance, perhaps following a multiple testing invoked by most authors. Sir Austin Bradford
process. Hill (1965) advocated the nine widely used crite-
Repeated demonstration of an association of ria listed in Table 6-​3, to distinguish causal from
similar direction and magnitude in several stud- noncausal associations.
ies, undertaken by different investigators in dif- The Hill criteria, although sensible and use-
ferent population groups, increases confidence ful, do not separately address the inherently dif-
in a genuine causal basis but cannot conclusively ferent issues that are posed by the results of a

TABLE 6-​3   THE HILL CRITERIA FOR INFERRING CAUSATION

Criteria Definition

Strength A strong association is more likely to be causal. The measure of strength of an association
is the relative risk and not statistical significance.
Consistency An association is more likely to be causal when it is observed in different population
groups.
Specificity When an exposure is associated with a specific outcome only (for example, a cancer site or
even better a particular histological type of this cancer), then it is more likely to be causal.
There are exceptions, however, for example, smoking causing several forms of cancer.
Temporality Not only should a cause precede the outcome (disease) but also the timing of the exposure
should be compatible with the latency period (in noninfectious diseases) or the incuba-
tion period (in infectious diseases).
Gradient This criterion refers to the presence of an exposure–​response relationship. If the fre-
quency or intensity of the outcome increases when an exposure is more intense or lasts
longer, then it is more likely that the association is causal.
Plausibility An association is more likely to be causal when it is biologically plausible.
Coherence A cause and effect interpretation of an association should not conflict with what is known
about the natural history and biology of the disease or its distribution in time and place.
Experimental If experimental evidence exists, then the association is more likely to be causal. Such
evidence ­evidence, however, is seldom available in human populations.
Analogy The existence of an analogy (for example, if a drug causes birth defects, then another
drug could also have the same effect) could strengthen the belief that an association
is causal.

Source: Hill,1965.
129

Concepts in Cancer Epidemiology and Etiology 129

single study, the results of several studies, and The Process of Causal Inference
the likelihood of causation in a certain indi- Criteria for causality can be invoked, explicitly
vidual. In reality, the perceived likelihood of a or implicitly, in evaluating the results of a single
causal association between a particular expo- epidemiologic study, although, in this instance,
sure and a specific disease moves forward or a firm conclusion is all but impossible. In the
backward in a continuous spectrum as research approach introduced by Cole (1997), this sit-
results accumulate. The evidence for causality is uation is denoted as single study level, or level
declared as sufficient when a particular thresh- I. Criteria for causality are more frequently used
old has been reached, but on occasion requires for the assessment of evidence accumulated from
reevaluation in the light of subsequent evidence several epidemiologic studies and other biomed-
(Cole, 1997). ical investigations. At this stage, the intellectual
process is inductive, moving from the specifics to
The IARC Classification generalization (several studies level, or level II).
The International Agency for Research on Cancer Finally, when causation has been established at
(IARC) evaluates the risk of specific agents to level II, then, and only then, can the cause of the
determine whether they are carcinogenic in disease in a particular individual be considered
humans. In order to come to a conclusion, the (specific person level, or level III). At this level,
IARC has implemented its own set of criteria for the intellectual process is deductive, moving
evaluating the carcinogenicity of agents. After from the general concept of disease causation to
considering all the evidence, the IARC working the examination of what might have caused dis-
group assigns the agent to one of five categories, ease in a particular individual.
summarized in Table 6-​4. Group 1 indicates that
there is sufficient evidence to conclude that the The Individual Study (Level I)
agent is carcinogenic to humans. A label of group Causality can never be inferred on the basis of
2A means that there are insufficient human data, a single epidemiologic study, but the likelihood
but there is strong evidence that the agent is car- that an observed association is causal is strength-
cinogenic in animal models. Agents for which ened when several of the following criteria are
there is limited evidence in humans and insuf- met:  (1) minimal confounding; (2)  minimal
ficient evidence in experimental animals are bias; (3)  limited chance variation; (4)  relatively
assigned to group 2B. Group 3 is used when there strong association; (5)  monotonic exposure–​
is inadequate human and animal data to come disease association, otherwise referred to as
to a conclusion. Group 4 indicates that the agent exposure-​response or dose-​response association;
is most likely not a carcinogen in humans based (6)  internal consistency, exemplified by similar-
on adequate evidence suggesting that it is not a ity of exposure-​response patterns among various
carcinogen in both animal models and human subgroups of study subjects; (7) compatibility of
studies. the temporal sequence of exposure and outcome
with the known or presumed latency of the dis-
ease; and (8) biologic plausibility, that is, a causal
TABLE 6-​4   INTERNATIONAL AGENCY link between the exposure and the disease should
FOR RESEARCH ON CANCER (IARC) be, at a minimum, biologically conceivable (it
CLASSIFICATION OF CARCINOGENICIT Y should not contradict physical theory or biolog-
OF AGENTS, MIXTURES OR PROCESSES ical principles).
Group 1 The agent is carcinogenic to humans The General Case
Group 2A The agent is probably carcinogenic to (Several Studies, Level II)
humans Establishment of the etiologic role of a particu-
lar exposure on the occurrence of a disease ide-
Group 2B The agent is possibly carcinogenic to
ally requires strong epidemiologic evidence, an
humans
appropriate and reproducible animal model, and
Group 3 The agent is not classifiable in terms documentation at the molecular or cellular level
of its carcinogenicity of the morphological or functional pathogenetic
Group 4 The agent is not carcinogenic to process. Sometimes, an intended or unintended
humans change or “natural experiment” greatly facili-
tates etiologic inference—​this happens when, for
130

130 Textbook of Cancer Epidemiology

example, an occupational group is exposed to was “more likely than not” caused by the speci-
high levels of compounds rarely encountered in fied exposure.
other settings, a religious group avoids an expo- For this conclusion to be drawn, all the fol-
sure that is otherwise widespread, or a vaccine lowing criteria must be met (Cole, 1997): (1) The
that creates herd immunity against a particular exposure under consideration, as an entity, must
virus turns out to reduce the incidence of a cer- be an established cause of the disease under con-
tain form of cancer. sideration, as an entity (level II); (2) The relevant
These conditions, however, are rarely col- exposure of the particular individual must have
lectively satisfied. Instead investigators have to properties comparable (in terms of intensity,
be guided by the best available biomedical evi- duration, associated latency, etc.) to those that
dence in order to interpret correctly epidemio- have been shown to cause the disease under con-
logic data from several studies. The following sideration; (3) The disease of the specified person
criteria need to be considered: (1) consistency, must be identical to, or within the symptomato-
that is similarity (lack of heterogeneity) of logical spectrum of the disease that, as an entity,
results obtained by different investigators using has been etiologically linked to the exposure;
different study designs in different populations; (4)  The patient must not have been exposed to
(2) overwhelming biomedical evidence for weak another established or likely cause of this disease.
associations, whereas for strong associations If the patient has been exposed to both the fac-
reliance on powerful biomedical knowledge tor under consideration (for example, smoking)
is less critical; (3)  compatibility of exposure-​ and to another causal factor (for example, asbes-
response patterns across different studies tos), individual attribution becomes a function
exploring the exposure-​disease association in of several relative risks, all versus the completely
different exposure ranges; (4) coherence, which unexposed:  (a) relative risk of those who only
requires results from analytic epidemiologic had the exposure under consideration, (b) rela-
studies to be compatible with ecologic patterns tive risk of those who had only been exposed to
and time trends, such as the increasing inci- the other causal factor(s), and, (c) relative risk of
dence of lung cancer over time, following the those who have had a combination of these expo-
increasing use of tobacco products by the pop- sures; (5) The relative risk should be reasonably
ulation; (5)  specificity, which exists when one elevated (e.g., 2 or more).
type of disease is consistently linked with one The last criterion stems from the fact that
type of exposure rather than several exposures the relative risk comprises a baseline component
all being associated with a certain disease, or equal to 1, which characterizes the unexposed,
one type of exposure being associated with sev- plus another component that applies only to the
eral diseases; and (6) biological analogy, which exposed. When the relative risk is higher than
exists when a similar exposure has been shown 1 but less than 2 the individual who has been
to cause a similar disease in another species or exposed and has developed the disease is more
a different form of the disease in humans. For likely than not to have developed the disease for
example, viruses have been shown to cause leu- reasons not entirely due to the exposure. For
kemia in several animal species and at least one instance, if the risk of a light-​smoking 55-​year-​
rare form of leukemia in humans. old man to suffer a first heart attack in the next
None of these criteria can be considered 5 years is 6%, and that of a same-​age nonsmok-
as absolutely necessary for causal inference, a ing man is 4% (relative risk 1.5), only 33% of the
sine qua non. But the evidence for causality is smoker’s risk (that is, one-​third of the total 6%)
strengthened when most of them are met. can be attributed to his smoking. When the rel-
ative risk is higher than 2, a particular individ-
Disease in a Specific Person (Level III) ual who has been exposed and has developed the
Causality can be conclusively established between disease under consideration is more likely than
a particular exposure as an entity and a particular not to have developed the disease because of the
disease as an entity. In contrast, it is not possi- exposure.
ble to establish such a link conclusively between
an exposure and a particular disease of a given CONCLUSION
individual, for example, smoking in a patient Manipulation of exposures in humans, many of
with lung cancer. It is possible, however, to infer which may be harmful, is frequently unfeasible,
deductively that the specific individual’s illness unethical, or both. Therefore, epidemiologists
13

Concepts in Cancer Epidemiology and Etiology 131

have to base their inferences on experiments that the passage of a certain time period (latency, or
humans subject themselves to intentionally, nat- induction period).
urally, or even unconsciously. The study of risk Closed cohort A closed cohort comprises a set of
for lung cancer among smokers compared with individuals who are followed for a defined period
nonsmokers is one classic example of a natural of time. After becoming a member of the cohort,
experiment. an individual remains in the cohort until the end
Because human life is characterized by a of the study, or development of the outcome.
myriad of complex, often interrelated behav-
Competing risks The risk of death from a cer-
iors and exposures -​ranging from genetic
tain disease competes with the risk of death
traits and features of the intrauterine environ-
from another disease by affecting time at risk.
ment to growth rate, physical activity, sexual
Competing risks generally bias risk ratios, but
practices, use of tobacco, alcohol and pharma-
not rate ratios, since person-​time allows for dif-
ceutical compounds, dietary intake, exposure
ferent follow-​up time.
to infections, environmental pollutants and
occupational hazards, and so on -​epidemio- Component cause An exposure that acts in con-
logic investigation is difficult and challenging. cert with other factors (component causes) to
Given this complexity, it is not surprising that produce disease. None of these factors is suffi-
from time to time epidemiologic studies gen- cient in itself to cause disease.
erate results that appear confusing, biologi- Confidence interval A statistical measure that
cally absurd, or contradictory. However, it is provides a range of possible values that include
reassuring that a wealth of new knowledge has the true measure of association with a particu-
been generated by epidemiologic studies over lar degree of certainty. For example, a 95% confi-
the last few decades. This knowledge now lays dence interval provides a range of values that will
the scientific ground for primary prevention of include the true value 95% of the time.
many major cancers and other chronic diseases Confounding A systematic error generated
among humans globally. when another factor that causes the disease
A detailed study of epidemiologic meth- under study, or is otherwise related with it, is
odology in any textbook (Hennekens & also related to the exposure under investiga-
Buring, 1987; Miettinen, 1985; Walker, 1991; tion, without being in the pathway that links
MacMahon & Trichopoulos, 1996; Rothman exposure under investigation with the disease
& Geenland, 1998; Rothman, 2002) can be fas- under study.
cinating and indeed necessary for those who
Ecologic study The study of exposure and the
want to pursue their own research. However,
disease at the population level, rather than at the
for the reader of this textbook, the general
individual level.
concepts introduced in this chapter should
provide a sufficient basis. We have tried to Epidemiology The nonexperimental investiga-
convey that the sometimes esoteric theory of tion of determinants of human disease.
modern epidemiology can be condensed to a Experimental study See randomized controlled
few central issues, namely (1) how to quantify trial.
and understand the impact of chance; (2) how Information bias A  random, or nonrandom,
to best harvest information on exposures and misclassification of information on either the
outcomes from a source population by using exposure, outcome, or confounding variables
a cohort design, a case-​control design, or vari- that leads to a biased estimation of the true effect.
ants thereof; (3)  how to achieve valid results
by minimizing the impact of confounding and Loss to follow-​up The inability to follow beyond
bias; and, (4) how to address the central issue a certain point in time and thus ascertain the ulti-
of causality in a structured way. mate fate of individuals in a cohort study.
Necessary cause A factor or exposure that is
G L O S S A RY essential in the etiology of the disease and with-
Cause A factor is a cause of a certain disease out which the disease cannot occur. For example,
when alterations in the frequency or intensity the human immunodeficiency virus is a neces-
of this factor—​without concomitant alterations sary cause of acquired immunodeficiency syn-
in other factors—​ are followed by changes in drome, although other factors may be involved
the frequency of occurrence of the disease, after in order for the disease to occur.
132

132 Textbook of Cancer Epidemiology

Nonexperimental study See observational study. history of cancer in a large case-​control study of
Observational study A study in which the inves- lymphoma. J Natl Cancer Inst 2006;98:61–​8.
Clemmesen J, Nielsen A. Comparison of age-​adjusted
tigator cannot control the circumstances of the
cancer incidence rates in Denmark and the United
exposure.
States. J Natl Cancer Inst 1957;19:989–​98.
Odds ratio A relative measure of association, Cole P. Causality in epidemiology, health pol-
which is calculated as the ratio of the odds of dis- icy and law. Environmental Law Reporter
ease among the exposed divided by the odds of 1997;27:10279–​85.
disease among the unexposed. Cordell HJ, Clayton DG. Genetic epidemiology 3: genetic
Open cohort A cohort of individuals whose association studies. Lancet 2005;366:1121–​31.
membership changes over time, with people Doll R, Hill AB. Smoking and lung cancer:  prelimi-
entering or exiting based on defining criteria. nary report. Brit Med J 1950;2:739–​48.
Doll R, Hill AB. Lung cancer and other causes of
Person-​time The sum of all time spent by each death in relation to smoking. Brit Med J 1956;
study participant at risk for a disease. 2:1071–​81.
P-​value A value that indicates the likelihood of Doll R, Peto R, Boreham J, Sutherland I. Mortality
observing an association as extreme as, or more from cancer in relation to smoking:  50  years
extreme than, the one found between a particular observations on British doctors. Br J Cancer
exposure and a certain disease, if there were in 2005;92:426–​9.
fact no association. Feinstein AR. Meta-​ analysis:  statistical alchemy
for the 21st century. J Clinical Epidemiology
Randomized controlled trial An experimental
1995;48:71–​9.
study design in which the researcher randomly Greenland S, Robins J. Invited commentary: ecologic
allocates subjects to groups that will be subjected studies—​biases, misconceptions, and counterex-
or not to a particular exposure. amples. Am J Epidemiol 1994;139:747–​60.
Recall bias A misclassification of an exposure, Hammond EC, Seikoff IJ, Seidman H. Asbestos expo-
common in case-​control studies, that occurs when sure, cigarette smoking and death rates. Ann NY
subjects with the disease remember or report their Acad Sci 1979;330:473–​90.
exposures differently than those without disease. Hansson LE, Nyren O, Hsing AW, Bergstrom R,
Josefsson S, Chow WH, et al. The risk of stomach
Relative risk A term that collectively describes
cancer in patients with gastric or duodenal ulcer
the various relative measures of association, that
disease. New Engl J Med 1996;335: 242–​9.
is, the risk ratio, the rate ratio, the odds ratio, and
Hennekens CH, Buring JE. Epidemiology in medi-
the standardized incidence or mortality ratio. cine. Boston: Little, Brown, 1987.
Selection bias A systematic error that results Hennekens CH, Speizer FE, Lipnick RJ, Rosner, Bain C,
from the process of selecting participants for Belanger C et al. A case–​control study of oral con-
the study or on account of factors that influence traceptive use and breast cancer. J Natl Cancer Inst
participation in the study. Selection bias occurs 1984;72:39–​42.
when the relationship between the exposure and Hill AB. The environment and disease:  asso-
the disease is different for those in the study than ciation or causation? Proc Roy Soc Med
for those not in the study. 1965;58:295–​300.
Hunter DJ, Morris JS, Stampfer MJ, Colditz GA,
Study base The person-​time of a group of indi-
Speizer FE, Willet WC. A prospective study of
viduals at risk for a disease from which an inves-
selenium status and breast cancer risk. JAMA
tigator aims to harvest information about disease 1990;264:1128–​31.
occurrence. International Agency for Research on Cancer. IARC
Sufficient cause A minimal set of factors or Monographs on the Evaluation of Carcinogenic
exposures that inevitably produce the disease Risks to Humans, Supplement 7, Overall Evaluations
after a certain period of time. of Carcinogenicity:  An Updating of IARC
Monographs, Volumes 1 to 42. Lyon: IARC, 1987.
REFERENCES MacMahon B. Epidemiological evidence on the nature
Brennan P. Chapter 12: Design and analysis issues in of Hodgkins disease. Cancer 1957;10:1045–​54.
case–​control studies addressing genetic suscepti- MacMahon B. Strengths and limitations of epide-
bility. IARC Sci Publ 1999;148:123–​32. miology. In:  The National Research Council in
Chang ET, Smedby KE, Hjalgrim H, Glimelius B, 1979:  current issues and studies. Washington,
Adami HO. Reliability of self-​ reported family DC: National Academy of Sciences, 1979: 91–​104.
13

Concepts in Cancer Epidemiology and Etiology 133

MacMahon B, Pugh TF, Ipsen J. Epidemiologic Wacholder S, McLaughlin JK, Silverman DT,
Methods. Boston: Little, Brown, 1960. Mandel JS. Selection of controls in case–​
MacMahon B, Trichopoulos D. Epidemiology:  prin- control studies. I.  Principles. Am J Epidemiol
ciples and methods. Boston: Little, Brown, 1996. 1992a;135:1019–​28.
Miettinen OS. Theoretical Epidemiology:  Principles Wacholder S, Chanock S, Garcia-​ Closas M, El
of Occurrence Research in Medicine. Ghormli L, Rothman N. Assessing the probabil-
New York: Wiley, 1985. ity that a positive report is false: an approach for
Morgenstern H. Uses of ecologic analysis in epi- molecular epidemiology studies. J Natl Cancer
demiologic research. Am J Public Health Inst 2004;96:434–​2.
1982;72:1336–​44. Wacholder S, Silverman DT, McLaughlin JK, Mandel JS.
Rothman KJ. Causes. Am J Epidemiol 1976;104:587–​92. Selection of controls in case–​control studies. II. Types
Rothman KJ, Modern Epidemiology. Boston, of controls. Am J Epidemiol 1992b;135:1029–​41.
MA: Little, Brown and Company, 1986. Wacholder S, Silverman DT, McLaughlin JK,
Rothman KJ. Epidemiology:  an introduction. Mandel JS. Selection of controls in case–​control
New York, NY, Oxford University Press, 2002. studies. III. Design options. Am J Epidemiol
Rothman KJ, Greenland S. Modern Epidemiology. 2 1992c;135:1042–​50.
ed. Philadelphia, PA: Lippincott-​Raven, 1998. Walker AM. Observation and Inference:  an
Sacks HS, Berrier J, Reitman D. Meta-​ analysis of Introduction to the Methods of Epidemiology.
randomized controlled trials. N Engl J Med Newton Lower Falls, MA. Epidemiology
1987;316:450–​55. Resources Inc, 1991.
Shapiro S. Meta-​ analysis/​
Shmeta-​ analysis. Am J Weiderpass E, Adami HO, Baron JA, Magnusson C,
Epidemiol 1994;140:771–​8. Bergstrom R, Lindgren A et al. Risk of endome-
Susser M. What is a cause and how do we know one? trial cancer following estrogen replacement with
A  grammar for pragmatic epidemiology. Am J and without progestins. J Natl Cancer Inst 1999;
Epidemiol 1991;133:635–​48. 91:1131–​7.
Taubes G. Epidemiology faces its limits. Science Wynder EL, Graham EA. Tobacco smoking as
1995;269:164–​69. a possible etiologic factor in bronchiogenic
Teare DM, Barrett JH. Genetic epidemiology 2: genetic carcinoma—​a study of 684 proved cases. JAMA
linkage studies. Lancet 2005;366:1036–​44. 1950;143:329–​36.
US Department of Health, Education and Welfare. Zhang SM, Hankinson SE, Hunter DJ, Giovannucci
Smoking and Health. Report of the Advisory EL, Colditz GA, Willett WC. Folate intake and risk
Committee to the Surgeon General of the Public of breast cancer characterized by hormone recep-
Health Service. Publication 1103. Washington, tor status. Cancer Epidemiol Biomarkers Prev
DC: US Government Printing Office, 1964. 2005;14:2004–​8.
134
135

PA R T   I I
 
136
 7
Oral and Pharyngeal Cancer
JENNIFER R. RIDER, PAUL BRENNAN, AND PAGONA LAG/OU

O ver the past decade, our understanding of

oral and pharyngeal cancer has changed dra­
matically. In 2007, the World Health Organization
play a causal role at many subsites, although the
strength of the association may differ.
Symptoms Early symptoms vary according to the
determined there was sufficient evidence to subsite within the oral cavity and pharynx, but
include human papilloma virus (HPV) type 16 as a sore throat or difficulty swallowing are typical.
a cause of oropharyngeal and oral cancer (IARC, In many cases, metastases to lymph nodes in the
2007). These new insights are due in large part neck bring the patient to medical attention.
to epidemiologic studies that elucidated distinct
Diagnosis Ca cers at these subsites are uniquely
etiologic pathways and the divergence of disease
accessible for direct inspection that allows a
outcomes according to the tumor's HPV status.
biopsy for a definite histopathologic diagno­
Appreciation of the HPV-positive and HPV­
sis. Comp,uted tomography (CT), magnetic res­
negative phenotypes has permitted a more accu­
onance imaging (MRI), or positron emission
rate assessment of trends in oral and pharyngeal
t0mograpliy (PET) are generally used for addi­
cancer rates over time, and more tailored strate­
tional staging and planning of treatment.
gies for disease management. It has also gener­
ated more promise for disease prevention. Treatment Primary treatment varies with the
Methodologically, these new insights have anatomic subsite, stage of disease, and human
produced new challenges for epidemiologic papilloma virus (HPV) status. For most early
research on oral and pharyngeal carnser. The cancers of the oral cavity, surgical resection is the
various anatomic subsites, whicn inclua.e the cornerstone of treatment. However, for certain
lip, tongue, salivary gland, mouth, oropHarynx, anatomic sites such as the tonsils, the base of the
nasopharynx, and hypopharrnx (1CD9 codes tongue, and the floor of the mouth, as well as for
140-148), are less homoge eous ith respect to all locally advanced cancers, radiotherapy is often
their descriptive epidemiology and risk factors used, either alone or combined with surgery and/
than once believed. Cons�uently, many previ­ or chemotherapy.
ous studies that focused on "oral cancer;' "head Prognosis About 60% of people with oral cav­
and neck cancer;' and "upper aerodigestive tract ity and oropharyngeal cancer survive 5 years,
cancer" that combined multiple sites are now although this is strongly dependent on stage
difficult to interpret. This chapter covers the (Howlander et al, 2014). Survival varies accord­
oral cavity and the oropharynx, which includes ing to tumor HPV status for oropharynx tumors
the base of the tongue, soft palate, tonsils, and and potentially for tumors of the oral cavity.
back and side walls of the throat, but not the Importantly, patients with HPV(+) tumors have
nasopharynx, which is described separately in a much more favorable outcome than those with
chapter 8. HPV(-) tumors, independent of stage (Fakhry
et al, 2008; Lassen et al, 2009; Ang et al, 2010;
CLINICAL SYNOPSIS Fakhry et al, 2014). Among HPV-positive oro­
Subgroups Cancers of the oral cavity and oro­ pharyngeal tumors, smoking history is strongly
pharynx, predominantly of a squamous cell type, associated with poor survival (Ang et al, 2010).
are characterized chiefly by their various ana­ Progress Survival is improving, in part because
tomic locations. While some risk factors appear a larger proportion of tumors are HPV-related.
to be uniquely linked with certain subsites, others HPV testing of oropharynx tumors is now
138

138 Textbook of Cancer Epidemiology

recommended. The lack of a precursor lesion 4.5 per 100,000 person-​years, nearly fivefold lower
and limited data on efficacy of the HPV vaccine than men in the same country, while rates among
in preventing oral HPV infection are barriers to women in Slovakia are more than seven times
primary and secondary prevention efforts. lower than among men.
Oral cavity and oropharyngeal cancer is rare
DESCRIPTIVE among men under the age of 35 in most countries
EPIDEMIOLOGY (Figs. 7-​3, 7-​4). The age-​specific rates are typical
An estimated 442,760 newly diagnosed cases of of many epithelial cancers and increase with age.
oral cavity and oropharyngeal cancer occurred Although the shape of the age incidence curves is
worldwide in 2012 (Ferlay et  al, 2013). These generally similar across countries, there are sub-
tumors account for approximately 7.9% of total stantial differences in the absolute rates at every
cancers among men and about 3.2% of total can- age (Figs. 7-​3, 7-​4).
cers among women. Mortality is quite high in Although data on the descriptive epidemiol-
high-​, middle-​, and low-​income countries, with ogy of tumors at specific subsites is more limited,
approximately 241,458 deaths occurring in 2012 combining tumors of the oropharynx and oral
(Ferlay et al, 2013). cavity cancers conceals important information
The incidence of oral cavity and oropharyngeal regarding the time trends and geographic varia-
cancer varies considerably worldwide (Figs. 7-​1, tion in tumors at specific sites.
7-​2)(Ferlay et al, 2013). Some of the highest rates In a study of health plan members in Northern
have been observed among men living in France, California, tumors at HPV-​associated anatomical
with an age-​standardized incidence of 30–​32 per sites of the head and neck increased 3.7% per year
100,000 person-​years in Somme and Calvados. between 1995 and 2010, while tumors at sites typi-
Overall age-​standardized incidence among men cally unrelated to HPV decreased by 2.4% per year
in France is 21 per 100,000 person-​years. Other during the same time period (Katzel et al, 2015).
countries with the highest age-​standardized inci- In an analysis of the IARC Cancer Incidence in
dence among men include Slovakia (23.1 per Five Continents data, oropharyngeal cancers
100,000 person-​ years) and Chennai, India (22 increased in middle-​and high-​income countries
per 100,000 person-​ years). Incidence among during the period of 1983–​2002 (Chaturvedi et al,
women is highest in Chennai, India, with age-​ 2013), particularly among adults less than 60 years
standardized incidence of nearly 11 per 100,000 of age. In the countries that experienced increases
person-​years, followed by Somme, France (9 per in incidence of oropharyngeal tumors among
100,000 person-​ years). Incidence rates in the men (United States, Australia, Canada, Japan,
same geographical area are generally much lower and Slovakia), oral cavity tumors among men
in women compared to men (Figs. 7-​1, 7-​2).  In decreased during the same time period. However,
France, age-​standardized rates among women are countries with increases in oropharyngeal tumors

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United states of America

0 0.5 1 1.5 2 2.5 3 3.5 4

ASR (world) per 100,000 person-years

FIGURE  7-​1 
Age-​
standardized (to the 2012 world population) incidence rates of lip and oral cavity cancer
among women.
Source: Ferlay et al, 2013.
139

Oral and Pharyngeal Cancer 139

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United states of America

0 2 4 6 8 10
ASR (world) per 100,000 person-years

FIGURE 7-​2  Age-​standardized (to the 2012 world population) incidence rates of lip and oral cavity cancer among men.
Source: Ferlay et al, 2013.

among women (Denmark, Estonia, France, the for oral and oropharyngeal cancer, which are
Netherlands, Poland, Slovakia, Switzerland, and reviewed in this section.
United Kingdom) also experienced simultaneous
increases in oral cavity tumors among women. Inherited Susceptibility
Of note, HPV(+) oropharyngeal cancers are Similar to other solid tumors, there appears to
expected to surpass the number of cervical can- be a familial susceptibility to oral cavity and oro-
cers by 2020, becoming the predominant HPV-​ pharynx cancer. In the ICARE study, having a
associated tumor in the United States (Chaturvedi first-​degree relative with a diagnosis of oral cav-
et al, 2011). ity and oropharynx cancer was associated with
a doubling of the risk in the index case, and the
GENETIC AND MOLECULAR risk increased with number of affected relatives
EPIDEMIOLOGY (Radoï et  al, 2013a). Similar results were found
Epidemiologic studies have contributed to our in a multicenter case-​control study in Italy and
understanding of the inherited risk factors Switzerland (Garavello et al, 2008).
New cases per 100,000 person-years

30

25

20

15

10

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 7-​3  Age-​specific incidence rates of lip and oral cavity cancer among women.
Source: Ferlay et al, 2013.
140

140 Textbook of Cancer Epidemiology

60

New cases per 100,000 person-years

50

40

30

20

10

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 7-​4  Age-​specific incidence rates of lip and oral cavity cancer among men.
Source: Ferlay et al, 2013.

High-​Penetrance Gene Mutations could represent one mechanism by which oral

Several inherited cancer syndromes are associ- and oropharyngeal cancer occurs. The most
ated with increased risk of oral and oropharyn- studied genes that have found associations with
geal cancer, including Fanconi anemia, which oral cavity and oropharyngeal cancer risk include
results from mutations in caretaker genes (Lustig the glutathione S-​ transferase (GST) family of
et al, 1995); Bloom syndrome, which results from enzymes, cytochrome P450 enzymes, and alco-
mutations in helicase genes; and xeroderma pig- hol dehydrogenases (ADH). In general, more
mentosum, resulting from mutations in the XP recent studies have not replicated initial findings,
genes in the nucleotide excision repair pathway but sample size is a frequent limitation.
(Keukens et al, 1989; Prime et al, 2001). The p53 The GST enzymes are expressed in oral tissue
gene, in its normal capacity, acts as a gatekeeper (Peters et al, 1993) and play a significant role in
by invoking cell-​ cycle arrest or cell death in the detoxification of activated metabolites of pol-
response to genetic mutations. Germline muta- ycyclic aromatic hydrocarbons (PAHs) such as
tions in p53 have been implicated in the etiology benzo(a)pyrene found in tobacco smoke, as well
of oral cancer. The Li-​Fraumeni syndrome, which as in the detoxification of ethanol and its metabo-
is caused by inherited defects in the p53 gene, lites. A pooled analysis of 42 case-​control studies
results in an excess risk of multiple malignancies on the effect of GSTM1 and T1 on head and neck
in affected family members, including oral cancer cancer (including cancer of the larynx) reported
(Prime et al, 2001). However, one analysis found an increased risk for both GSTM1 deficiency
no association between TP53 Arg72Pro variants (OR = 1.27, 95% CI 1.13–​1.42) and GSTT1 defi-
and risk of oral cavity and oropharyngeal cancer ciency (OR = 1.14, 95% CI 1.00–​1.31), with evi-
(Brunotto et al, 2013). dence of a combined effect for deficiency to both
genes (OR  =  1.99, 95% CI 1.74–​2.24) (Ye et  al,
Low-​Penetrance Polymorphisms 2004). When stratified by subsite, an increased
Early genetic epidemiologic studies of oral cav- risk among studies which included just oral can-
ity and oropharyngeal tumors focused primarily cer was still observed for GSTM1 (OR = 1.56, 95%
on polymorphisms in genes encoding enzymes CI 1.35–​1.80) although not GSTT1 (OR  =  1.16,
involved in the metabolism of tobacco, alcohol, 95% CI 0.91–​1.47). A  meta-​analysis published
and other environmental factors. Carcinogen-​ in 2010 based on 28 case-​control studies found
metabolizing enzymes are expressed in the oral a consistent association between the GSTM1 null
cavity (Peters et al, 1993; Dong et al, 1996), and genotype and oral and oropharyngeal cancer in
differences in enzyme activity in the oral cavity Asians but not in Caucasians, and no association
14

Oral and Pharyngeal Cancer 141

between GSTT1 genotype and cancer risk regard- Somatic Mutations

less of race (Zhang et al, 2011). A meta-​analysis The availability of next-​generation sequencing
including only more recent studies found sub- has permitted the confirmation of somatic muta-
stantial heterogeneity in studies of GSTM1 poly- tions already identified from earlier studies and
morphisms and tumors of the oral cavity and the identification of novel mutations (Leemans
oropharynx (Brunotto et al, 2014). et  al, 2011; Cancer Genome Atlas Network,
CYP1A1, of the CYP1 family, catabolizes 2015). The vast majority of the genetic altera-
PAHs, including benzo(a)pyrene and other tions in oral and oropharyngeal tumors can be
constituents of tobacco smoke (Guengerich & grouped into the following categories: (1) mito-
Shimada, 1991). A  meta-​ analysis of CYP1A1 genic signaling, particularly phosphoinositide
based on five studies identified a potential 3-​kinase (PIK3CA/​mTOR pathway); cell differ-
increased risk of oral cancer with the Val462 entiation (NOTCH1 pathway); cell-​cycle dereg-
allele (OR  =  1.48, 95% CI 0.77–​2.83) (Hashibe ulation resulting from inactivation of CDKNA2
et al, 2003). Associations are more consistent in (P16INK4A) and amplification of cyclin D1
Asian populations (Hernando-​Rodriguez et  al, (CCND1); and loss of TP53 or other genes result-
2012). However, two more studies found no ing in genomic instability (Iglesias-​ Bartolome
association (Amtha et  al, 2009; Cordero et  al, et al, 2013; Pickering et al, 2013).
2010). Another polymorphism in CYP1A1, the The first two whole-​exome sequencing studies
6235 T>C transition, has been extensively stud- of head and neck squamous cell cancer (includ-
ied, with conflicting findings in Asian popula- ing laryngeal tumors) both identified NOTCH1
tions and generally null findings in European mutations in 11%–​ 15% of tumors (Stransky
populations (Hernando-​Rodriguez et  al, 2012). et  al, 2011; Agrawal et  al, 2011). These results
Polymorphisms investigated in CYP1B1 have were confirmed by a report (Pickering et  al,
also yielded contradictory results (Hernando-​ 2013)  that also described the high frequency of
Rodriguez et al, 2012). tumor ploidy, copy number alterations that cor-
Ethanol is primarily oxidized to acetaldehyde, responded with changes in mRNA expression.
its carcinogenic form, by alcohol dehydrogenase Approximately three-​ quarters of the tumors
(ADH) (Bosron & Li, 1986). Although most eth- studied harbored at least 20 copy number altera-
anol metabolism occurs in the liver, ADH is also tions. The most common copy number gains
expressed in the oral cavity. The ADH1B and were on chromosome 8q and 3q and in regions
ADH1C genotypes are involved in metabolizing containing CCND1, EGFR, MYC, and TP63. This
ethanol to acetaldehyde, a recognized animal study also identified losses of chromosomes that
carcinogen (IARC, 1999). The ADH1C*1 and contain tumor suppressors, including 3p, 18q,
ADH1B*2 alleles encode for enzymes that result and 8p, as well as regional loss of CDKN2A. In
in the “fast” metabolism of ethanol. ADH1B and tumors from smokers, DNA methylation changes
ADH1C are located at only 16 kb of distance were also frequently observed.
on chromosome 4, and linkage disequilibrium The p53 gene, located on chromosome 17p,
between ADH1C*1 and ADH1B*2 has been dem- has multiple functions in controlling cellular
onstrated in several populations (Brennan et al, replication, including acting as a G1 checkpoint
2004). The ADHIB*2 allele is present in 93% of control and as a trigger of apoptosis in response
Japanese populations, but is less common (~5%–​ to cell damage. Next-​ generation sequencing
20%) in Africans and Caucasians (Malichalar-​ studies indicate that p53 mutations are the most
Mendia et  al, 2010). A  large study based on commonly occurring mutations in oral cavity
approximately 8,000 individuals of European and oropharyngeal tumors (Tabatabaeifar et  al,
ethnicity identified a strong protective associ- 2014). However, the occurrence of p53 mutations
ation for ASH1B*2 as well as an ADH7 variant appears to be associated with exposure to spe-
(Hashibe et al, 2008). These effects became more cific risk factors indicating different causal path-
apparent with increasing alcohol consumption. ways for oral cancer. Mutations have been found
In a Japanese study, ADHIB* homozygotes who to be more strongly associated with alcohol-​
consumed alcohol were more likely to develop and tobacco-​ associated oral cancer, and less
oral cavity and oropharyngeal tumors (Yokoyama often with infection with HPV (Tabatabaeifar
et al, 2001; Asakage et al, 2007). et al, 2014).
142

142 Textbook of Cancer Epidemiology

In a study of targetable pathways, the PI3K Tobacco

pathway contained the highest percentage of Although the proportion of total oral and oro-
mutated genes, with PIK3CA being the single pharyngeal cancers attributable to smoking is
most mutated gene in the pathway. In 20% of decreasing in high-​income countries, tobacco
HPV-​positive tumors, PI3K genes were the only remains an important risk factor for these
oncogenes identified (Lui et al, 2013). tumors globally. The form of tobacco used var-
Recent work has indicated that HPV-​positive ies in different parts of the world. In Europe, the
tumors are considerably more heterogeneous United States, Australia, and Japan, cigarettes,
and display higher mutation rates (Tabatabaeifar cigars, and pipes are the main types of smoked
et  al, 2014). A  targeted sequencing study tobacco, while chewing tobacco and snuff are
(Lechner et al, 2013) found that all HPV-​negative common smokeless forms. In India, Pakistan,
tumors exhibited mutations in TP53. Cyclin D1 China, and South East Asia, while cigarette con-
(CCND1) amplifications, CDKN2A/​B deletions, sumption may be high, tobacco is also chewed as
and PIK3CA mutations were also common in betel quid, paan, naswar, and nass (Gupta, 1996).
HPV-​negative tumors. In HPV-​positive tumors, Betel quid consists of the leaf of the betel vine
PIK3CA mutation or amplification, PTEN muta- wrapped around areca nut, lime, and tobacco.
tion or inactivation, and FBXW7 mutation were Paan is a mixture of areca nut, lime, spices, and
all commonly observed. FBXW7 mutation, pre- sweeteners and may or may not include tobacco.
viously reported in cell lines from other tumor Naswar is composed of tobacco and lime, while
types, was found to be mutated in 4% of tumors nass is an aqueous or oily mixture of tobacco,
in one of the original whole-​ exome sequenc- ash, and lime.
ing studies of head and neck tumors (Agrawal In ICARE, a large population-​based case-​
et al, 2011). control study in France, where age-​standardized
A major role of the p16 gene is to halt the rates of oral cavity and oropharynx cancer are
progression of cell growth cycle at G1. Loss of high, the population attributable risk of oral
the p16 gene leads to loss of p53 functions and and oropharyngeal cancer for tobacco alone was
thereby loss of cell-​cycle control, contributing to estimated to be 12.7% and 70% for the combined
unregulated cell growth. Because transcription effect of tobacco and alcohol (Radoï et al, 2015).
of HPV E7 oncogenic proteins upregulate p16 Tobacco appears to be a particularly important
expression, p16 mRNA and protein expression contributor to increases in incidence of oral
are used as surrogate markers of HPV status. P16 cavity and oropharynx cancers in women, who
expression was first discovered to be associated have experienced similar increases in incidence
with prognosis and response to therapy in oro- across many European countries (Chatervedi
pharyngeal tumors (Ang et al, 2010; Rischin et al, et al, 2013). This is in contrast to trends in men,
2010; Posner et al, 2011; Rietbergen et al, 2013), among whom only oropharyngeal tumors are
but more recently it has also been associated increasing in incidence, indicating a more prom-
with prognosis in oral cavity tumors where HPV inent role of HPV in increasing incidence of total
infection is less common (Chung et al, 2014). oral cavity and oropharynx tumors (Chatervedi
Epidermal growth factor receptor (EGFR) et al, 2013).
overexpression occurs in 90% of head and neck Smoking tobacco as cigarettes exposes the
cancers; higher expression levels and gene copy oral cavity to several carcinogens including PAHs,
number variation is associated with worse prog- aldehydes, and nitrosamines. A  positive associ-
nosis (Chung et al, 2006). Moreover, tumors with ation between cigarette smoking and oral and
low EGFR expression and high P16 expression oropharyngeal cancer has consistently been doc-
displayed more favorable clinical outcomes than umented in numerous studies (IARC, 2004a). The
tumors with high EGFR and low P16 expression risk increases substantially with duration of smok-
(Kumar et al, 2008). ing and number of cigarettes smoked. In Italy,
cigarette smokers had an 11-​fold higher risk of
RISK FACTORS oral and oropharyngeal cancer than nonsmokers
Epidemiologic studies over the past several (Franceschi et al, 1990). The risk increased with
decades have elucidated several important risk both the number of cigarettes smoked per day
factors for oral and oropharyngeal cancer, sum- and the number of years that a person smoked.
marized in Table 7-​1 and presented in detail in Moreover, high-​tar cigarettes confer a greater risk
this section. than low-​tar cigarettes (LaVecchia et  al, 1990),
143

Oral and Pharyngeal Cancer 143

TABLE 7-​1   EPIDEMIOLOGY OF ORAL AND OROPHARYNGEAL

CANCER: RISK FACTOR SUMMARY

Direction of risk factor Effect*

Well-​Confirmed Risk Factors

Cigarette smoking ↑↑
Smokeless tobacco ↑↑
Betel quid/​paan with and without tobacco ↑↑
Alcohol consumption among smokers ↑↑
Human papilloma virus type 16 ↑↑
Family history in first-​degree relative ↑
Probable Relationship Exists, Based on Substantial Data
Poor oral health ↑
HPV-​18 ↑↑
Weak, if any, Relationship Exists, Based on Substantial Data
Meat/​processed meat intake ↑
Coffee/​tea intake ↓
Alcohol consumption among nonsmokers ↑
Mouthwash use ↑
Mate drinking ↑
Inconsistent Findings or Limited Study to Date
BMI, waist-​to-​hip ratio —​
Total dietary fat intake during adulthood ↑
Fiber intake ↓
Saturated fat intake ↑
Carotenoids/​Carotenoid rich fruits and vegetables ↓
Physical activity ↓
Folate intake ↓
Vegetable fats, polyunsaturated fats, olive oil intake ↓
Intakes of vitamins A, C, E ↓
Intakes of retinol, thiamin, riboflavin, niacin —​
Intakes of calcium, selenium, and iron —​
Herpes simplex virus ↑
Gastroesophageal reflux disease ↑

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑,
moderate to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease
in risk; —​, no association

while the use of mentholated cigarettes does not 1989; Franceschi et  al, 1990). The decline in
appear to affect the risk (Kabat & Hebert, 1994). cancer risk after quitting occurs quickly, so that
Exclusive smoking in the form of cigars or pipes former smokers who quit for more than 10 years
contributes to a twofold increased risk of oral and approach the risk of never-​smokers. These find-
oropharyngeal cancer (Blot et al, 1988). ings have important public health implications
Quitting cigarette smoking dramatically and support smoking cessation efforts as a means
reduces the risk of oral and oropharyngeal tumors to reduce morbidity and mortality from oral and
(Blot et al, 1988; Franco et al, 1989; Merletti et al, oropharyngeal cancer.
14

144 Textbook of Cancer Epidemiology

Snuff was introduced in countries as and oropharyngeal cancer for black and blond
diverse as Sweden and Sudan almost 400 years tobacco is plausible because the concentra-
ago (Idris et  al, 1998). Chewing tobacco and tion of several carcinogens differs substantially
snuff contain tobacco-​specific N-​nitrosamines between black and blond tobacco; notably there
that are carcinogenic and have been shown to is a higher level of N-​nitrosamines and aromatic
cause oral cancer in mouse (Hecht et al, 1986; amines in black tobacco smoke (Boffetta, 1993;
Hecht & Hoffmann, 1988)  and rat models Sancho-​Garnier & Theobald, 1993; De Stefani
(Balbo et  al, 2013), but data among humans et al, 1993). A hospital-​based case control study
are not entirely consistent. The composition of from Uruguay reported a threefold increased
snuff varies across different cultures, and this risk of both oral cancer and pharyngeal cancer
has been posited as one reason for the con- among smokers of black compared to smok-
flicting evidence on snuff use. In Scandinavia, ers of blond tobacco (de Stefani et  al, 1998).
where moist snuff (snus) is common, no asso- Similarly, in a study from Spain there was a two-​
ciation between snuff use and oral and oropha- to fourfold increase in risk of oral and pharyn-
ryngeal tumors was apparent after adjustment geal cancer among smokers of black compared
for tobacco and alcohol in a meta-​ analysis to smokers of blond tobacco (Castellsague
(Lee & Hamling, 2009). The same analysis et al, 2004).
also identified changes over time in the asso-
ciation between smokeless tobacco use and
oral and oropharyngeal cancer in the United Diet
States, with no apparent increase in risk since Dietary Patterns
approximately 1990. Intake of fruits and vegetables has been con-
Approximately 90% of smokeless tobacco sistently inversely associated with oral cavity
consumed globally is estimated to occur in South and oropharynx cancer in studies conducted in
Asia (WHO, 2013). In India and other parts of geographically diverse populations (Franceschi
Asia, smokeless tobacco, in the form of betel et al, 1999a; Boeing et al, 2006; Freedman et al,
quid, paan, or nass, has long been suspected to 2008; Sapkota et  al, 2008; Lagiou et  al, 2009;
be related to the elevated rates of oral cancer seen Edefonti et  al, 2010; Bravi et  al, 2013). The
in these countries (Jayant & Deo, 1986). A review World Cancer Research Fund determined that
and meta-​analysis of six studies that controlled the evidence for nonstarchy vegetables, fruits,
for alcohol and smoking estimated an odds ratio and carotenoid-​rich foods was consistent with
of 6.3 (95% CI 3.9–​10.2) for oral cancer among a probable causal relationship with oral and
paan/​ betel quid users compared to nonusers oropharyngeal cancer (World Cancer Research
(Khan et al, 2014). Fund, 2007), a downgrading from their pre-
While tobacco appears to be the key agent vious report (World Cancer Research Fund,
in the etiology of oral and oropharyngeal can- 1997). The International Head and Neck Cancer
cer, other components in these products may Epidemiology (INHANCE) consortium initi-
enhance the risk (Gupta et al, 1996). In an eval- ated in 2004 and including populations from
uation of all available epidemiological studies Europe, North America, Latin America, India,
of betel quid by IARC, not only betel quid with Japan and Australia, has permitted large pooled
tobacco, but also betel quid without tobacco was analyses of dietary intake. Within INHANCE,
found to be causally associated with oral cancer oral and oropharyngeal cancer risk reductions
(IARC, 2004b). The areca nut itself, a constitu- of approximately 50% and 35% were identified
ent of paan and betel quid, contains arecoline comparing the top to the bottom quartile of
and other alkaloids, which can be converted fruit intake and vegetable intake, respectively
into N-​nitroso compounds. Even in the absence (Chuang et al, 2012).
of tobacco, consumption of paan has also been Meat consumption, on the other hand, has
reported to increase the risk of oral cancer shown less consistent associations with oral cav-
(Merchant et al, 2000). ity and oropharyngeal cancer. In INHANCE,
Black (air-​cured) tobacco, traditionally used higher intake of both red meat and processed
in Latin American and Mediterranean countries meat was associated with greater oral cavity and
in preference to blond (flue-​cured) tobacco, is oropharynx cancer risk (Chuang et  al, 2012).
now the predominant type in most industrial- A meta-​analysis including 12 case-​control stud-
ized countries. A potentially different risk of oral ies and one cohort study found that processed
145

Oral and Pharyngeal Cancer 145

meat—​but not total meat, red meat, or white to act in the chemoprevention of oral precancer-
meat—​was associated with an increased risk (Xu ous lesions. In Japan, men with 1 mmol/​L higher
et al, 2014). serum beta-​carotene had an 84% lower risk of
Tea and coffee, both substantial contributors developing the precancerous condition leukopla-
to antioxidant intake in a Western dietary pat- kia (Nagao et al, 2000).
tern, have been investigated with respect to oral Beta-​carotene has also been evaluated in
and oropharyngeal cancers in a number of stud- intervention studies. Among betel quid chewers,
ies, but most have been inconclusive due to small those randomized to a cocktail of beta-​carotene
sample size or the inclusion of multiple anatomic and retinol had been reported to have a threefold
sites with potentially divergent etiologies. In a decrease in the proportion of buccal cells with
case-​control study in Taiwan, a 40% reduction in micronuclei (Stich et al, 1984). Patients with leu-
oral and oropharyngeal cancer risk was observed koplakia who were randomized to beta-​carotene
when comparing five or more cups of tea versus supplements showed sustained remission of the
zero cups per day; the association was strongest lesions compared to those on placebo (Garewal
for pharyngeal cancer, with a nearly 70% reduc- et al, 1999), but these results were not confirmed
tion (Huang et al, 2014). In a French case-​control in a subsequent trial that randomized patients
study, greater coffee and tea intake were both with leukoplakia to low-​dose beta-​carotene and
associated with reduced oral and oropharyngeal vitamin C (Nagao et al, 2015). The chemopreven-
cancer risk, with no differences observed across tive nature of beta-​carotene may be related to its
anatomic sites (Radoï et al, 2013). In INHANCE, role as an antioxidant, with the ability to quench
coffee intake of more than four compared to oxygen-​free radicals.
zero cups per day showed an approximately 40% The World Cancer Research Fund determined
reduction in risk of oral cavity and oropharyn- that evidence was too limited to support a causal
geal cancers; no association was found for tea role of vitamin A, vitamin C, vitamin E, retinol,
intake (Galeone et  al, 2010). These results need thiamin, riboflavin, niacin, folate, calcium, iron,
to be interpreted with caution, given the poten- or selenium in oral and oropharyngeal cancer
tial biases known to occur when assessing dietary development (World Cancer Research Fund,
associations from case-​control studies. 2007). However, laboratory studies suggest that
the antioxidant vitamin C may act as a che-
Micronutrient Intake mopreventive agent, either alone or in concert
Although it can be difficult to disentangle the role with vitamin E, to prevent oral carcinogenesis
of individual micronutrients, observational and (Sawant & Kandarkar, 2000). A study within the
experimental studies indicate that carotenoids INHANCE consortium found that risk of oral
may be effective in reducing oral and oropha- and oropharyngeal cancer was reduced by nearly
ryngeal cancer risk. Carotenoid intake, measured 50% when comparing the top to the bottom
through dietary assessment or serum levels is quintile of natural vitamin C intake (Edefonti
associated with lower risk of oral and oropharyn- et al, 2015a). A case-​cohort study nested within
geal cancer in many (Negri et al, 2000; Bravi et al, the Netherlands Cohort Study found strong
2013) but not all studies (de Munter et al, 2015). inverse associations for highest compared to low-
The World Cancer Research Fund determined est quartile of vitamin C intake and risk of oral
foods containing carotenoids to be probably cavity cancer (OR = 0.35, 95% CI 0.16–​0.77) and
protective in reducing risk of oral and oropha- oropharyngeal cancers (OR = 0.29, 95% CI 0.12–​
ryngeal cancer (World Cancer Research Fund, 0.67) (de Munter et al, 2015). Total and natural
2007). In INHANCE, total carotenoid intake was folate intake was also associated with a reduc-
associated with an approximately 40% reduc- tion in risk of oral and oropharyngeal tumors
tion in risk of oral cavity cancer; intakes of beta-​ in INHANCE, but associations were stronger
carotene, lycopene, and letein plus zeaxanthin for oral cavity tumors (Galeone et  al, 2014).
were associated with 17%–​18% reductions in risk Moreover, the effect of folate on oral and oropha-
for both oral and pharyngeal tumors. Moreover, ryngeal cancer was modified by alcohol intake,
an interaction was found between carotenoid with heavy alcohol drinkers who consume low
intake and alcohol/​tobacco consumption, with levels of folate displaying a more than fourfold
a stronger inverse association seen for those increase in risk compared to never/​light drinkers
reporting greater levels of drinking and smoking. with high levels of folate intake. Despite identi-
(Leoncini et al, 2016). Beta-​carotene also appears fying a significant inverse association between
146

146 Textbook of Cancer Epidemiology

natural vitamin E intake and oral and oropha- are a rich source of dietary fiber, have also been
ryngeal cancers within INHANCE (OR  =  0.59, reported to be inversely associated with oral can-
95% CI 0.49–​0.71), the analysis revealed sub- cer risk in some (De Stefani et al, 1999), but not
stantial between-​study heterogeneity (Edefonti other studies (McLaughlin et al, 1988).
et  al, 2015b). Vitamin E was not associated
with risk of oral and oropharyngeal tumors in Alcohol
the Netherlands Cohort Study, but appeared to Epidemiologic data collected over the past
interact with alcohol consumption (de Munter 40 years provide consistent evidence that alcohol
et al, 2015). elevates the risk of oral and oropharyngeal can-
cer; it is the only dietary exposure for which the
Fat Intake World Cancer Research Fund determined that
Overall, the World Cancer Research Fund has the existing evidence of a causal association is
determined that evidence is too limited to sup- convincing (World Cancer Research Fund, 2007).
port a causal association of specific types of fat Studies from US, European, Chinese, and Korean
with oral and oropharyngeal cancer (World populations demonstrate an increased risk of
Cancer Research Fund, 2007). oral and oropharyngeal cancer with increasing
Several studies have examined these asso- amount of alcohol consumed or frequency of
ciations. Increased consumption of butter or drinking (Blot et al, 1988; Kabat & Wynder, 1989;
saturated fats has been reported to be positively, Merletti et al, 1989; Franceschi et al, 1990; Zheng
whereas that of olive oil has been reported to be et  al, 1990; Choi & Kahyo, 1991; Lubin et  al,
inversely, associated with the risk of these cancers 2011; Hashibe et  al, 2013). For example, in the
(Fioretti et  al, 1999; Franceschi, 1999a, 1999b). Prostate, Lung, Colorectal, and Ovarian (PLCO)
Consistent with these results, a case-​control study cancer screening trial, risk of total oral cavity
in Italy and Switzerland found intake of animal and oropharyngeal cancers more than doubled
fat, saturated fats, and cholesterol to be positively, with consumption of two or more compared
and intake of vegetable and polyunsaturated fats to zero drinks per day (Hashibe et  al, 2013). In
to be inversely, associated with the risk of these INHANCE, increases in risk were observed for
cancers (Bravi et  al, 2013). No association was tumors of the oral cavity, oropharynx, and hypo-
found for total fat or monounsaturated fatty acids pharynx in both men and women, but the magni-
in the same study. Intake of trans fatty acids, spe- tude of association associated with the same level
cifically ruminant fat, was associated with an of daily consumption was considerably higher in
increased risk of oral and oropharyngeal tumors women (Lubin et al, 2011).
in a study based in Norway (Laake et al, 2013). The effect of alcohol among never-​smokers,
if it exists, is weak and has been estimated
Fiber Intake to account for only 0.3% of oral cancer in the
Overall, the World Cancer Research Fund has ICARE study (Radoï et  al, 2015). The large
determined that evidence is too limited to sup- INHANCE pooled analysis of 17 studies,
port an association of fiber intake with oral and including 7,000 oral and pharynx cancer cases,
oropharyngeal cancer (World Cancer Research came to a similar conclusion and estimated that
Fund, 2007). alcohol on its own caused 0% of oral cancer
Fiber intake has been examined with respect cases and 5% of pharynx cancer cases (Hashibe
to oral and oropharyngeal cancer in a limited et al, 2009). However, perhaps the most striking
number of studies. It has been reported that it contribution of alcohol to the etiology of oral
may protect against both cancer (Gridley et  al, cancer is its consistently observed interaction
1990; Zheng et al, 1993; MacFarlane et al, 1995; with tobacco (Rothman & Keller, 1972). The
Horn-​Ross et  al, 1997)  and precancerous oral effect of the two risk factors suggests a greater
lesions (Gupta et al, 1998). In China, dietary fiber than multiplicative effect between tobacco and
derived from fruits and vegetables was inversely alcohol. While the relative risk of oral cancer
associated with oral cancer risk, while fiber among men who were intermediate cigarette
derived from other sources did not exhibit any smokers but who drank fewer than 35 drinks
protective effect (Zheng et al, 1993). Fiber from per week was 10.9, and was 2.3 for those who
vegetables, fruits, or grains was associated with a drank 60 or more alcoholic drinks per week
50% reduction in risk of oral cancer in an Italian but did not smoke, the risk for those who were
population (Soler et  al, 2001). Legumes, which both intermediate smokers and drank was more
147

Oral and Pharyngeal Cancer 147

than 36-​fold compared to that of men who did and greater waist-​to-​hip ratio; BMI was associ-
neither (Franceschi et  al, 1990). In the PLCO ated with higher risk in never smokers but not in
cohort, the relative risk of oral and oropharyn- former or current smokers. (Gaudet et al, 2015).
geal cancer was 1.4 comparing nonsmokers who In INHANCE, compared to those in the healthy
drank at least two alcoholic drinks per day to WHO-​defined BMI category (18.5–​24.9 kg/​m2),
never smokers who did not consume alcohol, risk was elevated for those in the underweight
but the relative risk increased to 11 when con- category and reduced for overweight/​obese, in
sidering those who drank at least two drinks per both men and women (Lubin et al, 2011). Similar
day and consumed at least one pack of cigarettes findings were observed in the ARCAGE study
per day (Hashibe et al, 2013). (Park et al, 2011), but importantly, both of these
The mechanism by which alcohol is a risk fac- studies did not see an association with BMI in
tor for oral and oropharyngeal cancer is unclear, never smokers or never drinkers. Oral and oro-
especially since ethanol in its pure form does pharyngeal cancer cases have been observed to be
not act as a carcinogen in experimental models leaner than controls in many studies, which may
(IARC, 1988). Although alcohol is not believed indicate a poor nutritional status (Kabat et  al,
to be a carcinogen per se, metabolism of alcohol 1994; D’Avanzo et al, 1996; IARC, 2001; Kreimer
by enzymes (ADH, CYP2E1) activates metabo- et al, 2006). Consistent with these results, height
lites with carcinogenic potential, such as acet- was inversely associated with risk in INHANCE
aldehyde. Many of the metabolic enzymes are (Leocini et al, 2014). However, in the NCI Cohort
expressed at the oral mucosa (Peters et al, 1993; Consortium height was associated with elevated
Dong et al, 1996), suggesting a mechanistic role risk in never and former smokers, but not in cur-
for alcohol in oral and oropharyngeal cancer eti- rent smokers (Gaudet et al, 2015).
ology. Alcohol may also contribute to carcino-
genesis by direct irritation of the oral mucosa or Infections
through the nutritional deficiencies associated Human papilloma viruses (HPVs) are an increas-
with heavy alcohol consumption (Lieber et  al, ingly important etiologic factor in oral cavity and
1979). Potential reasons why alcoholic beverages oropharyngeal cancers. HIV infection is also
are carcinogenic include their acting merely as a associated with increased risk of oral cavity and
solvent for tobacco carcinogens (Seitz et al, 1998; oropharynx cancer, particularly HPV-​associated
Wright & Ogden, 1998). Similarly impurities or tumors, resulting from immunosuppression.
contaminants in alcoholic drinks have been sug- Other infections have also been associated with
gested to represent the main carcinogenic agent. disease risk, but may be markers of immune sup-
For example, polycyclic aromatic hydrocarbons pression rather than causally related to tumor
are found in dark strong liquors such as whiskey, development.
and N-​nitrosodiethylamine has been detected in
some beers. Human Papilloma Virus
Human papilloma virus type 16 is a recognized
Reproductive Factors cause of oropharyngeal and oral cavity cancers
There is very limited evidence that menstrual and (IARC, 2007). The evidence comes from mul-
reproductive factors may be associated with the tiple epidemiologic studies and is supported by
risk of cancer of the salivary gland (Horn-​Ross molecular pathology studies, animal models, and
et al, 1999). No published studies have examined in vitro studies (Gillison et  al, 2012a). Tumors
these factors in relation to other anatomical sub- of the oropharynx, particularly the tonsil and
sites of oral cancer. base of the tongue, are most strongly associ-
ated with HPV (Gillison et  al, 2015). Estimates
Anthropometric Measures of HPV-​attributable proportions vary substan-
The association between adiposity and oral cavity tially depending on how HPV status was deter-
and oropharynx cancers is complex and depends mined and whether confounding by smoking is
on anatomic subsite or HPV status. Adiposity addressed, but range between 28%–​46% for oro-
measured by body mass index (BMI) was not pharyngeal cancer and 9%–​24% for oral cavity
associated with the risk in the PLCO cohort tumors (Gillison et al, 2015). HPV-​16 is detected
(Hashibe et  al, 2013). However, a study in the in the overwhelming majority of HPV-​positive
NCI Cancer Cohort Consortium found increased oral cavity and oropharynx cancers; HPV-​ 18
risk associated with greater waist circumference is uncommon in oropharyngeal tumors but
148

148 Textbook of Cancer Epidemiology

somewhat more common in tumors of the oral partners may be more strongly associated with
cavity (Kreimer et al, 2005). risk in men than in women (Chaturvedi et  al,
Serological biomarkers of HPV-​16 infection, 2015). Men have higher prevalence of oral high-​
which measure exposure to infection at any site, risk HPV in the United States (Gillison et  al,
include antibodies against HPV-​16 capsid antigen 2012b), and increases in incidence over time are
L1 and antibodies against HPV oncoproteins E6/​ consistent with rising incidence rates of oropha-
E7, each reflecting distinct biological phenom- ryngeal cancers in US men in recent decades
ena. Circulating L1 antibodies indicate serocon- (Chaturvedi et al, 2011).
version following HPV-​16 exposure, estimated Patients with HPV-​associated oral cavity and
to occur in 60%–​70% of individuals. However, oropharyngeal tumors vary in a number of ways
HPV E6/​H7 antibodies are overexpressed only that may influence disease outcomes (Marur
in cancer and are much more strongly associated et al, 2010). They tend to be younger at diagnosis
with HPV-​ associated tumors and HPV DNA than their HPV-​negative counterparts and have
detection in tumor tissue. (Gillison et al, 2012a). fewer comorbidities. HPV-​positive tumors also
Accordingly, studies that have used L1 serology tend to be more responsive to cisplatin-​based
have found more modest relative risks for oral chemotherapy and radiation.
cavity and oropharyngeal tumors than studies of
E6/​E7 serology. Regardless of the type of serology Other Infections
used, the strongest associations are consistently The higher risk of HPV-​positive malignancies
observed for tumors of the oropharynx, specifi- among patients infected with HIV as a result of
cally. Odds ratios for joint HPV-​16 E6/​E7 sero- immunosuppression has been consistently doc-
positivity and oropharyngeal tumors were 67.1 umented (Clifford et  al, 2005; Silverberg et  al,
(95% CI 12.9–​348.2) in one multicenter study 2011; Franzetti et al, 2013). Elevated risk of oral
of geographically diverse populations (Herrero and oropharyngeal cancer, specifically, for HIV-​
et al, 2003) and 179 (95% CI 35.8–​899) (Ribeiro positive persons compared to the general pop-
et al, 2011) in a large multicenter study from Latin ulation, is also quite consistent across studies
America. In another large multicenter study in (Frisch et al, 2000; Clifford et al, 2005; Silverberg
Europe, joint HPV-​16 E6/​E7 seropositivity was et al, 2011; Chaturvedi et al, 2009). HIV-​positive
rare, but the OR for HPV-​16 E6 seropositivity was individuals have an oral HPV infection preva-
132 (95% CI  =  65.29–​266.86) (Anantharaman lence of up to 40% (Beachler et al, 2012), approx-
et al, 2013). HPV-​16 E6 antibodies have also been imately two times higher than that of the general
detected in up to 10 years before onset of symp- population (Beachler et  al, 2012; D’Souza et  al,
toms for oropharynx cancer, raising the possibil- 2007). Importantly, tobacco use is also more
ity that it could be a sensitive and specific marker common among HIV-​infected populations, fur-
for HPV-​positive disease (Kreimer et  al, 2013). ther increasing the burden of oral and oropha-
Associations with tumors in the oral cavity were ryngeal tumors among these patients (Tesoriero
weaker and not always significant. et al, 2010).
HPV is primarily acquired through sexual Herpes simplex viruses 1 (HSV-​ 1) and
transmission. In epidemiologic studies, lifetime 2  (HSV-​2) have also been associated with oral
number of genital and oral sexual partners is cancer, but with associations much weaker than
generally associated with an increased risk of for HPV. In patients with oral cavity and oro-
oropharyngeal cancer; sexual behavior is not pharyngeal tumors, HSV-​ 1 antibodies have
consistently associated with cancers of the oral been observed to be higher compared to con-
cavity (Herrero et al, 2003; Rajkumar et al, 2003; trols (Shillitoe et  al, 1986; Larsson et  al, 1991).
Heck et  al, 2010). In INHANCE (Heck et  al, One study showed a threefold increased risk
2010), substantial variation in frequency of oral of oral cancer associated with HSV-​1 overall,
sexual behaviors was identified, geographically with a somewhat higher risk for cancer of the
and by gender, and increased among more recent lip (Schildt et  al, 1998); another study showed
birth cohorts, consistent with increases in inci- an almost twofold increased risk associated
dence of HPV-​ positive oropharyngeal cancer. with HSV-​2 (Maden et  al, 1992). Using SEER-​
Moreover, oral sexual behavior is associated with Medicare linkage in a study of elderly Americans,
both prevalent and incident oral HPV infection Herpes zoster diagnosis was associated with a
(Gillison et al, 2015). Data from NHANES sug- statistically significant 21% increase in odds of
gest that sexual behaviors and number of sexual oral cavity and pharyngeal cancer, with stronger
149

Oral and Pharyngeal Cancer 149

associations for advanced stage tumors (Mahale Support for a role of nitrosamines in the eti-
et  al, 2016). It is possible that HSV is acting as ology of oral cancer springs from experimental
a marker for immunosuppression rather than studies showing that non-​tobacco-​specific nitro-
being causally related to cancer development. samines result in the formation of DNA adducts
in oral mucosa. The rubber industry has some of
Physical Activity the highest exposures to nitrosamines, produced
In three large independent studies, greater levels in large part during salt bath curing and vulcan-
of physical activity were inversely associated with ization processing, and an IARC working group
incidence of tumors of the oral cavity and oro- concluded that for a variety of cancers, but not
pharynx. In the PLCO cohort, physical activity oral and oropharyngeal tumors, there was suf-
was associated inversely with oral cavity and oro- ficient evidence to support a causal association
pharyngeal cancer risk (Hashibe et al, 2013). In (IARC, 2012). While earlier studies showed
the NIH-​AARP Diet and Health Study, an inverse excess mortality from oral and oropharyngeal
association was also observed (Leitzmann et  al, cancer among rubber tire industry workers
2008). INHANCE has reported a risk reduc- (Delzell & Monson, 1981; Sorahan et  al, 1989;
tion for moderate physical activity (Nicolotti Straif et al, 2000), more recent analyses have not
et al, 2011). supported these original findings (Alder et  al,
2006; Pira et al, 2012).
Ionizing Radiation
Ionizing radiation is a risk factor for salivary Medical Conditions
gland cancer (Xu et  al, 2015). Risk for salivary and Treatment
gland tumors is increased in individuals exposed Limited data from a nationwide registry in
to environmental radiation (Saku et al, 1997) and Taiwan suggest that gastroesophageal reflux may
patients treated with radiotherapy for benign be linked to tumors of the oropharynx and hypo-
and malignant conditions of the head and neck pharynx (Kuo et al, 2015).
(Thompson et al, 1994; van der Laan et al, 1995;
Schneider et  al, 1998; Sale et  al, 2004; Whatley Other Risk Factors
et al, 2006). Mouthwash
Since the first report by Weaver and colleagues
Occupation (1979), several researchers have sought to examine
Higher rates of oral and oropharyngeal can- whether mouthwash increases the risk of oral can-
cer have been observed in some occupations. cer because it may contain up to 27% alcohol. In a
The European ARCAGE study, a multisite case-​ meta-​analysis including 18 studies of mouthwash
control study of including upper aerodigestive use and oral and oropharyngeal cancer, no signifi-
tract cancers in men and women, evaluated 283 cant association was identified for mouthwash use
occupations and found increased risk of oral overall or alcohol-​containing mouthwash (Gandini
and oropharyngeal cancer for workers in a vari- et al, 2012). Increased risk of oral and oropharyn-
ety of roles in the construction industry, includ- geal cancer, or combined upper aerodigestive tract
ing concrete workers, bricklayers, painters, and cancers, however, was observed in subsequent
those involved in road and building construc- case-​control studies evaluating type of mouthwash
tion (Richiardi et  al, 2012). In a study among use (Eliot et al, 2013; Ahrens et al, 2014) and non-
women in the ICARE study in France, risk of all alcoholic mouthwash (Eliot et al, 2013).
head and neck cancers combined was substan-
tially elevated for women employed more than Mate
10 years as welders and flame-​cutters, with a rel- Mate is an herb-​infused beverage typically drunk
ative risk of more than 20 (Carton et  al, 2014). at very high temperatures through a metal straw
Workers exposed to compounds including aro- in parts of South America, in particular south-
matic amines and phenoxy herbicides have been ern Brazil, Uruguay, and northern Argentina.
reported to be at elevated risk of oral cancer in There are consistent reports of an association
several cohort studies (Delzell & Monson, 1981; between esophageal cancer and consumption
Sorahan et  al, 1989; Merletti et  al, 1991; Foppa of mate, and it has been classified by the IARC
& Minder, 1992; Straif et al, 2000), as well as in as being “probably carcinogenic,” although only
a case-​control study that was able to adjust for when drunk at hot temperatures (IARC, 1991).
potential confounders (Coble et al, 2003). However, potentially carcinogenic chemical
150

150 Textbook of Cancer Epidemiology

properties of mate are conceivable. In 2007, the rubber-​ producing industry. Am J Epidemiol
World Cancer Research Fund determined that 2006;164:405–​20.
there was limited evidence suggesting that mate Amtha R, Ching CS, Zain R, Razak IA, Basuki B,
is a cause of oral and oropharyngeal cancer Roesian BO et al. GSTM1, GSTT1, and CYP1A1
(World Cancer Research Fund, 2007). In a meta-​ polymorphisms and risk of oral cancer:  a case-​
analysis including 566 cases of oral cancer, pub- control study in Jakarta, Indonesia. Asian Pac J
lished in 2010, mate drinking was associated with Cancer Prev 2009;10:21–​6.
a summary odds ratio of 2.1 (95% CI 1.4–​3.2) Anantharaman D, Gheit T, Waterboer T, Abedi-​
Ardekani B, Carreira C, McKay-​Chopin S et  al.
(Dasanayake et  al, 2010). In studies published
Human papillomavirus infections and upper
after the meta-​analysis, results have been mixed.
aero-​digestive tract cancers:  the ARCAGE study.
A multicenter hospital-​based case-​control study
J Natl Cancer Inst 2013;105:536–​45.
in Latin America found an association between Ang KK, Harris J, Wheeler R, Weber R, Rosenthal DI,
mate and esophageal cancer, but not oral cancer Nguyen-​Tân PF et al. Human papillomavirus and
(Szymańska et  al, 2010). A  case-​control study survival of patients with oropharyngeal cancer. N
in Uruguay reported a significant interaction Engl J Med 2010;363:24–​35.
between mate consumption and both alcohol and Asakage T, Yokoyama A, Haneda T, Yamazaki M,
tobacco on oral cancer risk, with substantially Muto M, Yokoyama T et  al. Genetic polymor-
higher risks for smokers and drinkers (Deneo-​ phisms of alcohol and aldehyde dehydrogenases,
Pellegrini et al, 2013). and drinking, smoking and diet in Japanese men
with oral and pharyngeal squamous cell carci-
CONCLUSION noma. Carcinogenesis 2007;28:865–​74.
Many important risk factors for oral and oro- Balbo S, James-​ Yi S, Johnson CS, O’Sullivan
pharyngeal cancer have been elucidated in the MG, Stepanov I, Wang M et  al.
epidemiologic literature, and a summary of (S)-N’-Nitro­ sonornicotine, a constituent of
the current state of knowledge is presented in smokeless tobacco, is a powerful oral cavity car-
Table 7-​1. Tobacco and alcohol remain impor- cinogen in rats. Carcinogenesis 2013;34:2178–​83.
tant modifiable risk factors for the disease, but Beachler DC, Weber KM, Margolick JB, Strickler HD,
Cranston RD, Burk RD et al. Risk factors for oral
the increasing incidence of HPV-​ associated
HPV infection among a high prevalence popu-
tumors is now evident. While these tumors are
lation of HIV-​positive and at-​risk HIV-​negative
more amenable to treatment than HPV-​negative
adults. Cancer Epidemiol Biomarkers Prev
tumors, they are still a source of considerable 2012;21:122–​33.
morbidity and mortality. Moreover, the lack of a Blot WJ, McLaughlin JK, Winn DM, Austin DF,
precursor lesion and limited data on efficacy of Greenberg RS, Preston-​Martin S et  al. Smoking
the HPV vaccine in preventing oral HPV infec- and drinking in relation to oral and pharyngeal
tion are barriers to primary and secondary pre- cancer. Cancer Res 1988;48:3282–​7.
vention efforts. Dietary patterns high in fruits Boeing H, Dietrich T, Hoffmann K, Pischon T, Ferrari
and vegetables and low in meats may confer P, Lahmann PH et al. Intake of fruits and vegeta-
some protection against oral and oropharyngeal bles and risk of cancer of the upper aero-​digestive
tumors. tract: the prospective EPIC-​study. Cancer Causes
Control 2006;17:957–​69.
REFERENCES Boffetta P, La Vecchia C, Levi F, Lucchini F. Mortality
Agrawal N, Frederick MJ, Pickering CR, Bettegowda patterns and trends for lung cancer and other
C, Chang K, Li RJ et  al. Exome sequencing of tobacco-​related cancers in the Americas, 1955–​
head and neck squamous cell carcinoma reveals 1989. Int J Epidemiol 1993;22:377–​84.
inactivating mutations in NOTCH1. Science Bosron WF, Li TK. Genetic polymorphism of human
2011;333:1154–​7. liver alcohol and aldehyde dehydrogenases, and
Ahrens W, Pohlabeln H, Foraita R, Nelis M, Lagiou their relationship to alcohol metabolism and alco-
P, Lagiou A et  al. Oral health, dental care and holism. Hepatology 1986;6:502–​10.
mouthwash associated with upper aerodigestive Bravi F, Bosetti C, Filomeno M, Levi F, Garavello
tract cancer risk in Europe:  the ARCAGE study. W, Galimberti S et  al. Foods, nutrients and the
Oral Oncol 2014;50:616–​25. risk of oral and pharyngeal cancer. Br J Cancer
Alder N, Fenty J, Warren F, Sutton AJ, Rushton L, 2013;109:2904–​10.
Jones DR et  al. Meta-​analysis of mortality and Brennan P, Lewis S, Hashibe M, Bell DA, Boffetta P,
cancer incidence among workers in the synthetic Bouchardy C et  al. Pooled analysis of alcohol
15

Oral and Pharyngeal Cancer 151

dehydrogenase genotypes and head and neck Clifford GM, Polesel J, Rickenbach M, Dal Maso
cancer:  a HuGE review. Am J Epidemiol L, Keiser O, Kofler A et  al. Swiss HIV Cohort.
2004;159:1–​16. Cancer risk in the Swiss HIV cohort study: asso-
Brunotto M, Zarate AM, Bono A, Barra JL, Berra S. ciations with immunodeficiency, smoking, and
Risk genes in head and neck cancer:  a system- highly active antiretroviral therapy. J Natl Cancer
atic review and meta-​analysis of last 5 years. Oral Inst 2005;97:425–​32.
Oncol 2014;50:178–​88. Coble JB, Brown LM, Hayes RB, Huang WY, Winn DM,
Cancer Genome Atlas Network. Comprehensive Gridley G et  al. Sugarcane farming, occupational
genomic characterization of head and neck solvent exposures, and the risk of oral cancer in
squamous cell carcinomas. Nature 2015;517: Puerto Rico. J Occup Environ Med 2003;45:869–​74.
576–​582. Cordero K, Espinoza I, Caceres D, Rocco A, Miranda
Carton M, Guida F, Paget-​Bailly S, Cyr D, Radoi L, C, Squicciarini V et  al. Oral cancer susceptibil-
Sanchez M et al. Occupation and head and neck ity associated with the CYP1A1 and GSTM1
cancer in women-​Results of the ICARE study. Am genotypes in Chilean individuals. Oncol Lett
J Ind Med 2014;57:1386–​97. 2010;1:549–​53.
Castellsague X, Quintana MJ, Martinez MC, Nieto D’Avanzo B, La Vecchia C, Talamini R, Franceschi S.
A, Sanchez MJ, Juan A et  al. The role of type of Anthropometric measures and risk of cancers of
tobacco and type of alcoholic beverage in oral car- the upper digestive and respiratory tract. Nutr
cinogenesis. Int J Cancer 2004;108:741–​9. Cancer 1996;26:219–​27.
Chaturvedi AK, Anderson WF, Lortet-​Tieulent J, de Munter L, Maasland DH, van den Brandt PA,
Curado MP, Ferlay J, Franceschi S et al. Worldwide Kremer B, Schouten LJ. Vitamin and carotenoid
trends in incidence rates for oral cavity and oro- intake and risk of head-​neck cancer subtypes in
pharyngeal cancers. J Clin Oncol 2013;31:4550–​9. the Netherlands Cohort Study. Am J Clin Nutr
Chaturvedi AK, Graubard BI, Broutian T, Pickard 2015;102:420–​32.
RK, Tong ZY, Xiao W et al. NHANES 2009–​2012 D’Souza G, Fakhry C, Sugar EA, Seaberg EC, Weber
findings:  association of sexual behaviors with K, Minkoff HL et al. Six-​month natural history of
higher prevalence of oral oncogenic human pap- oral versus cervical human papillomavirus infec-
illomavirus infections in U.S.  men. Cancer Res tion. Int J Cancer 2007;121:143–​50.
2015;75:2468–​77. Dasanayake AP, Silverman AJ, Warnakulasuriya S.
Chaturvedi AK, Engels EA, Pfeiffer RM, Hernandez Maté drinking and oral and oro-​pharyngeal can-
BY, Xiao W, Kim E et al. Human papillomavirus cer:  a systematic review and meta-​analysis. Oral
and rising oropharyngeal cancer incidence in the Oncol 2010;46:82–​6.
United States. J Clin Oncol 2011;29:4294–​301. De Stefani E, Barrios E, Fierro L. Black (air-​cured)
Chaturvedi AK, Madeleine MM, Biggar RJ, Engels and blond (flue-​ cured) tobacco and can-
EA. Risk of human papillomavirus-​ associated cer risk. III. Oesophageal cancer. Eur J Cancer
cancers among persons with AIDS. J Natl Cancer 1993;29A:763–​6.
Inst 2009;101:1120–​30. De Stefani E, Boffetta P, Oreggia F, Mendilaharsu M,
Choi SY, Kahyo H. Effect of cigarette smoking and Deneo-​Pellegrini H. Smoking patterns and can-
alcohol consumption in the aetiology of can- cer of the oral cavity and pharynx: a case-​control
cer of the oral cavity, pharynx, and larynx. Int J study in Uruguay. Oral Oncol 1998;34:340–​6.
Epidemiol 1991;20:878–​85. De Stefani E, Deneo-​ Pellegrini H, Mendilaharsu
Chuang SC, Jenab M, Heck JE, Bosetti C, Talamini R, M, Ronco A. Diet and risk of cancer of the
Matsuo K et al. Diet and the risk of head and neck upper aerodigestive tract. I.  Foods. Oral Oncol
cancer: a pooled analysis in the INHANCE con- 1999;35:17–​21.
sortium. Cancer Causes Control 2012;23:69–​88. Delzell E, Monson RR. Mortality among rubber work-
Chung CH, Ely K, McGavran L, Varella-​ Garcia ers. III. Cause-​specific mortality, 1940–​ 1978. J
M, Parker J, Parker N et  al. Increased epider- Occup Med 1981;23:677–​84.
mal growth factor receptor gene copy number Deneo-​Pellegrini H, De Stefani E, Boffetta P, Ronco
is associated with poor prognosis in head and AL, Acosta G, Correa P et  al. Maté consump-
neck squamous cell carcinomas. J Clin Oncol tion and risk of oral cancer: case-​control study in
2006;24:4170–​6. Uruguay. Head Neck 2013;35:1091–​5.
Chung CH, Zhang Q, Kong CS, Harris J, Fertig EJ, Dong YJ, Peng TK, Yin SJ. Expression and activities of
Harari PM et al. p16 protein expression and human alcohol dehydrogenase in human mouth. Alcohol
papillomavirus status as prognostic biomarkers of 1996;13:257–​62.
nonoropharyngeal head and neck squamous cell Edefonti V, Bravi F, La Vecchia C, Randi G, Ferraroni
carcinoma. J Clin Oncol 2014;32:3930–​8. M, Garavello W et  al. Nutrient-​ based dietary
152

152 Textbook of Cancer Epidemiology

patterns and the risk of oral and pharyngeal can- Franco EL, Kowalski LP, Oliveira BV, Curado MP,
cer. Oral Oncol 2010;46:343–​8. Pereira RN, Silva ME et  al. Risk factors for oral
Edefonti V, Hashibe M, Parpinel M, Ferraroni M, cancer in Brazil: a case-​control study. Int J Cancer
Turati F, Serraino D et al. Vitamin E intake from 1989;43:992–​1000.
natural sources and head and neck cancer risk: a Franzetti M, Adorni F, Parravicini C, Vergani B,
pooled analysis in the International Head and Antinori S, Milazzo L et  al. Trends and predic-
Neck Cancer Epidemiology consortium. Br J tors of non-​AIDS-​defining cancers in men and
Cancer 2015b;113:182–​92. women with HIV infection:  a single-​institution
Edefonti V, Hashibe M, Parpinel M, Turati F, Serraino retrospective study before and after the introduc-
D, Matsuo K et  al. Natural vitamin C intake tion of HAART. J Acquir Immune Defic Syndr
and the risk of head and neck cancer:  a pooled 2013;62:414–​20.
analysis in the International Head and Neck Freedman ND, Park Y, Subar AF, Hollenbeck AR,
Cancer Epidemiology Consortium. Int J Cancer Leitzmann MF, Schatzkin A et al. Fruit and veg-
2015a;137:448–​62. etable intake and head and neck cancer risk in
Eliot MN, Michaud DS, Langevin SM, McClean a large United States prospective cohort. Int J
MD, Kelsey KT. Periodontal disease and mouth- Cancer 2008;122:2330–​6.
wash use are risk factors for head and neck squa- Frisch M, Biggar RJ, Goedert JJ. Human
mous cell carcinoma. Cancer Causes Control papillomavirus-associated cancers in patients
2013;24:1315–​22. with human immunodeficiency virus infection
Fakhry C, Westra WH, Li S, Cmelak A, Ridge JA, and acquired immunodeficiency syndrome. J Natl
Pinto H et al. Improved survival of patients with Cancer Inst 2000;92:1500–​10.
human papillomavirus-​ positive head and neck Galeone C, Tavani A, Pelucchi C, Turati F, Winn
squamous cell carcinoma in a prospective clinical DM, Levi F et  al. Coffee and tea intake and risk
trial. J Natl Cancer Inst 2008;100:261–​269. of head and neck cancer:  pooled analysis in the
Fakhry C, Zhang Q, Nguyen-​Tan PF, Rosenthal D, international head and neck cancer epidemiology
El-​Naggar A, Garden AS et  al. Human papillo- consortium. Cancer Epidemiol Biomarkers Prev
mavirus and overall survival after progression of 2010;19:1723–​36.
oropharyngeal squamous cell carcinoma. J Clin Galeone C1, Edefonti V, Parpinel M, Leoncini E,
Oncol 2014;32:3365–​3373. Matsuo K, Talamini R et al. Folate intake and the
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, risk of oral cavity and pharyngeal cancer: a pooled
Mathers C, Rebelo M, Parkin DM, Forman D, Bray, analysis within the International Head and Neck
F. GLOBOCAN 2012 v1.0, Cancer Incidence and Cancer Epidemiology Consortium. Int J Cancer
Mortality Worldwide: IARC CancerBase No. 11 2015;136:904–​14.
[Internet]. Lyon, France: International Agency for Gandini S, Negri E, Boffetta P, La Vecchia C, Boyle
Research on Cancer; 2013. Available from: http:// P. Mouthwash and oral cancer risk quantita-
globocan.iarc.fr, accessed 2017. tive meta-​analysis of epidemiologic studies. Ann
Fioretti F, Bosetti C, Tavani A, Franceschi S, La Vecchia Agric Environ Med 2012;19:173–​80.
C. Risk factors for oral and pharyngeal cancer in Garavello W, Foschi R, Talamini R, La Vecchia C,
never smokers. Oral Oncol 1999;35:375–​8. Rossi M, Dal Masso L et al. Family history and the
Foppa I, Minder CE. Oral, pharyngeal and laryngeal risk of oral and pharyngeal cancer. Int J Cancer
cancer as a cause of death among Swiss cooks. 2008;122:1827–​31.
Scand J Work Environ Health 1992;18:287–​92. Garewal HS, Katz RV, Meyskens F, Pitcock J, Morse
Franceschi S, Favero A, Conti E, Talamini R, Volpe D, Friedman S et al. Beta-​carotene produces sus-
R, Negri E et al. Food groups, oils and butter, and tained remissions in patients with oral leukopla-
cancer of the oral cavity and pharynx. Br J Cancer kia: results of a multicenter prospective trial. Arch
1999a;80:614–​20. Otolaryngol Head Neck Surg 1999;125:1305–​10.
Franceschi S, Levi F, Conti E, Talamini R, Negri E, Dal Gaudet MM, Kitahara CM, Newton CC, Bernstein L,
Maso L et  al. Energy intake and dietary pattern Reynolds P, Weiderpass E et  al. Anthropometry
in cancer of the oral cavity and pharynx. Cancer and head and neck cancer:  a pooled analysis of
Causes Control 1999b;10:439–​44. cohort data. Int J Epidemiol 2015;44:673–​81.
Franceschi S, Talamini R, Barra S, Baron AE, Negri Gillison ML, Alemany L, Snijders PJ, Chaturvedi A,
E, Bidoli E et  al. Smoking and drinking in rela- Steinberg BM, Schwartz S et  al. Human papillo-
tion to cancers of the oral cavity, pharynx, lar- mavirus and diseases of the upper airway:  head
ynx, and esophagus in northern Italy. Cancer Res and neck cancer and respiratory papillomatosis.
1990;50:6502–​7. Vaccine 2012a;30 Suppl 5:F34–​54.
153

Oral and Pharyngeal Cancer 153

Gillison ML, Broutian T, Pickard RK, Tong ZY, Epidemiology (INHANCE) consortium. Int J
Xiao W, Kahle L et  al. Prevalence of oral HPV Epidemiol 2010;39:166–​81.
infection in the United States, 2009–​10. JAMA Hernando-​Rodriguez M, Rey-​Barja N, Marichalar-​
2012b;307:693–​703. Mendia X, Rodriguez-​Tojo MJ, Acha-​Sagredo A,
Gillison ML, Chaturvedi AK, Anderson WF, Fakhry Aguirre-​ Urizar JM. Role of cytochrome P-​ 450
C. Epidemiology of human papillomavirus-​ genetic polymorphisms in oral carcinogenesis. J
positive head and neck squamous cell carcinoma. Oral Pathol Med 2012;41:1–​8.
J Clin Oncol 2015;33:3235–​42. Herrero R, Castellsague X, Pawlita M, Lissowska J,
Gridley G, McLaughlin JK, Block G, Blot WJ, Winn Kee F, Balaram P et  al.; IARC Multicenter Oral
DM, Greenberg RS et al. Diet and oral and pha- Cancer Study Group. Human papillomavirus and
ryngeal cancer among blacks. Nutr Cancer oral cancer: the International Agency for Research
1990;14:219–​25. on Cancer multicenter study. J Natl Cancer Inst
Guengerich FP, Shimada T. Oxidation of toxic and car- 2003;95:1772–​83.
cinogenic chemicals by human cytochrome P-450 Horn-​Ross PL, Morrow M, Ljung BM. Diet and the
enzymes. Chem Res Toxicol 1991;4(4):391–​407. risk of salivary gland cancer. Am J Epidemiol
Gupta PC. Survey of demographic characteristics of 1997;146:171–​6.
tobacco use among 99,598 individuals in Bombay, Horn-​Ross PL, Morrow M, Ljung BM. Menstrual and
India using handheld computers. Tobacco Control reproductive factors for salivary gland cancer risk
1996;5:114–​20. in women. Epidemiology 1999;10:528–​30.
Gupta PC, Hebert JR, Bhonsle RB, Sinor PN, Mehta Howlader N, Noone AM, Krapcho M, Garshell J,
H, Mehta FS. Dietary factors in oral leukoplakia Miller D, Altekruse SF et al. SEER Cancer Statistics
and submucous fibrosis in a population-​ based Review. Bethesda:  National Cancer Institute;
case control study in Gujarat, India. Oral Dis 1975–​ 2011 [updated 1975–​ 2011; cited  2015];
1998;4:200–​6. Available from:  http://​seer.cancer.gov/​csr/​1975_​
Hashibe M, Brennan P, Chuang SC, Boccia S, 2011/​, based on November 2013 SEER data sub-
Castellsague X, Chen C et al. Interaction between mission, posted to the SEER website, April 2014.
tobacco and alcohol use and the risk of head and Huang CC, Lee WT, Tsai ST, Ou CY, Lo HI, Wong TY
neck cancer: pooled analysis in the International et al. Tea consumption and risk of head and neck
Head and Neck Cancer Epidemiology cancer. PLoS One 2014;9:e96507.
Consortium. Cancer Epidemiol Biomarkers Prev IARC. Alcohol Drinking. IARC Monographs on the
2009;18:541–​50. Evaluation of Carcinogenic Risks to Humans, Vol.
Hashibe M, Brennan P, Strange RC, Bhisey R, Cascorbi 44. Lyon: IARC, 1988.
I, Lazarus P et  al. Meta-​and pooled analyses of IARC. Coffee, Tea, Mate, Methylxanthines and Methyl­
GSTM1, GSTT1, GSTP1, and CYP1A1 geno- glyoxal. IARC Monographs on the Evaluation
types and risk of head and neck cancer. Cancer of Carcinogenic Risks to Humans, Vol. 51. Lyon:
Epidemiol Biomarkers Prev 2003;12:1509–​17. IARC Publications, 1991.
Hashibe M, Hunt J, Wei M, Buys S, Gren L, Lee YC. IARC. Tobacco Smoke and Involuntary Smoking. IARC
Tobacco, alcohol, body mass index, physical activ- Monographs on the Evaluation of Carcinogenic
ity, and the risk of head and neck cancer in the Risks to Humans. Vol. 83. Lyon: IARC Publications,
prostate, lung, colorectal, and ovarian (PLCO) 2004a.
cohort. Head Neck 2013;35:914–​22. IARC. Betel-​ Quid and Areca-​ Nut Chewing and
Hashibe M, McKay JD, Curado MP, Oliveira JC, Some Areca-​ Nut Related Nitrosamines. IARC
Koifman S, Koifman R et al. Multiple ADH genes Monographs on the Evaluation of Carcinogenic
are associated with upper aerodigestive cancers. Risks to Humans. Vol. 85. Lyon: IARC Publications,
Nat Genet 2008;40:707–​9. 2004b.
Hecht SS, Hoffman D. Tobacco specific nitrosamines, IARC. IARC Monograph on the Evaluation of
an important group of carcinogens in tobacco and Carcinogenic Risks to Humans. Vol. 90. Lyon:
tobacco smoke. Carcinogenesis 1988;9:875–​84. IARC Publications, 2007.
Hecht SS, Rivenson A, Braley J, DiBello J, Adams JD, IARC. Occupational Exposures in the Rubber
Hoffmann D. Induction of oral cavity tumors in Manufacturing Industry. IARC Monograph on the
F344 rats by tobacco-​specific nitrosamines and Evaluation of Carcinogenic Risks to Humans. Vol.
snuff. Cancer Res 1986;41:4162–​6. 100F. Lyon: IARC Publications, 2012.
Heck JE, Berthiller J, Vaccarella S, Winn DM, Smith Miller AB, Bartsch H, Boffetta P, Dragsted
EM, Shan’gina O et al. Sexual behaviours and the L, Vainio H, eds. Biomarkers in Cancer
risk of head and neck cancers:  a pooled analy- Chemoprevention. IARC Scientific Publication
sis in the International Head and Neck Cancer 154. Lyon: IARC, 2001.
154

154 Textbook of Cancer Epidemiology

Idris AM, Ibrahim SO, Vasstrand EN, Johannessen registry over 13  years. Auris Nasus Larynx
AC, Lillehaug JR, Magnusson B, et al. The Swedish 2015;42:401–​5.
snus and the Sudanese toombak:  are they differ- La Vecchia C, Bidoli E, Barra S, D’Avanzo B, Negri E,
ent? Oral Oncol 1998;34:558–​66. Talamini R et al. Type of cigarettes and cancers of
Iglesias-​
Bartolome R, Martin D, Gutkind JS. the upper digestive and respiratory tract. Cancer
Exploiting the head and neck cancer oncoge- Causes Control 1990;1:69–​74.
nome:  widespread P13K-​mTOR pathway altera- Laake I, Carlsen MH, Pedersen JI, Weiderpass E,
tions and novel molecular targets. Cancer Discov Selmer R, Kirkhus B et  al. Intake of trans fatty
2013;3:722–​5. acids from partially hydrogenated vegetable and
Jayant K, Deo MG. Oral cancer and cultural practices fish oils and ruminant fat in relation to cancer risk.
in relation to betel quid and tobacco chewing and Int J Cancer 2013;132:1389–​403.
smoking. Cancer Detect Prev 1986;9:207–​13. Lagiou P, Talamini R, Samoli E, Lagiou A, Ahrens W,
Kabat GC, Chanj CJ, Wynder EL. The role of tobacco, Pohlabeln H et  al. Diet and upper-​aerodigestive
alcohol use, and body mass index in oral and pha- tract cancer in Europe: the ARCAGE study. Int J
ryngeal cancer. Int J Epidemiol 1994;23:1137–​44. Cancer 2009;124:2761–​6.
Kabat GC, Hebert JR. Use of mentholated ciga- Larsson PA, Edström S, Westin T, Nordkvist A, Hirsch
rettes and oropharyngeal cancer. Epidemiology JM, Vahlne A. Reactivity against herpes simplex
1994;5:183–​8. virus in patients with head and neck cancer. Int J
Kabat GC, Wynder EL. Type of alcoholic beverage and Cancer 1991;49:14–​8.
oral cancer. Int J Cancer 1989;43:190–​4. Lassen P, Eriksen JG, Hamilton-​Dutoit S, Tramm T,
Katzel JA, Merchant M, Chaturvedi AK, Silverberg Alsner J, Overgaard J. Effect of HPV-​associated
MJ. Contribution of demographic and behav- p16INK4A expression on response to radiother-
ioral factors on changing incidence rates of oro- apy and survival in squamous cell carcinoma of
pharyngeal and oral cavity cancers in northern the head and neck. J Clin Oncol 2009;27:1992–​8.
California. Cancer Epidemiol Biomarkers Prev Lechner M, Frampton GM, Fenton T, Feber A, Palmer
2015;24:978–​84. G, Jay A et al. Targeted next-​generation sequenc-
Keukens F, van Voorst Vader PC, Panders AK, Vinks ing of head and neck squamous cell carcinoma
S, Oosterhuis JW et  al. Xeroderma pigmento- identifies novel genetic alterations in HPV+ and
sum: squamous cell carcinoma of the tongue. Acta HPV-​tumors. Genome Med 2013;5:49.
Derm Venereol 1989;69:530–​1. Lee PN, Hamling J. Systematic review of the relation
Khan Z, Tönnies J, Müller S. Smokeless tobacco and between smokeless tobacco and cancer in Europe
oral cancer in South Asia: a systematic review with and North America. BMC Med 2009;7:36.
meta-​analysis. J Cancer Epidemiol 2014;394696. Leemans CR, Braakhuis BJ, Brakenhoff RH. The
Kreimer AR, Clifford GM, Boyle P, Franceschi S. molecular biology of head and neck cancer. Nat
Human papillomavirus types in head and neck Rev Cancer 2011;11:9–​22.
squamous cell carcinomas worldwide:  a system- Leitzmann MF, Koebnick C, Freedman ND, Park Y,
atic review. Cancer Epidemiol Biomarkers Prev Ballard-​Barbash R, Hollenbeck AR et al. Physical
2005;14:467–​75. activity and head and neck cancer risk. Cancer
Kreimer AR, Johansson M, Waterboer T, Kaaks R, Causes Control 2008;19:1391–​9.
Chang-​Claude J, Drogen D et  al. Evaluation of Leoncini E, Edefonti V, Hashibe M, Parpinel M,
human papillomavirus antibodies and risk of Cadoni G, Ferraroni M et  al. Carotenoid intake
subsequent head and neck cancer. J Clin Oncol and head and neck cancer:  a pooled analy-
2013;31:2708–​15. sis in the International Head and Neck Cancer
Kreimer AR, Randi G, Herrero R, Castellsague X, La Epidemiology Consortium. Eur J Epidemiol
Vecchia C, Franceschi S; IARC Multicenter Oral 2016;31:369–​83.
Cancer Study Group. Diet and body mass, and Leoncini E, Ricciardi W, Cadoni G, Arzani D, Petrelli
oral and oropharyngeal squamous cell carcino- L, Paludetti G et al. Adult height and head and neck
mas: analysis from the IARC multinational case-​ cancer:  a pooled analysis within the INHANCE
control study. Int J Cancer 2006;118:2293–​7. Consortium. Eur J Epidemiol 2014;29:35–​48.
Kumar B, Cordell KG, Lee JS, Worden FP, Prince ME, Lieber CS, Seitz HK, Garro AJ, Worner TM. Alcohol-​
Tran HH, et al. EGFR, p16, HPV Titer, Bcl-​xL and related diseases and carcinogenesis. Cancer Res
p53, sex, and smoking as indicators of response to 1979;39:2844–​50.
therapy and survival in oropharyngeal cancer. J Lubin JH, Muscat J, Gaudet MM, Olshan AF, Curado
Clin Oncol 2008;26:3128–​37. MP, Dal Maso L et  al. An examination of male
Kuo CL, Chen YT, Shiao AS, Lien CF, Wang SJ. Acid and female odds ratios by BMI, cigarette smoking,
reflux and head and neck cancer:  a nationwide and alcohol consumption for cancers of the oral
15

Oral and Pharyngeal Cancer 155

cavity, pharynx, and larynx in pooled data from Negri E, Franceschi S, Bosetti C, Levi F, Conti
15 case-​control studies. Cancer Causes Control E, Parpinel M et  al. Selected micronutrients
2011;22:1217–​31. and oral and pharyngeal cancer. Int J Cancer
Lui VWY, Hedberg ML, Li H, Vangara BS, Pendleton 2000;86:122–​7.
K, Zeng Y et  al. Frequent mutation of the PI3K Nicolotti N, Chuang SC, Cadoni G, Arzani D, Petrelli
pathway in head and neck cancer defines predic- L, Bosetti C et  al. Recreational physical activity
tive biomarkers. Cancer Discov 2013;3:761–​9. and risk of head and neck cancer: a pooled anal-
Lustig JP, Lugassy G, Neder A, Sigler E. Head and ysis within the international head and neck can-
neck carcinoma in Franconi’s anaemia—​report of cer epidemiology (INHANCE) Consortium. Eur J
a case and review of the literature. Eur J Cancer B Epidemiol 2011;26:619–​28.
Oral Oncol 1995;31B:68–​72. Park SL, Lee YC, Marron M, Agudo A, Ahrens W,
MacFarlane GJ, Zheng T, Marshall JR, Boffetta P, Niu S, Barzan L et  al. The association between change
Brasure J et al. Alcohol, tobacco, diet and the risk of in body mass index and upper aerodigestive tract
oral cancer: a pooled analysis of three case-​control cancers in the ARCAGE project: multicenter case-​
studies. Oral Oncol Eur J Cancer 1995;31B:181–​7. control study. Int J Cancer 2011;128:1449–​61.
Maden C, Beckmann AM, Thomas DB, McKnight Peters WH, Wobbes T, Roelofs HM, Jansen JB.
B, Sherman KJ, Ashley RL et  al. Human papil- Glutathione S-​transferases in esophageal cancer.
lomaviruses, herpes simplex viruses and the Carcinogenesis 1993;14:1377–​80.
risk of oral cancer in men. Am J Epidemiol Pickering CR, Zhang J, Yoo SY, Bengtsson L, Moorthy
1992;135:1093–​1102. S, Neskey DM et  al. Integrative genomic charac-
Mahale P, Yanik EL, Engels EA. Herpes zoster and risk terization of oral squamous cell carcinoma iden-
of cancer in the elderly U.S.  population. Cancer tifies frequent somatic drivers. Cancer Discov
Epidemiol Biomarkers Prev 2016;25:28–​35. 2013;3:770–​81.
Marichalar-​ Mendia X, Rodriguez-​ Tojo MJ, Acha-​ Pira E, Pelucchi C, Romano C, Manzari M, Negri E, La
Sagredo A, Rey-​Barja N, Aguirre-​Urizar JM. Oral Vecchia C. Mortality from cancer and other causes
cancer and polymorphism of ethanol metabolis- in an Italian cohort of male rubber tire workers. J
ing genes. Oral Oncol 2010;46:9–​13. Occup Environ Med 2012;54:345–​9.
Marur S, D’Souza G, Westra WH, Forastiere AA. HPV-​ Posner MR, Lorch JH, Goloubeva O, Tan M,
associated head and neck cancer:  a virus-​related Schumaker LM, Sarlis NJ et  al. Survival and
cancer epidemic. Lancet Oncol 2010;11:781–​9. human papillomavirus in oropharynx cancer in
McLaughlin JK, Gridley G, Block G, Winn DM, TAX 324: a subset analysis from an international
Preston-​Martin S, Schoenberg JB et  al. Dietary phase III trial. Ann Oncol 2011;22:1071–​7.
factors in oral and pharyngeal cancer. J Natl Prime SS, Thakker NS, Pring M, Guest PG, Paterson
Cancer Inst 1988;80:1237–​43. IC. A review of inherited cancer syndromes and
Merchant A, Husain SS, Hosain M, Fikree FF, Pitiphat their relevance to oral squamous cell carcinoma.
W, Siddiqui AR et  al. Paan without tobacco:  an Oral Oncol 2001;37:1–​16.
independent risk factor for oral cancer. Int J Radoï L, Menvielle G, Cyr D, Lapôtre-​ Ledoux B,
Cancer 2000;86:128–​31. Stücker I, Luce D; ICARE Study Group. Population
Merletti F, Boffetta P, Ciccone G, Mashberg A, Terracini attributable risks of oral cavity cancer to behav-
B. Role of tobacco and alcoholic beverages in the ioral and medical risk factors in France: results of
etiology of cancer of the oral cavity/​oropharynx in a large population-​based case-​control study, the
Torino, Italy. Cancer Res 1989;49:4919–​24. ICARE study. BMC Cancer 2015;15:827.
Merletti F, Boffeta P, Ferro G, Pisani P, Terracini B. Radoï L, Paget-​Bailly S, Guida F, Cyr D, Menvielle G,
Occupation and cancer of the oral cavity or oro- Schmaus A et  al. Family history of cancer, per-
pharynx in Turin, Italy. Scand J Work Environ sonal history of medical conditions and risk of
Health 1991;17:248–​54. oral cavity cancer in France:  the ICARE study.
Nagao T, Ikeda N, Warnakulasuriya S, Fukano H, BMC Cancer 2013;13:560.
Yuasa H, Yano M et al. Serum antioxidant micro- Radoï L, Paget-​Bailly S, Menvielle G, Cyr D, Schmaus
nutrients and the risk of oral leukoplakia among A, Carton M, et  al. Tea and coffee consumption
Japanese. Oral Oncol 2000;36:466–​70. and risk of oral cavity cancer:  results of a large
Nagao T, Warnakulasuriya S, Nakamura T, Kato S, population-​based case control study, the ICARE
Yamamoto K, Fukano H et  al. Treatment of oral study. Cancer Epidemiol 2013;37:284–​9.
leukoplakia with a low-​dose of beta-​carotene and Rajkumar T, Sridhar H, Balaram P, Vaccarella S,
vitamin C supplements: a randomized controlled Gajalakshmi V, Nandakumar A et al. Oral cancer
trial. Int J Cancer 2015;136:1708–​17. in Southern India: the influence of body size, diet,
156

156 Textbook of Cancer Epidemiology

infections and sexual practices. Eur J Cancer Prev Shillitoe EJ, Greenspan D, Greenspan JS, Silverman S
2003;12:135–​43. Jr. Five-​year survival of patients with oral cancer
Ribeiro KB, Levi JE, Pawlita M, Koifman S, Matos and its association with antibody to herpes sim-
E, Eluf-​Neto J et  al. Low human papillomavirus plex virus. Cancer 1986;58:2256–​9.
prevalence in head and neck cancer: results from Silverberg MJ, Chao C, Leyden WA, Xu L, Horberg MA,
two large case-​control studies in high-​incidence Klein D et  al. HIV infection, immunodeficiency,
regions. Int J Epidemiol 2011;40:489–​502. viral replication, and the risk of cancer. Cancer
Richiardi L, Corbin M, Marron M, Ahrens W, Epidemiol Biomarkers Prev 2011;20:2551–​9.
Pohlabeln H, Lagiou P et al. Occupation and risk Soler M, Bosetti C, Franceschi S, Negri E, Zambon P,
of upper aerodigestive tract cancer: the ARCAGE Talamini R et al. Fiber intake and the risk of oral,
study. Int J Cancer 2012;130:2397–​406. pharyngeal and esophageal cancer. Int J Cancer
Rietbergen MM, Brakenhoff RH, Bloemena E, Witte 2001;91:283–​7.
BI, Snijders PJ, Heideman DA et al. Human pap- Sorahan T, Parkes HG, Veys CA, Waterhouse
illomavirus detection and comorbidity:  critical JA, Straughan JK, Nutt A. Mortality in the
issues in selection of patients with oropharyngeal British rubber industry 1946–​85. Br J Ind Med
cancer for treatment De-​ escalation trials. Ann 1989;46:1–​10.
Oncol 2013;11:2740–​5. Stich HF, Rosin MP, Vallejera MO. Reduction with
Rischin D, Young RJ, Fisher R, Fox SB, Le QT, Peters vitamin A  and beta-​ carotene administration
LJ et al. Prognostic significance of p16INK4A and of proportion of micronucleated buccal muco-
human papillomavirus in patients with oropha- sal cells in Asian betal nut and tobacco chewers.
ryngeal cancer treated on TROG 02.02 phase III Lancet 1984;1:1204–​6.
trial. J Clin Oncol 2010;28:4142–​8. Straif K, Weiland SK, Bungers M, Holthenrich D,
Rothman K, Keller A. The effect of joint exposure to Taeger D, Yi S et al. Exposure to high concentra-
alcohol and tobacco on risk of cancer of the mouth tions of nitrosamines and cancer mortality among
and pharynx. J Chronic Dis 1972;25:711–​16. a cohort of rubber workers. Occup Environ Med
Sancho-​ Garnier H, Theobald S. Black (air-​ cured) 2000;57:180–​7.
and blond (flue-​cured) tobacco and cancer risk. Stransky N, Egloff AM, Tward AD, Kostic AD,
II. Pharynx and larynx cancer. Eur J Cancer Cibulskis K, Sivachenko A et  al. The mutational
1993;29A:273–​6. landscape of head and neck squamous cell carci-
Saku T, Hayashi Y, Takahara O, Matsuura H, Tokunaga noma. Science 2011;333:1157–​60.
M, Tokunaga M et  al. Salivary gland tumors Szymańska K, Matos E, Hung RJ, Wünsch-​Filho V,
among atomic bomb survivors, 1950–​ 1987. Eluf-​Neto J, Menezes A et  al. Drinking of mate
Cancer 1997;79:1465–​75. and the risk of cancers of the upper aerodigestive
Sale KA, Wallace DI, Girod DA, Tsue TT. Radiation-​ tract in Latin America. Cancer Causes Control
induced malignancy of the head and neck. 2010;21:1799–​806.
Otolaryngol Head Neck Surg 2004;131:643–​5. Tabatabaeifar S, Kruse TA, Thomassen M, Larsen MJ,
Sapkota A, Hsu CC, Zaridze D, Shangina O, Szeszenia-​ Sorensen JA. Use of next generation sequencing
Dabrowska N, Mates D et al. Dietary risk factors in head and neck squamous cell carcinomas:  a
for squamous cell carcinoma of the upper aerodi- review. Oral Oncol 2014;50:1035–​40.
gestive tract in central and eastern Europe. Cancer Tesoriero JM, Gieryic SM, Carrascal A, Lavigne HE.
Causes Control 2008;19:1161–​70. Smoking among HIV positive New Yorkers: prev-
Sawant SS, Kandarkar SV. Role of vitamins C and E alence, frequency, and opportunities for cessation.
as chemopreventive agents in the hamster cheek AIDS Behav 2010;14:824–​835.
pouch treated with the oral carcinogen-​DMBA. Thompson DE, Mabuchi K, Ron E, Soda M, Tokunaga
Oral Dis 2000;6:241–​7. M, Ochikubo S et al. Cancer incidence in atomic
Schildt EB, Eriksson M, Hardell L, Magnuson A. Oral bomb survivors. Part II: Solid tumors, 1958–​1987.
infections and dental factors in relation to oral Radiat Res 1994;137(2 Suppl):S17–​67.
cancer: a Swedish case-​control study. Eur J Cancer van der Laan BF, Baris G, Gregor RT, Hilgers FJ, Balm
Prev 1998;7:201–​6. AJ. Radiation-​induced tumours of the head and
Schneider AB, Lubin J, Ron E, Abrahams C, Stovall neck. J Laryngol Otol 1995;109:346–​9.
M, Goel A et al. Salivary gland tumors after child- Weaver A, Fleming SM, Smith DB. Mouthwash and
hood radiation treatment for benign conditions of oral cancer:  carcinogen or coincidence? J Oral
the head and neck:  dose-​response relationships. Surg 1979;37:250–​3.
Radiat Res 1998;149:625–​30. Whatley WS, Thompson JW, Rao B. Salivary gland
Seitz HK, Poschl G, Simanowski UA. Alcohol and tumors in survivors of childhood cancer.
cancer. Recent Dev Alcohol 1998;14:67–​95. Otolaryngol Head Neck Surg 2006;134:385–​8.
157

Oral and Pharyngeal Cancer 157

Wight AJ, Ogden GR. Possible mechanisms by which Ye Z, Song H, Guo Y. Glutathione S-​transferase M1,
alcohol may influence the development of oral T1 status and the risk of head and neck cancer: a
cancer—​a review. Oral Oncol 1998;34:441–​7. meta-​analysis. J Med Genet 2004;41:360–​5.
World Cancer Research Fund in association with Yokoyama A, Muramatsu T, Omori T, Yokoyama T,
American Institute for Cancer Research. Food, Matsushita S, Higuchi S et al. Alcohol and aldehyde
Nutrition and the Prevention of Cancer: A Global dehydrogenase gene polymorphisms and oropha-
Perspective. Washington, DC, American Institute ryngolaryngeal, esophageal and stomach cancers in
for Cancer Research, 1997: 96–​106. Japanese alcoholics. Carcinogenesis 2001;22:433–​9.
World Cancer Research Fund/​ American Institute Zhang ZJ, Hao K, Shi R, Zhao G, Jiang GX, Song Y
for Cancer Research. Food, Nutrition, Physical et al. Glutathione S-​transferase MI (GSTM1) and
Activity, and the Prevention of Cancer: A Global glutathione S-​transferase TI (GSTT1) null poly-
Perspective. Washington DC: AICR, 2007. WHO morphisms, smoking, and their interaction in oral
South-​East Asia Region Press release 1563 2013, cancer:  a HuGE review and meta-​analysis. Am J
http://​www.searo.who.int/​mediacentre/​releases/​ Epidemiol 2011;173:847–​57.
2013/​pr1563/​en/​. Zheng TZ, Boyle P, Hu HF, Duan J, Jiang PJ, Ma DQ
Xu J, Yang XX, Wu YG, Li XY, Bai B. Meat consump- et al. Tobacco smoking, alcohol consumption and
tion and risk of oral cavity and oropharynx can- risk of oral cancer: a case-​control study in Beijing,
cer: a meta-​analysis of observational studies. PLoS People’s Republic of China. Cancer Causes
One 2014;9:e95048. Control 1990;1:173–​9.
Xu L, Tang H, El-​ Naggar AK, Wei P, Sturgis Zheng TZ, Boyle P, Willett WC, Hu H, Dan J, Evstifeeva
EM. Genetic variants in DNA double-​ strand TV et  al. A case-​control study of oral cancer in
break  repair genes and risk of salivary gland Beijing, People’s Republic of China. Associations
carcinoma: a case-​ control study. PLoS One with nutrient intakes, foods and food groups. Eur
2015;10:e0128753. J Cancer B Oral Oncol 1993;29B:45–​55.
158
159

8
Nasopharyngeal Cancer
E L L E N T. C H A N G A N D H A N S - O
​ LOV ADAMI

N asopharyngeal carcinoma (NPC) is endemic

in indigenous populations in East and
Southeast Asia, the Arctic, North Africa, and the
rare subtype constituting less than 0.2% of NPC,
in the 2005 WHO classification (Chan et  al,
2005). In high-​incidence areas, nearly all NPC
Middle East. Autopsy records show that NPC is of undifferentiated nonkeratinizing histol-
was prevalent in China by the early 20th century, ogy and most of the remainder is differentiated
and ancient Chinese medical literature describes keratinizing histology, whereas in low-​incidence
the existence of a disease resembling NPC sev- areas, as much as half of NPC is squamous cell
eral hundred years ago (Ho, 1976). Descriptive carcinoma (Vaughan et  al, 1996). In the future,
studies of ancient Egyptian mummies (Wells, updated histological classification systems are
1963)  and Iranian skulls (Krogman, 1940)  also expected to be based on tumor molecular profile
show bone changes that some have attributed to and may more explicitly incorporate information
NPC. Hence, genetic and/​or stable environmen- on disease prognosis (Wang et al, 2011).
tal risk factors may have persisted for centuries in Symptoms Most NPC patients present with a
NPC-​endemic regions and populations. Genetic cervical neck mass after the tumor has reached
and anthropological studies suggest that popula- a locally advanced disease stage with nodal
tions with elevated NPC risk may share common involvement. Patients may also present with local
ancestry with the aboriginal Bai-​Yue people of nasal, oral, or auricular symptoms. Symptoms of
southeastern China, who were dispersed by the more advanced disease can include cranial nerve
last ice age (Wee et al, 2010). The distinctive geo- palsy and neuropathies.
graphic distribution of NPC and the strong evi-
dence that it results from the interplay of genetic, Diagnosis Clinical examination may be followed
environmental, and infectious causes make it by computed tomography or magnetic resonance
a fascinating topic of epidemiologic research. imaging of the head and neck area. An endoscopic
However, despite many years of seeking to under- or fine needle aspiration biopsy from the pri-
stand the causes and prevention of NPC, with mary tumor is performed for definitive diagnosis.
substantial advances over recent decades, many Additional imaging studies are also indicated for
aspects of this disease remain enigmatic. evaluation of local or distant disease spread.
Treatment NPC is sensitive to both radiation
CLINICAL SYNOPSIS therapy and chemotherapy. The current standard
Subgroups NPC, a carcinoma arising in the epi- of care for early-​stage NPC is radiotherapy alone,
thelial lining of the nasopharynx, makes up the which has a high probability of cure, especially
vast majority of cancers of the nasopharynx. with the use of intensity-​modulated radiotherapy
Since the 1970s, the World Health Organization (IMRT) (Zhang et al, 2013). The standard of care
(WHO) has classified NPC histologically as for locoregionally advanced disease, which con-
keratinizing squamous cell carcinoma (type I) stitutes the majority of patients at presentation, is
or nonkeratinizing carcinoma, which is further cisplatin-​based concurrent chemoradiotherapy.
classified as differentiated (type II) or undiffer- Patients with distant metastases may be treated
entiated (type III) carcinoma (Shanmugaratnam, with chemotherapy and possibly definitive IMRT
1978). Since then, few changes have been made in and/​or immunotherapy targeting the Epstein
the histologic classification of NPC, including the Barr virus (EBV), which is present in virtually all
addition of basaloid squamous cell carcinoma, a nonkeratinizing NPC tumors. Surgery is usually
160

160 Textbook of Cancer Epidemiology

precluded by the tumor’s location in a remote levels of circulating EBV DNA (Lin et  al, 2004;
anatomic location, although nodal dissection or Wang et al, 2013).
excision can be performed in cases of persistent Progress The responsiveness of NPC to both
or recurrent regional nodal involvement. radiotherapy and chemotherapy distinguishes
Prognosis Among 2,687 NPC patients treated at it from other head and neck cancers, which
public oncology centers in Hong Kong between are typically insensitive to chemotherapy. The
1996 and 2000, 5-​year overall survival was 90% advent of IMRT and combined chemoradiother-
for stage I, 84% for stage II, 75% for stage III, apy regimens has appreciably improved survival
and 58% for stage IV disease (Lee et  al, 2005). with NPC in recent years. Standardization of
More recently, in 2001–​ 2008, 5-​ year survival the staging system, refinement of the histologi-
rates were higher (100%, 94%, 84%, and 71% for cal classification system to distinguish between
stages I, II, III, and IV a-​b, respectively) among prognostically distinct subtypes, better charac-
868 southern Chinese NPC patients treated terization of patients who would benefit from
with IMRT at an academic center (Sun et  al, adjuvant chemotherapy, and development of tar-
2014). These relatively favorable survival figures geted therapies will improve the precision and
probably are not representative of typical treat- efficacy of tailored treatments for different risk
ment outcomes, as indicated by the worldwide groups.
mortality-​to-​incidence-​rate ratio of 57% (Ferlay
et al, 2013) (see “Descriptive Epidemiology”) and DESCRIPTIVE
lower stage-​specific survival rates of 60%–​90% EPIDEMIOLOGY
for stage I, 57%–​79% for stage II, 44%–​77% for Although NPC is a rare malignancy through-
stage III, and 17%–​43% for stage IV elsewhere out most of the world, it is endemic in a few
in Asia (Dou et al, 2014; Leung et al, 2005; Phua well-​defined populations. In most regions, the
et al, 2011; Phua et al, 2013). In the United States, age-​
standardized incidence rate of NPC for
corresponding 5-​year overall survival rates by both males and females is below 1 per 100,000
stage for NPC patients diagnosed between 2004 person-​years (Figs. 8-​1 and 8-​2). However, dra-
and 2007 were 74.1%, 74.5%, 61.6%, and 36.9%, matically elevated rates are observed in the
respectively, indicating some overlap between Cantonese population of southern China, and
stages I and II (SEER, 2015). In addition to dis- intermediate rates are observed in several indig-
ease stage, other predictors of poorer prognosis enous populations in Southeast Asia, the Arctic
are male sex, undifferentiated nonkeratinizing region, North Africa, and the Middle East. Even
histology, and elevated pre-​and post-​treatment within China, more than 20-​fold variation in

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 0,2 0,4 0,6 0,8 1 1,2

ASR (world) per 100,000 person-years

FIGURE 8-​1  Age-​standardized (to the 2012 world population) incidence rates of nasopharyngeal cancer among women.
Source: Ferlay et al, 2013.
16

Nasopharyngeal Cancer 161

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 0,5 1 1,5 2 2,5 3

ASR (world) per 100,000 person-years

FIGURE 8-​2  Age-​standardized (to the 2012 world population) incidence rates of nasopharyngeal cancer among men.
Source: Ferlay et al, 2013.

NPC incidence exists across regions, with rates was the 10th most common new malignancy
generally increasing from northern China (e.g., in Southeast Asia and the 5th most common in
Beijing) to southern China (e.g., Hong Kong, Malaysia (Ferlay et al, 2013).
Macao, and Zhongshan) (Fig. 8-​3) (Forman et al, In almost all populations surveyed, the inci-
2013). In 2012, approximately 87,000 incident dence of NPC is two-​to threefold higher in males
cases of nasopharyngeal cancer were diagnosed than in females (Figs. 8-​1 and 8-​2). In low-​risk
worldwide and the estimated number of deaths populations, NPC incidence shows a minor peak
exceeded 50,000, making it the 24th most com- in young adulthood (ages 15–​24 years), plateaus
mon new cancer in the world; in contrast, NPC or declines slightly until age 35–​39  years, then

FIGURE 8-​3 Age-​standardized (to the world population) incidence rates of nasopharyngeal cancer among males and
females in China, 2003–​2007. Rates shown (listed for males and then females) are per 100,000 person-​years.
Source: Data from Forman et al, 2013.
162

162 Textbook of Cancer Epidemiology

New cases per 100,000

4

person-years
3

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 8-​4  Age-​specific incidence rates of nasopharyngeal cancer among women.

Source: Ferlay et al, 2013.

rises to age 65–​79  years (Bray et  al, 2008). The curve with a relatively early peak in high-​risk
early incidence peak is more apparent in com- populations suggest etiologic heterogeneity of
bined data (Bray et  al, 2008)  than individual NPC by region or histologic subtype, which
populations (Figs. 8-​4 and 8-​5), and suggests is closely linked to geography. However, it has
the involvement of genetic susceptibility and/​or been hypothesized that a bimodal age-​incidence
exposure to carcinogenic agents in early develop- pattern is present in high-​ risk populations,
ment, whereas the later peak may be due to long-​ but that young-​adulthood peak is obscured by
term environmental exposures in adulthood. In genetic and epigenetic risk factors for NPC (Bray
contrast, in high-​risk groups, NPC incidence et al, 2008).
rises monotonically to a peak at around ages Although geographic regions have generally
45–​59 years, then plateaus or declines modestly been classified as high-​or low-​incidence areas,
thereafter (Figs. 8-​4 and 8-​5), consistent with the the racial/​ethnic distribution of NPC within
involvement of EBV infection and other expo- regions is far from uniform. In the southeastern
sures in childhood. The bimodal age-​incidence Chinese province of Guangdong, where the over-
curve in low-​risk populations and the unimodal all NPC incidence rate is above 20 per 100,000

10
New cases per 100,000

9
8
person-years

7
6
5
4
3
2
1
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United states of America

FIGURE 8-​5  Age-​specific incidence rates of nasopharyngeal cancer among men.

Source: Ferlay et al, 2013.
163

Nasopharyngeal Cancer 163

person-​years among males, historical rates in that period (Tse et  al, 2006). In Taiwan, NPC
Cantonese-​speakers were double those in other incidence has decreased since the early 1980s
dialect groups such as the Hakka, Hokkien, and (Hsu et al, 2006), and in Singapore Chinese it
Chiu Chau, and highest among the Tanka boat has decreased since the mid-​1990s (National
people (Li et al, 1985). In Southeast Asia, NPC Registry of Diseases Office, 2013; National
risk appears to vary with degree of racial and Registry of Diseases Office, 2009). The lag in
social admixture with southern Chinese, espe- incidence trends may be attributable to differ-
cially historically with the Bai-​Yue people, who ences in the timing of the onset of rapid eco-
are ancestors of the Tanka (Wee et al, 2010). In nomic development, which occurred in the
Singapore, NPC incidence is low among Indians, mid-​1940s in Hong Kong, the 1950s in Taiwan,
who have little intermingling with Chinese; and in the 1960s in Singapore. In contrast,
intermediate among Malays, who have a his- the incidence rate of NPC did not change or
tory of intermarriage with Chinese; and highest even increased over recent decades among
among Chinese (Forman et al, 2013; Ho, 1976). males and females in Sihui County, Cangwu
Close ties have existed between Japan and China County, and Zhongshan City in southeast-
for thousands of years—​but mainly with north- ern China, where economic development has
ern China, where NPC is not endemic—​and occurred later (Jia et al, 2006; Wei et al, 2010).
the incidence of NPC in Japan is correspond- Between 1973 and 2012, the overall incidence
ingly low (Forman et al, 2013). NPC rates in the rate of NPC in the United States remained sta-
United States are highest among Chinese, fol- ble at around 1 per 100,000 in males and 0.4
lowed by Filipinos, then blacks, Koreans, non-​ per 100,000 in females (SEER, 2015). However,
Hispanic whites, Japanese, and Hispanic whites among Asian and Pacific Islander Americans,
(Forman et al, 2013). the incidence rate of NPC decreased signifi-
When high-​or intermediate-​ risk persons cantly by 2.2% per year in males and 2.4% per
migrate to lower-​risk countries, their incidence year in females between 1992 and 2012 (SEER,
of NPC remains higher than those of other races, 2015). At least the earlier half of this decrease
but markedly lower than in their native coun- (1992–​2002) appeared to be restricted to squa-
tries (Mousavi et  al, 2010; Parkin & Iscovich, mous cell NPC, but not differentiated or undif-
1997; Yu & Hussain, 2009). Conversely, risk of ferentiated nonkeratinizing NPC (Sun et  al,
NPC increases among males of low-​ risk par- 2005). No striking increase in NPC incidence
entage born in China, the Philippines (Buell, was noted in parallel with the onset of the HIV
1973), or North Africa (Jeannel et al, 1993). The epidemic and the ensuing era of highly active
decline in NPC incidence among Chinese after antiretroviral treatment.
migration to the West may be somewhat over-
estimated, since reported rates do not account GENETIC AND MOLECULAR
for the inclusion of both high-​risk and low-​risk EPIDEMIOLOGY
Chinese in the source population. Furthermore, A genetic component of NPC risk has long been
migrants are a self-​selected group and unlikely to indicated by the familial aggregation of the dis-
be representative of the residents of their native ease, and has been borne out more recently by
country. Because certain aspects of a traditional genetic studies that are remarkably consistent
Asian lifestyle are associated with elevated risk of in identifying an important role of the human
NPC, individuals who migrate overseas may be leukocyte antigen (HLA) genes in conferring
an inherently lower-​risk group. Thus, NPC inci- disease risk.
dence rates among migrants generally are not
directly comparable to those among natives of Inherited Susceptibility
their country of origin. Strong familial susceptibility to NPC is well
After several decades of high rates in established. In epidemiologic studies, the excess
Hong Kong, NPC incidence has declined risk of NPC is generally 4-​to 20-​fold among
steadily since the 1970s (Forman et  al, 2013). individuals who have a first-​degree relative with
However, the decrease between 1988 and 2002 NPC, compared with those lacking such a fam-
(at least) was limited primarily to squamous ily history (Friborg et al, 2005; Ji et al, 2011; Jia
cell NPC, whereas the rates of nonkeratiniz- et al, 2004; Zeng & Jia, 2002). In our 2010–​2014
ing NPC remained relatively stable during population-​ based case-​ control study of NPC
164

164 Textbook of Cancer Epidemiology

in southern China, including approximately of 20 Cantonese-​ speaking southern Chinese

2,500 cases and 2,500 controls, we found that families with multiple affected members identi-
the cumulative risk of NPC for males with an fied strong linkage to the D423002 marker on
affected relative was 5.0% (95% CI 3.6%–​7.0%) 4p12-​p15 (logarithm of odds for linkage [LOD]
and that for females was 1.9% (95% CI 1.1%–​ score = 4.2) (Feng et al, 2002). Follow-​up of this
1.2%) (Liu et al, 2017). Among relatives of cases, finding identified a functional variant in the pro-
the lifetime cumulative risk of NPC was 6.3% moter region of a novel gene, LOC344967, that
in brothers, 3.5% in sisters, 3.4% in fathers, might be responsible for the observed famil-
and 2.5% in mothers). An excess of NPC has ial susceptibility to NPC (Jiang et  al, 2006).
not been observed in any of the known inher- A  pedigree-​based linkage study of 18 high-​risk
ited cancer syndromes (Garber & Offit, 2005). southern Chinese families identified another
Instead, common genetic variation, particularly possible locus at D3S3624 on 3p21.31-​21.2 (LOD
in genes that play a role in the immune response score = 4.18) (Xiong et al, 2004), while another
to EBV infection or the metabolism of environ- pedigree study of 15 multiplex families from
mental carcinogens, is likely to be an important southern China found suggestive linkage to a
causal factor in NPC development. locus at D5S2021 on 5p13.1 (LOD score  =  2.1)
Familial aggregation of NPC has been widely (Hu et  al, 2008). The nonconcordant results of
documented in high-​, intermediate-​, and low-​ these four linkage studies might be due to dis-
incidence populations (Bei et al, 2012; Zeng & Jia, similar methods, population heterogeneity,
2002). For example, 15 of 109 family members differences in disease pathology, insufficient sta-
across four generations of a southern Chinese tistical power, or chance (Bei et al, 2012). Further
family were found to have developed NPC (Zhang sequencing, functional characterization, and rep-
& Zhang, 1999). The excess familial cancer risk lication studies are needed to better understand
appears to be limited to NPC, as risk of other these findings.
malignancies has been found not to be increased
in NPC families in high-​incidence regions (Jia Human Leukocyte Antigen Genes
et  al, 2004; Yu et  al, 2009). However, studies in Associations of NPC risk with certain HLA alleles,
Greenland natives reported an increased risk of which encode proteins required for antigen pres-
cancers of the salivary gland and uterine cervix entation to the immune system and other aspects
in NPC families, possibly suggesting familial sus- of immune function, have been reported since the
ceptibility to virus-​ associated malignancies or 1970s. It has long been hypothesized that indi-
shared exposure to environmental risk factors viduals who inherit HLA alleles with a reduced
(Albeck et al, 1993; Friborg et al, 2005). Despite ability to present EBV antigens to immune cells
the high relative risk (RR) of NPC in families, may have an increased risk of developing NPC,
familial clustering does not explain the majority whereas individuals with HLA alleles that pres-
of NPC cases, and a complex segregation analy- ent EBV efficiently may have a lower risk (Simons
sis of familial NPC in southern China concluded et al, 1975). In early studies in southern Chinese
that multiple genetic and environmental factors, and Southeast Asian populations, HLA-​A2-​B46
rather than a single major susceptibility gene, are and B17 were associated with a two-​to threefold
most likely to explain the observed pattern of increase in NPC risk (Simons et al, 1978; Simons
inheritance (Jia et al, 2005). et al, 1974). In contrast, 30%–​50% lower risk of
Familial linkage studies are used to map NPC was found in association with HLA-​A11
regions of the genome that are shared by affected in both Chinese and whites (Burt et  al, 1996;
relatives and therefore may confer disease sus- Simons et al, 1978), B13 in Chinese (Chan et al,
ceptibility. Four linkage studies of NPC have 1983), and A2 in Whites (Burt et al, 1994). More
identified putative susceptibility loci based on detailed HLA typing revealed a positive associa-
affected siblings or extended pedigrees (Feng tion with HLA-​A*0207 in Asians and an inverse
et al, 2002; Hu et al, 2008; Lu et al, 1990; Xiong association with HLA-​A*02:01 in Caucasians
et al, 2004). The first, a study of 30 affected sib- (Hildesheim et al, 2002).
lings from southern China and Southeast Asia, The high density of genes and strong linkage
focused on the highly polymorphic HLA region disequilibrium in the HLA region have thus far
of immune function genes, and identified a impeded the identification of specific causal vari-
locus at 6p22 that conferred an RR of 20.9 (Lu ants (Su et al, 2013). However, positive associa-
et  al, 1990). A  pedigree-​ based linkage study tions with HLA-​A*0207 (in linkage disequilibrium
165

Nasopharyngeal Cancer 165

with HLA-​B*4601) and HLA-​B*5801 (in link- A  meta-​ analysis (Huang et  al, 2011)  of five
age disequilibrium with HLA-​A*3303) and an case-​control studies of coding SNPs in the x-​
inverse association with HLA-​A*1101 (in link- ray repair cross-​complementing group 1 gene
age disequilibrium with HLA-​B*13) have been (XRCC1), including four studies in Asians and
detected consistently (Bei et al, 2012; Goldsmith one in North Africans, found no significant asso-
et  al, 2002; Hildesheim & Wang, 2012). Other ciation between the Arg194Trp or Arg280His
HLA class  I  genes (HLA-​C, -​E, -​G, and others) polymorphism and NPC risk, and a marginally
and class II genes (HLA-​DR, -​DP, and -​DQ) have significant association with Arg399Gln (odds
not consistently been associated with NPC risk. ratio [OR]  =  1.28, 95% CI 1.00–​1.65). A  study
Although linkage disequilibrium with other, of 1,052 NPC cases and 1,168 controls from
non-​HLA genes in the same region on 6p21.3 eastern and southern China reported a signifi-
cannot be ruled out, a causal interpretation of cant increase in risk associated with rs1805794
associations with HLA class I genes is supported in NBS1, a component of the MER11/​RAD50/​
by the biologic plausibility of the hypothesis that NBS1 (MRN) complex involved in double-​strand
binding and presentation of intracellular (includ- break repair (ORGC  =  1.69, 95% CI 1.26–​2.11;
ing viral) antigens to cytotoxic T cells is involved ORCC = 3.87, 95% CI 3.01–​5.03)—​a finding sup-
in NPC susceptibility. ported by in vitro cell migration studies (Zheng
et al, 2011). Another large study genotyped 676
Candidate-​Gene Studies haplotype-​tagging SNPs covering 88 DNA repair
The preponderance of epidemiologic research on genes in 755 matched case-​control pairs from
NPC in the past decade has been directed toward southern China, then validated 11 significant
identifying genetic associations. Most studies SNPs in a validation stage with 1,568 cases and
have been relatively modest in size and focused 1,297 controls (Qin et  al, 2011b). Two SNPs in
on a small number of candidate polymorphisms, strong linkage disequilibrium in the DNA repair
thereby limiting their ability to illuminate the protein RAD51 homolog 2 withstood valida-
genetic pathways underlying NPC development. tion (OR for rs927220 = 1.20, 95% CI 1.10–​1.30,
Consequently, causal associations have not yet Pcombined  =  5.55 × 10–​5); variants in XRCC1 and
been established. other DNA repair genes were not significantly
Motivated by the associations of NPC risk with associated.
nitrosamines from diet or tobacco smoking and A polymorphism in codon 72 of the tumor
possibly with other inhalant carcinogens, several suppressor gene TP53, which governs cell-​cycle
studies have evaluated genetic associations with arrest, has been investigated in at least seven
phase I  and II xenobiotic metabolism enzymes, case-​control studies of NPC (Cai et  al, 2014).
including the glutathione-​ S-​
transferase (GST) Based on a meta-​analysis of these studies, includ-
enzyme family. Five meta-​analyses have sum- ing a total of 1,133 NPC cases and 1,678 controls,
marized the results of up to 15 NPC case-​control the combined OR for rs1042522 (Arg72Pro) was
studies of GSTM1 and GSTT1 polymorphisms 1.32 (95% CI 1.18–​1.47) for the variant allele and
that result in decreased levels of detoxifying 1.90 (95% CI 1.51–​2.39) for the homozygous var-
enzymes (Jin et al, 2014; Murthy et al, 2013; Sun iant, with no significant heterogeneity between
et al, 2012; Wei et al, 2013; Zhuo et al, 2009). The Caucasians and Asians.
combined OR for the GSTM1 null genotype was
significantly increased in all analyses and ranged Genome-​Wide Association
between 1.42 and 1.54, while that for the GSTT1 Studies
null genotype was significantly increased in Five genome-​wide association studies (GWAS)
three of four analyses and ranged between 1.12 of NPC, each evaluating 400,000–​600,000 sin-
and 2.27, with the higher OR estimates coming gle nucleotide polymorphisms (SNPs), have been
from meta-​analyses (Jin et  al, 2014; Wei et  al, conducted to date (Bei et  al, 2010; Chin et  al,
2013)  that included twice the number of stud- 2015; Ng et al, 2009; Tang et al, 2012; Tse et al,
ies included in the meta-​analyses with lower OR 2009). In 2009, the first NPC GWAS, with a dis-
estimates (Murthy et al, 2013; Zhuo et al, 2009). covery set of 111 Malaysian Chinese NPC cases
DNA repair genes may be capable of coun- and 260 controls, identified a significant associ-
teracting the DNA-​damaging effects of EBV or ation (OR = 2.24, Pcombined = 8.27 × 10–​7) with the
environmental carcinogens and their reactive SNP rs2212020 in the integrin-​α 9 gene at 3p21
intermediates on the nasopharyngeal epithelium. (Ng et  al, 2009). Shortly thereafter, a GWAS of
16

166 Textbook of Cancer Epidemiology

562 cases/​controls from Taiwan in the discovery of the HLA-​A region revealed associations in the
phase and 2,275 in the replication phase identi- antigen peptide binding groove, the T-​cell recep-
fied two significant susceptibility loci (rs2517713 tor binding site, the 5’-​UTR, and other loci, these
and rs2975042; OR = 1.88, Pcombined = 3.9 × 10–​20; associations were eliminated after controlling for
and OR = 1.86, Pcombined = 1.6 × 10–​19, respectively) the effects of HLA-​A*11:01. A  meta-​analysis of
in the HLA-​A region at 6p21.3 (Tse et al, 2009). 2,152 cases and 3,740 controls from four of these
After adjusting for the HLA-​A SNPs, an addi- NPC GWAS confirmed many of the previously
tional association was identified with rs29232 in detected associations and additionally found a
the gamma-​aminobutyric acid b receptor 1 gene novel association between genetic variation in
(GABBR1; OR  =  1.67, Pcombined  =  8.97 × 10–​17, the CLPTM1L/​TERT region, which is involved in
Presidual < 5 × 10–​4), also located at 6p21.3. Except telomere length maintenance, and NPC risk (Bei
for the association with GABBR1, most of the et al, 2016).
observed associations in this study were due to
previously identified HLA-​A associations (Hsu Somatic Changes
et al, 2012b). Studies of loss of heterozygosity in NPC tumors
Subsequently, a GWAS with 3,477 discovery-​ detected a high frequency of allelic loss, especially
phase subjects and 6,570 replication-​phase sub- on chromosomes 3p, 9p, 11q, 13q, and 14q (Chen
jects from Guangdong and Guangxi in southern et  al, 1999; Hui et  al, 1999; Mutirangura et  al,
China identified three new susceptibility loci in 1997). Such findings suggest that tumor suppres-
the tumor necrosis factor receptor superfamily, sor genes at these loci may be involved in NPC
member 19, on 13q12 (rs9510787, OR  =  1.20, development. A  meta-​ analysis of comparative
Pcombined = 1.53 × 10–​9); the myelodysplasia 1 and genomic hybridization results revealed several
ecotropic viral insertion site 1 fusion proteins genomic hot spots where chromosomal losses
on 3q26 (rs6774494, OR  =  0.84, Pcombined  =  1.34 (3p, 8p, 11q, 13q, 14q, and 16q) and gains (1q,
× 10–​8); and the cyclin-​dependent kinase inhibi- 2q, 3q, 6q, 8q, 11q, 12p, 12q, 17p, and 17q) have
tors 2A and 2B on 9p21 (rs1412829, OR = 0.78, been detected at >15% frequency in NPC tumors
Pcombined = 4.84 × 10–​7) (Bei et al, 2010). This study (Li et  al, 2006b). Intensive screening in regions
also identified significant loci in HLA-​A, HLA-​ of copy amplification and copy number gain has
B/​C, and HLA-​DQ/​DR (rs2860580, OR  =  0.58, identified putative NPC-​ associated oncogenes,
Pcombined  =  4.88 × 10–​67; rs2894207, OR  =  0.61, while studies in regions of allelic deletions have
95% CI  =  0.57–​ 0.66, Pcombined  =  3.42 × 10–​33; identified potential NPC-​associated tumor sup-
rs28421666, OR  =  0.67, Pcombined  =  2.49 × 10–​18, pressor genes (Lo et al, 2012).
respectively). Somatic mutations of tumor suppressor genes,
Another NPC GWAS that included 1,043 such as TP53, pRb, p16, and VHL, appear to be
southern Chinese NPC subjects in the discov- relatively infrequent in NPC (Zeng & Jia, 2002).
ery phase and replicated findings in 985 subjects, However, aberrant promoter CpG hypermethyl-
confirmed the essential role of HLA by identifying ation of tumor suppressor genes or hypomethyl-
significant associations with HLA-​A (amino acid ation of proto-​oncogenes appears to be another
62, Pcombined = 7.38 × 10–​29), HLA-​B (amino acid means of inactivating gatekeeper genes in NPC
116, Pcombined = 6.51 × 10–​19), and HLA-​C (amino tumors (Li et al, 2011). Methylation studies using
acid 156, Pcombined = 6.84 × 10–​8) (Tang et al, 2012). nasopharyngeal brushing samples (Hutajulu
Follow-​up with high-​resolution HLA molecular et  al, 2011)  or cell-​free circulating DNA (Tian
typing revealed strong independent associations et al, 2013) indicate that hypermethylation of the
with HLA-​A*11:01 and with the HLA-​B*13:01, same tumor suppressor genes is also observed in
B*38:02, and B*55:02 alleles together. nontumor biospecimens, suggesting a possible
Finally, the association with HLA-​A, driven basis for early disease detection.
largely by an inverse association with HLA-​
A*11:01 and to a lesser extent by HLA-​A*02:07, RISK FACTORS
was again confirmed in the most recent NPC Few environmental risk factors for NPC are well
GWAS, based on 184 Malaysian Chinese NPC established; these include tobacco smoking, die-
cases and 236 controls in the discovery phase tary consumption of salt-​preserved fish, and EBV
and 260 cases and 245 controls in the replication infection. Identification of other risk factors has
phase (rs3869062, OR  =  2.37, Pcombined  =  1.73 × been hampered in part by the scarcity of rigor-
10–​9) (Chin et al, 2015). Although fine mapping ously designed, population-​ based case-​ control
167

Nasopharyngeal Cancer 167

studies of NPC, and even fewer informative Africa found that use of a compact charcoal oven
cohort studies. or a wood fire for cooking (but not heating) and
a lack of kitchen ventilation in childhood were
Tobacco and Other Smoke associated with 29%–​86% higher risk of NPC,
The International Agency for Research on Cancer but the same exposures in adulthood were not
(IARC) panel concluded that there is sufficient associated with risk (Feng et  al, 2009). Studies
evidence that tobacco smoking causes cancers examining burning incense or antimosquito coils
of the nasopharynx in humans (IARC, 2012d, and risk of NPC have also been equivocal, with
2004b). However, it remains unclear whether two studies finding up to a sixfold excess risk
tobacco smoking equally affects the develop- with use of antimosquito coils (Shanmugaratnam
ment of both nonkeratinizing and squamous et al, 1978; West et al, 1993), and one finding a
cell NPC. In our population-​based case-​control higher risk among individuals with religious
study of NPC in southern China, we found that altars at home (Geser et al, 1978), but other stud-
NPC risk was significantly elevated among men ies finding no association (Yu et al, 1990; Yu et al,
who currently smoked compared with those who 1986; Yu et al, 1988).
had never smoked (OR = 1.32, 95% CI 1.14–​1.53) Betel nut (Areca catechu) is a psychoactive
(Chang et al, 2017). Risk increased with smok- stimulant that is commonly chewed as betel quid
ing intensity (OR per 10 cigarettes/​day = 1.09, (with betel leaf, areca nut, and other spices or
95% CI 1.03–​1.16), smoking duration (OR per 10 ingredients, with or without tobacco) and is clas-
years = 1.11; 1.06, 1.16), and cumulative smok- sified by IARC a cause of oral cancer in humans
ing (OR per 10 pack-​years = 1.08; 1.04, 1.12), and (IARC, 2004a), but it has not been shown to have
decreased with later age at smoking initiation a causal effect on NPC. Betel nut use was not sig-
(OR per year = 0.97; 0.96, 0.98). These results nificantly associated with overall NPC risk in
are consistent with findings from a meta-​analy- three case-​control studies (Fachiroh et  al, 2012;
sis of 28 case-​control studies and four prospec- West et  al, 1993; Yang et  al, 2005), although a
tive cohort studies that yielded a meta-​RR of 1.60 positive association was detected for late-​onset
(95% CI 1.38–​1.87) for ever versus never smok- (≥ 40  years) familial NPC in Taiwan (Yang
ing, with positive trends by cumulative amount, et al, 2005).
intensity, and duration of smoking, but not age A hospital-​based case-​control study in North
at initiation (Xue et al, 2013). We also observed Africa detected a twofold increased risk of NPC
excess NPC risk associated with passive smoking in cannabis smokers, independent of tobacco
during childhood (OR = 1.24, 95% CI 1.03–​1.48) smoking, with positive trends by frequency of use
and spousal passive smoking during adulthood and cumulative consumption (Feng et al, 2009).
(OR = 1.30, 95% CI 1.03–​1.63) (Chang et al, The same study found no association of NPC risk
2017), although results from other studies have with use of snuff or shisha (hookah).
been inconsistent (Xue et al, 2013). Some stud-
ies conducted in low-​or intermediate-​incidence Diet
regions detected stronger associations of tobacco Dietary consumption of Chinese-​ style salted
smoking with squamous cell than nonkeratiniz- fish is classified by IARC as carcinogenic to
ing NPC (Feng et al, 2009; Polesel et al, 2011; humans based on its relationship to NPC, and
Vaughan et al, 1996). Given the modest magni- intake of traditional Asian pickled vegetables
tude of observed associations, the attributable is classified as possibly carcinogenic to humans
risk of NPC due to tobacco smoking appears to (IARC, 2012d, 1993). The early observation that
be an order of magnitude lower than that for lung Cantonese boat-​ dwelling fishermen and their
cancer and other respiratory tract malignancies families were at increased risk of NPC led to the
(IARC, 2012d; IARC, 2004b). hypothesis that their traditional diet, consist-
Whether inhalation of other types of smoke ing heavily of dried, salted fish, was responsi-
causes NPC is not certain. In three studies in ble for their excess risk (Ho, 1976). By contrast,
China (Guo et  al, 2009; Yao et  al, 1987; Zheng intake of fresh fruits and vegetables appears to
et  al, 1994b), NPC cases were up to five times decrease NPC risk, although confounding can-
more likely to be exposed to domestic wood fire not be excluded as an explanation for inverse
than controls, but others found no such associa- associations. Other dietary factors—​ including
tion (Yu et al, 1990; Yu et al, 1986; Yu et al, 1988). other individual foods and food groups, mac-
A  hospital-​based case-​ control study in North ronutrients, and micronutrients—​have not been
168

168 Textbook of Cancer Epidemiology

studied thoroughly, especially in high-​risk popu- Yuan et  al, 2000b), as well as in low-​incidence
lations, and causal relationships have not been northern Chinese (Ning et al, 1990) and North
established. Americans (excluding undifferentiated nonke-
ratinizing NPC) (Farrow et  al, 1998). A  meta-​
Salt-​Preserved Fish and Other Foods analysis of six studies in China, Southeast Asia,
The environmental exposure most consistently and North Africa found that high versus low
and strongly associated with risk of NPC is con- preserved vegetable intake was associated with
sumption of salt-​preserved fish. Salt-​preserved a twofold increase in NPC risk, with a stronger
foods have historically been a dietary staple in association for preserved green leafy vegetables
all NPC-​endemic populations; hence, this tra- (Gallicchio et  al, 2006), despite different pres-
ditional dietary pattern may explain part of the ervation methods and consumption amounts
international distribution of NPC incidence. In across studies.
studies of Chinese populations, the RR of NPC
associated with weekly consumption, compared Fresh Fruits and Vegetables
with no or rare consumption, generally ranged In contrast to preserved foods, frequent con-
from 1.1 to 4, while that for daily consumption sumption of fresh fruits and/​ or vegetables
ranged from 1.8 to 20 (Guo et al, 2009; IARC, has been associated with a lower risk of NPC
2012d; Jia et  al, 2010; Xu et  al, 2012; Yu et  al, (Armstrong et al, 1998; Jia et al, 2010; Liu et al,
1986; Yu et  al, 1989; Yuan et  al, 2000b; Zheng 2012; Polesel et al, 2013; Yu et al, 1989; Yuan et al,
et al, 1994a; Zheng et al, 1994b), although the 2000b), suggesting a possible protective effect.
magnitude of the estimated association appears In a meta-​analysis of five case-​control studies,
to have declined over time. Consumption of high versus low intake of various nonpreserved
salt-​preserved fish during weaning, childhood, vegetables was associated with a 36% lower risk
and adulthood are all positively associated of NPC, with slightly larger decrements in risk
with NPC risk, with RR estimates of compara- for green leafy vegetables and nonstarchy roots
ble magnitude and positive exposure-​response and tubers in particular (Gallicchio et al, 2006).
trends with increasing frequency of consump- Some studies found inverse associations with
tion. One study in northern China detected intake of specific fruits or vegetables, especially
an increasing trend in the RR with earlier age carrots and other red/​orange/​yellow vegetables
at first exposure to salt-​preserved fish (Ning (Armstrong et al, 1998; Liu et al, 2012; Ning et al,
et al, 1990). 1990; Polesel et al, 2013), green leafy vegetables
The carcinogenic potential of salt-​preserved (Liu et  al, 2012; Zheng et  al, 1994b), and citrus
fish is supported by experiments in rats, which fruit, oranges, or tangerines (Armstrong et  al,
develop malignant nasal and nasopharyngeal 1998; Farrow et  al, 1998; Liu et  al, 2012; Yuan
tumors after salted-​ fish consumption (Huang et al, 2000b), or associations with dietary intake
et  al, 1978). The process of salt-​ preservation of vitamin E (Lee et al, 1994), vitamin C (Farrow
is inefficient, allowing fish and other foods to et al, 1998), or carotenoids (Polesel et al, 2012).
become partially putrefied (Ho, 1972). As a The apparent protective effect of fruits and veg-
result, these foods accumulate significant levels etables, if causal, has been hypothesized to be
of nitrosamines, which are demonstrated car- attributable to antioxidant effects, prevention of
cinogens in animals (IARC, 2012d; IARC, 1993). nitrosamine formation, and other anticarcino-
Salt-​preserved fish also contains bacterial muta- genic properties.
gens, direct genotoxins, and EBV-​ reactivating
substances, any or all of which could also con- Alcohol and Tea
tribute to the observed association. Alcohol consumption is not strongly associ-
NPC risk has also been found to be elevated, ated with NPC risk, as most case-​control stud-
albeit with a lower magnitude and less consist- ies (Chen et al, 1990; Cheng et al, 1999; Fachiroh
ency, in association with other preserved food et al, 2012; Ji et al, 2011; Sriamporn et al, 1992;
items, including meats, eggs, fruits, and veg- Xu et al, 2012; Zheng et al, 1994a), albeit not all
etables, in southern Chinese, Southeast Asians, (Armstrong et al, 1998; Ruan et al, 2010; Vaughan
North Africans/​Middle Easterners, and Arctic et al, 1996), have reported no significant associ-
natives (Armstrong et  al, 1998; Fachiroh et  al, ation. A meta-​analysis of 11 case-​control studies
2012; Feng et  al, 2007; Jeannel et  al, 1990; Jia yielded a meta-​OR of 1.33 (95% CI 1.09–​1.62)
et  al, 2010; Lanier et  al, 1980a; Xu et  al, 2012; comparing the highest with the lowest category
169

Nasopharyngeal Cancer 169

of intake (Chen et al, 2009), although the amount Epstein Barr Virus

of alcohol consumed in the highest category The ubiquitous EBV infects and persists latently
varied across studies. Six studies with exposure-​ in nearly all humans, with infection usually
response data showed a J-​shaped trend, with risk occurring subclinically in childhood (Klein et al,
decreasing up to 15 drinks per week and rising 2010). Because children in NPC-​endemic areas
with higher intake. are typically infected during infancy or early
A few case-​control studies reported an inverse childhood (Kangro et al, 1994), very early infec-
association between tea consumption and NPC tion with EBV may play a critical role in NPC
risk, perhaps attributable to the antioxidant development. The ubiquity of EBV infection
and antimicrobial properties of catechins. One complicates epidemiologic research on infection
detected an association only with green tea (Hsu itself as a risk factor, and indicates that cofac-
et al, 2012a), another found inverse associations tors are almost certainly involved in mediating
with six different types of tea (Ruan et al, 2010), risk. Infectious mononucleosis, a manifestation
and a third did not distinguish among types of of late-​childhood or young-​adulthood primary
tea (Xu et al, 2012). infection with EBV, has not been well studied in
relation to NPC, perhaps because late infection
Reproductive Factors with EBV is rare in areas with high NPC inci-
There is no evidence of an important role for dence. Two US studies found that a history of
reproductive factors in the etiology of NPC. infectious mononucleosis was associated with a
statistically nonsignificant 60% decrease in NPC
Hormones and Hormone risk, (Levine et  al, 1998; Vaughan et  al, 1996),
Receptors with one study reporting the inverse association
There is no direct evidence of a role for hormones only for infectious mononucleosis diagnosed at
in the etiology of NPC, although the two-​to least 5 years earlier, whereas infectious mononu-
threefold higher risk in men suggests a sex-​based cleosis diagnosed within the last 5 years was asso-
contribution to risk. Moreover, age-​specific vari- ciated with a nonsignificant fourfold increase in
ation in the male excess, with an increase in the NPC risk (Levine et al, 1998).
male-​ to-​
female ratio during adolescence and Relative to controls, NPC patients have ele-
a decline after age 60  years, further supports a vated immunoglobulin G (IgG) and IgA anti-
possible influence of sex steroid hormones (Xie body titers to the EBV viral capsid antigen (VCA)
et al, 2013). and early antigen (EA), as well as increased IgG
against the latent viral nuclear antigens EBNA-​
Anthropometric Measures 1 and EBNA-​ 2 and neutralizing antibodies
There is little evidence of a role for anthropo- against EBV-​specific DNase (de-​The et al, 1978;
metric measures in the etiology of NPC, with Henderson et al, 1977; Lin et al, 1977). Moreover,
one study detecting higher BMI in cases than these antibody titers, especially of IgA, precede
controls among males, but not females, in unad- tumor development by several years (Chen et al,
justed analyses (Liu et al, 2012). 1985; Chien et  al, 2001; Ji et  al, 2007; Lanier
et  al, 1980b; Levine et  al, 1981)  and are corre-
Infections lated with tumor burden, remission, and recur-
IARC has determined that the scientific evi- rence (Chan et al, 1979; de Schryver et al, 1974;
dence is sufficient to classify EBV as a cause of Henle et  al, 1973). Based on these patterns,
NPC (IARC, 2012b, 1997). The etiologic role immunofluorescence-​ based tests for anti-​VCA
of other infectious agents, however, is uncer- IgA, anti-​EA IgA, and anti-​EBV DNase antibod-
tain. The involvement of EBV in NPC has been ies have long been established as NPC screening
postulated since 1966, when NPC patients were tools in high-​risk populations (Chien et al, 2001;
found to express antibodies against an antigen Zeng et al, 1985). Subsequently, circulating cell-​
later identified as that of EBV (Old et al, 1966). free EBV DNA has also been detected in a higher
This finding was confirmed in 1970, when anti-​ proportion of NPC patients than controls, and
EBV antibodies were observed to be higher in levels are correlated with disease stage and prog-
NPC patients than in controls (Henle et al, 1970). nosis, providing support for the use of plasma or
Subsequently, a large body of laboratory, clinical, serum EBV DNA for early NPC detection and
and epidemiologic research on the etiologic role disease management (Mutirangura et  al, 1998;
of EBV in NPC has accumulated. Nawroz et al, 1996). Prospective studies support
170

170 Textbook of Cancer Epidemiology

the use of plasma EBV DNA for early NPC detec- have a deficient CD8+ T-​cell response specific
tion (Chan et  al, 2013; Ji et  al, 2014), although to EBV nuclear antigen-​ 1 (EBNA-​ 1) epitopes
it appears to have limited predictive value as a presented by HLA class I alleles, compared with
screening tool in unaffected individuals. healthy donors (Fogg et al, 2009). Thus, pathways
EBV is further linked to the development linking EBV infection to NPC development are
of NPC through EBV DNA, RNA, and/​or gene beginning to be elucidated.
products in tumor cells of virtually all nonke-
ratinizing NPC cases, regardless of geographic Other Infections
origin, whereas EBV detection in squamous No infections other than EBV have been estab-
cell NPC is less consistent (Niedobitek, 2000). lished as causes of NPC. High-​risk human pap-
Because the EBV episome is identical in every illoma virus (HPV) DNA has been detected in a
tumor cell—​as assessed by the number of ter- minority of squamous cell NPC tumors, but rarely
minal repeats in the latent, circularized form of in nonkeratinizing NPC (Giannoudis et al, 1995;
the virus in NPC tumors (Raab-​Traub & Flynn, Hording et al, 1994; Lin et al, 2014). The positive
1986)—​NPC appears to originate from a single findings may be due at least in part to misclas-
progenitor cell infected with EBV prior to clonal sification of oropharyngeal squamous cell carci-
expansion. Clonal EBV has also been detected in nomas, which are caused by HPV (Singhi et  al,
severe dysplasia or carcinoma in situ of the naso- 2012). A case-​control study of NPC in southern
pharynx (Pathmanathan et al, 1995; Zhong et al, China, where hepatitis B virus (HBV) infection
2006), indicating a role for the virus in the early is also endemic, found a 40%–​50% increased risk
stages of tumor progression. of NPC in individuals with past HBV infection
Although certain sequence variants, includ- and a strong interaction with anti-​EBV-​VCA and
ing a 30-​ base-​pair deletion, in the oncogenic EA IgA positivity (Ye et  al, 2015). Incidence of
latent membrane protein 1 (LMP1) of EBV have both squamous cell and nonkeratinizing NPC is
been shown to enhance the virus’s oncogenic approximately doubled among individuals with
potential in vitro and are frequently detected in AIDS (Shebl et al, 2010), but risk is not increased
NPC tumors (Tsao et al, 2002), EBV strain vari- among solid organ transplant recipients (Engels
ation is inconsistently associated with NPC inci- et al, 2011).
dence or clinical characteristics and has not been Most studies investigating prior chronic ear,
established as a risk factor (Kwok et al, 2012). nose, throat, and lower respiratory tract condi-
A large case-​ control study of NPC in tions, including allergic rhinitis (Chung et  al,
Guangdong Province, China, found that cigarette 2014; Lin et  al, 2015), rhinitis or rhinosinus-
smoking was associated with a higher prevalence itis (Hung et al, 2014; Yu et al, 1990; Yuan et al,
of anti-​VCA IgA among controls, and that ciga- 2000a), otitis media (Huang et al, 2012; Yu et al,
rette smoke extract promoted EBV activation in 1990; Yu et al, 1988; Yuan et al, 2000a), and nasal
vitro (Xu et  al, 2012), suggesting that smoking polyps (Yu et al, 1990; Yuan et al, 2000a), found
may increase NPC risk by inducing EBV reactiv- positive associations with NPC risk, with RRs
ation. Two studies in Taiwan also found positive usually around 2.0 or higher. These findings sug-
interactions, albeit not statistically significant, gest that benign inflammation and infection of
between smoking and anti-​EBV antibody titers the respiratory tract may render the nasopharyn-
on NPC risk (Hsu et  al, 2009; Lin et  al, 1979). geal mucosa more susceptible to development
However, another study of unaffected relatives of NPC. In addition, some bacteria can reduce
of NPC patients in Taiwan found few significant nitrate to nitrite, which can then form carcino-
associations of sociodemographic, dietary, life- genic N-​nitroso compounds.
style, and occupational NPC risk factors with
EBV antibody patterns (Chang et al, 2012). The Physical Activity
higher EBV IgA seroreactivity in first-​ degree There is no evidence of an important role for
relatives of NPC cases than in general commun- physical activity in the etiology of NPC.
ity members (Hutajulu et al, 2012; Pickard et al,
2004; Qin et al, 2011a) may point to genetic sus- Ionizing Radiation
ceptibility as a key cofactor in EBV-​related NPC There is no evidence of an important role for ion-
risk—​a hypothesis supported by genetic studies izing radiation in the etiology of NPC. A  case-​
highlighting the involvement of the HLA region control study set in a high background-​radiation
in NPC. Indeed, NPC patients were found to area in China found that radiation exposure was
17

Nasopharyngeal Cancer 171

not associated with NPC risk, and that risk fac- Dust
tors for NPC were the same there as in other Several studies found that risk of NPC was ele-
regions (Zou et al, 2000). vated among wood workers and other individu-
als potentially exposed to wood dust (Armstrong
Occupation et al, 2000; Demers et al, 1995; Hildesheim et al,
IARC considers the scientific evidence sufficient 2001; Sriamporn et  al, 1992; Vaughan & Davis,
to conclude that formaldehyde (IARC, 2012c, 1991), with a few studies detecting positive dose-​
2006)  and wood dust (IARC, 2012a, 1995)  are response trends in accordance with longer dura-
nasopharyngeal carcinogens in humans. Because tion and higher average or cumulative exposure
specific occupational exposures tend to be to wood products. However, other studies did
uncommon in the general population, they are not confirm this association (Siew et  al, 2012;
unlikely to account for a substantial proportion Stellman et  al, 1998; Vaughan et  al, 2000), rais-
of NPC, especially in endemic areas. ing questions about whether the relationship is
causal. Wood dust could plausibly promote NPC
Formaldehyde development through irritation and inflamma-
Workplace exposure to formaldehyde has been tion of the nasopharynx, impaired ciliary clear-
examined in several studies of NPC, chiefly in ance and mucostasis, or direct genotoxicity
low-​incidence regions. An association with NPC from compounds carried on wood dust particles
is supported by experiments in rodents, which (IARC, 2012a, 1995).
can develop squamous cell carcinomas of the Whether other types of occupational dust are
nasal cavity after prolonged, high-​ level expo- associated with NPC risk is uncertain. In three
sure to formaldehyde vapors (Albert et al, 1982; studies from China, textile workers, who typ-
Swenberg et  al, 1980). The evidence in favor of ically have heavy exposure to cotton dust, were
an effect on NPC in humans comes mainly from at significantly increased risk of NPC (Li et  al,
a historical cohort of over 25,000 employees at 2006a; Ng, 1986; Zheng et  al, 1992), especially
10 US facilities that produced or used formalde- in association with long-​term employment or
hyde (Beane Freeman et al, 2013; Blair et al, 1986; high estimated cumulative exposure. In con-
Hauptmann et al, 2004). In the most recent update trast, investigators who found that NPC risk was
of this cohort, exposed workers did not experi- 70% lower in workers exposed to cotton dust
ence a significant excess of NPC mortality overall, suggested that bacterial endotoxins could have
but positive trends were observed by peak expo- a protective effect by potentiating an antitumor
sure, average exposure intensity, and cumulative immune response (Yu et al, 1990).
exposure among exposed workers (not including
unexposed workers). However, the positive asso- Other Exposures
ciation was driven exclusively by the findings at No other occupational exposures are confirmed as
a single plant where six of the 10 observed NPC causes of NPC. Occupational exposure to indus-
deaths occurred, whereas there was no excess risk trial heat (Armstrong et al, 2000) or combustion
in the remaining nine plants, and sensitivity anal- products (Yu et  al, 1990)  more than doubled
yses revealed instability in the statistical model the risk of NPC in some studies, although these
used (Marsh & Youk, 2005; Marsh et  al, 2007). categories may encompass disparate exposures.
No excess of NPC was observed in three other Similarly, the excess of NPC incidence or mortal-
cohort studies of formaldehyde-​exposed workers ity observed among welders (Lam & Tan, 1984;
(Coggon et al, 2014; Pinkerton et al, 2004; Siew Zheng et al, 1992), furnace men, boiler firemen,
et  al, 2012). A  meta-​analysis of six case-​control smiths and forging-​press operators, bakers, metal
studies and seven cohort studies found no sig- workers (Zheng et al, 1992), and restaurant wait-
nificant association with NPC in case-​ control staff (Yu et al, 2004) may be due to shared expo-
studies after adjusting for socioeconomic status, sure to heat and fumes, or to disparate exposures.
smoking, or study location (smoking-​adjusted Three studies reported an excess risk of NPC
meta-​OR = 1.10, 95% CI 0.80–​1.50), and no asso- among printing workers (Li et al, 2006a; Liu et al,
ciation in cohort studies after excluding the sin- 2002; Roush et al, 1987), but did not identify spe-
gle anomalous US plant (meta-​SMR = 0.72, 95% cific inks, solvents, or other substances that could
CI 0.40–​1.28) (Bachand et al, 2010). These incon- be responsible for the association. Another study
sistent findings do not confirm a causal effect of in China reported a doubling of NPC risk in asso-
formaldehyde on NPC. ciation with occupational exposure to nonspecific
172

172 Textbook of Cancer Epidemiology

solvents (Guo et  al, 2009), although risk was NPC risk, with ORs of 1.25 (95% CI 1.06–​1.49)
higher with shorter than with longer duration of and 1.55 (95% CI 1.13–​2.12) for having 1–​3 and
exposure, and a study in Taiwan detected no such more than 3 filled teeth versus none, and an OR
association (Hildesheim et  al, 2001). Although of 0.62 (95% CI 0.55–​0.70) for brushing teeth
an excess risk of NPC has been observed among twice or more per day versus once or less per day
agricultural workers (Lam & Tan, 1984; Ng, 1986; (Liu et al, 2016).
Sriamporn et  al, 1992), studies assessing overall In Asian populations, several case-​control stud-
use of nonspecific pesticides found no association ies reported a two-​to fourfold excess risk of NPC
with NPC risk (Li et al, 2006a; West et al, 1993; in association with use of traditional herbal medi-
Zhu et al, 2002; Zou et al, 2000). cines (Hildesheim et  al, 1992; Shanmugaratnam
et al, 1978; West et al, 1993; Zheng et al, 1994b),
Medical Conditions although others found no association (Yu et  al,
and Treatment 1986; Yu et al, 1989; Yu et al, 1988) or an inverse
There is no evidence of an important causal role association (Jia et al, 2010). Any association with
for other medical conditions other than infec- use of herbal drugs may be difficult to disentangle
tions in the etiology of NPC. from other aspects of a traditional lifestyle, such as
In our southern China population-​ based diet, and may also be influenced by reverse cau-
case-​control study, we found positive associations sation in case-​control studies. A  role of Chinese
between some indicators of poor oral health and herbal plants in NPC development is, however,

TABLE 8-​1   RISK FACTORS AND PREVENTIVE FACTORS FOR

NASOPHARYNGEAL CARCINOMA

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Epstein Barr virus infection (positive serology) ↑↑
Salt-​preserved fish ↑
Family history in a first-​degree relative ↑↑
Certain HLA-​A and HLA-​B alleles ↑
Tobacco smoke ↑
Possible Relationship Exists, Based on Substantial Data
Other preserved foods ↑
Fresh fruits and vegetables ↓
Chronic respiratory tract conditions ↑
Wood dust ↑
Formaldehyde ↑
HIV/​AIDS ↑
Inconsistent Findings or Limited Study to Date
Other types of smoke and dust ↑
Alcohol ↑
Tea ↓
Herbal medicines ↑
HLA class II alleles ↑

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in
risk; ↑↑, moderate to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to
large decrease in risk
Note: The magnitude of the risk can vary substantially, depending on the exact comparison being
made. For example, the magnitude of the association with consumption of salt-​preserved fish
appears to be larger for daily than weekly consumption.
173

Nasopharyngeal Cancer 173

biologically plausible because several such com- REFERENCES

monly used plants can induce viral lytic antigen Albeck H, Bentzen J, Ockelmann HH, Nielsen NH,
expression by activating EBV in vitro (Hirayama Bretlau P, Hansen HS. Familial clusters of nasopha-
& Ito, 1981; Zeng et al, 1983). ryngeal carcinoma and salivary gland carcinomas
In addition, EBV-​ inducers were detected in Greenland natives. Cancer 1993;72:196–​200.
in extracts of soils, as well as some vegetables Albert RE, Sellakumar AR, Laskin S, Kuschner M,
grown in these soils, from areas in southern Nelson N, Snyder CA. Gaseous formaldehyde and
China where NPC is endemic (Zeng et al, 1984). hydrogen chloride induction of nasal cancer in the
Although use of certain EBV-​ inducing herbs rat. J Natl Cancer Inst 1982:68:597–​603.
of the Euphorbiaceae family was not associated Armstrong RW, Imrey PB, Lye MS, Armstrong MJ,
with risk in southern China (Yu et  al, 1990; Yu Yu MC, Sani S. Nasopharyngeal carcinoma in
Malaysian Chinese:  occupational exposures to
et al, 1986; Yu et al, 1988), use of other specific
particles, formaldehyde and heat. Int J Epidemiol
EBV-​inducing herbal drugs has not been exam-
2000;29:991–​8.
ined in relation to NPC risk. In the Philippines,
Armstrong RW, Imrey PB, Lye MS, Armstrong MJ,
use of any herbal medicines was associated with Yu MC, Sani S. Nasopharyngeal carcinoma in
elevated NPC risk, especially among those who Malaysian Chinese:  salted fish and other dietary
used herbal drugs and had high anti-​EBNA anti- exposures. Int J Cancer 1998;77:228–​35.
body titers (Hildesheim et  al, 1992), suggest- Bachand AM, Mundt KA, Mundt DJ, Montgomery
ing a proliferative effect of herbal medicines on RR. Epidemiological studies of formaldehyde
EBV-​transformed cells. exposure and risk of leukemia and nasopharyn-
geal cancer:  a meta-​ analysis. Crit Rev Toxicol
CONCLUSION 2010;40:85–​100.
The distinctive racial/​ethnic and geographic dis- Beane Freeman LE, Blair A, Lubin JH, Stewart PA,
tribution of NPC worldwide suggests that both Hayes RB, Hoover RN et al. Mortality from solid
environmental factors and genetic traits are tumors among workers in formaldehyde indus-
major contributors to its development. Well-​ tries: an update of the NCI cohort. Am J Ind Med
established risk factors for NPC include elevated 2013;56:1015–​26.
EBV DNA and antibody titers, consumption of Bei JX, Jia WH, Zeng YX. Familial and large-​scale
salt-​preserved fish, a family history of NPC, cer- case-​control studies identify genes associated with
tain HLA-​A and HLA-​B class  I  genotypes, and nasopharyngeal carcinoma. Semin Cancer Biol
tobacco smoking (Table 8-​1). Consumption of 2012;22:96–​106.
other preserved foods, heavy alcohol intake, Bei JX, Li Y, Jia WH, Feng BJ, Zhou G, Chen LZ et al.
a history of chronic inflammatory respiratory A genome-​ wide association study of nasopharyn-
geal carcinoma identifies three new susceptibility
tract conditions, occupational exposure to wood
loci. Nat Genet 2010;42:599–​603.
or cotton dust, and certain polymorphisms in
Bei JX, Su WH, Ng CC, Yu K, Chin YM, Lou PJ
biotransformation and metabolism genes may
et  al. International Nasopharyngeal Carcinoma
be associated with elevated NPC risk, whereas Genetics Working G.  A  GWAS meta-​ analysis
consumption of fresh fruits and vegetables and and replication study identifies a novel locus
their antioxidant micronutrients may be associ- within CLPTM1L/​ TERT associated with naso-
ated with decreased risk. Evidence for a causal pharyngeal carcinoma in individuals of Chinese
role of various inhalants such as wood smoke, ancestry. Cancer Epidemiol Biomarkers Prev
heavy alcohol intake, herbal medicines, and 2016;25:188–​92.
formaldehyde and other occupational exposures Blair A, Stewart P, O’Berg M, Gaffey W, Walrath J,
is inconsistent. To date, few prospective studies Ward J et  al. Mortality among industrial work-
or strictly population-​based case-​control stud- ers exposed to formaldehyde. J Natl Cancer Inst
ies in highly endemic areas have been conducted 1986;76:1071–​84.
to identify nonviral environmental risk factors Bray F, Haugen M, Moger TA, Tretli S, Aalen OO,
for NPC. Comprehending how viral, genetic, Grotmol T. Age-​incidence curves of nasopharyn-
and environmental factors interact to cause geal carcinoma worldwide: bimodality in low-​risk
NPC will illuminate the pathways by which populations and aetiologic implications. Cancer
this malignancy—​a model for a chronic disease Epidemiol Biomarkers Prev 2008;17:2356–​65.
caused by genes, environment, and an infec- Buell P. Race and place in the etiology of nasopharyn-
tious agent—​develops, as well as how it may be geal cancer: a study based on California death cer-
prevented. tificates. Int J Cancer 1973;11:268–​72.
174

174 Textbook of Cancer Epidemiology

Burt RD, Vaughan TL, McKnight B, Davis S, Beckmann and the risk of nasopharyngeal carcinoma: a sys-
AM, Smith AG et al. Associations between human tematic review. Nutr Cancer 2009;61:1–​15.
leukocyte antigen type and nasopharyngeal carci- Chen YJ, Ko JY, Chen PJ, Shu CH, Hsu MT, Tsai SF
noma in Caucasians in the United States. Cancer et  al. Chromosomal aberrations in nasopharyn-
Epidemiol Biomarkers Prev 1996;5:879–​87. geal carcinoma analyzed by comparative geno-
Burt RD, Vaughan TL, Nisperos B, Swanson M, mic hybridization. Genes Chromosomes Cancer
Berwick M. A protective association between the 1999;25:169–​75.
HLA-​A2 antigen and nasopharyngeal carcinoma Cheng YJ, Hildesheim A, Hsu MM, Chen IH, Brinton
in US Caucasians. Int J Cancer 1994;56:465–​7. LA, Levine PH et  al. Cigarette smoking, alco-
Cai K, Wang Y, Zhao X, Bao X. Association hol consumption and risk of nasopharyngeal
between the P53 codon 72 polymorphism carcinoma in Taiwan. Cancer Causes Control
and nasopharyngeal cancer risk. Tumour Biol 1999;10:201–​7.
2014;35:1891–​7. Chien YC, Chen JY, Liu MY, Yang HI, Hsu MM,
Chan JKC, Bray F, McCarron P, Foo W, Lee AWM, Chen CJ et  al. Serologic markers of Epstein-​
Yip T et al. Nasopharyngeal carcinoma. In Barnes Barr virus infection and nasopharyngeal car-
L, Eveson JW, Reichart P, Sidransky D. World cinoma in Taiwanese men. N Engl J Med
Health Organization Classification of Tumours 2001;345:1877–​82.
Pathology and Genetics Head and Neck Tumours. Chin YM, Mushiroda T, Takahashi A, Kubo M,
Lyon:  International Agency for Research on Krishnan G, Yap LF et al. HLA-​A SNPs and amino
Cancer. 2005: 85–​97. acid variants are associated with nasopharyngeal
Chan KC, Hung EC, Woo JK, Chan PK, Leung SF, Lai FP carcinoma in Malaysian Chinese. Int J Cancer
et al. Early detection of nasopharyngeal carcinoma 2015;136:678–​87.
by plasma Epstein-​Barr virus DNA analysis in a Chung SD, Wu CS, Lin HC, Hung SH. Association
surveillance program. Cancer 2013;119:1838–​44. between allergic rhinitis and nasopharyngeal car-
Chan SH, Day NE, Kunaratnam N, Chia KB, cinoma: a population-​based study. Laryngoscope
Simons MJ. HLA and nasopharyngeal carci- 2014;124:1744–​9.
noma in Chinese—​a further study. Int J Cancer Coggon D, Ntani G, Harris EC, Palmer KT. Upper air-
1983;32:171–​6. way cancer, myeloid leukemia, and other cancers
Chan SH, Levine PH, de-​ The GB, Mulroney SE, in a cohort of British chemical workers exposed to
Lavoue MF, Glen SP et  al. A comparison of the formaldehyde. Am J Epidemiol 2014;179:1301–​11.
prognostic value of antibody-​dependent lympho- de-​The G, Lavoue MF, Muenz L. Differences in EBV
cyte cytotoxicity and other EBV antibody assays in antibody titres of patients with nasopharyn-
Chinese patients with nasopharyngeal carcinoma. geal carcinoma originating from high, interme-
Int J Cancer 1979;23:181–​5. diate and low incidence areas. IARC Sci Publ
Chang CM, Yu KJ, Hsu WL, Major JM, Chen JY, Lou 1978:20:471–​81.
PJ et al. Correlates of anti-​EBV EBNA1 IgA pos- de Schryver A, Klein G, Henle W, Henle G. EB virus-​
itivity among unaffected relatives from nasopha- associated antibodies in Caucasian patients with
ryngeal carcinoma multiplex families. Br J Cancer carcinoma of the nasopharynx and in long-​
2012;106:206–​9. term survivors after treatment. Int J Cancer
Chang ET, Liu Z, Hildesheim A, Liu Q, Cai Y, Zhang Z 1974;13:319–​25.
et al. Active and passive smoking and risk of naso- Demers PA, Boffetta P, Kogevinas M, Blair A, Miller
pharyngeal carcinoma: a population-based case- BA, Robinson CF et al. Pooled reanalysis of cancer
control study in southern China. Am J Epidemiol mortality among five cohorts of workers in wood-​
2017;185:1272–1280. related industries. Scand J Work Environ Health
Chen CJ, Liang KY, Chang YS, Wang YF, Hsieh T, Hsu 1995;21:179–​90.
MM et al. Multiple risk factors of nasopharyngeal Dou H, Hu D, Lam C, Liu Y, Wang X, Zhang W.
carcinoma:  Epstein-​ Barr virus, malarial infec- Retrospective analysis of results of treatment
tion, cigarette smoking and familial tendency. for nasopharyngeal carcinoma in Macao. Chin J
Anticancer Res 1990;10:547–​53. Cancer Res 2014;26:148–​58.
Chen JY, Hwang LY, Beasley RP, Chien CS, Yang CS. Engels EA, Pfeiffer RM, Fraumeni JF Jr, Kasiske BL,
Antibody response to Epstein-​Barr-​virus-​specific Israni AK, Snyder JJ et  al. Spectrum of cancer
DNase in 13 patients with nasopharyngeal carci- risk among US solid organ transplant recipients.
noma in Taiwan: a retrospective study. J Med Virol JAMA 2011;306:1891–​901.
1985;16:99–​105. Fachiroh J, Sangrajrang S, Johansson M, Renard H,
Chen L, Gallicchio L, Boyd-​ Lindsley K, Tao XG, Gaborieau V, Chabrier A et al. Tobacco consump-
Robinson KA, Lam TK et al. Alcohol consumption tion and genetic susceptibility to nasopharyngeal
175

Nasopharyngeal Cancer 175

carcinoma (NPC) in Thailand. Cancer Causes polymerase chain reaction technique. Cancer Lett
Control 2012;23:1995–​2002. 1995;89:177–​81.
Farrow DC, Vaughan TL, Berwick M, Lynch CF, Goldsmith DB, West TM, Morton R. HLA associa-
Swanson GM, Lyon JL. Diet and nasopharyn- tions with nasopharyngeal carcinoma in Southern
geal cancer in a low-​risk population. Int J Cancer Chinese: a meta-​analysis. Clin Otolaryngol Allied
1998;78:675–​9. Sci 2002;27:61–​7.
Feng BJ, Huang W, Shugart YY, Lee MK, Zhang F, Xia Guo X, Johnson RC, Deng H, Liao J, Guan L, Nelson GW
JC et al. Genome-​wide scan for familial nasopha- et al. Evaluation of nonviral risk factors for naso-
ryngeal carcinoma reveals evidence of linkage to pharyngeal carcinoma in a high-​risk population of
chromosome 4. Nat Genet 2002;31:395–​9. Southern China. Int J Cancer 2009;124:2942–​7.
Feng BJ, Jalbout M, Ayoub WB, Khyatti M, Dahmoul Hauptmann M, Lubin JH, Stewart PA, Hayes RB, Blair
S, Ayad M et al. Dietary risk factors for nasopha- A. Mortality from solid cancers among work-
ryngeal carcinoma in Maghrebian countries. Int J ers in formaldehyde industries. Am J Epidemiol
Cancer 2007;121:1550–​5. 2004;159:1117–​30.
Feng BJ, Khyatti M, Ben-​Ayoub W, Dahmoul S, Ayad Henderson BE, Louie EW, Jing JS, Alena B. Epstein-​
M, Maachi F et al. Cannabis, tobacco and domes- Barr virus and nasopharyngeal carcinoma:  is
tic fumes intake are associated with nasopha- there an etiologic relationship? J Natl Cancer Inst
ryngeal carcinoma in North Africa. Br J Cancer 1977;59:1393–​5.
2009;101:1207–​12. Henle W, Henle G, Ho HC, Burtin P, Cachin Y, Clifford
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser P et el. Antibodies to Epstein-​Barr virus in naso-
S, Mathers C, Rebelo M, Parkin DM, Forman D, pharyngeal carcinoma, other head and neck neo-
Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence plasms, and control groups. J Natl Cancer Inst
and Mortality Worldwide: IARC CancerBase No. 1970;44:225–​31.
11 [Internet]. Lyon, France: International Agency Henle W, Ho HC, Henle G, Kwan HC. Antibodies to
for Research on Cancer; 2013. Available from: Epstein-​Barr virus-​related antigens in nasopha-
http://globocan.iarc.fr, accessed 2017. ryngeal carcinoma:  comparison of active cases
Fogg MH, Wirth LJ, Posner M, Wang F. Decreased with long-​ term survivors. J Natl Cancer Inst
EBNA-​ 1-​
specific CD8+ T cells in patients 1973;51:361–​9.
with Epstein-​ Barr virus-​ associated nasopha- Hildesheim A, Apple RJ, Chen CJ, Wang SS, Cheng
ryngeal carcinoma. Proc Natl Acad Sci U S A YJ, Klitz W et al. Association of HLA class I and
2009;106:3318–​23. II alleles and extended haplotypes with nasopha-
Forman D, Bray F, Brewster DH, Gombe Mbalawa C, ryngeal carcinoma in Taiwan. J Natl Cancer Inst
Kohler B, Piñeros M et al, eds. Cancer Incidence 2002;94:1780–​9.
in Five Continents, Vol. 10 (electronic version). Hildesheim A, Dosemeci M, Chan CC, Chen CJ,
Lyon:  IARC, 2013. http://​ci5.iarc.fr (accessed Cheng YJ, Hsu MM et al. Occupational exposure
January 2015). to wood, formaldehyde, and solvents and risk of
Friborg J, Wohlfahrt J, Koch A, Storm H, Olsen OR, nasopharyngeal carcinoma. Cancer Epidemiol
Melbye M. Cancer susceptibility in nasopharyn- Biomarkers Prev 2001;10:1145–​53.
geal carcinoma families—​ a population-​ based Hildesheim A, Wang CP. Genetic predisposition
cohort study. Cancer Res 2005;65:8567–​72. factors and nasopharyngeal carcinoma risk:  a
Gallicchio L, Matanoski G, Tao XG, Chen L, Lam review of epidemiological association studies,
TK, Boyd K et  al. Adulthood consumption of 2000–​2011: Rosetta Stone for NPC: genetics, viral
preserved and nonpreserved vegetables and the infection, and other environmental factors. Semin
risk of nasopharyngeal carcinoma:  a systematic Cancer Biol 2012;22:107–​16.
review. Int J Cancer 2006;119:1125–​35. Hildesheim A, West S, DeVeyra E, De Guzman MF,
Garber JE, Offit K. Hereditary cancer predisposition Jurado A, Jones C et  al. Herbal medicine use,
syndromes. J Clin Oncol 2005;23:276–​92. Epstein-​Barr virus, and risk of nasopharyngeal
Geser A, Charnay N, Day NE, de-​The G, Ho HC. carcinoma. Cancer Res 1992;52:3048–​51.
Environmental factors in the etiology of naso- Hirayama T, Ito Y. A new view of the etiology of naso-
pharyngeal carcinoma:  report on a case-​ pharyngeal carcinoma. Prev Med 1981;10:614–​22.
control study in Hong Kong. IARC Sci Publ Ho HC. Current knowledge of the epidemiology of
1978;20:213–​29. nasopharyngeal carcinoma—​a review. IARC Sci
Giannoudis A, Ergazaki M, Segas J, Giotakis J, Publ 1972;2:357–​66.
Adamopoulos G, Gorgoulis V et  al. Detection Ho HC. Epidemiology of nasopharyngeal carcinoma.
of Epstein-​ Barr virus and human papilloma- In Hirayama T. Cancer Asia. Baltimore: University
virus in nasopharyngeal carcinoma by the Park Press, 1976: 49–​61.
176

176 Textbook of Cancer Epidemiology

Hording U, Nielsen HW, Daugaard S, Albeck H. Hutajulu SH, Ng N, Jati BR, Fachiroh J, Herdini C,
Human papillomavirus types 11 and 16 detected Hariwiyanto B et al. Seroreactivity against Epstein-​
in nasopharyngeal carcinomas by the polymerase Barr virus (EBV) among first-​degree relatives of
chain reaction. Laryngoscope 1994;104:99–​102. sporadic EBV-​ associated nasopharyngeal carci-
Hsu C, Shen YC, Cheng CC, Hong RL, Chang CJ, noma in Indonesia. J Med Virol 2012;84:768–​76.
Cheng AL. Difference in the incidence trend of IARC. Some Naturally Occurring Substances:  Food
nasopharyngeal and oropharyngeal carcinomas Items and Constituents, Heterocyclic Aromatic
in Taiwan:  implication from age-​ period-​
cohort Amines and Mycotoxins. IARC Monographs on
analysis. Cancer Epidemiol Biomarkers Prev the Evaluation of Carcinogenic Risks to Humans,
2006;15:856–​61. Vol. 56. Lyon: IARC, 1993.
Hsu WL, Chen JY, Chien YC, Liu MY, You SL, Hsu IARC. Wood Dust and Formaldehyde. IARC
MM et al. Independent effect of EBV and cigarette Monographs on the Evaluation of Carcinogenic
smoking on nasopharyngeal carcinoma: a 20-​year Risks to Humans, Vol. 62. Lyon: IARC, 1995.
follow-​up study on 9,622 males without family IARC. Epstein-​Barr Virus and Kaposi’s Herpesvirus/​
history in Taiwan. Cancer Epidemiol Biomarkers Human Herpesvirus 8. IARC Monographs on the
Prev 2009;18:1218–​26. Evaluation of Carcinogenic Risks to Humans, Vol.
Hsu WL, Pan WH, Chien YC, Yu KJ, Cheng YJ, Chen 70. Lyon: IARC, 1997.
JY et  al. Lowered risk of nasopharyngeal carci- IARC. Betel-​ Quid and Areca-​ Nut Chewing and
noma and intake of plant vitamin, fresh fish, green Some Areca-​ Nut-​Derived Nitrosamines. IARC
tea and coffee:  a case-​control study in Taiwan. Monographs on the Evaluation of Carcinogenic
PLoS One 2012a;7:e41779. Risks to Humans, Vol. 85 Lyon: IARC, 2004a.
Hsu WL, Tse KP, Liang S, Chien YC, Su WH, Yu KJ IARC. Tobacco Smoke and Involuntary Smoking.
et  al. Evaluation of human leukocyte antigen-​A IARC Monographs on the Evaluation of
(HLA-​A), other non-​HLA markers on chromo- Carcinogenic Risks to Humans, Vol. 83.
some 6p21 and risk of nasopharyngeal carcinoma. Lyon: IARC, 2004b;83:1–​1452.
PLoS One 2012b;7:e42767. IARC. Formaldehyde, 2-​ butoxyethanol and 1-​ tert-​
Hu LF, Qiu QH, Fu SM, Sun D, Magnusson K, He B butoxypropan-​ 2-​
ol. IARC Monographs on the
et al. A genome-​wide scan suggests a susceptibility Evaluation of Carcinogenic Risks to Humans, Vol.
locus on 5p 13 for nasopharyngeal carcinoma. Eur 88. Lyon: IARC, 2006.
J Hum Genet 2008;16:343–​9. IARC. Arsenic, Metals, Fibres, and Dusts. IARC
Huang DP, Ho JH, Saw D, Teoh TB. Carcinoma of the Monographs on the Evaluation of Carcinogenic
nasal and paranasal regions in rats fed Cantonese Risks to Humans, Vol. 100 C. Lyon: IARC, 2012a.
salted marine fish. IARC Sci Publ 1978;315–​28. IARC. Biological Agents. IARC Monographs on the
Huang GL, Guo HQ, Yu CY, Liu XY, Li BB, Wu JJ et al. Evaluation of Carcinogenic Risks to Humans, Vol.
XRCC1 polymorphisms and risk of nasopha- 100 B. Lyon: IARC, 2012b.
ryngeal carcinoma:  a meta-​analysis. Asian Pac J IARC. Chemical Agents and Related Occupations.
Cancer Prev 2011;12:2329–​33. IARC Monographs on the Evaluation of Carcino­
Huang WY, Lin CC, Jen YM, Lin KT, Yang MH, genic Risks to Humans, Vol. 100 F. Lyon: IARC,
Chen CM et  al. Association between adult otitis 2012c.
media and nasopharyngeal cancer:  a nationwide IARC. Part E: Personal Habits and Indoor Combustions.
population-​based cohort study. Radiother Oncol IARC Monographs on the Evaluation of Carcino­
2012;104:338–​42. genic Risks to Humans. Vol. 100. Lyon: IARC,
Hui AB, Lo KW, Leung SF, Teo P, Fung MK, To KF 2012d.
et  al. Detection of recurrent chromosomal gains Jeannel D, Ghnassia M, Hubert A, Sancho-​Garnier
and losses in primary nasopharyngeal carcinoma H, Eschwege F, Crognier E, et  al. Increased risk
by comparative genomic hybridisation. Int J of nasopharyngeal carcinoma among males of
Cancer 1999;82:498–​503. French origin born in Maghreb (North Africa).
Hung SH, Chen PY, Lin HC, Ting J, Chung SD. Int J Cancer 1993;54:536–​9.
Association of rhinosinusitis with nasopha- Jeannel D, Hubert A, de Vathaire F, Ellouz R, Camoun
ryngeal carcinoma:  a population-​ based study. M, Ben Salem M et al. Diet, living conditions and
Laryngoscope 2014;124:1515–​20. nasopharyngeal carcinoma in Tunisia—​ a case-​
Hutajulu SH, Indrasari SR, Indrawati LP, Harijadi A, control study. Int J Cancer 1990;46:421–​5.
Duin S, Haryana SM et al. Epigenetic markers for Ji MF, Huang QH, Yu X, Liu Z, Li X, Zhang LF et al.
early detection of nasopharyngeal carcinoma in a Evaluation of plasma Epstein-​ Barr virus DNA
high risk population. Mol Cancer 2011;10:48. load to distinguish nasopharyngeal carcinoma
17

Nasopharyngeal Cancer 177

patients from healthy high-​ risk populations in Kong from 1976–​81. Ann Acad Med Singapore
Southern China. Cancer 2014;120:1353–​60. 1984;13:361–​5.
Ji MF, Wang DK, Yu YL, Guo YQ, Liang JS, Cheng WM Lanier A, Bender T, Talbot M, Wilmeth S, Tschopp
et al. Sustained elevation of Epstein-​Barr virus anti- C, Henle W et  al. Nasopharyngeal carcinoma
body levels preceding clinical onset of nasopha- in Alaskan Eskimos Indians, and Aleuts:  a
ryngeal carcinoma. Br J Cancer 2007;96:623–​30. review of cases and study of Epstein-​Barr virus,
Ji X, Zhang W, Xie C, Wang B, Zhang G, Zhou F. HLA, and environmental risk factors. Cancer
Nasopharyngeal carcinoma risk by histologic type 1980a;46:2100–​6.
in central China: impact of smoking, alcohol and Lanier AP, Henle W, Bender TR, Henle G, Talbot
family history. Int J Cancer 2011;129:724–​32. ML. Epstein-​ Barr virus-​specific antibody titers
Jia WH, Collins A, Zeng YX, Feng BJ, Yu XJ, Huang in seven Alaskan natives before and after diag-
LX et  al. Complex segregation analysis of naso- nosis of nasopharyngeal carcinoma. Int J Cancer
pharyngeal carcinoma in Guangdong, China: evi- 1980b;26:133–​7.
dence for a multifactorial mode of inheritance Lee AW, Sze WM, Au JS, Leung SF, Leung TW,
(complex segregation analysis of NPC in China). Chua DT et  al. Treatment results for nasopha-
Eur J Hum Genet 2005;13:248–​52. ryngeal carcinoma in the modern era:  the Hong
Jia WH, Feng BJ, Xu ZL, Zhang XS, Huang P, Huang Kong experience. Int J Radiat Oncol Biol Phys
LX et al. Familial risk and clustering of nasopha- 2005;61:1107–​16.
ryngeal carcinoma in Guangdong, China. Cancer Lee HP, Gourley L, Duffy SW, Esteve J, Lee J, Day NE.
2004;101:363–​9. Preserved foods and nasopharyngeal carcinoma: a
Jia WH, Huang QH, Liao J, Ye W, Shugart YY, Liu Q case-​control study among Singapore Chinese. Int J
et al. Trends in incidence and mortality of naso- Cancer 1994;59:585–​90.
pharyngeal carcinoma over a 20–​25  year period Leung TW, Tung SY, Sze WK, Wong FC, Yuen KK, Lui
(1978/​1983–​2002) in Sihui and Cangwu counties CM et al. Treatment results of 1070 patients with
in southern China. BMC Cancer 2006;6:178. nasopharyngeal carcinoma: an analysis of survival
Jia WH, Luo XY, Feng BJ, Ruan HL, Bei JX, Liu WS et al. and failure patterns. Head Neck 2005;27:555–​65.
Traditional Cantonese diet and nasopharyngeal Levine PH, Pearson GR, Armstrong M, Bengali Z,
carcinoma risk: a large-​scale case-​control study in Berenberg J, Easton J et  al. The reliability of IgA
Guangdong, China. BMC Cancer 2010;10:446. antibody to Epstein-​ Barr virus (EBV) capsid
Jiang RC, Qin HD, Zeng MS, Huang W, Feng BJ, Zhang antigen as a test for the diagnosis of nasopha-
F et al. A functional variant in the transcriptional ryngeal carcinoma (NPC). Cancer Detect Prev
regulatory region of gene LOC344967 cosegre- 1981;4:307–​12.
gates with disease phenotype in familial nasopha- Levine R, Zhu K, Gu Y, Brann E, Hall I, Caplan L,
ryngeal carcinoma. Cancer Res 2006;66:693–​700. Baum M. Self-​reported infectious mononucleosis
Jin B, Dong P, Li K, Shen B, Xie J. Meta-​analysis of and 6 cancers:  a population-​based, case-​control
the association between GSTT1 null genotype study. Scand J Infect Dis 1998;30:211–​4.
and risk of nasopharyngeal carcinoma in Chinese. Li CC, Yu MC, Henderson BE. Some epidemiologic
Tumour Biol 2014;35:345–​9. observations of nasopharyngeal carcinoma in
Kangro HO, Osman HK, Lau YL, Heath RB, Yeung Guangdong, People’s Republic of China. Natl
CY, Ng MH. Seroprevalence of antibodies to Cancer Inst Monogr 1985;69:49–​52.
human herpesviruses in England and Hong Kong. Li LL, Shu XS, Wang ZH, Cao Y, Tao Q. Epigenetic
J Med Virol 1994;43:91–​6. disruption of cell signaling in nasopharyngeal car-
Klein G, Klein E, Kashuba E. Interaction of Epstein-​ cinoma. Chin J Cancer 2011;30:231–​9.
Barr virus (EBV) with human B-​ lymphocytes. Li W, Ray RM, Gao DL, Fitzgibbons ED, Seixas NS,
Biochem Biophys Res Commun 2010;396:67–​73. Camp JE et al. Occupational risk factors for naso-
Krogman WM. Study of four skulls from Seleucia-​on-​ pharyngeal cancer among female textile work-
the Tigris dating from 100 B.C. to 200 A.D. Hum ers in Shanghai, China. Occup Environ Med
Biol 1940;12:313–​22. 2006a;63:39–​44.
Kwok H, Tong AH, Lin CH, Lok S, Farrell PJ, Kwong Li X, Wang E, Zhao YD, Ren JQ, Jin P, Yao KT,
DL et  al. Genomic sequencing and comparative Marincola FM. Chromosomal imbalances in
analysis of Epstein-​ Barr virus genome isolated nasopharyngeal carcinoma:  a meta-​ analysis of
from primary nasopharyngeal carcinoma biopsy. comparative genomic hybridization results. J
PLoS One 2012;7:e36939. Transl Med 2006b:4:4.
Lam YM, Tan TC. Mortality from nasopharyn- Lin JC, Wang WY, Chen KY, Wei YH, Liang WM et al.
geal carcinoma and occupation in men in Hong Quantification of plasma Epstein-​Barr virus DNA
178

178 Textbook of Cancer Epidemiology

in patients with advanced nasopharyngeal carci- Mousavi SM, Sundquist J, Hemminki K.

noma. N Engl J Med 2004;350:2461–​70. Nasopharyngeal and hypopharyngeal carcinoma
Lin KT, Huang WY, Lin CC, Jen YM, Lin CS, Lo CH risk among immigrants in Sweden. Int J Cancer
et  al. Subsequent risk of nasopharyngeal carci- 2010;127:2888–​92.
noma among patients with allergic rhinitis:  a Murthy AK, Kumar V, Suresh KP. Meta-​analysis of
nationwide population-​based cohort study. Head GSTM1 and GSTT1 polymorphisms and risk of
Neck 2015;37:413–​7. nasopharyngeal cancer. Asian Pac J Cancer Prev
Lin TM, Yang CS, Chiou JF, Tu SM, Chen TY, Tu YC 2013;14:1697–​701.
et al. Antibodies to Epstein-​Barr virus capsid anti- Mutirangura A, Pornthanakasem W, Theamboonlers
gen and early antigen in nasopharyngeal carci- A, Sriuranpong V, Lertsanguansinchi P, Yenrudi S
noma and comparison groups. Am J Epidemiol et al. Epstein-​Barr viral DNA in serum of patients
1977;106:336–​9. with nasopharyngeal carcinoma. Clin Cancer Res
Lin TM, Yang CS, Tu SM, Chen CJ, Kuo KC, Hirayama 1998;4:665–​9.
T. Interaction of factors associated with cancer of Mutirangura A, Tanunyutthawongese C,
the nasopharynx. Cancer 1979;44:1419–​23. Pornthanakasem W, Kerekhanjanarong V,
Lin Z, Khong B, Kwok S, Cao H, West RB, Le QT et al. Sriuranpong V, Yenrudi S et al. Genomic altera-
Human papillomavirus 16 detected in nasopha- tions in nasopharyngeal carcinoma: loss of het-
ryngeal carcinomas in white Americans but not in erozygosity and Epstein-​Barr virus infection. Br
endemic Southern Chinese patients. Head Neck J Cancer 1997;76:770–​6.
2014;6:709–​14. National Registry of Diseases Office. Singapore
Liu Z, Chang ET, Liu Q, Cai Y, Zhang Z, Chen G et Cancer Registry Interim Annual Registry Report.
al. Oral hygiene and risk of nasopharyngeal car- Trends in Cancer Incidence in Singapore 2008–​
cinoma - a population-based case-control study 2012: Singapore Cancer Registry, 2013.
in China. Cancer Epidemiol Biomarkers Prev National Registry of Diseases Office. Trends in Cancer
2016;25:1201–1207. Incidence in Singapore 1968–​ 2007:  Singapore
Liu Z, Chang ET, Liu Q, Cai Y, Zhang Z, Chen G. Cancer Registry, 2009.
Quantification of familial risk of nasopharyn- Nawroz H, Koch W, Anker P, Stroun M, Sidransky
geal carcinoma in a high-incidence area. Cancer D. Microsatellite alterations in serum DNA
2017;123:2716–2725. of head and neck cancer patients. Nat Med
Liu YH, Du CL, Lin CT, Chan CC, Chen CJ, Wang JD. 1996;2:1035–​7.
Increased morbidity from nasopharyngeal carci- Ng CC, Yew PY, Puah SM, Krishnan G, Yap LF, Teo SH
noma and chronic pharyngitis or sinusitis among et al. A genome-​wide association study identifies
workers at a newspaper printing company. Occup ITGA9 conferring risk of nasopharyngeal carci-
Environ Med 2002;59:18–​22. noma. J Hum Genet 2009;54:392–​7.
Liu YT, Dai JJ, Xu CH, Lu YK, Fan YY, Zhang XL et al. Ng TP. A case-​referent study of cancer of the nasal
Greater intake of fruit and vegetables is associ- cavity and sinuses in Hong Kong. Int J Epidemiol
ated with lower risk of nasopharyngeal carcinoma 1986;5:171–​5.
in Chinese adults:  a case-​control study. Cancer Niedobitek G. Epstein-​ Barr virus infection in the
Causes Control 2012;23:589–​99. pathogenesis of nasopharyngeal carcinoma. Mol
Lo KW, Chung GT, To KF. Deciphering the molecular Pathol 2000;53:248–​54.
genetic basis of NPC through molecular, cytoge- Ning JP, Yu MC, Wang QS, Henderson BE.
netic, and epigenetic approaches. Semin Cancer Consumption of salted fish and other risk factors
Biol 2012;22:79–​86. for nasopharyngeal carcinoma (NPC) in Tianjin,
Lu SJ, Day NE, Degos L, Lepage V, Wang PC, Chan a low-​risk region for NPC in the People’s Republic
SH et al. Linkage of a nasopharyngeal carcinoma of China. J Natl Cancer Inst 1990;82:291–​6.
susceptibility locus to the HLA region. Nature Old LJ, Boyse EA, Oettgen HF, Harven ED, Geering G,
1990;346:470–​1. Williamson B, Clifford P. Precipitating antibody
Marsh GM, Youk AO. Reevaluation of mortality risks in human serum to an antigen present in cultured
from nasopharyngeal cancer in the formaldehyde Burkitt’s lymphoma cells. Proc Natl Acad Sci U S
cohort study of the National Cancer Institute. A 1966;56:1699–​704.
Regul Toxicol Pharmacol 2005;42:275–​83. Parkin DM, Iscovich J. Risk of cancer in migrants and
Marsh GM, Youk AO, Morfeld P. Mis-​specified and their descendants in Israel:  II. Carcinomas and
non-​robust mortality risk models for nasopha- germ-​cell tumours. Int J Cancer 1997;70:654–​60.
ryngeal cancer in the National Cancer Institute Pathmanathan R, Prasad U, Sadler R, Flynn K, Raab-​
formaldehyde worker cohort study. Regul Toxicol Traub N. Clonal proliferations of cells infected
Pharmacol 2007;47:59–​67. with Epstein-​ Barr virus in preinvasive lesions
179

Nasopharyngeal Cancer 179

related to nasopharyngeal carcinoma. N Engl J in relation to the risk of nasopharyngeal carci-

Med 1995;333:693–​8. noma in Guangdong, China. Front Med China
Phua CE, Tan BS, Yong TK, Govindasamy M. 2010;4:448–​56.
Retrospective analysis of results of treatment for SEER. Surveillance, Epidemiology, and End Results
nasopharyngeal carcinoma in Penang General (SEER) Program (www.seer.cancer.gov) Research
Hospital from 2001–​2005. Asian Pac J Cancer Data (1973–​ 2012), National Cancer Institute,
Prev 2011;12:3197–​200. DCCPS, Surveillance Research Program,
Phua VCE, Loo WH, Yusof MM, Wan Ishak WZ, Tho Surveillance Systems Branch, released April 2015,
LM, Ung NM. Treatment outcome for nasopha- based on the November 2014 submission.
ryngeal carcinoma in University Malaya Medical Shanmugaratnam K. Histological typing of nasopha-
Centre from 2004-​2008. Asian Pac J Cancer Prev ryngeal carcinoma. IARC Sci Publ 1978;20:3–​12.
2013:14:4567–​70. Shanmugaratnam K, Tye CY, Goh EH, Chia KB.
Pickard A, Chen CJ, Diehl SR, Liu MY, Cheng YJ, Hsu Etiological factors in nasopharyngeal carcinoma: a
WL et al. Epstein-​Barr virus seroreactivity among hospital-​based, retrospective, case-​control, ques-
unaffected individuals within high-​ risk naso- tionnaire study. IARC Sci Publ 1978;20:199–​212.
pharyngeal carcinoma families in Taiwan. Int J Shebl FM, Bhatia K, Engels EA. Salivary gland and
Cancer 2004;111:117–​23. nasopharyngeal cancers in individuals with
Pinkerton LE, Hein MJ, Stayner LT. Mortality acquired immunodeficiency syndrome in United
among a cohort of garment workers exposed to States. Int J Cancer 2010;126:2503–​8.
formaldehyde:  an update. Occup Environ Med Siew SS, Kauppinen T, Kyyronen P, Heikkila P, Pukkala
2004;61:193–​200. E. Occupational exposure to wood dust and form-
Polesel J, Franceschi S, Talamini R, Negri E, Barzan aldehyde and risk of nasal, nasopharyngeal, and
L, Montella M et  al. Tobacco smoking, alcohol lung cancer among Finnish men. Cancer Manag
drinking, and the risk of different histological Res 2012;4:223–​32.
types of nasopharyngeal cancer in a low-​risk pop- Simons MJ, Chan SH, Wee GB, Shanmugaratnam K,
ulation. Oral Oncol 2011;47:541–​5. Goh EH, Ho JH, et al. Nasopharyngeal carcinoma
Polesel J, Negri E, Serraino D, Parpinel M, Barzan L, and histocompatibility antigens. IARC Sci Publ
Libra M et al. Dietary intakes of carotenoids and 1978;20:271–​82.
other nutrients in the risk of nasopharyngeal car- Simons MJ, Day NE, Wee GB, Shanmugaratnam K,
cinoma: a case-​control study in Italy. Br J Cancer Ho HC, Wong SH et  al. Nasopharyngeal carci-
2012;107:1580–​3. noma V:  immunogenetic studies of Southeast
Polesel J, Serraino D, Negri E, Barzan L, Vaccher E, Asian ethnic groups with high and low risk for the
Montella M et  al. Consumption of fruit, vegeta- tumor. Cancer Res 1974;34:1192–​5.
bles, and other food groups and the risk of naso- Simons MJ, Wee GB, Chan SH, Shanmugaratnam K,
pharyngeal carcinoma. Cancer Causes Control Day NE, de-​The G. Immunogenetic aspects of
2013;24:1157–​65. nasopharyngeal carcinoma (NPC) III. HL-​a type
Qin HD, Jia WH, Zhang LL, Liu N, Zhou XX, Wang as a genetic marker of NPC predisposition to test
MH et  al. Elevated Epstein-​ Barr virus serore- the hypothesis that Epstein-​Barr virus is an etio-
activity among unaffected members of families logical factor in NPC. IARC Sci Publ 1975;11 Pt
with nasopharyngeal carcinoma. J Med Virol 2:249–​58.
2011a;83:1792–​8. Singhi AD, Califano J, Westra WH. High-​risk human
Qin HD, Shugart YY, Bei JX, Pan QH, Chen L, Feng papillomavirus in nasopharyngeal carcinoma.
QS et  al. Comprehensive pathway-​based associ- Head Neck 2012;34:213–​8.
ation study of DNA repair gene variants and the Sriamporn S, Vatanasapt V, Pisani P, Yongchaiyudha
risk of nasopharyngeal carcinoma. Cancer Res S, Rungpitarangsri V. Environmental risk factors
2011b;71:3000–​8. for nasopharyngeal carcinoma:  a case-​ control
Raab-​Traub N, Flynn K. The structure of the termini study in northeastern Thailand. Cancer Epidemiol
of the Epstein-​Barr virus as a marker of clonal cel- Biomarkers Prev 1992;1:345–​8.
lular proliferation. Cell 1986;47:883–​9. Stellman SD, Demers PA, Colin D, Boffetta P. Cancer
Roush GC, Walrath J, Stayner LT, Kaplan SA, Flannery mortality and wood dust exposure among par-
JT, Blair A. Nasopharyngeal cancer, sinonasal ticipants in the American Cancer Society Cancer
cancer, and occupations related to formalde- Prevention Study-​ II (CPS-​
II). Am J Ind Med
hyde:  a case-​ control study. J Natl Cancer Inst 1998;34:229–​37.
1987;79:1221–​4. Su WH, Hildesheim A, Chang YS. Human leuko-
Ruan HL, Xu FH, Liu WS, Feng QS, Chen LZ, cyte antigens and Epstein-​Barr virus-​associated
Zeng YX et  al. Alcohol and tea consumption nasopharyngeal carcinoma: old associations offer
180

180 Textbook of Cancer Epidemiology

new clues into the role of immunity in infection-​ Wang HY, Sun BY, Zhu ZH, Chang ET, To KF, Hwang
associated cancers. Front Oncol 2013;3:299. JS et al. Eight-​signature classifier for prediction of
Sun LM, Epplein M, Li CI, Vaughan TL, Weiss NS. nasopharyngeal carcinoma survival. J Clin Oncol
Trends in the incidence rates of nasopharyngeal 2011;29:4516–​25.
carcinoma among Chinese Americans living in Los Wang WY, Twu CW, Chen HH, Jiang RS, Wu CT,
Angeles County and the San Francisco metropolitan Liang KL et  al. Long-​term survival analysis of
area, 1992–​2002. Am J Epidemiol 2005;162:1174–​8. nasopharyngeal carcinoma by plasma Epstein-​
Sun X, Su S, Chen C, Han F, Zhao C, Xiao W et  al. Barr virus DNA levels. Cancer 2013;119:963–​70.
Long-​term outcomes of intensity-​modulated radi- Wee JT, Ha TC, Loong SL, Qian CN. Is nasopharyn-
otherapy for 868 patients with nasopharyngeal geal cancer really a “Cantonese cancer”? Chin J
carcinoma: an analysis of survival and treatment Cancer 2010;29:517–​26.
toxicities. Radiother Oncol 2014;110:398–​403. Wei KR, Yu YL, Yang YY, Ji MF, Yu BH, Liang ZH,
Sun ZF, Zhang J, Xu HM, Wang GL, Dong P. Association Reng X. Epidemiological trends of nasopharyn-
between GSTM1 polymorphism and nasopharyn- geal carcinoma in China. Asian Pac J Cancer Prev
geal cancer susceptibility:  a meta-​analysis. Asian 2010;11:29–​32.
Pac J Cancer Prev 2012;13:5817–​21. Wei Y, Zhou T, Lin H, Sun M, Wang D, Li H et  al.
Swenberg JA, Kerns WD, Mitchell RI, Gralla EJ, Pavkov Significant associations between GSTM1/​GSTT1
KL. Induction of squamous cell carcinomas of the polymorphisms and nasopharyngeal cancer risk.
rat nasal cavity by inhalation exposure to formal- Tumour Biol 2013;34:887–​94.
dehyde vapor. Cancer Res 1980;40:3398–​402. Wells C. Chronic sinusitis with alveolar fistulae of
Tang M, Lautenberger JA, Gao X, Sezgin E, medieval times. J Laryngol 1963;261:320–​22.
Hendrickson SL, Troyer JL et  al. The principal West S, Hildesheim A, Dosemeci M. Non-​viral risk
genetic determinants for nasopharyngeal carci- factors for nasopharyngeal carcinoma in the
noma in China involve the HLA class  I  antigen Philippines: results from a case-​control study. Int J
recognition groove. PLoS Genet 2012;8:e1003103. Cancer 1993;55:722–​7.
Tian F, Yip SP, Kwong DL, Lin Z, Yang Z, Wu VW. Xie SH, Yu IT, Tse LA, Mang OW, Yue L. Sex differ-
Promoter hypermethylation of tumor suppres- ence in the incidence of nasopharyngeal car-
sor genes in serum as potential biomarker for the cinoma in Hong Kong 1983–​ 2008:  suggestion
diagnosis of nasopharyngeal carcinoma. Cancer of a potential protective role of oestrogen. Eur J
Epidemiol 2013;37:708–​13. Cancer 2013;49:150–​5.
Tsao SW, Tramoutanis G, Dawson CW, Lo AK, Xiong W, Zeng ZY, Xia JH, Xia K, Shen SR, Li XL et al.
Huang DP. The significance of LMP1 expression A susceptibility locus at chromosome 3p21 linked
in nasopharyngeal carcinoma. Semin Cancer Biol to familial nasopharyngeal carcinoma. Cancer Res
2002;12:473–​87. 2004;64:1972–​4.
Tse KP, Su WH, Chang KP, Tsang NM, Yu CJ, Tang P Xu FH, Xiong D, Xu YF, Cao SM, Xue WQ, Qin HD
et al. Genome-​wide association study reveals mul- et al. An epidemiological and molecular study of
tiple nasopharyngeal carcinoma-​ associated loci the relationship between smoking, risk of naso-
within the HLA region at chromosome 6p21.3. pharyngeal carcinoma, and Epstein-​ Barr virus
Am J Hum Genet 2009;85:194–​203. activation. J Natl Cancer Inst 2012;104:1396–​410.
Tse LA, Yu IT, Mang OW, Wong SL. Incidence rate Xue WQ, Qin HD, Ruan HL, Shugart YY, Jia WH.
trends of histological subtypes of nasopha- Quantitative association of tobacco smoking
ryngeal carcinoma in Hong Kong. Br J Cancer with the risk of nasopharyngeal carcinoma:  a
2006;95:1269–​73. comprehensive meta-​ analysis of studies con-
Vaughan TL, Davis S. Wood dust exposure and squa- ducted between 1979 and 2011. Am J Epidemiol
mous cell cancers of the upper respiratory tract. 2013;178:325–​38.
Am J Epidemiol 1991;133:560–​4. Yang XR, Diehl S, Pfeiffer R, Chen CJ, Hsu WL,
Vaughan TL, Shapiro JA, Burt RD, Swanson GM, Dosemeci M et  al., Chinese, American Genetic
Berwick M, Lynch CF, Lyon JL. Nasopharyngeal Epidemiology of NPCST. Evaluation of risk factors
cancer in a low-​ risk population:  defining risk for nasopharyngeal carcinoma in high-​risk naso-
factors by histological type. Cancer Epidemiol pharyngeal carcinoma families in Taiwan. Cancer
Biomarkers Prev 1996;5:587–​93. Epidemiol Biomarkers Prev 2005;14:900–​5.
Vaughan TL, Stewart PA, Teschke K, Lynch CF, Yao KT, Wu PN, Jiang JW. The role of promot-
Swanson GM, Lyon JL et al. Occupational expo- ion in the carcinogenesis of nasopharyngeal
sure to formaldehyde and wood dust and naso- carcinoma. In Wagner G, Zhang LG. Cancer
pharyngeal carcinoma. Occup Environ Med of the Liver, Esophagus, and Nasopharynx.
2000;57:376–​84. New York: Springer-​Verlag, 1987: 187–​93.
18

Nasopharyngeal Cancer 181

Ye Y-​F, Xiang Y-​Q, Fang F, Gao R, Zhang L-​F, Xie S-​H antibody-​positive persons in Wuzhou City, China.
et al. Hepatitis B virus infection and risk of naso- Int J Cancer 1985;36:545–​7.
pharyngeal carcinoma in southern China. Cancer Zeng Y, Zhong JM, Mo YK, Miao XC. Epstein-​
Epidemiol Biomarkers Prev 2015;24:1766–​73. Barr virus early antigen induction in Raji cells
Yu IT, Chiu YL, Wong TW, Tang JL. Deaths from by Chinese medicinal herbs. Intervirology
nasopharyngeal cancer among waiters and wait- 1983;19:201–​4.
resses in Chinese restaurants. Int Arch Occup Zeng YX, Jia WH. Familial nasopharyngeal carci-
Environ Health 2004;77:499–​504. noma. Semin Cancer Biol 2002;12:443–​50.
Yu KJ, Hsu WL, Chiang CJ, Cheng YJ, Pfeiffer RM, Zhang F, Zhang J. Clinical hereditary characteristics
Diehl SR et  al. Cancer patterns in nasopharyn- in nasopharyngeal carcinoma through Ye-​Liang’s
geal carcinoma multiplex families in Taiwan. Int J family cluster. Chin Med J (Engl) 1999;112:185–​7.
Cancer 2009;124:1622–​5. Zhang L, Chen QY, Liu H, Tang LQ, Mai HQ. Emerging
Yu MC, Garabrant DH, Huang TB, Henderson BE. treatment options for nasopharyngeal carcinoma.
Occupational and other non-​dietary risk factors Drug Des Devel Ther 2013;7:37–​52.
for nasopharyngeal carcinoma in Guangzhou, Zheng J, Zhang C, Jiang L, You Y, Liu Y, Lu J et  al.
China. Int J Cancer 1990;45:1033–​9. Functional NBS1 polymorphism is associated
Yu MC, Ho JH, Lai SH, Henderson BE. Cantonese-​ with occurrence and advanced disease status
style salted fish as a cause of nasopharyngeal car- of nasopharyngeal carcinoma. Mol Carcinog
cinoma:  report of a case-​control study in Hong 2011;50:689–​96.
Kong. Cancer Res 1986;46:956–​61. Zheng W, McLaughlin JK, Gao YT, Gao RN, Blot WJ.
Yu MC, Huang TB, Henderson BE. Diet and naso- Occupational risks for nasopharyngeal cancer in
pharyngeal carcinoma:  a case-​ control study Shanghai. J Occup Med 1992;34:1004–​7.
in Guangzhou, China. Int J Cancer 1989; Zheng X, Yan L, Nilsson B, Eklund G, Drettner B.
43:1077–​82. Epstein-​Barr virus infection, salted fish and naso-
Yu MC, Mo CC, Chong WX, Yeh FS, Henderson BE. pharyngeal carcinoma:  a case-​ control study in
Preserved foods and nasopharyngeal carcinoma: a southern China. Acta Oncol 1994a;33:867–​72.
case-​control study in Guangxi, China. Cancer Res Zheng YM, Tuppin P, Hubert A, Jeannel D, Pan YJ,
1988;48:1954–​9. Zeng Y et al. Environmental and dietary risk fac-
Yu WM, Hussain SS. Incidence of nasopharyngeal tors for nasopharyngeal carcinoma: a case-​control
carcinoma in Chinese immigrants, compared with study in Zangwu County, Guangxi, China. Br J
Chinese in China and South East Asia:  review. J Cancer 1994b;69:508–​14.
Laryngol Otol 2009;123:1067–​74. Zhong BL, Zong YS, Lin SX, Zhang M, Liang YJ. Epstein-​
Yuan JM, Wang XL, Xiang YB, Gao YT, Ross RK, Yu Barr virus infection in precursor lesions of naso-
MC. Non-​ dietary risk factors for nasopharyn- pharyngeal carcinoma. Ai Zheng 2006;25:136–​42.
geal carcinoma in Shanghai, China. Int J Cancer Zhu K, Levine RS, Brann EA, Hall HI, Caplan LS,
2000a;85:364–​9. Gnepp DR. Case-​ control study evaluating the
Yuan JM, Wang XL, Xiang YB, Gao YT, Ross RK, Yu homogeneity and heterogeneity of risk factors
MC. Preserved foods in relation to risk of naso- between sinonasal and nasopharyngeal cancers.
pharyngeal carcinoma in Shanghai, China. Int J Int J Cancer 2002;99:119–​23.
Cancer 2000b;85:358–​63. Zhuo X, Cai L, Xiang Z, Li Q, Zhang X. GSTM1 and
Zeng Y, Miao XC, Jaio B, Li HY, Ni HY, Ito Y. Epstein-​ GSTT1 polymorphisms and nasopharyngeal can-
Barr virus activation in Raji cells with ether cer risk:  an evidence-​based meta-​analysis. J Exp
extracts of soil from different areas in China. Clin Cancer Res 2009;28:46.
Cancer Lett 1984;23:53–​9. Zou J, Sun Q, Akiba S, Yuan Y, Zha Y, Tao Z et al. A case-​
Zeng Y, Zhang LG, Wu YC, Huang YS, Huang NQ, control study of nasopharyngeal carcinoma in the
Li JY et  al. Prospective studies on nasopharyn- high background radiation areas of Yangjiang,
geal carcinoma in Epstein-​Barr virus IgA/​VCA China. J Radiat Res 2000;41(Suppl):53–​62.
182
183

9
Esophageal Cancer
C H R I S T I A N C . A B N E T, O L O F N Y R É N , A N D H A N S - ​O L O V   A D A M I

A n infamous “Asian esophageal cancer belt”

stretches in the east from the Caspian
Littoral in Iran via Turkmenistan, Uzbekistan,
Symptoms Dysphagia due to obstruction is the
typical presenting symptom. As disease pro-
gresses, weight loss, bleeding, inability to swal-
and Kazakhstan to the northern and central low, regurgitation, pain, and fistulas contribute
provinces of China. The Gonbad region in the to the generally disastrous disease development.
northeastern section of Iran and the Linxian and Diagnosis Esophagogastroscopy with biopsy con-
Cixian regions of China have the highest reported firms the diagnosis. Various imaging techniques
incidence rates in the world. In some of these are used for staging and treatment planning.
areas, the incidence rates of esophageal squa-
Treatment Radical surgical removal of a local-
mous cell carcinoma exceed by 100-​fold those in
ized tumor, an extensive procedure, is the only
low-​risk Western countries including the United
curative treatment when diagnosed at stage 2 or
States. Another apparent band stretches from the
greater. With the advent of systematic screening
Great Rift Valley in Eastern Africa to Southern
and the diagnosis of small tumors and advanced
Africa, and although lack of cancer registries
preneoplastic lesions, a number of more con-
prevents precise description of the cancer burden
servative therapies have allowed organ-​sparing
in these regions, reports suggest it is quite high.
treatments in cases with squamous cell carci-
Although studies have identified risk factors in
noma. Such screening and early intervention is
low and high-​incidence regions, no comprehen-
not widely deployed and is restricted to high-​risk
sive explanation exists for these large geographic
populations in China. Most esophageal cancers
differences. A new enigma in esophageal cancer
are diagnosed in incurable and rapidly progress-
epidemiology arose in the 1980s, when descrip-
ing stages. Various modalities including mechan-
tive data from several Western countries revealed
ical dilation, self-​expandable stents, radiotherapy,
a very rapid increase in the incidence of esopha-
and chemotherapy are used to improve passage
geal adenocarcinomas.
and achieve at least short-​term palliation. In very
The two histopathologic types of esophageal
resource limited populations, dilation and self-​
cancer, squamous cell carcinoma and adeno-
expanding metal stents may be the only pallia-
carcinoma, both have dismal clinical courses.
tion available.
Geographic and temporal patterns for these dis-
eases are quite discordant, and recent scholarship Prognosis Esophageal cancer is one of the most
has demonstrated pronounced differences in deadly of all malignancies. Currently, 18% of
etiology with little or no overlap in the primary the cases in the United States survive for at least
risk factors, whether environmental or genetic. 5  years. There has been some modest improve-
To best understand these two diseases research- ment in 5-​year survival in some high-​resource
ers should study them separately and in parallel settings. In most low-​resource settings, where
whenever possible. the preponderance of esophageal cancers are
diagnosed, 5-​year survival is substantially lower.
CLINICAL SYNOPSIS Progress Despite the absence of any revolutionary
Subgroups Esophageal cancer occurs mainly in breakthrough in diagnosis or treatment, there
two histologic types, squamous cell carcinoma are hopeful signs of an ongoing improvement.
and adenocarcinoma. Presently, the clinical man- In Sweden, the 5-​year relative survival after an
agement of these types is similar. esophageal adenocarcinoma diagnosis increased
184

184 Textbook of Cancer Epidemiology

from 5% in 1961–​89 to 13.7% in 1990–​96, while to the Eastern Cape Province of South Africa.
the corresponding survival after esophageal squa- The Gaucho areas of South America, centering
mous cell carcinoma rose from 5% in 1961–​69 to on Uruguay, historically had high incidence as
8.9%. Similar observations have been made in well, but excess rates in this region were never
the United States for adenocarcinoma as well as as extreme as the belts and seem to be dropping
for squamous cell carcinoma. But more than 80% rapidly in recent years.
of esophageal cancer is diagnosed in economi- In low-​ incidence areas, male predomi-
cally developing regions, and the worldwide bur- nance is the norm for both esophageal cancer
den of esophageal cancer will likely not be altered cell types, with up to a ninefold difference in
by these modest improvements in high-​resource risk for esophageal adenocarcinoma between
low-​incidence populations. men and women. Populations with high inci-
dence of esophageal squamous cell carcinoma
DESCRIPTIVE see much smaller sex ratios, with the most
EPIDEMIOLOGY extreme report coming in the 1970s from Iran,
where the cancer was in excess in women. The
Geographic and Demographic age-​standardized incidence rates among men
Distribution and women in the extreme-​risk Gonbad region
Worldwide, cancer of the esophagus ranks as the of Iran were 206 and 262 per 100,000 person-​
sixth leading cause of cancer death, with over years, respectively, in the 1970s (Hormozdiari
400,000 cases annually. There is a wide variation et al, 1975); while still being the most common
in risk among geographic areas (Figs. 9-​1, 9-​2,). cancer, the rates in 1996–​2000 seem to have
These figures present total esophageal cancer, but fallen to less than half (Semnani et  al, 2006).
geographic patterns show that nearly all esopha- The corresponding rates in Linxian, China, are
geal cancers in East Asian populations are squa- 138 and 99 and in Cixian, China, 209 and 120
mous cell carcinomas, while Western countries (He et al, 2005). China contributes almost half
have seen an increasing dominance of esopha- of all esophageal cancer cases occurring world-
geal adenocarcinoma with this cell type exceed- wide (Parkin et al, 1999).
ing 50% of cases in the United States. The age-​specific incidence rates for women
Two belts of high esophageal cancer inci- (Fig. 9-​3) and men (Fig. 9-​4) show that esoph-
dence have been described, with one stretch- ageal cancer occurs rarely before the age of
ing from northeastern Iran to north central 45  years, although in China this diagnosis
China across Central Asia and a second follow- needs to be seriously considered already at ages
ing the Great Rift Valley from Ethiopia south 45–​55. Noteworthy in Fig. 9-​3 is the relatively high

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 1 2 3 4 5 6 7 8
ASR (world) per 100,000

FIGURE 9-​1  Age-​standardized (to the 2012 world population) incidence rates of esophageal cancer among women.
Source: Ferlay et al, 2013.
185

Esophageal Cancer 185

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 5 10 15 20
ASR (world) per 100,000

FIGURE 9-​2  Age-​standardized (to the 2012 world population) incidence rates of esophageal cancer among men.
Source: Ferlay et al, 2013.

incidence among elderly women in the United Secular Trends by Histologic Types

Kingdom. Several reports from East Africa have The rapid change in incidence from squamous
identified populations where esophageal squa- cell carcinoma to adenocarcinoma in the United
mous cell carcinoma in young people is not unu- States and several other high-​ income coun-
sual (White et al, 2002; Parker et al, 2010). These tries with Western lifestyles, notably in North
regions are not covered by comprehensive cancer America, Europe, Australia, and New Zealand,
registries, thus rates in these age groups cannot remains an unexplained phenomenon. A detailed
be stated. Mean age at diagnosis in sub-​Saharan analysis from the Connecticut tumor registry,
Africa is lower, but this is not a useful method for one of the world’s oldest, shows evidence of an
assessing the presence of young cases, since the incipient increase stretches back many decades
dearth of aged people in these populations nec- (Abrams et al, 2011).
essarily leads to younger mean ages at diagno- The occurrence of this cancer in Scotland
sis. Nevertheless, the absolute number of young is notably high, where it appears that the inci-
cases in East Africa suggests a novel pattern of dence among men is approaching 10 per 100,000
incidence. person-​years (Pera et al, 2005). An incidence of

120
New cases per 100,000

100
person-years

80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 9-​3  Age-​specific incidence rates of esophageal cancer among women.

Source: Ferlay et al, 2013.
186

186 Textbook of Cancer Epidemiology

New cases per 100,000

250
200

person-years
150
100
50
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 9-​4  Age-​specific incidence rates of esophageal cancer among men.

Source: Ferlay et al, 2013.

7 per 100,000 person-​years has been reported the most common esophageal cancer type in
from England and Wales (Newnham et al, 2003). US white men (Vizcaino et  al, 2002;Brown &
As a comparison, the incidence rate among US Devesa, 2002); Australian men; English, Welsh,
white men in 2001, having increased sixfold since and Scottish men; and Swedish men (Ye & Nyren,
1973–​ 75 and thereby being the most rapidly 2003; Ji & Hemminki, 2006).
increasing cancer of all (Blot et al, 1991; Pohl & Few reports regarding trends in esophageal
Welch, 2005), was no more than 2.3 per 100,000 cancer by cell type have been published using
person-​year. data from South America, which is of substantial
Although changing classification practices interest given the high rates of esophageal squa-
might cloud the picture (Polh & Welch, 2005; mous cell carcinoma in and around Uruguay and
Botterweck et  al, 2000; Lindblad et  al, 2006), the notable changes in incidence patterns seen in
this seems unlikely, as does overdiagnosis. countries with populations of European descent.
Therefore, as noted in a global assessment of A hospital record pathology-​based report from far
the incidence of esophageal adenocarcinoma Southern Brazil reported that between 1993 and
there is wide consensus that we are witnessing 2012, there were no apparent changes in the rate
a real and important increase in the occurrence of diagnosis of these two cancers and that 94% of
of esophageal adenocarcinoma in Western all esophageal cancer diagnosed were esophageal
populations, that calendar period may be more squamous cell carcinoma (Fagundes et al, 2015).
important than birth cohort, and that this pat- This report highlights the need for population-​
tern is consistent with the phased introduction based registries to accurately track the changing
of a new etiologic factor in the 1950s (Edgren demographic patterns for cancer for a clear under-
et al, 2013). standing of challenges faced in local populations.
Simultaneously, the incidence of esopha- Within the US population, esophageal cancer
geal squamous cell carcinoma, although varying show dramatic differences by cell type for race
sevenfold across studied Western registries, has and sex. The rate of esophageal squamous cell
been more stable over time (Vizcaino et al, 2002); carcinoma in blacks is about twice that of whites,
increasing trends were observed in Denmark and this ratio remains unchanged during the
and the Netherlands among men and in Canada, gradual fall in rates for this specific cancer type.
Scotland, and Switzerland among women, but in And the recent dramatic rise of esophageal ade-
most of the countries analyzed, the rates in men nocarcinoma is noted in whites and blacks and
declined slowly, which may be due to long-​term in men and women, but overall this disease has
reductions in tobacco smoking. In contrast to much higher incidence in white males. Rates in
the marked male predominance in adenocarci- white men, white women, black men, and black
noma, a less extreme male/​female ratio (3–​4:1) women between 1977 and 2005 were 3.7, 0.5,
was observed for esophageal squamous cell car- 0.8, and 0.2 per 100,000, respectively (Cook et al,
cinoma (Vizcaino et al, 2002). Adenocarcinoma 2009). And rates of esophageal adenocarcinoma
has surpassed squamous cell carcinoma as are increasing in all groups, the 18-​fold difference
187

Esophageal Cancer 187

between white men and black women appears significant reductions in esophageal cancer mor-
to be growing when examining the ratios in the tality (Wei et al, 2015). Furthermore, the use of
most recent years. surgical and or destructive therapies on these
Recent efforts to correctly classify all inci- lesions reduced the incidence of the cancer.
dent cases in the high-​risk Golestan province
in Iran revealed that esophageal adenocarci- Barrett’s Esophagus and
noma remains fairly rare; only 6% of investigated Esophageal Adenocarcinoma
esophageal neoplasms were of this histological There is a broad consensus that Barrett’s esoph-
type (Islami et  al, 2004). Similar conclusions agus in most cases is a consequence of long-​
are drawn also in Linxian, China (Islami et  al, standing gastroesophageal reflux (Winters et  al,
2004) and Japan (Blaser & Saito, 2002), although 1987). This metaplasia is seen in approximately
there seems to be an increasing trend in esopha- 10% of patients who are investigated for reflux
geal adenocarcinoma mortality in the latter pop- disease. Three histologic subtypes of Barrett’s
ulation (Hongo, 2004). Registries in other parts of esophagus have been described: the fundic type,
the world have attempted to assess this phenome- the junctional type, and the specialized intestinal-​
non, with mixed results. The Hong Kong Tumor like type. Current definitions of Barrett’s esoph-
Registry reported a vanishingly small number of agus tend to require the presence of specialized
esophageal adenocarcinomas and that the trend intestinal metaplasia in addition to a macroscop-
was toward lower, not higher, incidence. ically visible lesion. Early clinical observations
indicated that the risk of developing adenocarci-
Preneoplasia noma in Barrett’s esophagus is substantial. This
The presence and identity of a distinct preneo- increased risk appears to be confined to patients
plastic phase in esophageal cancer of both cell with the specialized intestinal-​like subtype.
types is an important consideration for both The magnitude of the excess risk is a mat-
research and clinical practice. Preneoplastic ter of controversy, although review articles often
lesions are often used as independent outcomes mention a 50-​to 100-​fold increase (Streitz, 1994).
for etiologic studies of esophageal cancer, with Methodologically, studies of Barrett’s esopha-
some uncertainty about the appropriate inter- gus and adenocarcinoma risk are hampered by
pretations, and they serve as targets for screening small numbers (Shaheen et al, 2000) and by pos-
and early intervention. sible selection and publication bias. Selection bias
might inflate risk estimates in clinical case series
Esophageal Squamous Dysplasia coming from centers with a special interest in
and Esophageal Squamous esophageal cancer. And a systematic review gave
Cell Carcinoma clear indications of publication bias (Shaheen et al,
Early studies of esophageal cancer in Iran and 2000). Rather than consider only the relative risk of
China identified several common histologic developing cancer among subjects diagnosed with
changes in populations with high rates of this can- Barrett’s esophagus, a valuable metric is the annual
cer, including frequent evidence of esophagitis, rate of conversion to esophageal adenocarcinoma.
basal cell hyperplasia, and esophageal squamous Meta-​analyses of the existing literature have sug-
dysplasia. All these lesions were used as research gested that the conversion rate from Barrett’s
targets in older literature, but more recently sev- esophagus to esophageal adenocarcinoma was
eral long-​term follow studies have shown that approximately 5–​6 cases per 1,000 person-​years
only the presence of esophageal squamous dys- (Yousef et al, 2008; Sikkema et al, 2010).
plasia is linked to higher risk of future esophageal Because high-​grade dysplasia may trigger a
squamous cell carcinoma development (Want clinical intervention, estimates of the combined
et al, 2005;Taylor et al, 2013). The clear stepwise end point of high-​ grade dysplasia or cancer
increase in risk linked to higher histologic grades are typically offered as well, and these studies
of esophageal squamous dysplasia strongly impli- suggest that the rate is about twice that for the
cates this lesion as the true preneoplastic phase. cancer alone, or 9–​10 per 1,000 person-​years.
A community assignment trial showed upper Another population-​based study from Denmark
endoscopy with Lugol’s iodine staining, which reported a much lower rate of only 1.2 per 1,000
was specifically designed to locate and tar- person-​years (Hvid-​Jensen et  al, 2011), which,
get esophageal squamous dysplasia, can lead to if correct, would have a substantive impact on
18

188 Textbook of Cancer Epidemiology

clinical response to this diagnosis. These esti- (SNPs) at 10q23 that harbors the gene PLCE1.
mates are closer to a report from Ireland (1.3%) RS 2274223 has an affect allele frequency of 21%
(Bhat et al, 2011) than estimates from the prior in these studies, and the summary OR was 1.35
meta-​analysis. There are important differences (95% CI 1.26–​1.45) per allele with p = 3.65 × 10–​18
in basis of Barrett’s esophagus diagnosis between with no evidence for variation between popula-
countries and differences in risk based on sub- tions or by a limited number of other exposures,
classification of Barrett’s esophagus by grade of such as tobacco smoking or alcoholic beverage
histologic dysplasia or molecular differences, consumption. A mouse knockout model for the
which may provide valuable information on an PLCE1 has shown strong differences in suscep-
individual’s risk of progression, and which should tibility to various carcinogens including higher
be considered in clinical guidelines (Shaheen susceptibility to chemically induced skin cancer
et al, 2016). (Bai et al, 2004), higher susceptibility to geneti-
cally induced small bowel cancer (Li et al, 2009),
GENETIC AND MOLECULAR but greater resistance to UV-​induced skin cancer
EPIDEMIOLOGY (Oka et al, 2010).
A study from South Africa showed sub-
High Penetrance Syndromes stantially different linkage disequilibrium (LD)
Tylosis, an autosomal dominant trait character- patterns at 10q23, but there appeared to be an
ized by hyperkeratosis palmaris et plantaris, is association between SNPs in this gene and risk of
associated with a substantial excess risk of esoph- esophageal squamous cell carcinoma in the black
ageal squamous cell carcinoma. In a UK pedigree, population (Bye et al, 2012).
more than 90% of the affected family members Other confirmed overall associations include
developed esophageal cancer before age 70 (Ellis loci at 5q31.2, 17p13.1, 2q33.1, 21q22.12, and
et  al, 1994). Susceptibility has been mapped to 22q12.1, near genes TMEM173, TP53, CASP8,
missense mutations in RHBDF2 at chromo- RUNX1, and CHEK2, respectively. Each of
some 17q25 (Blaydon et  al, 2012). These muta- these loci has been linked to cancer risk at other
tions alter expression in both tumor and normal organ sites. For example, numerous studies have
esophageal tissue and may be working through reported links between SNPs near CASP8 and
epidermal growth factor receptor (EGFR). Apart risk of cancer in the lung, breast, pancreas, and
from tylosis, there are no reports of esophageal others (Abnet et al, 2012). The signal for esoph-
cancer families with defined germline mutations. ageal squamous cell carcinoma, which resides in
an adjacent LD block, suggests the biological rel-
Common Genetic Variation evance of the genomic region, but potential dif-
and Esophageal Cancer Risk—​ ferences in the biology of risk, which remains to
Esophageal Squamous Cell be explored in more detail. Two of the original
Carcinoma reports were subsequently updated, and the best
The advent of chip-​based genotyping revolution- estimates for these three studies are reported here
ized the practice of assessing the role of common (Wu et al, 2014).
genetic variants in disease risk. By 2012, three One geographically defined signal was
genome-​ wide association studies (GWAS) of reported at 6p21.32, which harbors the human
esophageal squamous cell carcinoma had been leukocyte antigen (HLA) class  II genes. This
conducted and analyzed in various combina- region also has been linked to several pathogen-​
tions (Abnet et  al, 2010; Wang et  al, 2010; Wu induced cancers, such as nasopharyngeal cancer
et al, 2011; Wu et al, 2012). All three studies used (Epstein Barr virus, or EBV) and hepatocellular
subjects of Chinese ethnicity and all three data- carcinoma (hepatitis B virus, or HBV), but also
sets have been jointly analyzed (Wu et al, 2014). cancers thought to have primarily other etio-
With just over 5,000 cases scanned, the power logic agents, such as never-​smoking lung cancer.
to detect associations with common genetic Again, the concentration of GWAS hits in this
variants remains modest, and this is further region strongly suggests its biological impor-
limited given some evidence for population-​ tance. But given that the finding was limited to
specific effects within China (Wu et al, 2012;Wu the two scans that used people from high-​risk
et al, 2014). regions and the unusal long-​range LD in this
The top hit in all three studies was a single sig- genomic region, our understanding of this find-
nal from set of single nucleotide polymorphisms ing is limited.
189

Esophageal Cancer 189

These studies reported distinct results for including hits at 19p14 in CRTC1, at 9q22 in
alcohol metabolizing genes, which likely reflects BARX1, and at 3p14 near FOXP1. Variants at
differences in alcohol consumption by the sub- 19p14 appear to have pleiotropic effects on phys-
jects in the individual studies. iology, including age at menarche and obesity.
The BARX1 gene has been linked to esophageal
Alcohol-​Metabolizing Gene development in mouse embryos and FOXP1 has
Polymorphisms also been linked to esophageal and lung develop-
The extensive literature on hypothesis-​ driven ment. This study also attempted to replicate find-
SNPs and risk of esophageal cancer has mostly ings from a GWAS for Barrett’s esophagus (Su
been repudiated by subsequent work using et al, 2012) and found evidence of an association
GWAS, with one notable exception being alcohol-​ at 16q24 near a possible tumor suppressor gene
metabolizing enzyme genes in East Asians. FOXF1, in both Barrett’s esophagus and esopha-
Common SNPs in alcohol-​metabolizing genes geal adenocarcinoma, suggesting a shared mech-
ADH and ALDH2 were known to convey extreme anism for the preneoplastic lesion and the tumor.
phenotypes on subjects with these variants that Other studies using this and other data sets
consumed alcohol, and these variants were con- have estimated the total effect of common genetic
sistently linked to esophageal squamous cell car- variants to esophageal adenocarcinoma and
cinoma risk (Brooks et al, 2009). These variants Barrett’s esophagus (Ek et al, 2013). The authors
are private to East Asian populations. GWAS of estimate the genetic heritability of Barrett’s
alcohol consumption have collected more than esophagus at 35% and of esophageal adenocarci-
100,000 subjects without defining SNPs clearly noma at 25%, and suggested that high correlation
linked to alcohol intake in whites. Nevertheless, between the two suggested a substantial set of
there is evidence that variants in ADH3 in whites shared genes to be uncovered. Other analyses have
modifying the risk of alcohol-​associated head shown independent Mendelian randomization–​
and neck cancer in whites (Hashibe et al, 2008). based evidence for the role of obesity (Thrift et al,
It is important to note that it appears that these 2014), height (Thrift et  al, 2014), and possibly
genes only convey risk for esophageal squamous androgen levels (Ek et al, 2016).
cell carcinoma in subjects with higher intake of
beverage alcohol. The three GWAS of esopha- Somatic Events
geal squamous cell carcinoma in China highlight Esophageal Squamous Cell Carcinoma
the importance of the gene–​environment inter- Sequencing of esophageal squamous cell car-
action. The GWASs from Abnet (Abnet et  al, cinoma tumor tissue has revealed a number of
2010)  and Wang (Wang et  al, 2010)  were con- frequent genetic alterations. These alterations
ducted using subjects from the high-​risk regions may represent molecular fingerprints of critical
of China where alcohol consumption is almost risk factors. It is therefore of interest to compare
uniformly low and neither showed evidence for not only cancers of different histologic types or
effect as ADH2 or ALDH2 polymorphisms on clinical expressions but also cancers arising in
esophageal squamous cell carcinoma risk. In geographic areas with different incidence pat-
contrast, the reports from Wu (Wu et  al, 2011; terns. In recent years, the continuing drop in
Wu et  al, 2012)  used subjects from Beijing and the cost of ultra-​high-​throughput sequencing is
other cities, where alcohol consumption is more leading and will lead to major advances in this
common and more intense, and in this group a area. Several databases, including COSMIC
clear effect of ADH2 and ALDH2 SNPs was evi- (http:// ​ c ancer.sanger.ac.uk/ ​ c ancergenome/​
dent, although differences within subpopulations projects/​cosmic), provide up-​ to-​
date compi-
were noted by the authors. lations of mutation results. IARC maintains a
specific p53 mutation database (http://​p53.iarc.
Common Genetic Variation fr/​) that is searchable by tumor and cell type.
and Esophageal Cancer Risk—​ A  major limitation of the older literature was
Esophageal Adenocarcinoma the requirement that researchers focus on select
A single GWAS with 2,390 esophageal adeno- targets to assess mutation. While this can pro-
carcinomas has examined the roles of common duce valid estimates of mutation prevalence in
genetic variants in risk of esophageal adenocar- the targeted genes, it does not give the overall
cinoma (Levine et al, 2013). This study reported view of mutation events currently offered via
three genome-​ wide significant associations, whole-​exome and whole-​genome sequencing.
190

190 Textbook of Cancer Epidemiology

Future progress will require transparent data Inactivation of the CDKN2A (p16INK4a)
sharing and new methods development, espe- gene appears to be the second most commonly
cially in bioinformatics and statistical analyses mutated gene in esophageal squamous cell car-
of these data that can show the true patterns of cinoma (Abedi-​Ardekani & Hainaut, 2014). In
interdependence of specific mutations in tumor the early stages of carcinogenesis, this occurs
biology. This vast and rapidly proliferating lit- mainly through methylation or loss of hetero-
erature is not reviewed in detail here, but a brief zygosity (LOH), while homozygous deletion is a
general overview seems appropriate. late event (Kuwano et  al, 2005). The prevalence
The highest mutation rates for esopha- of CDKN2A mutations ranged between 0% and
geal squamous cell carcinoma are reported to 68% in 12 studies. While the prevalence varied
occur in TP53 (Abedi-​ Ardekani & Hainaut, between 10% and 30% in most studies, it was
2014). Although this gene has also received the highest (>50%) among Chinese and Japanese
greatest degree of scrutiny, which may over- patients, p15INK4a alterations were common
estimate the dominance of mutations in this (50%) in one study (Xing et  al, 1999)  but rare
gene, whole-​genome sequences appear to con- (3%) in another (Suzuki et  al, 1995). It appears
firm that this is the most frequently mutated that p15INK4a is inactivated through abnormal
gene in esophageal squamous cell carcinoma. methylation or homozygous deletion at the same
Mutations in the TP53 gene may be an early time as CDKN2A. The simultaneous inactivation
event in the development of esophageal squa- of these genes results in loss of the pRb-​regulated
mous cell carcinoma because alterations in p53 restriction point, and plays an essential role in
protein expression are frequently observed in esophageal carcinogenesis.
precursor lesions (Tian et  al, 1998; Roshandel A small number of esophageal squamous
et al, 2014). Clear differences in TP53 mutation cell carcinoma tumors collected in the United
patterns occur between geographical regions, States (Agrawal et al, 2012) and over 200 esoph-
which likely reflects the primary carcinogens. ageal squamous cell carcinoma tumors from
Notable differences in TP53 mutations have China (Song et al, 2014; Lin et al, 2014; Gao et al,
been reported when comparing European 2014; Zhang et al, 2015) have been subjected to
tumors to Chinese tumors. It is hoped that either whole-​exome or whole-​genome sequenc-
investigation of mutation patterns in under- ing. Such focused analysis can provide the most
studied populations can lead to insight into pri- informative description of mutation patterns and
mary carcinogens. As one example, mutation mutation loads. Data in this area are developing
analysis in Iran produced a number of etiologic rapidly, making any summary quite temporary.
hypotheses regarding esophageal squamous Collectively these studies have reported muta-
cell carcinoma in this population (Sepehr et al, tions in 31 different genes with by far the highest
2001; Biramijamal et al 2001), which appears to prevalence in TP53 (about 77%). Five other genes
have the highest TP53 mutation rates anywhere have collectively shown mutations rates greater
(Abedi-​Ardekani et al, 2011). than 10% including CCND1, KMT2D, NOTCH1,
Cell cycling is controlled strictly through FAT1, and FAT3.
two major pathways, the p53 (p14-​MDM2-​p53-​ Some of these are known cancer-​associated
p21) and pRb (p16-​cyclin D1-​pRb) pathways. genes, while others appear to be novel findings,
The prevalence of TP53 gene aberrations ranges including FAM135B, which has been shown to
between 10% and 85% and appears higher in be mutated in 6 out of 88 tumors tested in func-
high-​than in low-​incidence areas. In the United tional assays (Song et  al, 2014). This study also
States and Europe, TP53 mutations are seen more highlighted some overlap with mutation in squa-
frequently in smokers than in nonsmokers. The mous head and neck cancer and the potential
most common site of the mutations varies with role of alcohol consumption.
geographic area (Lam et  al, 2000). In Western
and Japanese populations, the TP53 mutations Esophageal Adenocarcinoma
are evenly distributed in exons 5, 6, and 8.  In Adenocarcinomas of the esophagus are thought
China, they tend to occur more often in exon 5, to develop through a stepwise process termed
in Taiwan in exons 6 and 7, and in Hong Kong the metaplasia-​dysplasia-​carcinoma sequence
in exons 7 and 8. Moreover, in Hong Kong and (Souza & Spechler, 2005). Metaplasia is the proc-
mainland China, they are seen as often in non- ess whereby one adult cell type replaces another
smokers as in smokers. adult cell type. In the esophagus the squamous
19

Esophageal Cancer 191

epithelium is replaced by metaplastic specialized The value of comprehensive sequencing stud-

intestinal-​
type columnar epithelium, termed ies is evident in the first major report on esopha-
Barrett’s esophagus. This metaplasia constitutes geal adenocarcinoma (Dulak et al, 2013), which
a possibly obligatory precursor lesion for esoph- confirmed frequent mutation in five known
ageal adenocarcinoma, but this remains to be genes, but also reported 21 new mutation tar-
determined conclusively. gets. The popularity of the “Vogelgram” has built
It appears that cell-​cycle abnormalities are expectations that the accrual of gene mutations
mainly caused by inactivation of genes that block may follow a predictable stage-​ specific model
Rb function, such as P53 and CDKN2A. With a during the precursor and early tumor sequence.
prevalence of up to 90%, mutations of P53 appear But several studies dispute this notion and sug-
to be more common in adenocarcinoma than in gest that the accumulation of mutations may be
squamous cell carcinoma (Taniere et  al, 2001). more complex (Weaver et al, 2014) and that these
Such mutations have already been observed in changes may not be well correlated with tradi-
Barrett’s esophagus. The prevalence of P53 altera- tional histologic categorizations (Ross-​ Innes
tions increases with increasing severity of epithe- et al, 2015). And the model of orderly accumu-
lial dysplasia from 5% to 10% in indeterminate lation of mutations has also been challenged for
dysplasia to 75% of cases with high-​grade dyspla- esophageal adenocarcinoma by studies show-
sia (Younes et al, 1993). Further, the prevalence ing differences in mutations with either overall
increases with increasing DNA ploidy abnor- genome doubling (Stachler et al, 2015) or by so-​
malities from 1% to 5% in diploid cell popula- called genomic catastrophes of chromothripsis
tions to almost 100% in aneuploid cells (Galipeau (Nones et al, 2014).
et al, 1996). Another genetic alteration of uncertain sig-
As in squamous cell carcinoma, mutations nificance is the frequent loss of the whole Y chro-
of genes TP53 also occur frequently in esoph- mosome, which was observed in one study in 9%
ageal adenocarcinoma, and the mutation rate of cases with Barrett’s metaplasia without dys-
may even be higher (Abedi-​Ardekani & Hainaut, plasia, in 38% of cases with indefinite dysplasia,
2014), but the overall pattern of mutations and in 100% of cases with high-​grade dysplasia
appears quite distinct between the two diseases. (Barrett et al, 1996). APC tumor suppressor gene
The pattern of TP53 mutations in esopha- mutations and beta-​ catenin accumulation are
geal adenocarcinoma differs markedly from also common findings in esophageal adenocarci-
that in squamous cell carcinoma (Montesano noma (Wu et al, 1998), and they appear to occur
et al, 1996). Adenocarcinomas show one of the late in the metaplasia-​dysplasia-​adenocarcinoma
highest frequencies (up to 69%) of transition sequence (Ransford & Jankowski, 2000).
at CpG dinucleotides observed in any cancer,
while those transitions are seen in only about RISK FACTORS
18% of squamous cell carcinomas. Other com-
mon aberrations of cell-​cycle genes in esoph- Tobacco and Alcohol—​Esophageal
ageal adenocarcinoma, observable also in the Squamous Cell Carcinoma
Barrett precursor state, include mutations in The positive association between heavy alcohol
CDKN2A and increased cyclin D1 expression. use, tobacco smoking, and the risk of esopha-
As opposed to the squamous cell carcinomas, geal cancer is well established. A  large number
promoter hypermethylation with loss of heter- of studies, including around 60 with case-​control
ozygosity (LOH) appears to be the dominating designs and several prospective cohort studies,
mechanism behind the inactivation of CDKN2A have addressed the association with esophageal
in adenocarcinomas. Most likely contributing to squamous cell carcinoma or unspecified esoph-
this inactivation and the inactivation of P53 is ageal cancer, the latter presumably consisting
the LOH at 9p and 17p that is frequently seen mainly of squamous cell carcinoma, given the
in both Barrett’s esophagus and esophageal ade- years and geography of the completed studies.
nocarcinoma (Galipeau et al, 1999). Inactivation Although some variation does exist, the over-
of the adenomatous polyposis coli gene (APC) all results are unusually consistent; even in the
by promoter methylation and LOH of 5q21 have absence of the other (Tuyns, 1983; Tavani et  al,
been found both in esophageal adenocarcinoma 1994; Castellsague et al, 1999), both smoking and
and in its precursor Barrett’s esophagus with and alcohol habits are strong and independent risk
without high-​grade dysplasia. factors for esophageal squamous cell carcinoma
192

192 Textbook of Cancer Epidemiology

(Fig. 9-​5). In Western (including South America) expresses the percent of the studied outcome dis-
and African populations, the relative risk among ease that can be attributed to the exposure under
heavy smokers, relative to nonsmokers, is typi- study, or in other words, the percent of all cases
cally on the order of 4–​7 (Castellsague et al, 1999; that might be prevented by eliminating this risk
Wynder & Bross, 1961; Cheng et al, 1992; Parkin factor.
et  al, 1994). Some studies in high-​risk regions One review indicated that the risk of esopha-
(Tran et al, 2005; Nasrollahzadeh et al, 2008; Dar geal cancer remains elevated many years (at least
et al, 2012) do not find strong associations with 10) after cessation of smoking, to decline by about
ever smoking or ever drinking, but that seems to 40% only thereafter (Bosetti et  al, 2006). After
be primarily due to modest consumption. 10 years since cessation of smoking, ex-​smokers
All quantitative aspects of tobacco use still have a twofold increased risk as compared to
seem to be dose-​dependently related to the risk never smokers.
(Launoy et  al, 1997; Lagergren et  al, 2000). The There are consistent indications that the use
risk increases throughout the whole dose range, of pipes, cigars, hand-​rolled and/​or high-​tar ciga-
possibly in a less than linear manner (Launoy rettes, bidi, or dark or black tobacco—​all perceived
et al, 1997; Segal et al, 1988). Since cohort studies as “strong” or unrefined tobacco products—​is
yield results similar to those of case-​control stud- associated with steeper risk increases than is the
ies, the association cannot be attributed to recall use of commercially available cigarettes made
bias. A  thorough investigation of how delivery from blond tobacco (Castellsague et  al, 1999;
rate (the momentary smoking “intensity”) mod- Launoy et al, 2000). Various other chewed prod-
ifies the strength of the association suggests that ucts like betel quid (aka areca nut), paan (betel
longer duration with lower intensity may convey with tobacco), and nass (tobacco with lime and
greater risk than higher intensity with shorter ash) (Dar et al, 2012) were repeatedly associated
duration, even when this leads to the same total with an increased esophageal cancer risk. The
exposure (i.e., pack-​years) (Lubin et al, 2012). wide variety of consumption styles and relative
In a US case-​control study, the population dearth of data makes these products difficult
attributable risk percent (PAR) for smoking vis-​ to classify, but excess risks seems highly likely
à-​vis esophageal squamous cell carcinoma was (Akhtar, 2013). Interestingly, for chewing betel
estimated to be 56.9% (Engel et  al, 2003). PAR quid without tobacco, excess risk was at least

25
23,1

20
20
16,2
Relative risk

15

10 8,8
6,9

5 3,2
3,1
2 2,3 2
0,6
1,1 1 1,3
0
Smoking Heavy Heavy Severe “normal”over- obese
>35 years alcohol smoking reflux weight
intake +alcohol Body mass index
Squamous cell carcinoma Adenocarcinoma

FIGURE 9-​5 Major risk factors—​and evidence for different etiologies—​for squamous cell cancer and adenocarcinoma
of the esophagus.
Source: Lagergren J. Cancer of the Esophagus and Gastric Cardia Etiological Aspects. Doctoral thesis (ISBN 91-​628-​3397-​9), Karolinska
Institutet, Stockholm, 1999.
193

Esophageal Cancer 193

as great as for chewing betel quid with tobacco among women than among men. Birth cohort
(Jussawalla & Deshpande, 1971; Nandakumar differences in age at smoking initiation seems
et al, 1996). to explain previously reported differences in
The dose-​risk trend with alcohol consumption tobacco risk estimates by sex for lung cancer,
has been described as linear (Launoy et al, 1997; and the same may be applicable to esophageal
Tuyns et al, 1979) or more than linear (Segal et al, squamous cell carcinoma. In US populations,
1988; Tuyns et al, 1977; Graham et al, 1990), or only recently did men and women have similar
J-​shaped with no excess risk with modest intake ages at initiation (17  years). In western Europe
(Freedman et  al, 2007). Substantial excesses in and North America, 90% or more of the risk
risk are generally seen only in the heavy con- of esophageal cancer can be attributed to alco-
sumption categories (>3 standard drinks/​day), hol and tobacco (Engel et  al, 2003). In a large
where the relative risk in comparison with that South American case-​control study, the fraction
of nonusers typically exceeds 5.  For alcohol, as of all cases that could be attributed to smoking
opposed to tobacco, the current weekly or daily and/​or alcohol use was estimated at 90% among
dose, and not the duration of the habit, appears men and 29% among women (Castellsague et al,
to be the most important determinant of risk 1999), which reflects lower exposure to these car-
(Castellsague et al, 1999; Launoy et al, 1997; Yu cinogens in women.
et al, 1988; Hu et al, 1994). But similar to smok- As pointed out previously, the strong and
ers, moderate and heavy alcohol consumers can multiplicative effects of smoking and drinking
apparently reduce their risk of esophageal squa- are not universally important in the etiology
mous cell carcinoma substantially by restricting of esophageal squamous cell carcinoma. Some
their alcohol use (Cheng et al, 1995). studies in high-​risk regions (Tran et  al, 2005;
The results regarding the importance of Nasrollahzadeh et  al, 2008; Dar et  al, 2012)  do
type of beverage are contradictory. While many not find strong associations with ever smoking or
authors claim that the risk of squamous cell car- ever drinking, but that seems to be primarily due
cinoma is most strongly associated with hard liq- to modest consumption. But in low-​to moderate-​
uor consumption (Lagergren et al, 2000; Brown incidence areas in China, the “Western” risk pat-
et  al, 1994; Gammon et  al, 1997), others report tern prevails (Hu et  al, 1994; Wu et  al, 2006).
that wine and/​or beer drinkers have the highest Subjects in the high-​incidence regions tend to
risks (Segal et al, 1988; Zambon et al, 2000). Most have low numbers of cigarettes per day, which
authors, however, find no important variation in makes comparisons of risk estimates for ever
risk with type of alcoholic beverage. smoking likely to reflect different lifetime expo-
Early on, it was noted that the joint deleteri- sure to tobacco carcinogens. Furthermore, many
ous effects of alcohol and tobacco are not additive people in these populations have high exposures
but act synergistically in a multiplicative manner to smoke from other sources. Nonsmoking sub-
(Tuyns et al, 1977). There is now abundant evi- jects in China, Iran, Kenya, and Brazil have all
dence to support this multiplicative relationship high exposure to polycyclic aromatic hydrocar-
in most populations. The risk among the heavi- bons using urinary biomarkers. This may dilute
est users of both alcohol and tobacco, relative to the effects of tobacco when it is added to high
nonusers of both, is typically increased 20-​to 50-​ background exposure to one tobacco carcinogen.
fold, but the excess may be more than 100-​fold
(Zambon et  al, 2000; Brown et  al, 1994). Some Tobacco and Alcohol—​
recent studies have not reported evidence of Esophageal Adenocarcinoma
interactions, but high consumption of both prod- Compared with the esophageal squamous cell
ucts will lead to high risk for disease regardless of carcinomas, the esophageal adenocarcinomas
the absolute risk estimate. are less strongly associated with tobacco use
Even though the incidence of esophageal (Fig. 9-​5). The interpretation of published data
squamous cell carcinoma is considerably lower on risk factors for the latter histologic type is
among women than among men in Western complicated by the fact that some early studies
populations, the combined literature seems to have lumped adenocarcinomas of the esopha-
indicate that tobacco and alcohol are strong and gus and gastric cardia into one category. This
important risk factors in both sexes (Castellsague complication notwithstanding, relative risk esti-
et al, 1999; Franceschi et al, 1994). However, the mates associated with smoking cluster around
relative risk estimates are systematically lower 1.5–​2.5 in the studies published to date, all with a
194

194 Textbook of Cancer Epidemiology

hospital-​based or population-​based case-​control greater than that of persons with a BMI below
design (Lagergren et  al, 2000; Gammon et  al, 22 (Lagergren et  al, 1999). Tall stature, possibly
1997; Kabat et al, 1993; Wu et al, 2001; Lindblad reflecting better socioeconomic conditions dur-
et al, 2005) and in the modest number of cohort ing childhood and adolescence, appears to be
studies reporting on this disease (Freedman et al, associated with a low risk. Detailed analyses of
2007). The BEACON consortium of case-​control body fat distribution have been limited, but cur-
and cohort studies produced a summary estimate rent reports suggest that different weight gain
for esophageal adenocarcinoma of 1.96 (95% CI patterns may confer differences in risk not cap-
1.64–​2.34) using over 1,500 cases, strongly sug- tured by BMI for esophageal adenocarcinoma
gesting that the risks conferred for esophageal (O’Doherty et  al, 2012)  and Barrett’s esophagus
adenocarcinoma are lower than those reported (Kubo et al, 2013).
for esophageal squamous cell carcinoma (Cook
et al, 2010). Yet in the United States, PAR pertain- Diet
ing to esophageal adenocarcinoma was estimated Esophageal Squamous Cell Carcinoma
at 39.7% (Engel et al, 2003). A review paper with Dietary factors have been addressed in numerous
a combined evaluation of the effect of smoking etiologic studies of esophageal cancer, but recent
cessation did not unveil a clear reduction of risk comparisons between prospective cohort stud-
(Bosetti et al, 2006). ies and case-​control studies suggest that results
It also appears that alcohol use is not a risk from studies collecting diet data at the time of
factor for these tumors. Wine has been reported diagnosis should be interpreted with caution
to be associated with lower risk of esophageal (World Cancer Research Fund, 2007). Recall bias
adenocarcinoma or Barrett’s esophagus in sev- may have led to overestimation of the beneficial
eral studies (Anderson et al, 2009; Pandeya et al, effects of apparently healthier diets. Substantial
2009; Kubo et al, 2009). But the BEACON con- differences in diet limit generalizability of stud-
sortium reported no excess risk for esophageal ies from one population to another, and few pro-
adenocarcinoma even in subjects that consumed spective studies in low-​incidence populations are
≥7 drinks per day and no differences by bever- of sufficient size to study esophageal squamous
age type in the combined analysis (Freedman cell carcinoma.
et al, 2011). Dietary pattern analyses show some common
themes. High-​risk populations in economically
Anthropometric Measures underdeveloped areas typically exhibit lack of
There are some reports of an inverse relation variation in the diet, which is mainly vegetarian,
between relative weight, body mass index, (BMI), with a relatively small contribution from animal
and the risk of esophageal squamous cell carci- products (Hormozdiari et  al, 1975; Thurnham
noma (Fig. 9-​5). Since weight loss is an early et  al, 1985; Jaskiewicz et  al, 1988). Corn (van
consequence of the disease, the possibility of Rensburg, 1981), wheat, or maize (Thurnham
reversed causation or confounding by socioeco- et  al, 1985), is the chief staple, and cereals and
nomic status cannot be ruled out. Other studies bread dominate the diet. Recent studies in high-​
found elevated BMI well before disease onset to income populations also suggest that dietary
be unrelated to the risk of squamous cell carci- pattern analysis may be useful for cancer risk
noma (Chow et al, 1998; Lagergren et al 1999). studies.
Esophageal adenocarcinomas, on the other
hand, are positively, independently, and strongly Vitamins and Micronutrients
associated with higher BMI in both case-​control The corn-​or wheat-​based diet has been esti-
(Chow et  al, 1998; Lagergren et  al 1999)  and mated to be marginal or deficient in vitamins
cohort studies (MacInnis et al, 2006), including A and C, riboflavin, nicotinic acid, magnesium,
several with separate estimates for esophageal and zinc (Hormozdiari et al, 1975; van Rensburg,
adenocarcinoma without contamination with 1981). Deficiency of vitamin A  (retinol), essen-
gastric cardia adenocarcinoma, which is also tial for growth, development, and differentiation
associated with high BMI. It appears that the of all tissues, has been associated with cancers
risk gradient goes through the whole spectrum of epithelial origin, and riboflavin (vitamin B2)
of relative weights, from the leanest to the most is essential for maintaining the integrity of the
obese (Abnet et  al, 2008). In a Swedish study, skin and the squamous epithelium of the esoph-
persons with a BMI above 30 had a risk 16 times agus. Zinc, which protects against free-​radical
195

Esophageal Cancer 195

formation and peroxidation, is also critical for lower selenium intake in the past than did con-
the maintenance of normal bioactive retinol lev- trols (Cai et al, 2006).
els in blood. Zinc deficiency further increases the The nutritional hypotheses led to the launch-
CYP2E1-​dependent microsomal metabolism of ing of three randomized, placebo-​ controlled
nitrosamines and enhances the carcinogenicity intervention trials among high-​risk subjects in
of methylbenzylnitrosamine in animal models China. An intervention trial in Linxian, China,
(Barch, 1989). used a factorial design to compare four combina-
Deficiencies in vitamin or microelement lev- tions of vitamin/​mineral supplements in 29,584
els have, indeed, been documented in high-​risk subjects aged 40–​69 (Blot et  al, 1993). In the
countries like China (Thurnham et al, 1985;Yang group receiving a combination of beta-​carotene,
et  al, 1984)  and South Africa (Jaskiewicz et  al, vitamin E, and selenium daily for about 5 years,
1988). The most striking and consistent finding subjects experienced significantly lower total
in China and Uzbekistan is the high (up to 97%) mortality and overall cancer mortality, lower gas-
prevalence of riboflavin deficiency. This prev- tric cancer mortality, and a nonsignificant 4%
alence seems to be lower (about 40%) in South decline in esophageal cancer mortality. A bord-
Africa. Vitamin A  deficiencies were common erline reduction in esophageal cancer incidence
(20% to 40%) in some but not all high-​risk areas was found in the group receiving riboflavin and
in China, while the vitamin A  levels were gen- niacin. Two postintervention follow-​up analyses
erally acceptable in South Africa. Zinc deficien- revealed that effects reported at the end of the
cies were common (about 25%) in a few areas in intervention were durable, with a notable effect
both China and South Africa but less common on esophageal cancer mortality. There was a sig-
in others. Data regarding carotenoids, vitamin nificant quantitative interaction by age such that
C, and vitamin E were patchy and variable. Low those under age 55 at baseline assigned to receive
levels were noted in some but not all high-​risk daily beta-​ carotene, vitamin E, and selenium
areas. Other deficiencies reported in single stud- daily showed a significant 17% (2%–​29%) reduc-
ies include vitamin B12, folic acid, magnesium, tion in esophageal squamous cell carcinoma
calcium, and selenium. Noteworthy is the nic- mortality, while those aged 55 or over had signif-
otinic acid deficiencies observed in about three icantly increased mortality 14% (0%–​30%) (Qiao
out of four subjects in high-​risk Transkei, South et al, 2009). This interaction resulted in a net null
Africa. In an ecological study from China (Chen effect and was interpreted as potential benefit in
et al, 1992), dietary intake of a long list of trace those at the earlier stages of carcinogenesis and
elements was associated with esophageal can- an adverse effect analogous to “tumor-​feeding,”
cer mortality, and zinc, in particular, emerged as similar to the hypothesis for the provision of
inversely related to mortality risk. Further, cop- folate to subjects at high risk for developing colo-
per intake was inversely and calcium intake posi- rectal cancer.
tively related to this risk. The second intervention trial compared daily
Data on individual intake, derived from case-​ supplementation consisting of 14 vitamins and
control studies, indicate that a low intake of vita- 12 minerals with placebo among 3,318 persons
mins A, C, and B2 is associated with an increased with cytologic diagnosis of esophageal dyspla-
risk in both high-​and low-​risk populations. On sia (Li et  al, 1993). No difference was found in
the other hand, foods containing retinol were the cumulative incidence of esophageal can-
positively associated with a risk in several stud- cer between the two treatment groups. A small,
ies from Europe and the United States. Thus, the nonsignificant reduction in the occurrence of
combined evidence from observational studies esophageal dysplasia (Dawsey et al, 1994) and a
concerned with vitamin and mineral intake is modest reduction in squamous cell proliferation
fairly strong with regard to the protective effect (Rao et al, 1994) were noted, though. Although it
of vitamins B2 and C, but less convincing for vita- is impossible to tell what would have happened
min A  and most trace elements. Folate intake if other doses had been used, or if the supple-
was consistently inversely associated with risk for ments had been given earlier, these intervention
both squamous cell carcinoma and adenocarci- trials did not provide the critical evidence that
noma of the esophagus in several studies in west- would definitely establish the vitamin/​mineral
ern populations (Mayne et  al, 2001; Chen et  al, deficiency hypothesis in esophageal carcinogen-
2002). Moreover, Chinese cases with esophageal esis. Long-​term follow-​up of this study showed
squamous cell carcinoma reported a significantly no late effects, and reported that even in this
196

196 Textbook of Cancer Epidemiology

nutrient-​deficient population, there is no benefit fruits and vegetables, compared to cohort studies
to 6 years of supplementation with multivitamins (World Cancer Research Fund, 2007).
for esophageal cancer, all cancers, or total mor- Three cohort studies showed no benefit for
tality (Wang et al, 2013). fruit intake in relation to esophageal adenocar-
A third intervention study was conducted cinoma (Steevens et  al, 2011; Freedman et  al,
among subjects in Linxian with mild or moderate 2007; Gonzales et al, 2006), which suggests that
dysplasia (Limburg et al, 2005); in a randomized the previously declared benefit of fruit intake for
controlled trial with factorial design, participants this disease may have been due to overreliance
were given selenomethionine 200 μg daily and/​ on data from case-​control studies, which are par-
or celecoxib 200 mg twice daily for 10  months. ticularly susceptible to information bias. In con-
Endoscopies were performed before and after the trast, three published cohort studies (Steevens
treatment period. Although selenomethionine et  al, 2011; Freedman et  al, 2007; Yamaji et  al,
treatment resulted in a trend toward increased 2008)  suggest a beneficial effect for esophageal
dysplasia regression (43% versus 32% among squamous cell carcinoma. For each cell type, the
subjects who did not receive this treatment), and total amount of information available is limited,
a trend toward decreased progression (14% ver- but contradictory results may also reflect etio-
sus 19%), the overall effect was nonsignificant. logic heterogeneity between these cancers.
However, the effect was greater (and statistically Other exceptions exist among studies from
significant) among subjects with mild dysplasia high-​risk countries in the Asian esophageal can-
at baseline, compared to those with moderate cer belt; Li et al (1989) reported an increased risk
dysplasia. Since the latter stratification was an with consumption of fresh vegetables and corn
unplanned ad hoc analysis, this result has to be in a large case-​control study in Linxian, China,
interpreted with caution. and Wang et al (1992) associated intake of boiled
vegetables with a considerable excess risk. Others
Fruits and Vegetables were unable to confirm any positive or negative
Previously, the most consistent finding in the association between fruit/​vegetable intake and
numerous case-​control studies of dietary fac- esophageal cancer risk (Wu et al, 2006;Guo et al,
tors is a strong inverse relationship between 1994;Takezaki et al, 2001) or found only a weak
fruit and/​or vegetable intake and the risk of inverse association (Yu et  al, 1993). High con-
esophageal cancer. Like most older literature, sumption of the staple corn may reflect low soci-
this is limited by lack of specificity in associa- oeconomic status and lack of dietary variation.
tions by cell type, but much of early literature
likely included mainly esophageal squamous Other Food Groups
cell carcinoma. Most of more than 50 cohort The overall evidence regarding the association
and case-​control studies have corroborated this with food groups others than fruits and veg-
inverse association, and the more recent addi- etables is even more inconsistent and thus less
tions constitute no exceptions; the overwhelm- persuasive. One reason might be that dietary
ing majority have demonstrated a protective habits are closely linked to ethnic, cultural, and
effect of fruits and/​or vegetables, or of their socioeconomic factors that are difficult to meas-
microconstituents (Mayne et  al, 2001; Chen ure. They may also covary with the distinct geo-
et al, 2002; Kjaerheim et al, 1998; Launoy et al, graphic patterns of these diseases. Hence, care is
1998; Castellsague et al, 2000; Franceschi et al, needed to avoid the ecologic fallacy. If there is a
2000; Takezaki et  al, 2000; Brown et  al, 2001; correlation between mean exposure level and the
Terry et al, 2001; De Stefani et al, 2003; Boeing background rate of disease across studied groups,
et al, 2006). confounding by group may be unavoidable in
This literature has been reviewed repeatedly estimating the effects. While meat and fish were
(World Cancer Research Fund, 2007; Cheng & seemingly protective in some studies (Launoy
Day, 1996; Steinmetz & Potter, 1996). The most et al, 1998)—​but not invariably so (Bosetti et al,
recent World Cancer Research Fund report sug- 2000)—​and red meat was associated with a mod-
gested that there was some benefit for both cell erate risk increase in others (Castellsague et  al,
types. The magnitude is, however, modest, and 2000; De Stefani et al, 2003; Bosetti et al, 2000),
careful review of the literature also suggests that several studies showed no relationship with meat
case-​control studies may have overestimated intake (Cheng & Day, 1996; Tavani et al, 2000).
the benefits of apparently healthy foods, such as A summary of only cohort studies would suggest
197

Esophageal Cancer 197

a potential modest increase in risk for higher red (De Jong et  al, 1972), and water instillation in
meat esophageal adenocarcinoma and a likely rodent esophageal cancer models suggests that
increase for esophageal squamous cell carcinoma, this exposure can potentiate the risk conferred by
but again the data is limited (Cross et al, 2011). chemical carcinogens (Yioris et  al, 1984). Thus,
hot temperature may exert its effects in combi-
Pickled Vegetables nation with other exposures (Islami et al, 2009).
A comprehensive meta-​analysis of pickled veg- Numerous studies from Uruguay and the
etables and gastric cancer risk concluded that surrounding regions of South America have
there was consistent evidence for an association documented an apparent association between
with gastric cancer, which highlights the carcino- consumption of the herbal infusion maté and
genic potential of this exposure (Ren et al, 2012). esophageal cancer risk. Although not docu-
A similar analysis for esophageal squamous cell mented in every study, these South American
carcinoma in Asians also concluded that pickled studies likely include primarily esophageal squa-
vegetables were a strong risk factor for esopha- mous cell carcinoma cases. Duration, amount,
geal squamous cell carcinoma (Islami et al, 2009). and typical temperature of consumption of maté
Furthermore, intake of moldy food and pickled have been linked to increased risk in several stud-
vegetable juice was positively linked to esopha- ies (Castellsague et al, 2000; Vassallo et al, 1985;
geal cancer. However, several other studies in Sewram et al, 2003).
China, among them a cohort study, were unable Tea is the primary beverage in both the Asian
to verify this positive association (Yu et al, 1993). and African esophageal cancer regions. A  case-​
Strong public health messages in this population control study from Golestan Province, Iran, a
inculcated a knowledge of the potential health region with notably elevated esophageal squa-
hazards of local pickles, which differ substantially mous cell carcinoma rates, has shown substantial
from vinegar-​based Western pickles, and may increase in esophageal squamous cell carcinoma
have led to substantial underreporting. A system- risk among drinkers of exceedingly hot black tea.
atic review (Jakszyn & Gonzales, 2006) attempted These studies used a questionnaire regarding tea
to summarize the available literature with regard temperature that was simultaneously validated
to the relationship between estimated nitros- in healthy subjects. Subjects in the highest cat-
amine exposure and risk of esophageal can- egory of tea consumption (denominated “very
cer. Exposure to preformed nitrosamines was hot” in the questionnaire and equivalent to
estimated from external sources by examining >70°C) had increased risk of 8.16 (3.93–​16.91)
intake of processed meat, beer, pickled and dried compared to subjects that drank “warm” tea
vegetables, smoked fish or meat, and salted or (<65°C) and those in the intermediate category
dried fish or meat. The quantification of endog- had risk of 2.07 (1.62–​3.83) (Islami et al, 2009).
enous nitrosamine formation was based on the This is the most compelling evidence to date that
intake of heme-​containing red meat. Although high-​temperature beverages can increase risk
the authors concluded that there is insufficient of esophageal squamous cell carcinoma, but the
evidence for a role of nitrosamines in esophageal case-​control design makes it difficult to rule out
carcinogenesis, they noted that a majority of the that the presence of developing esophageal can-
studies showed point estimates that supported cers may have affected temperature assessments
the nitrosamine hypothesis. This was particularly by the cases. Studies of hot beverages in Western
true for processed meat, but less for preserved populations have generally been null (Ren et al,
fish, preserved vegetables, assessed nitrite expo- 2010), but the consumption of such beverages
sure, and measured nitrosamine exposure. in these populations tends to be modest and the
temperatures are less extreme.
Hot Drinks For all hot beverages, the chemical composi-
Thermal injury from unusually hot beverage and tion of the product may provide either beneficial
food consumption has been hypothesized to con- or additional adverse exposures. It is difficult to
vey risk in several populations with high rates disentangle the independent effects of the bever-
of esophageal squamous cell carcinoma. Formal age from the effects of the temperature at which it
studies of this behavior have been hampered by is drunk. Green tea has been shown to have anti-
the difficulty in capturing quantitative data on carcinogenic effects in animal models and one
temperature of consumption. Hot beverages do study from Shanghai yielded supportive evidence
raise the temperature of the esophagus notably in humans (Gao et  al, 1994). However, a study
198

198 Textbook of Cancer Epidemiology

from Japan (Inoue et al, 1998) found no associ- in South African black patients with esophageal
ation, while a third study reported a statistically cancer have led to the hypothesis that iron over-
significant positive dose-​risk trend of increasing load is a potential causal factor for these cancers
esophageal squamous cell carcinoma risk with (Isaacson et al, 1985). The iron is thought to ema-
increasing number of cups per day (Ishikawa nate from alcoholic drinks that are home-​brewed
et al, 2006). Maté has been shown to contain sub- in cast iron containers.
stantial amounts of polycyclic aromatic hydro-
carbons (Kamangar et  al, 2008), which could Reproductive Factors
convey risk independent of the temperature at and Hormones
which it is consumed. In fact, at least some stud- Neither esophageal squamous cell carcinoma
ies suggest that maté drunk cold is a risk factor nor adenocarcinoma is perceived as a hormone-​
for esophageal squamous cell carcinoma. dependent cancer. There are limited data on hor-
Heterocyclic amines are carcinogenic sub- mones and squamous cell carcinoma. Limited
stances formed through pyrolysis of amino acids results have suggested a benefit to a large number
and creatine or creatinine when meats are cooked of reproductive years or breastfeeding (Yu et al,
at high temperature, particularly by pan-​frying. 2011), use of hormone replacement therapy with
Heterocyclic amine levels increase with cook- estrogen and progestin (Bodelon et al, 2011), or
ing temperature, with the type and shape of the overall for squamous cancers of the upper diges-
cooked piece of meat, with the degree of brown- tive tract (Freedman et al, 2010). But a study in
ing on the surface, and with the cooking method. Sweden showed that some apparent associations
Several investigators have tried to shed light on a for esophageal squamous cell carcinoma may be
possible relationship between heterocyclic amine due to correlated nonhormonal factors, since
levels and risk of esophageal cancer, with cooking similar results were obtained in men and women
method, doneness preferences, or pictures of sur- (Lu et al, 2012).
face browning as markers of exposure. Frequent The strong and unexplained male predom-
frying of food (Galeone et  al, 2005)  or broiling inance among esophageal adenocarcinoma
of meat (De Stefani et  al, 1998)  was associated patients has attracted some attention. It was
with risk of esophageal cancer or upper aerodi- hypothesized that high estrogen levels might
gestive tract cancer (of which approximately half be protective. However, a study among more
were esophageal) in two European studies, while than 100,000 men receiving estrogen treatment
cooking method was unrelated to esophageal for prostate cancer found no evidence of any
cancer in a third study from the United States protective effect (Lagergren et  al, 1998). The
(Ward et al, 1997). A Swedish case-​control study hypothesis postulating protection by female
with a more detailed quantification of heterocy- sex hormones has been tested further among
clic amine levels showed a statistically significant women receiving hormone replacement ther-
2.4-​fold excess risk for esophageal squamous cell apy (Lindblad et al, 2006), women with breast
carcinoma, but not for esophageal adenocarci- cancer being treated with the selective estrogen
noma, among individuals with the highest intake receptor modulator Tamoxifen (Chandanos
of all three major heterocyclic amines, relative to et  al, 2006)  and among women with varying
those with the lowest intake (Terry et al, 2003). parity (Lagergren et al, 2005) without yielding
For the sake of completeness, a few other any evidence in support of this theory. Given the
hypotheses about food-​related reasons for the already very low baseline incidence of esopha-
excess of esophageal cancer in the high-​incidence geal adenocarcinoma among women, protec-
areas should be mentioned. Contamination of tive effects by any intervention are difficult to
mycotoxins in the crop is one possibility and confirm. Breast-​feeding was associated with a
has, indeed, been documented in South Africa. 60% reduction in esophageal adenocarcinoma
However, its etiologic relevance has not been risk among women in a study from Scotland,
addressed in any analytical epidemiologic study. though (Cheng et  al, 2000). Alternatively, the
Another possible mechanism is an action of sil- higher incidence of esophageal adenocarci-
ica fragments that come from the grinding of noma in men could be linked to higher levels
millet bran. Silica fragments have been observed of androgens, and results from one study have
in the mucosa surrounding esophageal tumors suggested that higher levels of androgens may
of patients in northern China (O’Neill et  al, be associated with increased risk of Barrett’s
1982). Finally, high hepatic iron concentrations esophagus (Cook et al, 2015).
19

Esophageal Cancer 199

Infections gastric atrophy). Gastric atrophy is one late con-

Human Papilloma Virus sequence of H.  pylori infection that can lead to
Infection with human papilloma virus (HPV), disappearance of the causative microorganisms.
especially HPV type 16, has been implicated as A  similar analysis in China, however, did not
a risk factor for squamous cell carcinoma of the reproduce these results (Kamangar et  al, 2007).
esophagus, but the overall picture appears to sug- A  meta-​analysis reported a possible increased
gest that this is not involved in the majority of risk of esophageal squamous cell carcinoma in
esophageal squamous cell carcinoma cases. There subjects with atrophy (Islami et  al, 2011)  with
are reports of HPV DNA detected in squamous subsequent reports of associations in Finland
cell carcinoma, although other studies are nega- (Cook et  al, 2010), Iran (Nasrollahzadeh et  al,
tive. Among six seroepidemiologic studies, three 2012), and Brazil (Almodova et al, 2013). Other
of which were nested within prospective cohorts, markers of certain types of gastric atrophy such
three found positive associations but with wide as low serum ghrelin (Murphy et  al, 2012)  or
confidence intervals (Han et al, 1996; Bjorge et al, prior diagnosis of pernicious anemia are asso-
1997), whereas two retrospective studies, includ- ciated with higher risk of esophageal squamous
ing the largest published to date (Lagergren et al, cell carcinoma (Murphy et  al 2015). Clearly,
1999; Van Doornum et al, 2003) and a prospec- as H.  pylori does not colonize the esophagus,
tive study in a Chinese high-​risk area with precise and gastric atrophy is confined to the stomach,
odds ratio estimates for HPV-​16, HPV-​18, and esophageal carcinogenesis seems to be in some
HPV-​73 (Kamanger et  al, 2006), were negative. way dependent on events that take place in the
One study in the same region reported evidence stomach.
of HPV-​16 DNA in the tumor coinciding with An even stronger indication of the interplay
HPV-​16 E7 seropositivity, that is, nonconcordant between the intragastric environment and carcin-
infections (Guo et al, 2012). The reasons for the ogenesis in the esophagus is the inverse relation-
inconsistent results remain conjectural. A  large ship between H. pylori seropositivity and the risk
international consortium with over 1,500 cases of adenocarcinoma of the esophagus and gastric
from six geographic regions showed through cardia that has been confirmed in several studies
serology (Kamangar et al, 2007) and tissue analy- conducted in Western populations (Ye et al, 2004;
ses (Islami et  al, 2011)  little evidence that HPV Chow et al, 1998; Hansen et al, 1999; Siman et al,
could be a common cause of esophageal squa- 2001; de Martel et al, 2005; Kamangar et al, 2006).
mous cell carcinoma. The assumed mechanism involves gastric atrophy,
The geographic distribution, male predom- hypoacidity, and reduced acidic reflux into the
inance, racial differences in risk, and secular esophagus. However, gastric atrophy, measured
trends in esophageal adenocarcinoma resemble through the pepsinogen 1 biomarker, was unre-
the recent rise to dominance of HPV as a risk fac- lated to the risk of esophageal adenocarcinoma
tor for oropharyngeal cancer in Western, white, in Sweden (Ye et al, 2004) and a large prospective
males. A  modest number of studies have tested study in China (Ren et al, 2009). Overall, the role
for and found some evidence for an association of gastric atrophy in esophageal carcinogenesis—​
between esophageal adenocarcinoma and HPV both squamous cell carcinoma and adenocarci-
(Rajendra et al, 2013), while others have showed noma—​seems unclear, yet moderately high ORs
no evidence for a link to Barrett’s esophagus (El-​ and the ability to measure this state serologically
Serag et al, 2013). No strong conclusions can be suggest that more research is warranted.
drawn with regard to this hypothesis.
Occupation
Helicobacter Pylori Some occupational groups seem to be at
Two studies from Nordic countries have increased risk for squamous cell carcinoma of
addressed the possible link between Helicobacter the esophagus, but recently there has been little
pylori seropositivity and risk of esophageal scholarship on this question. There is substan-
squamous cell carcinoma; one found a twofold tial evidence that persons who work with vul-
increased risk among subjects seropositive for canization in the rubber or automobile industry
H. pylori CagA (Ye et al, 2004), whereas the other run an increased risk of this cancer (Norell et al,
(Cook et al, 2010) could not confirm this serolog- 1983). Other occupations associated with an
ical association (but found a fourfold increased increased risk include chimney sweeps (Evanoff
risk among subjects with serological evidence of et al, 1993), mine workers (Vizcaino et al, 1995),
20

200 Textbook of Cancer Epidemiology

chemical products workers (Sathiakumar et  al, responses to it may be conducive to the develop-
1992), butchers (Besson et al, 2006), and medical ment of esophageal cancer.
x-​ray workers. Significant positive associations
have also been identified among workers in the Medical Conditions
cement industry (Jakobsson et al, 1990; Jansson Some medical conditions confer a higher risk
et al, 2006), the plastics and composites industry of esophageal cancer. The classic example is
(Wong et  al, 1994), the dye production indus- Plummer-​ Vinson or Paterson-​ Kelly syndrome
try (Bulbulyan et  al, 1995), and bookbindery (Ahlbom et  al, 1936), characterized by hypo-
(Bulbulyan et al, 1999). chromic anemia and changes in the hypophar-
Specific exposures tentatively linked to ynx. Formerly common in northern Europe, this
esophageal cancer include metal dust (Yu et  al, syndrome is rarely seen nowadays. It is thought
1988), asbestos (Kang et  al, 1997), silica dust to be caused by nutritional deficiencies and was
(Pan et  al, 1999;Wernli et  al, 2006), combus- often accompanied by esophageal squamous cell
tion products (Gustavsson et  al, 1993), metals carcinomas. Celiac disease has also been claimed
(Wernli et al, 2006), organic solvents (Lynge et al, to be associated with an increased risk of esoph-
1997;Hansen et  al, 2001)—​particularly perchlo- ageal cancer, although properly designed epide-
rethylene in dry-​cleaning industries (Vaughan miologic studies are lacking. Hemochromatosis
et  al, 1997)—​and polycyclic aromatic hydrocar- was followed by an almost 50-​fold excess risk of
bon (PAH) (Gustavsson et al, 1998). Cumulative esophageal cancer in one cohort study (Hsing
exposure to endotoxin, a contaminant of cotton et al, 1995) with only two observed cases. A three-
dust, was found to be inversely related to esopha- fold increased risk of esophageal squamous cell
geal cancer risk among female textile workers in carcinoma, but no significant risk elevation for
China (Wernli et  al, 2006). A  problem in most adenocarcinoma, was noted among patients who
of these studies was the inability to adjust for had been hospitalized with pernicious anemia
possible confounders, particularly smoking and (Ye et al, 2003; Murphy et al, 2015).
alcohol use. Achalasia, a paralysis of the lower esophagus,
The strong gender imbalance among ade- has been linked to elevated risk for both esoph-
nocarcinoma cases indicates that occupational ageal adenocarcinoma and squamous cell car-
exposures also may be important for this cancer cinoma, which may be due to irritation caused
type, but hard data are scarce. One case-​control by prolonged contact between swallowed mate-
study reported a fourfold increased risk among rial and the esophageal mucosa (Zendehdel et al,
white women with administrative jobs, while a 2011). Other studies also reported a substantial
twofold increase was found among female health relative risk when examining esophageal squa-
professionals. Another study implicated employ- mous cell carcinoma, but the low baseline risk
ments in administrative support, financial, insur- in most Western populations suggests that the
ance, and real estate and health services (Engel clinical relevance of achalasia in an individual
et al, 2002). A third study found pesticide expo- may not necessitate screening (Leeuwenburgh
sure to be borderline significantly associated with et al, 2010).
risk of esophageal adenocarcinoma (Jansson et al,
2006). More studies will be needed, but the rar- Gastroesophageal Reflux
ity of esophageal adenocarcinoma constitutes an Gastroesophageal reflux disease (GERD) is the
essentially insurmountable hurdle in the assess- strongest known risk factor for adenocarcinoma
ment of occupation-​specific risks. of the esophagus (Fitzgerald, 2005). In a large
In a Swedish population-​based case-​control study, the risk gradient between those without
study, self-​reported job strain, work place sat- any symptoms of reflux and those with the most
isfaction, and coping were investigated in rela- severe and long-​standing symptoms was more
tion to risk of esophageal cancer by histology than 40-​fold (Lagergren et al, 1999). The associ-
(Jansson et al, 2004). Subjects reporting low work ation seems to be independent of other known
place satisfaction had an almost fourfold elevated or suspected risk factors for esophageal adeno-
risk of esophageal squamous cell carcinoma and carcinoma, including BMI. The presence of a hia-
a nearly threefold increased risk of esophageal tal hernia may interact synergistically with the
adenocarcinoma. Although work-​related stress reflux condition to inflate the relative risks more
did not seem to be important, the interaction of than multiplicatively (Wu et al, 2003). Although
a stressful work environment and the individual’s the strength of the association varies among
201

TABLE 9-​1   EPIDEMIOLOGY OF ESOPHAGEAL SQUAMOUS CELL

CARCINOMA: RISK FACTOR SUMMARY

Well-​confirmed risk factors Direction of effect*

Tobacco ↑↑
Alcoholic beverage consumption ↑↑
Achalasia ↑↑
Poverty ↑
Betel quid with or without tobacco ↑
Probable Relationship Exists, Based on Substantial Data
Hot beverages ↑↑
Pickled vegetables ↑
Fresh fruits and vegetables ↓
Selenium deficiency ↑
Poor oral health ↑
Gastric atrophy ↑
Weak, if Any, Relationship Exists, Based on Substantial Data
Human papilloma virus —​
Inconsistent Findings or Limited Study to Date
Mycotoxins ↑
Low zinc status ↑
Occupation ↑
Helicobacter pylori ↑
Polycyclic aromatic hydrocarbons ↑
Gastroesophageal reflux disease ↑↑
Higher BMI ↑↑
Male sex ↑↑
European ancestry ↑↑
Tobacco smoking ↑
Probable Relationship Exists, Based on Substantial Data
Hiatal hernia ↑
Helicobacter pylori ↓
Gastric atrophy ↓
Red meat ↑
Aspirin ↓
Weak, if Any, Relationship Exists, Based on Substantial Data
Fresh fruits and vegetables ↓
Alcoholic beverage consumption —​
Inconsistent Findings or Limited Study to Date
Sex steroid hormones —​

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate to large
increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association
Note: The magnitude of the risk can vary substantially, depending on the exact comparison being made.
20

202 Textbook of Cancer Epidemiology

studies, the results are consistent and biologically epidemiologists. Such a rapid change is bound to
plausible. Gastroesophageal reflux is also con- depend on changes in the exposure to important
sistently linked to Barrett’s esophagus (Taylor & causal factors. The well-​documented increase in
Rubenstein, 2010). the prevalence of obesity in many populations
might be one such factor. However, given some
Poor Oral Health conflicting observations, for instance the much
A modest number of studies have examined the lower risk among women despite the equally
association between poor oral health, poor oral high prevalence of obesity, there is clearly room
hygiene, or tooth loss and risk of esophageal for alternative hypotheses. Furthermore, current
squamous cell carcinoma and shown an asso- risk factors cannot account for the racial differ-
ciation in most (Abnet et  al, 2008; Abnet et  al, ences in esophageal adenocarcinoma risk as is
2005; Abnet et al, 2001; Dar et al, 2013; Hiraki clearly evident in US cancer registries.
et al, 2008) but not all studies (Abnet et al, 2005). Overall, esophageal cancer remains a major
Most of these studies have demonstrated inde- public health problem worldwide and unfor-
pendence of this association vis-​à-​vis obvious tunately remains a cancer with only modest
confounders such as tobacco smoking, alco- attention from researchers. Given the dramatic
holic beverage consumption, or low socioeco- differences in esophageal cancer incidence within
nomic status, but few mechanistic studies have and between countries and the huge differences
been completed. One study reported an associ- in risk by simple demographic factors such as
ation between a measure of bacterial population sex and race, major strides in understanding
diversity (a principal component of a bacterial the etiology of esophageal cancer etiology seem
beta-​ diversity matrix—​ a pairwise measure of within reach.
difference in microbiota communities) in the
upper digestive tract and esophageal squamous REFERENCES
dysplasia (Yu et al, 2014), but this result is pre- Abedi-​Ardekani B, Hainaut P. Cancers of the upper
liminary and requires further investigation. gastro-​intestinal tract:  a review of somatic
Limited data has also linked the microbiome of mutation distributions. Arch Iranian Med
the distal esophagus to intestinal metaplasia and 2014;17:286–​92.
potentially esophageal adenocarcinoma (Yang Abedi-​Ardekani B, Kamangar F, Sotoudeh M, Villar
et al, 2009). S, Islami F, Aghcheli K et al. Extremely high Tp53
mutation load in esophageal squamous cell car-
CONCLUSION cinoma in Golestan Province, Iran. PLoS One
Esophageal cancer remains a challenge to health- 2011;6:e29488.
care and public health authorities. Esophageal Abnet CC, Kamangar F, Dawsey SM, Stolzenberg-​
cancer is classified as rare in high-​income coun- Solomon RZ, Albanes D, Pietinen P et  al. Tooth
tries, but remains a common cancer in many loss is associated with increased risk of gastric
low-​and middle-​income countries, where the non-​cardia adenocarcinoma in a cohort of Finnish
prospect for appropriate care is most challenging. smokers. Scand J Gastroenterol 2005;40:681–​7.
The remarkably uneven geographic distribution, Abnet CC, Freedman ND, Hollenbeck AR, Fraumeni
and the apparently small proportion that can be JF Jr, Leitzmann M, Schatzkin A. A prospective
study of BMI and risk of oesophageal and gastric
attributed to genetic factors, seem to imply that
adenocarcinoma. Eur J Cancer 2008;44:465–​71.
strong external risk factors—​possibly reinforced
Abnet CC, Freedman ND, Hu N, Wang Z, Yu K, Shu
by genetically determined susceptibility—​are in
XO et al. A shared susceptibility locus in PLCE1 at
operation (Table 9-​1). This should raise hope
10q23 for gastric adenocarcinoma and esophageal
that preventive measures will become a realistic squamous cell carcinoma. Nat Genet 2010;42:764–​7.
option in the foreseeable future. If the recently Abnet CC, Kamangar F, Islami F, Nasrollahzadeh D,
reported potential utility of endoscopic screen- Brennan P, Aghcheli K et al. Tooth loss and lack of
ing and early intervention in populations at very regular oral hygiene are associated with higher risk
high risk of esophageal squamous cell carcinoma of esophageal squamous cell carcinoma. Cancer
(Wei et  al, 2015)  is further confirmed, cancer Epidemiol Biomarkers Prev 2008;17:3062–​8.
screening may become a potentially useful inter- Abnet CC, Qiao YL, Dawsey SM, Dong ZW, Taylor PR,
vention in selected populations. Mark SD. Tooth loss is associated with increased
The dramatic rise in incidence of the ade- risk of total death and death from upper gastro-
nocarcinomas poses a special challenge to intestinal cancer, heart disease, and stroke in a
203

Esophageal Cancer 203

Chinese population-​based cohort. Int J Epidemiol in Barrett’s esophagus patients:  results from a
2005;34:467–​74. large population-​based study. J Natl Cancer Inst
Abnet CC, Qiao YL, Mark SD, Dong ZW, Taylor PR, 2011;103:1049–​57.
Dawsey SM. Prospective study of tooth loss and Biramijamal F, Allameh A, Mirbod P, Groene HJ,
incident esophageal and gastric cancers in China. Koomagi R, Hollstein M. Unusual profile and high
Cancer Causes Control 2001;12:847–​54. prevalence of p53 mutations in esophageal squa-
Abnet CC, Wang Z, Song X, Hu N, Zhou FY, Freedman mous cell carcinomas from northern Iran. Cancer
ND et al. Genotypic variants at 2q33 and risk of Res 2001;61:3119–​23.
esophageal squamous cell carcinoma in China: a Bjorge T, Hakulinen T, Engeland A, Jellum E, Koskela
meta-​analysis of genome-​wide association stud- P, Lehtinen M et al. A prospective, seroepidemio-
ies. Hum Mol Genet 2012;21:2132–​41. logical study of the role of human papillomavi-
Abrams JA, Sharaiha RZ, Gonsalves L, Lightdale CJ, rus in esophageal cancer in Norway. Cancer Res
Neugut AI. Dating the rise of esophageal ade- 1997;57:3989–​92.
nocarcinoma:  analysis of Connecticut Tumor Blaser MJ, Saito D. Trends in reported adenocarcino-
Registry data, 1940–​ 2007. Cancer Epidemiol mas of the oesophagus and gastric cardia in Japan.
Biomarkers Prev 2011;20:183–​6. Eur J Gastroenterol Hepatol 2002;14:107–​13.
Agrawal N, Jiao Y, Bettegowda C, Hutfless SM, Wang Blaydon DC, Etheridge SL, Risk JM, Hennies HC, Gay
Y, David S et al. Comparative genomic analysis of LJ, Carroll R et al. RHBDF2 mutations are associ-
esophageal adenocarcinoma and squamous cell ated with tylosis, a familial esophageal cancer syn-
carcinoma. Cancer Disc 2012;2:899–​905. drome. Am J Hum Gen 2012;90:340–​6.
Ahlbom HE. Simple achlorhydric anaemia, Plummer-​ Blot WJ, Devesa SS, Kneller RW, Fraumeni JF Jr.
Vinson syndrome, and carcinoma of the mouth, Rising incidence of adenocarcinoma of the esoph-
pharynx, and oesophagus in women:  obser- agus and gastric cardia. JAMA 1991;265:1287–​9.
vations at Radiumhemmet, Stockholm. BMJ Blot WJ, Li JY, Taylor PR, Guo W, Dawsey S, Wang
1936;2:331–​3. GQ et al. Nutrition intervention trials in Linxian,
Akhtar S. Areca nut chewing and esophageal China:  supplementation with specific vitamin/​
squamous-​cell carcinoma risk in Asians: a meta-​ mineral combinations, cancer incidence, and
analysis of case-​control studies. Cancer Causes disease-​specific mortality in the general popula-
Control 2013;24:257–​65. tion. J Natl Cancer Inst 1993;85:1483–​92.
Almodova Ede C, de Oliveira WK, Machado LF, Grejo Bodelon C, Anderson GL, Rossing MA, Chlebowski
JR, da Cunha TR, Colaiacovo W et  al. Atrophic RT, Ochs-​Balcom HM, Vaughan TL. Hormonal
gastritis:  risk factor for esophageal squamous factors and risks of esophageal squamous cell car-
cell carcinoma in a Latin-​American population. cinoma and adenocarcinoma in postmenopausal
World J Gastroenterol 2013;19:2060–​4. women. Cancer Prev Res 2011;4:840–​50.
Anderson LA, Cantwell MM, Watson RG, Johnston Boeing H, Dietrich T, Hoffmann K, Pischon T, Ferrari
BT, Murphy SJ, Ferguson HR et al. The association P, Lahmann PH et al. Intake of fruits and vegeta-
between alcohol and reflux esophagitis, Barrett’s bles and risk of cancer of the upper aero-​digestive
esophagus, and esophageal adenocarcinoma. tract: the prospective EPIC-​study. Cancer Causes
Gastroenterol 2009;136:799–​805. Control 2006;17:957–​69.
Bai Y, Edamatsu H, Maeda S, Saito H, Suzuki N, Satoh Bosetti C, Gallus S, Garavello W, La Vecchia C.
T et al. Crucial role of phospholipase Cepsilon in Smoking cessation and the risk of oesophageal
chemical carcinogen-​induced skin tumor devel- cancer:  an overview of published studies. Oral
opment. Cancer Res 2004;64:8808–​10. Oncol 2006;42:957–​64.
Barch DH. Esophageal cancer and microelements. J Bosetti C, La Vecchia C, Talamini R, Simonato L,
Am Col Nutr 1989;8:99–​107. Zambon P, Negri E et al. Food groups and risk of
Barrett MT, Galipeau PC, Sanchez CA, Emond MJ, squamous cell esophageal cancer in northern Italy.
Reid BJ. Determination of the frequency of loss Int J Cancer 2000;87:289–​94.
of heterozygosity in esophageal adenocarcinoma Botterweck AA, Schouten LJ, Volovics A, Dorant E,
by cell sorting, whole genome amplification van Den Brandt PA. Trends in incidence of ade-
and microsatellite polymorphisms. Oncogene nocarcinoma of the oesophagus and gastric car-
1996;12:1873–​8. dia in ten European countries. Int J Epidemiol
Besson H, Banks R, Boffetta P. Cancer mortality 2000;29:645–​54.
among butchers: a 24-​state death certificate study. Brooks PJ, Enoch MA, Goldman D, Li TK, Yokoyama
J Occup Environ Med 2006;48:289–​93. A. The alcohol flushing response: an unrecognized
Bhat S, Coleman HG, Yousef F, Johnston BT, McManus risk factor for esophageal cancer from alcohol
DT, Gavin AT et al. Risk of malignant progression consumption. PLoS Med 2009;6:e50.
204

204 Textbook of Cancer Epidemiology

Brown LM, Devesa SS. Epidemiologic trends in Chen H, Tucker KL, Graubard BI, Heineman EF,
esophageal and gastric cancer in the United States. Markin RS, Potischman NA et al. Nutrient intakes
Surg Oncol Clin North Am 2002;11:235–​56. and adenocarcinoma of the esophagus and distal
Brown LM, Hoover R, Silverman D, Baris D, Hayes stomach. Nutr Cancer 2002;42:33–​40.
R, Swanson GM et  al. Excess incidence of squa- Cheng KK, Day NE, Davies TW. Oesophageal can-
mous cell esophageal cancer among US Black cer mortality in Europe:  paradoxical time trend
men: role of social class and other risk factors. Am in relation to smoking and drinking. Br J Cancer
J Epidemiol 2001;153:114–​22. 1992;65:613–​7.
Brown LM, Hoover RN, Greenberg RS, Schoenberg Cheng KK, Day NE. Nutrition and esophageal cancer.
JB, Schwartz AG, Swanson GM et  al. Are racial Cancer Causes Control 1996;7:33–​40.
differences in squamous cell esophageal can- Cheng KK, Duffy SW, Day NE, Lam TH, Chung
cer explained by alcohol and tobacco use? J Natl SF, Badrinath P. Stopping drinking and risk of
Cancer Inst 1994;86:1340–​5. oesophageal cancer. BMJ 1995;310:1094–​7.
Brown LM, Silverman DT, Pottern LM, Schoenberg Cheng KK, Sharp L, McKinney PA, Logan RF, Chilvers
JB, Greenberg RS, Swanson GM et  al. CE, Cook-​Mozaffari P et al. A case-​control study
Adenocarcinoma of the esophagus and esopha- of oesophageal adenocarcinoma in women: a pre-
gogastric junction in white men in the United ventable disease. Br J Cancer 2000;83:127–​32.
States:  alcohol, tobacco, and socioeconomic fac- Chow WH, Blaser MJ, Blot WJ, Gammon MD,
tors. Cancer Causes Control 1994;5:333–​40. Vaughan TL, Risch HA et al. An inverse relation
Bulbulyan MA, Figgs LW, Zahm SH, Savitskaya T, between cagA+ strains of Helicobacter pylori
Goldfarb A, Astashevsky S et al. Cancer incidence infection and risk of esophageal and gastric car-
and mortality among beta-​ naphthylamine and dia adenocarcinoma. Cancer Res 1998;58:588–​90.
benzidine dye workers in Moscow. Int J Epidemiol Chow WH, Blot WJ, Vaughan TL, Risch HA, Gammon
1995;24:266–​75. MD, Stanford JL et al. Body mass index and risk
Bulbulyan MA, Ilychova SA, Zahm SH, Astashevsky of adenocarcinomas of the esophagus and gastric
SV, Zaridze DG. Cancer mortality among women cardia. J Natl Cancer Inst 1998;90:150–​5.
in the Russian printing industry. Am J Industr Cook MB, Chow WH, Devesa SS. Oesophageal can-
Med 1999;36:166–​71. cer incidence in the United States by race, sex,
Bye H, Prescott NJ, Lewis CM, Matejcic M, Moodley and histologic type, 1977–​ 2005. Br J Cancer
L, Robertson B et al. Distinct genetic association 2009;101:855–​9.
at the PLCE1 locus with oesophageal squamous Cook MB, Dawsey SM, Diaw L, Blaser MJ, Perez-​
cell carcinoma in the South African population. Perez GI, Abnet CC et  al. Serum pepsinogens
Carcinogenesis 2012;33:2155–​61. and Helicobacter pylori in relation to the risk
Cai L, You NC, Lu H, Mu LN, Lu QY, Yu SZ et  al. of esophageal squamous cell carcinoma in the
Dietary selenium intake, aldehyde dehydrogenase-​ alpha-​ tocopherol, beta-​
carotene cancer preven-
2 and X-​ray repair cross-​complementing 1 genetic tion study. Cancer Epidemiol Biomarkers Prev
polymorphisms, and the risk of esophageal squa- 2010;19:1966–​75.
mous cell carcinoma. Cancer 2006;106:2345–​54. Cook MB, Kamangar F, Whiteman DC, Freedman
Castellsague X, Munoz N, De Stefani E, Victora CG, ND, Gammon MD, Bernstein L et  al. Cigarette
Castelletto R, Rolon PA et  al. Independent and smoking and adenocarcinomas of the esophagus
joint effects of tobacco smoking and alcohol and esophagogastric junction:  a pooled analysis
drinking on the risk of esophageal cancer in men from the international BEACON consortium. J
and women. Int J Cancer 1999;82:657–​64. Natl Cancer Inst 2010;102:1344–​53.
Castellsague X, Munoz N, De Stefani E, Victora Cook MB, Wood SN, Cash BD, Young P, Acosta
CG, Castelletto R, Rolon PA. Influence of mate RD, Falk RT et  al. Association between circulat-
drinking, hot beverages and diet on esopha- ing levels of sex steroid hormones and Barrett’s
geal cancer risk in South America. Int J Cancer esophagus in men:  a case-​control analysis. Clin
2000;88:658–​64. Gastroenterol Hepatol 2015;13:673–​82.
Chandanos E, Lindblad M, Jia C, Rubio CA, Ye Cross AJ, Freedman ND, Ren J, Ward MH, Hollenbeck
W, Lagergren J. Tamoxifen exposure and risk AR, Schatzkin A et al. Meat consumption and risk
of oesophageal and gastric adenocarcinoma:  a of esophageal and gastric cancer in a large prospec-
population-​based cohort study of breast cancer tive study. Am J Gastroenterol. 2011;106:432–​42.
patients in Sweden. Br J Cancer 2006;95:118–​22. Dar NA, Bhat GA, Shah IA, Iqbal B, Makhdoomi MA,
Chen F, Cole P, Mi Z, Xing L. Dietary trace elements Nisar I et al. Hookah smoking, nass chewing, and
and esophageal cancer mortality in Shanxi, China. oesophageal squamous cell carcinoma in Kashmir,
Epidemiol 1992;3:402–​6. India. Br J Cancer 2012;107:1618–​23.
205

Esophageal Cancer 205

Dar NA, Islami F, Bhat GA, Shah IA, Makhdoomi MA, Engel LS, Chow WH, Vaughan TL, Gammon MD,
Iqbal B et al. Poor oral hygiene and risk of esoph- Risch HA, Stanford JL et al. Population attributa-
ageal squamous cell carcinoma in Kashmir. Br J ble risks of esophageal and gastric cancers. J Natl
Cancer 2013;109:1367–​72. Cancer Inst 2003;95:1404–​13.
Dawsey SM, Wang GQ, Taylor PR, Li JY, Blot WJ, Li B Engel LS, Vaughan TL, Gammon MD, Chow WH,
et al. Effects of vitamin/​mineral supplementation Risch HA, Dubrow R et  al. Occupation and risk
on the prevalence of histological dysplasia and of esophageal and gastric cardia adenocarcinoma.
early cancer of the esophagus and stomach: results Am J Industr Med 2002;42:11–​22.
from the Dysplasia Trial in Linxian, China. Cancer Evanoff BA, Gustavsson P, Hogstedt C. Mortality and
Epidemiol Biomarkers Prev 1994;3:167–​72. incidence of cancer in a cohort of Swedish chim-
De Jong UW, Day NE, Mounier-​Kuhn PL, Haguenauer ney sweeps:  an extended follow up study. Br J
JP. The relationship between the ingestion of hot Industr Med 1993;50:450–​9.
coffee and intraoesophageal temperature. Gut Fagundes RB, de Carli D, Xaubet RV, Cantarelli JC Jr.
1972;13:24–​30. Unchanging pattern of prevalence of esophageal
de Martel C, Llosa AE, Farr SM, Friedman GD, cancer, overall and by histological subtype, in the
Vogelman JH, Orentreich N et  al. Helicobacter endoscopy service of the main referral hospital
pylori infection and the risk of development in the central region of Rio Grande do Sul State,
of esophageal adenocarcinoma. J Infect Dis in Southern Brazil. Dis Esophagus 2015 May 13.
2005;191:761–​7. doi: 10.1111/​dote.12350. [Epub ahead of print]
De Stefani E, Deneo-​Pellegrini H, Ronco AL, Boffetta Fitzgerald RC. Complex diseases in gastroenterology
P, Brennan P, Munoz N et  al. Food groups and and hepatology:  GERD, Barrett’s, and esopha-
risk of squamous cell carcinoma of the oesopha- geal adenocarcinoma. Clin Gastroenterol Hepatol
gus: a case-​control study in Uruguay. Br J Cancer 2005;3:529–​37.
2003;89:1209–​14. Franceschi S, Bidoli E, Negri E, Barbone F, La Vecchia
De Stefani E, Ronco A, Mendilaharsu M, Deneo-​ C. Alcohol and cancers of the upper aerodiges-
Pellegrini H. Case-​control study on the role of tive tract in men and women. Cancer Epidemiol
heterocyclic amines in the etiology of upper Biomarkers Prev 1994;3:299–​304.
aerodigestive cancers in Uruguay. Nutr Cancer Franceschi S, Bidoli E, Negri E, Zambon P, Talamini
1998;32:43–​8. R, Ruol A et  al. Role of macronutrients, vita-
Dulak AM, Stojanov P, Peng S, Lawrence MS, Fox mins and minerals in the aetiology of squamous-​
C, Stewart C et  al. Exome and whole-​genome cell carcinoma of the oesophagus. Int J Cancer
sequencing of esophageal adenocarcinoma identi- 2000;86:626–​31.
fies recurrent driver events and mutational com- Freedman ND, Abnet CC, Leitzmann MF, Mouw T,
plexity. Nat Genet 2013;45:478–​86. Subar AF, Hollenbeck AR et  al. A prospective
Edgren G, Adami HO, Weiderpass E, Nyren O. A study of tobacco, alcohol, and the risk of esopha-
global assessment of the oesophageal adenocarci- geal and gastric cancer subtypes. Am J Epidemiol
noma epidemic. Gut 2013;62:1406–​14. 2007;165:1424–​33.
Ek WE, Lagergren K, Cook M, Wu AH, Abnet CC, Freedman ND, Lacey JV Jr, Hollenbeck AR, Leitzmann
Levine D et  al. Polymorphisms in genes in the MF, Schatzkin A, Abnet CC. The association of
androgen pathway and risk of Barrett’s esopha- menstrual and reproductive factors with upper
gus and esophageal adenocarcinoma. Int J Cancer gastrointestinal tract cancers in the NIH-​AARP
2016;138:1146–​52. cohort. Cancer 2010;116:1572–​81.
Ek WE, Levine DM, D’Amato M, Pedersen NL, Freedman ND, Murray LJ, Kamangar F, Abnet CC,
Magnusson PK, Bresso F et al. Germline genetic Cook MB, Nyren O et  al. Alcohol intake and
contributions to risk for esophageal adenocarci- risk of oesophageal adenocarcinoma:  a pooled
noma, Barrett’s esophagus, and gastroesophageal analysis from the BEACON Consortium. Gut
reflux. J Natl Cancer Inst 2013;105:1711–​8. 2011;60:1029–​37.
Ellis A, Field JK, Field EA, Friedmann PS, Fryer Freedman ND, Park Y, Subar AF, Hollenbeck AR,
A, Howard P et  al. Tylosis associated with car- Leitzmann MF, Schatzkin A et  al. Fruit and
cinoma of the oesophagus and oral leukoplakia vegetable intake and esophageal cancer in a
in a large Liverpool family—​a review of six gen- large prospective cohort study. Int J Cancer
erations. Eur J Cancer Part B, Oral Oncology 2007;121:2753–​60.
1994;30b:102–​12. Galeone C, Pelucchi C, Talamini R, Levi F, Bosetti C,
El-​Serag HB, Hollier JM, Gravitt P, Alsarraj A, Younes Negri E et  al. Role of fried foods and oral/​pha-
M. Human papillomavirus and the risk of Barrett’s ryngeal and oesophageal cancers. Br J Cancer
esophagus. Dis Esophagus 2013;26:517–​21. 2005;92:2065–​9.
206

206 Textbook of Cancer Epidemiology

Galipeau PC, Cowan DS, Sanchez CA, Barrett MT, cancer in Shaanxi Province, China. J Natl Cancer
Emond MJ, Levine DS et  al. 17p (p53) allelic Inst 1996;88:1467–​71.
losses, 4N (G2/​ tetraploid) populations, and Hansen J, Raaschou-​ Nielsen O, Christensen JM,
progression to aneuploidy in Barrett’s esoph- Johansen I, McLaughlin JK, Lipworth L et  al.
agus. Proceedings of the National Academy Cancer incidence among Danish workers exposed
of Sciences of the United States of America. to trichloroethylene. J Occup Environ Med
1996;93:7081–​4. 2001;43:133–​9.
Galipeau PC, Prevo LJ, Sanchez CA, Longton GM, Hansen S, Melby KK, Aase S, Jellum E, Vollset SE.
Reid BJ. Clonal expansion and loss of heterozy- Helicobacter pylori infection and risk of car-
gosity at chromosomes 9p and 17p in premalig- dia cancer and non-​ cardia gastric cancer:  a
nant esophageal (Barrett’s) tissue. J Natl Cancer nested case-​control study. Scand J Gastroenterol
Inst 1999;91:2087–​95. 1999;34:353–​60.
Gammon MD, Schoenberg JB, Ahsan H, Risch HA, Hashibe M, McKay JD, Curado MP, Oliveira JC,
Vaughan TL, Chow WH et  al. Tobacco, alcohol, Koifman S, Koifman R et al. Multiple ADH genes
and socioeconomic status and adenocarcinomas are associated with upper aerodigestive cancers.
of the esophagus and gastric cardia. J Natl Cancer Nat Genet 2008;40:707–​9.
Inst 1997;89:1277–​84. He YT, Hou J, Chen ZF, Qiao CY, Song GH, Meng
Gao YB, Chen ZL, Li JG, Hu XD, Shi XJ, Sun ZM et al. FS et  al. Decrease in the esophageal cancer inci-
Genetic landscape of esophageal squamous cell dence rate in mountainous but not level parts
carcinoma. Nat Genet 2014;46:1097–​102. of Cixian County, China, over 29  years. APJCP
Gao YT, McLaughlin JK, Blot WJ, Ji BT, Dai Q, 2005;6:510–​4.
Fraumeni JF Jr. Reduced risk of esophageal can- Henrik Siman J, Forsgren A, Berglund G, Floren CH.
cer associated with green tea consumption. J Natl Helicobacter pylori infection is associated with
Cancer Inst 1994;86:855–​8. a decreased risk of developing oesophageal neo-
Gonzalez CA, Pera G, Agudo A, Bueno-​de-​Mesquita plasms. Helicobacter 2001;6:310–​6.
HB, Ceroti M, Boeing H et  al. Fruit and vegeta- Hiraki A, Matsuo K, Suzuki T, Kawase T, Tajima K.
ble intake and the risk of stomach and oesopha- Teeth loss and risk of cancer at 14 common sites
gus adenocarcinoma in the European Prospective in Japanese. Cancer Epidemiol Biomarkers Prev
Investigation into Cancer and Nutrition (EPIC-​ 2008;17:1222–​7.
EURGAST). Int J Cancer 2006;118:2559–​66. Hongo M. Review article:  Barrett’s oesophagus and
Graham S, Marshall J, Haughey B, Brasure J, carcinoma in Japan. Aliment Pharmacol Ther
Freudenheim J, Zielezny M et  al. Nutritional 2004;20 Suppl 8:50–​4.
epidemiology of cancer of the esophagus. Am J Hormozdiari H, Day NE, Aramesh B, Mahboubi E.
Epidemiol 1990;131:454–​67. Dietary factors and esophageal cancer in the Caspian
Guo W, Blot WJ, Li JY, Taylor PR, Liu BQ, Wang W, Littoral of Iran. Cancer Res 1975;35:3493–​8.
et al. A nested case-​control study of oesophageal Hsing AW, McLaughlin JK, Olsen JH, Mellemkjar
and stomach cancers in the Linxian nutrition L, Wacholder S, Fraumeni JF Jr. Cancer risk fol-
intervention trial. International journal of epide- lowing primary hemochromatosis:  a population-​
miology. 1994;23(3):444–​50. based cohort study in Denmark. Int J Cancer
Guo F, Liu Y, Wang X, He Z, Weiss NS, Madeleine 1995;60:160–​2.
MM et  al. Human papillomavirus infection and Hu J, Nyren O, Wolk A, Bergstrom R, Yuen J, Adami
esophageal squamous cell carcinoma:  a case-​ HO et  al. Risk factors for oesophageal cancer in
control study. Cancer Epidemiol Biomarkers Prev northeast China. Int J Cancer 1994;57:38–​46.
2012;21:780–​5. Hvid-​Jensen F, Pedersen L, Drewes AM, Sorensen HT,
Gustavsson P, Evanoff B, Hogstedt C. Increased risk Funch-​ Jensen P. Incidence of adenocarcinoma
of esophageal cancer among workers exposed among patients with Barrett’s esophagus. New
to combustion products. Arch Environ Health Engl J Med 2011;365:1375–​83.
1993;48:243–​5. Inoue M, Tajima K, Hirose K, Hamajima N, Takezaki
Gustavsson P, Jakobsson R, Johansson H, Lewin F, T, Kuroishi T et  al. Tea and coffee consumption
Norell S, Rutkvist LE. Occupational exposures and and the risk of digestive tract cancers: data from a
squamous cell carcinoma of the oral cavity, phar- comparative case-​referent study in Japan. Cancer
ynx, larynx, and oesophagus: a case-​control study Causes Control 1998;9:209–​16.
in Sweden. Occup Enviro Med 1998;55:393–​400. Isaacson C, Bothwell TH, MacPhail AP, Simon M. The
Han C, Qiao G, Hubbert NL, Li L, Sun C, Wang Y iron status of urban black subjects with carcinoma
et  al. Serologic association between human pap- of the oesophagus. South African Med Journal
illomavirus type 16 infection and esophageal 1985;67:591–​3.
207

Esophageal Cancer 207

Ishikawa A, Kuriyama S, Tsubono Y, Fukao A, Kamangar F, Dawsey SM, Blaser MJ, Perez-​Perez GI,
Takahashi H, Tachiya H et  al. Smoking, alcohol Pietinen P, Newschaffer CJ et  al. Opposing risks
drinking, green tea consumption and the risk of of gastric cardia and noncardia gastric adenocar-
esophageal cancer in Japanese men. J Epidemiol cinomas associated with Helicobacter pylori sero-
2006;16:185–​92. positivity. J Natl Cancer Inst 2006;98:1445–​52.
Islami F, Boffetta P, Ren JS, Pedoeim L, Khatib D, Kamangar F, Qiao YL, Blaser MJ, Sun XD, Katki H,
Kamangar F. High-​ temperature beverages and Fan JH et  al. Helicobacter pylori and oesopha-
foods and esophageal cancer risk—​a systematic geal and gastric cancers in a prospective study in
review. Int J Cancer 2009;125:491–​524. China. Br J Cancer 2007;96:172–​6.
Islami F, Kamangar F, Aghcheli K, Fahimi S, Semnani Kamangar F, Qiao YL, Schiller JT, Dawsey SM, Fears T,
S, Taghavi N et al. Epidemiologic features of upper Sun XD et al. Human papillomavirus serology and
gastrointestinal tract cancers in Northeastern the risk of esophageal and gastric cancers: results
Iran. Br J Cancer. 2004;90:1402–​6. from a cohort in a high-​risk region in China. Int J
Islami F, Pourshams A, Nasrollahzadeh D, Kamangar Cancer 2006;119:579–​84.
F, Fahimi S, Shakeri R et  al. Tea drinking hab- Kamangar F, Schantz MM, Abnet CC, Fagundes RB,
its and oesophageal cancer in a high risk area in Dawsey SM. High levels of carcinogenic polycyclic
northern Iran:  population based case-​ control aromatic hydrocarbons in mate drinks. Cancer
study. BMJ 2009;338:b929. Epidemiol Biomarkers Prev 2008;17:1262–​8.
Islami F, Ren JS, Taylor PR, Kamangar F. Pickled veg- Kang SK, Burnett CA, Freund E, Walker J, Lalich N,
etables and the risk of oesophageal cancer: a meta-​ Sestito J. Gastrointestinal cancer mortality of
analysis. Br J Cancer 2009;101:1641–​7. workers in occupations with high asbestos expo-
Islami F, Sheikhattari P, Ren JS, Kamangar F. Gastric sures. Am J Industr Med 1997;31:713–​8.
atrophy and risk of oesophageal cancer and gas- Kjaerheim K, Gaard M, Andersen A. The role of alcohol,
tric cardia adenocarcinoma—​a systematic review tobacco, and dietary factors in upper aerogastric tract
and meta-​analysis. Annals Oncol 2011;22:754–​60. cancers:  a prospective study of 10,900 Norwegian
Jakobsson K, Attewell R, Hultgren B, Sjoland K. men. Cancer Causes Control 1998;9:99–​108.
Gastrointestinal cancer among cement work- Kubo A, Cook MB, Shaheen NJ, Vaughan TL,
ers: a case-​referent study. Int Arch Occup Environ Whiteman DC, Murray L et al. Sex-​specific asso-
Health 1990;62:337–​40. ciations between body mass index, waist circum-
Jakszyn P, Gonzalez CA. Nitrosamine and related food ference and the risk of Barrett’s oesophagus:  a
intake and gastric and oesophageal cancer risk: a pooled analysis from the international BEACON
systematic review of the epidemiological evi- consortium. Gut 2013;62:1684–​91.
dence. World J Gastroenterol 2006;12:4296–​303. Kubo A, Levin TR, Block G, Rumore GJ, Quesenberry
Jansson C, Johansson AL, Jeding K, Dickman PW, CP Jr, Buffler P et al. Alcohol types and sociodemo-
Nyren O, Lagergren J. Psychosocial working con- graphic characteristics as risk factors for Barrett’s
ditions and the risk of esophageal and gastric car- esophagus. Gastroenterol 2009;136:806–​15.
dia cancers. Eur J Epidemiol 2004;19:631–​41. Kuwano H, Kato H, Miyazaki T, Fukuchi M, Masuda
Jansson C, Plato N, Johansson AL, Nyren O, Lagergren N, Nakajima M et al. Genetic alterations in esoph-
J. Airborne occupational exposures and risk of ageal cancer. Surgery Today 2005;35:7–​18.
oesophageal and cardia adenocarcinoma. Occup Lagergren J, Bergstrom R, Lindgren A, Nyren O.
Environ Med 2006;63:107–​12. Symptomatic gastroesophageal reflux as a risk fac-
Jaskiewicz K, Marasas WF, Lazarus C, Beyers AD, tor for esophageal adenocarcinoma. N Engl J Med
Van Helden PD. Association of esophageal cyto- 1999;340:825–​31.
logical abnormalities with vitamin and lipotrope Lagergren J, Bergstrom R, Lindgren A, Nyren O. The
deficiencies in populations at risk for esophageal role of tobacco, snuff and alcohol use in the aetiol-
cancer. Anticancer Res 1988;8:711–​5. ogy of cancer of the oesophagus and gastric car-
Ji J, Hemminki K. Familial risk for esophageal can- dia. Int J Cancer 2000;85:340–​6.
cer: an updated epidemiologic study from Sweden. Lagergren J, Bergstrom R, Nyren O. Association
Clin Gastroenterol Hepatol 2006;4:840–​5. between body mass and adenocarcinoma of the
Jussawalla DJ, Deshpande VA. Evaluation of cancer esophagus and gastric cardia. Annals Int Med
risk in tobacco chewers and smokers:  an epide- 1999;130:883–​90.
miologic assessment. Cancer 1971;28:244–​52. Lagergren J, Jansson C. Sex hormones and oesopha-
Kabat GC, Ng SK, Wynder EL. Tobacco, alcohol geal adenocarcinoma:  influence of childbearing?
intake, and diet in relation to adenocarcinoma of Br J Cancer 2005;93:859–​61.
the esophagus and gastric cardia. Cancer Causes Lagergren J, Nyren O. Do sex hormones play a role
Control 1993;4:123–​32. in the etiology of esophageal adenocarcinoma?
208

208 Textbook of Cancer Epidemiology

A  new hypothesis tested in a population-​based esophageal squamous cell carcinoma. Nat Genet
cohort of prostate cancer patients. Cancer 2014;46:467–​73.
Epidemiol Biomarkers Prev 1998;7:913–​5. Lindblad M, Garcia Rodriguez LA, Chandanos E,
Lagergren J, Wang Z, Bergstrom R, Dillner J, Nyren O. Lagergren J. Hormone replacement therapy and
Human papillomavirus infection and esophageal risks of oesophageal and gastric adenocarcino-
cancer: a nationwide seroepidemiologic case-​control mas. Br J Cancer 2006;94:136–​41.
study in Sweden. J Natl Cancer Inst 1999;91:156–​62. Lindblad M, Rodriguez LA, Lagergren J. Body mass,
Lam AK. Molecular biology of esophageal squa- tobacco and alcohol and risk of esophageal, gastric
mous cell carcinoma. Crit Rev Oncol Hematol cardia, and gastric non-​cardia adenocarcinoma
2000;33:71–​90. among men and women in a nested case-​control
Launoy G, Milan C, Day NE, Pienkowski MP, Gignoux study. Cancer Causes Control 2005;16:285–​94.
M, Faivre J. Diet and squamous-​cell cancer of the Lindblad M, Ye W, Lindgren A, Lagergren J. Disparities
oesophagus:  a French multicentre case-​ control in the classification of esophageal and cardia ade-
study. Int J Cancer 1998;76:7–​12. nocarcinomas and their influence on reported
Launoy G, Milan C, Faivre J, Pienkowski P, Gignoux incidence rates. Annals Surg 2006;243:479–​85.
M. Tobacco type and risk of squamous cell cancer Lu Y, Lagergren J. Reproductive factors and risk of
of the oesophagus in males: a French multicentre oesophageal cancer, a population-​ based nested
case-​control study. Int J Epidemiol 2000;29:36–​42. case-​control study in Sweden. Br J Cancer
Launoy G, Milan CH, Faivre J, Pienkowski P, Milan 2012;107:564–​9.
CI, Gignoux M. Alcohol, tobacco and oesophageal Lubin JH, Cook MB, Pandeya N, Vaughan TL, Abnet
cancer:  effects of the duration of consumption, CC, Giffen C et  al. The importance of exposure
mean intake and current and former consump- rate on odds ratios by cigarette smoking and alco-
tion. Br J Cancer 1997;75:1389–​96. hol consumption for esophageal adenocarcinoma
Leeuwenburgh I, Scholten P, Alderliesten J, Tilanus and squamous cell carcinoma in the Barrett’s
HW, Looman CW, Steijerberg EW et al. Long-​term Esophagus and Esophageal Adenocarcinoma
esophageal cancer risk in patients with primary Consortium. Cancer Epidemiol 2012;36:306–​16.
achalasia: a prospective study. Am J Gastroenterol Lynge E, Anttila A, Hemminki K. Organic solvents and
2010;105:2144–​9. cancer. Cancer Causes Control 1997;8:406–​19.
Levine DM, Ek WE, Zhang R, Liu X, Onstad L, Sather MacInnis RJ, English DR, Hopper JL, Giles GG.
C et al. A genome-​wide association study identi- Body size and composition and the risk of gastric
fies new susceptibility loci for esophageal ade- and oesophageal adenocarcinoma. Int J Cancer
nocarcinoma and Barrett’s esophagus. Nat Genet 2006;118:2628–​31.
2013;45:1487–​93. Mayne ST, Risch HA, Dubrow R, Chow WH, Gammon
Li JY, Ershow AG, Chen ZJ, Wacholder S, Li GY, MD, Vaughan TL et al. Nutrient intake and risk of
Guo W et  al. A case-​control study of cancer of subtypes of esophageal and gastric cancer. Cancer
the esophagus and gastric cardia in Linxian. Int J Epidemiol Biomarkers Prev 2001;10:1055–​62.
Cancer 1989;43:755–​61. Montesano R, Hollstein M, Hainaut P. Genetic altera-
Li JY, Taylor PR, Li B, Dawsey S, Wang GQ, Ershow tions in esophageal cancer and their relevance to
AG et al. Nutrition intervention trials in Linxian, etiology and pathogenesis: a review. Int J Cancer
China:  multiple vitamin/​ mineral supplementa- 1996;69:225–​35.
tion, cancer incidence, and disease-​specific mor- Murphy G, Dawsey SM, Engels EA, Ricker W, Parsons
tality among adults with esophageal dysplasia. J R, Etemadi A et  al. Cancer risk after perni-
Natl Cancer Inst 1993;85:1492–​8. cious anemia in the US elderly population. Clin
Li M, Edamatsu H, Kitazawa R, Kitazawa S, Kataoka Gastroenterol Hepatol 2015;13:2282–​9 e4.
T. Phospholipase Cepsilon promotes intestinal Murphy G, Kamangar F, Albanes D, Stanczyk FZ,
tumorigenesis of Apc(Min/​+) mice through aug- Weinstein SJ, Taylor PR et  al. Serum ghrelin
mentation of inflammation and angiogenesis. is inversely associated with risk of subsequent
Carcinogenesis 2009;30:1424–​32. oesophageal squamous cell carcinoma. Gut
Limburg PJ, Wei W, Ahnen DJ, Qiao Y, Hawk ET, 2012;61:1533–​7.
Wang G et  al. Randomized, placebo-​controlled, Nandakumar A, Anantha N, Pattabhiraman V,
esophageal squamous cell cancer chemopreven- Prabhakaran PS, Dhar M, Puttaswamy K et  al.
tion trial of selenomethionine and celecoxib. Importance of anatomical subsite in correlat-
Gastroenterol 2005;129:863–​73. ing risk factors in cancer of the oesophagus—​
Lin DC, Hao JJ, Nagata Y, Xu L, Shang L, Meng X report of a case—​ control study. Br J Cancer
et al. Genomic and molecular characterization of 1996;73:1306–​11.
209

Esophageal Cancer 209

Nasrollahzadeh D, Kamangar F, Aghcheli K, Sotoudeh 1963–​1977. Cancer Epidemiol Biomarkers Prev

M, Islami F, Abnet CC et al. Opium, tobacco, and 1994;3:537–​47.
alcohol use in relation to oesophageal squamous Pera M, Manterola C, Vidal O, Grande L. Epidemiology
cell carcinoma in a high-​risk area of Iran. Br J of esophageal adenocarcinoma. J Surg Oncol
Cancer 2008;98:1857–​63. 2005;92:151–​9.
Nasrollahzadeh D, Malekzadeh R, Aghcheli K, Pohl H, Welch HG. The role of overdiagnosis and
Sotoudeh M, Merat S, Islami F et  al. Gastric reclassification in the marked increase of esopha-
atrophy and oesophageal squamous cell carci- geal adenocarcinoma incidence. J Natl Cancer Inst
noma: possible interaction with dental health and 2005;97:142–​6.
oral hygiene habit. Br J Cancer 2012;107:888–​94. Qiao YL, Dawsey SM, Kamangar F, Fan JH, Abnet
Newnham A, Quinn MJ, Babb P, Kang JY, Majeed CC, Sun XD et  al. Total and cancer mortality
A. Trends in the subsite and morphology of after supplementation with vitamins and miner-
oesophageal and gastric cancer in England and als:  follow-​up of the Linxian General Population
Wales 1971–​ 1998. Aliment Pharmacol Ther Nutrition Intervention Trial. J Natl Cancer Inst
2003;17:665–​76. 2009;101:507–​18.
Nones K, Waddell N, Wayte N, Patch AM, Bailey P, Rajendra S, Wang B, Snow ET, Sharma P, Pavey D,
Newell F et  al. Genomic catastrophes frequently Merrett N et  al. Transcriptionally active human
arise in esophageal adenocarcinoma and drive papillomavirus is strongly associated with Barrett’s
tumorigenesis. Nat Com 2014;5:5224. dysplasia and esophageal adenocarcinoma. Am J
Norell S, Ahlbom A, Lipping H, Osterblom L. Gastroenterol 2013;108:1082–​93.
Oesophageal cancer and vulcanisation work. Ransford RA, Jankowski JA. Genetic versus environ-
Lancet 1983;1:462–​3. mental interactions in the oesophagitis-​metapla-
O’Doherty MG, Freedman ND, Hollenbeck AR, sia-​dysplasia-​adenocarcinoma sequence (MCS) of
Schatzkin A, Abnet CC. A prospective cohort Barrett’s oesophagus. Acta Gastro-​Enterol Belgica
study of obesity and risk of oesophageal and gas- 2000;63:18–​21.
tric adenocarcinoma in the NIH-​AARP Diet and Rao M, Liu FS, Dawsey SM, Yang K, Lipkin M, Li JY
Health Study. Gut 2012;61:1261–​8. et  al. Effects of vitamin/​ mineral supplementa-
Oka M, Edamatsu H, Kunisada M, Hu L, Takenaka tion on the proliferation of esophageal squamous
N, Dien S et  al. Enhancement of ultraviolet epithelium in Linxian, China. Cancer Epidemiol
B-​induced skin tumor development in phos- Biomarkers Prev 1994;3:277–​9.
pholipase Cepsilon-​ knockout mice is associ- Ren JS, Freedman ND, Kamangar F, Dawsey SM,
ated with decreased cell death. Carcinogenesis Hollenbeck AR, Schatzkin A et  al. Tea, coffee,
2010;31:1897–​902. carbonated soft drinks and upper gastrointestinal
O’Neill C, Pan Q, Clarke G, Liu F, Hodges G, Ge M tract cancer risk in a large United States prospec-
et al. Silica fragments from millet bran in mucosa tive cohort study. Eur J Cancer 2010;46:1873–​81.
surrounding oesophageal tumours in patients in Ren JS, Kamangar F, Forman D, Islami F. Pickled
northern China. Lancet 1982;1:1202–​6. food and risk of gastric cancer—​ a systematic
Pan G, Takahashi K, Feng Y, Liu L, Liu T, Zhang S et al. review and meta-​analysis of English and Chinese
Nested case-​control study of esophageal cancer in literature. Cancer Epidemiol Biomarkers Prev
relation to occupational exposure to silica and 2012;21:905–​15.
other dusts. Am J Industr Med 1999;35:272–​80. Ren JS, Kamangar F, Qiao YL, Taylor PR, Liang H,
Pandeya N, Williams G, Green AC, Webb PM, Whiteman Dawsey SM et al. Serum pepsinogens and risk of
DC. Alcohol consumption and the risks of adeno- gastric and oesophageal cancers in the General
carcinoma and squamous cell carcinoma of the Population Nutrition Intervention Trial cohort.
esophagus. Gastroenterol 2009;136:1215–​24, e1–​2. Gut 2009;58:636–​42.
Parker RK, Dawsey SM, Abnet CC, White RE. Roshandel G, Merat S, Sotoudeh M, Khoshnia M,
Frequent occurrence of esophageal cancer in Poustchi H, Lao-​Sirieix P et al. Pilot study of cyto-
young people in western Kenya. Dis Esophagus logical testing for oesophageal squamous cell dys-
2010;23:128–​35. plasia in a high-​risk area in Northern Iran. Br J
Parkin DM, Pisani P, Ferlay J. Estimates of the world- Cancer 2014;111:2235–​41.
wide incidence of 25 major cancers in 1990. Int J Ross-​ Innes CS, Becq J, Warren A, Cheetham RK,
Cancer 1999;80:827–​41. Northen H, O’Donovan M et  al. Whole-​genome
Parkin DM, Vizcaino AP, Skinner ME, Ndhlovu sequencing provides new insights into the clonal
A. Cancer patterns and risk factors in the architecture of Barrett’s esophagus and esophageal
African population of southwestern Zimbabwe, adenocarcinoma. Nat Genet 2015;47:1038–​46.
210

210 Textbook of Cancer Epidemiology

Sathiakumar N, Delzell E, Austin H, Cole P. A fol- Su Z, Gay LJ, Strange A, Palles C, Band G, Whiteman
low-​up study of agricultural chemical production DC et al. Common variants at the MHC locus and
workers. Am J Indust Med 1992;21:321–​30. at chromosome 16q24.1 predispose to Barrett’s
Segal I, Reinach SG, de Beer M. Factors associated esophagus. Nat Genet 2012;44:1131–​6.
with oesophageal cancer in Soweto, South Africa. Suzuki H, Zhou X, Yin J, Lei J, Jiang HY, Suzuki Y et al.
Br J Cancer 1988;58:681–​6. Intragenic mutations of CDKN2B and CDKN2A
Semnani S, Sadjadi A, Fahimi S, Nouraie M, Naeimi in primary human esophageal cancers. Hum Mol
M, Kabir J et  al. Declining incidence of esoph- Genet. 1995;4:1883–​7.
ageal cancer in the Turkmen Plain, eastern part Takezaki T, Gao CM, Wu JZ, Ding JH, Liu YT, Zhang Y
of the Caspian Littoral of Iran:  a retrospec- et al. Dietary protective and risk factors for esoph-
tive cancer surveillance. Cancer Detect Prev ageal and stomach cancers in a low-​ epidemic
2006;30:14–​9. area for stomach cancer in Jiangsu Province,
Sepehr A, Taniere P, Martel-​Planche G, Zia’ee AA, China: comparison with those in a high-​epidemic
Rastgar-​Jazii F, Yazdanbod M et  al. Distinct pat- area. Jpn J Cancer Res 2001;92:1157–​65.
tern of TP53 mutations in squamous cell car- Takezaki T, Shinoda M, Hatooka S, Hasegawa Y,
cinoma of the esophagus in Iran. Oncogene Nakamura S, Hirose K et al. Subsite-​specific risk
2001;20:7368–​74. factors for hypopharyngeal and esophageal cancer
Sewram V, De Stefani E, Brennan P, Boffetta P. Mate (Japan). Cancer Causes Control 2000;11:597–​608.
consumption and the risk of squamous cell Taniere P, Martel-​Planche G, Maurici D, Lombard-​
esophageal cancer in Uruguay. Cancer Epidemiol Bohas C, Scoazec JY, Montesano R et al. Molecular
Biomarkers Prev 2003;12:508–​13. and clinical differences between adenocarcinomas
Shaheen NJ, Crosby MA, Bozymski EM, Sandler RS. of the esophagus and of the gastric cardia. Am J
Is there publication bias in the reporting of can- Pathol 2001;158:33–​40.
cer risk in Barrett’s esophagus? Gastroenterol Tavani A, La Vecchia C, Gallus S, Lagiou P, Trichopoulos
2000;119:333–​8. D, Levi F et al. Red meat intake and cancer risk: a
Shaheen NJ, Falk GW, Iyer PG, Gerson LB. ACG study in Italy. Int J Cancer 2000;86:425–​8.
clinical guideline:  diagnosis and management of Tavani A, Negri E, Franceschi S, La Vecchia C. Risk fac-
Barrett’s esophagus. American J Gastroenterol tors for esophageal cancer in lifelong nonsmokers.
2016;111:30–​50. Cancer Epidemiol Biomarkers Prev1994;3:387–​92.
Sikkema M, de Jonge PJ, Steyerberg EW, Kuipers EJ. Taylor JB, Rubenstein JH. Meta-​analyses of the effect
Risk of esophageal adenocarcinoma and mortal- of symptoms of gastroesophageal reflux on the
ity in patients with Barrett’s esophagus: a system- risk of Barrett’s esophagus. Am J Gastroenterol
atic review and meta-​analysis. Clin Gastroenterol 2010;105:1729, 30–​7; quiz 38.
Hepatol 2010;8:235–​44. Taylor PR, Abnet CC, Dawsey SM. Squamous
Song Y, Li L, Ou Y, Gao Z, Li E, Li X et al. Identification dysplasia—the precursor lesion for esophageal
­
of genomic alterations in oesophageal squamous squamous cell carcinoma. Cancer Epidemiol
cell cancer. Nature 2014;509:91–​5. Biomarkers Prev 2013;22:540–​52.
Souza RF, Spechler SJ. Concepts in the prevention of Terry P, Lagergren J, Hansen H, Wolk A, Nyren O.
adenocarcinoma of the distal esophagus and prox- Fruit and vegetable consumption in the preven-
imal stomach. CA 2005;55:334–​51. tion of oesophageal and cardia cancers. Eur J
Stachler MD, Taylor-​Weiner A, Peng S, McKenna A, Cancer Prev 2001;10:365–​9.
Agoston AT, Odze RD et al. Paired exome analysis Terry PD, Lagergren J, Wolk A, Steineck G, Nyren O.
of Barrett’s esophagus and adenocarcinoma. Nat Dietary intake of heterocyclic amines and can-
Genet 2015;47:1047–​55. cers of the esophagus and gastric cardia. Cancer
Steevens J, Schouten LJ, Goldbohm RA, van den Epidemiol Biomarkers Prev 2003;12:940–​4.
Brandt PA. Vegetables and fruits consumption Thrift AP, Risch HA, Onstad L, Shaheen NJ, Casson
and risk of esophageal and gastric cancer subtypes AG, Bernstein L et al. Risk of esophageal adeno-
in the Netherlands Cohort Study. Int J Cancer carcinoma decreases with height, based on con-
2011;129:2681–​93. sortium analysis and confirmed by Mendelian
Steinmetz KA, Potter JD. Vegetables, fruit, and randomization. Clin Gastroenterol Hepatol
cancer prevention:  a review. J Am Diet Ass 2014;12:1667–​76.e1.
1996;96:1027–​39. Thrift AP, Shaheen NJ, Gammon MD, Bernstein L,
Streitz JM Jr. Barrett’s esophagus and esophageal can- Reid BJ, Onstad L et al. Obesity and risk of esoph-
cer. Chest Surg Clin North Am. 1994;4:227–​40. ageal adenocarcinoma and Barrett’s esophagus: a
21

Esophageal Cancer 211

Mendelian randomization study. J Natl Cancer Wang JB, Abnet CC, Fan JH, Qiao YL, Taylor PR.
Inst 2014;106. The randomized Linxian Dysplasia Nutrition
Thurnham DI, Zheng SF, Munoz N, Crespi M, Grassi Intervention Trial after 26 years of follow-​up: no
A, Hambidge KM et al. Comparison of riboflavin, effect of multivitamin supplementation on mor-
vitamin A, and zinc status of Chinese populations tality. JAMA Intern Med 2013;173:1259–​61.
at high and low risk for esophageal cancer. Nutr Wang LD, Zhou FY, Li XM, Sun LD, Song X, Jin Y
Cancer 1985;7:131–​43. et  al. Genome-​wide association study of esoph-
Tian D, Feng Z, Hanley NM, Setzer RW, Mumford JL, ageal squamous cell carcinoma in Chinese sub-
DeMarini DM. Multifocal accumulation of p53 jects identifies susceptibility loci at PLCE1 and
protein in esophageal carcinoma:  evidence for C20orf54. Nat Genet 2010;42:759–​63.
field cancerization. Int J Cancer 1998;78:568–​75. Wang YP, Han XY, Su W, Wang YL, Zhu YW, Sasaba
Tran GD, Sun XD, Abnet CC, Fan JH, Dawsey SM, T et  al. Esophageal cancer in Shanxi Province,
Dong ZW et al. Prospective study of risk factors People’s Republic of China:  a case-​control study
for esophageal and gastric cancers in the Linxian in high and moderate risk areas. Cancer Causes
general population trial cohort in China. Int J Control 1992;3:107–​13.
Cancer 2005;113:456–​63. Ward MH, Sinha R, Heineman EF, Rothman N,
Tuyns AJ, Pequignot G, Abbatucci JS. Oesophageal Markin R, Weisenburger DD et  al. Risk of ade-
cancer and alcohol consumption; importance of nocarcinoma of the stomach and esophagus with
type of beverage. Int J Cancer 1979;23:443–​7. meat cooking method and doneness preference.
Tuyns AJ, Pequignot G, Jensen OM. Esophageal can- Int J Cancer 1997;71:14–​9.
cer in Ille-​et-​Vilaine in relation to levels of alcohol Weaver JM, Ross-​Innes CS, Shannon N, Lynch AG,
and tobacco consumption. Risks are multiplying. Forshew T, Barbera M et al. Ordering of mutations
Bulletin Cancer 1977;64:45–​60. in preinvasive disease stages of esophageal carcin-
Tuyns AJ. Oesophageal cancer in non-​smoking drink- ogenesis. Nat Genet 2014;46:837–​43.
ers and in non-​drinking smokers. Int J Cancer Wei WQ, Chen ZF, He YT, Feng H, Hou J, Lin DM et al.
1983;32:443–​4. Long-​term follow-​up of a community assignment,
Van Doornum GJ, Korse CM, Buning-​ Kager JC, one-​time endoscopic screening study of esopha-
Bonfrer JM, Horenblas S, Taal BG et al. Reactivity geal cancer in China. J Clin Oncol 2015;33:1951–​7.
to human papillomavirus type 16 L1 virus-​like Wernli KJ, Fitzgibbons ED, Ray RM, Gao DL, Li
particles in sera from patients with genital can- W, Seixas NS et  al. Occupational risk factors
cer and patients with carcinomas at five different for esophageal and stomach cancers among
extragenital sites. Br J Cancer 2003;88:1095–​100. female textile workers in Shanghai, China. Am J
van Rensburg SJ. Epidemiologic and dietary evidence Epidemiol 2006;163:717–​25.
for a specific nutritional predisposition to esoph- White RE, Abnet CC, Mungatana CK, Dawsey SM.
ageal cancer. J Natl Cancer Inst 1981;67:243–​51. Oesophageal cancer:  a common malignancy in
Vassallo A, Correa P, De Stefani E, Cendan M, young people of Bomet District, Kenya. Lancet
Zavala D, Chen V et  al. Esophageal cancer in 2002;360:462–​3.
Uruguay: a case-​control study. J Natl Cancer Inst Winters C Jr, Spurling TJ, Chobanian SJ, Curtis
1985;75:1005–​9. DJ, Esposito RL, Hacker JF 3rd et  al. Barrett’s
Vaughan TL, Stewart PA, Davis S, Thomas DB. Work esophagus:  a prevalent, occult complication of
in dry cleaning and the incidence of cancer of gastroesophageal reflux disease. Gastroenterol
the oral cavity, larynx, and oesophagus. Occup 1987;92:118–​24.
Environ Med 1997;54:692–​5. Wong O, Trent LS, Whorton MD. An updated cohort
Vizcaino AP, Moreno V, Lambert R, Parkin DM. Time mortality study of workers exposed to styrene in
trends incidence of both major histologic types the reinforced plastics and composites industry.
of esophageal carcinomas in selected countries, Occup Environ Med 1994;51:386–​96.
1973–​1995. Int J Cancer 2002;99:860–​8. World Cancer Research Fund. American Institute
Vizcaino AP, Parkin DM, Skinner ME. Risk factors for Cancer Research. Food, Nutrition, Physical
associated with oesophageal cancer in Bulawayo, Activity and the Prevention of Cancer: A Global
Zimbabwe. Br J Cancer 1995;72:769–​73. Perspective: A Project of World Cancer Research
Wang GQ, Abnet CC, Shen Q, Lewin KJ, Sun XD, Roth Fund International. Washington, DC:  American
MJ et  al. Histological precursors of oesophageal Institute for Cancer Research, 2007: xxv, 517.
squamous cell carcinoma:  results from a 13  year Wu AH, Tseng CC, Bernstein L. Hiatal hernia, reflux
prospective follow up study in a high risk popula- symptoms, body size, and risk of esophageal and
tion. Gut 2005;54:187–​92. gastric adenocarcinoma. Cancer 2003;98:940–​8.
21

212 Textbook of Cancer Epidemiology

Wu AH, Wan P, Bernstein L. A multiethnic popula- and squamous-​cell carcinoma of the esophagus
tion-​based study of smoking, alcohol and body and adenocarcinoma of the gastric cardia. J Natl
size and risk of adenocarcinomas of the stomach Cancer Inst 2004;96:388–​96.
and esophagus (United States). Cancer Causes Ye W, Nyren O. Risk of cancers of the oesopha-
Control 2001;12:721–​32. gus and stomach by histology or subsite in
Wu C, Hu Z, He Z, Jia W, Wang F, Zhou Y et  al. patients hospitalised for pernicious anaemia. Gut
Genome-​wide association study identifies three 2003;52:938–​41.
new susceptibility loci for esophageal squamous-​ Yioris N, Ivankovic S, Lehnert T. Effect of thermal
cell carcinoma in Chinese populations. Nat Genet injury and oral administration of N-​methyl-​N’-​
2011;43:679–​84. Nitro-​N-​nitrosoguanidine on the development
Wu C, Kraft P, Zhai K, Chang J, Wang Z, Li Y et  al. of esophageal tumors in Wistar rats. Oncology
Genome-​ wide association analyses of esopha- 1984;41:36–​8.
geal squamous cell carcinoma in Chinese identify Younes M, Lebovitz RM, Lechago LV, Lechago J.
multiple susceptibility loci and gene-​environment p53 protein accumulation in Barrett’s metapla-
interactions. Nat Genet 2012;44:1090–​7. sia, dysplasia, and carcinoma:  a follow-​up study.
Wu C, Wang Z, Song X, Feng XS, Abnet CC, He J et al. Gastroenterol 1993;105:1637–​42.
Joint analysis of three genome-​wide association Yousef F, Cardwell C, Cantwell MM, Galway K,
studies of esophageal squamous cell carcinoma in Johnston BT, Murray L. The incidence of esoph-
Chinese populations. Nat Genet 2014;46:1001–​6. ageal cancer and high-​grade dysplasia in Barrett’s
Wu M, Zhao JK, Hu XS, Wang PH, Qin Y, Lu YC et al. esophagus: a systematic review and meta-​analysis.
Association of smoking, alcohol drinking and die- Am J Epidemiol 2008;168:237–​49.
tary factors with esophageal cancer in high-​and Yu G, Gail MH, Shi J, Klepac-​Ceraj V, Paster BJ, Dye
low-​risk areas of Jiangsu Province, China. World J BA et  al. Association between upper digestive
Gastroenterol 2006;12:1686–​93. tract microbiota and cancer-​predisposing states
Wu TT, Watanabe T, Heitmiller R, Zahurak M, in the esophagus and stomach. Cancer Epidemiol
Forastiere AA, Hamilton SR. Genetic alterations Biomarkers Prev 2014;23:735–​41.
in Barrett esophagus and adenocarcinomas of the Yu H, Liu G, Zhao P, Zhu L. Hormonal and repro-
esophagus and esophagogastric junction region. ductive factors and risk of esophageal cancer in
Am J Pathol 1998;153:287–​94. Chinese postmenopausal women:  a case-​control
Wynder EL, Bross IJ. A study of etiological factors in study. APJCP 2011;12:1953–​6.
cancer of the esophagus. Cancer 1961;14:389–​413. Yu MC, Garabrant DH, Peters JM, Mack TM. Tobacco,
Xing EP, Nie Y, Wang LD, Yang GY, Yang CS. Aberrant alcohol, diet, occupation, and carcinoma of the
methylation of p16INK4a and deletion of esophagus. Cancer Res 1988;48:3843–​8.
p15INK4b are frequent events in human esoph- Yu Y, Taylor PR, Li JY, Dawsey SM, Wang GQ, Guo
ageal cancer in Linxian, China. Carcinogenesis WD et al. Retrospective cohort study of risk-​fac-
1999;20:77–​84. tors for esophageal cancer in Linxian, People’s
Yamaji T, Inoue M, Sasazuki S, Iwasaki M, Kurahashi Republic of China. Cancer Causes Control
N, Shimazu T et  al. Fruit and vegetable con- 1993;4:195–​202.
sumption and squamous cell carcinoma of the Zambon P, Talamini R, La Vecchia C, Dal Maso
esophagus in Japan: the JPHC study. Int J Cancer L, Negri E, Tognazzo S et  al. Smoking, type of
2008;123:1935–​40. alcoholic beverage and squamous-​ cell oesoph-
Yang CS, Sun Y, Yang QU, Miller KW, Li GY, Zheng SF ageal cancer in northern Italy. Int J Cancer
et al. Vitamin A and other deficiencies in Linxian, 2000;86:144–​9.
a high esophageal cancer incidence area in north- Zendehdel K, Nyren O, Edberg A, Ye W. Risk of
ern China. J Natl Cancer Inst 1984;73:1449–​53. esophageal adenocarcinoma in achalasia patients,
Yang L, Lu X, Nossa CW, Francois F, Peek RM, Pei Z. a retrospective cohort study in Sweden. Am J
Inflammation and intestinal metaplasia of the dis- Gastroenterol 2011;106:57–​61.
tal esophagus are associated with alterations in the Zhang L, Zhou Y, Cheng C, Cui H, Cheng L, Kong
microbiome. Gastroenterol 2009;137:588–​97. P et  al. Genomic analyses reveal mutational sig-
Ye W, Held M, Lagergren J, Engstrand L, Blot WJ, natures and frequently altered genes in esopha-
McLaughlin JK et  al. Helicobacter pylori infec- geal squamous cell carcinoma. Am J Hum Genet
tion and gastric atrophy: risk of adenocarcinoma 2015;96:597–​611.
213

10
Stomach Cancer
W E I M I N Y E , O L O F N Y R É N , A N D H A N S - ​O L O V   A D A M I

W hereas cancer researchers have often faced

the challenge of explaining dramatic tem-
poral increases, for stomach cancer the situa-
stomach cancer. Anemia due to bleeding and
obstruction of the gastric outlet due to tumor
growth in the distal stomach are also common.
tion is the reverse. As an “unplanned triumph,” Diagnosis Endoscopic examination with biopsy
its incidence declined dramatically in the 20th and histopathology is the primary diagnostic tool.
century—​in the United States by almost 90% Computed tomography, ultrasound, and other
since the 1930s. Yet on a global scale stomach modern imaging techniques may further clarify
cancer remains one of the most common causes the extent of local tumor growth and reveal metas-
of cancer death. Stomach cancer is the latest, but tases in lymph nodes, the liver, and elsewhere.
perhaps not the last, malignancy for which a pre-
Treatment Cure can be achieved only by radical
dominant infectious etiology has been revealed.
surgical removal—​typically a subtotal or total
Because the epidemiology of proximal gas-
gastrectomy—​of a localized cancer. There is no
tric tumors, or cardia cancer, differs from that of
established adjuvant systemic treatment (Marrelli
distal tumors, reference is made to these specific
et  al, 2015; Elimova et  al, 2015). In advanced
features where relevant. As there is no firm con-
stages, chemotherapy and radiotherapy may offer
sensus about the anatomic definition of cardia
short-​term palliation.
cancer, available data on the frequency of this
cancer vary by 10% to 35% of all gastric tumors Prognosis Because most deaths occur within the
and a corresponding variation in risk factor pat- first couple of months to years following diag-
terns would be expected. It appears, though, that nosis, 5-​year relative survival is a good approx-
the measures of association with putative risk imation of the long-​ term cure rate. Five-​ year
factors are relatively robust with regard to these relative survival varied between 10% and 20%
variations, at least as long as the lower borderline among patients diagnosed during the 1980s in
of what is considered to be the cardia does not the United States and Europe. Based on SEER and
exceed 3 cm below the gastroesophageal junction EUROCARE data, the 5-​ year relative survival
(Lagergren et al, 1999). rates are now 28.8% in the United States (2007)
and 25.1% in Europe (2005–​2007) (De Angelis
CLINICAL SYNOPSIS et al, 2014; Pinheiro et al, 1999; Parkin et al, 2005).
Subgroups Because adenocarcinoma constitutes Progress Despite developments in diagnostic
more than 95% of all stomach neoplasms, we methods, notably the widespread use of endos-
focus solely on this cancer. Other rare types, such copy, and constant efforts to improve surgical
as gastric lymphomas, carcinoids, and leiomyo- treatment, the prognosis has remained grim.
sarcomas, are not covered here. Adenocarcinomas Although the prognosis may, in fact, have
arising in the columnar epithelium of the esopha- changed somewhat for the better, it appears that
gogastric junction (cardia cancer) are histologi- prevention is the most realistic way to reduce the
cally identical to adenocarcinomas in the middle burden of this dreaded disease.
and distal parts of the stomach and are classified
as stomach cancers. DESCRIPTIVE
Symptoms Abdominal pain—​often indistinguish- EPIDEMIOLOGY
able from that caused by peptic ulcer—​anorexia, Notwithstanding that a remarkable spontane-
and weight loss are symptoms characteristic of ous global decline has halved the age-​specific
214

214 Textbook of Cancer Epidemiology

incidence of stomach (gastric) cancer in most is that the Republic of Korea has replaced Japan
Western countries in the past 30–​40  years, this as reportedly having the highest gastric can-
cancer is still the fifth most frequent one world- cer incidence due to significant drop in Japan.
wide, surpassed only by cancer of the lung, breast, The age-​ standardized incidence in Japan has
colorectum, and prostate. Due to its poor prog- declined from 62.1 in 2002 to 45.7 in 2012 per
nosis, stomach cancer ranks number 3 among 100,000 person-​years in men and correspond-
all causes of cancer death. In fact, it accounts ingly from 26.1 to 16.5 per 100,000 person-​years
around 9% of all cancer deaths. More than two-​ in women. This declining trend might be attrib-
thirds of the cases occur in developing countries. utable to declining prevalence of Helicobacter
The worldwide estimates of age adjusted inci- pylori (H.  pylori) infection and a retardation of
dence (17.4 per 100,000 person-​years in men the progression rate of gastric atrophy among
and 7.5 per 100,000 person-​years in women in H.  pylori–​infected persons due to improvement
2012) are about 30%–​40% lower than the values of socioeconomic status in the whole society
estimated in 1985. However, due to the aging of (Kamada et al, 2015). In the United States, stom-
the world’s population and the steep age gradient ach cancer now ranks number 12 among men
in incidence among the elderly, stomach cancer and number 16 among women as far as inci-
continues to claim an increasing number of vic- dence is concerned (American Cancer Society,
tims: 900,000–​1,000,000 per year, corresponding 2015)  with a corresponding 15,540 male and
to a 26% increase between 1985 and 2012. 9,050 female incident cases in 2015. In terms of
deaths, the ranking is number 13 and 12 among
Geographic and Demographic US men and women, respectively, with 6,500 and
Distribution 4,220 deaths.
International comparisons need to be interpreted With few exceptions, the incidence among
cautiously in view of possible differences in the women is approximately half of that among men,
availability of medical services, diagnostic meth- regardless of geographic area, culture, or reli-
ods, and registration practices. The national inci- gion, and approximately half of the overall mag-
dence rates of stomach cancer vary approximately nitude of incidence (Figs. 10-​1, 10-​2). Stomach
10-​fold, with the lowest reliable rates observed cancer is a malignancy of the elderly. Except in
among North Americans (age-​standardized inci- eastern Asia, it is rarely seen before the age of 50.
dence of 5.5 per 100,000 person-​years in men The age-​distribution patterns are similar in most
and 2.7 per 100,000 person-​ years in women, countries, and the 2:1 male/​female ratio persists
2012)  and the highest in East Asia (Torre et  al, throughout all age groups in all countries (Figs.
2015). One noteworthy change in eastern Asia 10-​3, 10-​4).

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 2 4 6 8 10 12 14 16 18
ASR (world) per 100,000 person-years

FIGURE 10-​1  Age-​standardized (to the 2012 world population) incidence rates of stomach cancer among women.
Source: Ferlay et al, 2013.
215

Stomach Cancer 215

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 10 20 30 40 50
ASR (world) per 100,000 person-years

FIGURE 10-​2  Age-​standardized (to the 2012 world population) incidence rates of stomach cancer among men.
Source: Ferlay et al, 2013.

While the risk of stomach cancer seems to and in eastern Europe (Torre et al, 2015; Parkin
covary with socioeconomic conditions, there et al, 2005).
is no clear correlation between national level of Differences in incidence within nations fur-
economic development and national incidence ther add to the variability. There is an increas-
rates, but suspected underreporting may distort ing south-​to-​north gradient in incidence in the
comparisons. For instance, it has previously been Northern Hemisphere. In Sweden, the differ-
suggested that gastric cancer is rare in Africa ence between the lowest and highest incidence
(Holcombe, 1992), but a closer look revealed is approximately twofold (Swedish Oncological
that the incidence in Middle Africa is not very Centres, 1995). A  similar 1.6-​fold, south–​north
low (IARC, 2003). Although the highest rates gradient also exists in China (Wong et al, 1998).
are observed in eastern Asia, low rates (<10 per Likewise, the incidence rates observed in 13 can-
100,000 person-​years) are reported from South cer registries in Italy varied threefold both in
and Southeast Asia (e.g., India, Thailand, and women and in men (Parkin et  al, 1997). In the
the Philippines). High incidence rates also are United States, the incidence is twice as high in
found in tropical Central and South America blacks as in whites (Parkin et al, 1997), and the
New cases per 100,000

250
200
person-years

150
100
50
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 10-​3  Age-​specific incidence rates of stomach cancer among women.

Source: Ferlay et al, 2013.
216

216 Textbook of Cancer Epidemiology

600

New cases per 100,000

500

person-years
400
300
200
100
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 10-​4  Age-​specific incidence rates of stomach cancer among men.

Source: Ferlay et al, 2013.

incidence among Japanese Americans is three by a fixed percentage each year (Hansson et al,
to six times higher than that of US-​born whites 1991). In multivariate modeling, with control
(Kamineni et al, 1999). Immigrant Koreans have for the effects of age, calendar period, and/​or
an incidence that is eightfold higher than that period of birth, this decline is best explained
among whites (Cho et  al, 1996), while the inci- by a marked fall in incidence in successive
dence among Filipino men, regardless of birth- birth cohorts (Hansson et  al, 1991; Aragones
place, is only 60% that of US-​born white men et al, 1997).
(Kamineni et  al, 1999). Another example of Figure 10-​5 depicts the age-​adjusted relative
marked differences within a limited geographic risk of stomach cancer among Swedes in 1960–​84
area comes from Singapore, where the incidence by year of birth, with the risk among those born
rates among men of Malay and Chinese descent in 1876–​84 as a reference. With control for age,
are 8.7 and 29.3 per 100,000 person-​years, respec- the relative risk among Swedish men and women
tively (Parkin et al, 1997). born in 1906–​14 was 0.46 and 0.39, respectively,
When people move between populations while the corresponding figure for men born in
with different risks of stomach cancer, their risk 1956–​64 was no more than 0.07 (Hansson et al,
patterns are usually retained or only slightly 1991). In European countries engaged in World
modified, regardless of their country of origin War II, there are transient rises in risk among the
and country of destination. In the succeeding generations born around the 1940s (Aragones
generation, the rates adjust to that prevail- et  al, 1997), consistent with the importance of
ing in the new environment, but this adapta- exposures, possibly dietary, early in life.
tion appears to be slower for stomach cancer Contrary to the overall declining incidence
than for colorectal and some other cancers, in Western populations, an unexpected increase
particularly among Japanese emigrants to the of noncardia gastric cancer was recently noted in
United States (Kamineni et  al, 1999). Though US whites aged 25 to 39  years (Anderson et  al,
the patterns of risk in relation to migration 2010). This observation was also corroborated by
are complex and defy simple dietary or other a study in Sweden, which showed that the prev-
interpretation, it appears that early-​life expo- alence of gastric corpus atrophy, a precancerous
sures are important for the future risk of gas- lesion for gastric cancer, was increasing among
tric cancer. young Swedes (Song et  al, 2015). If this unfa-
vorable trend can be verified and continues, we
Secular Trends expect to observe a nadir of the declining inci-
The decline in the age-​ specific incidence of dence curve in next 30–​40 years, followed by an
stomach cancer seems to have begun in the early upward trend.
1930s in the Western Hemisphere and thereafter
spread eastward. In the United States, stomach Distribution of Histologic Types
cancer was the leading cause of cancer death in There are several classifications of gastric ade-
1930. The secular trend seems to fit well with a nocarcinoma. Most often used in epidemiologic
log-​linear model, that is, the incidence decreases research is the one proposed by Lauren (1965).
217

Stomach Cancer 217

1.0

0.8

0.6
Relative risk

0.4

0.2

FIGURE 10-​5 Relative risk of stomach cancer by birth

0
1880 1900 1920 1940 1960 year among men and women in Sweden 1960–​84. The
Midyear of birth birth year 1888 was set as the reference. The estimates
were adjusted for attained age.
Males Females
Source: Hansson et al, 1991.

It distinguishes between two main groups:  (1) trends for the two main histologic types when
the intestinal type, with glandular epithelium adjustments were made for age, gender, and local
composed of absorptive cells and globlet cells, geographic differences in incidence (Ekstrom
and (2)  the diffuse type, with poorly differenti- et al, 2000). Although data from the United States
ated small cells in a dissociated, noncohesive have shown divergent secular trends by histolog-
growth pattern. In addition, mixed and unclassi- ical type, the diffuse type also started to decrease
fiable tumors occur. Lauren noted that the intes- since 2000 (Wu et al, 2009). Thus, the descriptive
tinal type was more common among men than epidemiology does not clearly support the notion
among women and it was seen less often in young that the occurrence of one histologic type of
patients than in the elderly. stomach cancer is more environmentally deter-
Some authors have attributed the overall mined than the other. But etiologic heterogene-
decline in the incidence of stomach cancer to a ity among histopathological subtypes of stomach
selective decrease in the intestinal type (Munoz & cancer should be considered in future studies.
Connelly, 1971; Nikulasson et al, 1992; Lauren &
Nevalainen, 1993). This led to the hypothesis that Opposing Secular Trend
the intestinal type, with its variation over time for Cardia Cancer?
and across populations (Coleman et  al, 1993), It appears that the decline in the incidence of gas-
has a predominantly environmental etiology, tric carcinoma overall has abated in the United
while the diffuse type was postulated to be more States (Devesa et  al, 1987; Devesa et  al, 1998).
genetically predetermined. Other investiga- For instance, the incidence among men (age-​
tors, however, have also noted falling incidence adjusted against the 1970 US standard) fell by no
rates for the diffuse type (Sipponen et  al, 1987; more than 0.2 per 100,000 person-​years between
Nikulasson et  al, 1992). There are reports from 1974–​6 and 1992–​4 (Devesa et al, 1998). A closer
Sweden and Mexico of an unchanged ratio of look at the data, however, reveals that the steep
intestinal to diffuse cancers in case series extend- downward trend persisted for distal stomach
ing over decades (Lundegardh et al, 1991; Mohar cancer in white men and women. Among the
et al, 1997). And a study in Sweden, with metic- former, the incidence fell from 5.1 to 3.7 per
ulous case ascertainment and histopathologic 100,000 person-​years. But this decline was bal-
classification of all tumors, showed identical time anced by an increase in the incidence of cardia
218

218 Textbook of Cancer Epidemiology

cancer (from 2.1 to 3.3 per 100,000 person-​years Second, recall bias may be substantial if the study
among white men). is extended beyond first-​degree relatives. Third,
Increasing incidence rates of cardia cancer because having a large number of siblings is an
have been noted in a number of cancer regis- established risk factor for gastric cancer (dis-
ters in Europe and the United States in the past cussed in what follows), and because the num-
20–​30  years (Blot et  al, 1991; Hansson et  al, ber of siblings is an important determinant of the
1993b). This deviating trend coincided with likelihood of having a positive first-​degree family
markedly increasing incidence rates of esoph- history of any disease, there is a clear potential
ageal adenocarcinoma reported from several for confounding by family size.
Western countries (Blot et  al, 1991; Hansson Most studies have not adjusted for number
et al, 1993a). Given that a separate International of siblings. But even with more stringent study
Classification of Diseases code for cardia can- designs, familial aggregation is observed. In an
cer was not introduced until after 1970 in most analysis of 23,386 Swedish twin pairs, with case
countries, and considering the potential for ascertainment in the virtually complete Swedish
misclassification of esophageal adenocarcinoma Cancer Registry, the risk of stomach cancer
as cardiac, there are reasons to be somewhat among twins with a partner who developed
skeptical about the alleged rise in cardia cancer the same cancer was five times as high as it was
incidence. among twins whose partner did not have stom-
Indeed, considerable misclassification of the ach cancer (Ahlbom et al, 1997). In a subsequent,
site within the stomach has been demonstrated in bigger Scandinavian study including 80,309
the otherwise high-​quality Swedish cancer regis- monozygotic and 123,382 same-​ sex dizygotic
tration (Ekstrom et al, 1999b); following careful twin individuals, it was estimated that inherited
classification of all tumors, no increasing trend genes contribute 22% to the risk of gastric can-
could be confirmed for cardia cancer (Ekstrom cer, shared environmental effects contribute 6%,
et al, 2000). Some other studies have also failed to and nonshared environmental factors make up
verify any upward trend, and even in the United the remaining 72% of the risk (Mucci et al, 2016).
States, the trend seems to have leveled off in the Therefore, studies on twins predict the involve-
1990s (Devesa et  al, 1998), and even started to ment of major environmental factors plus minor
decrease since 2000 (Wu et al, 2009). Regardless genetic components.
of whether the incidence curve for cardia can-
cer is flat or turning up, it is quite clear that it Inherited Susceptibility
looks different from the descending one for distal Germline Mutations
stomach cancer. An aggregation of two or more stomach cancers
in the same family is noted in about 10% of all
GENETIC AND MOLECULAR stomach cancer cases. Among them, a number
EPIDEMIOLOGY of syndromes can be identified:  (1) hereditary
diffuse gastric cancer (HDGC—​ requiring two
Familial Occurrence or more documented cases of diffuse stomach
Familial clustering of stomach cancer occurrence cancer in first-​/​second-​degree relatives, with at
has long been noticed. Historically, stomach can- least one diagnosed before the age of 50; or three
cer is known for its unique, perhaps genetically or more cases of documented diffuse stomach
determined, clustering in the family of Napoleon cancer in first-​ /s​econd-​
degree relatives, inde-
Bonaparte. More recently, familial aggregation pendently of age) (Caldas et al, 1999); (2) famil-
of stomach cancer has been reported in the epi- ial diffuse gastric cancer (FDGC—​families with
demiologic literature. Typically, a 50% to 130% aggregation of stomach cancer and an index case
excess risk was observed among subjects with with diffuse stomach cancer, but not fulfilling the
a positive family history. Interestingly, a some- criteria for HDGC, for instance due to unknown
what higher risk was noted when the mother histological type of the related cases); (3) familial
was affected. Most epidemiologic studies have intestinal gastric cancer (FIGC—​families with a
used an abbreviated family history approach and similar aggregation as in HDGC, but with intes-
must, therefore, be cautiously interpreted. First, tinal type histology) (Shinmura et al, 1999); and
many laypeople refer to the whole abdomen as (4) familial gastric cancer (FGC—​familial aggre-
the stomach, so the specificity of positive reports gation of stomach cancer but with unknown
of stomach cancer among relatives may be low. histopathology).
219

Stomach Cancer 219

Germline truncating mutations in the gene genes MTX1, TGFBR2, PKLR, GSTP1, CASP8,
for the cell–​ cell adhesion protein E-​ cadherin and TNF were also found highly significant
(CDH1) were first discovered in three Maori (Mocellin et al, 2015).
families with autosomal dominant diffuse stom-
ach cancer but have since been detected in an Somatic Mutations
increasing proportion of families with aggrega- Unlike in the adenoma-​ colorectal cancer
tion of stomach cancer, notably HDGC (Guilford sequence, no clear linear accumulation of genetic
et  al, 1998; Gayther et  al, 1998). In a review of defects has been identified in gastric carcino-
the accumulated literature to 2006 (Oliveira genesis (Vauhkonen et al, 2006). One important
et al, 2006), HDGC accounted for 26.9% of 439 reason for the rather confusing overall picture is
screened families with any of the syndromes the fact that the histological subtypes of stomach
described previously. Another 23.7% had FDGC, cancer appear to arise through at least partly dis-
8.7% had FIGC, and 40.8% had FGC. Germline parate molecular pathways. Moreover, Lauren’s
CDH1 mutations were found in 36.4% of families intestinal type may develop along three subpath-
with HDGC and in 12.5% of families with FDGC. ways (Tahara, 2004): (1) an intestinal metaplasia
An additional small proportion has other (IM) → adenoma → carcinoma sequence; (2) an
familial cancer syndromes, for instance heredi- IM → carcinoma sequence; and (3) de novo. For
tary nonpolyposis colorectal cancer (HNPCC) the diffuse histological type, no premalignant
(Aarnio et al, 1997) and Li-​Fraumeni syndrome lesion is known. However, most—​ if not all—​
(Varley et  al, 1995), which are both linked to stomach cancers, regardless of histological type,
an increased risk of stomach cancer. A  TP53-​ are preceded by H.  pylori infection with entail-
truncating germline mutation was described in a ing gastritis. The use of partly different morpho-
family with characteristics of both HDGC and Li-​ logical classifications further compounds the
Fraumeni syndrome (Kim et al, 2004). However, interpretation of existing data. But even among
in about two-​thirds of HDGC families, a large stomach cancers with apparently identical histo-
proportion of FDGC and in the majority of FGC pathological morphology there is substantial het-
families, cancer susceptibility is caused by pres- erogeneity in expression of molecular markers.
ently unknown genetic defects. Quick progress In general, the number of genes in which
in sequencing technology has brought hope to mutations (i.e., base substitutions, insertions, or
cover this gap, including the finding of CTNNA1-​ deletions of a small number of base pairs) are
truncating germline mutation (Majewski et  al, regularly found is low in stomach cancer. As in
2013), and germline mutations in the other three several other cancers, however, mutation of the
genes (INSR, FBXO24, and DOT1L) (Donner cellular gatekeeper and tumor suppressor gene
et al, 2015) that are associated with HDGC risk, TP53—​largely missense point mutation—​is com-
as well as MAP3K6 germline mutations that are mon. The gene is usually inactivated through the
associated with FGC risk (Gaston et al, 2014). classic two-​hit mechanism, with loss of hetero-
zygosity (LOH) and mutation of the remaining
Polymorphisms allele. The frequency of TP53 mutations in intesti-
The literature on genetic polymorphisms and nal type stomach cancers is equally high (approx-
stomach cancer risk is limited by a common imately 40%) regardless of whether the tumors
lack of appropriate control of potential sources are early or advanced. Therefore, the mutations
of bias; few studies are population based, and are thought to be early events in the development
the sample sizes are often insufficient even for of this histological subtype. TP53 mutations are
the statistical verification of moderate main rare in early diffuse type cancers, but when the
effects, let  alone gene–​ environment interac- tumors are advanced the mutations occur with
tions. Besides, information on exposure to rel- the same frequency as in the intestinal type
evant cofactors such as H. pylori infection, diet, (Tamura, 2006).
and smoking is often lacking. Earlier studies Similarly, mutation of the CDH1/​E-​cadherin
usually employed a candidate gene approach, gene, involved in calcium-​dependent cell-​to-​cell
while several genome-​wide association studies adhesion and apparently important for morpho-
(GWAS) have been conducted. The main find- genesis, is rare in early but common (approxi-
ings from GWAS are summarized in Table 10-​1. mately 50%) in advanced diffuse type stomach
In a comprehensive meta-​analysis of 824 studies, cancer. This mutation is always uncommon in
besides those found in GWAS, several SNPs in the intestinal type, though, and when it occurs in
 20
TABLE 10-​1   ASSOCIATION BET WEEN SELECT SNPS AND RISK FOR STOMACH CANCER FROM PREVIOUSLY PUBLISHED
GENOME-WIDE ASSOCIATION STUDIES

First author Population SNP Location Mapped Total gastric Cardia Non-​cardia
year gene(s) cancer

OR P value OR P value OR P value

Sakamoto H* Japan/​Korea rs2976392 chr8 PSCA 1.62 1.1 × 10–​9

2008

Abnet CC 2010 China rs3781264 chr10 PLCE1 1.36 3.8 × 10–​9 1.60 1.1 × 10–​13 1.11 0.14

rs4072037 chr1 MUC1 0.75 4.2 × 10 –​7

0.75 9.5 × 0–​5
0.72 5.7 × 10–​5

Shi Y 2011 China rs9841504 chr3 ZBTB20 0.76 1.7 × 10–​9

rs13361707 chr5 PRKAA1 1.41 7.6 ×10–​29

Jin G 2012 China rs2494938 chr6 LRFN2 1.18 4.9 ×10–​9

Hu N 2015 China/​Korea rs10074991 chr5 PRKAA1 0.80 4.8 × 10–​26 0.83 7.4 × 10–​12 0.77 2.4 × 10–​23

rs2294693 chr6 UNC5CL TSPO2 1.14 7.2 × 10–​8 1.08 1.5 × 10–​2 1.18 2.5 × 10–​8

rs4072037 chr1 MUC1 0.76 6.6 × 10–​8 0.74 5.0 × 10–​5 0.76 1.1 × 10–​5

Wang Z 2015 China rs80142782 chr1 ASH1L 0.62 1.7 × 10–​19

rs4072037 chr1 MUC1 0.74 6.3 × 10–​17

rs7712641 chr5 NA 0.84 1.2 × 10–​11

rs2294008 chr8 PSCA 1.20 6.0 × 10–​11

* Diffuse-​type stomach cancer.

21

Stomach Cancer 221

tumors with mixed histology (both intestinal and between H. pylori gastritis and CpG island meth-
diffuse), it is only seen in the diffuse component. ylation of genes is unclear, the microorganism
These observations have fostered the hypoth- may act as a trigger factor for hyperplasia in
esis that the differentiated intestinal type may hTERT positive “stem cells” in intestinal metapla-
sometimes become “dedifferentiated” and turn sia (Tahara, 2004).
into a diffuse type tumor, and that acquisition of Epigenetic methylation-​ associated inactiva-
CDH1/​E-​cadherin mutations may be involved in tion of the mismatch repair gene mutL homolog
this transition (Tamura, 2006). 1 colon cancer (hMLH1) is of particular interest.
Point mutations in the KRAS oncogene have It is a potent trigger of microsatellite instabil-
been observed in up to one-​fifth of stomach can- ity (MSI), defined as the presence of replication
cers (Vauhkonen et al, 2006), mainly in the intes- errors in simple tandemly repeated stretches of
tinal type (Smith et al, 2006). Other oncogenes or DNA (microsatellites). MSI, classified as either
proto-​oncogenes may be amplified; ERBB2/​HER-​ high-​frequency (MSI-​H) or low-​frequency
2 (c-​erbB2) is amplified in approximately 20% of (MSI-​L), is found in 15%–​50% of stomach can-
intestinal type cancers but not in cancers of the cers, more often in the intestinal type than in the
diffuse type, while MET (c-​met) and FGFR2 (K-​ diffuse. Conversely, among the microsatellite sta-
sam) are each preferentially amplified in about ble (MSS) tumors, that is, replication error nega-
one-​third of diffuse type cancers. tive, the diffuse histological type is dominating. In
Deficient regulation of gene expression may a study among early intestinal cancers, 20% of the
also be caused by epigenetic changes. Tumor sup- tumors were classified as MSI-​H (Ohmura et al,
pressor genes may be silenced by hypermethyl- 2000), while another study among early diffuse
ation of the promoter CpG islands. Accordingly, cancers found no evidence of MSI at all (Tamura
the promoter regions of DCC and CDKN2A et al, 2001). Up to 90% of MSI-​H tumors are asso-
(p16) exhibit frequent methylation. DAP-​kinase ciated with hMLH1 promoter hypermethylation.
promoter methylation is more frequent in the In MSI-​ H stomach cancers several genes
diffuse than in the intestinal type. RASSF1A, undergo alterations in the repetitive segments of
RUNX3, and TCLC1 are other tumor suppressor the coding regions leading to reading frame muta-
genes that are frequently silenced through hyper- tions, for instance transforming growth factor β
methylation in stomach cancer (Tamura, 2006). receptor II gene (TGFβRII—​63%–​92%), insulin-​
In fact, increasing numbers of genes involved in like growth factor II receptor gene (IGFIIR—​5%–​
cell-​cycle control, DNA repair, cell–​cell adhesion, 25%), proapoptotic BAX gene (15%–​68%), DNA
apoptosis, and angiogenesis have been described repair genes hMSH6 (22%) and hMSH3 (38%),
as being silenced by hypermethylation; meth- and transcription factor gene E2F-​4 (37%–​61%)
ylation of CDH1/​E-​cadherin, associated with (Vauhkonen et al, 2006). The MSI-​L phenotype,
decreased expression, is observed in more than also linked to the intestinal type of stomach can-
50% of early stage diffuse cancers and is also cer, is associated with the LOH phenotype. Target
observable in surrounding noncancerous gas- gene mutations are infrequent in MSI-​L tumors.
tric epithelia. Thus, the epigenetic inactivation Frequent occurrence of LOH in a tumor has
of CDH1/​E-​cadherin via promoter methylation been suggested to indicate chromosomal insta-
may play a major role in the development of the bility (CIN). LOH is thought to represent the
diffuse type of stomach cancer. The present epi- second inactivating hit in tumor suppressor
genetic data suggest that much, if not most, of genes, characterizing the so-​ called suppressor
the decrease (or loss) of gene expression may be pathway of carcinogenesis. In stomach cancer,
explained by promoter hypermethylation. LOH is mostly found on chromosomes 2q, 4p,
In some cases promoter hypermethylation 5q, 6p, 7q, 11q, 14q, 17p, 18q, and 21q. It can be
leads to activation of the gene. This is true for the detected in up to 80% of all tumors. It appears
human telomerase transcriptase gene (hTERT), that the frequency is similar in intestinal and dif-
an important determinant of telomerase activity. fuse cancers, but few studies have addressed the
Reactivation of telomerase activity is necessary occurrence of LOH in the diffuse histological
for cell immortality and is normally repressed in type. The LOH rate in TP53 is reportedly 26%–​
gastric cells. Such reactivation is seen in virtu- 83%, predominantly occurring in the intestinal
ally all gastric cancers (Tahara et al, 1995). Over histological type.
50% of intestinal metaplasias express low lev- Development of sequencing technology has
els of telomerase activity. While the association added a lot to our understanding of mutation
2

222 Textbook of Cancer Epidemiology

landscape in stomach cancer. A landmark study prevalence, and thus between men and women,
from The Cancer Genome Atlas (TCGA) proj- but within sexes the variation between American
ect proposed a novel molecular classification and European data is surprisingly small; thus,
for stomach cancer, including tumors positive PAR among men varied between 21.5% and
for Epstein Barr virus, microsatellite unstable 28.6% and among women between 11% and 14%
tumors, genomically stable tumors, and tumors in three studies emanating from the United States
with chromosomal instability (Cancer Genome and Europe (Chao et al, 2002; Engel et al, 2003;
Atlas Research Network, 2014). This classifi- Gonzalez et al, 2003).
cation system can be used for patient strati-
fication for proper targeted therapies, as well Alcohol
as for searching etiologic factors by molecular The most authoritative review of the literature
subtypes. published up until the mid-​2000s (World Cancer
Research Fund—​American Institute of Cancer
RISK FACTORS Research, 2007) noted that the bulk of evidence
weighed against the possibility of a substan-
Tobacco tial effect of alcohol consumption on the risk of
A relationship between smoking and risk of gas- stomach cancer. Although the association might
tric cancer is well established (Tredaniel et  al, be stronger for cardia cancer, more recent studies
1997; Ladeiras-​ Lopes et  al, 2008). The excess suggest no important differences between cardia
risk among current smokers is 1.5-​to 2.5-​fold. and noncardia stomach cancers. A meta-​analysis
Although most earlier studies failed to confirm in 2001 (Bagnardi et al, 2001) arrived at a sum-
a dose-​response relationship, more recent inves- mary relative risk estimate of 1.15 (95% CI 1.09–​
tigations have generally shown increasing risks 1.22) and 1.32 (95% CI 1.18–​1.49) for an alcohol
with higher doses and/​or longer duration of cig- intake of 50 and 100 g/​day, respectively, relative
arette smoking (Gonzalez et  al, 2003; Koizumi to no intake.
et al, 2004; Ladeiras-​Lopes et al, 2008). It appears Two cohort studies observed a link between
that the risk returns to baseline relatively soon alcohol consumption and the development of
after quitting smoking (Chow et al, 1999), but in gastric cancer among men (Duell et  al, 2011;
a pooled analysis of two Japanese cohorts, a sig- Everatt et al, 2012). A Swedish prospective study
nificant risk elevation remained for up to 14 years among women found no overall association
after cessation (Koizumi et al, 2004). of alcohol intake with stomach cancer risk but
While some studies suggest that smoking noted that frequent consumption of medium-​
is more strongly related to cardia cancer risk strong/​strong beer, which contains the animal
(Lagergren et al, 2000; Mao et al, 2002; Gonzalez carcinogen N-​nitrosodimethylamine (NDMA),
et al, 2003; Nomura et al, 2012), others indicate was associated with a statistically significant
that the link with distal stomach cancer is not twofold risk elevation (Larsson et  al, 2007b).
appreciably weaker, and in Japan it might even The link to beer intake, however, was not in
be stronger (Koizumi et al, 2004). In a Swedish accord with another Scandinavian cohort study
study, the association between smoking and (Barstad et  al, 2005), which instead showed a
stomach cancer risk was largely confined to significant 40% risk reduction per glass of wine
subjects with a low intake of fresh fruits and drunk per day, a weak and nonsignificant trend
vegetables (Ye et  al, 1999). There is also some toward positive association of stomach cancer
evidence suggesting a multiplicative interaction with hard liquor consumption, no association
between cigarette smoking and CagA positive with beer consumption, and no association with
H. pylori infection, leading to 17-​fold increased overall alcohol intake. Published data from pro-
risks among people exposed to both risk factors, spective cohort studies on specific relationships
compared to those exposed to neither (Brenner between different types of alcoholic beverages
et al, 2002). and stomach cancer risk was recently reviewed
Population attributable risk percent (PAR) (Fang et  al, 2015), with a summarized relative
expresses the percent of the studied outcome dis- risk estimate of 1.21 (95% CI 1.02–​1.43) and 1.22
ease that can be attributed to the exposure under (95% CI 1.05–​1.43), respectively, for the higher
study, or in other words, the percent of all cases consumption of beer and liquor compared to
that might be prevented by eliminating this risk no intake, but no such positive association was
factor. PAR for smoking varies with the exposure observed for wine.
23

Stomach Cancer 223

Tea and Coffee of phase I  metabolic enzymes (e.g., P450 cyto-

No causal association is established between tea chromes), which may create reactive oxidative
or coffee and risk of stomach cancer. However, molecules, or induction of phase II enzymes (e.g.,
frequent intake of green tea, hypothesized to have glutathione transferase), which detoxify residual
a protective effect by virtue of its polyphenol con- electrophilic metabolites generated by the phase
tent, was shown to be inversely associated, albeit I  enzymes (flavonoids and allium compounds,
sometimes without a clear dose-​effect pattern, isothiocyanates, glucosinolates, indoles, thiocya-
with risk of stomach cancer in the majority—​but nates, and some carotenoids).
not all—​ of 10 published case-​ control studies. In a meta-​analysis (Riboli & Norat, 2003),
Of eight prospective studies, however, only one however, the protective effect seemed to be
(Sasazuki et  al, 2004)  found an inverse associa- weaker in cohort than in case-​control studies,
tion, and it was confined to distal stomach can- suggesting that in the latter recall bias might have
cer among women. Coffee consumption has been pushed the relative risk estimates away from the
unrelated to stomach cancer risk in most stud- null value. A systematic review of 24 cohort stud-
ies (World Cancer Research Fund—​ American ies also suggested only a 4%–​10% reduction of
Institute of Cancer Research, 2007), but four stomach cancer risk associated with high intake
(Jacobsen et al, 1986; Galanis et al, 1998; Larsson of fruits and vegetables (Wang et al, 2014).
et al, 2006b; Ren et al, 2010) out of 13 prospective Even though the estimation of portion size
studies showed an excess incidence of border- and frequency of consumption of a wide range
line statistical significance among highly exposed of vegetables is difficult and nondifferential mis-
individuals, with relative risks in the highest ver- classification may bias the relative risk estimates
sus lowest exposure category ranging from 1.46 toward the null value, it is reasonable to assume
to 1.86. that cohort studies with increasingly sophisti-
cated dietary assessments, are less affected by
Diet such bias compared to earlier studies. Therefore,
Fruits, Vegetables, and Antioxidants previous research may have overestimated the
Until recently, the most consistent nutritional protection conferred by these plant foods.
epidemiology finding in relation to stomach can-
cer has been inverse associations with fruit and Vitamin C
vegetable intake. In 2007, an international expert Whereas there is almost total consensus among
panel at World Cancer Research Fund–​American case-​control studies that vitamin C intake is
Institute for Cancer Research concluded that strongly protective, only one (Botterweck et  al,
high intake of nonstarchy vegetables and allium 2000) out of five prospective studies (Zheng et al,
vegetables probably reduces the risk of stomach 1995; Botterweck et al, 2000; Nouraie et al, 2005;
cancer (World Cancer Research Fund–​American Jenab et al, 2006b; Epplein et al, 2010) reported
Institute for Cancer Research, 2007). A  similar a significant inverse association between esti-
conclusion was also drawn with regard to high mated vitamin C intake and stomach cancer.
fruit intake. The protective effects of these plant The summary of relative risk in a meta-​analysis,
foods have been attributed to their high concen- however, was still statistically significant (relative
trations of antioxidant substances, notably carot- risk among subjects with the highest intake, rel-
enoids, vitamin C, and vitamin E compounds. ative to those with the lowest, was 0.89 (95% CI
Additionally, there are many nonessential 0.85–​0.93)).
bioactive compounds with anticarcinogenic
properties in fruits and vegetables, for exam- Vitamin E
ple, phenolics (Steinmetz & Potter, 1996). These Vitamin E (tocopherol), another important anti-
substances could conceivably act through sev- oxidant in plant foods, has been investigated
eral mechanisms. The most important ones with regard to its relationship with stomach can-
being (1)  scavenging of potentially mutagenic cer risk in at least 18 case-​control studies. Close
free radicals (vitamins C and E, beta-​carotene, to half of these reported a statistically significant
and some phenolic compounds such as flavo- inverse association while the others were una-
noids); (2)  inhibition of carcinogenic nitrosa- ble to statistically confirm any relationship at all.
mine formation in the stomach (vitamins C and Among five prospective studies that related esti-
E); (3) regulation of cell differentiation and DNA mated dietary vitamin E intake to stomach can-
synthesis (carotenoids and folate); (4) inhibition cer risk, only one—​conducted among Finnish
24

224 Textbook of Cancer Epidemiology

smokers (Nouraie et al, 2005)—​showed a signif- evidence of an inverse relationship (Stahelin et al,
icantly reduced risk of noncardia stomach can- 1991; Taylor et al, 2003; Yuan et al, 2004; Nouraie
cer among individuals with the highest intake. et  al, 2005; Jenab et  al, 2006a) but only two
This study also noted an increased risk for can- yielded statistically significant results (Stahelin
cer of the gastric cardia. The other four (Chyou et al, 1991; Jenab et al, 2006a). The summary of
et  al, 1990; Zheng et  al, 1995; Botterweck et  al, relative risk (highest versus lowest category) in
2000; Carman et al, 2009) found no association. these seven studies was 0.80 (95% CI 0.66–​0.96)
However, in one cohort study, supplemental vita- (Larsson, 2006).
min E was associated with a reduced risk of non- With regard to beta-​carotene intake (provita-
cardia stomach cancer (Carman et al, 2009). min A, mainly coming from plant foods and with
Six prospective studies that proceeded from additional antioxidative actions), case-​ control
predisease blood levels of tocopherols yielded studies have generally shown inverse associations
mixed results; a large European study reported with stomach cancer risk, but the prospective
a strong and statistically significant inverse rela- studies have been less persuasive. After exclud-
tionship with stomach cancer risk, albeit seem- ing one outlier, the summary of relative risk in
ingly limited to the diffuse histologic type (Jenab a meta-​analysis related to predisease blood levels
et  al, 2006a), while a Chinese study showed a of beta-​carotene was 0.82 (95% CI 0.60–​1.10).
positive association with noncardia cancer risk
but no relationship with cardia cancer (Taylor Intervention Trials
et al, 2003). The other four studies found no asso- Not even randomized intervention trials have
ciation (Stahelin et al, 1991; Nomura et al, 1995; been able to provide an unambiguous answer
Yuan et al, 2004; Ito et al, 2005). Thus, the effect regarding the protective effect of the antioxidative
of vitamin E on risk of stomach cancer remains vitamins in plant foods (Mayne et al, 2012). Two
uncertain. such studies argue in favor of a protective effect; a
Chinese study among subjects who were likely to
Vitamin A be vitamin deficient, showed a reduced incidence
At least 28 case-​control studies have addressed of gastric cancer mortality after administra-
the relationship between intake of total vita- tion of a combination of beta-​carotene, vitamin
min A  (retinol and provitamin A  carotenoids) E, and selenium (Blot et  al, 1993). In the other
and risk of stomach cancer. The overwhelm- study, carried out in South America (Correa
ing majority of them found an inverse associ- et al, 2000), treatment with either beta-​carotene
ation, statistically significant in six of these. Of or ascorbic acid significantly increased the rates
13 case-​control studies that specifically exam- of regression of atrophic gastritis and intestinal
ined retinol—​mainly present in foods of animal metaplasia. A  third randomized, double-​blind,
origin—​most reported either a null finding or a placebo-​controlled trial from China was una-
trend toward a positive association. Five prospec- ble to statistically verify any beneficial effects of
tive studies have related estimated dietary intake beta-​carotene on cancer incidence or progression
of retinol to stomach cancer risk: one observed of precancerous changes in the gastric mucosa
a significant inverse association (Larsson et  al, but reported a significantly reduced incidence of
2007a), two found a nonsignificant but similar all gastrointestinal cancers combined, and more
trend (Zheng et al, 1995; Nouraie et al, 2005) (in reversions of gastric dysplasia, with folic acid in
one study this trend was significant for cardia combination with vitamin B12 (Zhu et al, 2003).
cancer: Nouraie et al, 2005), while the other two Two other randomized intervention studies,
suggested a positive association (Botterweck conducted among Finnish male smokers (Malila
et  al, 2000; Miyazaki et  al, 2012). The pooled et  al, 2002; Virtamo et  al, 2003)  and American
estimate in a meta-​analysis was however sig- male physicians (Hennekens et al, 1996), respec-
nificant or marginally significant; relative risk tively, showed no reduction in stomach can-
among subjects with the highest intake com- cer incidence during or after supplementation
pared to those with the lowest was 0.66 (95% CI with either beta-​carotene or alpha-​tocopherol,
0.52–​0.84) for total vitamin A  intake and 0.94 the most active form of vitamin E.  Moreover,
(95% CI 0.87–​1.03) for retinol intake. two additional randomized Chinese interven-
Among seven prospective studies that inves- tion studies observed no reductions in stomach
tigated stomach cancer risk in relation to pre- cancer incidence or mortality after daily supple-
disease blood levels of retinol, all but two found mentation with 14 vitamins and 12 minerals for
25

Stomach Cancer 225

6 years (Li et al, 1993; Dawsey et al, 1994) or in abundant case-​control and cohort study data on
the prevalence of precancerous gastric lesions by intake of salt or salty foods and risk of stomach
a combination of vitamin C, vitamin E, and sele- cancer. Although the results are somewhat diver-
nium every second day for 7.3  years (You et  al, gent, the bulk of evidence weighs toward a posi-
2006). A  similar lack of effect on precancerous tive association, albeit not particularly strong.
gastric lesions by treatment with a combination However, confounding is a major concern; in some
of vitamin C, vitamin E, and beta-​carotene was of the studied populations, consumption of salted
reported from a randomized, double-​blind che- foods may have correlated inversely with socioec-
moprevention trial in a Venezuelan high-​ risk onomic status, access to refrigeration, consump-
population (Plummer et al, 2007). tion of fruits and vegetables, and positively with
the prevalence of H.  pylori infection. Moreover,
Other Plant Foods salted foods tend to contain significant amounts of
Chili pepper, although generally rich in antioxi- N-​nitroso compounds (NOCs), which may be the
dative vitamins, has been linked to a moderately true culprits. The relative risk estimates in cohort
increased stomach cancer risk, allegedly due to and case-​control studies are mostly in the range
its capsaicin content (Lopez-​Carrillo et al, 2003). where undetected confounding might well explain
Soy foods, rich in presumably anticarcinogenic the association. It should also be noted that there is
isoflavones, were found to confer protection no laboratory evidence that salt per se is a carcino-
according to several studies and a meta-​analysis gen for any site of the body (Cohen & Roe, 1997),
(Nagata et al, 2002; Wu et al, 2000) but appeared although salt might act as promoter or interact
to act as a moderately strong risk factor (26% with H. pylori infection to promote gastric carcin-
increased odds ratio according to the pooled ogenesis (D’Elia et al, 2014).
analysis) when the product was fermented (Wu
et al, 2000). Possible confounding from salt, fruit/​ N-​nitroso Compounds
vegetables, and other dietary factors precludes N-​nitroso compounds are carcinogenic in multi-
confident causal inference, though. ple organs in at least 40 animal species. Humans
are exposed to NOCs from diet (processed meats,
Dietary Fibers smoked preserved foods, pickled and salty pre-
Several investigators found a decreased risk of served foods, and foods dried at high tempera-
stomach cancer among people with a high con- tures such as the constituents of beer, whisky,
sumption of fiber. In a meta-​analysis (Zhang et al, and dried milk), tobacco smoke, and other envi-
2013), a summary odds ratio of 0.58 (95% CI ronmental sources. But a large proportion of
0.49–​0.67) was reported for gastric cancer when NOCs (typically more than 50%) comes from
comparing the highest to the lowest dietary fiber endogenous synthesis. Of 11 case-​control studies
intake. The inverse association was consistent addressing the possible association between esti-
across geographic area, fiber source, fiber type, mated nitrite exposure and stomach cancer risk,
histological subtype, and anatomic subtype of seven showed relative risks (highest versus low-
gastric cancer. However, significant heterogeneity est exposure category) above 1 (range 1.2 to 1.7),
was found by study design; the only two cohort and the risk elevations were statistically signifi-
studies reported nonsignificant reduction of risk cant in five. Three prospective studies showed
(summary odds ratio 0.89, 95% CI 0.68–​1.17). mixed results, but none of the estimates were sta-
Such heterogeneity by study design might be tistically significant (van Loon et al, 1998a; Knekt
explained partly by recall bias in case-​control et al, 1999; Keszei et al, 2013).
studies. On the other hand, residual confounding All but one of six case-​control studies that
by socioeconomic status might distort the associ- estimated NOC intake in relation to stomach can-
ation toward null. cer reported relative risks above 1 (range 1.1–​7.0)
and three of them yielded statistically significant
Salt results. Of four prospective studies that esti-
Most textbooks list salt intake as an established mated NOC intake, three were negative (Knekt
risk factor for stomach cancer. Ecological studies et al, 1999; Jakszyn et al, 2006; Keszei et al, 2013),
provide support for a relatively strong correlation while a Swedish prospective study among women
between urinary salt excretion and stomach can- showed a positive association (relative risk among
cer mortality (Kono & Hirohata, 1996; Joossens highest versus lowest exposure category 1.96, 95%
et  al, 1996; Tsugane, 2005). Further, there are CI 1.08–​3.58) (Larsson et al, 2006a).
26

226 Textbook of Cancer Epidemiology

Proceeding from the estimated intake of iron and red meat consumption, respectively. Because
from meat, an index for endogenous NOC expo- the results from previous epidemiological studies
sure was determined for each participant in the have been mixed, the relationships with red meat
European Prospective Investigation into Cancer and total meat intake need to be further explored
and Nutrition (EPIC) study (Jakszyn et al, 2006); before they are accepted as causal.
when this measure was modeled with stomach
cancer occurrence as the outcome, a statistically Total Energy and Fat
significant association emerged with noncardia There is no consistent data regarding the rela-
stomach cancer (relative risk associated with a 40 tionship between total energy intake and the
μg/​day increase in endogenous NOC exposure risk of stomach cancer. A weak positive associa-
was 1.42, 95% CI 1.14–​1.78) but not with cardia tion has been noted for total intake, as well as an
cancer. But a Dutch study could not confirm this inverse association for total energy intake during
finding (Keszei et  al, 2013). Thus, the epidemi- adolescence.
ological literature has been unable to unequiv- Fat intake has been positively linked to stom-
ocally confirm a link between nitrite or NOC ach cancer risk. In some instances the excess risk
exposure and risk of gastric cancer. was confined to persons consuming large amounts
of animal fat or saturated lipids and cholesterol,
Meat Intake whereas those on a diet rich in vegetable lipids or
While meat consumption has been associated polyunsaturated lipids had an up to 70% reduced
with increased risks of cancer of the colorectum, risk. One can only speculate about the extent to
the epidemiologic evidence for a relationship which residual confounding from fruit and vegeta-
with stomach cancer risk has so far been insuffi- ble intake might have contributed to these findings.
cient. However, more recent cohort studies have
reported up to twofold risk elevations among Trace Elements
subjects in the highest intake categories, relative In animal and human studies, iron deficiency has
to those in the lowest (Ngoan et al, 2002; van den been found to accelerate H. pylori–​induced car-
Brandt et al, 2003; Gonzalez et al, 2006; Larsson cinogenesis (Noto et al, 2013). Two cohort stud-
et  al, 2006a). These and other studies were ies also support this notion by measuring serum
included in a meta-​analysis that encompassed biomarkers of body iron load (Cook et al, 2012;
six prospective cohort studies and nine case-​ Fonseca-​Nunes et al, 2015). Selenium appears to
control studies (Larsson et  al, 2006c). The esti- have the potential for inhibiting the development
mated summary relative risks of stomach cancer of several tumors, but its intake is impossible to
for an increase in processed meat consumption assess through dietary questionnaires in epide-
of 30 g/​day—​approximately half of an average miologic studies because the selenium content of
serving—​were 1.15 (95% CI 1.04–​1.27) for the a specific food depends on the selenium content
cohort studies and 1.38 (95% CI 1.19–​1.60) for of the soil where it was produced. Toenail sele-
the case-​control studies. nium levels were inversely associated with stom-
In three cohort and four case-​control stud- ach cancer risk in a Dutch prospective study.
ies that examined the association between bacon Four prospective studies using determinations
consumption and stomach cancer, the summary of serum levels of selenium reached different
relative risk was 1.37 (95% CI 1.17–​1.61) for conclusions; one reported an inverse association
highest versus lowest intake. Thus, high intake (Knekt et  al, 1990)  while others found no rela-
of processed meat should be added to the list of tionship (e.g., You et al, 2000). The possible link
known—​but moderately strong—​risk factors for to calcium intake has been addressed in a few
stomach cancer. In the European EPIC cohort, studies. Their conflicting results are most likely
the association with processed meat was con- an indication of no overall association.
fined to noncardia stomach cancer; for every 50
grams per day increase in processed meat intake, Frying/​Grilling
the noncardia stomach cancer risk increased Heterocyclic amines formed when fish or meat
2.45-​fold (95% CI 1.43–​4.21) (Gonzalez et  al, is fried are mutagenic in experimental studies
2006). Positive associations were also noted for (IARC, 1993). Polycyclic aromatic hydrocarbons
nonprocessed meat, with significant 3.52-​and (PAHs) are formed during incomplete combus-
1.73-​fold increases in noncardia stomach can- tion of organic material, for instance when food
cer risk per 50 grams/​day increase in total meat is smoked or fried over an open fire. Many PAHs
27

Stomach Cancer 227

are regarded as carcinogenic to humans. Some stomach cancer risk may be still underestimated.
investigators, but not all, found a link between In studies that restricted the outcome to noncar-
high consumption of hard-​fried food and the dia stomach cancer and that took measures to
risk of stomach cancer. In a case-​control study overcome the misclassification of exposure due to
from Uruguay, the subjects most exposed to 2-​ spontaneous disappearance of the microorgan-
amino-​1-​methyl-​6-​phenylimidazo(4,5-​b)pyr- ism, the relative risk linked to the infection was
idine had an almost fourfold increased risk of 20-​fold or more (Ekstrom et  al, 2001; Brenner
stomach cancer compared to those in the lowest et al, 2004; Persson et al, 2011). The sizable risk
exposure category (De Stefani et  al, 1998). The associated with H.  pylori infection was further
effect of this heterocyclic amine interacted with clearly demonstrated in a Japanese cohort study,
the effect of nitrosodimethylamine in a mul- where 2.9% of 1,246 infected individuals devel-
tiplicative manner, yielding an almost 13-​fold oped gastric cancer during 7.8 years of follow-​up,
increased risk among persons highly exposed while no cancer was observed among 280 nonin-
to both substances. However, studies addressing fected cohort members (Uemura et al, 2001).
the relationship between exposure to PAH and Although the ultimate proof of causality is still
gastric cancer risk have been negative, with one missing, a growing number of randomized trials
exception. Interpretation is complicated by the have either shown trends toward reduced gastric
fact that grilling over an open fire seems to be cancer incidence or indications of slowing pro-
closely linked to a high intake of fruits and veg- gression of precancerous lesions after H.  pylori
etables in Western settings. eradication, thus gradually adding to our confi-
dence in a causal inference. In a meta-​analysis of
Pickled Food six randomized controlled trials, a summarized
Studies from China, Japan, Hawaii, and Spain relative risk of 0.66 for H. pylori eradication was
reported an increased risk of gastric and/​or car- reported; the number needed to treat for pre-
dia cancer among frequent consumers of pickled vention of stomach cancer ranges from 15 for
food. Extracts from pickled vegetables in China Chinese men and 245 for US women (Gaston
have been shown to be mutagenic and contain N-​ et al, 2014). An oral recombinant H. pylori vac-
nitroso compounds and benzopyrenes. cine to reduce the infection among children also
raises the hope of a primary prevention measure
Infections and Microbiome for stomach cancer (Zeng et al, 2015).
Helicobacter pylori Some H.  pylori virulence factors may mod-
Already in 1994, the International Agency for ify their carcinogenic capacity. It seems that the
Research on Cancer concluded that the evi- stomach cancer risk is particularly high among
dence was sufficient that H. pylori causes stom- carriers of CagA-​positive strains (and among car-
ach cancer and the microorganism was classified riers of CagA-​positive strains in those with strains
as a human carcinogen class I according to the having the “A-​B-​D-​type” CagA typically seen in
established protocol (IARC, 1994). Most of the Asian high-​risk populations (Hatakeyama, 2004)),
numerous subsequent observational studies although CagA-​negative strains are not without
confirmed this claim. It appears that the asso- risk (Held et al, 2004). The vacA gene of H. pylori,
ciation is confined to noncardia gastric cancer, encoding a vacuolating cytotoxin, comprises two
while the risk is similarly elevated for both the variable regions: the s (signaling) and the m (mid)
intestinal and diffuse histological types of stom- regions. H.  pylori vacA type s1 and m1 strains
ach cancer. It was estimated that in year 2008, appear more carcinogenic than strains with other
worldwide 660,000 out of 989,000 new cases vacA types (Figueiredo et al, 2002). In studies using
of stomach cancer (66.7%) can be attributed a multiplex serology assay, several other serologi-
to H.  pylori infection (de Martel et  al, 2012). cal markers have been associated with an increased
This estimate was further renewed to 780,000 or decreased risk of stomach cancer; however no
(78.9%) based on study results using a more consistent pattern was found across studies con-
sensitive assay for detecting H. pylori antibodies ducted in different populations (Epplein et al, 2014;
(Plummer et al, 2015). Epplein et al, 2012; Gao et al, 2009; Murphy et al,
However, because some infections disap- 2015; Shakeri et  al, 2015; Song et  al, 2014). This
pear spontaneously due to changes in the gas- lack of consistency might highlight the importance
tric microenvironment during the precancerous of host–​pathogen interaction in the carcinogen-
stages, the strength of the association with esis process. A  study from Columbia found that
28

228 Textbook of Cancer Epidemiology

mismatches between host and H.  pylori genetic between the serum level of ascorbic acid and the
traits might affect stomach cancer risk in local risk of progression to mucosal dysplasia or gas-
populations (Kodaman et al, 2014). tric cancer was noted in a cohort study (You et al,
Proposed mechanisms behind the puta- 2000). In a population-​based case-​control study
tive carcinogenicity of H.  pylori include:  (1) an (Ekstrom et  al, 2001), the association between
intense inflammatory response with increased H. pylori infection and the risk of noncardia gas-
production of proinflammatory cytokines tric cancer was significantly modified by ascorbic
(tumor necrosis factor alpha [TNF-​ α], IL-​1β, acid intake, so that the risk gradient was consid-
IL-​6, IL-​8, and IL-​12) followed by attraction and erably steeper among subjects with a low intake
activation of macrophages and polymorphonu- of ascorbic acid compared to those with a high
clear cells, with degranulation and production intake.
of mutagenic superoxide radicals; (2)  induction Experimental data in a mouse model dem-
of inducible nitric oxide synthase (iNOS) and onstrated that treatment with the antioxidant
proinflammatory cyclooxygenase (particularly astaxanthin reduces gastric inflammation and
COX-​2) enzymes, resulting in release of reac- causes a shift of the T-​ lymphocyte response
tive, mutagenic substances and possible altera- from a predominant Th1 response dominated by
tions in epithelial cell growth; (3) increased cell interferon-​γ to a mixed Th1/​Th2 response with
proliferation, possibly due to the activation of interferon-​γ and IL-​4 (Bennedsen et  al, 1999).
the proto-​oncogene c-​fos, reduced expression Most persuasive is a report from a randomized
of the cell-​ cycle regulatory protein p27, and chemoprevention trial in Colombia (Correa et al,
CCK-​β receptor activation; (4) decreased secre- 2000), in which treatment with beta-​carotene or
tion of vitamin C into the gastric lumen, lead- ascorbic acid significantly increased the rates of
ing to a poorer inhibition of the mutagenicity regression of H.  pylori–​associated atrophy and
of preformed NOCs and deficient scavenging of intestinal metaplasia.
reactive oxygen metabolites, nitrite ions, nitric Despite significant progresses in under-
oxide, and other free radicals; (5) development of standing H.  pylori–​ related stomach cancer, a
atrophic gastritis, with decreased production of number of intriguing questions remain to be
hydrochloric acid entailing colonization of bacte- addressed: Why is the incidence of gastric cancer
rial species that catalyze the nitrosation of dietary twice as high in men as in women when the prev-
amines and amides to form carcinogenic NOCs alence of the infection is similar in both genders?
(unopposed by vitamin C); (6)  altered mucus This gender difference has not been explained by
production and reduced mucosal blood flow, known risk factors. Similarly, worldwide many
resulting in abnormal gastric surface permeabil- regions have a high prevalence of H. pylori infec-
ity; (7) an upward expansion of the proliferative tion, while a surprisingly low incidence of stom-
zone, paving the way for greater carcinogen pen- ach cancer is noted there (Graham et  al, 2009).
etration to the stem cells; (8) presence of larger The reasons for this contradictory distribution
pool of cells in S-​phase, sensitive to mutation; are not well understood. Further, why is the
(9)  increased likelihood that mutations escape risk of gastric cancer in duodenal ulcer patients
DNA repair and become fixed in progeny cells 40% lower than that in the general population
due to the increased cell turnover; (10) possi- (Hansson et al, 1996) despite an infection prev-
ble inhibition of the transcription of epithelial alence of almost 100% among the former? It
DNA repair genes; and (11) inhibition of apop- appears that characteristics of the infecting strain
tosis. Strains that are cagA-​positive and harbor or the host, or factors in the environment, or a
the vacA s1a m1 genotype stimulate mucosal cell combination thereof might radically modify the
proliferation in the host while simultaneously carcinogenicity of H. pylori.
inhibiting apoptosis. The resulting hyperprolifer-
ation, not balanced by cell death, may contribute Non–​H. pylori Gastric Microbiota
to malignant transformation. The existence of bacteria other than H. pylori in
It appears that mechanisms that potentially the stomach was first reported in the 1980s based
can be blocked by antioxidants play a quantita- on culturing techniques. However, because a
tively important role. The production of reac- substantial part of gut bacteria might be difficult
tive oxygen metabolites in the gastric mucosa to cultivate, the understanding of the composi-
following H.  pylori infection is well substanti- tion and diversity of gastric microbiota was lim-
ated (Farinati et  al, 1998). An inverse relation ited then. With the emergence of modern DNA
29

Stomach Cancer 229

sequencing technology, the existence of a com- for metagenomic analysis, and incorporating
plex indigenous gastric bacterial community was metabolomic and transcriptomic analyses.
revealed containing five major phyla (Firmicutes,
Bacteroidetes, Proteobacteria, Actinobacteria, Epstein Barr Virus
Fusobacteria) (Bik et  al, 2006; Monstein et  al, About 10% (2%–​20%) of gastric cancers are EBV-​
2000). However, whether non–​H.  pylori micro- associated (so-​called EBVaGC) (Camargo et  al,
biota plays a role in gastric carcinogenesis is 2011). Unlike other EBV-​ related malignancies,
still unclear, although it is reasonable to postu- no endemic area was reported for EBVaGC. They
late their importance. While H.  pylori tends to are more likely to affect the proximal stomach,
disappear in the hypochlorhydric precancerous and have a unique pattern of molecular changes
stomach, other microbes might survive and pro- including recurrent PIK3CA mutations, extreme
liferate. Some species in the thus-​created micro- DNA hypermethylation, and amplification of
biota may reduce nitrate to nitrite by nitrate JAK2, PD-​L1, and PD-​L2 (Cancer Genome Atlas
reductase or produce enzymes capable of cata- Research Network, 2014). Compared to EBV-​
lyzing N-​nitrosation, and hence may act inde- negative gastric cancer, EBVaGC also has a bet-
pendently or synergistically to promote stomach ter prognosis (Camargo et  al, 2014). However,
cancer development. whether EBV plays a causal role in EBVaGC devel-
Animal studies demonstrate that H. pylori-​ opment, and if so, how it interacts with H. pylori
infected mice are less likely to develop gastric to promote gastric carcinogenesis is still unknown.
neoplastic lesions when their enteric micro-
biota was absent (Lertpiriyapong et  al, 2014; Oral Health
Lofgren et al, 2011). So far, human studies are The oral cavity harbors a number of microbes.
scarce. A  study from Mexico used microarray Through constant and high volume salivary
G3 phyloChip to compare gastric microbiota flow these microbes or their products (e.g.,
between patients with nonatrophic gastritis, formed carcinogens) can easily reach the stom-
intestinal metaplasia, and intestinal-​type gas- ach. Previous cross-​sectional studies reported
tric cancer (five patients in each group). They that some specific oral health conditions and
found that bacterial diversity decreased from behaviors, as well as high levels of colonization
gastritis to cancer, and further principal coordi- of periodontal pathogens are associated with
nate analysis revealed that nonatrophic gastritis an increased prevalence of gastric precancerous
microbiota separated significantly from that of lesions (Salazar et al, 2012; Salazar et al, 2013).
gastric cancer, while intestinal metaplasia over- Direct supportive evidence comes from two
lapped with both (Aviles-​Jimenez et  al, 2014). cohort studies showing that tooth loss increases
Another study based on a Chinese Cancer the risk of noncardia stomach cancer (Abnet
Screening cohort in Linxian also revealed that et al, 2001, Abnet et al, 2005). However, in one
bacterial diversity was low in patients with cohort study from the United States, tooth loss
serology-​defined atrophic gastritis, and gastric or periodontal disease did not increase stomach
microbiota were significantly different between cancer risk (Michaud et al, 2008). Confounding
patients with and without atrophic gastritis (Yu by other risk factors, such as age, smoking, and
et al, 2014). socioeconomic status, might explain the con-
In a comparative study between subjects flicting results and caution the interpretation of
residing in two areas in Colombia with con- any observed associations.
trasting incidence rates of gastric cancer
(25-​ fold difference), significant difference of Physical Activity
gastric microbiota was also noted, and exclu- The relationship between physical activity and
sive genuses were observed in both areas. But the risk of gastric cancer has been addressed in
convincing evidence that non–​H.  pylori bac- few studies. The considerable misclassification
teria are involved in gastric carcinogenesis is due to imperfect instruments for exposure assess-
still lacking, as previous studies are limited by ment has hampered investigation of the exact
small sample size or poor study design, includ- association between physical activity and gas-
ing contamination and lack of prospective tric cancer risk. An analysis including 16 cohorts
and longitudinal follow-​up. Other challenges from 7 countries revealed that adherence to the
include distinguishing resident and transient European Code Against Cancer recommenda-
bacteria, avoiding human DNA interference tions regarding physical activity was associated
230

230 Textbook of Cancer Epidemiology

with increased stomach cancer mortality (Ocke but several careful studies failed to observe
et  al, 1998). However, in a large meta-​analysis,
any important association (Swaen et  al, 1985;
physical activity was associated with a signifi- Weinberg et al, 1985; Coggon et al, 1990). If a
cantly reduced risk of gastric cancer (summary true association exists, it is unclear which sub-
RR  =  0.82), which was evident for both cardia stance accounts for the observed excess.
cancer and noncardia gastric cancer (Behrens The overall evidence of associations between
et al, 2014). exposure to asbestos or silica and stomach can-
cer risk has been deemed to be uncertain (Cocco
Ionizing Radiation et  al, 1996). However, data seem to support a
Among atomic bomb survivors in Hiroshima, role of asbestos in stomach carcinogenesis. In
Japan, the standardized mortality ratio for a cohort study in the Netherlands, prolonged
stomach cancer rose steadily with increasing high exposure to asbestos was associated with a
radiation dose (Nakamura, 1977). The same 2.7-​fold excess risk of stomach cancer as com-
pattern was seen in patients treated with x-​ pared with nonexposure (Offermans et al, 2014).
rays for ankylosing spondylitis (Brown & Doll, Similarly, a study from China among chrysotile
1965). Further, among long-​term survivors of miners reported that those with middle or high-
patients with testicular cancer, Hodgkin lym- est level of cumulative asbestos dust exposure
phoma, or cervical cancer who were treated showed a 5-​fold and 12-​fold increased mortality
with radiotherapy, dose-​ dependent increased from stomach cancer, compared with those with
risk of stomach cancer has been reported the lowest level (Lin et al, 2014). A study among
(Hauptmann et  al, 2015; Kleinerman et  al, Swedish construction workers found exposure
2013; Morton et  al, 2013). Persons exposed to to quartz dust—​but not asbestos—​to be signifi-
radiation at younger ages had a higher risk of cantly and dose-​dependently related to stomach
stomach cancer, as were nonuranium miners cancer risk (Sjodahl et al, 2007), albeit with mod-
exposed to alpha-​emissions from radon daugh- erate strength (relative risk 1.3 in highest expo-
ters (Darby et al, 1995) and female radium dial sure tertile relative to lowest). An increased risk
workers (Stebbings et al, 1984). Although based of stomach cancer among workers exposed to
on small numbers of observed cases, the con- wood dust has often been noted (e.g., Robinson
sistency of the findings strongly suggests that et al, 1996; Jansson et al, 2005), but the findings
the risk of gastric cancer increases with expo- across occupations are inconsistent, and sup-
sure to ionizing radiation. port from exposure–​response analyses is lack-
ing. Textile dust has been implicated by some
Occupation investigators (Simpson et  al, 1999), while oth-
Small numbers and/​ or considerable hetero- ers found no association between textile work
geneity with regard to factual exposure typ- and stomach cancer risk (Aragones et  al, 2002;
ically hamper the interpretation of data on Weiderpass et  al, 2003; Ji & Hemminki, 2006;
site-​
specific cancer risks from occupational Wernli et al, 2006).
studies. Because the large literature on occupa- A small excess of stomach cancer was
tional exposures and gastric cancer risk is not revealed in a large meta-​analysis of studies con-
strikingly consistent, the data need cautious cerned with cancer risks among farmers (Blair
interpretation. Occupations commonly asso- et al, 1992). The reasons for this excess remain
ciated with an increased risk of gastric can- to be clarified, but exposure to pesticides/​her-
cer include work in dusty industries such as bicides and fertilizers has been specifically
foundry, steel, and mining (Cocco et  al, 1996; evaluated. Self-​reported pesticide exposure was
Ji & Hemminki, 2006). The reasoning is that a moderately strong (odds ratio 2.1) risk fac-
when dust is inhaled, the particles are cleared tor for cardia cancer in a Swedish population-​
from the lungs by the mucociliary system and based case-​control study (Jansson et al, 2006).
swallowed. The majority of studies concerned A  cohort study among farmers revealed that
with exposure to mineral and metal dusts have exposure to methyl bromide, a genotoxic
been positive (Cocco et al, 1996; Ekstrom et al, soil fumigant, is associated with exposure-​
1999a; Aragones et  al, 2002; Ji & Hemminki, dependent increase of stomach cancer risk
2006; Kreuzer et  al, 2012). The evidence also (Barry et al, 2012). Large studies among farm-
weighs in favor of an increased gastric cancer ers and others with professional exposure to
risk among coal miners (Cocco et  al, 1996), pesticides did not reveal any excess incidence
231

Stomach Cancer 231

or mortality of stomach cancer (Alavanja et al, Medical Conditions

2005; Blair et al, 2005; Lee et al, 2004). Although and Treatment
a link to phenoxy acids has been suggested, the Peptic Ulcer
epidemiologic evidence for a causal associa- Although gastric cancer and peptic ulcer share
tion between chlorophenols and gastric cancer the same dominating risk factor, namely H. pylori
risk is weak (Cocco et al, 1996). A 70% excess infection, which might be a necessary cause of
risk among subjects exposed to phenoxyacetic both diseases, not all peptic ulcer patients have
acids was revealed in one study with detailed an increased risk of gastric cancer. A large retro-
exposure assessment and control for potential spective cohort study among unoperated peptic
confounding (Ekstrom et  al, 1999a), but there ulcer patients hospitalized in Sweden from 1965
was no clear trend with increasing duration of and onward revealed—​as expected—​an approx-
the exposure. Reports on fertilizers and gas- imately twofold risk elevation among gastric
tric cancer risk were considered inconclusive ulcer patients, relative to the age-​, sex-​, and cal-
(Cocco et al, 1996). endar period–​matched Swedish general popula-
Exposure to preformed nitrosamines has tion (Hansson et al, 1996). Unoperated patients
been a concern in agriculture, the rubber hospitalized for duodenal ulcer, on the other
industry, and foundries, as well as in the metal, hand, had a risk that was almost halved, despite
leather, and chemical industries. Excess cases a close to 100% H.  pylori infection rate. A  sub-
of gastric cancer have been observed, notably sequent follow-​up of the same cohort showed
among workers in the rubber industry, where that this risk reduction pertained to noncardia
the epidemiologic evidence is considered more gastric cancer only, while the risk of cardia can-
consistent. cer was close to, or slightly above, expectation
A large meta-​analysis reported a moderately (Bahmanyar et al, 2007). In contrast, the risk ele-
increased risk of stomach cancer associated with vation linked to gastric ulcer was similar for both
exposure to chromium VI, an established lung cardia and noncardia gastric cancer.
carcinogen (Welling et al, 2015). Chromium VI Gastric ulcer is associated with H.  pylori
exposure is noted among chromium production colonization of both the lowermost part of the
or plating, stainless steel, cement, and leather stomach (antrum) and the middle-​upper part
workers. (corpus), where the hydrochloric acid is pro-
Exposure to diesel exhausts among construc- duced. This leads to corpus-​predominant gastri-
tion workers (Sjodahl et  al, 2007), motor vehi- tis, and a tendency toward atrophy with decreased
cle operators, transport workers, or gas station acid production, similar to the intragastric milieu
workers (Aragones et  al, 2002; Ji & Hemminki, in cases with stomach cancer. Conversely, duo-
2006; Engel et  al, 2002)  has been shown with denal ulcer is associated with H. pylori coloniza-
some consistency to be linked to an increased tion exclusively in the antrum, where the gastric
risk of stomach cancer. Although potentially car- acidity is monitored and regulated. The antrum-​
cinogenic PAHs and combustion particles might predominant gastritis, therefore, is linked to a
be involved, the causality remains uncertain. less efficient control of acid production and a
An increased risk of gastric cancer was tendency toward hyperacidity. It is hypothesized
observed for workers in oil refineries, but work that the acidity in some way contributes to the
in the petroleum industry overall did not show apparent protection against noncardia stomach
an excess risk in a meta-​analysis (Cocco et al, cancer, possibly through prevention of superin-
1996). There is suggestive evidence that grind- fection by other microorganisms.
ing operations, which can entail either min-
eral oil-​based or ethanolamine-​based oils, are Partial Gastrectomy
associated with an excess risk of cancer of the Partial gastrectomy, but not vagotomy alone, is
esophagus, stomach, and pancreas (Tolbert, followed by an increased risk of gastric cancer.
1997). An excess gastric cancer risk also has Probably due to the reduction in the amount of
been seen among workers exposed to sulfites tissue at risk, the excess becomes apparent only
and sulfates in the pulping process of the paper after prolonged follow-​up. But according to the
industry (Milham & Demers, 1984; Rix et  al, largest cohort study published on the subject
1997). A decreased risk, however, was reported (Lundegardh et  al, 1988), the risk increased by
among sulfite mill workers (Andersson 28% for each successive 5-​year interval after sur-
et al, 1998). gery. Moreover, Billroth II reconstructions, and
23

232 Textbook of Cancer Epidemiology

operations performed for gastric rather than for et  al, 1996). A  reverse causation arose because,
duodenal ulcer, conferred a higher risk. The rel- in some patients, these drugs were prescribed for
ative excess appears greater in women than in symptoms emanating from a yet-​undiagnosed
men. Meta-​analyses of all published studies con- gastric cancer, and an excessive incidence of gas-
cerning partial gastrectomy and the risk of gastric tric cancer was invariably observed during the
cancer show significantly elevated risks following first years after the start of treatment. But with
both Billroth I  and Billroth II reconstructions, longer follow-​up, the relative risk estimates fell to
but only 15  years or more after the operation. unity (Moller et al, 1992).
The risk was even greater if a vagotomy had been It is tempting to generalize these findings
added to the procedure (Tersmette et al, 1990). to the even more potent proton pump inhibi-
A retrospective cohort study among all tors. However, indications of a more rapid
patients hospitalized for peptic ulcer in Sweden development of atrophy when the drug was
since 1965 (Bahmanyar et  al, 2007)  has added administered to H. pylori–​infected subjects, and
significant insights though; when compared alterations in the gastric milieu favoring bacte-
to their unoperated peers, surgically resected rial N-​nitrosation (Kuipers et  al, 1996; Mowat
duodenal ulcer patients had a significant 60% et  al, 2000)  are concerns. Such concerns were
increased risk of noncardia gastric cancer in the corroborated by the finding of an increased risk
long term following the operation. This meant of stomach cancer observed among subjects
that their advantage (as unoperated) relative to with most prescriptions for proton pump inhib-
the general population (see above) was reduced, itors or longest follow-​ up duration (Poulsen
albeit not entirely canceled (SIR among unoper- et  al, 2009). No such trend was noted among
ated duodenal ulcer patients  =  0.5; SIR among histamine-​ 2-​
antagonist users, however. This
resected ones = 0.8). study was hampered by a limited statistical
In contrast, patients who underwent gastric power, thus future epidemiologic studies with
resection due to gastric ulcer, with an elevated larger sample sizes or longer follow-​up duration
baseline risk relative to the general population, are warranted.
experienced a risk reduction. In comparison
with their unoperated gastric ulcer peers, the Aspirin and Nonsteroidal
risk reduction was 40% (95% CI 20%–​50%). This Anti-​Inflammatory Drugs
changed the SIR, that is, the relative risk in rela- Several epidemiological studies have noted small
tion to the general Swedish population, from reductions in risk of stomach cancer among users
2.1 among the unoperated gastric ulcer patients of aspirin and/​or nonsteroidal anti-​inflammatory
to 1.2 among the operated. Thus it appears that drugs (NSAIDs). A meta-​analysis of eight case-​
following surgical intervention stomach cancer control studies yielded a summary estimate of
risks are shifted toward normality, regardless relative risk that indicated a statistically signif-
of underlying ulcer type. However, a long-​term icant 22% risk reduction (Wang et  al, 2003).
(>300 months) increased risk of stomach cancer An expanded meta-​analysis including 13 case-​
after distal gastrectomy for peptic ulcers cannot control studies and 8 cohort studies confirmed
be ignored, which might be related to H. pylori or the inverse association, which was evidenced
Epstein Barr virus infection in the gastric stump irrespective of study design, cancer subsite, and
(Takeno et al, 2014). type of drug (Tian et al, 2010). However, a ran-
domized trial among 40,000 US women suggests
Acid Inhibition that low-​dose aspirin use may not lower the
After the introduction of cimetidine—​a potent risk of stomach cancer after a 10-​year follow-​
H2 receptor antagonist used for treatment of up (Cook et  al, 2005). Notwithstanding, after
peptic ulcer—​several gastroenterologists voiced an extension of 8-​year follow-​up, a statistically
the concern that the acid inhibition might lead nonsignificant 25% reduction of mortality from
to bacterial overgrowth, formation of NOCs, stomach cancer was noted among the aspirin
growth stimulation by increased gastrin levels, group (Cook et al, 2013). This long latent period
and ultimately the development of gastric cancer. for aspirin to protect against stomach cancer was
Therefore, cancer risks among patients treated also observed in a combined analysis of seven
with H2 receptor antagonists have been closely randomized trials that showed a significant pro-
monitored. All studies, regardless of their design, tective effect only after 10  years of follow-​up
have been reassuringly negative (e.g., Johnson (Rothwell et al, 2011).
23

Stomach Cancer 233

Other Risk Factors many siblings may increase the risk of transmis-
Although the biological correlates are yet to be sion of H. pylori. Family size and birth order have
determined, socioeconomic status has been a also been positively associated with the risk of gas-
strong risk indicator for gastric cancer in epi- tric cancer (Hansson et al, 1994; Blaser et al, 1995).
demiologic studies. Generally, people who have
low education and/​ or are economically under- CONCLUSION
privileged have at least a twofold higher risk than Efforts to improve the treatment of stomach can-
persons who are better off (Hansson et  al, 1994; cer, the second most common cancer worldwide,
Brown et  al, 1998; van Loon et  al, 1998b; Ji & have had limited success. In contrast, consider-
Hemminki, 2006). The risk gradient across soci- able knowledge about potentially causal factors
oeconomic strata has never been fully explained. has been gathered during the past few decades,
It is conceivable that occupational and dietary fac- and primary prevention no longer appears to
tors may play a role. Crowding during childhood be an unattainable goal (Table 10-​2). H.  pylori,
due to poor living conditions and the presence of the strongest and most important risk factor, is

TABLE 10-​2   EPIDEMIOLOGY OF STOMACH CANCER:

RISK FACTOR SUMMARY

Well-​Confirmed Risk Factors Direction of effect*

H. pylori ↑↑↑
Tobacco smoking ↑↑
Ionizing radiation ↑↑
Probable Relationship Exists Based on Substantial Data
Fresh fruits and vegetables ↓
Dietary fiber ↓
Salt or salty foods ↑
Occupation ↑
Aspirin and NSAIDs ↓
Partial gastrectomy ↑
Weak if any Relationship Exists Based on Substantial Data
—​ —​
Inconsistent Findings or Limited Study to Date
Alcoholic beverage consumption ↑
Meat intake ↑
Tea ↓
Coffee ↑
Vitamins ↓
Iron deficiency ↑
Selenium ↓
N-​nitro compounds ↑
EBV ↑
Poor oral health ↑
Acid inhibition ↑
Non–​H. pylori microbiota ↑
Physical activity ↓

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate
to large increase in risk; ↑↑↑, large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large
decrease in risk; —​, no association.
234

234 Textbook of Cancer Epidemiology

likely to become the first target in future pre- Aragones N, Pollan M, Gustavsson P. Stomach can-
vention strategies. A  deeper understanding of cer and occupation in Sweden:  1971–​89. Occup
effect-​modifying factors or circumstances in the Environ Med 2002;59:329–​37.
microorganism, in the host, and/​or in the envi- Aragones N, Pollan M, Lopez-​Abente G, Ruiz M,
ronment may help us to design precisely targeted Vergara A, Moreno C et al. Time trend and age-​
interventions. period-​cohort effects on gastric cancer incidence
Dietary factors, particularly the intake in Zaragoza and Navarre, Spain. J Epidemiol
of antioxidant-​rich fruits and vegetables, are Community Health 1997;51:412–​7.
Aviles-​Jimenez F, Vazquez-​ Jimenez F, Medrano-​
likely to be important, albeit perhaps some-
Guzman R, Mantilla A, Torres J. Stomach micro-
what overrated in older case-​ control data.
biota composition varies between patients with
Interesting interactions with H.  pylori have
non-​atrophic gastritis and patients with intestinal
been observed. New insights in bacterial and type of gastric cancer. Sci Rep 2014;4:4202.
human genetics may enable us to focus even Bagnardi V, Blangiardo M, La Vecchia C, Corrao G. A
more precisely on the relevant high-​ risk meta-​analysis of alcohol drinking and cancer risk.
groups. They may also allow selective preven- Br J Cancer 2001;85:1700–​5.
tion of infections predestined to trigger the Bahmanyar S, Zendehdel K, Nyren O, Ye W. Risk of oesoph-
carcinogenic process. To achieve these goals, ageal cancer by histology among patients hospitalised
epidemiologists, gastroenterologists, patholo- for gastroduodenal ulcers. Gut 2007;56:464–​8.
gists, tumor biologists, microbiologists, geneti- Barry KH, Koutros S, Lubin JH, Coble JB, Barone-​
cists, and biochemists need to join forces in a Adesi F, Beane Freeman LE et al. Methyl bromide
coordinated effort. exposure and cancer risk in the Agricultural Health
Study. Cancer Causes Control 2012;23:807–​18.
REFERENCES Barstad B, Sorensen TI, Tjonneland A, Johansen D,
Aarnio M, Salovaara R, Aaltonen LA, Mecklin JP, Becker U, Andersen IB et al. Intake of wine, beer
Jarvinen HJ. Features of gastric cancer in hered- and spirits and risk of gastric cancer. Eur J Cancer
itary non-​polyposis colorectal cancer syndrome. Prev 2005;14:239–​43.
Int J Cancer 1997;74:551–​5. Behrens G, Jochem C, Keimling M, Ricci C, Schmid
Abnet CC, Kamangar F, Dawsey SM, Stolzenberg-​ D, Leitzmann MF. The association between phys-
Solomon R Z, Albanes D et al. Tooth loss is asso- ical activity and gastroesophageal cancer: system-
ciated with increased risk of gastric non-​cardia atic review and meta-​analysis. Eur J Epidemiol
adenocarcinoma in a cohort of Finnish smokers. 2014;29:151–​70.
Scand J Gastroenterol 2005;40:681–​7. Bennedsen M, Wang X, Willen R, Wadstrom T, Andersen
Abnet CC, Qiao YL, Mark SD, Dong ZW, Taylor PR, LP. Treatment of H. pylori infected mice with anti-
Dawsey SM. Prospective study of tooth loss and oxidant astaxanthin reduces gastric inflammation,
incident esophageal and gastric cancers in China. bacterial load and modulates cytokine release by
Cancer Causes Control 2001;12:847–​54. splenocytes. Immunol Lett 1999;70:185–​9.
Ahlbom A, Lichtenstein P, Malmstrom H, Feychting Bik EM, Eckburg PB, Gill SR, Nelson KE, Purdom EA,
M, Hemminki K, Pedersen NL. Cancer in twins: Francois F et al. Molecular analysis of the bacterial
genetic and nongenetic familial risk factors [see microbiota in the human stomach. Proc Natl Acad
comments]. J Natl Cancer Inst 1997;89:287–​93. Sci U S A 2006;103:732–​7.
Alavanja MC, Sandler DP, Lynch CF, Knott C, Lubin Blair A, Sandler DP, Tarone R, Lubin J, Thomas K,
JH, Tarone R et al. Cancer incidence in the agri- Hoppin JA et  al. Mortality among participants
cultural health study. Scand J Work Environ in the agricultural health study. Ann Epidemiol
Health 2005;31 Suppl 1:39–​45; Discussion 5–​7. 2005;15:279–​85.
American Cancer Society. Cancer Facts and Figures Blair A, Zahm SH, Pearce NE, Heineman EF, Fraumeni
2015. Atlanta: American Cancer Society, 2015. JF. Clues to cancer etiology from studies of farm-
Anderson WF, Camargo MC, Fraumeni JF Jr, Correa ers. Scand J Work Environ Health 1992;18:209–​15.
P, Rosenberg PS, Rabkin CS. Age-​specific trends Blaser MJ, Chyou PH, Nomura A. Age at establish-
in incidence of noncardia gastric cancer in US ment of helicobacter pylori infection and gastric
adults. JAMA 2010;303;1723–​8. carcinoma, gastric ulcer, and duodenal ulcer risk.
Andersson E, Nilsson T, Persson B, Wingren G, Toren Cancer Res 1995;55:562–​5.
K. Mortality from asthma and cancer among sul- Blot W, Devesa S, Kneller R, Fraumeni JJ. Rising inci-
fite mill workers. Scand J Work Environ Health dence of adenocarcinoma of the esophagus and
1998;24:12–​7. gastric cardia. JAMA 1991;265:1287–​1289.
235

Stomach Cancer 235

Blot WJ, Li JY, Taylor PR, Guo W, Dawsey S, Wang cancer in Illinois 1986–​ 1988. Am J Epidemiol
GQ et al. Nutrition intervention trials in Linxian, 1996;144:661–​4.
China:  supplementation with specific vitamin/​ Chow WH, Swanson CA, Lissowska J, Groves FD,
mineral combinations, cancer incidence, and Sobin LH, Nasierowska-​Guttmejer A et al. Risk of
disease-​specific mortality in the general popula- stomach cancer in relation to consumption of cig-
tion. J Natl Cancer Inst 1993;85:1483–​92. arettes, alcohol, tea and coffee in Warsaw Poland.
Botterweck AA, Van Den Brandt PA, Goldbohm Int J Cancer 1999;81:871–​6.
RA. Vitamins, carotenoids, dietary fiber, and the Chyou PH, Nomura AM, Hankin JH, Stemmermann
risk of gastric carcinoma: results from a prospec- GN. A case-​cohort study of diet and stomach can-
tive study after 6.3  years of follow-​up. Cancer cer. Cancer Res 1990;50:7501–​4.
2000;88:737–​48. Cocco P, Ward MH, Buiatti E. Occupational risk fac-
Brenner H, Arndt V, Bode G, Stegmaier C, Ziegler tors for gastric cancer:  an overview. Epidemiol
H, Stumer T. Risk of gastric cancer among smok- Rev 1996;18:218–​34.
ers infected with helicobacter pylori. Int J Cancer Coggon D, Barker DJ, Cole RB. Stomach can-
2002;98:446–​9. cer and work in dusty industries. Br J Ind Med
Brenner H, Arndt V, Stegmaier C, Ziegler H, 1990;7:298–​301.
Rothenbacher D. Is Helicobacter pylori infection Cohen AJ, Roe FJ. Evaluation of the aetiological role of
a necessary condition for noncardia gastric can- dietary salt exposure in gastric and other cancers
cer? Am J Epidemiol 2004;159 252–​8. in humans. Food Chem Toxicol 1997;35:271–​93.
Brown J, Harding S, Bethune A, Rosato M. Coleman MP, Esteve J, Damiecki P, Arslan A, Renhard
Longitudinal study of socio-​economic differences H. Trends in Cancer Incidence and Mortality.
in the incidence of stomach, colorectal and pan- Lyon:  International Agency for Research on
creatic cancers. Popul Trends 1998;35–​41. Cancer, 1993.
Brown WM, Doll R. Mortality from cancer and other Cook MB, Kamangar F, Weinstein SJ, Albanes D,
causes after radiotherapy for ankylosing spondyli- Virtamo J, Taylor PR et  al. Iron in relation to
tis. BMJ 1965;5474:1327–​32. gastric cancer in the Alpha-​ Tocopherol Beta-​
Caldas C, Carneiro F, Lynch HT, Yokota J, Wiesner Carotene Cancer Prevention Study. Cancer
GL, Powell SM et al. Familial gastric cancer: over- Epidemiol Biomarkers Prev 2012;21:2033–​42.
view and guidelines for management. J Med Genet Cook NR, Lee IM, Gaziano JM, Gordon D, Ridker
1999;36:873–​80. PM, Manson JE et al. Low-​dose aspirin in the pri-
Camargo MC, Kim WH, Chiaravalli AM, Kim KM, mary prevention of cancer:  the Women’s Health
Corvalan AH, Matsuo K et al. Improved survival Study:  a randomized controlled trial. JAMA
of gastric cancer with tumour Epstein-​Barr virus 2005;294:47–​55.
positivity:  an international pooled analysis. Gut Cook NR, Lee IM, Zhang SM, Moorthy MV, Buring
2014;63:236–​43. JE. Alternate-​ day low-​dose aspirin and cancer
Camargo MC, Murphy G, Koriyama C, Pfeiffer risk: long-​term observational follow-​up of a ran-
RM, Kim WH, Herrera-​ Goepfert R, et  al. domized trial. Ann Intern Med 2013;159:77–​85.
Determinants of Epstein-​Barr virus-​positive gas- Correa P, Fontham ET, Bravo JC, Bravo LE, Ruiz B,
tric cancer: an international pooled analysis. Br J Zarama G et al. Chemoprevention of gastric dys-
Cancer 2011;105:38–​43. plasia:  randomized trial of antioxidant supple-
Cancer Genome Atlas Research Network. ments and anti-​Helicobacter pylori therapy. J Natl
Comprehensive molecular characterization of Cancer Inst 2000;92:1881–​8.
gastric adenocarcinoma. Nature 2014;513:202–​9. D’elia L, Galletti F, Strazzullo P. Dietary salt intake
Carman S, Kamangar F, Freedman ND, Wright ME, and risk of gastric cancer. Cancer Treat Res
Dawsey SM, Dixon LB et  al. Vitamin E intake 2014;159:83–​95.
and risk of esophageal and gastric cancers in the Darby SC, Radford EP, Whitley E. Radon exposure
NIH-​AARP Diet and Health Study. Int J Cancer and cancers other than lung cancer in Swedish
2009;125:165–​70. iron miners. Environ Health Perspect 1995;103
Chao A, Thun MJ, Henley SJ, Jacobs EJ, Mccullough Suppl 2:45–​7.
ML, Calle EE. Cigarette smoking, use of other Dawsey SM, Wang GQ, Taylor PR, Li JY, Blot WJ, Li B
tobacco products and stomach cancer mortality et al. Effects of vitamin/​mineral supplementation
in us adults: the cancer prevention study II. Int J on the prevalence of histological dysplasia and
Cancer 2002;101:380–​9. early cancer of the esophagus and stomach: results
Cho NH, Moy CS, Davis F, Haenszel W, Ahn YO, Kim from the dysplasia trial in Linxian, China. Cancer
H. Ethnic variation in the incidence of stomach Epidemiol Biomarkers Prev 1994;3:167–​72.
236

236 Textbook of Cancer Epidemiology

De Angelis R, Sant M, Coleman MP, Francisci S, risks of esophageal and gastric cancers. J Natl
Baili P, Pierannunzio D et  al. Cancer survival in Cancer Inst 2003;95:1404–​13.
Europe 1999–​2007 by country and age:  results Engel LS, Vaughan TL, Gammon MD, Chow WH,
of Eurocare–​5; a population-​based study. Lancet Risch HA, Dubrow R, Mayne ST et al. Occupation
Oncol 2014;15:23–​34. and risk of esophageal and gastric cardia adeno-
De Martel C, Ferlay J, Franceschi S, Vignat J, Bray F, carcinoma. Am J Ind Med 2002;42:11–​22.
Forman D et al. Global burden of cancers attribut- Epplein M, Shu XO, Xiang YB, Chow WH, Yang G, Li
able to infections in 2008: a review and synthetic HL et al. Fruit and vegetable consumption and risk
analysis. Lancet Oncol 2012;13:607–​15. of distal gastric cancer in the Shanghai Women’s
De Stefani E, Boffetta P, Mendilaharsu M, Carzoglio J, and Men’s Health Studies. Am J Epidemiol
Deneo-​Pellegrini H. Dietary nitrosamines, het- 2010;172:397–​406.
erocyclic amines, and risk of gastric cancer:  a Epplein M, Zheng W, Li H, Peek RM Jr, Correa P, Gao
case-​control study in Uruguay. Nutr Cancer J et  al. Diet, Helicobacter pylori strain-​specific
1998;30:158–​62. infection, and gastric cancer risk among Chinese
Devesa SS, Blot WJ, Fraumeni JF Jr. Changing patterns men. Nutr Cancer 2014;66:550–​7.
in the incidence of esophageal and gastric carci- Epplein M, Zheng W, Xiang YB, Peek RM Jr, Li H,
noma in the United States. Cancer 1998;83:2049–​53. Correa P et al. Prospective study of Helicobacter
Devesa SS, Silverman DT, Young JL Jr, Pollack ES, pylori biomarkers for gastric cancer risk among
Brown CC, Horm JW et al. Cancer incidence and Chinese men. Cancer Epidemiol Biomarkers Prev
mortality trends among whites in the United States 2012;21:2185–​92.
1947–​84. J Natl Cancer Inst 1987;79:701–​70. Everatt R, Tamosiunas A, Kuzmickiene I, Virviciute D,
Donner I, Kiviluoto T, Ristimaki A, Aaltonen LA, Radisauskas R, Reklaitiene R et  al. Alcohol con-
Vahteristo P. Exome sequencing reveals three sumption and risk of gastric cancer: a cohort study
novel candidate predisposition genes for diffuse of men in Kaunas, Lithuania, with up to 30 years
gastric cancer. Fam Cancer 2015;14:241–​6. follow-​up. BMC Cancer 2012;12:475.
Duell EJ, Travier N, Lujan-​ Barroso L, Clavel-​ Fang XX, Wei JY, He XY, An P, Wang H, Jiang L et al.
Chapelon F, Boutron-​Ruault MC, Morois S et al. Landscape of dietary factors associated with risk
Alcohol consumption and gastric cancer risk of gastric cancer:  a systematic review and dose-​
in the European Prospective Investigation into response meta-​ analysis of prospective cohort
Cancer and Nutrition (EPIC) Cohort. Am J Clin studies. Eur J Cancer 2015;51;2820–​2832.
Nutr 2011;94:1266–​75. Farinati F, Cardin R, Degan P, Rugge M, Mario
Ekstrom AM, Eriksson M, Hansson LE, Lindgren FD, Bonvicini P, et  al. Oxidative DNA dam-
A, Signorello LB, Nyren O et  al. Occupational age accumulation in gastric carcinogenesis. Gut
exposures and risk of gastric cancer in a 1998;42:351–​6.
population-​based case-​control study. Cancer Res Figueiredo C, Machado JC, Pharoah P, Seruca R,
1999a;59:5932–​7. Sousa S, Carvalho R, et al. Helicobacter pylori and
Ekstrom AM, Hansson LE, Signorello LB, Lindgren interleukin 1 genotyping: an opportunity to iden-
A, Bergstrom R, Nyren O. Decreasing incidence tify high-​risk individuals for gastric carcinoma. J
of both major histologic subtypes of gastric Natl Cancer Inst 2002;94:1680–​7.
adenocarcinoma—​ a population-​based study in Fonseca-​Nunes A, Agudo A, Aranda N, Arija V, Cross
Sweden. Br J Cancer 2000;83:391–​6. A, J, Molina E et al. Body iron status and gastric
Ekstrom AM, Held M, Hansson LE, Engstrand L, cancer risk in the EURGAST study. Int J Cancer
Nyren O. Helicobacter pylori in gastric cancer 2015;137:2904–​14.
established by CagA immunoblot as a marker of Galanis DJ, Kolonel LN, Lee J, Nomura A. Intakes of
past infection. Gastroenterology 2001;121:784–​91. selected foods and beverages and the incidence
Ekstrom AM, Signorello LB, Hansson LE, Bergstrom of gastric cancer among the Japanese residents
R, Lindgren A, Nyren O. Evaluating gastric cancer of Hawaii:  a prospective study. Int J Epidemiol
misclassification:  a potential explanation for the 1998;27:173–​80.
rise in cardia cancer incidence [see comments]. J Gao L, Michel A, Weck MN, Arndt V, Pawlita M,
Natl Cancer Inst 1999b;91:786–​90. Brenner H. Helicobacter pylori infection and gas-
Elimova E, Ajani J. Surgical resection first for local- tric cancer risk: evaluation of 15 H. pylori proteins
ized gastric adenocarcinoma:  are there adjuvant determined by novel multiplex serology. Cancer
options? J Clin Oncol 2015;33:3085–​91. Res 2009;69:6164–​70.
Engel LS, Chow WH, Vaughan TL, Gammon MD, Gaston D, Hansford S, Oliveira C, Nightingale M,
Risch HA, Stanford JL et al. Population attributable Pinheiro H, Macgillivray C et  al. Germline
237

Stomach Cancer 237

mutations in Map3k6 are associated with familial Hennekens CH, Buring JE, Manson JE, Stampfer M,
gastric cancer. PloS Genet 2014;10:E1004669. Rosner B, Cook NR et al. Lack of effect of long-​
Gayther SA, Gorringe KL, Ramus SJ, Huntsman term supplementation with beta carotene on the
D, Roviello F, Grehan N et  al. Identification of incidence of malignant neoplasms and cardiovas-
germ-​line e-​cadherin mutations in gastric can- cular disease. N Engl J Med 1996;334:1145–​9,
cer families of European origin. Cancer Res Holcombe C. Helicobacter pylori: the African enigma.
1998;58:4086–​9. Gut 1992;33:429–​31.
Gonzalez CA, Jakszyn P, Pera G, Agudo A, Bingham S, IARC. Some Natural Occurring Substances:  Food
Palli D et al. Meat intake and risk of stomach and Items and Constituents, Heterocyclic Aromatic
esophageal adenocarcinoma within the European Amines and Mycotoxins. IARC Monographs on
prospective investigation into cancer and nutri- the Evaluation of Carcinogenic Risks to Humans,
tion (EPIC). J Natl Cancer Inst 2006;98:345–​54. Vol. 56. Lyon: IARC, 1993.
Gonzalez C, A, Pera G, Agudo A, Palli D, Krogh V, IARC. Schistosomes, Liver Flukes and Helicobacter
Vineis P et  al. Smoking and the risk of gastric Pylori. IARC Monographs on the Evaluation
cancer in the European prospective investigation of Carcinogenic Risks to Humans, Vol. 61.
into cancer and nutrition (EPIC). Int J Cancer Lyon: IARC, 1994.
2003;107:629–​34. IARC. Cancer in Africa:  epidemiology and preven-
Graham D, Y, Lu H, Yamaoka Y. African, Asian or Indian tion. IARC Sci Publ Vol. 153. Lyon: IARC, 2003.
enigma, the East Asian Helicobacter pylori:  facts Ito Y, Kurata M, Hioki R, Suzuki K, Ochiai J, Aoki K.
or medical myths. J Dig Dis 2009;10:77–​84. Cancer mortality and serum levels of carotenoids,
Guilford P, Hopkins J, Harraway J, Mcleod M, retinol and tocopherol:  a population-​ based
Mcleod N, Harawira P et  al. E-​cadherin germ- follow-​up study of inhabitants of a rural area of
line mutations in familial gastric cancer. Nature Japan. Asian Pac J Cancer Prev 2005;6:10–​5.
1998;392:402–​5. Jacobsen B, K, Bjelke E, Kvale G, Heuch I. Coffee
Hansson LE, Baron J, Nyren O, Bergstrom R, Wolk A, drinking, mortality, and cancer incidence: results
Lindgren A, Adami HO. Early-​life risk indicators from a Norwegian prospective study. J Natl
of gastric cancer. A population-​based case-​control Cancer Inst 1986;76:823–​31.
study in Sweden. Int J Cancer 1994;57:32–​7. Jakszyn P, Bingham S, Pera G, Agudo A, Luban R,
Hansson LE, Bergstrom R, Sparen P, Adami HO. The Welch A et  al. Endogenous versus exogenous
decline in the incidence of stomach cancer in exposure to n-​nitroso compounds and gastric can-
Sweden 1960–​1984: a birth cohort phenomenon. cer risk in the European Prospective Investigation
Int J Cancer 1991;47:499–​503. into Cancer and Nutrition (Epic-​Eurgast) Study.
Hansson LE, Nyren O, Hsing AW, Bergstrom R, Carcinogenesis 2006;27:1497–​501.
Josefsson S, Chow WH et  al. The risk of stom- Jansson C, Johansson AL, Bergdahl IA, Dickman PW,
ach cancer in patients with gastric or duode- Plato N, Adami J et  al. Occupational exposures
nal ulcer disease [see comments]. N Engl J Med and risk of esophageal and gastric cardia can-
1996;335:242–​9. cers among male Swedish construction workers.
Hansson LE, Sparen P, Nyren O. Increasing incidence Cancer Causes Control 2005;16:755–​64.
of both major histological types of esophageal Jansson C, Plato N, Johansson AL, Nyren O, Lagergren
carcinomas among men in Sweden. Int J Cancer J. Airborne occupational exposures and risk of
1993a;54: 402–​7. oesophageal and cardia adenocarcinoma. Occup
Hansson LE, Sparen P, Nyren O. Increasing incidence Environ Med 2006;63:107–​12.
of carcinoma of the gastric cardia in Sweden from Jenab M, Riboli E, Ferrari P, Friesen M, Sabate J, Norat
1970 to 1985. Br J Surg 1993b;80:374–​7. T et al. Plasma and dietary carotenoid, retinol and
Hatakeyama M. Oncogenic mechanisms of the tocopherol levels and the risk of gastric adenocarci-
Helicobacter pylori CagA protein. Nat Rev Cancer nomas in the European prospective investigation into
2004;4:688–​94. cancer and nutrition. Br J Cancer 2006a;95:406–​15.
Held M, Engstrand L, Hansson LE, Bergstrom R, Jenab M, Riboli E, Ferrari P, Sabate J, Slimani N, Norat
Wadstrom T, Nyren O. Is the association between T et al. Plasma and dietary vitamin C levels and
Helicobacter pylori and gastric cancer confined to risk of gastric cancer in the European Prospective
CagA-​positive strains? Helicobacter 2004;9:271–​7. Investigation into Cancer and Nutrition (Epic-​
Hauptmann M, Fossa SD, Stovall M, Van Leeuwen Eurgast). Carcinogenesis 2006b;27:2250–​7.
FE, Johannesen TB, Rajaraman P et al. Increased Ji J, Hemminki K. Socio-​economic and occupational
stomach cancer risk following radiotherapy for risk factors for gastric cancer:  a cohort study in
testicular cancer. Br J Cancer 2015;112:44–​51. Sweden. Eur J Cancer Prev 2006;15:391–​7.
238

238 Textbook of Cancer Epidemiology

Johnson AG, Jick SS, Perera DR, Jick H. Histamine-​ Kuipers EJ, Lundell L, Klinkenberg-​Knol EC, Havu
2 receptor antagonists and gastric cancer. N, Festen HP, Liedman B et al. Atrophic gastritis
Epidemiology 1996;7:434–​6. and helicobacter pylori infection in patients with
Joossens JV, Hill MJ, Elliott P, Stamler R, Lesaffre E, reflux esophagitis treated with omeprazole or fun-
Dyer A et al. Dietary salt, nitrate and stomach can- doplication. N Engl J Med 1996;334:1018–​22.
cer mortality in 24 countries, European Cancer Ladeiras-​Lopes R, Pereira AK, Nogueira A, Pinheiro-​
Prevention (ECP) and the Intersalt Cooperative Torres T, Pinto I, Santos-​ Pereira R, Lunet N.
Research Group. Int J Epidemiol 1996;25:494–​504. Smoking and gastric cancer:  systematic review
Kamada T, Haruma K, Ito M, Inoue K, Manabe N, and meta-​ analysis of cohort studies. Cancer
Matsumoto H et al. Time trends in Helicobacter Causes Control 2008;19:689–​701.
pylori infection and atrophic gastritis over Lagergren J, Bergstrom R, Lindgren A, Nyren O.
40 years in Japan. Helicobacter 2015;20:92–​8. Symptomatic gastroesophageal reflux as a risk
Kamineni A, Williams MA, Schwartz SM, Cook factor for esophageal adenocarcinoma [see com-
LS, Weiss NS. The incidence of gastric carci- ments]. N Engl J Med 1999;340:825–​31.
noma in Asian migrants to the United States Lagergren J, Bergstrom R, Lindgren A, Nyren O. The
and their descendants. Cancer Causes Control role of tobacco, snuff and alcohol use in the aetiol-
1999;10:77–​83. ogy of cancer of the oesophagus and gastric car-
Keszei AP, Goldbohm RA, Schouten LJ, Jakszyn P, Van dia. Int J Cancer 2000;85:340–​6.
Den Brandt PA. Dietary n-​nitroso compounds, Larsson SC. Diet and Gastrointestinal Cancer. One
endogenous nitrosation, and the risk of esophageal Carbon Metabolism and Other Aspects (Thesis).
and gastric cancer subtypes in the Netherlands Karolinska Institutet, 2006.
Cohort Study. Am J Clin Nutr 2013;97:135–​46. Larsson SC, Bergkvist L, Naslund I, Rutegard J, Wolk
Kim IJ, Kang HC, Shin Y, Park HW, Jang SG, Han A. Vitamin A, retinol, and carotenoids and the
SY et  al. A Tp53-​truncating germline mutation risk of gastric cancer: a prospective cohort study.
(e287x) in a family with characteristics of both Am J Clin Nutr 2007a;85:497–​503.
hereditary diffuse gastric cancer and Li-​Fraumeni Larsson SC, Bergkvist L, Wolk A. Processed meat
Syndrome. J Hum Genet 2004:49:591–​5. consumption, dietary nitrosamines and stomach
Kleinerman RA, Smith SA, Holowaty E, Hall P, cancer risk in a cohort of Swedish women. Int J
Pukkala E, Vaalavirta L et al. Radiation dose and Cancer 2006a;119:915–​9.
subsequent risk for stomach cancer in long-​term Larsson SC, Giovannucci E, Wolk A. Coffee con-
survivors of cervical cancer. Int J Radiat Oncol sumption and stomach cancer risk in a cohort of
Biol Phys 2013;86:922–​9. Swedish women. Int J Cancer 2006b;119:2186–​9.
Knekt P, Aromaa A, Maatela J, Alfthan G, Aaran RK, Larsson SC, Giovannucci E, Wolk A. Alcoholic bev-
Hakama M et al. Serum selenium and subsequent erage consumption and gastric cancer risk: a pro-
risk of cancer among Finnish men and women. J spective population-​based study in women. Int J
Natl Cancer Inst 1990;82:864–​8. Cancer 2007b;120:373–​7.
Knekt P, Jarvinen R, Dich J, Hakulinen T. Risk of colo- Larsson SC, Orsini N, Wolk A. Processed meat con-
rectal and other gastro-​intestinal cancers after expo- sumption and stomach cancer risk:  a meta-​
sure to nitrate, nitrite and n-​nitroso compounds: a analysis. J Natl Cancer Inst 2006c;98:1078–​87.
follow-​up study. Int J Cancer 1999;80:852–​6. Lauren P. The two histological main types of gastric
Kodaman N, Sobota RS, Mera R, Schneider BG, carcinoma:  diffuse and so-​called intestinal type
Williams SM. Disrupted human-​ pathogen co-​ carcinoma, an attempt at histoclinical classifica-
evolution:  a model for disease. Front Genet tion. Acta Pathol Microbiol Scand 1965;64:31–​49.
2014;5:290. Lauren P, Nevalainen TJ. Epidemiology of intestinal
Koizumi Y, Tsubono Y, Nakaya N, Kuriyama S, and diffuse types of gastric carcinoma, a time-​
Shibuya D, Matsuoka H et al. Cigarette smoking trend study in Finland with comparison between
and the risk of gastric cancer:  a pooled analysis studies from high-​and low-​risk areas [see com-
of two prospective studies in Japan. Int J Cancer ments]. Cancer 1993;71:2926–​33.
2004;112:1049–​55. Lee WJ, Lijinsky W, Heineman EF, Markin RS,
Kono S, Hirohata T. Nutrition and stomach cancer. Weisenburger DD, Ward MH. Agricultural pesti-
Cancer Causes Control 1996;7:41–​55. cide use and adenocarcinomas of the stomach and
Kreuzer M, Straif K, Marsh J, W, Dufey F, Grosche oesophagus. Occup Environ Med 2004;61;743–​9.
B, Nosske D et  al. Occupational dust and radia- Lertpiriyapong K, Whary MT, Muthupalani S, Lofgren
tion exposure and mortality from stomach can- JL, Gamazon ER, Feng Y et al. Gastric colonisation
cer among German uranium miners 1946–​2003. with a restricted commensal microbiota replicates
Occup Environ Med 2012;69:217–​23. the promotion of neoplastic lesions by diverse
239

Stomach Cancer 239

intestinal microbiota in the Helicobacter Pylori incidence of gastric cancer in a general Japanese
Ins-​Gas mouse model of gastric carcinogenesis. population:  the Hisayama Study. Gastric Cancer
Gut 2014;63:54–​63. 2012;15:162–​9.
Li JY, Taylor PR, Li B, Dawsey S, Wang GQ, Ershow Mocellin S, Verdi D, Pooley KA, Nitti D. Genetic var-
AG et el. Nutrition intervention trials in Linxian, iation and gastric cancer risk: a field synopsis and
China:  multiple vitamin/​ mineral supplementa- meta-​analysis. Gut 2015;64:1209–​19.
tion, cancer incidence, and disease-​specific mor- Mohar A, Suchil-​Bernal L, Hernandez-​Guerrero A,
tality among adults with esophageal dysplasia. J Podolsky-​Rapoport I, Herrera-​Goepfert R, Mora-​
Natl Cancer Inst 1993;85:1492–​8. Tiscareno A et al. Intestinal type: diffuse type ratio
Lin S, Wang X, Yano E, Yu I, Lan Y, Courtice MN, of gastric carcinoma in a Mexican population. J
Christiani DC. Exposure to chrysotile mining dust Exp Clin Cancer Res 1997;16:189–​94.
and digestive cancer mortality in a Chinese Miner/​ Moller H, Nissen A, Mosbech J. Use of cimetidine and
Miller Cohor. Occup Environ Med 2014;71:323–​8. other peptic ulcer drugs in Denmark 1977–​1990
Lofgren JL, Whary MT, Ge Z, Muthupalani S, Taylor with analysis of the risk of gastric cancer among
NS, Mobley M et al. Lack of commensal flora in cimetidine users. Gut 1992;33:1166–​9.
Helicobacter pylori-​infected ins-​gas mice reduces Monstein HJ, Tiveljung A, Kraft CH, Borch K,
gastritis and delays intraepithelial neoplasia. Jonasson J. Profiling of bacterial flora in gastric
Gastroenterology 2011;140:210–​20. biopsies from patients with helicobacter pylori-​
Lopez-​Carrillo L, Lopez-​Cervantes M, Robles-​Diaz associated gastritis and histologically normal
G, Ramirez-​ Espitia A, Mohar-​ Betancourt A, control individuals by temperature gradient gel
Meneses-​Garcia A et al. Capsaicin consumption, electrophoresis and 16s rDNA sequence analysis.
Helicobacter pylori positivity and gastric cancer J Med Microbiol 2000;49:817–​22.
in Mexico. Int J Cancer 2003;106:277–​82. Morton LM, Dores GM, Curtis RE, Lynch CF, Stovall
Lundegardh G, Adami HO, Helmick C, Zack M, M, Hall P et  al. Stomach cancer risk after treat-
Meirik O. Stomach cancer after partial gastrec- ment for Hodgkin lymphoma. J Clin Oncol
tomy for benign ulcer disease. N Engl J Med 2013;31:3369–​77.
1988;319:195–​200. Mowat C, Williams C, Gillen D, Hossack M, Gilmour
Lundegardh G, Hansson LE, Nyren O, Adami HO, D, Carswell A et  al. Omeprazole, Helicobacter
Krusemo UB. The risk of gastrointestinal and pylori status, and alterations in the intragas-
other primary malignant diseases following gas- tric milieu facilitating bacterial n-​ nitrosation.
tric cancer. Acta Oncol 1991;30:1–​6. Gastroenterology 2000;119:339–​47.
Malila N, Taylor PR, Virtanen MJ, Korhonen P, Mucci LA, Hjelmborg JB, Harris JR, Czene K,
Huttunen JK, Albanes D et  al. Effects of alpha-​ Havelick DJ, Scheike T et al. Nordic Twin Study
tocopherol and beta-​ carotene supplementation of Cancer. Familial risk and heritability of can-
on gastric cancer incidence in male smokers cer among twins in Nordic countries. JAMA
(ATBC Study Finland). Cancer Causes Control 2016;315:68–​76.
2002;13:617–​23. Munoz N, Connelly R. Time trends of intestinal and
Mao Y, Hu J, Semenciw R, White K. Active and passive diffuse types of gastric cancer in the United States.
smoking and the risk of stomach cancer by subsite Int J Cancer 1971;8:158–​64.
in Canada. Eur J Cancer Prev 2002;11:27–​38. Murphy G, Freedman ND, Michel A, Fan JH, Taylor
Marrelli D, Polom K, De Manzoni G, Morgagni P, PR, Pawlita M et al. Prospective study of helico-
Baiocchi G, Roviello F. Multimodal treatment of bacter pylori antigens and gastric noncardia can-
gastric cancer in the West:  where are we going? cer risk in the Nutrition Intervention Trial Cohort.
World J Gastroenterol 2015;21:7954–​69. Int J Cancer 2015;137:1938–​46.
Mayne ST, Ferrucci LM, Cartmel B. Lessons learned Nagata C, Takatsuka N, Kawakami N, Shimizu H. A
from randomized clinical trials of micronutrient prospective cohort study of soy product intake and
supplementation for cancer prevention. Ann Rev stomach cancer death. Br J Cancer 2002;87:31–​6.
Nutr 2012;32:369–​90. Nakamura K. Stomach cancer in atomic-​bomb survi-
Michaud DS, Liu Y, Meyer M, Giovannucci E, vors. Lancet 1977;2:866–​7.
Joshipura K. Periodontal disease, tooth loss, and Ngoan LT, Mizoue T, Fujino Y, Tokui N, Yoshimura T.
cancer risk in male health professionals:  a pro- Dietary factors and stomach cancer mortality. Br J
spective cohort study. Lancet Oncol 2008;9:550–​8. Cancer 2002;87:37–​42.
Milham SJ, Demers RY. Mortality among pulp and Nikulasson S, Hallgrimsson J, Tulinius H, Sigvaldason
paper workers. J Occup Med 1984;26:844–​46. H, Olafsdottir G. Tumours in Iceland. 16.
Miyazaki M, Doi Y, Ikeda F, Ninomiya T, Hata J, Malignant tumours of the stomach, histological
Uchida K et  al. Dietary vitamin A  intake and classification and description of epidemiological
240

240 Textbook of Cancer Epidemiology

changes in a high-​risk population during 30 years. attributable to Helicobacter pylori. Int J Cancer
APMIS 1992;100:930–​41. 2015;136:487–​90.
Nomura AM, Stemmermann GN, Chyou PH. Gastric Plummer M, Vivas J, Lopez G, Bravo JC, Peraza S,
cancer among the Japanese in Hawaii. Jpn J Cancer Carillo E et al. Chemoprevention of precancerous
Res 1995;86:916–​23. gastric lesions with antioxidant vitamin supple-
Nomura AM, Wilkens LR, Henderson BE, Epplein M, mentation: a randomized trial in a high-​risk pop-
Kolonel LN. The association of cigarette smoking ulation. J Natl Cancer Inst 2007;99:37–​46.
with gastric cancer: the multiethnic cohort study. Poulsen AH, Christensen S, Mclaughlin JK,
Cancer Causes Control 2012;23:51–​8. Thomsen RW, Sorensen HT, Olsen JH et  al.
Noto JM, Gaddy JA, Lee JY, Piazuelo MB, Friedman Proton pump inhibitors and risk of gastric can-
DB, Colvin DC et  al. Iron deficiency accel- cer: a population-​based cohort study. Br J Cancer
erates Helicobacter pylori-​ induced carcino- 2009;100:1503–​7.
genesis in rodents and humans. J Clin Invest Ren JS, Freedman ND, Kamangar F, Dawsey SM,
2013;123:479–​92. Hollenbeck AR, Schatzkin A et  al. Tea, coffee,
Nouraie M, Pietinen P, Kamangar F, Dawsey SM, carbonated soft drinks and upper gastrointestinal
Abnet CC, Albanes D et  al. Fruits, vegetables, tract cancer risk in a large United States prospec-
and antioxidants and risk of gastric cancer among tive cohort study. Eur J Cancer 2010;46:1873–​81.
male smokers. Cancer Epidemiol Biomarkers Prev Riboli E, Norat T. Epidemiologic evidence of the pro-
2005;14:2087–​92. tective effect of fruit and vegetables on cancer risk.
Ocke MC, Bueno-​ De-​Mesquita HB, Feskens EJ, Am J Clin Nutr 2003;78:559s–​69s.
Kromhout D, Menotti A, Blackburn H, Adherence Rix BA, Villadsen E, Lynge E. Cancer incidence of
to the European code against cancer in relation to sulfite pulp workers in Denmark, Scand J Work
long-​term cancer mortality: intercohort compari- Environ Health 1997;23:458–​61.
sons from the seven countries study. Nutr Cancer Robinson CF, Petersen M, Sieber WK, Palu S,
1998;30:14–​20. Halperin WE. Mortality of Carpenters’ Union
Offermans NS, Vermeulen R, Burdorf A, Goldbohm members employed in the U.S.  construction or
RA, Keszei AP, Peters S et al. Occupational asbes- wood products industries, 1987–​1990. Am J Ind
tos exposure and risk of esophageal, gastric and Med 1996;30:674–​94.
colorectal cancer in the prospective Netherlands Rothwell PM, Fowkes FG, Belch JF, Ogawa H, Warlow
cohort study. Int J Cancer 2014;135:1970–​7. CP, Meade TW. Effect of daily aspirin on long-​
Ohmura K, Tamura G, Endoh Y, Sakata K, Takahashi term risk of death due to cancer: analysis of indi-
T, Motoyama T. Microsatellite alterations in vidual patient data from randomised trials. Lancet
differentiated-​type adenocarcinomas and pre- 2011;377:31–​41.
cancerous lesions of the stomach with special Salazar CR, Francois F, Li Y, Corby P, Hays R, Leung
reference to cellular phenotype. Hum Pathol C et  al. Association between oral health and
2000;31:1031–​5. gastric precancerous lesions. Carcinogenesis
Oliveira C, Seruca R, Carneiro F. Genetics, pathology, 2012;33:399–​403.
and clinics of familial gastric cancer. Int J Surg Salazar CR, Sun J, Li Y, Francois F, Corby P, Perez-​
Pathol 2006;14:21–​33. Perez G et  al. Association between selected oral
Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer pathogens and gastric precancerous lesions, Plos
statistics 2002. CA Cancer J Clin 2005;55:74–​108. One 2013;8:E51604.
Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J. Sasazuki S, Inoue M, Hanaoka T, Yamamoto S, Sobue
Cancer Incidence in Five Continents, Vol. 7. T, Tsugane S. Green tea consumption and subse-
Lyon:  International Agency for Research on quent risk of gastric cancer by subsite: the JPHC
Cancer, 1997. study. Cancer Causes Control 2004;15:483–​91.
Persson C, Jia Y, Pettersson H, Dillner J, Nyren O, Ye Shakeri R, Malekzadeh R, Nasrollahzadeh D, Pawilta
W. H. Pylori seropositivity before age 40 and sub- M, Murphy G, Islami F et al. Multiplex H. pylori
sequent risk of stomach cancer: a glimpse of the serology and risk of gastric cardia and noncardia
true relationship? Plos One 2011;6:E17404. adenocarcinomas. Cancer Res 2015;75:4876–​83.
Pinheiro P, Van Der Heijden L, Colbergh J. Unchanged Shinmura K, Kohno T, Takahashi M, Sasaki A, Ochiai
survival of gastric cancer in the Southeastern A, Guilford P et al. Familial gastric cancer: clini-
Netherlands since 1982:  result of differential copathological characteristics, RER phenotype
trends in incidence according to Laurén type and and germline P53 and E-​ cadherin mutations.
subsite. Int J Cancer (Pred Oncol) 1999;84:28–​32. Carcinogenesis 1999;20:1127–​31.
Plummer M, Franceschi S, Vignat J, Forman D, Simpson J, Roman E, Law G, Pannett B. Women’s
De Martel C. Global burden of gastric cancer occupation and cancer:  preliminary analysis of
241

Stomach Cancer 241

cancer registrations in England and Wales 1971–​ Tamura G, Sato K, Akiyama S, Tsuchiya T, Endoh
1990. Am J Ind Med 1999;36:172–​85. Y, Usuba O et  al. Molecular characterization of
Sipponen P, Jarvi O, Kekki M, Siurala M. Decreased undifferentiated-​type gastric carcinoma. Lab
incidences of intestinal and diffuse types of gas- Invest 2001;81:593–​8.
tric carcinoma in Finland during a 20-​year period. Taylor PR, Qiao YL, Abnet CC, Dawsey SM, Yang CS,
Scand J Gastroenterol 1987;22:865–​71. Gunter EW et al. Prospective study of serum vita-
Sjodahl K, Jansson C, Bergdahl IA, Adami J, Boffetta min E levels and esophageal and gastric cancers. J
P, Lagergren J. Airborne exposures and risk of Natl Cancer Inst 2003;95:1414–​6.
gastric cancer:  a prospective cohort study. Int J Tersmette AC, Offerhaus GJ, Tersmette KW, Giardiello
Cancer 2007;120:2013–​8. FM, Moore GW, Tytgat GN et  al. Meta-​analysis
Smith MG, Hold GL, Tahara E, El-​Omar EM. Cellular of the risk of gastric stump cancer:  detection of
and molecular aspects of gastric cancer. World J high risk patient subsets for stomach cancer after
Gastroenterol 2006;12:2979–​90. remote partial gastrectomy for benign conditions.
Song H, Ekheden IG, Zheng Z, Ericsson J, Nyren O, Cancer Res 1990;50:6486–​9.
Ye W. Incidence of gastric cancer among patients Tian W, Zhao Y, Liu S, Li X. Meta-​analysis on the rela-
with gastric precancerous lesions:  observational tionship between nonsteroidal anti-​inflammatory
cohort study in a low risk western population, drug use and gastric cancer. Eur J Cancer Prev
BMJ 2015;351:H3867. 2010;19:288–​98.
Song H, Michel A, Nyren O, Ekstrom AM, Pawlita Tolbert PE. Oils and cancer. Cancer Causes Control
M, Ye W. A Caga-​independent cluster of anti- 1997;8:386–​405.
gens related to the risk of noncardia gastric Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-​Tieulent J,
cancer:  Associations between Helicobacter Jemal A. Global cancer statistics 2012. CA Cancer
pylori antibodies and gastric adenocarcinoma J Clin 2015;65:87–​108.
explored by multiplex serology. Int J Cancer Tredaniel J, Boffetta P, Buiatti E, Saracci R, Hirsch A.
2014;134:2942–​50. Tobacco smoking and gastric cancer: review and
Stahelin HB, Gey KF, Eichholzer M, Ludin E, meta-​analysis. Int J Cancer 1997;72:565–​73.
Bernasconi F, Thurneysen J et al. Plasma antiox- Tsugane S. Salt, salted food intake, and risk of gas-
idant vitamins and subsequent cancer mortality tric cancer:  epidemiologic evidence. Cancer Sci
in the 12-​year follow-​up of the prospective Basel 2005;96:1–​6.
study. Am J Epidemiol 1999;133:766–​75. Uemura N, Okamoto S, Yamamoto S, Matsumura
Stebbings JH, Lucas HF, Stehney AF. Mortality from N, Yamaguchi S, Yamakido M et al. Helicobacter
cancers of major sites in female radium dial work- pylori infection and the development of gastric
ers. Am J Ind Med 1984;5:435–​59. cancer. N Engl J Med 2001;345:784–​9.
Steinmetz KA, Potter JD. Vegetables, fruit, and Van Den Brandt PA, Botterweck AA, Goldbohm RA.
cancer prevention:  a review. J Am Diet Assoc Salt intake, cured meat consumption, refrigera-
1996;96:1027–​39. tor use and stomach cancer incidence: a prospec-
Swaen GM, Aerdts CW, Sturmans F, Slangen JJ, tive cohort study (Netherlands). Cancer Causes
Knipschild P. Gastric cancer in coal miners:  a Control 2003;14:427–​38.
case-​control study in a coal mining area. Br J Ind Van Loon AJ, Botterweck AA, Goldbohm RA, Brants
Med 1985;42:627–​30. HA, Van Klaveren JD, Van Den Brandt PA. Intake
Swedish Oncological Centres. Atlas of Cancer of nitrate and nitrite and the risk of gastric can-
Incidence in Sweden. Stockholm:  Swedish cer:  a prospective cohort study. Br J Cancer
Oncological Centres, 1995. 1998a;78:129–​35.
Tahara E. Genetic pathways of two types of gastric Van Loon AJ, Goldbohm RA, Van Den Brandt PA.
cancer. IARC Sci Publ 2004;(157):327–​49. Socioeconomic status and stomach cancer inci-
Tahara H, Kuniyasu H, Yokozaki H, Yasui W, Shay JW, dence in men:  results from the Netherlands
Ide T et  al. Telomerase activity in preneoplastic Cohort Study. J Epidemiol Community Health
and neoplastic gastric and colorectal lesions. Clin 1998b;2:166–​71.
Cancer Res 1995;1:245–​51. Wang QB, Chen Y, Wang XL, Gong GQ, Li GP, Li
Takeno S, Hashimoto T, Maki K, Shibata R, Shiwaku CY. Consumption of fruit but not vegetables
H, Yamana I et  al. Gastric cancer arising from may reduce risk of gastric cancer:  results from
the remnant stomach after distal gastrectomy:  a a meta-​analysis of cohort studies. Eur J Cancer
review. World J Gastroenterol 2014;20:13734–​40. 2014;50:1498–​1509.
Tamura G. Alterations of tumor suppressor and tumor- Wang WH, Huang JQ, Zheng GF, Lam SK, Karlberg J,
related genes in the development and progression of Wong BC. Non-​steroidal anti-​inflammatory drug
gastric cancer. World J Gastroenterol 2006;12:192–​8. use and the risk of gastric cancer:  a systematic
24

242 Textbook of Cancer Epidemiology

review and meta-​ analysis. J Natl Cancer Inst Wu H, Rusiecki JA, Zhu K, Potter J, Devesa SS. Stomach
2003;95:1784–​91. carcinoma incidence patterns in the United States
Varley JM, Mcgown G, Thorncroft M, Tricker KJ, by histologic type and anatomic site. Cancer
Teare MD, Santibanez-​ Koref MF et  al. An Epidemiol Biomarkers Prev 2009;18:1945–​52.
extended Li-​Fraumeni kindred with gastric carci- Ye W, Ekstrom AM, Hansson LE, Bergstrom R, Nyren
noma and a codon 175 mutation in tp53. J Med O. Tobacco alcohol and the risk of gastric can-
Genet 1995;32:942–​5. cer by sub-​site and histologic type. Int J Cancer
Vauhkonen M, Vauhkonen H, Sipponen P. Pathology 1999;83:223–​9.
and molecular biology of gastric cancer. Best Pract You W, Zhang L, Gail MH, Chang Y, Liu W, Ma J et al.
Res Clin Gastroenterol 2006;20:651–​74. Gastric dysplasia and gastric cancer: Helicobacter
Weiderpass E, Vainio H, Kauppinen T, Vasama-​ pylori, serum vitamin C, and other risk factors. J
Neuvonen K, Partanen T, Pukkala E. Occupa­ Natl Cancer Inst 2000;92:1607–​12.
tional exposures and gastrointestinal cancers You WC, Brown LM, Zhang L, Li JY, Jin ML, Chang
among Finnish women. J Occup Environ Med YS et al. Randomized double-​blind factorial trial
2003;45:305–​15. of three treatments to reduce the prevalence of
Weinberg GB, Kuller LH, Stehr PA. A case-​control precancerous gastric lesions. J Natl Cancer Inst
study of stomach cancer in a coal mining region 2006;98:974–​83.
of Pennsylvania. Cancer 1985;56:703–​13. Yu G, Gail MH, Shi J, Klepac-​Ceraj V, Paster BJ, Dye
Welling R, Beaumont JJ, Petersen SJ, Alexeeff BA et  al. Association between upper digestive
GV, Steinmaus C. Chromium VI and stom- tract microbiota and cancer-​predisposing states
ach cancer:  a meta-​analysis of the current epi- in the esophagus and stomach. Cancer Epidemiol
demiological evidence. Occup Environ Med Biomarkers Prev 2014;23:735–​41.
2015;72:151–​9. Yuan JM, Ross RK, Gao YT, Qu YH, Chu XD, Yu
Wernli KJ, Fitzgibbons ED, Ray RM, Gao DL, Li MC. Prediagnostic levels of serum micronutrients
W, Seixas NS et  al. Occupational risk factors in relation to risk of gastric cancer in Shanghai,
for esophageal and stomach cancers among China. Cancer Epidemiol Biomarkers Prev
female textile workers in Shanghai China. Am J 2004;13:1772–​80.
Epidemiol 2006;163:717–​25. Zeng M, Mao XH, Li JX, Tong WD, Wang B, Zhang
Virtamo J, Pietinen P, Huttunen J, K, Korhonen P, YJ, Guo G. Efficacy, safety, and immunogenicity
Malila N, Virtanen MJ et al. Incidence of cancer of an oral recombinant Helicobacter pylori vac-
and mortality following alpha-​ tocopherol and cine in children In China: A randomised, double-​
beta-​carotene supplementation:  a postinterven- blind, placebo-​controlled, phase 3 trial. Lancet
tion follow-​up. JAMA 2003;290:476–​85. 2015;386: 1457–​64.
Wong BC, Ching CK, Lam SK, Li ZL, Chen BW et al. Zhang Z, Xu G, Ma M, Yang J, Liu X. Dietary fiber
Differential north to south gastric cancer-​duodenal intake reduces risk for gastric cancer: a meta-​
ulcer gradient in China, China Ulcer Study Group. analysis. Gastroenterology 2013;145:113–​120 E3.
J Gastroenterol Hepatol 1998;13:1050–​7. Zheng W, Sellers TA, Doyle TJ, Kushi LH, Potter JD,
World Cancer Research Fund—​ American Institute Folsom AR. Retinol, antioxidant vitamins, and
of Cancer Research. Food Nutrition and the cancers of the upper digestive tract in a prospec-
Prevention of Cancer:  A Global Perspective. tive cohort study of postmenopausal women. Am
Washington DC: AICR, 2007. J Epidemiol 1995;42:955–​60.
Wu A, H, Yang D, Pike MC. A meta-​analysis of soy- Zhu S, Mason J, Shi Y, Hu Y, Li R, Wahg M et al. The
foods and risk of stomach cancer:  the problem effect of folic acid on the development of stom-
of potential confounders. Cancer Epidemiol ach and other gastrointestinal cancers. Chin Med
Biomarkers Prev 2000;9:1051–​8. J (Engl) 2003;116:15–​9.
243

11
Colorectal Cancer
N A N A K E U M , K A N A W U , E D W A R D G I O V A N N U C C I , A N D D AV I D J .   H U N T E R

C olorectal cancer (CRC), typically adenocarci-

noma, arises from epithelial cells lining the
large bowel or intestine. Colorectal adenomas are
to the ileum) to distally, into the cecum, ascend-
ing colon, hepatic flexure, transverse colon,
splenic flexure, descending colon, and sigmoid
well-​established precursor lesions for the major- colon (which connects to the rectum). By con-
ity of CRCs. Relatively uncommon prior to 1900, vention, the parts of the colon up to and includ-
CRC has become the third most commonly diag- ing the transverse colon are considered the right
nosed cancer, as well as the fourth leading cause sided or proximal colon, and the splenic flexure,
of malignant death globally (IARC/​WHO, 2015). descending, and sigmoid colon are classified as
This cancer is largely preventable. Modifiable the left sided or distal colon (NIH, 2014). Most
causes are demonstrated by the large variation in of the cancers that occur in the large bowel are
incidence across countries, rapid changes in inci- adenocarcinomas, cancers of the intestinal gland
dence within some populations, and the transi- cells producing mucus (Levin & Raijman, 1995),
tion in disease risk for immigrants toward that of and are the focus of this chapter.
the host country in migration studies. Typically, Symptoms Early in its natural history, CRC is
exposure to certain aspects of economic devel- typically asymptomatic. It may be detected at a
opment and the Western lifestyle and diet in a screening colonoscopy or sigmoidoscopy entirely
population leads to a dramatic rise in incidence. without symptoms or signs. As CRC progresses,
A number of lifestyle and dietary factors are now symptoms may develop including change in
established as convincing or probable causes. In bowel habits, such as constipation or diarrhea,
addition, the ability to access and remove adeno- abdominal discomfort, and blood in the stool
mas can lower cancer incidence through second- (often not visually obvious). The mass of the
ary prevention. Thus, a combination of primary lesion, especially when distally located, may
and secondary prevention can greatly lower inci- cause a narrowing of stools. Fatigue and uninten-
dence and mortality from CRC. tional weight loss are ominous signs of advanced
As the etiologies of colon and rectal cancer disease (Tsai & Gearhart, 2011).
overlap largely, this chapter generally considers
Diagnosis Colonoscopy is the main test to diag-
CRC (Table 11-​1). When there is evidence of dif-
nosis CRC. Due to the asymptomatic nature of
ferences between colon and rectal cancer, these
early-​stage CRCs, CRCs tend to be diagnosed in
are pointed out. In addition, in some cases there
advanced stages, with only 40% of CRC patients
may be etiologic differences between proximal
diagnosed at a localized stage (Siegel et al, 2014).
and distal colon cancer and these are described.
Yet, the increasing prevalence of CRC screening
Finally, much has been learned about the molec-
has contributed to increased detection of asymp-
ular underpinnings of CRC. This knowledge is
tomatic CRCs at an early stage.
now beginning to offer new insights into the eti-
ology of CRC. Treatment When the cancer is localized, sur-
gery may be curative; chemotherapy and radia-
CLINICAL SYNOPSIS tion therapy and more recently immunotherapy
Subgroups The large bowel or intestine, also are other modes of treatment (Cunningham
called the colorectum, is generally divided into et al, 2010).
the colon and rectum. The colon can be further Prognosis Not surprisingly, the prognosis of CRC
subdivided, moving proximally (which connects is largely dependent on the stage and grade at
24

TABLE 11-​1   EPIDEMIOLOGY OF CRC: RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Adenomas (adenomatous polyps) (direct precursors) ↑↑
Familial adenomatous polyposis ↑↑
Lynch syndrome (hereditary nonpolyposis CRC) ↑↑
Untreated Inflammatory bowel disease ↑↑
Family history in first-​degree relative ↑↑
Older age ↑↑
Male sex ↑
Taller adult height ↑
Tobacco smoking ↑
Excessive adiposity ↑
Abdominal adiposity (particularly visceral fat) ↑
Physical activity ↓
Western dietary pattern ↑
Alcohol use (in men) ↑
Processed meat intake ↑
Aspirin ↓
Screening endoscopy ↓↓
Probable Relationship Exists, Based on Substantial Data
Serrated polyps ↑↑
Metabolic syndrome ↑
Type 2 diabetes mellitus ↑
Excessive adiposity at early life (in women) ↑
Sedentary lifestyle ↑
Unprocessed red meat intake ↑
Fiber intake ↓
Folate intake ↓
Calcium intake ↓
Vitamin D status ↓
High plasma insulin-​like growth factor 1 levels in adulthood ↑
High plasma insulin levels in adulthood ↑
Ionizing radiation exposure ↑
Alcohol use (in women) ↑
Weak, if Any, Relationship Exists,
Based on Substantial Data
Asbestos exposure ↑
Fusobacterium nucleatum ↑
Inconsistent Findings or Limited Study to Date
Cholecystectomy ↑
Acromegaly ↑
Garlic intake ↓

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate to large
increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association
Note: The magnitude of the risk can vary substantially, depending on the exact comparison being made. For example, having
a family history of CRC in a first-​degree relative is a consistent CRC risk factor. However, the magnitude of the association
increases substantially the earlier the age at diagnosis in the relative(s) and with the number of relatives affected.
245

Colorectal Cancer 245

diagnosis, though more recently, tumor molecu- rates are low in Africa and South-​Central Asia
lar characteristics may add prognostic informa- (Figs. 11-​1, 11-​2). The incidence rates closely
tion. Survival rates do not vary appreciably by mirror economic development in the countries,
anatomic location. In the United States (between increasing in economically transitioning coun-
2005 and 2011), the 5-​year survival rates range tries while stabilizing at high incidence in the
from 90% for stage I cancer to 13% for stage IV majority of high-​income countries (Center et al,
cancer (SEER, 2015). The bulk of recurrence and 2009). Now, about 52% of all CRC deaths occur
mortality occurs within the first 5 years after diag-
in low-​ income countries (IARC/​ WHO, 2015).
nosis. The primary site for metastasis is the liver.There may be inherent ethnic differences in the
Progress With advances in molecular character- rates, but these are relatively subtle and hard to
ization of CRC, ongoing research efforts aim to confirm, given the strong environmental com-
provide evidence for individualized cancer treat- ponent. For example, half a century ago it might
ment tailored to tumor molecular characteristics. have been thought that Asian populations nat-
urally have very low rates. However, following
DESCRIPTIVE economic development, rates have increased
EPIDEMIOLOGY dramatically. Specifically, while rates used to be
CRC is one of the most common cancers world- very low in Japan and South Korea, these two
wide, with approximately 1.4  million cases countries now have among the highest rates in
diagnosed globally each year. Excluding non- the world (especially in men).
melanoma skin cancer, CRC accounts for almost The relatively rapid changes in incidence rates
10% of all cancers diagnosed annually and about over time also strongly point to environmental
8.5% of all cancer deaths worldwide (IARC/​ factors being important. In general, economic
WHO, 2015). This malignancy was relatively development and the introduction of a Western
rare before lifestyle and dietary changes resulting lifestyle to a population inevitably lead to an
from industrialization. Several lines of evidence increase in the rates of CRC. Migration stud-
including geographic variation, secular trends, ies also show the same pattern (Parkin, 2004).
and migration studies indicate a strong environ- For example, over time, CRC incidence rates in
mental component to CRC. Japanese moving to the United States increased,
By geographic region, age-​standardized inci- merging closer to those of the host country
dence rates of CRC vary approximately 10-​fold (Haenszel & Kurihara, 1968). Interestingly,
across the world. In general, the highest incidence even the distribution of anatomic subsites of
rates are observed in Australia/​ New Zealand, CRC shifted toward the pattern of the United
Europe, and Northern America, whereas the States, which had a relatively greater tendency

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 5 10 15 20 25 30 35 40
ASR (world) per 100,000 person-years

FIGURE 11-​1  Age-​standardized (to the 2012 world population) incidence rates of colorectal cancer among women.
Source: Ferlay et al, 2013.
246

246 Textbook of Cancer Epidemiology

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 10 20 30 40 50
ASR (world) per 100,000 person-years

FIGURE 11-​2  Age-​standardized (to the 2012 world population) incidence rates of colorectal cancer among men.
Source: Ferlay et al, 2013.

for left-​sided cancers. Similar patterns have also role in promoting the progression) (Muzny et al,
been observed among Chinese (Flood et  al., 2012; Wood et al, 2007). Specific driver mutations
2000)  and Europeans (Lilienfeld et  al, 1972a, likely underlie the major subtypes of CRC. The
1972b) migrating to the United States and among three key global molecular classifiers for CRC
European immigrants to Australia (McCredie include chromosomal instability (CIN), CpG
et al, 1999). An important observation is that the island methylator phenotype (CIMP), and mic-
immigrants seem to adopt the rates of the new rosatellite instability (MSI) (Carethers & Jung,
country within a few decades or within first-​ 2015; Kang, 2011; Markowitz & Bertagnolli,
generation immigrants (Kune, 2010; Maskarinec 2009; Samowitz, 2008). These pathways suggest
& Noh, 2004). increased genetic and epigenetic alterations such
as chromosomal aberrations, widespread DNA
GENETIC AND MOLECULAR promoter hypermethylation, or genomic insta-
EPIDEMIOLOGY bility, which can ultimately promote colorectal
CRC is not a single disease, but represents a carcinogenesis (Carethers & Jung, 2015; Simons
heterogeneous group of cancers with different et  al, 2013). Susceptibility to the three CRC
genetic and epigenetic alterations (i.e., molecu- subtypes differ by anatomic site, sex, and age
lar profiles) that are sporadic or inherited (Ogino at onset. For example, CIN-​positive CRC tends
& Goel, 2008). The initial characterization of to be more common in the distal colon than in
molecular events along colorectal carcinogen- the proximal colon and in men than in women.
esis was pioneered by Fearon and Vogelstein The CIMP-​high CRC and MSI-​high CRC on the
(Fearon & Vogelstein, 1990; Vogelstein et  al, other hand, are more frequently observed in the
1988). Because of the accessibility to adenoma proximal colon and in women. They also tend to
tissue, CRC was particularly amenable to studies have a later age at onset than CIN-​positive CRC
to identify molecular steps involved in the pro- (Carethers & Jung, 2015).
gression from normal to hyperproliferative epi-
thelium, to early, intermediate, and late adenoma Chromosomal Instability
and finally to CRC. The most common molecular subtype of sporadic
CRC, CIN-​positive CRC, occurs in 60%–​80% of
Somatic Mutations CRC. In this subtype, there are frequent karyo-
Only a limited number of the alterations are con- type abnormalities such as chromosomal gains
sidered to be driver mutations (i.e., genetic aber- and losses. These broad chromosomal aberra-
rations driving cancer progression as opposed to tions strongly indicate mutations in genes criti-
random passenger mutations having no causal cal to maintaining chromosomal integrity during
247

Colorectal Cancer 247

cell division. Both chromosomal gain increas- genes or oncogenes. Indeed, the MSI-​high CRC
ing oncogene expression and chromosomal loss subtype, which encompasses about 10%–​20% of
silencing critical tumor suppressor genes pro- sporadic CRC, is characterized by inactivation of
mote carcinogenesis (Ogino & Goel, 2008). More critical mismatch-​repair genes (e.g., MLH1 gene)
than 90% of CIN-​ positive CRC relates to the (Wheeler et  al, 2000)  and exhibits an increased
Adenomatous Polyposis Coli (APC) gene muta- rate of mutations. Genetic mutations in MSI-​
tion, which is considered as the first step of colo- high CRC not only promote colorectal carcino-
rectal carcinogenesis in the adenoma-​carcinoma genesis but also can result in the production of
sequence (Pino & Chung, 2010; Tanaka, 2009). abnormal proteins that may be recognized by
The APC protein is involved in the Wnt signal- the host immune system as foreign. Referred
ing pathway, by degrading cytoplasmic CTNNB1 to as neoantigens, these proteins tend to elicit
(or beta-​catenin), a key downstream effector of strong antitumor immune responses, which may
the pathway that mediates transcription of crit- explain why MSI-​high CRCs have better progno-
ical oncogenes such as C-​MYC (He et  al, 1998; sis and better responsiveness to immunotherapy
Rennoll & Yochum, 2015). The APC gene muta- such as immune-​checkpoint blockade (Le et  al,
tions lead to an accumulation of beta-​catenin in 2015; Webber et al, 2015).
the nucleus, thereby enhancing oncogene expres- Interestingly, sporadic MSI-​high CRC tends
sion. Once the APC gene is mutated, the progres- to be CIMP-​high. This suggests that hypermeth-
sion from premalignant lesions to CRC occurs ylation is the major mechanism by which DNA
with additional genetic and epigenetic alterations repair genes are silenced in MSI-​high CRC. Of
that activate oncogenes (e.g., KRAS) or inacti- note, most CRCs occurring in individuals with
vate tumor suppressor genes (e.g., TP53) (Pino & Lynch syndrome are also MSI-​high cancers, but
Chung, 2010). unlike sporadic MSI-​ high CRC, the underly-
ing basis is inherited mutations (i.e., germline
CpG Island Methylator Phenotype mutations) in mismatch repair genes (see “Lynch
Often located around the promoter sites of genes, Syndrome” section) (Herman et al, 1998; Muller &
CpG islands are short regions of DNA rich in Fishel, 2002; Poynter et al, 2008).
the dinucleotide CG sequence (i.e., cytosine and
guanine on one DNA strand separated by one Natural History
phosphate). Hypermethylation of CpG islands in The natural history of CRC is fairly well under-
promotor regions of specified genes (e.g., RUNX3, stood, largely because precursor lesions are
CACNA1G, IGF2, MLH1, NEUROG1, CRABP1, detectable and removed through endoscopy.
SOCS1, and CDKN2A) defines CIMP-​high CRC Depending on the type of precursor, CRC has
(Lochhead et al, 2014; Ogino et al, 2006; Ogino three distinctive natural histories, which include
et  al, 2013). DNA methylation is a key epige- adenoma-​ cancer sequence, serrated pathway,
netic modification by which cells control gene and inflammatory bowel disease (IBD) pathway.
expression; methylating promoter CpG islands
of tumor suppressor genes that are not typically Adenoma-​Cancer Sequence
methylated can inactivate the genes, promoting It is now well established that most CRCs (>95%)
carcinogenesis. Accounting for approximately arise in adenomas. Adenomas (or adenomatous
10%–​ 15% of CRC, CIMP-​ high CRCs tend to polyps) are a type of polyp which are benign neo-
have poorer differentiation than other subtypes plasms arising from the mucosal layer of the large
(Ogino et al, 2006; Ogino et al, 2007; Ogino et al, bowel (Neugut et al, 1993). The development of
2013; Samowitz et  al, 2005; Toyota et  al 2000; CRCs in this pathway starts from hyperprolif-
Weisenberger et al, 2006). eration of crypt cells in the colorectum, leading
to abnormal foci called aberrant crypts. Initially
Microsatellite Instability these may form microadenomas but, as genetic
“Microsatellites” refers to short repetitive and epigenetic abnormalities accumulate, they
sequences of DNA. During DNA replication, grow to visible adenomas, which in turn may
failure to correct errors (e.g., addition or dele- progress to CRC. This adenoma-​ carcinoma
tion of nucleotides) can lead to changes in the sequence was first proposed by Hill and Morson
number of repeats, termed MSI. Thus, MSI indi- about 4 decades ago (Hill et  al, 1978). The key
cates an impaired DNA mismatch-​repair system, observations were that foci of cancer were often
which can lead to mutations in tumor suppressor seen in adenomas, and that individuals with
248

248 Textbook of Cancer Epidemiology

syndromes leading to many adenomas were at adenoma-​carcinoma sequence (Rosty et al, 2013;
high risk for CRC. With identification of molec- Snover, 2011; Tadros & Anderson, 2013).
ular events involved in the progression from
normal to adenomas and finally to CRC (see Inflammatory Bowel Disease Pathway
“Somatic Mutations” section), the adenoma-​ A distinct carcinogenesis pathway probably exists
carcinoma sequence is accepted as the major for patients with ulcerative colitis, an IBD (see
pathway of colorectal carcinogenesis. “Inflammatory Bowel Disease” section) (Murthy
Based on histology, adenomas are charac- et  al, 2002; Wong & Harrison, 2001). Some
terized as tubular, tubulovillous, and villous. patients with IBD, particularly those with ulcer-
Between 75% and 90% of adenomas are tubular ative colitis, suffer from chronic inflammation in
adenomas, while villous adenomas account for the large bowel. In some cases, areas exposed to
3%–​10% of the total (Lee, 1995). Overall, less long-​term inflammation can develop dysplasia,
than 10% of adenomas progress to CRC if not which may often lead to CRC without an obvious
removed (Lev, 1990) but the potential to prog- precursor (e.g., adenomas) (Murthy et  al, 2002;
ress varies by histologic patterns of growth, Wong & Harrison, 2001). Genetic alterations
with villous pattern being an adverse indicator. found in IBD-​associated CRC differ from those
Other predictors include larger size (>1 cm in common in other CRCs (Murthy et al, 2002). For
surface diameter), multiplicity, and high-​grade example, mutations in the TP53, IDH, and MYC
dysplasia (Carethers & Jung, 2015; Cotton et al, genes are more common, and mutations in the
1996; Morson, 1984). A small, tubular adenoma APC gene less common, than in sporadic CRC
should not be considered to indicate high risk (Yaeger et al, 2016).
for CRC, while a large, villous adenoma with
high-​grade dysplasia should be considered to Inherited Susceptibility
be a very high-​risk lesion (Cotton et al, 1996; Several genetic factors increase the baseline risk
Morson, 1984). of CRC or vulnerability to the harmful effects of
The time it takes for an adenoma to progress environmental risk factors of CRC.
to CRC is generally thought to be at least 10 years,
though individual progression rates vary greatly Familial Adenomatous Polyposis
(Carethers & Jung, 2015; Lev, 1990). In high-​risk Familial adenomatous polyposis (FAP) is a rare,
countries such as the United States, approxi- autosomal dominant syndrome, estimated to
mately one-​ third of asymptomatic individuals afflict approximately 1 in 8,000 in the popula-
aged 50–​75  years have one or more adenomas tion (Bisgaard et al 1994). FAP accounts for less
(Heitman et al, 2009). than 1% of all CRCs. This syndrome, also called
familial polyposis coli or adenomatous polyposis,
Serrated Pathway is characterized by multiple colorectal adenomas
A different type of polyp is the serrated polyp, throughout the entire colorectum. There is usu-
so named because of the sawtooth pattern of the ally a minimum of 100 adenomas, but up to sev-
crypts in histological cross-​sections. The World eral thousands of adenomas can occur, causing a
Health Organization now recommends classi- carpet-​like appearance. Tragically, the adenomas,
fying serrated lesions into the following three and even sometimes cancer, occur in afflicted
groups, based on histology and genetic altera- individuals in their 20s and 30s. The adenomas
tions:  (1) (simple) hyperplastic polyp (about individually are not believed to be at higher risk
70%), (2)  sessile serrated adenoma and polyp for progression, but because there are so many
(SSA/​ SSP, 25%), and (3)  traditional serrated adenomas, some inevitably progress to CRC by
adenoma (TSA <2%) (Snover, 2011). Of note, age 50 or typically earlier (Fearnhead et al, 2001;
the terms “SSA” and “SSP” are often used inter- Lynch, 1995), with the risk proportional to the
changeably. Recent studies have established that number of adenomas (Debinski et al, 1996). It is
SSA/​ SSP and TSA may have the potential to thus important to identify afflicted individuals
develop into CRC, with large serrated lesions and family members as early as possible and treat
conferring a 2-​to 3-​fold increased risk of CRC with colectomy or proctocolectomy.
(Holme et al, 2015). Due to the distinct molecu- FAP is caused by an inherited genetic muta-
lar features of CRC that has progressed from ser- tion in the APC gene (Fearnhead et  al, 2001;
rated lesions, the serrated pathway is suggested Lynch, 1995). As discussed previously, a somatic
to be an alternative pathway to the traditional mutation in the APC gene is an important early
249

Colorectal Cancer 249

event in the development of the majority of spo- an inherited mutation in DNA mismatch repair
radic CRC (see “Somatic Mutations” section) genes, while sporadic MSI-​high CRC is attributa-
(Fearon & Vogelstein, 1990). Depending on the ble to hypermethylation of promoter CpG islands
specifics of the mutation, differences in pheno- of DNA mismatch repair genes. In approximately
typic expression may occur (Fearnhead et  al, 70% of CRC in individuals with Lynch syndrome,
2001). Of note, extracolonic tumors may occur genetic mutations have been identified in the
in some affected individuals, notably, gastroin- hMLH1 and hMSH2 genes, with mutations in the
testinal cancers in the duodenum (periampullary MSH6 and PMS2 genes accounting for most of
carcinomas) (Fearnhead et al, 2001; Houlston & the rest (Peltomaki, 2016; Wheeler et al, 2000).
Tomlinson, 2001). In a form of FAP called The penetrance of Lynch syndrome has
Turcot’s syndrome, the central nervous system is been estimated to be up to 80% (Wheeler
prone to tumor formation (Hamilton et al, 1995; et al, 2000), varying by which gene is mutated.
Lynch, 1995). Interestingly, the penetrance of mutations in
the hMLH1/​hMSH2 genes is higher in men
Lynch Syndrome, or Hereditary than in women (Mitchell et al, 2002). Although
Nonpolyposis Colorectal Cancer the reasons are unclear, the difference may
Lynch syndrome, previously known as heredi- explain in part the higher risk of CRC in men.
tary nonpolyposis colorectal cancer (HNPCC), is The risk of CRC in individuals with Lynch syn-
an autosomal dominant disease predisposing to drome also varies by other risk factors of CRC.
CRC and sometimes other cancers. When can- Men in general may be exposed to more risk
cer is confined to colorectum, this is called Lynch factors. Estrogen in women may confer protec-
I syndrome. When extracolonic cancers develop, tion against CRC.
this is called Lynch II syndrome, which com-
monly affects the endometrium, ovary, stomach, Family History of CRC
pancreas, small intestine, hepatobiliary tract, in First-​Degree Relatives
ureter, and renal pelvis (Wheeler et  al, 2000). As for many malignancies, a family history of
Despite the difference in the anatomic site of CRC in the first-​degree relatives (parents, sib-
cancer, Lynch I and Lynch II syndrome may not lings, or children) is associated with an increased
actually have distinct genetic causes. The lifetime risk of CRC. The risk further increases when
risk of CRC in individuals with Lynch syndrome CRC develops in a relative at a young age or
is about 70%–​85%, which is not as high in those in multiple relatives (Butterworth et  al, 2006;
with FAP. However, because Lynch syndrome is Johns & Houlston, 2001). In a meta-​analysis of
the most common form of hereditary CRC, it 27 case-​control and prospective studies (Johns &
accounts for a higher proportion (1%–​3%) of total Houlston, 2001), the risk of CRC was increased
CRCs in a population than FAP does (Lynch & 2.25-​fold for individuals with one first-​degree rel-
Smyrk, 1996; Peltomaki, 2016). ative with CRC; the increase was slightly higher
“Nonpolyposis” in HNPCC is somewhat if the relative was diagnosed with colon cancer
of a misnomer, because most of these can- (2.4-​fold) than with rectal cancer (1.9-​fold).
cers do indeed progress through an adenoma Of note, from twin studies, the heritability of
stage. Unlike those with FAP, individuals with CRC was estimated to be 35%, which suggests
Lynch syndrome may have only several polyps. that genetics contributes less than environmental
However, the adenomas in those with Lynch syn- factors to the variation in cancer risk in a popu-
drome are more likely to be villous and severely lation (Mucci et al, 2016). In fact, most CRCs are
dysplastic (Lynch & Smyrk, 1996; Vasen et  al, sporadic rather than inherited, and thus sporadic
1999). Thus, patients with Lynch syndrome have for most CRCs in a family member. In a high-​risk
an accelerated carcinogenic process and usually country such as the United States, about 10% of
develop CRC at an early age, approximately in adults without CRC and 15%–​20% of CRC cases
the mid-​40s. will have a positive family history of CRC.
The vast majority of CRCs associated with
Lynch syndrome are MSI-​ high CRC, and Low-​Penetrance Genes
these have a predisposition for the proximal and Susceptibility to CRC
colon (Wheeler et  al, 2000)  (see “Microsatellite In the candidate gene era, one of the few con-
Instability” section). MSI-​high CRC in individu- sistent associations observed was a lower risk
als with Lynch syndrome is typically caused by of CRC associated with the reduced functional
250

250 Textbook of Cancer Epidemiology

variant of the MTHFR gene (Chen et  al, 1996; (Du et  al, 2014). In general, the presence of
Kennedy et  al, 2012). Recently, genome-​wide G × E interaction in the etiology of CRC sug-
association studies (GWAS) have identified gested by studies based on suspected suscepti-
more than 40 common susceptibility loci associ- bility loci (i.e., candidate gene G × E interaction
ated with CRC incidence (Hsu et al, 2015; Zheng analysis) has not been replicated in studies lever-
et al, 2016). Most studies have been conducted in aging the entire genome (i.e., genome-​wide G × E
persons of European descent and have reported analysis) (Kantor & Giovannucci, 2015). As very
relatively small effect sizes (odds ratios gener- large sample sizes are required to detect genome-​
ally less than 1.2 per risk allele). Given the small wide G × E interactions, further large studies
effect size, individual low-​ penetrant genetic may elucidate more associations.
variants predisposing to CRC only account for
a limited fraction of the estimated heritability of
CRC. However, in aggregation, they may be use- RISK FACTORS
ful for predicting CRC (Dunlop et al, 2013; Hsu Tobacco
et al, 2015). For example, using a risk prediction Carcinogens in tobacco smoke, which could
model incorporating 27 single-​nucleotide poly- reach the colorectal mucosa either through the
morphisms, the genetic risk score increased the digestive tract or through the circulatory sys-
area under the receiver-​operating characteristic tem, contribute to colorectal carcinogenesis by
curve from 0.51 to 0.59 in men and from 0.52 inducing genetic mutations in cancer-​ related
to 0.56 in women, compared to that using only genes (Giovannucci & Martinez, 1996)  or epi-
family history (Hsu et al, 2015). Further identifi- genetic field defects (DeMarini, 2004). CRC was
cation of new genetic susceptibility variants and not considered a tobacco-​ related cancer ini-
incorporation of lifestyle and clinical informa- tially (Doll et  al, 1980; Rogot & Murray, 1980),
tion into the genetic score may increase the pre- but several lines of epidemiologic evidence now
dictive capabilities of such models. strongly support a role for tobacco in the patho-
Given the numerous known or suspected genesis of CRC. First, a meta-​analysis found that
modifiable risk factors (especially dietary fac- CRC risk increased with earlier initiation and
tors) for CRC, it had been hoped that account- longer duration of smoking (Liang et  al, 2009).
ing for gene–​environment (G × E) interaction These findings suggest that smoking may act
could help explain some of the missing herita- on early stages of carcinogenesis (Giovannucci
bility of CRC and offer insights into the modi- et  al, 1994; Giovannucci et  al, 1994). Second,
fiable factors (Kantor et al, 2014; Thomas, 2010). long-​term heavy smokers had an increased risk
In one large consortium of 13 studies (GECCO) of colorectal adenoma, a well-​established precur-
that tested for a G × E interaction between pre- sor lesion for CRC (Giovannucci, 2001). Third, a
viously identified genetic susceptibility loci for positive association between smoking and colo-
CRC and lifestyle risk/​protective factors for CRC rectal neoplasia was pronounced for specific
(e.g., body mass index, alcohol, smoking, aspi- molecular subtypes of CRCs such as CIMP-​high
rin, postmenopausal hormone use in women, and MSI-​high (Curtin et  al, 2009; Limsui et  al,
and various dietary factors), no evidence of an 2010; Poynter et al, 2009) and for serrated lesions
interaction was observed (Kantor et  al, 2014). (Figueiredo et al, 2015). The relative risk for these
In the same consortium, however, the applica- subtypes is at least 2, which is higher than that for
tion of a genome-​wide G × E interaction analysis total CRC or adenomas. Strong associations (e.g.,
discovered that the inverse association of aspirin relative risks of 2)  are less likely attributable to
and CRC risk differed by the genetic variation residual confounding than modest associations
of rs2965667 at chromosome 12p12.3 near the (e.g., relative risks of 1.2), which strengthens evi-
MGST1 gene and rs16973225 at 15q25.2 near the dence for causality.
IL16 gene (Nan et  al, 2015). In a genome-​wide An important question remaining unre-
G × E interaction investigation of dietary factors solved has been whether the adverse effect of
such as meat, vegetables, fruit, and fiber, only smoking on CRC is reversible after cessation
one genome-​wide statistically significant inter- of smoking. Some studies indicated no benefit of
action was found between processed meat intake cessation (Giovannucci et al, 1994a; Giovannucci
and rs4143094 (10p14/​ near the GATA3 gene) et  al, 1994b), but this conclusion has been dis-
(Figueiredo et  al, 2014). No genome-​wide G × puted by other studies (Hannan et al, 2009). The
E interactions were detected for calcium intake inconsistent findings can be explained in part by
251

Colorectal Cancer 251

accounting for tumor subtype. In a recent study, an increased risk of colon cancer (Magalhaes
compared with continuing smokers, smokers et al, 2012).
who quit for at least 10  years had an approx- A priori methods create dietary scores to
imately 50% reduced risk of CRC for CIMP-​ measure adherence to expert dietary guidelines
high subtype, but not for CIMP-​low subtype (e.g., the Healthy Eating Index and the Dietary
(Nishihara et al, et al, 2013). Thus, reversibility Approaches to Stop Hypertension diet score). As
of an elevated CRC risk associated with smok- reviewed by Fung and Brown (Fung & Brown,
ing may be confined to CIMP-​high CRC, which 2013), a lower risk of CRC was consistently
occurs mostly in the proximal colon and at older observed in individuals with a better adherence
ages (Curtin et  al, 2011). Unfortunately, rectal to a plant-​based diet with abundant intakes of
cancer, with which smoking forms a stronger fruits and vegetables; moderate intakes of dairy
association, is rarely CIMP-​ high and tends products; and limited intakes of red and pro-
to develop at considerably earlier ages. Taken cessed meats, refined grains, and added sugar.
together, this evidence indicates that smoking The hybrid approach of a posteriori and a
may have dual effects on colorectal carcinogen- priori methods derives a dietary pattern associ-
esis:  a reversible effect for proximally enriched ated with established or suspected biomarkers of
CIMP-​ high colon cancer occurring at older disease risk. For instance, the C-​peptide dietary
ages and an irreversible effect for rectal cancer pattern, a dietary pattern correlated with high C-​
appearing at earlier ages. peptide (an estimate of hyperinsulinemia) con-
centrations is characterized by higher intakes of
Diet meat, fish, and sweetened beverages and lower
Dietary Patterns intakes of coffee, high fat dairy products, and
Studies of diet and cancer focus on individ- whole grains. Higher C-​peptide dietary pattern
ual food groups, food items, and/​or nutrients. score was associated with an increased colon can-
While it is important to identify single nutrients cer risk (Fung et al, 2012). Similarly, higher die-
or foods that affect CRC risk, specific items may tary inflammatory index (i.e., diet with greater
simply be acting as a marker of an overall die- proinflammatory properties) was associated with
tary pattern that truly influences CRC risk in a a higher CRC risk (Steck et al, 2015). Physiologic
complex way. There are multiple ways to define markers are likely influenced by the broad effects
a dietary pattern such as a posteriori methods of internal dose of overall dietary intakes and
(e.g., factor analysis), a priori methods (e.g., thus, a dietary pattern derived based on biologi-
dietary score), and a hybrid of the two meth- cal pathways rather than eating behaviors is likely
ods (e.g., reduced rank regression) (Hu, 2002). more predictive of CRC.
In addition, groups following a special diet may
also be studied (e.g., vegans, vegetarians) (Orlich Red Meat and Processed Meat
et al, 2015). Of note, there is considerable over- Higher intakes of red meat and processed meat
lap across different dietary patterns. For example, are considered important risk factors for CRC.
various patterns representing a “prudent” diet In the 2007 report, the World Cancer Research
may commonly involve high consumption of Fund/​ American Institute for Cancer Research
fruits, vegetables, and whole grains. Thus, these (WCRF/​AICR) panel concluded that both red
dietary patterns may provide similar information meat (e.g., beef, pork, lamb, and goat) and pro-
in some circumstances. cessed meat (any meats, usually red meats,
Principal component analysis, an example of preserved by smoking, curing, salting, or add-
a posteriori methods, has identified a so-​called ing chemical preservatives such as nitrites and
Western dietary pattern. This pattern has gen- nitrates:  e.g., bacon, sausages, hot dogs, ham,
erally been characterized by high consumption salami, and pepperoni) are “convincing” causes
of red meats, processed meats, refined grains, of CRC (WCRF/​ AICR, 2007). In 2015, the
high-​ fat dairy products, sugar-​ sweetened bev- International Agency for Research on Cancer
erages, and French fries (Fung & Brown, 2013). (IARC) Monograph Working Group declared the
While the specifics varied across studies, this die- consumption of processed meat as “carcinogenic
tary pattern has emerged in many populations, to humans (Group 1) based on sufficient evidence
especially in North American populations. In a in humans that the consumption of processed
recent meta-​analysis of 11 observational studies, meat causes colorectal cancer,” and the consump-
a Western dietary pattern was associated with tion of red meat as “probably carcinogenic to
25

252 Textbook of Cancer Epidemiology

humans (Group  2A) based on limited evidence suggesting that cooking of chicken may some-
that the consumption of red meat causes cancers times produce even higher levels of some HCAs
in humans and strong mechanistic evidence” than red meat (Byrne et  al, 1998); yet chicken
(Bouvard et al, 2015). In contrast, other sources consumption is not associated with CRC risk.
of animal protein including low-​fat dairy prod- Some studies examined potential interactions
ucts, fish, and poultry have not been generally of meat intake with genotypes or phenotypes
associated with CRC risk, and in some studies of xenobiotic metabolizing enzymes, such as
these have been inversely associated with the risk NAT1 and NAT2, but these have not yielded con-
(Carr et al, 2016; Shi et al, 2015). sistent findings (Ananthakrishnan et  al, 2015;
It is noteworthy that the level of evidence Andersen et al, 2013; Andersen & Vogel, 2015).
for red meat has somewhat weakened over time, The stronger evidence for processed meat sug-
while that for processed meat has remained gests a role for N-​nitroso compounds (NOCs),
strong. For example, from the 2007 WCRF/​AICR known carcinogens in experimental models of
report to the 2011 updated meta-​analysis that colorectal carcinogenesis. Processed meats are
included 10 additional studies (Chan et al, 2011), often preserved with nitrate or nitrite, which are
the estimated increase in CRC risk per 100 g/​day precursors for NOCs (Cross & Sinha, 2004; Loh
intake of red meat reduced from 29% to 17%. In et al, 2011). To date, only limited epidemiologic
contrast, the risk increase per 50 g/​day intake studies have examined dietary intakes of nitrate,
of processed meat did not change appreciably nitrite, and NOCs in relation to CRC risk and
(from 21% to 18%). Other recent meta-​analyses findings are inconsistent (Cross et al, 2010; Cross
have reached similar conclusions, supporting a & Sinha, 2004; Dellavalle et al, 2014; Knekt et al,
stronger association of CRC risk with processed 1999). Measurement error in dietary assessment
meat than with red meat (Alexander et al, 2010; may be a potential explanation for the discrepant
Hjartaker et al, 2013; Huxley et al, 2009; Johnson findings. Additionally, endogenous production
et  al, 2013; Larsson & Wolk, 2006; Norat et  al, of NOCs could be induced by foods other than
2002; Pham et  al, 2014; Sandhu et  al, 2001). Of processed meats; facilitated by heme iron in red
note, there is some suggestion that the positive meat; inhibited by other dietary factors includ-
association may be more pronounced for distal ing calcium, vitamin C, and some polyphenols;
colon and rectal cancers than for proximal colon or influenced by the composition of gut bacterial
cancer. However, because many studies did not flora (Cross & Sinha, 2004; Hord et al, 2009). All
report the results by anatomic subsites, there is of the aforementioned modifiers complicate the
considerable potential for publication bias to interpretation of the findings across the studies.
explain the heterogeneous findings (Hjartaker
et al, 2013). Folate (Vitamin B9)
The mechanisms linking red meat and pro- Folate, as a carrier of one-​ carbon units, plays
cessed meat to CRC risk have not been estab- essential roles in nucleic acid synthesis (e.g., DNA
lished. Meat is a major source of saturated fat and synthesis) and methylation reactions (e.g., DNA
protein, but other major sources of these nutri- methylation). Because aberrations in DNA synthe-
ents have not been associated with an increased sis and DNA methylation are implicated in carcin-
risk of CRC. Recent hypotheses have turned to ogenesis, poor folate status has been hypothesized
unique constituents of meat such as meat muta- to increase the risk of CRC (as well as other can-
gens. Heterocyclic amines (HCAs) are carcino- cers). The folate hypothesis has received consider-
genic compounds formed when meat is cooked able study, and epidemiologic data tend to support
at high temperatures for long durations (Sinha an inverse association between folate intake and
& Norat, 2002; Sinha et al, 2005). To date, there CRC risk. In a pooled analysis of 13 prospective
is no established biomarker of HCAs for use in studies, total folate intake of ≥560 mcg/​day, com-
large epidemiologic studies. Thus, most stud- pared to that of <240 mcg/​day, was associated with
ies have estimated exposure to HCAs through a 13% lower risk of colon cancer (Kim et al, 2010).
questionnaires inquiring about cooking prac- A functional polymorphism in the gene encoding
tices of meat (Cross et  al, 2010; Ollberding methylenetetrahydrofolate reductase, an enzyme
et  al, 2012). While the studies have not yielded that controls the balance between different forms
convincing results, it is unclear if measure- of folate for DNA synthesis and DNA methylation,
ment error obscures any true association. The is generally inversely associated with CRC risk
HCAs hypothesis is also inconsistent with data (Kennedy et al, 2012).
253

Colorectal Cancer 253

However, in the recent decade, the folate been observed in patients with adenomas: higher
hypothesis has been questioned. Based on some uracil misincorporation into DNA, higher global
animal data and the ability of folate to provide DNA hypomethylation, and lower colonocyte
DNA precursors, some have proposed that a folate level were seen in histologically normal
high folate intake may promote the growth and tissue adjacent to the polyp compared to that
progression of premalignant neoplastic lesions distal to the polyp (McGlynn et  al, 2013). An
(Kim, 2006; Mason et  al, 2007; Osterhues et  al, early phase of colorectal carcinogenesis at which
2009; Ulrich, 2007). In broad terms, this hypoth- folate appears to exert an effect and consequent
esis suggests a potential dual role of folate:  at requirement of a long-​time lag, perhaps decades,
early stages of colorectal carcinogenesis (before to observe an effect of folate status on CRC risk
adenoma development), folate deficiency may may explain why existing RCTs of CRC, which
increase the risk; but at later stages, for exam- only ran for about 5 years or less, have not seen
ple in the presence of large, advanced adenomas a benefit of folic acid supplementation. Of note,
and early-​stage CRC, excessive folate may pro- in a recent RCT conducted in China for which
vide nutrients supporting the growth of tumors. adenomas was the primary endpoint, a benefit of
Consistent with the putative dual role of folate folic acid supplementation on the primary pre-
in colorectal carcinogenesis, the WCRF/​ AICR vention of adenomas was observed after 3 years
panel concluded that evidence is too inconsistent of intervention (Gao et al, 2013).
to declare a benefit of folate (WCRF/​AICR 2011).
Moreover, in a recent meta-​analysis of random- Calcium Intake
ized controlled trials (RCTs), no benefit of folic Based on considerable evidence from prospective
acid supplementation on CRC was observed dur- observational studies, RCTs, and mechanistic
ing an average of 5.2 years of follow-​up (Vollset studies, the 2011 WCRF/​AICR report concluded
et al, 2013). that calcium “probably” protects against CRC
It is worth to note that several lines of evi- (WCRF/​ AICR, 2011). Observational studies
dence refute the hypothesis that a high folate generally support an inverse association of cal-
intake may increase the risk of CRC. First, most cium intake with the risk of colorectal adenomas
long-​term observational studies do not suggest a (particularly high-​risk adenomas) (Keum et  al
positive association between a long-​term folate 2015)  and CRC (Keum et  al, 2014). For both
intake and CRC risk. Second, population-​wide endpoints, the dose-​response curves were fairly
increases in folic acid intake in the United States, linear, though there was an apparent flattening
first from supplementation and then from for- at higher calcium intakes. For high-​risk adeno-
tification, were not associated with an increase mas (≥1 cm in diameter, (tubulo)villous growth
in CRC incidence rates (Keum & Giovannucci, pattern, high-​grade dysplasia, or multiplicity),
2014). Third, a meta-​analysis of 13 RCTs of folic compared to 550 mg/​day of total calcium intake,
acid supplementation at considerably high doses the risk was reduced by 23% at 1,000 mg/​day and
(500–​40,000 mcg/​day across the studies) did not by 31% at 1,450 mg/​day (Keum et al, 2015). For
lead to an increased (nor decreased) risk of CRC CRC, compared to 250 mg/​day of total calcium
during the first 5  years of intervention (Vollset intake, the risk was reduced by 18% at 1000 mg/​
et al, 2013). day and by 26% at 1,750 mg/​day (Keum et  al,
A critical element in understanding the role 2014). An inverse association was observed for
of folate in colorectal carcinogenesis may be the both dietary and supplemental calcium and
timing of folate exposure. The Nurses’ Health across diverse populations.
Study and Health Professionals Follow-​up Study Results from RCTs have been mixed. Initial
cohorts are the only cohort studies with multiple RCTs for recurrent adenomas were promising.
repeated measures of diet over several decades. In the Calcium Polyp Prevention Study, cal-
In both cohorts, a statistically significant inverse cium supplementation of 1,200 mg/​ day over
association emerged when a time lag of at least 4  years appeared to lower the risk of histologi-
12–​16  years was assumed between folate intake cally advanced neoplasms (Wallace et  al, 2004).
and CRC diagnosis (Lee et al, 2011). In view of In a meta-​analysis of RCTs that included trials
this finding, folate deficiency may be etiologi- published up to 2002, subjects assigned to take
cally relevant early in the carcinogenesis process 1,200–​ 2,000 mg/​day of calcium supplements
(e.g., before adenoma formation). Indeed, field over 3–​4  years had a 20% reduced risk of ade-
defects that precede adenoma formation have noma recurrence compared to those assigned to
254

254 Textbook of Cancer Epidemiology

receive placebo (Shaukat et  al, 2005). However, the northern United States, where exposure to
a recent RCT of calcium supplementation (alone solar ultraviolet B radiation, the major source
or in combination with vitamin D supplement) of vitamin D for most people, is low (Garland
did not find a benefit against recurrence of total & Garland, 1980). In the subsequent decades,
or advanced adenomas (Baron et  al, 2015). No extensive epidemiologic studies have been con-
clear explanation is available for the inconsist- ducted to examine the chemopreventive poten-
ent findings among the RCTs of adenomas. In tial of vitamin D against CRC (Giovannucci,
the recent RCT, a protective association was 2009, 2013). The findings have been generally
observed only among those who were not over- supportive of an association. Additionally, stud-
weight or obese (Baron et al, 2015); yet, obesity ies that measured circulating 25-​(OH) vitamin
does not appear to modify the effect of calcium D (25(OH)D) concentrations, a marker of over-
on CRC risk (Keum et  al, 2017). For the CRC all vitamin D status, suggest that the optimal
endpoint, a meta-​analysis of eight RCTs did not levels for CRC protection (if causal) may be in
find a protective effect of calcium supplements the range of 30–​40 ng/​mL (Giovannucci, 2013).
(Bristow et al, 2013). Yet, the trials included had Of note, determinants of 25(OH)D overlap with
insufficient follow-​up times, failing to account many lifestyle factors associated with CRC risk.
for a potentially long latency period between an Two major examples are staying lean and being
increased calcium intake and change in CRC risk. physically active outdoors, which are associated
Experimental studies conducted in both ani- with higher 25(OH)D levels as well as lower CRC
mals and humans suggest a number of poten- risk. Thus, confounding by adiposity and phys-
tial mechanisms. The initial hypothesis was ical activity is a major threat to the validity of
that calcium binds and inactivates carcino- studies on 25(OH)D levels and CRC risk.
genic secondary bile acids or ionized fatty acids Although evidence for a role of vitamin D in
in the colorectal lumen (Newmark et  al, 1984; CRC prevention seems strong, an RCT is required
Wargovich et al, 1984). A series of animal stud- to prove causality. There are ongoing RCTs of
ies indicated that calcium directly reduces cell vitamin D supplementation that may provide
proliferation and promotes cell differentiation some evidence for its chemopreventive poten-
(Buset et  al, 1986; Lamprecht & Lipkin, 2001). tial against CRC (Pradhan & Manson, 2016),
Interestingly, in the normal-​appearing mucosa although the required follow-​up period for the
of adenoma patients, calcium supplementation cancer endpoint is unknown and may be long.
of 2,000 mg/​day appeared to induce favorable For adenoma recurrence, an RCT of daily sup-
changes on the expression of APC and beta-​ plementation with 1,000 IU of vitamin D (alone
catenin (Ahearn et al, 2012), which are important or in combination with calcium supplementa-
in the early stages of colorectal carcinogenesis tion) over a 3-​to 5-​year period did not observe
(see “Chromosomal Instability” section). This a beneficial effect (Baron et  al, 2015). The find-
study suggests that calcium may act on the pre- ing is disappointing, because the mean increase
adenoma phase and thus, a long follow-​up per- in serum 25(OH)D levels in the treatment group
iod (at least 10 years) is required to observe the compared to that in the control group was 7.83
benefit of calcium intake on CRC risk (Keum ng/​mL, which was expected to yield a relative
et al, 2014). Finally, calcium appeared to normal- risk between 0.80 and 0.88 based on observa-
ize increased fecal NOCs due to processed meat tional studies. Yet, the negative finding should be
consumption in humans, thereby reducing their interpreted cautiously, accounting for limitations
potentially carcinogenic effects (see “Red Meat of using recurrent adenomas as the endpoint (see
and Processed Meat” section) (Pierre et al, 2013; “Chemoprevention” section).
Santarelli et al, 2013).
Fiber
Vitamin D Status Almost half a century ago, upon observing a
While vitamin D has been proposed as a protec- low CRC incidence in East Africa where die-
tive agent against numerous cancer sites (Feldman tary fiber intake was high, Burkitt proposed that
et al, 2014), to date, epidemiologic data are most fiber may protect against CRC (Burkitt, 1969).
supportive for CRC. A  potential link between This hypothesis, while appealing given the obvi-
vitamin D and CRC was first proposed in 1980 ous effects of fiber on the colonic microbiota and
by Garland and Garland following their obser- function, has been difficult to confirm in epide-
vation of higher colon cancer mortality rates in miologic investigations. Nonetheless, the recent
25

Colorectal Cancer 255

report coordinated by WCRF/​AICR concluded isolated from foods, which can help increase fiber
that evidence for a protective effect of foods con- consumption, may not be as beneficial as fiber in
taining dietary fiber on CRC risk is “convincing” the original foods. Second, fiber is not a single
(WCRF/​AICR, 2011). This conclusion was largely nutrient but there are various types (e.g., pectin,
based on a dose-​response meta-​analysis of pro- gum, mucilage, psyllium, inulin, cellulose, lig-
spective observational studies, which found an nin, etc.) of fiber, each of which may contribute
approximately 10% reduced risk of CRC for a 10 to health in unique ways. Thus, fiber-​rich whole
g/​day–​increment of total fiber intake (Aune et al, foods containing a mixture of different fibers ver-
2011). By food sources of fiber (i.e., fruits, vegeta- sus foods with added fibers of limited variety are
bles, cereals, and legumes) cereal fiber appeared unlikely to confer the same health benefit. Third,
most strongly associated with a lower risk of CRC in light of recent strong findings concerning
(Aune et al, 2011). (abdominal) obesity, physical activity, sedentary
However, many large well-​conducted stud- lifestyle, smoking, alcohol, and overall dietary
ies do not show a clear association between fiber patterns, fiber is unlikely to be a dominant deter-
intake and CRC risk, after adjustment for impor- minant of CRC risk. Nevertheless, even if a
tant confounders such as physical activity and specific role of fiber in CRC development is ques-
folate and calcium intakes (Michels et  al, 2005; tionable, high fiber intake can be considered as
Otani et  al, 2006; Schatzkin et  al, 2007; Wakai a part of healthy dietary patterns that can help
et al, 2007). Similarly, in a large pooling project prevent CRC (see “Dietary Patterns” section). In
of 13 cohort studies, an association was observed conclusion, fiber may have a role in CRC preven-
in age-​adjusted analyses, but not in multivari- tion, but its contribution is likely to be smaller
able analyses except for a moderately elevated than that initially hypothesized by Burkitt.
risk of CRC among individuals with very low
fiber intakes (<10 g/​ day) (Park et  al, 2005). Garlic
Furthermore, most RCTs of fiber supplements Experimental studies have indicated potential
have not supported an association with recur- mechanisms by which garlic consumption may
rent adenomas (Baron, 2005). Yet, whether fiber help prevent CRC (Ngo et al, 2007). The WCRF/​
supplements would have the same effect as foods AICR panel judged garlic intake to be a “proba-
rich in fiber is debatable (see the last paragraph in ble” protective factor against CRC (WCRF/​AICR,
“Fiber” section). 2011), though there were a limited number of
Despite the mixed results of epidemiologic cohort studies at the time of the review. Some
studies, several mechanisms have been suggested subsequent prospective data are not supportive
for a protective effect of fiber on CRC develop- of a benefit (McCullough et al, 2012; Meng et al,
ment. The role of fiber on colorectal function is 2013). In recent meta-​analyses, an inverse associ-
clear. Insoluble fibers, which increase stool bulk ation between garlic intake and CRC risk was not
and shorten stool transit time, may reduce or supported by cohort studies, though some case-​
dilute exposure from carcinogens (Kritchevsky, control studies suggested a benefit (Chiavarini
1995). For soluble fibers that are fermentable by et al, 2016; Turati et al, 2014).
the anaerobic intestinal microbiota, the break-
down products including short-​chain fatty acids Alcohol
(e.g., acetate, propionate, and butyrate) may play Alcohol (specifically, ethanol) consumption is
a role in preventing CRC. Butyrate, in particular, considered an established risk factor for CRC by
may suppress the growth of colorectal neoplas- IARC. The WCRF/​AICR report concluded that
tic cells by inhibiting proliferation and induc- “high alcohol consumption probably increases
ing apoptosis (Kritchevsky, 1995; Sebastian & the risk of cancers of the colon and rectum” and
Mostoslavsky, 2014). that the association is likely to be “related to total
In assessing the relationship between fiber ethanol intake, irrespective of the type of drink”
consumption and CRC risk, several points are (WCRF/​AICR, 2007). Recent meta-​analyses con-
important to consider. First, the patterns of fiber sistently showed a positive association between
consumption in modern times are different than alcohol consumption and CRC risk (Cai et  al,
the patterns in the preindustrial era:  the range 2014; Fedirko et al, 2011; Jayasekara et al, 2016).
of fiber intake is much lower and fiber is more In one meta-​analysis, light drinking (≤1 drink/​
processed (e.g., crushing and drying), which day) was not associated with an increased risk
may lower the quality of fiber. Fiber supplements of CRC, but moderate (2–​ 3 drinks/​ day) and
256

256 Textbook of Cancer Epidemiology

heavy (≥4 drinks/​day) drinking was associated women with nulliparous women; there was no
with a 21% and 52% increased risk, respectively dose-​response apparent with increasing parity.
(Fedirko et  al, 2011). In the Pooling Project of However, a prior meta-​analysis did not observe
Prospective Studies of Diet and Cancer, indi- an association with parity or any evidence of a
viduals who consumed ≥45 g/​day of alcohol (i.e., dose-​response based on results from 11 cohort
ethanol content in three drinks/​day in the United studies (Guan et al, 2013). In a meta-​analysis of
States) had a 42% elevated risk of CRC compared 11 case-​control studies and 11 cohort studies no
to those who abstained from alcohol (Cho et al, association was observed between age and men-
2004). Overall, the association was stronger arche and CRC risk (Lu et al, 2013).
among men than among women (Fedirko et al, A reduction in risk associated with oral con-
2011), which may be in part due to higher levels traceptive use has also been reported in some
of alcohol intake in men (even within a speci- studies; in the Women’s Health Initiative study
fied category of drinking). An increased risk of ever versus never use was associated with a mul-
CRC associated with heavy drinking was notably tivariate relative risk of 0.74 (95% CI 0.63–​0.86),
pronounced in studies conducted in Asia (81% althoughno duration response was apparent.
increased risk) (Fedirko et al, 2011). However, in the larger Nurses’ Health Study
The carcinogenicity of alcohol is established. cohorts essentially no association was seen for
The first metabolite of ethanol, acetaldehyde, is a ever use compared with never use (Charlton
known carcinogen (Jokelainen et  al, 1997; Seitz et al, 2015).
& Becker, 2007). Once consumed, ethanol is rap-
idly absorbed from the small intestine into the Anthropometric Measures
blood stream and circulating ethanol reaches the Metabolic syndrome and its individual com-
colonic mucosa. Bacteria oxidizes the ethanol to ponents (e.g., hypertriglyceridemia, low HDL
acetaldehyde (Jokelainen et  al, 1997; Salaspuro, cholesterol, hyperglycemia, insulin resistance,
2009; Seitz & Becker, 2007), which accumulates and hyperinsulinemia) are consistently asso-
in the colonocytes. The accumulated acetalde- ciated with CRC, especially colon cancer (see
hyde is implicated in colorectal carcinogenesis “Metabolic Syndrome and Hyperinsulinemia”
by damaging DNA and promoting cell growth section). The key underlying cause of meta-
(Jokelainen et al, 1997; Seitz & Becker, 2007). By bolic syndrome is a positive energy balance (i.e.,
this mechanism, Asian populations, who have a greater energy intake than energy expenditure),
high prevalence of the genetic variant that leads which leads to excess adiposity. In epidemiologic
to a slower elimination of acetaldehyde, are par- studies, adiposity has been assessed with a num-
ticularly vulnerable to the detrimental effect of ber of related and complementary metrics: body
alcohol on CRC. mass index (BMI), which captures overall body
Another mechanism by which the accumu- fatness; waist circumference (WC) or waist-​to-​
lated acetaldehyde contributes to colorectal car- hip ratio (WHR), which reflect overall or rela-
cinogenesis is through interactions with folate tive abdominal fatness; and adult weight change,
(see “Folate” section), such as impairing intesti- which estimates time-​integrated fat accumula-
nal folate absorption and destroying intracellu- tion. All of the measures have been positively
lar folate (Giovannucci, 2004). This is consistent associated with colon cancer and, to a lesser
with the finding that CRC risk was particularly degree, with rectal cancer. Yet, evidence indicates
high when folate intake was low and alcohol abdominal fatness better captures colon cancer
intake was high (Giovannucci, 2004). Some evi- risk than overall body fatness. For instance, an
dence suggests that adverse effects of alcohol on elevated colon cancer risk associated with WC
CRC may be stronger in those with a family his- (or WHR) persisted even after statistical adjust-
tory of CRC (Cho et al, 2012). ment for BMI (MacInnis et  al, 2004; Pischon
et al, 2006). A gain in WC during adulthood was
Reproductive Factors a particularly strong risk factor for colon cancer,
A reduced risk of CRC associated with par- particularly in men (Aleksandrova et  al, 2013;
ity has been reported in some studies, most Song et al, 2016).
recently from the Womens’ Health Initiative While WC and WHR are measures of
Observational Cohort (Murphy et  al, 2016)  in abdominal fatness, they cannot distinguish vis-
which a multivariate relative risk of 0.82 (95% CI ceral adiposity tissue (VAT), which surrounds
0.67–​1.00) was observed comparing ever parous the internal organs, from subcutaneous adipose
257

Colorectal Cancer 257

tissue (SAT), which is located beneath the skin. appears to be an important time window. Body
The distinction between VAT and SAT is impor- fatness in early life was a stronger risk factor for
tant, because VAT is more strongly associated CRC in women than in men (Zhang et al, 2015).
with insulin resistance and thus, metabolically In a Mendelian randomization study, genetic
more harmful (Ibrahim, 2010; Kaaks et al, 2002; determinants of BMI, which are mainly pre-
Wajchenberg, 2000). This distinction could be dictive of obesity at younger ages, were more
especially important at relatively low BMIs (i.e., strongly associated with CRC risk in women than
within the normal or only slightly elevated range in men (Zhang et al, 2015). All of the aforemen-
of BMIs such as <27.5  kg/​m2), where individu- tioned heterogeneous patterns by sex may have a
als with relatively low WC could still have sub- hormonal basis (Lin et al, 2013), but the specific
stantial levels of VAT. If insulin resistance and reason is yet to be elucidated.
accompanying hyperinsulinemia are the prin-
cipal factors elevating CRC risk (see “Metabolic Infections
Syndrome and Hyperinsulinemia” section), more The large intestine is colonized by a wide vari-
direct measures of VAT could better predict CRC ety of microorganisms, the majority of which are
risk. Indeed, in a study that measured VAT and bacterial species. Consisting of both beneficial
SAT using computed tomography, VAT, but not and harmful organisms, the intestinal microbi-
SAT, predicted the risk of advanced colorectal ota plays a pivotal role in host health and dis-
adenomas (Kang et  al, 2010; Keum et  al, 2015; ease conditions by influencing host physiology,
Nagata et  al, 2014). Furthermore, when BMI, metabolism, and the immune system. A growing
WC, and VAT were simultaneously adjusted for, body of evidence suggests that specific bacterial
only VAT significantly predicted the occurrence pathogens and imbalance in the composition of
of adenomas (Yamaji et al, 2014). Of note, many the bacterial microbiota (i.e., dysbiosis) may be
of the studies on VAT were conducted among implicated in the pathogenesis of CRC (Burnett-​
Asians, whose overall BMI was much lower than Hartman et  al, 2008; Schwabe & Jobin, 2013).
that in North American and European popula- One of the most investigated factors includes
tions. Although the majority of the study par- Fusobacterium nucleatum. While F. nucleatum is
ticipants were in a relatively low range of BMI a bacterial strain commonly found in the mouth,
(<27.5 kg/​m2), excess VAT was a strong risk fac- studies have demonstrated that F.  nucleatum
tor for CRC. Therefore, VAT may serve as a clue are enriched in colorectal neoplasms compared
to explain why Asians have high incidence rates with adjacent normal tissue; CRC enriched with
of colorectal neoplasia despite having a relatively F.  nucleatum was associated with an increased
low BMI (Ning et al, 2010a). CRC mortality (Mima et al, 2015). Additionally,
It is noteworthy that several aspects of the rela- there was suggestive evidence for an increased
tionship between excess adiposity and CRC risk risk of CRC for women with moderate or severe
differ between men and women. First, although periodontal disease whose well-​known patho-
excess adiposity is a risk factor for CRC for both gens include F. nucleatum (Momen-​Heravi et al,
sexes, the association is consistently stronger 2017). According to animal and experimental
(about two-​fold higher) in men than in women studies, F.  nucleatum in the mouth may travel
(Ning et  al, 2010a). This sex difference may be through the bloodstream to reach colorectal
in part because men have a tendency toward neoplasms (Abed et al, 2016) and promote colo-
abdominal obesity and higher VAT (Despres, rectal carcinogenesis by activating the oncogenic
2006; Geer & Shen, 2009). Second, the timing Wnt signaling pathway and inhibiting antitumor
at which adiposity exerts an effect on colorectal immune response (Kostic et al, 2013; Rubinstein
carcinogenesis seems to differ. For men, adult- et al, 2013).
hood seems to be an etiologically important time
window. Greater increases in weight (Keum et al, Physical Activity
2015)  and WC (Song et  al, 2016)  that occurred Available evidence consistently shows that a
in adulthood were much stronger risk factors higher level of physical activity is associated with
of CRC in men than in women. A  similar pat- a lower risk of colon cancer as well as colon ade-
tern was suggested in a trajectory analysis, which noma (especially large and advanced adenoma)
examined patterns of weight gain over the life (Boyle et al, 2012; Wolin et al, 2009; Wolin et al,
course (Song et  al, 2016). On the contrary, for 2011). These findings suggest that physical activ-
women early life (childhood and/​or adolescence) ity inhibits the growth of adenomas into cancer.
258

258 Textbook of Cancer Epidemiology

Although there is no definitive data from an RCT, This finding suggests that some of the beneficial
this association is generally considered as causal effects of physical activity are independent of
in nature due to its consistency (WCRF/​AICR, adiposity. Of note, controlled trials showed that
2007). For example, while recreational physical physical activity, particularly high-​intensity exer-
activity is the type of activity most investigated, cise, reduces VAT (Irving et al, 2008; Ross et al,
due to its association with affluence and a healthy 2000; Ross et al, 2004) and VAT loss is achieved
lifestyle that could affect CRC risk, confounding even in the absence of measureable weight loss
is a threat to the causal validity. Yet, an associa- (Ross et al, 2000; Ross et al, 2004). Thus, if VAT
tion has been observed in activities in different is the parameter of adiposity that is most rele-
domains (i.e., occupation, transportation, and vant to colorectal carcinogenesis, studies using
household), which are likely to have different, BMI as the sole measure of adiposity may lead to
sometimes opposing, patterns. Thus, confound- misleading conclusions on the degree to which
ing by other healthy lifestyle factors seems adiposity mediates the effect of physical activity
unlikely to entirely account for the observed asso- on colon cancer risk.
ciation between physical activity and CRC risk.
An association is also observed among diverse Sedentary Lifestyle
subgroups defined by sex, ethnicities, socioeco- As described previously, incidence rates of CRC
nomic status, and countries, which argues against are high in economically developed countries.
confounding by these stratifying factors. In fact, In high-​income countries, the majority of indi-
in the WCRF/​AICR report, the level of evidence viduals lead a mostly sedentary lifestyle:  occu-
supporting a causal role of physical activity is pations generally do not require considerable
concluded as “convincing” only for colon cancer physical activity; cars, buses, and subways are
out of all cancer sites (WCRF/​AICR, 2007). major means for commuting; and watching TV
In observational studies, physical activity is represents a major form of leisure time activity.
assessed mostly by questionnaires or estimated Thus, individuals spend a considerable propor-
crudely by occupational status. The heterogeneity tion of their waking hours engaging in seden-
in methods for assessing physical activity across tary behavior, which is typically characterized
studies makes it difficult to precisely summarize by prolonged (unbroken) sitting. By definition,
the association by levels and types of activities. sedentary behavior is distinct from physical
Yet, according to meta-​analyses, the summary inactivity, which describes the absence of struc-
estimate comparing the highest and lowest level tured and purposive exercise. Thus, a person
of activity suggests a 20%–​30% reduction in risk can be both sedentary and physically active. In
of colon cancer (Wolin et al, 2009) and of large/​ high-​income countries, people engage in physi-
advanced adenoma (Wolin et  al, 2011). Studies cal activity mostly through recreational exercise,
based on a composite score incorporating inten- but even those in the most active category do not
sity, duration, and frequency of activity suggest spend more than 5%–​10% of their waking hours
a dose-​response relationship (Giovannucci et al, in purposive exercise and the rest of the time are
1995; Martinez et  al, 1997), which strengthens likely to be occupied with sedentary behaviors.
evidence for causality. In recent years, there has been added focus on
A variety of mechanisms have been proposed sedentary behavior in cancer epidemiologic
to explain the association between physical research. Within a generally sedentary popula-
activity and colon cancer risk. Physical activ- tion, it is challenging to demonstrate sedentary
ity may lower chronic inflammation, enhance lifestyle as an individual risk factor. Nonetheless,
immunity (e.g., natural killer cells), reduce prolonged sitting has emerged as a risk fac-
insulin resistance and hyperinsulinemia, and tor for CRC, which appears to be independent
stimulate colonic motility (Harriss et  al, 2007; of adiposity measures and structured physical
Slattery, 2004). While each of these mechanisms activity (Howard et al, 2008; Keum et al, 2016).
has some support, none has been proven. An In a recent meta-​analysis that pooled risk ratios
important question is whether physical activity comparing the highest versus lowest level of sed-
decreases colon cancer risk in part by reducing entary time, colon cancer risk increased by 54%
adiposity. Generally, the association between with TV viewing time, by 24% with occupational
physical activity and colon cancer risk was not sitting time, and by 24% with total sitting time
entirely eliminated when controlled for BMI (Schmid & Leitzmann, 2014). It is important to
(Giovannucci et al, 1995; Martinez et al, 1997). note that some of the studies in the meta-​analysis
259

Colorectal Cancer 259

are limited by lack of control for behaviors asso- sedentary occupation was shown to be posi-
ciated with sedentary lifestyle (e.g., unhealthful tively associated with CRC risk (see “Sedentary
eating patterns). Lifestyle” section).

Ionizing Radiation Medical Conditions

Epidemiologic evidence to suggest ionizing radi- and Treatment
ation as a risk factor for CRC originates from Inflammatory Bowel Disease
a long-​ term cohort study of survivors of the IBD encompasses ulcerative colitis and Crohn’s
Hiroshima and Nagasaki atomic bombings in disease (granulomatous colitis), both of which
Japan in 1945 (Nakatsuka et al, 1992). Over the are relatively uncommon and distinct dis-
period of 30  years after radiation exposure, the eases. Ulcerative colitis affects the large bowel,
survivors had an elevated colon cancer risk com- whereas Crohn’s disease can occur in any part of
pared to people not exposed to atomic bomb the gastrointestinal tract, with the small intes-
radiation, with the vulnerability increasing with tine most commonly affected. Ulcerative coli-
younger ages at the time of bombing. Time since tis is more important as a risk factor for CRC,
radiation exposure did not modify the relation- but cancer risk can also be increased in Crohn’s
ship. Ionizing radiation (e.g., x-​rays and gamma disease. Persistent chronic inflammation is the
rays) refers to a type of radiation that has enough underlying mechanism driving colorectal car-
energy to remove ions (i.e., ionize) from mol- cinogenesis in IBD patients (Coussens & Werb,
ecules and thus, to produce ionized molecules 2002), and the carcinogenic process typically
that are unstable. When ionizing radiation proceeds through dysplasia rather than the
passes through cells, particularly highly mitotic classical precursor, adenomas (see “Natural
or undifferentiated cells, it can damage DNA History” section).
by directly ionizing DNA or by ionizing other Individuals with ulcerative colitis, if not
molecules such as water to produce free radi- treated with colectomy, have a 4-​to 20-​ fold
cals that can attack DNA. The ability of ionizing increased risk of CRC (Triantafillidis et al, 2009),
radiation to cause DNA damage and mutations with the lifetime risk approaching 50% (Podolsky,
is the mechanistic basis for its association with 2002). The mean age at diagnosis of CRC in IBD
CRC risk. patients is younger (40–​50  years) than that of
In developed countries, a major source of sporadic CRC (Mattar et al, 2011). For an indi-
exposure to ionizing radiation is medical proce- vidual with ulcerative colitis, the actual risk of
dures (e.g., x-​rays, CT scans, nuclear scans, and CRC is a function of multiple factors associated
radiation therapy), which may be implicated in with the condition, including the age at onset,
colorectal carcinogenesis. For instance, radia- the duration, extent of mucosal involvement, and
tion therapy for a previous cancer may increase number of dysplastic foci. Due to the relative rar-
the risk of second primary CRC. In a recent ity of IBD (especially very severe ulcerative coli-
meta-​analysis, prostate cancer patients treated tis) and availability of treatment, it is estimated
with radiotherapy had an approximately 80% that about 1% of CRC is attributable to IBD.
increased risk of second primary CRC (Wallis
et al, 2016). With growing use of medical imag- Cholecystectomy
ing tests, future studies are warranted to identify A long-​standing hypothesis has been that chole-
cumulative effect of long-​term exposure to low-​ cystectomy leads to constant release of bile acids,
dose ionizing radiation on cancer risk. which in the presence of colonic bacteria, are con-
verted to potentially toxic secondary bile acids.
Occupation Exposure to the toxic bile acids may increase
While CRC is not generally considered to have the risk of CRC over time. In a meta-​analysis of
an occupational origin, occupational exposure case-​control studies, cholecystectomy was posi-
to asbestos, a group of fibrous minerals that are tively associated with colon cancer, particularly
carcinogenic to humans, is classified as a prob- proximal colon cancer (RR = 1.34) (Giovannucci
able cause of CRC based on limited evidence. et al, 1993). A causal explanation of this potential
In a meta-​analysis, workers in the sector of association may reflect that bile acids are con-
repair and installation of machinery exposed to centrated proximally and then become diluted
asbestos had an approximately 40% increased distally. Alternatively, this association could be
risk of CRC (Oddone et al, 2014). Additionally, attributable to selection bias, given that most of
260

260 Textbook of Cancer Epidemiology

the included case-​control studies used hospital-​ and antiapoptosis (Kaaks & Lukanova, 2001;
based controls, or to detection bias, because pain Pollak, 2008). Insulin also increases bioavaila-
from gallstones and proximal colon cancers could bility of insulin-​like growth factor-​1 (IGF-​1) by
be referred to the right side of the abdomen, lead- stimulating the hepatic production of IGF-​1 and
ing to more diagnostic procedures resulting in an by decreasing hepatic secretion of insulin-​like
increased detection of incidental proximal colon growth factor binding proteins (IGFBPs) (Kaaks
cancers. In fact, in the meta-​analysis, the pos- & Lukanova, 2001; Pollak, 2008). Moreover,
itive association was not observed when analy- IGF-​1 and insulin may interact with each other
sis was confined to cohort studies. A recent large at the pathway level, because insulin receptors,
study also reported no association (Shang, 2016). IGF-​1 receptors, and hybrid insulin/​IGF-​1 recep-
Furthermore, there does not seem to be a consist- tors may respond to insulin or IGF-​1 (Kaaks &
ent association between cholecystectomy and the Lukanova, 2001; Pollak, 2008). Indeed, in two
risk of colorectal adenoma (Zhao et al, 2012). Yet, cohorts, individuals high in either IGF-​ 1 or
in a recent study, serum glycochenodeoxycholate, insulin (as indicated by C-​peptide) had an ele-
a bile acid metabolite, was positively associated vated CRC risk but risk was not further elevated
with CRC in women (Cross et  al, 2014). Taken if they were high in both (Wei et  al, 2005; Wu
together, the evidence is not strongly supportive et al, 2011). These findings suggest saturation of
of an association between cholecystectomy and a common cancer-​promoting pathway by high
colorectal neoplasia, but a small association can- concentrations of either insulin or IGF-​1, but if
not be ruled out. one of these is elevated, a higher level of the other
does not confer additional risk.
Metabolic Syndrome and
Hyperinsulinemia Type 2 Diabetes Mellitus
A unifying mechanism for CRC involves meta- Given the link between metabolic syndrome
bolic syndrome. A large and reasonably consist- and CRC (see “Metabolic Syndrome and
ent body of literature indicates the risk of colon Hyperinsulinemia” section), it is not surprising
cancer (and adenoma) is increased in those with that T2DM, the ultimate consequence of met-
the major determinants of metabolic syndrome abolic syndrome, is associated with the risk of
(e.g., obesity, abdominal obesity, physical inac- CRC. In general, T2DM patients are at approx-
tivity, and sedentary lifestyle); clinical conse- imately 30% increased risk of colorectal neopla-
quences of the syndrome (e.g., type 2 diabetes sia compared with individuals without T2DM
mellitus [T2DM] and hypertension); and circu- (Luo et  al, 2016). This association has been
lating markers of the defining components of the remarkably consistent for both cancer and ade-
syndrome (i.e., hypertriglyceridemia, low HDL noma endpoints, across case-​control and cohort
cholesterol, hyperglycemia, insulin resistance study designs, for both men and women, and
as indicated by HOMA-​IR index, or hyperin- across populations (e.g., Asia, Europe, and North
sulinemia as indicated by insulin or C-​peptide America).
levels) (Chen et  al, 2013; Giovannucci, 2007; The relationship between T2DM and CRC
Yao & Tian, 2015; Yoon et  al, 2015b). Some of risk may vary by several factors. First, time since
these individual components are discussed else- T2DM diagnosis may modify the relationship.
where in this chapter (i.e., obesity, abdominal Along the natural history of T2DM, hyperin-
obesity, physical inactivity, sedentary lifestyle, sulinemia predominates in earlier stages but, as
and T2DM). Of note, it appears highly unlikely pancreatic beta-​cells become depleted, hypoin-
that these factors are acting entirely independ- sulinemia and hyperglycemia predominate in
ently, but rather they represent a constellation later stages (DeFronzo, 1992). Thus, accounting
of interrelated risk factors. For example, certain for time since T2DM diagnosis in the examina-
behaviors such as physical inactivity are inevita- tion of T2DM and CRC risk may help identify
bly associated with abdominal adiposity, which is the pathophysiological feature of T2DM that is
also highly related to triglyceride and HDL levels. particularly critical to carcinogenesis. In one
Yet, it remains as an important question study, the risk of CRC appeared to attenuate
which factors are most causally related to CRC with time since T2DM diagnosis, which sug-
risk. Some evidence indicates that hyperinsu- gests that hyperinsulinemia at earlier stages may
linemia is the underlying risk factor for CRC. be important to the pathogenesis of CRC (Hu
Insulin is a direct mediator of cell proliferation et  al, 1999). In contrast, another study found
261

Colorectal Cancer 261

increasing risk of CRC at >15  years after diag- with acromegaly, the very high IGF-​1 levels lead
nosis of T2DM (Campbell et al, 2010). Second, to a profound insulin resistance, which may fur-
treatments for T2DM might be another modi- ther elevate the risk of colorectal neoplasia. In a
fying factor. In one study, a positive associa- small study, an increase in fasting insulin levels
tion between T2DM and CRC risk was stronger among acromegalic patients was associated with
among men using insulin than among men not a 9-​to-​15-​fold elevated risk of colonic adenomas
using insulin (Campbell et  al, 2010). In fact, (Colao et al, 2007). Yet, some recent data do not
among diabetic patients, insulin treatment support this association (Petroff et al, 2015). In
(increasing insulin exposure) appears associated summary, given the relatively rarity of the dis-
with an increased risk of CRC, while metformin ease, the relationship between acromegaly and
treatment (decreasing insulin exposure) might CRC can be considered as suggestive rather than
lower the risk (Singh et  al, 2013). Yet, conclu- definitive.
sions should be drawn with caution, given the
complexity of pharmacoepidemiologic studies Other Risk Factors
(e.g., confounding by indication especially in the Age
context of multiple drugs) as well as the evolving Excluding those affected with hereditary syn-
natural history of T2DM, which itself may have dromes, CRC develops rarely in individuals
a complex association with CRC. aged under 40  years (Figures 11-​3 and 11-​4).
The incidence rates increase exponentially with
Acromegaly age, with the vast majority of CRC diagnosed
As discussed earlier, high levels of circulat- after the age of 50 years (Siegel et al, 2014). The
ing IGF-​1 appear to increase the risk of CRC. exponential increase with age suggests that mul-
Acromegaly is a hormonal disease in which tiple genetic and epigenetic events need to accu-
a benign tumor in the pituitary gland causes mulate for a CRC to develop. The median age at
excessive production of growth hormone, lead- diagnosis of colon cancer is approximately 6 to
ing to an elevated IGF-​1 level. The rarity of this 8  years older than that of rectal cancer (Siegel
medical condition makes it difficult to study et al, 2014). Proximal colon cancers tend to be
its relationship with CRC risk. Nonetheless, a diagnosed at a higher average age, and have a
meta-​analysis found that acromegalic patients higher frequency of CIMP-​high and MSI-​high
had a statistically significant elevated risk of subtypes (Nosho et al, 2008). In high-​risk indi-
hyperplastic polyps (3.6-​fold), colon adenomas viduals with inherited syndromes, the carcin-
(2.5-​fold) and colon cancer (4.4-​fold) (Rokkas ogenesis process is accelerated and the average
et  al, 2008)  compared with individuals without age at diagnosis tends to be lower. The afore-
acromegaly. In one study, acromegalic patients mentioned patterns by age are similar between
had an increased colonic epithelial cell prolifer- men and women, but incidence rates of CRC are
ation (Cats et al, 1996). Interestingly, in patients higher in men.

400
New cases per 100,000
person-years

300

200

100

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 11-​3  Age-​specific incidence rates of colorectal cancer among women.

Source: Ferlay et al, 2013.
26

262 Textbook of Cancer Epidemiology

New cases per 100,000

600
500

person-years
400
300
200
100
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 11-​4  Age-​specific incidence rates of colorectal cancer among men.

Source: Ferlay et al, 2013.

Sex Race and Ethnicity

In most populations, men have higher age-​ In some populations, CRC incidence and mortal-
standardized incidence rates than women. ity rates differ by race and ethnicity. For example,
Worldwide, the age-​ adjusted incidence rate in the United States, the highest age-​standardized
is 1.44 times higher in men (20.6 per 100,000 incidence rate has been observed in African
person-​years) than in women (14.3 per 100,000 Americans, followed by American Indian/​
person-​ years) (IARC/​ WHO, 2015). Across Alaska Natives, Whites, Hispanics, and Asian/​
the age span, the male to female ratio for CRC Pacific Islander people (CDC, 2015). It is unclear
incidence increases with age. Despite the rel- whether these differences are primarily due to
ative increase in CRC for men at older ages, biologic differences or due to socioeconomic
the median age at diagnosis is lower in men disparities resulting in a different spectrum of
(69  years compared to 73  years in women) risk factors and screening practices (Manser &
(IARC/​WHO, 2015), because more women gen- Bauerfeind, 2014). Individuals with lower socio-
erally attain older ages in the population. Also, economic status are more likely to be obese, to
despite stronger influence of competing causes have diets that predispose to CRC, and to smoke;
of death on men, more men die of CRC (373,639/​ and they are less likely to engage in leisure-​time
year) than women (320,294/​year) (IARC/​WHO, physical activity and to undergo screening tests.
2015). By anatomic location, men have a higher All these factors contribute to the higher risk of
propensity to develop cancer in the distal colon CRC. As discussed previously, CRC incidence
and rectum. The male to female ratio is close to rates in some Asian populations have increased
1 in the cecum, but then increases progressively dramatically with changes in lifestyle. It will be
as moving distally, and reaches up to 1.8 at the important in the future to determine whether
rectum. differential vulnerability to risk factors, such as
Causes of the sex differences in CRC inci- insulin resistance, contributes to any apparent
dence and death are unclear but may be multifac- racial and ethnic differences.
torial. Sex hormones may play a role (McMichael
& Potter, 1980). Both observational studies and Height
RCTs have shown that estrogen replacement Height of a population or an individual can be
therapy in postmenopausal women reduces the viewed from several perspectives. The average
risk of CRC (Lin et al, 2012). Additionally, some height in most populations has increased over
risk factors (especially smoking and high alcohol the past century, with improved nutrition status
intake) are more common in men; obesity has a and reduced rates of childhood infectious dis-
stronger positive association with CRC in men eases. Yet, within a generally well fed population
due to their tendency to accumulate metaboli- with low rates of childhood infectious diseases,
cally harmful visceral fat (Ning et al, 2010b) (see genetics contribute substantially to individual
“Excess Adiposity” section). variation in height. It is noteworthy that CRC (as
263

Colorectal Cancer 263

well as many other cancers) rates increase as a 2003). Thus, in theory IGF-​1 levels (like height)
population become taller on average over time; can be manipulated, but the variation is relatively
population-​average height correlates with CRC inflexible in a well-​fed population. From the per-
incidence across countries; and at the individual spective of a generally well nourished population,
level within a well-​fed population, height cor- IGF-​1 levels can be considered an inherent host
relates with CRC risk. For example, in a recent factor.
meta-​analysis, relative risk per 10  cm increase A final complexity in considering IGF-​1 is
in height was 1.25 for colon cancer and 1.14 for its potential interactions with insulin, which
rectal cancer (Green et al, 2011). Height may be is more readily modifiable by diet and life-
a marker of linear growth and factors related to style changes (see “Metabolic Syndrome and
greater linear growth (e.g., circulating IGF-​ 1) Hyperinsulinemia” section).
appear to be associated with higher CRC risk.
Chemoprevention
Circulating Insulin-​Like Growth Factor Chemoprevention describes an attempt to inter-
IGF-​1, through its mitogenic and antiapoptotic vene in the carcinogenic process with agents
properties, is likely to be an important factor in ranging from nutrients (e.g., vitamins and min-
colorectal carcinogenesis. The roles of IGF-​1 are erals, mostly at nondietary doses) to natural
complex and multifaceted. It can be considered (e.g., phytochemicals) and synthetic (e.g., drugs)
a host factor, but on the other hand IGF-​1 lev- chemical compounds. Nutritional factors (some
els respond to nutrition. IGF-​1 levels peak dur- discussed earlier), such as folate, calcium, vita-
ing the growth period in adolescence, where min D, and marine omega-​3 fatty acids, can be
levels may be four-​fold higher than in adults. considered within the realm of chemopreven-
Adolescent IGF-​1 contributes to attained adult tion, though they are parts of a normal diet
height. In contrast, adult IGF-​1 levels tend to with other functions. For drugs, some evidence
have very weak correlations with height, which exists for aspirin and other nonsteroidal anti-​
suggests that adult IGF-​1 levels cannot be con- inflammatory drugs (NSAIDs) (e.g., sulindac
sidered as a surrogate of adolescent IGF-​1 lev- sulfone, celecoxib), estrogens/​progestin used in
els. Thus, it may be adolescent IGF-​1 levels that hormone replacement therapy for postmeno-
underlie the association between height and risk pausal women, ursodeoxycholic acid (a bile acid
of CRC (see “Height” section), although adult modifier), and difluoromethylornithine (Alberts,
height may contribute to the risk independent of 2005; Asano, 2004; Meyskens et  al, 2008; Lin
adolescent IGF-​1 levels. et al, 2012).
Not only adolescent IGF-​ 1 levels but also While CRC is among the most commonly
adulthood IGF-​1 levels are relevant to CRC risk. diagnosed cancers, the absolute risk is still rel-
Many studies have found that a measure of cir- atively low for an average-​ risk person. Thus,
culating IGF-​1 during adulthood prospectively chemopreventive agents, which mostly have
predicts the risk of CRC (Rinaldi et al, 2010) as side effects, are not generally recommended for
well as “high risk” adenomas (≥1 cm in diameter, average-​risk individuals, but are targeted to high-​
villous growth pattern, or high-​grade dysplasia, risk individuals (e.g., individuals with familial
≥3 in numbers, in situ cancers) (Renehan et  al, syndrome) for whom chemopreventive benefits
2001; Yoon et  al, 2015a). Further supporting a may outweigh the risk of the agents.
role for adulthood IGF-​1 levels in colorectal car- To assess the chemopreventive potential of
cinogenesis is that acromegaly (see “Acromegaly” an agent against CRC, while observational data
section) appears to increase the risk of CRC. may be informative, RCTs are the study design
Energy and protein deprivation that is sub- of choice because chemopreventive agents can
stantial enough to seriously limit height of an be provided in pill form and the agents may not
individual may lower IGF-​1 levels, but variation be widely used in the general population. In tri-
in IGF-​1 levels in a well-​fed population is not als, the most commonly used endpoint is ade-
overly sensitive to dietary pattern. While high noma recurrence, which require much shorter
protein intake, particularly from dairy sources, follow-​up periods than CRC incidence to accu-
was reported to increase IGF-​1 levels, this effect mulate an adequate number of cases for statisti-
was relatively modest and much of the variation cal power. Although short-​term (e.g., 3–​5 years)
was due to genetic or undetermined epigenetic RCTs of recurrent adenomas have some utility,
factors (Giovannucci, 2007; Giovannucci et  al, there are some important limitations or caveats
264

264 Textbook of Cancer Epidemiology

that contribute to false negative findings. First, Moreover, the optimal dose remains to be further
duration of intervention or study follow-​up per- investigated. Some studies suggest higher doses
iod may not be long enough to exert an effect are required, but other data suggest that even low
or to account for a latency period, respectively. doses (e.g., 100 mg) on alternate days are ade-
Second, the most commonly detected neoplasms quate (Cao et al, 2016; Cook et al, 2013; Garcia-​
in recurrent adenoma trials are small and tubular Albeniz & Chan 2011).
adenomas, which may not adequately represent
“high-​ risk” adenomas that are likely to prog- Screening Methods
ress to CRC. Third, because recurrent adenomas Screening may have two effects on CRC. First,
that are detected about 3–​5  years after baseline an early diagnosis and subsequent treatment
endoscopy generally reflect stages of carcinogen- of CRC decrease mortality from this can-
esis between macroscopically invisible adenomas cer. Second, an early detection and removal of
and small detectable adenomas, the effects of colorectal adenomas can reduce incidence of
chemopreventive agents acting at other carcino- CRC (Winawer et al, 1993; Zauber et al, 2012).
genic stages (e.g., adenoma formation, progres- Screening modalities based on the detection
sion from large adenomas to cancer) are missed. of blood in the stool, such as the guaiac-​based
The chemopreventive agent that has received fecal occult blood test (gFOBT) and the fecal
by far the most attention from observational immunochemical test (FIT), mainly detect
studies, RCTs, and mechanistic studies is aspi- cancer because adenomas often do not bleed.
rin. Initially observed in case-​control and cohort Indeed, in meta-​ analysis of RCTs, annual or
studies, the link between aspirin use and lower biennial screening with gFOBT was shown to
risk of adenoma and CRC is now established decrease CRC mortality but not CRC incidence
in RCTs (Chan et  al, 2005; Cook et  al, 2013; (Hewitson et al, 2008). Of note, the detection rate
Giovannucci et al, 1995; Rothwell et al, 2010). of colorectal neoplasms is higher for FIT than
Aspirin, as the most traditionally used gFOBT (Hol et  al, 2010). A  significant reduc-
NSAID, influences colorectal carcinogene- tion in CRC mortality emerged about 4  years
sis by inhibiting cyclooxygenase-​ 2 (COX-​ 2, after implementing biennial FIT-​based screen-
also known as prostaglandin-​ endoperoxide ing programs, while the effect manifested much
synthase 2), which is overexpressed in the later in gFOBT-​based trials (Zorzi et  al, 2015).
majority of CRC (Soumaoro et al, 2004). COX-​2 A multitargeted test that detects a combination
is an enzyme that mediates proinflammatory of genetic mutations, methylation markers, and
responses by converting arachidonic acid to human hemoglobin in stool had a higher sensi-
prostaglandins. The COX-​2 pathway is impli- tivity for advanced precancerous lesions as well
cated in colorectal carcinogenesis by promoting as CRC than FIT, although it exhibited more
cell proliferation, inhibiting cell apoptosis, and false positives (Imperiale et al, 2014).
stimulating angiogenesis (Rajakariar et al, 2006; Concerning screening endoscopy, flexible
Sostres et  al, 2014). Consistent with the role of sigmoidoscopy directly visualizes the lower half
aspirin in inhibiting COX-​2, regular aspirin use or third of the colorectum, while colonoscopy
was associated with a 40% lower risk of CRC inspects the entire colon and rectum. RCTs of
that overexpressed COX-​2, but not with CRC screening sigmoidoscopy showed a reduction in
that did not overexpress COX-​2 (Chan et  al, both CRC incidence (presumably by removing
2007). Similarly, aspirin use after CRC diagnosis adenomas) and CRC mortality (Elmunzer et al,
improved overall survival of patients with COX-​ 2012; Littlejohn et al, 2012; Schoen et al, 2012).
2 overexpression, but not of patients without With regard to the efficacy of screening colonos-
COX-​2 overexpression (Chan et al, 2009). copy in reducing CRC incidence and mortality,
Despite the strong evidence on aspirin and although observational studies provide sup-
CRC chemoprevention, its prophylactic use in porting evidence (Brenner et  al, 2012; Brenner
average-​ risk populations is not generally rec- et  al, 2011; Nishihara et  al, 2013), no RCTs are
ommended. Aspirin results in gastrointestinal available. Yet, it can be inferred from the RCTs
bleeding. In addition, at least 10 years of aspirin of screening sigmoidoscopy that visualizing the
use may be required for average-​risk individuals entire colorectum must provide at least simi-
to benefit from aspirin in CRC prevention (Chan lar benefits. Other screening methods such as
et  al, 2005; Giovannucci et  al, 1995). Thus, the double-​contrast barium enema (DCBE) and CT
risk-​benefit ratio of aspirin may not be favorable. colonography are less frequently used.
265

Colorectal Cancer 265

For adults at average risk for CRC, screening medical conditions (e.g., FAP, Lynch syndrome,
is generally recommended beginning at age 50 and IBD). To better address variation in the
(Qaseem et al, 2012). Screening may not be suit- optimal age from which screening colonos-
able for those over age 75 years with a life expect- copy should be initiated, future research should
ancy less than 10 years. The screening guidelines focus on ways to better stratify individual risk
for average-​ risk individuals suggest annual (e.g., risk based on genetic, serologic, and life-
gFOBT or FIT, every 5 years of flexible sigmoid- style factors). Some chemopreventive agents,
oscopy, DCBE, or CT colonography, or every particularly aspirin, are promising, but the risk-​
10  years of colonoscopy (Qaseem et  al, 2012). benefit ratio needs to be considered.
More frequent screenings starting from an ear- Finally, emerging data suggest that some of
lier age are recommended for high-​risk individu- the CRC risk factors such as excess adiposity
als: those with FAP, HNPCC, or IBD; those with and the Western dietary pattern may be acting as
a history of previous colorectal adenoma or CRC; early as in childhood and adolescence (Levi et al,
or those with a family history of colorectal ade- 2011; Nimptsch et al, 2013; Nimptsch et al, 2011;
noma or CRC in first-​degree relatives (Roncucci Nimptsch et  al, 2014; Ruder et  al, 2011). More
& Mariani, 2015). research is needed to identify CRC risk factors
at each life stage (i.e. infancy, childhood, adoles-
CONCLUSION cence, early adulthood, middle adulthood, and
CRC is a cancer largely related to the Western late adulthood), which will further our knowl-
diet and sedentary lifestyle. The average life- edge about the complex biological mechanisms
time risk of CRC in high-​ income countries underlying colorectal carcinogenesis spanning
is approximately 5%, though the risk is much several decades. Profound insights into etiology
higher in certain subgroups such as those with and pathogenesis of CRC will also stem from
familial syndromes. Yet, CRC is largely ame- ongoing research efforts to account for etio-
nable to primary prevention by modifying a logic heterogeneity of CRC based on tumor tis-
number of dietary and lifestyle risk factors, sue biomarkers including genetic and epigenetic
including tobacco smoking, excess adiposity, alterations, infiltration of immune cells, and the
physical inactivity, high alcohol intake, and microbiome (Hamada et al, 2016).
poor dietary habits (high intake of refined sug-
ars, red meats, and processed meats and low REFERENCES
intake of fiber and micronutrients). Individuals Abed J, Emgard JE, Zamir G, Faroja M, Almogy G,
with multiple risk factors are likely to have a Grenov A et  al. Fap2 mediates Fusobacterium
lifetime risk greatly exceeding 5%, whereas nucleatum colorectal adenocarcinoma enrich-
those who avoid these risk factors have a lower ment by binding to tumor-​expressed Gal-​GalNAc.
lifetime risk. On the population level, the num- Cell Host Microbe 2016;20:215–​25.
ber of CRC cases potentially preventable by Ahearn TU, Shaukat A, Flanders WD, Rutherford
avoiding several of the important lifestyle fac- RE, Bostick RM. A randomized clinical trial of
tors (i.e., smoking, excess adiposity, and physi- the effects of supplemental calcium and vitamin
cal inactivity) has been estimated to range from D3 on the APC/​B-​catenin pathway in the normal
16% to 70% (Aleksandrova et al, 2014; Erdrich mucosa of colorectal adenoma patients. Cancer
et  al, 2015; Platz et  al, 2000). Micronutrients Prev Res 2012;5:1247–​56.
including folate, calcium, and vitamin D may Alberts DS, Hess LM, Einspahr JG, Green SB,
additionally contribute to the prevention of Bhattacharyya AK, Guillen J et  al., Phoenix and
CRC, but more study is required for definitive Tucson Gastroenterologist Networks. Phase III
recommendations. A  hindrance to our under- trial of ursodeoxycholic acid to prevent colo-
rectal adenoma recurrence. J Natl Cancer Inst
standing is that a long-​time lag may be required
2005;97:846–​53.
to observe benefits from some micronutrients,
Aleksandrova K, Pischon T, Buijsse B, May AM,
and thus conventional RCTs may not be helpful.
Peeters PH et  al. Adult weight change and risk
CRC is also amenable to secondary prevention of colorectal cancer in the European Prospective
through screening tests that remove precursor Investigation into Cancer and Nutrition. Eur J
lesions or detect CRC early at a treatable stage. Cancer 2013;49:3526–​36.
Of note, current guidelines on CRC screen- Aleksandrova K, Pischon T, Jenab M, Bueno-​ de-​
ing differ by individual risk defined by family Mesquita H, Fedirko V, Norat T et al. Combined
history and personal history of predisposing impact of healthy lifestyle factors on colorectal
26

266 Textbook of Cancer Epidemiology

cancer: a large European cohort study. BMC Med Burkitt DP. Related disease—​ related cause? Lancet
2014;12:168. 1969;2:1229–​31.
Alexander DD, Miller A J, Cushing CA, Lowe KA. Burnett-​ Hartman AN, Newcomb PA, Potter JD.
Processed meat and colorectal cancer: a quantita- Infectious agents and colorectal cancer:  a review
tive review of prospective epidemiologic studies. of Helicobacter pylori, Streptococcus bovis, JC virus,
Eur J Cancer Prev 2010;19:328–​41. and human papillomavirus. Cancer Epidemiol
Ananthakrishnan AN, Du M, Berndt SI, Brenner H, Biomarkers Prev 2008;17:2970–​79.
Caan BJ, Casey G et  al. Red meat intake, NAT2, Buset M, Kipkin M, Winawer S, Swaroop S, Friedman
and risk of colorectal cancer: a pooled analysis of E. Inhibition of human colonic epithelial cell pro-
11 studies. Cancer Epidemiol Biomarkers Prev liferation in vivo and in vitro by calcium. Cancer
2015;24:198–​205. Res 1986;46:5426–​30.
Andersen V, Holst R, Vogel U. Systematic review: diet-​ Butterworth AS, Higgins JP, Pharoah P. Relative and
gene interactions and the risk of colorectal can- absolute risk of colorectal cancer for individu-
cer. Aliment Pharmacol Ther 2013;37:383–​91. als with a family history:  a meta-​analysis. Eur J
Andersen V, Vogel U. Interactions between meat Cancer 2006;42:216–​27.
intake and genetic variation in relation to colorec- Byrne C, Sinha R, Platz EA, Giovannucci, E, Colditz
tal cancer. Genes Nutr 2015;10:448. GA, Hunter DJ et  al. Predictors of dietary het-
Asano TK, Mcleod RS. Nonsteroidal anti-​ erocyclic amine intake in three prospective
inflammatory drugs and aspirin for the prevention cohorts. Cancer Epidemiol Biomarkers Prev
of colorectal adenomas and cancer:  a systematic 1998;7:523–​29.
review. Dis Colon Rectum 2004;47:665–​73. Cai S, Li Y, Ding Y, Chen K, Jin M. Alcohol drinking
Aune D, Chan DS, Lau R, Vieira R, Greenwood DC, and the risk of colorectal cancer death:  a meta-​
Kampman E, Norat T. Dietary fibre, whole grains, analysis. Eur J Cancer Prev 2014;23:532–​39.
and risk of colorectal cancer:  systematic review Campbell PT, Deka A, Jacobs EJ, Newton CC,
and dose-​response meta-​analysis of prospective Hildebrand JS, McCullough ML et al. Prospective
studies. BMJ 2011;343:d6617. study reveals associations between colorectal can-
Baron JA. Dietary fiber and colorectal cancer:  an cer and type 2 diabetes mellitus or insulin use in
ongoing saga. JAMA 2005;294:2904–​06. men. Gastroenterology 2010;139:1138–​46.
Baron JA, Barry EL, Mott LA, Rees JR, Sandler RS, Cao Y, Nishihara R, Wu K, Wang M, Ogino S, Willett
Snover DC et al. A trial of calcium and vitamin D WC et  al. Population-​wide impact of long-​term
for the prevention of colorectal adenomas. N Engl use of aspirin and the risk for cancer. JAMA Oncol
J Med 2015;373;1519–​30. 2016;2:762–​9.
Bisgaard ML, Fenger K, Bulow S, Niebuhr E, Mohr Carethers JM, Jung BH. Genetics and genetic biomark-
J. Familial adenomatous polyposis (FAP):  fre- ers in sporadic colorectal cancer. Gastroenterology
quency, penetrance, and mutation rate. Hum 2015;149:1177–​90.
Mutat 1994;3:121–​25. Carr PR, Walter V, Brenner H, Hoffmeister M. Meat
Bouvard V, Loomis D, Guyton KZ, Grosse Y, Ghissassi subtypes and their association with colorectal
FE, Benbrahim-​Tallaa L et al. Carcinogenicity of cancer: systematic review and meta-​analysis. Int J
consumption of red and processed meat. Lancet Cancer 2016;138:293–​302.
Oncol 2015;16:1599–​600. Cats A, Dullaart R, Kleibeuker J, Kuipers F, Sluiter W,
Boyle T, Keegel T, Bull F, Heyworth J, Fritschi L. Hardonk M et al. Increased epithelial cell prolif-
Physical activity and risks of proximal and dis- eration in the colon of patients with acromegaly.
tal colon cancers:  a systematic review and meta-​ Cancer Res 1996;56: 523–​26.
analysis. J Natl Cancer Inst 2012;104:1548–​61. CDC. Colorectal Cancer Rates by Race and Ethnicity.
Brenner H, Chang-​Claude J, Rickert A, Seiler CM, 2015. http://​www.cdc.gov/​cancer/​colorectal/​sta-
Hoffmeister M. Risk of colorectal cancer after tistics/​race.htm.
detection and removal of adenomas at colonos- Center MM, Jemal A, Ward E. International trends
copy: population-​based case-​control study. J Clin in colorectal cancer incidence rates. Cancer
Oncol 2012;30:2969–​76. Epidemiol Biomarkers Prev 2009;18:1688–​94.
Brenner H, Chang-​ Claude J, Seiler CM, Chan AT, Giovannucci EL, Meyerhardt JA,
Hoffmeister M. Long-​ term risk of colorectal Schernhammer ES, Curhan GC et  al. Long-​term
cancer after negative colonoscopy. J Clin Oncol use of aspirin and nonsteroidal anti-​inflammatory
2011;29:3761–​67. drugs and risk of colorectal cancer. JAMA
Bristow SM, Bolland MJ, MacLennan GS, Avenell A, 2005;294:914–​23.
Grey A et  al. Calcium supplements and cancer Chan AT, Ogino S, Fuchs CS. Aspirin and the risk of
risk: a meta-​analysis of randomised controlled tri- colorectal cancer in relation to the expression of
als. Br J Nutr 2013;110:1384–​93. COX-​2. N Engl J Med 2007;356:2131–​42.
267

Colorectal Cancer 267

Chan AT, Ogino S, Fuchs CS. Aspirin use and sur- Cross AJ, Sinha R. Meat-​related mutagens/​carcino-
vival after diagnosis of colorectal cancer. JAMA gens in the etiology of colorectal cancer. Environ
2009;302:649–​59. Mol Mutagen 2004;44:44–​55.
Chan DS, Lau R, Aune D, Vieira R, Greenwood DC, Cunningham D, Atkin W, Lenz HJ, Lynch HT, Minsky
Kampman E et  al. Red and processed meat and B, Nordlinger B et  al. Colorectal cancer. Lancet
colorectal cancer incidence: meta-​analysis of pro- 2010;375:1030–​47.
spective studies. PLoS One 2011;6:e20456. Curtin K, Samowitz WS, Wolff RK, Herrick J, Caan
Charlton BM, Wu K, Zhang X, Giovannucci EL, BJ, Slattery ML. Somatic alterations, metabolizing
Fuchs CS, Missmer SA et  al. Oral contraceptive genes and smoking in rectal cancer. Int J Cancer
use and colorectal cancer in the Nurses’ Health 2009;125;158–​64.
Study I and II. Cancer Epidemiol Biomarkers Prev Curtin K, Slattery ML, Samowitz WS. CpG island
2015;24:1214–​21. methylation in colorectal cancer: past, present and
Chen J, Giovannucci E, Kelsey K, Rimm E, Stampfer future. Patholog Res Int 2011;902674.
MJ, Colditz GA et al. A methylenetetrahydrofolate Debinski HS, Love S, Spigelman AD, Phillips RK.
reductase polymorphism and risk of colorectal Colorectal polyp counts and cancer risk in famil-
cancer. Cancer Res 1996;56:4862–​4. ial adenomatous polyposis. Gastroenterology
Chen L, Li L, Wang Y, Li P, Luo L, Yang B et  al. 1996;110:1028–​30.
Circulating C-​peptide level is a predictive factor DeFronzo RA. Pathogenesis of type 2 (non-​insulin
for colorectal neoplasia: evidence from the meta-​ dependent) diabetes mellitus:  a balanced over-
analysis of prospective studies. Cancer Causes view. Diabetologia 1992;35;389–​97.
Control 2013;24:1837–​47. Dellavalle CT, Xiao Q, Yang G, Shu XO, Aschebrook-​
Chiavarini M, Minelli L, Fabiani R. Garlic consump- Kilfoy B, Zheng W et  al. Dietary nitrate and
tion and colorectal cancer risk in man: a system- nitrite intake and risk of colorectal cancer in the
atic review and meta-​analysis. Public Health Nutr Shanghai Women’s Health Study. Int J Cancer
2016;19:308–​17. 2014;134: 2917–​26.
Cho E, Lee J, Rimm E, Fuchs C, Giovannucci E. DeMarini DM. Genotoxicity of tobacco smoke and
Alcohol consumption and the risk of colon cancer tobacco smoke condensate:  a review. Mutat Res
by family history of colorectal cancer. Am J Clin 2004;567:447–​74.
Nutr 2012;95:413–​19. Despres JP. Abdominal obesity:  the most prevalent
Cho E, Smith-​Warner SA, Ritz J, van den Brandt PA, cause of the metabolic syndrome and related
Colditz GA, Folsom AR et al. Alcohol intake and cardiometabolic risk. European Heart Journal
colorectal cancer:  a pooled analysis of 8 cohort Supplements 2006;8(B):B4–​B12.
studies. Ann Intern Med 2004;140:603–​13. Doll R, Gray R, Hafner B, Peto R. Mortality in rela-
Colao A, Pivonello R, Auriemma R, Galdiero M, Ferone tion to smoking: 22 years’ observations on female
D, Minuto F et al. The association of fasting insu- British doctors. Br J Med 1980;280:967–​71.
lin concentrations and colonic neoplasms in acro- Du M, Zhang X, Hoffmeister M, Schoen R, Baron J,
megaly: a colonoscopy-​based study in 210 patients. Berndt S et al. Colon Cancer Family Registry and
J Clin Endocrinol Metab 2007;92:3854–​60. Genetics and Epidemiology of Colorectal Cancer
Cook NR, Lee IM, Zhang SMM, Moorthy MV, Consortium. No evidence of gene-​calcium inter-
Buring JE. Alternate-​day, low-​dose aspirin and actions from genome-​wide analysis of colorectal
cancer risk:  long-​ term observational follow-​ cancer risk. Cancer Epidemiol Biomarkers Prev
up of a randomized trial. Ann Intern Med 2014;23:2971–​76.
2013;159:77–​W32. Dunlop MG, Tenesa A, Farrington SM, Ballereau S,
Cotton S, Sharp L, Little J. The adenoma-​carcinoma Brewster DH, Koessler T et al. Cumulative impact
sequence and prospects for the prevention of colo- of common genetic variants and other risk factors
rectal neoplasia. Crit Rev Oncogen 1996;7:293–​342. on colorectal cancer risk in 42,103 individuals.
Coussens LM, Werb Z. Inflammation and cancer. Gut 2013;62:871–​81.
Nature 2002;420:860–​67. Elmunzer BJ, Hayward RA, Schoenfeld PS, Saini
Cross A, Moore S, Boca S, Huang W, Xiong X, SD, Deshpande A, Waljee AK. Effect of flexible
Stolzenberg-​Solomon R et al. A prospective study sigmoidoscopy-​based screening on incidence and
of serum metabolites and colorectal cancer risk. mortality of colorectal cancer: a systematic review
Cancer 2014;120:3049–​57. and meta-​analysis of randomized controlled tri-
Cross AJ, Ferrucci LM, Risch A, Graubard BI, Ward als. PLoS Med 2012;9:e1001352.
MH, Park Y et  al. A large prospective study of Erdrich J, Zhang X, Giovannucci E, Willett W.
meat consumption and colorectal cancer risk: an Proportion of colon cancer attributable to lifestyle
investigation of potential mechanisms underlying in a cohort of US women. Cancer Causes Control
this association. Cancer Res 2010;70:2406–​14. 2015;26:1271–​79.
268

268 Textbook of Cancer Epidemiology

Fearnhead NS, Britton MP, Bodmer WF. The ABC of Giovannucci E. Alcohol, one-​carbon metabolism, and
APC. Hum Mol Genet 2001;10:721–​33. colorectal cancer: recent insights from molecular
Fearon ER, Vogelstein B. A genetic model for colorec- studies. J Nutr 2004;134:2475S–​81S.
tal tumorigenesis. Cell 1990;61:759–​67. Giovannucci E. Metabolic syndrome, hyperinsu-
Fedirko V, Tramacere I, Bagnardi V, Rota M, Scotti L, linemia, and colon cancer:  a review. American
Islami F et al. Alcohol drinking and colorectal can- Journal of Clinical Nutrition 2007;86:s836–​42.
cer risk: an overall and dose-​response meta-​analysis Giovannucci E. Vitamin D and cancer incidence in the
of published studies. Ann Oncol 2011;22:1958–​72. Harvard cohorts. Ann Epidemiol 2009;19:84–​8.
Feldman D, Krishnan AV, Swami S, Giovannucci Giovannucci E. Epidemiology of vitamin D and colo-
E, Feldman BJ. The role of vitamin D in reduc- rectal cancer. Anticancer Agents Med Chem
ing cancer risk and progression. Nat Rev Cancer 2013;13:11–​9.
2014;14:342–​57. Giovannucci E, Ascherio A, Rimm EB, Colditz GA,
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Stampfer MJ, Willett WC. Physical activity, obe-
Mathers C, Rebelo M, Parkin DM, Forman D, Bray, sity, and risk for colon cancer and adenoma in
F. GLOBOCAN 2012 v1.0, Cancer Incidence and men. Ann Intern Med 1995;122:327–​34.
Mortality Worldwide: IARC CancerBase No. 11 Giovannucci E, Colditz GA, Stampfer MJ. A meta-​
[Internet]. Lyon, France: International Agency for analysis of cholecystectomy and risk of colorectal
Research on Cancer; 2013. Available from: http:// cancer. Gastroenterology 1993;105:130–​41.
globocan.iarc.fr, accessed 2017. Giovannucci E, Colditz GA, Stampfer MJ, Hunter D,
Figueiredo J, Crockett S, Snover D, Morris C, Rosner BA, Willett WC et al. A prospective study
McKeown-​Eyssen G, Sandler R et  al. Smoking-​ of cigarette smoking and risk of colorectal ade-
association risks of conventional adenomas and noma and colorectal cancer in U.S. women. J Natl
serrated polyps in the colorectum. Cancer Causes Cancer Inst 1994a;86:192–​99.
Control 2015;26:377–​89. Giovannucci E, Egan KM, Hunter DJ, Stampfer MJ,
Figueiredo J, Hsu L, Hutter C, Lin Y, Campbell P, Colditz GA, Willett WC et  al. Aspirin and the
Baron J et  al. GECCO. Genome-​wide diet-​gene risk of colorectal cancer in women. N Engl J Med
interaction analyses for risk of colorectal cancer. 1995;333:609–​14.
PLoS Genet 2014;10:e1004228. Giovannucci E, Martinez ME. Tobacco, colorectal
Flood DM, Weiss NS, Cook LS, Emerson JC, Schwartz cancer, and adenomas: a review of the evidence. J
SM, Potter JD. Colorectal cancer incidence in Asian Natl Cancer Inst 1996;88:1717–​30.
migrants to the United States and their descen- Giovannucci E, Pollak M, Liu Y, Platz EA, Majeed N,
dants. Cancer Causes Control 2000;11:403–​11. Rimm EB, Willett WC. Nutritional predictors of
Fung T, Hu F, Schulze M, Pollak M, Wu T, Fuchs C, insulin-​like growth factor I and their relationships
Giovannucci E. A dietary pattern that is associated to cancer in men. Cancer Epidemiol Biomarkers
with C-​peptide and risk of colorectal cancer in Prev 2003;12:84–​9.
women. Cancer Causes Control 2012;23:959–​65. Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA,
Fung TT, Brown LS. Dietary Patterns and the Risk of Ascherio A, Kearney J, Willett WC. A prospective
Colorectal Cancer. Curr Nutr Rep 2013;2:48–​55. study of cigarette smoking and risk of colorectal
Gao QY, Chen HM, Chen YX, Wang YC, Wang ZH, adenoma and colorectal cancer in U.S. men. J Natl
Tang JT et al. Folic acid prevents the initial occur- Cancer Inst 1994b:86:183–​91.
rence of sporadic colorectal adenoma in Chinese Green J, Cairns B, Casabonne D, Wright F, Reeves G,
older than 50 years of age: a randomized clinical Beral V, Collaborators MWS. Height and cancer
trial. Cancer Prev Res 2013;6:744–​52. incidence in the Million Women Study:  prospec-
Garcia-​ Albeniz X, Chan AT. Aspirin for the pre- tive cohort, and meta-​analysis of prospective stud-
vention of colorectal cancer. Best Pract Res Clin ies of height and total cancer risk. Lancet Oncol
Gastroenterol 2011;25:461–​72. 2011;12:785–​94.
Garland CF, Garland FC. Do sunlight and vitamin Guan HB, Wu QJ, Gong TT, Lin B, Wang YL, Liu
D reduce the likelihood of colon cancer? Int J CX. Parity and risk of colorectal cancer:  a dose-​
Epidemiol 1980;9:227–​31. response meta-​ analysis of prospective studies.
Geer EB, Shen W. Gender differences in insulin resist- PLOS One 2013;8:e75279.
ance, body composition, and energy balance. Haenszel W, Kurihara M. Studies of Japanese migrants.
Gend Med 2009;6 Suppl 1:60–​75. I. Mortality from cancer and other diseases among
Giovannucci E. An updated review of the epidemio- Japanese in the United States. J Natl Cancer Inst
logical evidence that cigarette smoking increases 1968;4:43–​68.
risk of colorectal cancer. Cancer Epidemiol Hamada T, Keum N, Nishihara R, Ogino S. Molecular
Biomarkers Prev 2001;10:725–​31. pathological epidemiology:  new developing
269

Colorectal Cancer 269

frontiers of big data science to study etiologies and Howard RA, Freedman DM, Park Y, Hollenbeck A,
pathogenesis. J Gastroenterol 2016;52:265–​75. Schatzkin A, Leitzmann MF. Physical activity, sed-
Hamilton SR, Liu B, Parsons RE, Papadopoulos N, Jen entary behavior, and the risk of colon and rectal
J, Powell SM et al. The molecular basis of Turcot’s cancer in the NIH-​AARP Diet and Health Study.
syndrome. N Engl J Med 1995;332:839–​47. Cancer Causes Control 2008;19:939–​53.
Hannan LM, Jacobs EJ, Thun MJ. The association Hsu L, Jeon J, Brenner H, Gruber SB, Schoen RE,
between cigarette smoking and risk of colorec- Berndt SI et al. A model to determine colorectal
tal cancer in a large prospective cohort from the cancer risk using common genetic susceptibility
United States. Cancer Epidemiol Biomarkers Prev loci. Gastroenterology 2015;148:1330–​39.
2009;18:3362–​67. Hu FB. Dietary pattern analysis:  a new direction in
Harriss DJ, Cable NT, George K, Reilly T, Renehan nutritional epidemiology. Curr Opin Lipidol
AG, Haboubi N. Physical activity before and after 2002;13:3–​9.
diagnosis of colorectal cancer: disease risk, clini- Hu FB, Manson JE, Liu S, Hunter D, Colditz GA,
cal outcomes, response pathways and biomarkers. Michels KB et al. Prospective study of adult onset
Sports Med 2007;37:947–​60. diabetes mellitus (type 2) and risk of colorectal can-
He TC, Sparks AB, Rago C, Hermeking H, Zawel L, da cer in women. J Natl Cancer Inst 1999;91:542–​47.
Costa, LT et al. Identification of c-​MYC as a target Huxley RR, Ansary-​ Moghaddam A, Clifton P,
of the APC pathway. Science 1998;281:1509–​12. Czernichow S, Parr CL, Woodward M. The
Heitman SJ, Ronksley PE, Hilsden RJ, Manns BJ, impact of dietary and lifestyle risk factors on
Rostom A, Hemmelgarn BR. Prevalence of adeno- risk of colorectal cancer: a quantitative overview
mas and colorectal cancer in average risk individ- of the epidemiological evidence. Int J Cancer
uals: a systematic review and meta-​analysis. Clin 2009;125:171–​180.
Gastroenterol Hepatol 2009;7:1272–​78. IARC/​ WHO. GLOBOCAN 2012:  estimated cancer
Herman JG, Umar A, Polyak K, Graff JR, Ahuja N, incidence, mortality and prevalence worldwide in
Issa JP et  al. Incidence and functional conse- 2012. http://​globocan.iarc.fr/​Pages/​fact_​sheets_​
quences of hMLH1 promoter hypermethylation cancer.aspx (accessed 2015).
in colorectal carcinoma. Proc Natl Acad Sci U S A Ibrahim MM. Subcutaneous and visceral adipose tis-
1998;95:6870–​75. sue:  structural and functional differences. Obes
Hewitson P, Glasziou P, Watson E, Towler B, Irwig Rev 2010;11:11–​8.
L. Cochrane systematic review of colorectal can- Imperiale TF, Ransohoff DF, Itzkowitz SH. Multitarget
cer screening using the fecal occult blood test stool DNA testing for colorectal-​cancer screening.
(hemoccult):  an update. Am J Gastroenterol N Engl J Med 2014;371:187–​88.
2008;103:1541–​49. Irving BA, Davis CK, Brock DW, Weltman JY, Swift D,
Hill MJ, Morson BC, Bussey HJ. Aetiology of Barrett EJ et al. Effect of exercise training intensity
adenoma—​ carcinoma sequence in large bowel. on abdominal visceral fat and body composition.
Lancet 1978;1:245–​47. Med Sci Sports Exerc 2008;40:1863–​72.
Hjartaker A, Aagnes B, Robsahm TE, Langseth H, Jayasekara H, MacInnis R, Room R, English D. Long-​
Bray F, Larsen IK. Subsite-​specific dietary risk fac- term alcohol consumption and breast, upper
tors for colorectal cancer: a review of cohort stud- aero-​digestive tract and colorectal cancer risk:  a
ies. J Oncol 2013;703854. systematic review and meta-​ analysis. Alcohol
Hol L, van Leerdam ME, van Ballegooijen M, van Alcohol 2016;51:315–​30.
Vuuren AJ, van Dekken H, Reijerink JC et  al. Johns LE, Houlston RS. A systematic review and
Screening for colorectal cancer: randomised trial meta-​analysis of familial colorectal cancer risk.
comparing guaiac-​ based and immunochemical Am J Gastroenterol 2001;96:2992–​3003.
faecal occult blood testing and flexible sigmoidos- Johnson CM, Wei C, Ensor JE, Smolenski DJ, Amos
copy. Gut 2010;59:62–​8. CI, Levin B, Berry DA. Meta-​analyses of colo-
Holme O, Bretthauer M, Eide TJ, Loberg EM, Grzyb rectal cancer risk factors. Cancer Causes Control
K, Loberg M et  al. Long-​term risk of colorectal 2013;24:1207–​22.
cancer in individuals with serrated polyps. Gut Jokelainen K, Nosova T, Koivisto T, Vakevainen
2015;64:929–​36. S, Jousimies-​ Somer H, Heine R, Salaspuro M.
Hord NG, Tang Y, Bryan NS. Food sources of nitrates Inhibition of bacteriocolonic pathway for eth-
and nitrites: the physiologic context for potential anol oxidation by ciprofloxacin in rats. Life Sci
health benefits. Am J Clin Nutr 2009;90:1–​10. 1997;61:1755–​62.
Houlston RS, Tomlinson IP. Polymorphisms Kaaks R, Lukanova A. Energy balance and cancer: the
and colorectal tumor risk. Gastroenterology role of insulin and insulin-​like growth factor-​I.
2001;121:282–​301. Proc Nutr Soc 2001;60:91–​106.
270

270 Textbook of Cancer Epidemiology

Kaaks R, Lukanova A, Kurzer MS. Obesity, endoge- analyses of 13 prospective cohort studies on folate
nous hormones, and endometrial cancer risk:  a intake and colon cancer. Cancer Causes Control
synthetic review. Cancer Epidemiol Biomarkers 2010;21:1919–​30.
Prev 2002;11:1531–​43. Kim YI. Folate:  a magic bullet or a double edged
Kang GH. Four molecular subtypes of colorectal can- sword for colorectal cancer prevention? Gut
cer and their precursor lesions. Arch Pathol Lab 2006;55:1387–​89.
Med 2011;135:698–​703. Knekt P, Jarvinen R, Dich J, Hakulinen T. Risk of
Kang HW, Kim D, Kim HJ, Kim CH, Kim YS, Park colorectal and other gastro-​ intestinal cancers
MJ et  al. Visceral obesity and insulin resistance after exposure to nitrate, nitrite and N-​nitroso
as risk factors for colorectal adenoma:  a cross-​ compounds:  a follow-​ up study. Int J Cancer
sectional, case-​control study. Am J Gastroenterol 1999;80:852–​56.
2010;105:178–​87. Kostic AD, Chun E, Robertson L, Glickman JN,
Kantor E, Giovannucci E. Gene-​diet interactions and Gallini CA, Michaud M et  al. Fusobacterium
their impact on colorectal cancer risk. Curr Nutr nucleatum potentiates intestinal tumorigenesis
Rep 2015;4:13–​21. and modulates the tumor-​immune microenviron-
Kantor ED, Hutter CM, Minnier J, Berndt SI, Brenner ment. Cell Host Microbe 2013;14:207–​15.
H, Caan BJ et  al. Gene-​environment interaction Kritchevsky D. Epidemiology of fibre, resistant
involving recently identified colorectal cancer starch and colorectal cancer. Eur J Cancer Prev
susceptibility Loci. Cancer Epidemiol Biomarkers 1995;4:345–​52.
Prev 2014;23:1824–​33. Kune GA. The Melbourne Colorectal Cancer
Kennedy D, Stern S, Matok I, Moretti M, Sarkar Study: reflections on a 30-​year experience. Med J
M, Adams-​ Webber T, Koren G. Folate intake, Aust 2010;193:648–​52.
MTHFR polymorphisms, and the risk of colorec- Lamprecht SA, Lipkin M. Cellular mechanisms of cal-
tal cancer: a systematic review and meta-​analysis. cium and vitamin D in the inhibition of colorectal
J Cancer Epidemiol 2012;952508. carcinogenesis. Ann N Y Acad Sci 2001;952:73–​87.
Keum N, Aune D, Greenwood DC, Ju W, Giovannucci Larsson SC, Wolk A. Meat consumption and risk of
EL. Calcium intake and colorectal cancer risk: colorectal cancer:  a meta-​analysis of prospective
dose-​response meta-​analysis of prospective obser- studies. Int J Cancer 2006;119:2657–​64.
vational studies. Int J Cancer 2014;135:1940–​48. Le DT, Uram JN, Wang H, Bartlett BR, Kemberling
Keum N, Cao Y, Oh H, Smith-​Warner SA, Orav J, Wu H, Eyring AD et  al. PD-​1 blockade in tumors
K et  al. Sedentary behaviors and light-​intensity with mismatch-​repair deficiency. N Engl J Med
activities in relation to colorectal cancer risk. Int 2015;372:2509–​20.
J Cancer 2016;138:2109–​17. Lee JE, Willett WC, Fuchs CS, Smith-​Warner SA, Wu
Keum N, Giovannucci EL. Folic acid fortifica- K, Ma J, Giovannucci E. Folate intake and risk of
tion and colorectal cancer risk. Am J Prev Med colorectal cancer and adenoma:  modification by
2014;46:S65–​72. time. American J Clin Nutr 2011;93:817–​25.
Keum N, Greenwood DC, Lee DH, Kim R, Aune D, Ju Lee RG. Benign tumors of the colon. In: Haubrich WS,
W et  al. Adult weight gain and adiposity-​related Schaffner F, Berk J (eds.) Bockus gastroenterology.
cancers:  a dose-​response meta-​analysis of pro- 5 ed. Philadelphia: Saunders 1995: 1715–​1730.
spective observational studies. J Natl Cancer Inst Lev R. Adenomatous Polyps of the Colon.
2015;107. New York: Springer-​Verlag, 1990: 61.
Keum N, Kim H, Giovannucci EL. Calcium as a Levi Z, Kark JD, Barchana M, Liphshitz I, Zavdi O,
chemopreventive agent against colorectal neo- Tzur D et al. Measured body mass index in ado-
plasm:  does obesity play a role? Cancer Causes lescence and the incidence of colorectal cancer in
Control 2017;28:853–​56. a cohort of 1.1  million males. Cancer Epidemiol
Keum N, Lee D, Greenwood D, Zhang X, Giovannucci Biomarkers Prev 2011;20:2524–​31.
E. Calcium intake and colorectal adenoma risk: Levin B, Raijman I. Malignant tumors of the colon
dose-​response meta-​analysis of prospective obser- and rectum. In:  Haubrich WS, Schaffner F,
vational studies. Int J Cancer 2015;136:1680–​87. Berk J (eds.) Bockus Gastroenterology. 5e ed.
Keum N, Lee DH, Kim R, Greenwood DC, Philadelphia: Saunders, 1995: 1744–​1772.
Giovannucci EL. Visceral adiposity and Liang PS, Chen TY, Giovannucci E. Cigarette smok-
colorectal adenomas:  dose-​ response meta-​ ing and colorectal cancer incidence and mortal-
analysis of observational studies. Ann Oncol ity:  systematic review and meta-​ analysis. Int J
2015;26:1101–​09. Cancer 2009;124:2406–​15.
Kim DH, Smith-​ Warner SA, Spiegelman D, Yaun Lilienfeld A, Levin M, Kessler I. Mortality among
SS, Colditz GA, Freudenheim JL et  al. Pooled the foreign born and their countries of
271

Colorectal Cancer 271

origin. In: Cancer in the United States. Cambridge, colorectal cancer patients:  a systematic review.
MA: Harvard University Press, 1972a: 233–​278. Gastrointest Endosc 2014;80:42–​60.
Lilienfeld A, Levin M, Kessler I. Summary of mortal- Markowitz S, Bertagnolli M. Molecular origins of can-
ity data for selected sites. In: Cancer in the United cer: molecular basis of colorectal cancer. N Engl J
States. Cambridge, MA: Harvard University Press, Med 2009;361:2449–​60.
1972b: 279–​297. Martinez ME, Giovannucci E, Spiegelman D, Hunter
Limsui D, Vierkant RA, Tillmans LS, Wang AH, DJ, Willett WC, Colditz GA. Leisure-​time physi-
Weisenberger DJ, Laird PW et  al. Cigarette cal activity, body size, and colon cancer in women.
smoking and colorectal cancer risk by molec- Nurses’ Health Study Research Group. J Natl
ularly defined subtypes. J Natl Cancer Inst Cancer Inst 1997;89:948–​55.
2010;102:1012–​22. Maskarinec G, Noh JJ. The effect of migration on can-
Lin JH, Zhang SMM, Rexrode KM, Manson JE, Chan cer incidence among Japanese in Hawaii. Ethn Dis
AT, Wu KN et al. Association between sex hormones 2004;14:431–​39.
and colorectal cancer risk in men and women. Mason JB, Dickstein A, Jacques PF, Haggarty P,
Clinical Gastroenterol Hepatol 2013;11:419–​24. Selhub J, Dallal G, Rosenberg IH. A temporal
Lin K, Cheung W, Lai J, Giovannucci E. The effect of association between folic acid fortification and
estrogen vs. combined estrogen-​progestogen ther- an increase in colorectal cancer rates may be
apy on the risk of colorectal cancer. Int J Cancer illuminating important biological principles:  a
2012;130:419–​30. hypothesis. Cancer Epidemiol Biomarkers Prev
Littlejohn C, Hilton S, Macfarlane GJ, Phull P. 2007;16:1325–​29.
Systematic review and meta-​analysis of the evi- Mattar MC, Lough D, Pishvaian MJ, Charabaty A.
dence for flexible sigmoidoscopy as a screening Current management of inflammatory bowel dis-
method for the prevention of colorectal cancer. Br ease and colorectal cancer. Gastrointest Cancer
J Surg 2012;99:1488–​1500. Res 2011;4:53–​61.
Lochhead P, Chan AT, Giovannucci E, Fuchs CS, Wu McCredie M, Williams S, Coates M. Cancer mortality
K, Nishihara R et  al. Progress and opportunities in migrants from the British Isles and continen-
in molecular pathological epidemiology of colo- tal Europe to New South Wales, Australia, 1975–​
rectal premalignant lesions. Am J Gastroenterol 1995. Int J Cancer 1999;83:179–​85.
2014;109:1205–​14. McCullough M, Jacobs E, Shah R, Campbell P, Gapstur
Loh YH, Jakszyn P, Luben RN, Mulligan AA, Mitrou S. Garlic consumption and colorectal cancer risk
PN, Khaw KT. N-​Nitroso compounds and cancer in the CPS-​II nutrition cohort. Cancer Causes
incidence: the European Prospective Investigation Control 2012;23:1643–​51.
into Cancer and Nutrition (EPIC)-​Norfolk Study. McGlynn AP, Wasson GR, O’Reilly SL, McNulty H,
Am J Clin Nutr 2011;93:1053–​61. Downes CS, Chang CK et  al. Low colonocyte
Lu CY, Song B, Wang YY, Meng H, Guo SB, Liu NL folate is associated with uracil misincorporation
et al. Age at menarche and risk of colorectal can- and global DNA hypomethylation in human col-
cer: a meta-​analysis. PLOS One 2013;8:e65645. orectum. J Nutr 2013;143:27–​33.
Luo S, Li JY, Zhao LN, Yu T, Zhong W, Xia ZS et al. McMichael AJ, Potter JD. Reproduction, endogenous
Diabetes mellitus increases the risk of colorectal and exogenous sex hormones, and colon can-
neoplasia:  an updated meta-​ analysis. Clin Res cer:  a review and hypothesis. J Natl Cancer Inst
Hepatol Gastroenterol 2016;40:110–​23. 1980;65:1201–​07.
Lynch HT, Smyrk T. Hereditary nonpolyposis colorec- Meng S, Zhang X, Giovannucci E, Ma J, Fuchs C,
tal cancer (Lynch syndrome): an updated review. Cho E. No association between garlic intake
Cancer 1996;78:1149–​67. and risk of colorectal cancer. Cancer Epidemiol
Lynch PM. Polyposis syndromes. In:  Haubrich WS, 2013;37:152–​55.
Schaffner F, Berk J (eds.) Bockus Gastroenterology. Meyskens FL Jr, McLaren C, Pelot D, Fujikawa-​
5 ed. Philadelphia: Saunders, 1995: 1731–​1743. Brooks S, Carpenter P, Hawk E et  al. Difluoro­
MacInnis RJ, English DR, Hopper JL, Haydon AM, methylornithine plus sulindac for the prevention
Gertig DM, Giles GG. Body size and composition of sporadic colorectal adenomas:  a randomized
and colon cancer risk in men. Cancer Epidemiol placebo-​controlled, double-​blind trial. Cancer Prev
Biomarkers Prev 2004;13:553–​9. Res 2008;1:32–​8.
Magalhaes B, Peleteiro B, Lunet N. Dietary patterns Michels KB, Fuchs CS, Giovannucci E, Colditz GA,
and colorectal cancer:  systematic review and Hunter DJ, Stampfer MJ, Willett WC. Fiber intake
meta-​analysis. Eur J Cancer Prev 2012;21:15–​23. and incidence of colorectal cancer among 76,947
Manser C, Bauerfeind P. Impact of socioeconomic women and 47,279 men. Cancer Epidemiol
status on incidence, mortality, and survival of Biomarkers Prev 2005;14:842–​49.
27

272 Textbook of Cancer Epidemiology

Mima K, Nishihara R, Qian ZR, Cao Y, Sukawa Y, NIH. Tests to Detect Colorectal Cancer and Polyps.
Nowak JA et al. Fusobacterium nucleatum in colo- http://​ w w w.cancer.gov/​ t yp es/​ c olorec t al/​
rectal carcinoma tissue and patient prognosis. Gut screening-​fact-​sheet (accessed 2014).
2015. [Epub ahead of print]. Nimptsch K, Bernstein A, Giovannucci E, Fuchs C,
Mitchell RJ, Farrington SM, Dunlop MG, Campbell H. Willett W, Wu K. Dietary intakes of red meat,
Mismatch repair genes hMLH1 and hMSH2 and poultry, and fish during high school and risk of
colorectal cancer: a HuGE review. Am J Epidemiol colorectal adenomas in women. Am J Epidemiol
2002;156:885–​902. 2013;178:172–​83.
Momen-​Heravi F, Babic A, Tworoger SS, Zhang L, Wu Nimptsch K, Giovannucci E, Willett W, Fuchs C, Wei
K, Smith-​Warner SA et al. Periodontal disease, tooth E, Wu K. Body fatness during childhood and ado-
loss and colorectal cancer risk:  results from the lescence, adult height, and risk of colorectal ade-
Nurses’ Health Study. Int J Cancer 2017;140:646–​652. noma in women. Cancer Prev Res 2011;4:1710–​18.
Morson BC. The evolution of colorectal carcinoma. Nimptsch K, Malik V, Fung T, Pischon T, Hu F, Willett
Clin Radiol 1984;35:425–​31. W et al. Dietary patterns during high school and
Mucci LA, Hjelmborg J, Harris J, Czene K, Havelick risk of colorectal adenoma in a cohort of middle-​
D, Scheike T et  al. Familial risk and heritabil- aged women. Int J Cancer 2014;134:2458–​67.
ity of cancer in the Nordic Twin Cohorts. JAMA Ning Y, Wang L, Giovannucci EL. A quantitative
2016;315:68–​76. analysis of body mass index and colorectal can-
Muller A, Fishel R. Mismatch repair and the hered- cer: findings from 56 observational studies. Obes
itary non-​polyposis colorectal cancer syndrome Rev 2010a;11:19–​30.
(HNPCC). Cancer Invest 2002;20:102–​9. Nishihara R, Morikawa T, Kuchiba A, Lochhead P,
Murphy N, Xu L, Zervoudakis A, Xue X, Kabat G, Yamauchi M, Liao X et al. A prospective study of
Rohan TE et al. Reproductive and menstrual fac- duration of smoking cessation and colorectal can-
tors and colorectal cancer in the Womens’ Health cer risk by epigenetics-​related tumor classifica-
Initiative Observational Cohort. Br J Cancer 2016. tion. Am J Epidemiol 2013;178:84–​100.
[Epub ahead of print]. Nishihara R, Wu K, Lochhead P, Morikawa T, Liao X,
Murthy S, Flanigan A, Clearfield H. Colorectal can- Qian ZR et al. Long-​term colorectal-​cancer inci-
cer in inflammatory bowel disease: molecular and dence and mortality after lower endoscopy. N
clinical features. Gastroenterol Clin North Am Engl J Med 2013;369:1095–​1105.
2002;31:551–​64. Norat T, Lukanova A, Ferrari P, Riboli E. Meat con-
Muzny D, Bainbridge M, Chang K, Dinh H, sumption and colorectal cancer risk:  dose-​
Drummond J, Fowler G et  al. Comprehensive response meta-​analysis of epidemiological studies.
molecular characterization of human colon and Int J Cancer 2002;98:241–​56.
rectal cancer. Nature 2012;487:330–​37. Nosho K, Irahara N, Shima K, Kure S, Kirkne, GJ,
Nagata N, Sakamoto K, Arai T, Niikura R, Shimbo T, Schernhammer ES et al. Comprehensive biostatis-
Shinozaki M et  al. Visceral abdominal fat meas- tical analysis of CpG island methylator phenotype
ured by computed tomography is associated with in colorectal cancer using a large population-​
an increased risk of colorectal adenoma. Int J based sample. PLoS One 2008;3:e3698.
Cancer 2014;135:2273–​81. Oddone E, Modonesi C, Gatta G. Occupational
Nakatsuka H, Shimizu Y, Yamamoto T, Sekine I, Ezaki exposures and colorectal cancers:  a quantitative
H, Tahara E et  al. Colorectal cancer incidence overview of epidemiological evidence. World J
among atomic bomb survivors, 1950–​80. J Radiat Gastroenterol 2014;20:12431–​44.
Res 1992;33:342–​61. Ogino S, Brahmandam M, Cantor M, Namgyal C,
Nan H, Hutter CM, Lin Y, Jacobs EJ, Ulrich CM, White Kawasaki T, Kirkner G et  al. Distinct molecular
E et al. Association of aspirin and NSAID use with features of colorectal carcinoma with signet ring
risk of colorectal cancer according to genetic vari- cell component and colorectal carcinoma with
ants. JAMA 2015; 313:1133–​42. mucinous component. Mod Pathol 2006;19:59–​68.
Neugut AI, Jacobson JS, De Vivo I. Epidemiology of Ogino S Goel A. Molecular classification and correlates
colorectal adenomatous polyps. Cancer Epidemiol in colorectal cancer. J Mol Diagn 2008;10:13–​27.
Biomarkers Prev 1993;2:159–​76. Ogino S, Kawasaki T, Kirkner GJ, Kraft P, Loda M,
Newmark HL, Wargovich MJ, Bruce WR. Colon can- Fuchs CS. Evaluation of markers for CpG island
cer and dietary fat, phosphate, and calcium:  a methylator phenotype (CIMP) in colorectal can-
hypothesis. J Natl Cancer Inst 1984;72:1323–​25. cer by a large population-​based sample. J Mol
Ngo SN, Williams DB, Cobiac L, Head RJ. Does gar- Diagn 2007;9:305–​14.
lic reduce risk of colorectal cancer? A systematic Ogino S, Lochhead P, Chan AT, Nishihara R, Cho
review. J Nutr 2007;137:2264–​69. E, Wolpin BM et  al. Molecular pathological
273

Colorectal Cancer 273

epidemiology of epigenetics: emerging integrative cohort of middle-​aged US men. Cancer Causes

science to analyze environment, host, and disease. and Control 2000;11:579–​88.
Mod Pathol 2013;26:465–​84. Podolsky DK. Inflammatory bowel disease. N Engl J
Ollberding NJ, Wilkens LR, Henderson BE, Med 2002;347:417–​29.
Kolonel LN, Le Marchand L. Meat consump- Pollak M. Insulin and insulin-​like growth factor signal-
tion, heterocyclic amines and colorectal cancer ling in neoplasia. Nat Rev Cancer 2008;8:915–​28.
risk:  the Multiethnic Cohort Study. Int J Cancer Poynter JN, Haile RW, Siegmund KD, Campbell PT,
2012;131:E1125–​33. Figueiredo JC, Limburg P et  al. Registry CCF.
Orlich MJ, Singh PN, Sabate J, Fan J, Sveen L, Associations between smoking, alcohol consump-
Bennett H et  al. Vegetarian dietary patterns and tion, and colorectal cancer, overall and by tumor
the risk of colorectal cancers. JAMA Intern Med microsatellite instability status. Cancer Epidemiol
2015;175:767–​76. Biomarkers Prev 2009;18:2745–​50.
Osterhues A, Holzgreve W, Michels KB. Shall we put Poynter JN, Siegmund KD, Weisenberger DJ, Long
the world on folate? Lancet 2009;74:959–​61. TI, Thibodea SN, Lindor N et  al., Investigators
Otani T, Iwasaki M, Ishihara J, Sasazuki S, Inoue M, CCFR. Molecular characterization of MSI-​ H
Tsugane S, Japan Public Health Center-​ Based colorectal cancer by MLHI promoter methylation,
Prospective Study, G. Dietary fiber intake and immunohistochemistry, and mismatch repair
subsequent risk of colorectal cancer:  the Japan germline mutation screening. Cancer Epidemiol
Public Health Center-​based prospective study. Int Biomarkers Prev 2008;17:3208–​15.
J Cancer 2006;119:1475–​80. Pradhan A, Manson J. Update on the vitamin D and
Park Y, Hunter D, Spiegelman D, Bergkvist L, Berrino omegaA-​3 trial (vital). J Steroid Biochem Mol Biol
F, Brandt P et al. Dietary fiber intake and risk of 2016;155:252–​6.
colorectal cancer: a pooled analysis of prospective Qaseem A, Denberg TD, Hopkins RH Jr, Humphrey
cohort studies. JAMA 2005;294:2849–​57. LL, Levine J, Sweet DE, Shekelle P. Screening for
Parkin DM. International variation. Oncogene colorectal cancer: a guidance statement from the
2004;23:6329–​40. American College of Physicians. Ann Intern Med
Peltomaki P. Update on Lynch syndrome genomics. 2012;156:378–​86.
Fam Cancer 2016;15:385–​93. Rajakariar R, Yaqoob MM, Gilroy DW. COX-​ 2
Petroff D, Tonjes A, Grussendorf M, Droste M, in inflammation and resolution. Mol Interv
Dimopoulou C, Stalla G et  al. The incidence of 2006;6:199–​207.
cancer among acromegaly patients:  results from Renehan AG, Painter JE, Atkin WS, Potten CS, Shalet
the german acromegaly registry. J Clin Endocrinol SM, O’Dwyer ST. High-​risk colorectal adenomas
Metab 2015;100:3894–​3902. and serum insulin-​like growth factors. Br J Surg
Pham NM, Mizoue T, Tanaka K, Tsuji I, Tamakoshi 2001;88:107–​13.
A, Matsuo K et  al. Meat consumption and colo- Rennoll S, Yochum G. Regulation of MYC gene expres-
rectal cancer risk:  an evaluation based on a sion by aberrant Wnt/​β-​catenin signaling in colo-
systematic review of epidemiologic evidence rectal cancer. World J Biol Chem 2015;6:290–​300.
among the Japanese population. Jpn J Clin Oncol Rinaldi S, Cleveland R, Norat T, Biessy C, Rohrmann
2014;44:641–​50. S, Linseisen J et al. Serum levels of IGF-​I, IGFBP-​3
Pierre FH, Martin OC, Santarelli RL, Tache S, Naud and colorectal cancer risk: results from the EPIC
N, Gueraud F et al. Calcium and alpha-​tocopherol cohort, plus a meta-​analysis of prospective stud-
suppress cured-​ meat promotion of chemically ies. Int J Cancer 2010;126:1702–​15.
induced colon carcinogenesis in rats and reduce Rogot E, Murray JL. Smoking and causes of death
associated biomarkers in human volunteers. Am J among US veterans:  16  years of observation.
Clin Nutr 2013;98:1255–​62. Public Health Rep 1980;95:213–​22.
Pino MS, Chung DC. The chromosomal instabil- Rokkas T, Pistiolas D, Sechopoulos P, Margantinis
ity pathway in colon cancer. Gastroenterology G, Koukoulis G. Risk of colorectal neoplasm in
2010;138:2059–​72. patients with acromegaly: a meta-​analysis. World
Pischon T, Lahmann PH, Boeing H, Friedenreich C, J Gastroenterol 2008;14:3484–​89.
Norat T, Tjonneland A et al. Body size and risk of Roncucci L, Mariani F. Prevention of colorectal can-
colon and rectal cancer in the European Prospective cer: how many tools do we have in our basket? Eur
Investigation Into Cancer and Nutrition (EPIC). J J Intern Med 2015;26:752–​6.
Natl Cancer Inst 2006;98:920–​31. Ross R, Dagnone D, Jones PJ, Smith H, Paddags A,
Platz EA, Willett WC, Colditz GA, Rimm EB, Hudson R, Janssen I. Reduction in obesity and
Spiegelman D, Giovannucci E. Proportion of related comorbid conditions after diet-​ induced
colon cancer risk that might be preventable in a weight loss or exercise-​ induced weight loss in
274

274 Textbook of Cancer Epidemiology

men:  a randomized, controlled trial. Ann Intern incidence and mortality with screening flexible
Med 2000;133:92–​103. sigmoidoscopy. N Engl J Med 2012;366:2345–​57.
Ross R, Janssen I, Dawson J, Kungl AM, Kuk JL, Wong Schwabe RF, Jobin C. The microbiome and cancer. Nat
SL et  al. Exercise-​induced reduction in obesity Rev Cancer 2013;13:800–​12.
and insulin resistance in women:  a randomized Sebastian C, Mostoslavsky R. Untangling the fiber
controlled trial. Obes Res 2004;12:789–​98. yarn: butyrate feeds Warburg to suppress colorec-
Rosty C, Hewett DG, Brown IS, Leggett BA, Whitehall tal cancer. Cancer Discov 2014;4:1368–​70.
VL. Serrated polyps of the large intestine: current SEER. Surveillance, Epidemiology, and End Results
understanding of diagnosis, pathogenesis, and clin- (SEER) Stat Fact Sheets:  Colon and Rectum
ical management. J Gastroenterol 2013;48:287–​302. Cancer. Available from:  http://​seer.cancer.gov/​
Rothwell PM, Wilson M, Elwin CE, Norrving B, Algra statfacts/​html/​colorect.html (accessed 2015).
A, Warlow CP, Meade TW. Long-​term effect of Seitz HK, Becker P. Alcohol metabolism and cancer
aspirin on colorectal cancer incidence and mor- risk. Alcohol Res Health 2007;30:38–​41.
tality: 20-​year follow-​up of five randomised trials. Shang J, Reece JC, Buchanan DD, Giles GG,
Lancet 2010;376:1741–​50. Figueiredo JC, Casey G et  al. Cholecystectomy
Rubinstein MR, Wang X, Liu W, Hao Y, Cai G, Han and the risk of colorectal cancer by tumor mis-
YW. Fusobacterium nucleatum promotes colo- match repair deficiency status. Int J Colorect Dis
rectal carcinogenesis by modulating E-​cadherin/​ 2016;31:1451–​57.
beta-​catenin signaling via its FadA adhesin. Cell Shaukat A, Scouras N, Schunemann HJ. Role of
Host Microbe 2013;14:195–​206. supplemental calcium in the recurrence of
Ruder E, Thiebaut A, Thompson F, Potischman colorectal adenomas:  a metaanalysis of ran-
N, Subar A, Park Y et  al. Adolescent and mid-​ domized controlled trials. Am J Gastroenterol
life diet:  risk of colorectal cancer in the NIH-​ 2005;100:390–​4.
AARP diet and health study. Am J Clin Nutr Shi Y, Yu PW, Zeng DZ. Dose-​response meta-​analysis
2011;94:1607–​19. of poultry intake and colorectal cancer incidence
Salaspuro M. Acetaldehyde as a common denom- and mortality. Eur J Nutr 2015;54;243–​50.
inator and cumulative carcinogen in diges- Siegel R, Desantis C, Jemal A. Colorectal cancer statis-
tive tract cancers. Scand J Gastroenterol tics, 2014. CA Cancer J Clin 2014;64:104–​17.
2009;44:912–​25. Simons CC, Hughes LA, Smits KM, Khalid-​de Bakker
Samowitz W. Genetic and epigenetic change in colon CA, de Bruine AP, Carvalho B et al. A novel clas-
cancer. Exp Mol Pathol 2008;85:64–​7. sification of colorectal tumors based on micro-
Samowitz WS, Albertsen H, Herrick J, Levin TR, satellite instability, the CpG island methylator
Sweeney C, Murtaugh MA et  al. Evaluation of a phenotype and chromosomal instability: implica-
large, population-​based sample supports a CpG tions for prognosis. Ann Oncol 2013;24:2048–​56.
island methylator phenotype in colon cancer. Singh S, Singh H, Singh P, Murad M, Limburg P.
Gastroenterology 2005;129:837–​45. Antidiabetic medications and the risk of colo-
Sandhu MS, White IR, McPherson K. Systematic rectal cancer in patients with diabetes mellitus: a
review of the prospective cohort studies on meat systematic review and meta-​ analysis. Cancer
consumption and colorectal cancer risk:  a meta-​ Epidemiol Biomarkers Prev 2013;22:2258–​68.
analytic approach. Cancer Epidemiol Biomarkers Sinha R, Norat T. Meat cooking and cancer risk. IARC
Prev 2001;10:439–​46. Sci Publ 2002;156:181–​6.
Santarelli RL, Naud N, Tache S, Gueraud F, Vendeuvre Sinha R, Peters U, Cross AJ, Kulldorff M, Weissfeld
JL, Zhou L et  al. Calcium inhibits promotion JL, Pinsky PF et al. Meat, meat cooking methods
by hot dog of 1,2-​ dimethylhydrazine-​induced and preservation, and risk for colorectal adenoma.
mucin-​depleted foci in rat colon. Int J Cancer Cancer Res 2005;65:8034–​41.
2013;133:2533–​41. Slattery ML. Physical activity and colorectal cancer.
Schatzkin A, Mouw T, Park Y, Subar AF, Kipnis V, Sports Med 2004;34:239–​52.
Hollenbeck A et al. Dietary fiber and whole-​grain Snover DC. Update on the serrated pathway to colo-
consumption in relation to colorectal cancer in rectal carcinoma. Hum Pathol 2011;42:1–​10.
the NIH-​AARP Diet and Health Study. Am J Clin Song M, Hu F, Spiegelman D, Chan A, Wu K, Ogino S
Nutr 2007;5:1353–​60. et al. Long-​term status and change of body fat dis-
Schmid D, Leitzmann MF. Television viewing and tribution, and risk of colorectal cancer: a prospec-
time spent sedentary in relation to cancer risk: a tive cohort study. Int J Epidemiol 2016;45:871–​83.
meta-​analysis. J Natl Cancer Inst 2014;106. Song M, Willett W, Hu F, Spiegelman D, Must A, Wu K
Schoen RE, Pinsky PF, Weissfeld JL, Yokochi LA, et al. Trajectory of body shape across the lifespan
Church T, Laiyemo AO et  al. Colorectal-​cancer and cancer risk. Int J Cancer 2016;138:2383–​95.
275

Colorectal Cancer 275

Sostres C, Gargallo CJ, Lanas A. Aspirin, cyclooxy- Wajchenberg BL. Subcutaneous and visceral adipose
genase inhibition and colorectal cancer. World J tissue:  their relation to the metabolic syndrome.
Gastrointest Pharmacol Ther 2014;5:40–​9. Endocr Rev 2000;21:697–​738.
Soumaoro LT, Uetake H, Higuchi T, Takagi Y, Enomoto Wakai K, Date C, Fukui M, Tamakoshi K, Watanabe Y,
M, Sugihara, K. Cyclooxygenase-​2 expression: a sig- Hayakawa N et al. Dietary fiber and risk of colorec-
nificant prognostic indicator for patients with colo- tal cancer in the Japan collaborative cohort study.
rectal cancer. Clin Cancer Res 2004;10:8465–​71. Cancer Epidemiol Biomarkers Prev 2007;16:668–​75.
Steck S, Guinter M, Zheng J, Thomson, C. Index-​based Wallace K, Baron JA, Cole BF, Sandler RS, Karagas
dietary patterns and colorectal cancer risk: a sys- MR, Beach MA et  al. Effect of calcium supple-
tematic review. Adv Nutr 2015;6:763–​73. mentation on the risk of large bowel polyps. J Natl
Tadros M, Anderson JC. Serrated polyps:  clini- Cancer Inst 2004;96:921–​5.
cal implications and future directions. Curr Wallis CJ, Mahar AL, Choo R, Herschorn S, Kodama
Gastroenterol Rep 2013;15:342. RT, Shah PS et al. Second malignancies after radi-
Tanaka T. Colorectal carcinogenesis: Review of human otherapy for prostate cancer:  systematic review
and experimental animal studies. J Carcinog and meta-​analysis. BMJ 2016;352:i851.
2009;8:5. Wargovich MJ, Eng VWS, Newmark H. Calcium
Thomas D. Gene–​ environment-​ wide association inhibits the damaging and compensatory prolifer-
studies:  emerging approaches. Nat Rev Genet ative effects of fatty acids on mouse colon epithe-
2010;11:259–​72. lium. Cancer Lett 1984;23:253–​8.
Toyota M, Ohe-​Toyota M, Ahuja N, Issa JP. Distinct Webber EM, Kauffman TL, O’Connor E, Goddard KA.
genetic profiles in colorectal tumors with or with- Systematic review of the predictive effect of MSI sta-
out the CpG island methylator phenotype. Proc tus in colorectal cancer patients undergoing 5FU-​
Natl Acad Sci U S A 2000;97:710–​15. based chemotherapy. BMC Cancer 2015;15:156.
Triantafillidis JK, Nasioulas G, Kosmidis PA. Wei EK, Ma J, Pollak MN, Rifai N, Fuchs CS,
Colorectal cancer and inflammatory bowel dis- Hankinson SE et  al. A prospective study of C-​
ease:  epidemiology, risk factors, mechanisms peptide, insulin-​like growth factor-​I, insulin-​like
of carcinogenesis and prevention strategies. growth factor binding protein-​1, and the risk of
Anticancer Res 2009;29:2727–​37. colorectal cancer in women. Cancer Epidemiol
Tsai S, Gearhart S. Presentation and initial eval- Biomarkers Prev 2005;14:850–​5.
uation of colorectal cancer. In:  Gearhart S, Weisenberger DJ, Siegmund KD, Campan M, Young
Ahuja N (eds.) Early Diagnosis and Treatment J, Long TI, Faasse MA et  al. CpG island meth-
of Cancer. Philadelphia, PA:  Saunders Elsevier, ylator phenotype underlies sporadic microsat-
2011: 13–​19. ellite instability and is tightly associated with
Turati F, Guercio V, Pelucchi C, Vecchia CL, Galeone BRAF mutation in colorectal cancer. Nat Genet
C. Colorectal cancer and adenomatous polyps 2006;38:787–​93.
in relation to allium vegetables intake:  a meta-​ Wheeler JM, Bodmer WF, Mortensen NJ. DNA mis-
analysis of observational studies. Mol Nutr Food match repair genes and colorectal cancer. Gut
Res 2014;58:1907–​14. 2000;47:148–​53.
Ulrich CM. Folate and cancer prevention:  a closer Winawer SJ, Zauber AG, O’Brien MJ, Ho MN, Gottlieb
look at a complex picture. American J Clin Nutr L, Sternberg SS et al. Randomized comparison of
2007;86:271–​73. surveillance intervals after colonoscopic removal
Vasen HF, Watson P, Mecklin JP, Lynch HT. New clin- of newly diagnosed adenomatous polyps:  the
ical criteria for hereditary nonpolyposis colorectal National Polyp Study Workgroup. N Engl J Med
cancer (HNPCC, Lynch syndrome) proposed by 1993;328:901–​6.
the International Collaborative group on HNPCC. Wolin K, Yan Y, Colditz G, Lee I. Physical activity and
Gastroenterology 1999;116:1453–​56. colon cancer prevention:  a meta-​ analysis. Br J
Vogelstein B, Fearon ER, Hamilton, SR, Kern SE, Cancer 2009;100:611–​6.
Preisinger AC, Leppert M et al. Genetic alterations Wolin KY, Yan Y, Colditz GA. Physical activity and
during colorectal-​tumor development. N Engl J risk of colon adenoma:  a meta-​ analysis. Br J
Med 1998;319:525–​32. Cancer 2011;104:882–​5.
Vollset SE, Clark R, Lewington S, Ebbing M, Halsey Wong NA, Harrison DJ. Colorectal neoplasia in ulcer-
J, Lonn E et  al. Collaboration BVTT. Effects of ative colitis-​ recent advances. Histopathology
folic acid supplementation on overall and site-​ 2001;39:221–​34.
specific cancer incidence during the randomised Wood L, Parsons D, Jones S, Lin J, Sjoblom T, Leary
trials: meta-​analyses of data on 50,000 individuals. R et al. The genomic landscapes of human breast
Lancet 2013;381:1029–​36. and colorectal cancers. Science 2007;318:1108–​13.
276

276 Textbook of Cancer Epidemiology

World Cancer Research Fund/​American Institute for IGF-​1/​IGFBP-​ 3 molar ratio and colorectal
Cancer Research (WCRF/​ AICR). Continuous adenomas:  a meta-​ analysis. Cancer Epidemiol
Update Project Report. Food, Nutrition, Physical 2015a;39:1026–​35.
Activity, and the Prevention of Colorectal Cancer, Yoon YS, Keum N, Zhang X, Cho E, Giovannucci EL.
2011: 1–​40. Hyperinsulinemia, insulin resistance and colo-
World Cancer Research Fund/​American Institute for rectal adenomas:  a meta-​ analysis. Metabolism
Cancer Research (WCRF/​AICR). Food, Nutrition, 2015b;64:1324–​33.
Physical Activity, and the Prevention of Cancer: a Zauber AG, Winawer SJ, O’Brien MJ, Lansdorp-​
Global Perspective. Washington, DC:  AICR, Vogelaar I, van Ballegooijen M, Hankey BF et al.
2007: 1–​517. Colonoscopic polypectomy and long-​term pre-
Wu K, Feskanich D, Fuchs CS, Chan AT, Willett WC, vention of colorectal-​cancer deaths. N Engl J Med
Hollis BW et al. Interactions between plasma lev- 2012;366:687–​96.
els of 25-​hydroxyvitamin D, insulin-​like growth Zhang X, Wu K, Giovannucci E, Ma J, Colditz G, Fuchs
factor (IGF)-​1 and C-​peptide with risk of colorec- C et al. Early life body fatness and risk of colorec-
tal cancer. PLoS One 2011;6:e28520. tal cancer in U.S.  women and men-​results from
Yamaji Y, Mitsushima T, Koike K. Pulse-​wave velocity, two large cohort studies. Epidemiol Biomarkers
the ankle-​brachial index, and the visceral fat area Prev 2015;24:690–​97.
are highly associated with colorectal adenoma. Zeng C, Matsuda K, Jia WH, Chang J, Kweon SS,
Dig Liver Dis 2014;46:943–​9. Xiang YB et  al. Identification of Susceptibility
Yaeger R, Shah MA, Miller VA, Kelsen JR, Wang K, Heins Loci and Genes for Colorectal Cancer Risk.
ZJ et  al. Genomic alterations observed in colitis-​ Gastroenterology 2016;150:1633–​45.
associated cancers are distinct from those found Zhao C, Ge Z, Wang Y, Qian J. Meta-​analysis of obser-
in sporadic colorectal cancers and vary by type vational studies on cholecystectomy and the
if inflammatory bowel disease. Gastroenterology risk of colorectal adenoma. Eur J Gastroenterol
2016;151:278–​87. Hepatol 2012;24:375–​81.
Yao X, Tian Z. Dyslipidemia and colorectal can- Zorzi M, Fedeli U, Schievano E, Bovo E, Guzzinati
cer risk:  a meta-​analysis of prospective studies. S, Baracco S et  al. Impact on colorectal can-
Cancer Causes Control 2015;26:257–​68. cer mortality of screening programmes based
Yoon YS, Keum N, Zhang X, Cho E, Giovannucci on the faecal immunochemical test. Gut
EL. Circulating levels of IGF-​ 1, IGFBP-​ 3, and 2015;64:784–​90.
27

12
Cancer of the Liver and Biliary Tract
CHRISTINA BAMIA, SHERRI STUVER, AND LORELEI MUCCI

P rimary liver cancer is an uncommon dis-

ease in the Western world, but liver can-
cer is important for several reasons. It is quite
Predominant are HCC and cholangiocarcinoma
of the liver and adenocarcinomas of the gallblad-
der and extrahepatic bile ducts.
commonly diagnosed in Asia and sub-​Saharan Symptoms The main symptoms are those of an
Africa. It is a rapidly and almost uniformly fatal advanced abdominal malignancy such as pain,
disease for which treatment is generally ineffec- anorexia, weight loss, weakness, fever, and
tive, but there is already sufficient knowledge abdominal swelling. Jaundice is particularly
and technology for effective primary prevention. common in bile duct cancers.
Indeed, more important causal factors have been
Diagnosis Clinical examination and analysis of
identified for three histologic forms of primary
blood levels of alpha-​fetoprotein, as well as ultra-
liver cancer (hepatocellular carcinoma [HCC],
sound, computed tomography, and other imag-
cholangiocarcinoma, and angiosarcoma) than
ing techniques, are used for diagnosis and staging.
for any of the common cancers except lung and
Histopathologic confirmation requires liver biopsy
cervix cancer. Liver cancer is also the first com-
specimens obtained percutaneously, laparoscop-
mon human cancer that was found to have an
ically, or via open surgery. Percutaneous transhe-
infectious etiology. Following clinical observa-
patic or endoscopic retrograde cholangiography
tions made in the 1940s, a causal role of hepati-
documents biliary anatomy in bile duct tumors.
tis B virus (HBV) was documented in the 1970s,
while the corresponding role of hepatitis C virus Treatment Hepatobiliary cancers are curable
(HCV) emerged about a decade later. Hence, the only through radical surgical resection or, occa-
discovery of the causes of primary liver cancer, sionally, hepatectomy followed by liver trans-
and in particular HCC, is a success story of epi- plantation. For intrahepatic bile duct cancer,
demiology with rich input from ecologic, case-​ the surgical approach is similar to that of HCC.
control, and cohort studies. Progress has been made in the development of
Metastatic liver cancers are 18 to 40 times ablative localized treatments, which have dem-
more common than primary liver cancer. The onstrated some effectiveness among patients
most common primary cancers of origin are with early stage, limited disease; these therapies
breast, lung, or colorectal (Namasivayam et  al, include percutaneous ethanol injection, radiofre-
2007). The high incidence of metastases to the quency ablation, and transarterial embolization
liver has been attributed both to the dual blood and chemoembolization. However, most cancers
supply of the liver from the portal and systemic are diagnosed in advanced stages, when curative
circulation (increasing the chances for metastatic treatment is beyond reach.
deposits) and to the fenestrations in hepatic Prognosis The prognosis for patients with hepa-
sinusoidal epithelium that enable easier penetra- tobiliary cancer is extraordinarily gloomy. The
tion of metastatic cells into the liver parenchyma median survival time following diagnosis is
(Kew, 2002). This chapter focuses on the epide- short, and 5-​year relative survival rates in Europe
miology of primary liver cancer. for those diagnosed with liver cancer between
2000 and 2007 are approximately 12% (De
CLINICAL SYNOPSIS Angelis et  al, 2014). The overall 5-​year survival
Subgroups Hepatobiliary cancers encompass a of gallbladder cancer is less than 5%, with a mean
variety of anatomic and histopathologic types. survival time of 6 months (Lai and Lu, 2008).
278

278 Textbook of Cancer Epidemiology

Progress Notwithstanding the development of a higher rate of liver cancer than women (Figs.
numerous, often advanced, techniques, there has 12-​1, 12-​2), and this excess is more evident in
been no substantial progress in the treatment of populations with elevated liver cancer rates
hepatobiliary cancer. The most promising devel- (Stuver & Trichopoulos, 1994; Parkin et al, 2005;
opment in the control of this dreaded disease is Ferlay et al, 2013). In addition, the occurrence
primary prevention of HCC through vaccination of liver cancer in persons less than 45  years of
against HBV and interruption of HCV transmis- age is higher in high-​risk countries for both
sion and chronic infection. sexes (Figs. 12-​3, 12-​4). Due to revisions in the
classification and coding of primary liver can-
cer, as well as changes in diagnostic and certi-
LIVER CANCER fication practices, secular trends in liver cancer
Descriptive Epidemiology are difficult to describe (Stuver & Trichopoulos,
Primary liver cancer is one of the most fre- 1994). Nonetheless, liver cancer rates seem to
quently occurring malignancies in the world. be increasing in a number of Western countries
Incidence rate estimates in 2012 rank primary including the United States (El-​Serag et al, 2014;
liver cancer sixth with regard to the global can- Altekruse et  al, 2009), the United Kingdom
cer burden (Ferlay et  al, 2013, 2015). It has a (El-​Serag & Rudolph, 2007), and France (La
very high fatality rate and thus is an even more Vecchia et al, 2000; McGlynn et al, 2015).
important cause of cancer mortality world- Trends in liver cancer incidence rates from
wide, ranked as second (Ferlay et  al, 2013). 1973 to 2007 have been reported (Stuver &
International liver cancer rates are extremely Trichopoulos, 1994; McGlynn et  al, 2001; Zhang
variable (Stuver & Trichopoulos, 1994; Bosch et  al, 2015), with evidence of decreasing trends
et  al, 2005; Ferlay et  al, 2013), and about 83% from 1982 to 2007 in some historically high-​risk
of all primary liver cancer cases occur in low-​ areas such as China, Singapore, and Japan (only
and middle-​ income countries (Ferlay et  al, after 1993), possibly due to the HBV vaccine and
2013). Rates are highest in eastern and south- surveillance programs that have been applied
eastern Asia and western Africa and lowest in in these countries (Chang et  al, 2009; Zhang
the south-​central Asia, Americas, Australia, and et al 2015).
northern Europe (Figs. 12-​1, 12-​2) (Jemal et al, Although the global burden of liver cancer
2011; Ferlay et al 2013). Intracountry variation is substantial, this malignancy is largely pre-
in liver cancer occurrence also can be quite ventable. We know as much about its etiology
pronounced (McGlynn et  al, 2001). Men have as about that of any other common form of

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 2 4 6 8 10 12
ASR (world) per 100,000 person-years

FIGURE 12-​1  Age-​standardized (to the 2012 world population) incidence rates of liver cancer among women.
Source: Ferlay et al, 2013.
279

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 5 10 15 20 25 30 35 40
ASR (world) per 100,000 person-years

FIGURE 12-​2  Age-​standardized (to the 2012 world population) incidence rates of liver cancer among men.
Source: Ferlay et al, 2013.

160
New cases per 100,000

140
120
person-years

100
80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 12-​3  Age-​specific incidence rates of liver cancer among women.

Source: Ferlay et al, 2013.

350
New cases per 100,000

300
person-years

250
200
150
100
50
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 12-​4  Age-​specific incidence rates of liver cancer among men.

Source: Ferlay et al, 2013.
280

280 Textbook of Cancer Epidemiology

human cancer; more than 10 distinct biolog- those in tobacco smoke. Alcohol consumption
ical, chemical, and physical agents have been can induce CYP2E1, and the highest levels of the
conclusively incriminated. Indeed, HCC, the enzyme are found in the liver (Lee et  al, 1997;
most frequent histologic type, is the first com- Boccia et  al, 2015). The GSTs and EPHX are
mon cancer for which a viral etiology has been involved in the process of aflatoxin B1 detoxifi-
conclusively established. It is also important cation in hepatocytes. In a 2008 meta-​analysis,
that the evidence basis that revealed the etiol- genetic variants in GSTM1 or GSTT1 were asso-
ogy of liver cancer has been overwhelmingly ciated with an increased risk of HCC (White
from epidemiological studies. et al, 2008).
To date, few genome-​wide association stud-
Genetic and Molecular ies (GWAS) of primary liver cancer have been
Epidemiology undertaken. The first GWAS for HCC in chronic
Inherited Susceptibility HBV carriers (Zhang et al, 2010) included 355
Most clustering of HCC observed within fami- HBV, HCC cases and 360 HBV controls in China.
lies likely reflects shared risk factors, most nota- Rs17401966 located on chromosome 1p36.22
bly chronic HBV infection. However, family was identified as a new susceptibility variant for
history has remained associated with risk of HBV-​related HCC. GWAS of HBV-​related HCC
HCC even with adjustment for other known risk in Chinese individuals identified 4 SNPs in 8p12
factors (Chen et al, 1997). Several inherited dis- chromosome (Chan et al, 2011), as well as the
orders, including alpha1-​antitrypsin deficiency, rs3771333 SNP in HJURP gene (Huang et al,
hemochromatosis, and porphyria cutanea tarda, 2016). Qu et al (2016) undertook a two-​phase
have been associated with HCC (Deugnier & GWAS and found genes of common susceptibil-
Turlin, 1997; McGlynn & London, 2005; Fix ity loci for HCC in signaling pathways, including
& Kowdley, 2008). Most of the evidence for an those associated with transforming growth factor-​
association with these diseases comes from case beta (TGF-​β), insulin/​phosphoinositide 3 kinase,
series studies, although some analytic epidemi- Wnt, and epidermal growth factor receptor.
ologic investigations have provided important
data (Sparos et  al, 1984). Several studies have Somatic Events
demonstrated an association between HCC and Hepatocarcinogenesis likely involves a multi-
porphyria in Europe (Linet et al, 1999; Andant step process representing an accumulation of
et  al. 2000; Schneider-​Yin et  al. 2009; Innala somatic genetic changes. Cirrhosis, observed
& Andersson 2011; Elder et  al. 2013; Sardh in the majority of HCC cases, appears to play
et al. 2013). Among 122 individuals with acute an important role in mutation events through
hepatic porphyrias from a single European por- the associated cycles of liver necrosis and regen-
phyria specialist center, there was an estimated eration (Tabor, 1998; Thorgeirsson & Grisham,
35-​fold increased risk of HCC (Lang et al, 2015). 2002; Lee 2013). DNA damage response, cell
The inherited metabolic disorders often result in cycle control, and apoptosis have been related
inflammation, fibrosis, and/​or cirrhosis of the to the development of HCC (Ozturk, 1999;
liver, which likely is the mechanism by which Thorgeirsson & Grisham, 2002; Lee 2015). The
they may be associated with the development of application of microarray technology to the
liver cancer (Deugnier & Turlin, 1997; Hussain study of the molecular pathogenesis of HCC has
et al, 2007). further advanced genetic research in hepatocar-
Specific genetic studies have yielded conflict- cinogenesis (Shirota et al, 2001; Thorgeirsson &
ing results. Directly opposite effects have been Grisham, 2002; Lee 2015).
reported for the association of specific genotypes Mutations of tumor-​suppressor genes, in par-
of enzymes, including cytochrome P450 2E1 ticular p53, and inactivation of their proteins are
(CYP2E1) (Yu et al, 1995a; Lee et al, 1997), the often observed in HCC (Tornesello et al, 2013).
glutathione S-​transferases (GSTM1 and GSTT1) About 30% of HCC cases worldwide, and up to
(McGlynn et  al, 1995, 2003; Yu et  al, 1995; Liu 60% of cases in high-​risk areas, are reported to
et al, 2013), and the epoxide hydrolases (EPHX) have p53 mutations (Tabor, 1998; Ozturk, 1999).
(McGlynn et al, 1995, 2003; Zhong et al, 2013). The mutation of p53 codon 72 has been associ-
CYP2E1 is involved in the oxidation and met- ated with excess risk of liver cancer in individ-
abolic activation of N-​nitrosamines, including ual studies, as well as in a pooled analysis of 11
281

Cancer of the Liver and Biliary Tract 281

case-​control studies (Lv et  al, 2013). Aflatoxin alcohol consumption—​or that may be directly
B1 exposure has been linked to specific muta- genotoxic—​for example, tobacco smoke or die-
tions of codon 249 of the p53 gene (IARC, 1993; tary aflatoxin—​ appear to be important. Such
Montesano et  al, 1997). It is hypothesized that environmental exposures, as well as host-​related
these p53-​related events occur late in the pro- factors such as male sex, may also act to mod-
gression of tumor development, resulting in the ify the development of virus-​ associated liver
loss of cell growth control through p53-​mediated malignancy.
apoptosis (Tabor, 1997; Thorgeirsson & Grisham, Unlike the situation with most other can-
2002). Loss of heterozygosity of the retinoblas- cers, a substantial fraction of primary liver can-
toma gene (RB1), which plays a role in cell-​cycle cer cases and deaths could actually be prevented
regulation, has been found in some HCC cases given the existing knowledge and technology.
(Tabor, 1994; Ozturk, 1999; Kummee et al, 2007). Eradication of HBV infection, reduction of HCV
In addition, RB1 mutations, which result in sup- transmission and its treatment, and improve-
pression of transcription of the c-​myc and c-​fos ment of socioeconomic conditions could dra-
promoters in vitro, may lead to uncontrolled matically diminish the occurrence of this cancer
expression of these oncogenes. Other genes in the affected populations. Public health pro-
important to cell-​ cycle regulation, including grams for the surveillance, diagnosis, and treat-
p15INK4A and cyclin D and A gene expression, ment of liver cancer have been applied in Japan
have been found to be mutated in 10%–​20% of with promising results regarding the incidence
HCCs (Ozturk, 1999; Shiraha et al, 2013). Tumor of and mortality from liver cancer (Chang et al,
necrosis factor (TNF)-​α cytokine has been also 2009; Kudo, 2015). However, because liver can-
frequently studied in relation to the risk of devel- cer remains a malignancy primarily of low-​
oping HCC, and a positive association between income countries and disadvantaged groups,
the 308 position (TNF-​ α-​308) and HCC risk efforts to reduce its burden are advancing slowly.
has been confirmed in various reports and In addition, moderation of alcohol consumption,
meta-​analyses (Guo et al, 2010; Qin et al, 2010; prevention of obesity, and avoidance of smoking
Tavakolpour and Sali, 2016). may have an impact on the occurrence of liver
Several growth factors have been implicated in cancer, particularly in high-​income countries,
hepatocarcinogenesis (Tabor, 1994; Thorgeirsson where incidence rates seemed to increase over
& Grisham, 2002; Shiraha et al, 2013). The level the last decades.
of transforming growth factor alpha (TGF-​α) fre-
quently is increased in HCC tissue, as detected by Tobacco
immunohistochemistry. Overexpression of TGF-​ In some industrialized countries where the prev-
α in tumorous livers may result from liver regen- alence of chronic hepatitis virus infection is low,
eration and lead to enhanced tumor progression. tobacco smoking may be an important risk factor
Cyclooxygenase 2 (COX-​2), a proangiogenic and for HCC.
growth-​promoting factor, may play a role in the The association of smoking with HCC risk was
development of HCC, particularly in the HBV-​ reported as early as in the 1980s (Trichopoulos
related disease (Tang et al, 2005; Kwon et al, 2012). et al, 1980; Lam et al, 1982). Subsequent studies
(some undertaken only in men) have found a sig-
Risk Factors nificant effect of smoking on HCC risk after con-
Like cancers of the lung and the cervix, liver can- trolling for the potentially confounding effect of
cer is one of the malignancies for which the major chronic HBV and/​or HCV infections (Yu & Chen,
etiologic factors have been identified (Table 12-​1). 1993; Chen et  al, 1997; Yang et  al, 2002; Evans
These factors appear to act through a chronic et  al, 2002; Wang et  al, 2003; Batty et  al, 2008;
process of liver damage and regeneration, pos- Lee et  al, 2009; Koh et  al, 2011; Trichopoulos
sibly through the accumulation of mutations in et al, 2011; Koh et al, 2011). In a commentary by
a stochastic manner over a long latent period epidemiologists of the International Agency for
(Tabor, 1998). Persistent infection with HBV or Research on Cancer (IARC) Working Group on
HCV, which accounts for over three-​ quarters the carcinogenicity of tobacco smoke, the relative
of all liver cancer cases in the world, results in risk estimates of HCC for smokers compared to
this type of injury to the liver. Exposures that nonsmokers range from 1.5 to 2.5 (IARC, 2004;
also damage the liver—​ for example, excessive Vineis et al, 2004; IARC, 2012d).
28

TABLE 12-​1   EPIDEMIOLOGY OF LIVER CANCER: RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Chronic HBV infection ↑↑↑
Chronic HCV infection ↑↑↑
Alcoholic cirrhosis ↑↑
Alcohol intake ↑
Dietary aflatoxins ↑↑
Opisthorchis viverrini infection (cholangiocarcinoma) ↑
Clonorchis sinensis infection (cholangiocarcinoma) ↑
Diagnostic thorium dioxide [alpha-​particles] (mostly angiosarcoma) ↑↑
Vinyl chloride (angiosarcoma) ↑
Organic trichloroethylene solvent ↑
Inorganic arsenic (angiosarcoma) ↑
Gallstones/​gallbladder disease (gallbladder cancer) ↑↑
Tobacco smoking ↑

Probable Relationship Exists, Based on Substantial Data

Inherited metabolic disorders—​alpha 1-​antitrypsin deficiency
hemochromatosis, porphyria cutanea tarda ↑↑
Nonalcoholic fatty liver disease (NAFLD) ↑
Nonalcoholic steatohepatitis (NASH) ↑
Cirrhosis of any etiology ↑
Obesity ↑
Diabetes mellitus ↑
Increased coffee consumption ↓
Primary sclerosing cholangitis (cholangiocarcinoma) ↑
Ionizing radiation [x-​and gamma-​rays] ↑

Inconsistent Findings or Limited Study to Date

Oral contraceptives ↑
Increased consumption of tea/​green tea ↓
Increased consumption of vegetables ↓
Increased consumption of fish ↓
Increased physical activity ↓
Increased serum retinol ↓
Increased selenium intake ↓
High parity ↓↑–​
Increased age at first birth –↑
Increased age at menarche ↓
Increased age at menopause ↑–​
Increased testosterone ↑–​
Increased estrogen levels ↓

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate to
large increase in risk; ↑↑↑, very large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease
in risk;—​, no association.
283

Cancer of the Liver and Biliary Tract 283

In some studies, the association between of aflatoxin exposure are highest in sub-​Saharan
smoking and HCC risk was observed only among Africa and southern China (Montesano et  al,
those negative for HBV infection (Trichopoulos 1997; Williams et  al, 2004). Ecologic studies
et al, 1980; Lam et al, 1982), Moreover, significant conducted in those areas revealed significant
interactions between cigarette smoking and HBV associations between estimated aflatoxin intake
infection, HCV infection, or both have been and HCC incidence and mortality (IARC, 1993,
observed in some studies (Mori et al, 2000; Sun 2012f). Analytic epidemiologic studies using
et al, 2003), although not in others (Wang et al, food-​ frequency questionnaires in conjunction
2003; Yuan et al, 2004). In a meta-​analysis, a more-​with data from aflatoxin-​ contamination food
than-​additive interaction between HBV and cig- surveys supported an association between afla-
arette smoking, and a more-​than-​multiplicative toxin consumption and liver cancer risk (IARC,
interaction between HCV and cigarette smoking 1993, 2012f), although some studies found no
was suggested (Chuang et al, 2010). A synergis- such relationship (Lam et al, 1982; Srivatanakul
tic effect of tobacco smoking and heavy alcohol et al, 1991).
consumption has been also reported (Kuper et al, Aflatoxin and its oxidative metabolites can
2000b; Marrero et al, 2005). be measured in urine and serum and can be use-
ful as markers of dietary intake and biologically
Diet effective dose. Epidemiological studies have indi-
A positive association of aflatoxins with the devel- cated a positive association of aflatoxin biomark-
opment of HCC has been supported by a number ers with liver cancer. A meta-​analysis of studies
of studies, as has a protective effect of vegetables, undertaken in populations of aflatoxin-​endemic
fish and, possibly, high levels of the antioxidants areas estimated a pooled relative risk (RR) of
retinol and selenium. A consistent inverse asso- 4.75 (95% confidence interval [CI] 2.78–​8.11) for
ciation with coffee intake has been reported in HCC risk, associated with biomarkers of aflatoxin
a number of studies including European and exposure after adjustment for hepatitis B surface
US multicenter cohort studies, and this asso- antigen (BsAg) positivity (Liu et al, 2012).
ciation has been considered “probable” in the Molecular evidence for a role of aflatoxin in
2015 report of the World Cancer Research Fund/​ the development of HCC has been convincingly
American Institute for Cancer Research (WCRF/​ presented (Montesano et al, 1997; Sun et al, 1999).
AICR, 2015). In the 1991 IARC Monographs for Consumption of aflatoxin B1 induces the G:C to
the evaluation of coffee as a carcinogen, this bev- T:A transversion at codon 249 of p53 observed in
erage had been classified as possibly carcinogenic liver cancer cells from cases in southern China
to humans (Group 2B) (IARC, 1991). This classi- and sub-​Saharan Africa. This p53 mutation is
fication changed in the most recent report of the rarely observed in HCC cases in the United
Working Group for the update of the 1991 IARC States or Europe, where aflatoxin exposure is low.
Monographs (IARC, 2017), according to which In vitro experiments involving human cells show
“drinking coffee was judged as not classifiable the same mutational effect of aflatoxin B1, with
as to its carcinogenicity to humans (Group  3)” the N7-​guanine adduct being primarily responsi-
(https://​www.iarc.fr/​en/​media-​centre/​pr/​2016/​ ble. This one-​base-​pair change results in a serine
pdfs/​pr244_​E.pdf). In the respective publication, substitution of the arginine amino acid, which
the Working Group also notes that “reduced is important for the proper folding of the DNA-​
risks were seen for cancers of the liver” (Loomis binding domain of the p53 protein.
et al, 2016). Excretion of urinary aflatoxins and HBsAg
interacted strongly positively with respect to risk
Aflatoxin Contamination of HCC (Ross et al, 1992; Qian et al, 1994; Wang
Aflatoxins are produced primarily by two et  al, 1996; Lunn et  al, 1997). In a 2012 meta-​
Aspergillus species of mold that can contaminate analysis (Liu et  al, 2012), the estimated relative
dietary staples including corn, peanuts, and rice risks for aflatoxin exposure (any) were 2.39 (95%
stored in hot, humid environments. Aflatoxins CI 1.50–​3.82) in HBV+ individuals, 5.91 (95% CI
have been classified as established carcinogens 3.66–​9.55) in HBV-​individuals, and 54.1 (95% CI
(IARC, 1993, 2012f) and are believed to be an 21.3–​137.7) for the combined effect of aflatoxin
important contributing cause of HCC in some and HBV infection. A similar strong interaction
low-​income countries (Montesano et  al, 1997; between aflatoxin B1 and HBV chronic infection
Williams et  al, 2004). The level and prevalence also was observed in cohort studies in Taiwan.
284

284 Textbook of Cancer Epidemiology

Early exposure to aflatoxin and HBV may explain vitamin D supplements, vitamin C, and low fat
the younger age peak in HCC incidence observed diet) are limited and with inconsistent results.
in some parts of Africa and China. Epidemiological studies of vegetable intake
Data from an ecological study (Chen et  al, support an inverse association with risk of liver
1996) suggests that a reduction of aflatoxin expo- cancer/​HCC, including cohort studies in Asia
sure from 100% to 23% would result in an esti- (Yu et al, 1995; Sauvaget et al, 2003; Kurozawa
mated population attributable benefit of 65% et al, 2004; Pham et al, 2006; Kurahashi et al,
reduction in liver cancer incidence overall and 2009; Zhang et al, 2013), case-​control studies
17% among those without HBV infection. The in Europe (La Vecchia et al, 1988; Negri et al,
2012 meta-​analysis (Liu et al, 2012) also estimated
1991; Braga et al, 1997; Talamini et al, 2006) and
that, by reducing aflatoxin below detectable lim- Serbia (Kanazir et al, 2010), and a cohort study
its, an HCC incidence reduction by 14%–​19% in Europe (Bamia et al, 2015a) based on the mul-
in the general population, and by 10%–​29% in ticenter European Prospective Investigation into
HBV+ populations could be achieved. Indeed, Cancer and nutrition (EPIC).
changes in food policy and economic develop- Higher consumption of fish has been associ-
ment may have led to a dramatic decline in the ated with a modest decreased risk of liver cancer
levels of aflatoxin exposure in some endemic in some studies, although evidence is limited. In
regions, such as Taiwan and selected Chinese a meta-​analysis of four cohort studies in Europe
regions. (Fedirko et  al, 2013a), Japan (Iso et  al, 2007;
The exact nature of the apparent interaction Sawada et al, 2012), and the United States (Daniel
between aflatoxin B1 and HBV is not clear. It et al, 2011), a summary RR of 0.94 (95% CI 0.89–​
is possible that chronic HBV infection and the 0.99) per 20 g/​day fish intake was estimated,
associated liver damage may lead to the selec- although there was heterogeneity across studies.
tive proliferation of aflatoxin-​mutated hepato- Notably three of the four studies included did
cytes, via modulation of relevant metabolizing or not show statistical significant associations when
detoxifying enzymes, increased hepatocyte turn- comparing the highest to the lowest category of
over, increased oxidative stress, or interference fish consumption (WCRF/​AICR, 2015).
with DNA repair pathways by the HBV X gene The association of tea consumption and liver
(Kew, 2003, 2006, 2013). A  higher frequency of cancer risk has been investigated in studies con-
p53 mutation in HBV + HCC cases, as well as ducted almost exclusively in Asia (as green tea)
data from experimental studies in woodchucks with inconclusive results. A  meta-​ analysis of
and HBV-​ transgenic mice, also support an four cohort studies published until 2013 (all in
interaction between chronic HBV infection and Japanese populations) found a summary rela-
aflatoxin exposure in the development of liver tive risk per cup/​day of 0.99 (95% CI 0.94–​1.03)
cancer (Montesano et al, 1997). A meta-​analysis (WCRF/​ AICR, 2014). A  meta-​ analysis of five
of 48 studies, however, revealed no significant case-​control and six prospective cohort studies
modifying effect of HBV infection on the associ- found a summary relative risk comparing highest
ation between aflatoxin exposure and frequency versus the lowest consumption of 0.79 (95% CI
of the p53 249ser mutation in HCC cases (Stern 0.68–​0.93) for green tea and 0.77 for any type of
et al, 2001). tea (95% CI 0.57–​1.03) (Sing et al, 2011). A meta-​
analysis of prospective studies in Asian popula-
Food Groups, Macronutrients, Micronutrients tions suggested a significant association with
The association between diet and risk of liver liver cancer of highest versus lowest green tea
cancer was evaluated by the WCRF (1997, consumptions, but with no linear trend (Huang
2007)  and updated in the 2015 report (WCRF/​ et al, 2015). In EPIC, an inverse association of tea
AIRC 2015). The strongest evidence is for alco- (mostly other than green tea) with HCC risk was
hol, for which the positive association with liver observed (Bamia et al, 2015b).
cancer risk is considered strong/​convincing, and With respect to micronutrient intake, an
coffee, for which evidence of an inverse associ- inverse relationship between dietary and serum
ation is considered strong/​probable. Studies of retinol levels and HCC risk has been observed in
other foods/​ foods groups and nutrients (e.g., studies of men in Taiwan (Yu et  al, 1995b) and
cereals (grains) and their products, nonstarchy China (Yuan et al, 2006), most apparent among
vegetables, peanuts (groundnuts), salted fish, those with chronic HBV infection. In Italy, low
meat and poultry, glycemic index, calcium and serum retinol was strongly associated with HCC
285

Cancer of the Liver and Biliary Tract 285

risk in patients with cirrhosis (Clemente et  al, underlying the inverse association between cof-
2002). Evidence is emerging to support a protec- fee and liver cancer.
tive effect of high selenium intake on HCC risk
(McGlynn & London, 2005). Nested case-​control Alcohol
studies in Taiwan (Yu et  al, 1999b) and China Results of several epidemiological studies
(Sakoda et  al, 2005)  reported an inverse asso- strongly support the hypothesis that heavy alco-
ciation between low selenium levels and HCC hol drinking and alcoholic cirrhosis increase the
incidence. However, the two studies differed risk of HCC (Austin, 1991; Adami et  al, 1992;
with respect to the subgroups in which the pro- Kuper et  al, 2000b, 2001a; Donato et  al, 2002;
tective effect was most apparent: cigarette smok- Yuan et al, 2004; Morgan et al, 2004; Schutze et al,
ers and those with low levels of serum retinol or 2011; Simazu et al, 2012; Persson et al, 2013). In
carotenoids, in Taiwan; women and nondrink- 1988, IARC classified alcohol as a human car-
ers, in China. In a meta-​analysis of randomized cinogen and specifically implicated alcohol con-
controlled trials in China (Vinceti et  al, 2014), sumption in the etiology of liver cancer (IARC,
selenium supplementation was associated with 1988); this classification was updated in 2012
a decreased risk for liver cancer, although the (IARC, 2012e). In the 2015 WCRF/​AICR report,
number of events was small. the evidence regarding the positive association
of alcohol consumption and liver cancer risk
Coffee was upgraded from “probable” (WCRF, 2015) to
The association between coffee consumption and “convincing.”
liver cancer risk has been investigated in several Although alcohol metabolism results in the
case-​control and cohort studies with an inverse generation of acetaldehyde and reactive oxygen
association observed in most studies (Inoue species, which could directly damage hepatocytes
et al, 2005, Ohishi et al, 2008; Johnson et al, 2011; (Morgan et  al, 2004), alcohol intake might be a
Hu et al, 2008; Lai et al, 2013; Bamia et al, 2015b; liver carcinogen mainly, or, only by being causally
Petrick et al, 2015; Setiawan et al, 2015). The involved in the development of cirrhosis (Adami
2015 WCRF report cited the evidence regarding et al, 1992). A nationwide cohort study in Sweden
the association of coffee with liver cancer for the found a relative risk of 22.4 for the incidence of
first time as “strong/​ probable” (WCRF/​ AICR, HCC among patients hospitalized for alcohol-
2015). In their meta-​analysis of six studies, the ism and cirrhosis compared to the general pop-
summary RR per 1 cup/​day was 0.86 (95% CI ulation, but only a 2.4-​fold increased risk among
0.81–​0.90). Similarly, in the 2016 report of the patients hospitalized for alcoholism without cir-
IARC Working Group for the evaluation of coffee rhosis (Kuper et  al, 2001a). Thus, alcohol abuse
as a carcinogen, the inverse associations of liver leading to cirrhosis may be an important cause
cancer with coffee drinking, evident in many of HCC (Vall Mayans et al, 1990), whereas light
cohort and case-​control studies in Asia, Europe, and moderate drinking may have no major influ-
and North America, are highlighted (Loomis ence on HCC risk (Trichopoulos et al, 1987; La
et al, 2016). A recent meta-​analysis of prospec- Vecchia et al, 1988). In a 2014 meta-​analysis of 19
tive cohort studies estimated a 15% lower risk of prospective studies, a relative risk of 1.16 (95% CI
liver cancer (Bravi et al, 2016) for an increment of 1.01–​1.34) was reported for alcohol drinkers of >
1 cup of coffee per day. 3 drinks per day and 1.22 (95% CI 1.10–​1.35), for
It has been suggested that the inverse associ- those consuming >6 drinks per day (Turati et al,
ation of coffee intake and liver cancer may be at 2014). In addition, a more-​than-​additive effect of
least partly explained by its inverse association heavy alcohol consumption and hepatitis virus
with chronic liver disease, cirrhosis, or diabetes infection has been reported (Donato et al, 2002;
mellitus (Corrao et al, 2001; Huxley et al, 2009; Hassan et al, 2002; Yuan et al, 2004).
Saab et al, 2014). Coffee contains several hun-
dred chemical compounds including caffeine, Reproductive Factors
diterpenes, chlorogenic acids, and flavonoids, Evidence regarding the association of repro-
which have been shown to induce the endoge- ductive factors with risk of primary liver cancer
nous defense system and DNA repair capacity is limited. Stanford et al (1992) postulated that
and reduce the expression of genes involved in endogenous hormones, primarily estrogens, may
inflammation (Bohn et al, 2014). More research play a role in the development of HCC among
is needed in order to understand the mechanisms women. Studies that investigated this hypothesis,
286

286 Textbook of Cancer Epidemiology

by means of the association of parity with liver Caperell-​ Grant, 2011). Notwithstanding the
cancer risk, showed inconsistent results. In ear- progress in this research area, the precise role
lier case-​ control studies, increased parity was of sex-​specific hormones as well as the related
positively (La Vecchia et al, 1992; Stanford et receptors and pathways leading to liver cancer
al, 1992; Tzonou et al, 1992) or not associated are still poorly understood. Three cohort studies
(Hsing et al, 1992; Lambe et al, 1993; Mucci et of Asian men have investigated the relationship
al, 2001) with liver cancer risk. In contrast, two between baseline serum testosterone level and
studies in Taiwan found that the risk of HCC (Yu HCC risk, with mixed results. In two studies, tes-
et al, 2003) or liver cancer mortality (Wu et al, tosterone levels in the highest tertile were associ-
2011) decreased with increasing number of full-​ ated with a significant two-​to fourfold increased
term pregnancies. risk of HCC (Yu & Chen, 1993; Yu et al, 2000).
Only four epidemiologic studies have inves- A  study in Shanghai reported a nonsignificant
tigated the association between age at first birth 50% increased risk of HCC in the highest tertile
and liver cancer; three of those reported no asso- of testosterone levels among HBV carriers (Yuan
ciation (La Vecchia et al, 1992; La Vecchia et al, et  al, 1995). In a case-​control study in Greece
1993; Lambe et al, 1993), whereas a cohort study (Mucci et  al, 2001), levels of testosterone and
in Taiwan (Wu et al, 2011) reported an increased sex hormone binding protein among women
risk of liver cancer with increasing age at first were found to be slightly higher for HCC cases
birth. Regarding age at menarche/​menopause, a than for controls; however, analysis of sex steroid
case-​control study conducted in Greece found levels among men suggested that the differences
that the risk of HCC was higher for women with observed were more likely due to liver damage
lower age at menarche or higher age at men- per se (Kuper et al, 2001b). Nonetheless, the con-
opause (Mucci et  al, 2001); the HCC cases also sistently higher incidence of HCC in men than in
had somewhat higher levels of estradiol than women, particularly in low-​and middle-​income
did the controls. An increased risk of HCC with countries where liver cancer rates are high and
earlier age at menarche, but not with older age at chronic HBV infection is endemic, suggests a
menopause or increasing parity, was reported for role for testosterone in hepatocarcinogenesis,
women in Taiwan (Yu et al, 2003). which is also supported by evidence from ani-
In a pooled analysis of data from 11 cohorts mal studies (Yu & Chen, 1993). Of interest, HBV-​
participating in the Liver Cancer Pooling Project, carrier Taiwanese men with shorter CAG repeats
there were no associations of HCC risk with par- in androgen receptor (AR), indicating increased
ity, age at first birth, age at natural menopause, or receptor transactivation, had an increased risk of
duration of fertility (McGlynn et al, 2015). Bilateral developing HCC, which was further elevated for
oophorectomy was, however, associated with a those men with the highest level of testosterone
significantly increased risk of HCC (RR 2.67, 95% (Yu et al, 2000). In addition, limited case-​report
1.22–​5.85) with no evidence of interaction with data point to an effect of androgenic-​anabolic
duration of exposure to endogenous hormones or steroids on liver cancer risk (IARC, 1987;
to menopausal hormone therapy use. Trichopoulos, 1992).
Exogenous sex hormones also may have an
Hormones effect on the occurrence of HCC. Use of oral con-
Overall, the evidence is suggestive of a role for traceptives has been associated with increased
steroid hormones in the incidence of liver can- risk of liver adenomas and may have an effect
cer, which may be further modified by HBV sta- on the occurrence of HCC (IARC, 2012a). Case-​
tus in a complex manner that differs for men and control studies, primarily conducted in high-​
women (Yeh & Chen, 2010; Montella et al, 2015). income countries, investigated oral contraceptive
Epidemiological and animal studies suggest that use as a risk factor and found a two-​to three-
the higher incidence of liver cancer among men, fold increased risk of HCC among oral contra-
as compared to women might be due to the stim- ceptive users (Yu et  al, 1991; The Collaborative
ulatory effects of androgens and the protective MILTS Project Team, 1997). In contrast, among
effects of estrogen (Nakatani et al, 2001; Shi et al, women from HBV-​endemic areas, users showed
2014). It has also been argued that the protec- a nonsignificant decrease in this risk (WHO
tion conveyed by estrogens against chemically Collaborative Study of Neoplasia and Steroid
induced liver tumors may be mediated by pro- Contraceptives, 1989b; Yu et  al, 2003). Other
lactin through liver prolactin receptors (Bigsby & reports from high-​income countries also have
287

Cancer of the Liver and Biliary Tract 287

not supported an effect of oral contraceptive use Physical Activity

on liver cancer in general (Colditz, 1994; Bosch Epidemiological studies on the association of
et  al, 2004). A  meta-​analysis of 12 case-​control physical activity and liver cancer risk gener-
studies (Maheshwari et  al, 2007), estimated ally report an inverse association. A  cohort
a nonsignificant pooled RR of liver cancer in study in Japanese men and women (Inoue et al,
relation to oral contraceptives of 1.57 (95% CI 2008) showed a nonsignificant decreased risk of
0.96–​2.54) with some evidence of dose (dura- liver cancer (RR 0.54, 95% CI 0.23–​1.29) when
tion)-​response relationship. comparing high and low levels of total physical
activity. Yum et al (2008) reported a risk of 0.88
Anthropometric Measures (95% CI 0.81–​0.95) when comparing moderate-​
The association of body mass index (BMI) with high with low levels of leisure-​ time activity.
liver cancer incidence/​mortality has been exten- Duration of walking (2007) was also associated
sively studied. In a report of 15 studies published with a decreased risk of liver cancer mortality
from 2003 to 2013 in Europe, the United States, in both men and women for the highest versus
and Asia (WCRF/​AICR, 2015), a positive associ- lowest levels of walking per day. In a 2013 cohort
ation was observed comparing the highest to the study (Behrens et  al, 2013)  a statistically signif-
lowest category of BMI. Relative risks were statis- icant decreased risk of liver cancer was associ-
tically significant in eight studies (in two studies ated with vigorous physical activity (>5 times per
only among men), with reported estimates rang- week). Although evidence from existing studies
ing from 1.39 among women (Jee et al, 2008) to is rather consistent in that higher levels of activ-
4.57 for both men and women (Ohishi et  al, ity are accompanied with decreased liver cancer
2008). A  meta-​analysis of the indicated studies risk, the studies are few and have varying defini-
suggested a significant positive dose-​ response tions and assessments of physical activity. More
relationship between BMI and liver cancer, with studies are warranted for any firm conclusion
an estimated RR of 1.30 (95% CI 1.16–​1.46) per regarding the association between physical activ-
5-​unit increase in BMI (WCRF 2015). The asso- ity and liver cancer to be drawn.
ciations were stronger for liver cancer incidence
rather than mortality, and weaker associations for Infections
Asian (RR 1.18, 95% CI 1.04–​1.34) as compared There is convincing evidence for a causal role of
to European (RR 1.59, 95% CI 1.35–​1.87) stud- chronic HBV and HCV infections in the devel-
ies were observed. Results from pooled analyses opment of HCC, and viral infection is involved
(Batty et al, 2009; Whitlock et al, 2009; Parr et al, in the majority of liver cancers. In addition,
2010; Borena et al, 2012), as well as from meta-​ infection with the liver flukes Opisthorchis viver-
analyses (Larsson et  al, 2007; Chen et  al, 2012; rini and Clonorchis sinensis, endemic in cer-
Rui et al, 2012; Wang et al, 2012) also report pos- tain parts of Asia, is associated with a risk of
itive associations for BMI and liver cancer risk, cholangiocarcinoma.
although not always statistically significant. In
the 2015 WCRF/​AICR report, the existing evi- Hepatitis B Virus
dence regarding the association of body fatness Hepatitis B virus, a DNA virus, requires a reverse
(as assessed through BMI) with liver cancer risk transcription step, similar to retroviruses, in
was judged as “strong/​convincing.” order to replicate. When replicating, HBV does
The effect of obesity on liver cancer incidence not directly copy its DNA genome, as all other
may be mediated, at least in part, through the DNA viruses do. Instead, a new DNA genome
strong relationship of overweight/obese with risk is made via reverse transcription of a full-​length
of type 2 diabetes mellitus, which is, itself, associ- RNA copy of the viral DNA. The virus prefer-
ated with increased risk of HCC carcinoma (Yang entially infects hepatocytes and can establish a
et al, 2011). Body fatness is also related to many chronic, persistent infection within the liver. The
circulating hormones, such as insulin, growth chronic carrier state of HBV infection is char-
factors, and estrogens, which may be involved in acterized by the serologic detection of HBsAg.
carcinogenesis (Hursting et al, 2003). Obesity is The development of chronic infection is directly
a risk factor for nonalcoholic fatty liver disease/​ related to the age at first infection with HBV.
steatohepatitis, both of which are associated with Thus, 80% to 90% of babies born to infected
HCC risk (Calle & Kaaks, 2004; Alzahrani & mothers become chronic carriers of HBV com-
Hebbard, 2014). pared to fewer than 10% of exposed adolescents
28

288 Textbook of Cancer Epidemiology

and adults. With regard to the expression of the disease is quite frequent in the younger age
acute, symptomatic disease upon initial infec- groups (Bosch et al, 2004).
tion, the reverse situation is true. Both phenom- On a molecular level, HBV DNA has been
ena are the result of the host’s immune response detected in the liver tissue specimens from over
to the virus, with neonates and children having a 90% of HBsAg-​positive HCC cases as well as in
weaker immune response that leads to minimal some cases with no HBV serum markers (Tabor,
symptoms and chronic infection and adults hav- 1998; Bréchot et al, 2010). These HBV sequences
ing a stronger response that leads to overt disease appear to be clonally integrated in the tumor
and viral clearance. cells. Although the woodchuck hepatitis virus
The global distribution of chronic HBV causes liver cancer through insertional mutagen-
infection varies. Low levels of endemicity (i.e., esis at the myc gene locus (Tennant et al, 2004),
<2% HBsAg prevalence) can be found in North no targeted integration site for HBV-​associated
America, Central America, western and north- human HCC had been identified (Tabor, 1998).
ern Europe, and Australia. Areas with a more Some studies have shown that HBV integration
moderate prevalence of HBV carriers (2% to 7%) may occur more frequently in or adjacent to host
include eastern and southern Europe, the Middle genes related to cell growth (Bréchot et al, 2010;
East, Japan, southern Asia, and parts of South Momosaki et  al, 2003; Murakami et  al, 2005).
America. China, Southeast Asia, and sub-​Saharan Moreover, insertion of multiple copies of the
Africa have the highest levels of chronic HBV HBV genome can generate various random chro-
infection (≥8%) (IARC, 2012b; Ott et  al, 2012). mosomal abnormalities (Bréchot et al, 2010).
As would be expected, transmission around the In addition to the cys-​ acting potential of
time of birth and during childhood are impor- HBV integration, the X and preS2/​S regions of
tant routes of infection with regard to HBV ende- the viral genome appear able to transactivate
micity and are major sources of infection in areas cellular oncogenes (Bréchot et al, 2000; Bréchot
where the prevalence of HBsAg is high. A strong et al, 2010). The HBV X gene encodes a viral pro-
geographic correlation has been demonstrated tein that is important viral replication and that
between the prevalence of chronic HBV infection can transactivate the transcription of viral and
and the incidence of HCC (Maupas and Melnick, host genes in vitro. The preS2/​S gene encodes
1981), with populations in the United States and for the middle protein of the HBV surface anti-
Europe having low HCC rates as well as a low gen. Sequences from both viral genes are sys-
prevalence of HBV carriers. tematically retained in HBV inserts in tumor
The most convincing evidence for a causal role cells (Bréchot et  al, 2000; Bréchot et  al, 2010).
of chronic HBV infection in HCC development The deletions occurring in the integrated subvi-
is in the overwhelmingly consistent results from ral DNA likely are important for transactivation,
the extensive number of epidemiological studies, since the native preS2/​S protein has no detecta-
representing over four decades of epidemiologic ble transactivation potential and the truncated
research in more than 25 countries (IARC, 1994a, X protein is a stronger transactivator than the
2012b). Most studies have been case-​control in wild-​type form. The X protein has been shown to
design and have yielded RR estimates from 1.5 to inhibit the function of p53 by binding this tumor
87 for the association of HCC with HBsAg posi- suppressor in HCC tissue and in transgenic mice
tivity. The RR estimates are consistently higher in (Bréchot et  al, 2010). Furthermore, this viral
cohort studies (IARC, 1994a, 2012b; Mori et al, protein may play a role in the hepatocarcino-
2000; Trichopoulos et  al, 2011), exceeding 100-​ genic effect of HBV via its interaction with host
fold in a study from Taiwan (Beasley, 1988). In genes involved in DNA repair, cell-​cycle regula-
addition, early establishment of chronic HBV tion, and apoptosis (Bréchot et al, 2010; Neuvet,
infection during childhood appears to increase et al, 2010).
the future risk of HCC (Muñoz et al, 1989; Kuper The potential etiologic relevance of so-​called
et  al, 2000a). This hypothesis is supported by occult or silent HBV infection and HCC risk has
the differential risk observed for the younger been the focus of several investigations (Bréchot
age groups in high-​and low-​risk populations. et al, 2010; Tabor, 2002). Low levels of HBV DNA
In populations from low-​and intermediate-​risk have been found in both HBsAg-​negative HCC
areas, liver cancer is rarely observed in persons cases and noncases including blood donors,
less than 40 years old, in contrast to populations whose only marker of infection is antibody to
with a high incidence of liver cancer, in which the HBV core protein. Existing data support that
289

Cancer of the Liver and Biliary Tract 289

occult HBV infection does play a role in HCC (IARC, 2012b; Zhou et al, 2014; Lee et al, 2015).
(Bréchot et al, 2010; Tabor, 2002; Pollicino et al, Most of the studies reported a positive associa-
2004), most notably in HCV-​associated disease tion, indicating a 2-​to 10-​fold increased risk.
(Bréchot et al, 2010; Tabor, 2002; Pollicino et al, Two cohort studies provide additional evidence
2004; Momosaki et  al, 2005). The presence of for a role of HBV and intrahepatic cholangiocar-
HBsAg-​negative HBV infection in HCC cases cinoma, with relative risks of 4.80 among preg-
may be related to reduced virus replication asso- nant women in Taiwan (Fwu et  al, 2011)  and
ciated with integration of the virus into the host 8.56 among blood donors in Japan (Tanaka et al,
genome or to mutations to the virus genome itself. 2010). In a meta-​analysis of 16 case-​control stud-
Of interest, a specific double mutation within ies and the two cohort studies, an overall relative
the HBV core promoter and X genes has been risk of 2.66 was obtained for cholangiocarci-
reported in HBsAg-​ negative HCV-​ associated noma; the association with HBV infection was
HCC cases (Momosaki et  al, 2005). This same slightly stronger for intrahepatic (RR 3.42) than
mutation has been found in HBsAg-​ positive for extrahepatic cholangiocarcinoma (RR 1.46)
tumors, and its detection in stored blood samples (Li et al, 2012).
has been shown to predict the incidence of HCC, In 1994, IARC concluded that there was suf-
up to 5  years before diagnosis, in a prospective ficient evidence to classify HBV as a human car-
cohort study in China (Kuang et al, 2004). cinogen (IARC, 1994a). An IARC Monograph
Other markers of HBV infection also may specified that chronic HBV infection causes
prove to be useful in identifying carriers at high HCC; the observation of “positive associations”
risk of progression to HCC. High serum levels of between HBV and cholangiocarcinoma was
HBV DNA are associated with a five-​to sevenfold noted (IARC, 2012b). Globally, an estimated 54%
increased incidence of HCC in cohort studies in of all liver cancer could be attributed to HBV
Taiwan, suggesting that increased virus replica- infection (Parkin, 2006), and 77% for all can-
tion might play a role in HBV-​induced hepato- cers combined (de Martel at al, 2015). Most of
carcinogenesis (Chan et al, 2008; Chen et al, 2006; the HBV-​associated liver cancer occurs in low-​
Yang et al, 2002; Yu et al, 2005). Moreover, some and middle-​income countries (attributable frac-
studies have shown that HBV carriers infected tion, 59%); those areas with the greatest burden
with genotype C may be at increased risk of HCC include Melanesia (77%), western and central
(Chan et  al, 2008; Fujie et  al, 2001; Yang et  al, Africa (70%), and China (70%). In high-​income
2008; Yu et al, 2005). Of interest, specific muta- countries, only 23% of liver cancer is likely due
tions within the HBV preS gene and the viral core to HBV infection, with as little as 9% in North
promoter have been identified in conjunction America and western and northern Europe
with the development of HCC among carriers (Parkin, 2006).
with genotype C infection (Liu et al, 2009; Kuang The existence of a safe, effective vaccine
et al, 2004; Yang et al, 2008). against HBV infection provides the opportu-
In addition to having a direct mutagenic nity to prevent a substantial proportion of HCC.
effect, chronic HBV infection likely also pro- Childhood vaccination programs initiated in
motes tumorogenesis through the continual HBV-​endemic areas have led to distinct reduc-
turnover of hepatocytes resulting from con- tions in the prevalence of HBV carriers (Chen,
stant inflammation and damage to the liver. 2009; IARC, 2012b). According to the WHO,
In fact, HCC frequently arises within cirrhotic programs aimed at the universal immunization
livers (Tabor, 1998; Bréchot et  al, 2000). The of infants had been established in 184 countries
host immune response to the virus can lead to by the end of 2014 (WHO, 2015). Of singular
the generation of cytokines and reactive oxy- importance was the initial report of a signifi-
gen species and the induction of genetic muta- cant decline in HCC incidence among children
tions (Bréchot et  al, 2010; Neuvet, et  al, 2010). in Taiwan between 1981 and 1994 following the
Thus, due to its potential ability to both initiate introduction of a nationwide HBV immunization
and promote HCC development, HBV has been program (Chang et al, 1997). Similar reductions
viewed as a complete carcinogen (Trichopoulos in HCC rates in children and adults, subse-
et al, 1987; Tabor, 1994). quent to HBV immunization efforts, have been
The relationship between chronic HBV infec- observed in other endemic populations (Gwack
tion and risk of cholangiocarcinoma has been et al, 2011; Locarnini et al, 2015; Sun et al, 2013).
investigated in multiple case-​ control studies The Qidong Hepatitis B Intervention Study
290

290 Textbook of Cancer Epidemiology

randomized controlled trial demonstrated an McMahon, 2000). Pockets of relatively high HCV
efficacy of 84% for neonatal immunization for the seroprevalence also have been reported in areas of
prevention of primary liver cancer, after 30 years Japan (Kiyosawa et al, 1994; Tanaka et al, 1997).
of follow-​up (Qu et  al, 2014). These data dem- Percutaneous transmission of the virus appears
onstrate the effectiveness of HBV vaccination to be the major route of infection, with expo-
in preventing the occurrence of HCC. Antiviral sure to contaminated blood products (prior to
treatment also has been reported to reduce the screening), intravenous drug use (in the United
risk of HCC in HBV carriers (Dienstag, 2008). States and other Western populations), and use
of nondisposable needles/​ syringes (previously,
Hepatitis C Virus in Japan, other Southeast Asia populations, some
Hepatitis C virus, an RNA virus in the Flaviviridae European countries, Egypt) representing impor-
family, was first identified in 1988 (Choo et  al, tant modes of transmission (Alter, 2007).
1989). With no reverse transcriptase, HCV does Epidemiologic studies consistently support
not integrate into the host’s DNA (Tabor, 1998). a strong association between HCV and HCC.
The current multiantigen tests have sufficient sen- Relative risks from 2.5 to greater than 80 have
sitivity and specificity to reliably detect antibody been reported in cohort studies (IARC, 1994a,
to HCV (anti-​HCV). The presence of viremia is 2012b; Trichopoulos et al, 2011), most of which
determined using reverse transcriptase-​PCR for were performed in Asia. Similarly strong asso-
the detection of HCV RNA. ciations have been observed in the more than
Characterization of the natural history of 30 case-​control studies conducted in a variety of
HCV has been challenging, primarily due to the populations (IARC, 1994a, 2012b; Donato et al,
largely asymptomatic nature of the infection. 1998). In addition, some studies have reported
Based for the most part on studies of patients that HCV genotype 1b may contribute dispropor-
identified by the presence of clinical hepatitis or tionately to the virus hepatocarcinogenic effect
by transfusion exposure, as well as on population-​ (Tanaka et al, 1996; Tagger et al, 1999). A meta-​
based cross-​sectional studies of HCV RNA prev- analysis of 21 studies estimating a summary RR
alence, about 70% to 85% of those infected with of 1.78 for the association of HCV infection with
HCV will develop a chronic, persistent infec- genotype 1, compared to other genotypes, and
tion (Alter et  al, 1999; Bellentani et  al, 1999; the risk of HCC (Raimondi et al, 2009).
Hoofnagle, 2002); persistence of infection may The mechanisms by which HCV may cause
be less frequent in women and those who are liver cancer are not well understood. Since most
younger or who experience symptomatic acute HCV-​associated HCC occurs in the presence
infection (Alter et al, 1999; Bellentani et al, 1999; of cirrhosis, HCV infection may lead to cancer
Hoofnagle, 2002). through the indirect mechanism of immune-​
Among persons with chronic HCV infection, mediated damage and subsequent liver cell turn-
50% to 80% progress to chronic hepatitis (Alter over, resulting in the accumulation of genetic
et  al, 1992; Takahashi et  al, 1993; Hoofnagle, mutations (Tabor, 1998; Lin et  al, 2015). As a
2002). The rate of subsequent progression to nonintegrating virus, HCV would be unlikely to
cirrhosis appears to be quite variable and to be have a direct role in HCC initiation. However,
related to characteristics of the infected individ- several HCV proteins have been shown to exhibit
ual, including age at infection and underlying oncogenic properties in vitro and in transgenic
disease status (Poynard et al, 1997). The 20-​year mice models (Lin et  al, 2015). In particular,
incidence of cirrhosis was estimated to range the core protein may affect cell-​cycle control
from 4% in blood donors and 7% in community-​ through repression of p53 (Ray et  al, 1997; Lin
based populations to 22% in liver disease patients et al, 2015) and interference with tumor necro-
and 24% in persons with transfusion-​ related sis factor alpha–​inducted apoptosis (Block et al,
infection (Freeman et al, 2001). After the devel- 2003); the HCV core protein also inactivates
opment of cirrhosis, 3%–​5% per year progress to the p73 and RB1 tumor suppressors (Lin et  al,
malignancy (Fattovich et al, 2004). 2015). In addition, the nonstructural virus pro-
In most populations throughout the world, the teins NS3 and NS5A may play a role in HCV-​
prevalence of anti-​HCV is <2.5% (Mohd Hanafiah inducted hepatocarcinogenesis via interaction
et al, 2013); however, infection is high (>10%) in with p53 and intracellular signaling pathways
Egypt and appears to be linked to mass schis- (Block et al, 2003; Lin et al, 2015). HCV infec-
tosomiasis treatment campaigns (Lavanchy & tion has been associated with increased levels
291

Cancer of the Liver and Biliary Tract 291

of TGF-​β, which could induce transformation cholangiocarcinoma (Li et al, 2015). In a cohort of

of hepatocytes (Nelson et  al, 1997; IARC et  al, US veterans, the hazard ratio was 2.6 for the asso-
2012b). Last, the induction of hepatic steatosis, ciation between HCV and intrahepatic cholan-
oxidative stress, and insulin resistance associ- giocarcinoma risk over 16 years of follow-​up; no
ated with chronic HCV infection may contribute significant association was observed for extrahe-
to the development of HCC (IARC et al, 2012b; patic cholangiocarcinoma (El-​Serag et al, 2009).
Lin et al, 2015). HCV was designated as carcinogenic to
HCV-​associated HCC appears to affect older humans by IARC in 1994 (IARC, 1994a). In the
patients and to follow more severe liver disease 100B Monograph update, a causal relationship
than does HBV-​associated HCC (Shiratori et al, was indicated for HCV with HCC and a “positive
1995). Thus, a prolonged period of hepatocellu- association” with cholangiocarcinoma (IARC,
lar damage progressing from chronic hepatitis to 2012b). Just under one-​third of the liver cancers
cirrhosis to HCC may be required for the devel- occurring in the world have been estimated to be
opment of HCV-​related malignancy (Kiyosawa attributable to HCV infection, based on 2002 data
et  al, 1990). There is considerable evidence for (Parkin, 2006). The impact of HCV-​associated
a strong interaction between heavy alcohol con- HCC appears to be greatest in Asia, where the
sumption and HCV infection, more so than for attributable fraction is 30% in Japan, 36% in
HBV infection, on the occurrence of liver dam- China, and 41% in Southeast Asia, as well as in
age and the progression of liver disease (Poynard northern (61%) and central Africa (55%) (Parkin,
et  al, 1997; Tagger et  al, 1999; Mori et  al, 2000; 2006). Overall, the proportion of liver cancer due
Donato et  al, 2002). Several studies have found to HCV is greater in low-​and middle-​income
that heavy alcohol consumption may mark- (33%) than in high-​income (20%) countries.
edly enhance the risk of HCC among those with Although at present there is no vaccine to
HCV-​ associated cirrhosis (Chen et  al, 1997; prevent HCV infection, evidence exists to sup-
Okuda, 1997). port significant reductions in new infections
HBV and HCV infections appear to interact through the interruption of parenteral transmis-
to increase the development of HCC (Kaklamani sion of the virus. The testing of blood donations
et al, 1991; Tagger et al, 1999; Kew, 2003, 2006). for anti-​HCV has led to substantial decreases in
In a meta-​analysis of 32 case-​control studies con- the number of post-​transfusion hepatitis C cases
ducted in 1993–​1997, the synergism was super- in Taiwan (Wang et  al, 1995), Japan (Japanese
additive, with summary RRs of 165 for dual Red Cross Non A, Non B-​Hepatitis Research
infection, 22.5 for infection with HBV alone, Group, 1991), and the United States (Tobler &
and 17.3 for infection with HCV alone (Donato Busch, 1997). The use of disposable needles and
et  al, 1998). A  meta-​ analysis of case-​ control syringes and other changes in medical proce-
studies conducted in China between 1990 and dures also have helped to reduce HCV infec-
2004 obtained an RR of 35.7 for HBV/​HCV dual tion in Japan (Okuda, 1997) and Taiwan (Sung,
infection, 15.6 for HBV infection alone, and 1997). Since parenteral exposure is a major route
8.1 for HCV infection alone (Shi et  al, 2005). of transmission, such interventions likely will
Coinfection with human immunodeficiency be important for the future decline of HCV-​
virus also appears to increase the risk of HCV-​ associated liver cancer. In addition, more than
induced liver disease progression, most likely due 30 clinical studies have shown a relatively strong
to insufficient immune control of chronic HCV protective effect of interferon treatment on the
infection and increased hepatocellular damage development of HCC among HCV carriers, with
(Graham et al, 2001; Thomas, 2002). and without cirrhosis, who experienced a sus-
HCV infection may be associated with the risk tained virologic response (SVR) (Smith-​Palmer
of developing cholangiocarcinoma (IARC, 2012b; et  al, 2015). A  meta-​analysis by Morgan and
Zhou et al, 2014; Li et al, 2015). Of the case-​control colleagues (2013) estimated a 76% lower risk of
studies examining the relationship between HCV HCC for patients who achieved SVR in compar-
infection and cholangiocarcinoma, most have ison to those who did not. A major advance in
observed an increase risk ranging from three-​ the treatment of HCV has been the development
to 10-​fold. In a meta-​analysis of 16 case-​control of oral direct-​acting antiviral agents. Sustained
studies published through September 2011, the virologic response rates greater than 90% have
pooled relative risks were 3.38 for intrahepatic been observed in clinical trials of regimens
cholangiocarcinoma and 1.75 for extrahepatic based on two or more classes of these oral agents
29

292 Textbook of Cancer Epidemiology

(Li & Chung, 2015). It is anticipated that effective cholangiocarcinoma, with RRs ranging from 2.7
treatment of HCV will lead to further decreases to 13.6 (Choi et  al, 2006; Lee et  al, 2008; Shin
in HCC incidence. et al, 1996). Studies from China also support an
effect of C. sinensis infection on the occurrence of
Liver Flukes cholangiocarcinoma (Xia et al, 2015).
Cholangiocarcinoma, occurring predominantly Based on the available data in 1994, an
in Asia, has been linked to infection with the IARC Working Group concluded that O.  viver-
parasitic liver flukes Opisthorchis viverrini and rini infection was carcinogenic to humans and
Clonorchis sinensis (Parkin et  al, 1991; IARC, C.  sinensis infection was probably carcinogenic
1994b; 2012b). High endemic levels of liver (IARC, 1994b). In the subsequent 100B IARC
fluke infection can be found in Thailand and Monograph, C.  sinensis was reclassified as a
Laos (O.  viverrini) as well as in China and the human carcinogen, and a causal relationship
Republic of Korea (C.  sinensis) (IARC, 2012b). with cholangiocarcinoma was specified for both
Transmission occurs via consumption of raw or liver flukes (IARC, 2012b).
undercooked fish (Fried et  al, 2011). Chronic
infection with these food-​borne parasites usu- Ionizing Radiation
ally takes place in the intrahepatic bile ducts. Thorotrast (thorium dioxide) was a radio-
O. viverrini and C. sinensis likely exert a carcin- contrast agent used in the 1930s to the 1950s.
ogenic effect through inflammatory processes Exposure to thorotrast has been strongly associ-
leading to cell turnover and induction of reactive ated with an increased risk of liver cancer (IARC,
substances, which can result in the generation of 1987). Data on the relationship of Thorotrast
mutations in the host DNA; the helminths also exposure to cancer occurrence have originated
release excretory/​secretory products that could from follow-​up studies of patients injected with
have a direct effect on cell proliferation and the this radioactive contrast medium, in which RR
development of fibrosis and biliary hyperplasia estimates ranging from 13 to 126 were reported
(Fried et al, 2011). (Andersson & Storm, 1992; Mori et  al, 1999;
The epidemiologic evidence for a role of dos Santos Silva et  al, 2003; Travis et  al, 2003).
O.  viverrini and C.  sinensis in the etiology of Despite the potential limitations of these study
cholangiocarcinoma is based on a limited num- cohorts, including selection of an appropriate
ber of case-​ control studies using biomark- unexposed group, differential loss to follow-​
ers of infection. The prevalence of O.  viverrini up, quality of the cancer outcome information,
was evaluated among 18,393 adults living in and lack of data on important confounders, it is
Northeast Thailand, who participated in a mobile unlikely that bias could account for the strong
screening project (1990-​ 2001) and was found associations observed. The malignancy appears
to be positively associated with the incidence to occur long after the exposure, with the major-
of cholangiocarcinoma in the same areas at the ity of cases being of the cholangiocarcinoma and
population level (Sriamporn et  al, 2004). Two angiosarcoma subtypes.
hospital-​based studies in Thailand reported RRs Results from the Life Span Study of the
of 3.1–​5.0 for the presence of elevated antibodies atomic bomb survivors cohort in Hiroshima and
to O.  viverrini and cholangiocarcinoma (Parkin Nagasaki, Japan, have shown a dose-​response
et  al, 1991; Manwong et  al, 2013). In a nested relationship between whole-​ body radiation
case-​ control study, a similar association was exposure and the incidence of primary liver can-
found for a history of having O. viverrini eggs in cer (Thompson et  al, 1994). In contrast to the
stool (Poomphakwaen et al, 2009). A population-​ findings of the studies of Thorotrast, the effect
based case-​control study in Thailand reported a of the atomic bomb radiation exposure appears
more than 20-​fold higher risk of cholangiocar- to be related to the development of HCC and
cinoma with elevated antibody levels against not to the other histologic subtypes (Cologne
O. viverrini (Honjo et al, 2005). With respect to et  al, 1999). Additional analyses suggest that
C.  sinensis, endemicity of infection in Korea is the radiation-​ related excess risk is higher in
strongly associated with incidence of cholangio- men and in persons who were exposed in their
carcinoma (Lim et  al, 2006). In three hospital-​ 20s; however, the investigators noted that con-
based case-​control studies conducted in Korea, founding by known risk factors for HCC might
significant increased associations were observed account, in part, for these observations (Cologne
between evidence of C.  sinensis infection and et al, 1999).
293

Cancer of the Liver and Biliary Tract 293

Occupation countries, including in a meta-​ analysis of 18

A limited number of epidemiological studies studies published in 2012 (Wang et al, 2012). The
of occupational exposures to specific chemi- estimated RRs associated with liver cancer were
cals have suggested positive associations with 2.01 (95% CI 1.61–​2.51) for incidence and 1.56
risk of primary liver cancer. In the 2012 IARC (95% CI 1.30–​1.87) for mortality. Importantly,
report (2012c), vinyl chloride (VCM) and tri- associations were independent of geographic
chloroethylene (TCE) were suggested to have a location, alcohol consumption, history of cir-
carcinogenic effect on human liver. Uccello and rhosis, or infections with HBV or HCV. Similar
colleagues (2012) undertook a comprehensive findings were evident in cohort studies under-
review and found a moderately increased risk of taken among HBV-​(Fu et  al, 2015)  and HCV-​
HCC in individuals exposed to VCM, organic positive subjects (Huang et  al, 2015)  as well as
solvents such as TCE, pesticides, polychlorinated across different ethnic groups (Setiawan et  al
biphenyls, and arsenic. 2015). Moreover, a meta-​analysis of 21 studies
Vinyl chloride has been considered to of HCC patients indicated that diabetes melli-
increase the risk of developing angiosarcoma in tus may be associated with poorer prognosis,
the liver (Chen et  al, 1997; Ishak, 1997; IARC, reporting an RR of 1.46 (95% CI, 1.29–​1.66) for
2012c), and this has been supported by occupa- overall survival and 1.57 (95% CI, 1.21–​2.05) for
tional epidemiologic studies (Lewis & Rempala, disease-​free survival for HCC patients with ver-
2003; Boffetta et  al, 2003; Sherman 2009). In sus without diabetes mellitus (Wang et al, 2014).
addition, specific p53 mutation patterns occur Although the nature of this association and
in VCM–​associated angiosarcomas (Ishak, 1997; its biological significance are not well under-
Weihrauch et al, 2000). Exposure to VCM also has stood, exposure of the liver to elevated levels of
been linked to the occurrence of HCC (Boffetta insulin may lead to increased synthesis of IGF-​1,
et al, 2003; Mastrangelo et al, 2004; Anwar et al, which promotes cell growth and proliferation
2008; Lindbohm et al, 2009; Soliman et al, 2010). and inhibits apoptosis and contributes to carcin-
Trichloroethylene is an environmental and ogenesis (Weng et al, 2010). Both hepatitis virus
industrial pollutant and a common contaminant infection, particularly HCV, and heavy alcohol
of groundwater and air; however, its worldwide consumption have been reported to act synergis-
industrial use has decreased over time. TCE tox- tically with diabetes to increase the occurrence
icity mainly affects the kidney and liver (Caldwell of HCC (Donato et al, 2002; Hassan et al, 2002;
et  al, 2006, Wartenberg 2009). In an exten- Davila et al, 2005; Yuan et al, 2004).
sive review and meta-​analysis of the epidemio- Another possible mechanism is through the
logic evidence for the carcinogenicity of TCE, positive association of diabetes mellitus with non-
Wartenberg and colleagues (2000) estimated an alcoholic fatty liver disease (NAFLD) through a
overall relative risk of 1.9 (95% CI 1.0–​3.4) for the process of metabolic dysregulation and insulin
association of TCE and primary liver cancer inci- resistance. Case series reports have suggested a
dence, based on data from occupational cohorts possible association between NAFLD, and its
with the best-​quality exposure assessment. more serious manifestation—​nonalcoholic ste-
The possible association of angiosarcoma atohepatitis (NASH) and HCC mainly through
with arsenic includes exposure not only in occu- subsequent fibrosis and cirrhosis (Bugianesi
pational settings but also via medical treatment et al, 2002, Shimada et al, 2002; Baffy et al, 2012;
(e.g., Fowler’s solution and contaminated drink- Dongiovanni et  al, 2014). Population-​ based
ing water) (Chen et al, 1997; Ishak, 1997; IARC cohort studies in patients with NAFLD/​NASH
2012c). Intake of inorganic arsenic from well showed no, or moderate increased incidence
water also has been associated with HCC in epi- of HCC, respectively, over 6  years of follow-​up
demiologic studies conducted in Asia where high (Ong et al, 2008; Sanyal et al, 2010). In particu-
levels of arsenic contamination are present (Luo lar, NAFLD/​NASH was associated with increased
et  al, 1995; Tsuda et  al, 1995; Chen et  al, 1997; HCC risk when cirrhosis or fibrosis was also
Tsai et al, 1999). apparent (Ratziu et al, 2002; Sorensen et al, 2003;
Sanyal et al, 2006; Kojima et al, 2006; Yatsuji et al,
Medical Conditions 2009; Ascha et al, 2010; Bhala et al, 2011).
A fairly strong positive association of diabe- Two medical conditions that directly affect the
tes mellitus and liver cancer has been noted in biliary ducts also appear to be associated with liver
epidemiologic studies conducted in different cancer risk. Based on studies in Europe, patients
294

294 Textbook of Cancer Epidemiology

with primary sclerosing cholangitis, a relatively gallstone disease; for example, a high prevalence of
rare disease, have a high incidence of cholangio- gallstones has been observed in Native Americans
carcinoma (Bergquist et al, 1998; Chalasani et al, and Hispanics in the United States, populations in
2000; Razumilava and Gores 2014). Smoking and which gallbladder cancer rates are high (Lazcano-​
alcohol consumption seemed to enhance the risk Ponce et al, 2001; Randi et al, 2006).
of the malignancy among patients with this con- A number of case-​control studies (Kato et al,
dition (Bergquist et  al, 1998)  (Chalasani et  al, 1989; WHO Collaborative Study of Neoplasia
2000). An increased occurrence of HCC also has and Steroid Contraceptives, 1989a; Zatonski
been observed in patients with primary biliary cir- et al, 1997), and a few cohort studies (Chow et al,
rhosis (Lööf et al, 1994; Bishayee 2014), and it is 1999; Randi et al, 2006), have evaluated the asso-
likely that any type of cirrhosis, although to a var- ciation between history of gallstones or gallblad-
iable degree, may act as a growth-​enhancing fac- der disease and the risk of gallbladder cancer.
tor in hepatocarcinogenesis (Kew & Popper 1984; In a review and meta-​analysis of risk factors for
Trichopoulos et al, 1987). gallbladder cancer by Randi et  al (2006), sum-
mary RR estimates of 7.1 (95% CI 4.5–​11.2) for
CANCER OF THE the case-​control studies and 2.2 (95% CI 1.2–​4.2)
B I L I A RY   T R A C T for the cohort studies were obtained for history
of gallbladder disease.
Descriptive Epidemiology A family history of gallbladder disease or
Cancers of the biliary tract primarily include cancer appears to increase the risk of develop-
malignancies occurring in the gallbladder and ment of gallbladder cancer (Kato et  al, 1989;
extrahepatic bile ducts. Gallbladder cancer Fernandez et al, 1994; Hundal & Shaffer, 2014).
occurs more frequently than bile duct cancer, The frequency of cholecystectomy operations
accounting for 80% to 95% of biliary tract can- would affect the incidence of biliary tract cancer
cers, and is more common in women than in and likely explains some of the observed varia-
men (Eldon and Shaffer, 2014). On a global level, tions among countries and over time (Lazcano-​
the incidence of these cancers is relatively low. Ponce et al, 2001; Shaffer 2008). Cholecystectomy
The highest incidence rates of gallbladder cancer also is associated with a reduced risk of bile duct
are found in India, Japan, Korea, Latin America, cancer (Ekbom et  al, 1993; Chow et  al, 1999;
and eastern Europe; low incidence rates occur in Shaffer 2008).
northern Europe, the United States, and Canada
(Lazcano-​Ponce et  al, 2001; Randi et  al, 2006). Obesity and Diet
Differences in rates are also observed within Obesity may be another major risk factor for
countries, with respect to area of residence and biliary tract cancer, probably through the strong
racial group. Biliary tract cancer rates appear to relationship between overweight and the occur-
be increasing in Japan, Chile, Italy, and Spain rence of gallstones (WCRF, 1997; Calle & Kaaks,
and decreasing in the United States, Canada, the 2004). An association of obesity with gallblad-
United Kingdom, and Colombia (Lazcano-​Ponce der cancer has been observed for both men and
et al, 2001; Randi et al, 2006). women (Calle et al, 2003; Samanic et al, 2004),
although sometimes the association is more
Risk Factors evident for women (Zatonksi et al, 1997; Wolk
Gallbladder Disease et al, 2001; Randi et al, 2006). A meta-​analysis of
Probably due to its infrequent occurrence and rap- studies up to 2007 (Larsson & Wolk, 2007) esti-
idly fatal outcome, little epidemiologic research mated a summary RR of 1.15 (95% CI 1.01–​1.30)
has been done on the etiology of biliary tract can- for overweight and 1.66 (95% CI 1.47–​1.88) for
cer. The most consistently observed association is obesity compared to normal weight, and the
between a history of gallstones or gallbladder dis- association was stronger among women.
ease and the occurrence of gallbladder cancer. On With respect to diet and biliary tract cancer
a global level, a strong geographic correlation exists occurrence, frequent consumption of fruits and
between gallstone prevalence and gallbladder can- vegetables (Kato et al, 1989; Negri et al, 1991;
cer incidence (Lazcano-​Ponce et  al, 2001; Randi Pandey 2002) and fiber (Fedirko et al, 2013b)
et al, 2006). Variations in incidence within countries appears to be protective. There is inconsistent
also correspond to differences in the frequency of evidence of an association between intake of
295

Cancer of the Liver and Biliary Tract 295

high-​fat foods with gallbladder cancer (Kato et al, CONCLUSION

1989; Zatonksi et al, 1997). An increased level of Primary liver cancer encompasses HCC (>90%),
total energy intake also may be related to an ele- cholangiocarcinoma (<10%), and some other rare
vated occurrence of gallbladder cancer (Zatonksi histologic types. HCC is one of the most common
et al, 1997). cancers in Southeast Asia and in western Africa,
and it is almost uniformly fatal. During the last
Hormonal Factors 40 years, epidemiologic research has revealed the
Reproductive factors appear to play a role in the major causes of most HCC cases, most of which
development of biliary tract cancer. Increasing are preventable. Chronic infection with HBV
parity, which also is associated with gallstones, and HCV dominates the etiology of the disease
has been linked to an increased risk of gallblad- in low-​and middle-​income countries. Aflatoxin
der cancer (Lambe et al, 1993; Zatonski et al, exposure also likely contributes to the causation
1997; Randi et al, 2006; Andreotti et al, 2010). of HCC in Africa. In high-​ income countries,
Older age at first birth, however, may decrease however, these factors are less prevalent, and
risk (Lambe et al, 1993; Zatonski et al, 1997; alcoholic cirrhosis, cirrhosis of other etiology,
Andreotti et al, 2010). Reports of oral con- and tobacco smoking assume a relatively more
traceptive use in relation to biliary tract can- important role. Diet may also affect the devel-
cer do not support a strong association (WHO opment of liver cancer, but the relevant evidence
Collaborative Study of Neoplasia and Steroid is not conclusive, except perhaps with respect
Contraceptives, 1989a; Zatonski et al, 1997). In to the consumption of coffee, which seems to
a large claims database study of US women using be associated with a decrease of HCC risk. Rare
oral contraceptives, a small, statistically signif- chemical exposures, such as inorganic arsenic
icant increase in the risk of gallbladder disease and VCM, and, in the past, internal radiation
was apparent for specific formulations of oral from the contrast medium thorotrast, have been
contraceptives (Etminan et al, 2011). The limited responsible for a small proportion of cases. Some
number of studies conducted have not found a inherited conditions, notably alpha1-​antitrypsin
consistent association of hormone replacement deficiency, hemochromatosis, and porphyria
therapy with biliary tract cancer risk (Zatonski et cutanea tarda, also increase the risk of HCC.
al, 1997; Gallus et al, 2002). Cholangiocarcinoma has been linked to chronic
infection with O.  viverrini and C.  sinensis, both
Mechanisms food-​borne parasites. Cancer of the extrahepatic
Although all of the potential risk factors identi- bile ducts is closely associated with cholelithiasis
fied for biliary tract cancer are also associated and its attending risk factors, in particular female
with the development of gallstones (Lambe et gender, obesity, and high parity.
al, 1993; Zatonski et al, 1997; Lazcanoy-​Ponce
et al, 2001), it is not established whether there REFERENCES
is a common etiology. Genetic susceptibility Adami HO, Hsing AW, McLaughlin JK, Trichopoulos
has been also postulated (Lazcano-​Ponce et al, D, Hacker D, Ekbom A et al. Alcoholism and liver
2001). The mechanism by which cholelithia- cirrhosis in the etiology of primary liver cancer.
sis contributes to biliary tract carcinogenesis is Int J Cancer 1992;51:898–​902.
not clear. Chronic inflammation probably plays Altekruse SF, McGlynn KA, Reichman ME. Hepato­
cellular carcinoma incidence, mortality, and sur-
a role (Lazcano-​Ponce et al, 2001), a hypothesis
vival trends in the United States from 1975 to
that is supported by the strong long-​term pro-
2005. J Clin Oncol 2009;27:1485–​91.
tective effective of removal of the gallbladder on
Alter MJ. Epidemiology of hepatitis C virus infection.
the development of bile duct cancer (Ekbom et World J Gastroenterol 2007;13:2436–​41.
al, 1993). Alternatively, metabolic abnormalities Alter MJ, Kruszon-​Moran D, Nainan OV, McQuillan
or changes in bile composition are relevant to GM, Gao F, Moyer LA et al. The prevalence of hep-
both gallstone formation and biliary tract cancer atitis C virus infection in the United States, 1988
development (Lambe et al, 1993; Lazcano-​Ponce through 1994. N Engl J Med 1999;341:556–​62.
et al, 2001). Nonetheless, avoidance of behaviors Alter MJ, Margolis HS, Krawczynski K, Judson FN,
and exposures that may increase the risk of gall- Mares A, Alexander WJ et  al. The natural his-
bladder disease would likely reduce the incidence tory of community-​acquired hepatitis C in the
of biliary tract cancer. United States:  the Sentinel Counties Chronic
296

296 Textbook of Cancer Epidemiology

non-​A, non-​B Hepatitis Study Team. N Engl J associations using extended follow-​up of the origi-
Med 1992;327:1899–​905. nal Whitehall study. Ann Oncol 2008;19:996e1002.
Alzahrani B, Iseli TJ, Hebbard LW. Non-​viral causes of Batty GD, Shipley MJ, Kivimaki M, Barzi F, Smith
liver cancer: does obesity led inflammation play a GD, Mitchell R et  al. Obesity and overweight in
role? Cancer Lett 2014;345:223–​9. relation to liver disease mortality in men: 38-​year
Andant C, Puy H, Bogard C, Faivre J, Soulé JC, follow-​up of the original Whitehall study. Int J
Nordmann Y et  al. Hepatocellular carcinoma in Obes 2008;32:1741–​4.
patients with acute hepatic porphyria:  frequency Behrens G, Matthews CE, Moore SC, Freedman
of occurrence and related factors. J Hepatol ND, McGlynn KA, Everhart JE et  al. The asso-
2000;32:933–​9. ciation between frequency of vigorous physical
Andersson M, Storm HH. Cancer incidence among activity and hepatobiliary cancers in the NIH-​
Danish thorotrast-​exposed patients. J Natl Cancer AARP Diet and Health Study. Eur J Epidemiol
Inst 1992;84:1318–​25. 2013;28:55–​66.
Andreotti G, Hou L, Gao Y-​T, Brinton L A, Rashid Bellentani S, Pozzato G, Saccoccio G, Crovatto M,
A, Chen J et  al. Reproductive factors and risks Croce LS, Mazzoran L et  al. Clinical course and
of biliary tract cancers and stones: a population-​ risk factors of hepatitis C virus related liver dis-
based study in Shanghai, China Br J Cancer ease in the general population:  report from the
2010;102:1185–​9. Dionysos study. Gut 1999;44:874–​80.
Anwar WA, Khaled HM, Amra HA, El-​Nezami H, Bergquist A, Glaumann H, Persson B, Broome U.
Loffredo CA. Changing pattern of hepatocel- Risk factors and clinical presentation of hepato-
lular carcinoma (HCC) and its risk factors in biliary carcinoma patients with primary sclero-
Egypt:  possibilities for prevention. Mutat Res sing cholangitis: a case-​control study. Hepatology
2008;659:176–​84. 1998;27:311–​6.
Ascha MS, Hanouneh IA, Lopez R, Tamimi TA, Bhala N, Angulo P, van der Poorten D, Lee E, Hui JM,
Feldstein AF, Zein NN. The incidence and risk Saracco G et  al. The natural history of nonalco-
factors of hepatocellular carcinoma in patients holic fatty liver disease with advanced fibrosis or
with nonalcoholic steatohepatitis. Hepatology cirrhosis:  an international collaborative study.
2010;51:1972–​8. Hepatology 2011;54:1208–​16.
Austin H. The role of tobacco use and alcohol con- Bigsby RM, Caperell-​Grant A. The role for estrogen
sumption in the etiology of hepatocellular car- receptor-​ alpha and prolactin receptor in sex-​
cinoma. In:  Tabor E, Di Bisceglie AM, Purcell dependent DEN-​ induced liver tumorigenesis.
RH (eds.). Etiology, Pathology, and Treatment Carcinogenesis 2011;32:1162–​6.
of Hepatocellular Carcinoma in North America. Bishayee A. The role of inflammation and liver cancer.
Advances in Applied Biotechnology Series, Vol. Adv Exp Med Biol 2014;816:401–​35.
13. Gulf, Houston, 1991: 57–​75. Boccia S, Miele L, Panic N, Turati F, Arzani D, Cefalo
Baffy G, Brunt EM, Caldwell SH. Hepatocellular car- C et  al. The effect of CYP, GST, and SULT poly-
cinoma in non-​alcoholic fatty liver disease:  an morphisms and their interaction with smoking on
emerging menace. J Hepatol 2012;56:1384e91. the risk of hepatocellular carcinoma. Biomed Res
Bamia C, Lagiou P, Jenab M, Aleksandrova K, Fedirko Int 2015;2015:179867.
V, Trichopoulos D et al. Fruit and vegetable con- Block TM, Mehta AS, Fimmel CJ, Jordan R. Molecular
sumption in relation to hepatocellular carcinoma viral oncology of hepatocellular carcinoma.
in a multi-​centre, European cohort study. Br J Oncogene 2003;22:5093–​107.
Cancer 2015a;112:1273–​82. Boffetta P, Matisane L, Mundt KA, Dell LD. Meta-​
Bamia C, Lagiou P, Jenab M, Trichopoulou A, analysis of studies of occupational exposure to
Fedirko V, Aleksandrova K et al. Coffee, tea and vinyl chloride in relation to cancer mortality.
decaffeinated coffee in relation to hepatocellu- Scand J Work Environ Health 2003;29:220–​9.
lar carcinoma in a European population:  mul- Bohn SK, Blomhoff R, Paur I. Coffee and cancer risk,
ticentre, prospective cohort study. Int J Cancer epidemiological evidence and molecular mecha-
2015b;136:1899–​908. nisms. Mol Nutr Food Res 2014;58:915–​30.
Batty GD, Barzi F, Huxley R, Chang CY, Jee SH, Borena W, Strohmaier S, Lukanova A, Bjørge T,
Jamrozik K et  al. Obesity and liver cancer mor- Lindkvist B, Hallmans G et al. Metabolic risk fac-
tality in Asia:  the Asia Pacific Cohort Studies tors and primary liver cancer in a prospective study
Collaboration. Cancer Epidemiol 2009;33:469–​72. of 578,700 adults. Int J Cancer 2012;131:193–​200.
Batty GD, Kivimaki M, Gray L, Smith GD, Marmot Bosch FX, Ribes J, Cleries R, Diaz M. Epidemiology
MG, Shipley MJ. Cigarette smoking and site-​ of hepatocellular carcinoma. Clin Liver Dis
specific cancer mortality:  testing uncertain 2005;9:191–​211.
297

Cancer of the Liver and Biliary Tract 297

Bosch FX, Ribes J, Diaz M, Cleries R. Primary risk of hepatocellular carcinoma. Hepatology
liver cancer:  worldwide incidence and trends. 1996;24:38–​42.
Gastroenterol 2004;127:S5–​16. Chen CJ, Yang HI, Su J et  al. REVEAL-​HBV Study
Braga C, La Vecchia C, Negri E, Franceschi S. Group. Risk of hepatocellular carcinoma across
Attributable risks for hepatocellular carcinoma in a biological gradient of serum hepatitis B virus
northern Italy. Eur J Cancer 1997;33:629–​34. DNA level. JAMA 2006;295:65–​73.
Bravi F, Tavani A, Bosetti C, Boffetta P, La Vecchia C. Chen CJ, Yu M-​W, Liaw Y-​F. Epidemiological char-
Coffee and the risk of hepatocellular carcinoma acteristics and risk factors of hepatocellular
and chronic liver disease: a systematic review and carcinoma. J Gastroenterol Hepatol 1997;12
meta-​analysis of prospective studies. Eur J Cancer Suppl:S294–​308.
Prev 2016; Apr 22. [Epub ahead of print]. Chen Y, Wang X, Wang J, Yan Z, Luo J. Excess body
Brechot C, Gozuacik D, Murakami Y, Paterlini-​Brechot weight and the risk of primary liver cancer:  an
P. Molecular bases for the development of hepatitis updated meta-​analysis of prospective studies. Eur
B virus (HBV)-​related hepatocellular carcinoma J Cancer 2012;48:2137–​45.
(HCC). Sem Cancer Biology 2000;10:211–​31. Choi D, Lim JH, Lee KT et  al. Cholangiocarcinoma
Bugianesi E, Leone N, Vanni E, Marchesini G, and Clonorchis sinensis infection:  a case-​control
Brunello F, Carucci P et al. Expanding the natu- study in Korea. J Hepatol 2006;44:1066–​73.
ral history of nonalcoholic steatohepatitis:  from Choo QL, Kuo G, Weiner AJ, Overby LR, Bradley DW,
cryptogenic cirrhosis to hepatocellular carcinoma. Houghton M. Isolation of a cDNA clone derived
Gastroenterol 2002;123:134–​40. from a blood-​borne non-​A, non-​B viral hepatitis
Caldwell JC, Keshava N. Key issues in the modes of genome. Science 1989;244:359–​62.
action and effects of trichloroethylene metabo- Chow WH, Johansen C, Gridley G, Mellemkjaer L,
lites for liver and kidney tumorigenesis. Environ Fraumeni Jr JF. Gallstones, cholecystectomy and
Health Perspect 2006;114:1457–​63. risk of cancers of the liver, biliary tract and pan-
Calle EE, Kaaks R. Overweight, obesity and can- creas. Br J Cancer 1999; 79:640–​4.
cer:  epidemiological evidence and proposed Chuang SC, Lee YC, Hashibe M, Dai M, Zheng T,
mechanisms. Nat Rev Cancer 2004;4:579–​91. Boffetta P. Interaction between cigarette smoking
Calle EE, Rodriguez C, Walker-​Thurmond K, Thun and hepatitis B and C virus infection on the risk
MJ. Overweight, obesity, and mortality from can- of liver cancer: a meta-​analysis. Cancer Epidemiol
cer in a prospectively studied cohort of U.S. adults. Biomarkers Prev 2010;19:1261e8.
N Engl J Med 2003;348:1625–​38. Clemente C, Elba S, Buongiorno G, Berloco P, Guerra
Chalasani N, Baluyut A, Ismail A, Zaman A, Sood G, V, Di Leo A. Serum retinol and risk of hepatocel-
Ghalib R et  al. Cholangiocarcinoma in patients lular carcinoma in patients with child-​Pugh class
with primary sclerosing cholangitis: a multicenter A cirrhosis. Cancer Letters 2002;178:123–​9.
case-​control study. Hepatology 2000;31:7–​11. Cologne JB, Tokuoka S, Beebe GW, Fukuhara T,
Chan HL, Tse CH, Mo F et al. High viral load and hep- Mabuchi K. Effects of radiation on incidence of
atitis B virus subgenotype ce are associated with primary liver cancer among atomic bomb survi-
increased risk of hepatocellular carcinoma. J Clin vors. Radiation Res 1999;152:364–​73.
Oncol 2008;26:177–​82. Corrao G, Zambon A, Bagnardi V, D’Amicis A, Klatsky
Chan KY, Wong CM, Kwan JS, Lee JM, Cheung KW, A. Coffee, caffeine, and the risk of liver cirrhosis.
Yuen MF et  al. Genome-​wide association study Ann Epidemiol 2001:11:458–​65.
of hepatocellular carcinoma in Southern Chinese Daniel CR, Cross AJ, Graubard BI, Hollenbeck AR,
patients with chronic hepatitis B virus infection. Park Y, Sinha R et al. Prospective investigation of
PLoS One. 2011;6(12):e28798. poultry and fish intake in relation to cancer risk.
Chang MH, Chen CJ, Lai MS, Hsu HM, Wu TC, Kong Cancer Prev Res 2011;4:1903–​11.
MS et  al. Universal hepatitis B vaccination in Davila JA, Morgan RO, Shaib Y, McGlynn KA, El-​
Taiwan and the incidence of hepatocellular carci- Serag HB. Diabetes increases the risk of hepatocel-
noma in children. Taiwan Childhood Hepatoma lular carcinoma in the United States: a population
Study Group. N Engl J Med 1997;336:1855–​9. based case control study. Gut 2005;54:533–​9.
Chang MH, You SL, Chen CJ, Liu CJ, Lee CM, De Angelis R, Sant M, Coleman MP, Francisci S,
Lin SM et  al. Decreased incidence of hepato- Baili P, Pierannunzio D et  al. Cancer survival in
cellular carcinoma in hepatitis B vaccinees:  a Europe 1999-​2007 by country and age:  results
20-​year follow-​ up study. J Natl Cancer Inst of EUROCARE—​5-​a population-​based study.
2009;101:1348–​55. Lancet Oncol 2014;15:23–​34.
Chen CJ, Wang LY, Lu SN, Wu MH, You SL, Zhang Dienstag JL. Hepatitis B virus infection. N Engl J Med
YJ et al. Elevated aflatoxin exposure and increased 2008;359:1486–​1500.
298

298 Textbook of Cancer Epidemiology

Donato F, Boffetta P, Puoti M. A meta-​analysis of Investigation into Cancer and Nutrition (EPIC).
epidemiologic studies on the combined effect Ann Oncol 2013a;24:2166–​73.
of hepatitis B and C virus infections in caus- Fedirko V, Lukanova A, Bamia C, Trichopolou A,
ing hepatocellular carcinoma. Int J Cancer Trepo E, Nöthlings U et al. Glycemic index, glyce-
1998;75:347–​54. mic load, dietary carbohydrate, and dietary fiber
Donato F, Tagger A, Gelatti U, Parrinello G, Boffetta intake and risk of liver and biliary tract cancers in
P, Albertini A et  al. Alcohol and hepatocellular Western Europeans. Ann Oncol 2013b;24:543–​53.
carcinoma:  the effect of lifetime intake and hep- Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser
atitis virus infections in men and women. Am J S, Mathers C, Rebelo M, Parkin DM, Forman D,
Epidemiol 2002;255:323–​31. Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence
Dongiovanni P, Romeo S, Valenti L. Hepatocellular and Mortality Worldwide: IARC CancerBase No.
carcinoma in nonalcoholic fatty liver:  role of 11 [Internet]. Lyon, France: International Agency
environmental and genetic factors. World J for Research on Cancer; 2013. Available from:
Gastroenterol 2014;20:12945–​55. http://globocan.iarc.fr, accessed 2017.
dos Santos Silva I, Malveiro F, Jones ME, Swerdlow Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C,
AJ. Mortality after radiological investigation Rebelo M, Parkin DM, Forman D, Bray F. Cancer
with radioactive thorotrast:  a follow-​ up study incidence and mortality worldwide:  sources,
of up to fifty years in Portugal. Radiation Res methods and major patterns in GLOBOCAN
2003;159:521–​34. 2012. Int J Cancer 2015 Mar;136:E359–​86.
Ekbom A, Hsieh CC, Yuen J, Trichopoulos D, Fernandez E, La Vecchia C, D’Avanzo B, Negri E,
McLaughlin JK, Lan SJ et al. Risk of extrahepatic Franceschi S. Family history and the risk of
bile duct cancer after cholecystectomy. Lancet liver, gallbladder, and pancreatic cancer. Cancer
1993;342:1262–​5. Epidemiol Biomarkers Prev 1994;3:209–​12.
Elder G, Harper P, Badminton M, Sandberg S, Fix OK, Kowdley KV. Hereditary hemochromatosis.
Deybach JC. The incidence of inherited porphyr- Minerva Med 2008;99:605–​17.
ias in Europe. J Inherit Metab Dis 2013;36:849–​57. Fon Sing M, Yang WS, Gao S, Gao J, Xiang YB.
Eldon RH Shaffer A. Gallbladder cancer:  epidemiol- Epidemiological studies of the association
ogy and outcome. Clin Epidemiol 2014;6:99–​109. between tea drinking and primary liver cancer: a
El-​
Serag HB, Engels EA, Landgren O, Chiao E, meta-​analysis. Eur J Cancer Prev 2011;20:157–​65.
Henderson L, Amaratunge HC, Giordano TP. Risk Freeman AJ, Dore GJ, Law MG, Thorpe M, Von
of hepatobiliary and pancreatic cancers after hep- Overbeck J, Lloyd AR et  al. Estimating progres-
atitis C virus infection: a population-​based study sion to cirrhosis in chronic hepatitis C virus infec-
of U.S. veterans. Hepatology 2009;49:116–​23. tion. Hepatology 2001;34:809–​16.
El-​Serag HB, Kanwal F. Epidemiology of hepatocellu- Fried B, Reddy A, Mayer D. Helminths in human car-
lar carcinoma in the United States: where are we? cinogenesis. Cancer Lett 2011;305:239–​49.
Where do we go? Hepatology 2014;60:1767–​75. Fu SC, Huang YW, Wang TC, Hu JT, Chen DS, Yang
El-​
Serag HB, Rudolph L. Reviews in basic and SS. Increased risk of hepatocellular carcinoma in
clinical gastroenterology. Gastroenterology chronic hepatitis B patients with new onset diabe-
2007;132:2557–​76. tes: a nationwide cohort study. Aliment Pharmacol
Etminan M, Delaney JAC, Bressler B, Brophy JM. Ther 2015;41:1200–​9.
Oral contraceptives and the risk of gallblad- Fujie H, Moriya K, Shintani Y, Yotsuyanagi H, Iino S,
der disease:  a comparative safety study. CMAJ Koike K. Hepatitis B virus genotypes and hepa-
2011;183:899–​904. tocellular carcinoma in Japan. Gastroenterology
Evans AA, Chen G, Ross EA, Sheu F-​M, Lin W-​Y, 2001;120:1564–​5.
London WT. Eight-​year follow-​up of the 90,000-​ Fwu C-​W, Chien Y-​C, You S-​L, Nelson KE, Kirk GD,
person Haimen City Cohort. I.  Hepatocellular Kuo H-​S, Feinleib M, Chen C-​J. Hepatitis b virus
carcinoma mortality, risk factors, and gender infection and risk of intrahepatic cholangiocar-
differences. Cancer Epidemiol, Biomarkers Prev cinoma and non-​Hodgkin lymphoma:  a cohort
2002;11:369–​76. study of parous women in Taiwan. Hepatology
Fattovich G, Stroffolini T, Zagni I, Donato F. Hepato­ 2011;53:1217–​25.
cellular carcinoma in cirrhosis: incidence and risk Gallus S, Negri E, Chatenoud L, Bosetti C, Franceschi
factors. Gastroenterology 2004;127:S35–​50. S, La Vecchia C. Post-​ menopausal hormonal
Fedirko V, Trichopolou A, Bamia C, Duarte-​Salles therapy and gallbladder cancer risk. Int J Cancer
T, Trepo E, Aleksandrova K, Nöthlings U et  al. 2002;99:762–​3.
Consumption of fish and meats and risk of hepa- Graham CS, Baden LR, Yu E, Mrus JM, Carnie J, Heeren
tocellular carcinoma:  the European Prospective T et  al. Influence of human immunodeficiency
29

Cancer of the Liver and Biliary Tract 299

virus infection on the course of hepatitis C carcinoma: insights into the etiology and pathogen-
virus infection:  a meta-​analysis. Clin Infect Dis esis of liver cancer. Oncogene 2007:2;26:2166–​76.
2001;33:562–​9. Huxley R, Lee CM, Barzi F, Timmermeister L,
Guo YM, Wei WY, Shen XZ. Tumour necrosis factor Czernichow S, Perkovic V et  al. Coffee, decaf-
308 polymorphisms and hepatocellular carcinoma feinated coffee, and tea consumption in relation
risk:  a meta-​ analysis. Hepatogastroenterology to incident type 2 diabetes mellitus:  a system-
2010;57:926–​931. atic review with meta-​analysis. Arch Intern Med
Gwack J, Park SK, Lee E-​U, Park B, Choi Y, Yoo K-​Y. 2009;169:2053–​63.
Hepatitis B vaccination in liver cancer mortality Innala E, Andersson C. Screening for hepatocellular
reduction in Korea in children and adolescents. carcinoma in acute intermittent porphyria: a 15-​
Asian Pacific J Cancer Prev 2011;12:2205–​8. year follow-​up in northern Sweden. J Intern Med
Hassan MM, Hwang L-​Y, Hatten CJ, Swaim M, Li 2011;269:538–​45.
D, Abbruzzese JL et  al. Risk factors for hepato- Inoue M, Yoshimi I, Sobue T, Tsugane S. For the
cellular carcinoma:  synergism of alcohol with JPHC Study Group. Influence of coffee drinking
viral hepatitis and diabetes mellitus. Hepatology on subsequent risk of hepatocellular carcinoma: a
2002;36:1206–​1213. prospective study in Japan. J Natl Cancer Inst
Honjo S, Srivatanakul P, Sriplung H, Kikukawa H, 2005;97:293–​300.
Hanai S, Uchida K et al. Genetic and environmen- Inoue M, Yamamoto S, Kurahashi N, Iwasaki M,
tal determinants of risk for cholangiocarcinoma Sasazuki S, Tsugane S. Japan Public Health
via Opisthorchis viverrini in a densely infested Center-​based prospective study group: daily total
area in Nakhon Phanom, northeast Thailand. Int J physical activity level and total cancer risk in men
Cancer 2005;117:854–​60. and women: results from a large-​scale population-​
Hoofnagle JH. Course and outcome of hepatitis C. based cohort study in Japan. Am J Epidemiol
Hepatology 2002;36 Suppl 1:S21–​9. 2008;168:391–​403.
Hsing AW, McLaughlin JK, Hoover RN, Co-​Chien International Agency for Research on Cancer Mono­
HT, Blot WJ, Fraumeni Jr JF. Parity and primary graphs on the Evaluation of Carcinogenic Risks to
liver cancer among young women. J Natl Cancer Humans, Overall Evaluations of Carcinogenicity:
Inst 1992;84:1118–​9. An Updating of IARC Monographs Volumes 1 to
Hu G, Tuomilehto J, Pukkala E, Hakulinen T, 42, Suppl. 7, Lyon: IARC Press, 1987, pp. 313–​316.
Antikainen R, Vartiainen E et  al. Joint effects International Agency for Research on Cancer.
of coffee consumption and serum gamma-​ Alcohol drinking. IARC Monographs on the
glutamyltransferase on the risk of liver cancer. Evaluation of Carcinogenic Risks to Humans, Vol.
Hepatology 2008;48:129–​36. 44: Lyon: IARC, 1988.
Huang W, Zhang H, Hao Y, Xu X, Zhai Y, Wang S International Agency for Research on Cancer. Coffee,
et  al. A non-​synonymous single nucleotide pol- Tea, Mate, Methylxanthines and Methylglyoxal.
ymorphism in the HJURP gene associated with IARC Monographs on the Evaluation of Carcino­
susceptibility to hepatocellular carcinoma among genic Risks to Humans, Vol. 51. Lyon: IARC, 1991.
Chinese. PLoS One 2016;11:e0148618. International Agency for Research on Cancer.
Huang YQ, Lu X, Min H, Wu QQ, Shi XT, Bian KQ Hepatitis viruses. IARC Monographs on the
et  al. Green tea and liver cancer risk:  a meta-​ Evaluation of Carcinogenic Risks to Humans, Vol.
analysis of prospective cohort studies in Asian 59: Lyon: IARC, 1994a; p. 1–​255.
populations. Nutrition 2016;32:3–​8. International Agency for Research on Cancer Working
Huang YW, Wang TC, Yang SS, Lin SY, Fu SC, Hu JT Group on the Evaluation of Carcinogenic Risks
et  al. Increased risk of hepatocellular carcinoma to Humans. Schistosomes, liver flukes and
in chronic hepatitis C patients with new onset Helicobacter pylori. IARC Monographs on the
diabetes:  a nation-​ wide cohort study. Aliment Evaluation of Carcinogenic Risks to Humans. Vol.
Pharmacol Ther 2015;42:902–​11. 61: Lyon: IARC, 1994b; p. 1–​241.
Hundal R, Shaffer EA. Gallbladder cancer:  epi- International Agency for Research on Cancer.
demiology and outcome. Clin Epidemiol Pharmaceuticals. IARC Monographs on the
2014;6:99–​109. Evaluation of Carcinogenic Risks to Humans, Vol.
Hursting SD, Lavigne JA, Berrigan D, Perkins SN, 100A. Lyon:  IARC, 2012a. monographs.iarc.fr/​
Barrett JC. Calorie restriction, aging, and cancer ENG/​Monographs/​vol100A/​index.php.
prevention: mechanisms of action and applicabil- International Agency for Research on Cancer. Hepatitis
ity to humans. Ann Rev Med 2003;54:131–​52. B Virus; Hepatitis C Virus; Opisthorchis viverrini
Hussain SP, Schwank J, Staib F, Wang XW, Harris and Clonorchis sinensis. IARC Monographs on
CC. TP53 mutations and hepatocellular the Evaluation of Carcinogenic Risks to Humans,
30

300 Textbook of Cancer Epidemiology

Vol. 100B. Lyon:  IARC, 2012b. monographs.iarc. Kato K, Akai S, Tominaga S, Kato I. A case-​control
fr/​ENG/​Monographs/​vol100B/​. study of biliary tract cancer in Niigata Prefecture,
International Agency for Research on Cancer. Arsenic, Japan. Jpn J Cancer Res 1989;80:932–​8.
Metals, Fibres and Dusts. IARC Monographs on Kew M C. Hepatic tumors and cysts. In: Feldman M,
the Evaluation of Carcinogenic Risks to Humans, Friedman LS, Sleisenger MH, Scharschmidt BF
Vol. 100C. Lyon:  IARC, 2012c. monographs.iarc. (eds.) Sleisenger and Fordtran’s Gastrointestinal
fr/​ENG/​Monographs/​vol100C/​. and Liver Disease:  Pathophysiology/​Diagnosis/​
International Agency for Research on Cancer. Tobacco Management. 7 ed. Philadelphia:  Saunders,
Smoking. IARC Monographs on the Evaluation 2002: 1589–​90.
of Carcinogenic Risks to Humans, Vol. 100D. Kew MC. Synergistic interaction between aflatoxin
Lyon:  IARC, 2012d. monographs.iarc.fr/​ENG/​ B1 and hepatitis B virus in hepatocarcinogenesis.
Monographs/​vol100E/​. Liver Intl 2003;23:405–​9.
International Agency for Research on Cancer. Kew MC. Interaction between hepatitis B and C
Consumption of Alcoholic Beverages. IARC viruses in hepatocellular carcinogenesis. J Viral
Monographs on the Evaluation of Carcinogenic Hepat 2006;13:145–​9.
Risks to Humans, Vol. 100E. Lyon: IARC, 2012e. Kew MC. Aflatoxins as a cause of hepatocellular carci-
monographs.iarc.fr/​ENG/​Monographs/​vol100E/​. noma. J Gastrointestin Liver Dis 2013;22:305–​10.
International Agency for Research on Cancer. Kiyosawa K, Sodeyama T, Tanaka E, Gibo Y, Yoshizawa
Aflatoxins. IARC Monographs on the Evaluation K, Nakano Y et al. Interrelationship of blood trans-
of Carcinogenic Risks to Humans, Vol. 100F. fusion, non-​A, non-​B hepatitis and hepatocellular
Lyon:  IARC, 2012f. monographs.iarc.fr/​ENG/​ carcinoma:  analysis by detection of antibody to
Monographs/​vol100F/​index.php. hepatitis C virus. Hepatology 1990;12:671–​5.
International Agency for Research on Cancer. Kiyosawa K, Tanaka E, Sodeyama T, Yoshizawa K,
Coffee, Mate and Very Hot Beverages. IARC Yabu K, Furuta K et al. Transmission of hepatitis C
Working Group. Lyon, France; 24–​ 31 May, in an isolated area in Japan: community-​acquired-​
2016. IARC Monographs on the Evaluation infection: the South Kiso Hepatitis Study Group.
of Carcinogenic Risks to Humans, Vol. 116. Gastroenterology 1994;106:1596–​602.
Lyon: IARC (in press). Koh WP, Robien K, Wang R, Govindarajan S, Yuan JM,
Iso H, Kubota Y. Nutrition and disease in the Japan Yu MC. Smoking as an independent risk factor for
Collaborative Cohort Study for Evaluation of hepatocellular carcinoma: the Singapore Chinese
Cancer (JACC). Asian Pac J Cancer Prev 2007;8 Health Study. Br J Cancer 2011;105:1430e5.
Suppl:35–​80. Kojima H, Sakurai S, Matsumura M, Umemoto N,
Japanese Red Cross Non A, Non B-​Hepatitis Research Uemura M, Morimoto H et  al. Cryptogenic cir-
Group. Effect of screening for hepatitis C virus rhosis in the region where obesity is not prevalent.
antibody and hepatitis B virus core antibody on World J Gastroenterol 2006;12:2080–​5.
incidence of post-​ transfusion hepatitis. Lancet Kuang SY, Jackson PE, Wang JB, Lu PX, Munoz A,
1991;338:1040–​1. Qian GS et  al. Specific mutations of hepatitis B
Jee SH, Yun JE, Park EJ, Cho ER, Park IS, Sull JW et al. virus in plasma predict liver cancer development.
Body mass index and cancer risk in Korean men Proc Natl Acad Sci U.S.A. 2004;101:3575–​80.
and women. Int J Cancer 2008;123:1892–​6. Kudo M. Liver cancer:  surveillance, diagnosis, treat-
Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman ment, and outcome of liver cancer in Japan. Liver
D. Global cancer statistics. CA Cancer J Clin Cancer 2015;4:39–​50.
2011;61:69–​90. Kummee P, Tangkijvanich P, Poovorawan Y,
Johnson S, Koh WP, Wang R, Govindarajan S, Yu MC, Hirankarn N. Association of HLA-​DRB1*13 and
Yuan JM. Coffee consumption and reduced risk TNF-​alpha gene polymorphisms with clearance
of hepatocellular carcinoma:  findings from the of chronic hepatitis B infection and risk of hepa-
Singapore Chinese Health Study. Cancer Causes tocellular carcinoma in Thai population. J Viral
Control 2011;22:503–​10. Hepat 2007;14:841–​8.
Kaklamani E, Trichopoulos D, Tzonou A, Zavitsanos X, Kuper H, Hsieh CC, Stuver SO, Mucci LA, Tzonou A,
Koumantaki Y, Hatzakis A et al. Hepatitis B and C Zavitsanos X et al. Birth order, as a proxy for age
viruses and their interaction in the origin of hepato- at infection, in the etiology of hepatocellular car-
cellular carcinoma. JAMA 1991;265:1974–​6. cinoma. Epidemiology 2000a;11:680–​3.
Kanazir M, Boricic I, Delic D, Tepavcevic DK, Kuper H, Ye W, Broomé U, Romelsjö A, Mucci LA,
Knezevic A, Jovanovic T, Pekmezovic T. Risk fac- Ekbom A et al. The risk of liver and bile duct can-
tors for hepatocellular carcinoma:  a case-​control cer in patients with chronic viral hepatitis, alco-
study in Belgrade (Serbia). Tumori 2010;96:911–​7. holism, or cirrhosis. Hepatology 2001a;34:714–​8.
301

Cancer of the Liver and Biliary Tract 301

Kuper H, Mantzoros C, Lagiou P, Tzonou A, Tamimi La Vecchia C, Negri E, Franceschi S, D’Avanzo B.

R, Mucci L et al. Estrogens, testosterones and sex Reproductive factors and the risk of hepatocellular
hormone binding globulin in relation to liver can- carcinoma in women. Int J Cancer 1992;52:351–​4.
cer in men. Oncology 2001b;60:355–​60. Lazcano-​Ponce EC, Miquel JF, Munoz N, Herrero R,
Kurahashi N, Inoue M, Iwasaki M, Tanaka Y, Ferrecio C, Wistuba II et  al. Epidemiology and
Mizokami M, Tsuqane S, JPHC Study Group. molecular pathology of gallbladder cancer. CA
Vegetable, fruit and antioxidant nutrient con- Cancer J Clin 2001;51:349–​64.
sumption and subsequent risk of hepatocellular Lee H-​S, Yoon J-​W, Kamimura S, Iwata K, Watanabe
carcinoma:  a prospective cohort study in Japan. H, Kim CY. Lack of association of cytochrome
Br J Cancer 2009;100:181–​4. P450 2E1 genetic polymorphisms with the risk
Kurozawa Y, Ogimoto I, Shibata A, Nose T, of human hepatocellular carcinoma. Int J Cancer
Yoshimura T, Suzuki H et al. Dietary habits and 1997;71:737–​40.
risk of death due to hepatocellular carcinoma Lee BS, Park E-​C, Park SW, Nam CM, Roh J. Hepatitis
in a large scale cohort study in Japan:  univari- B virus infection, diabetes mellitus, and their syn-
ate analysis of JACC study data. Kurume Med J ergism for cholangiocarcinoma development:  a
2004;51:141–​9. case-​control study in Korea. World J Gastroenterol
Kwon SH, Jeong SW, Jang JY, Lee JE, Lee SH, Kim SG 2015;21:502–​10.
et  al. Cyclooxygenase-​ 2 and vascular endothe- Lee JS. Genomic profiling of liver cancer. Genomics
lial growth factor in chronic hepatitis, cirrhosis Inform 2013;11:180–​5.
and hepatocellular carcinoma. Clin Mol Hepatol Lee JS. The mutational landscape of hepatocellular
2012;18:287–​94. carcinoma. Clin Mol Hepatol 2015;21:220–​9.
Lai CH, Lau WY. Gallbladder cancer—​a comprehen- Lee TY, Lee SS, Jung SW et al. Hepatitis B virus infec-
sive review. Surgeon 2008;6:101–​10. tion and intrahepatic cholangiocarcinoma in
Lai GY, Weinstein SJ, Albanes D, Taylor PR, McGlynn Korea:  a case-​control study. Am J Gastroenterol
KA, Virtamo J et  al. The association of coffee 2008;103:1716–​20.
intake with liver cancer incidence and chronic Lee YC, Cohet C, Yang YC, Stayner L, Hashibe M,
liver disease mortality in male smokers. Br J Straif K. Meta-​ analysis of epidemiologic stud-
Cancer 2013;109:1344–​51. ies on cigarette smoking and liver cancer. Int J
Lam KC, Yu MC, Leung JWC, Henderson BE. Epidemiol 2009;38:1497e511.
Hepatitis B virus and cigarette smoking: risk fac- Lewis R, Rempala G. A case-​cohort study of angio-
tors for hepatocellular carcinoma in Hong Kong. sarcoma of the liver and brain cancer at a poly-
Cancer Res 1982;42:5246–​8. mer production plant. J Occup Environ Med
Lambe M, Trichopoulos D, Hsieh C-​C, Ekbom A, 2003;45:538–​45.
Adami H-​O, Pavia M. Parity and cancer of the Li DK, Chung RT. Impact of hepatitis C virus erad-
gall bladder and the extrahepatic bile ducts. Int J ication on hepatocellular carcinogenesis. Cancer
Cancer 1993;54:941–​4. 2015;121:2874–​82.
Lang E, Schäfer M, Schwender H, Neumann NJ, Frank Li M, Li J, Li P, Li H, Su T, Zhu R, Gong J. Hepatitis B
J. Occurrence of malignant tumours in the acute virus infection increases the risk of cholangiocar-
hepatic porphyrias. JIMD Rep 2015;22:17–​22. cinoma: a meta-​analysis and systematic review. J
Larsson SC, Wolk A. Obesity and the risk of gall- Gastroenterol Hepatol 2012;27:1561–​8.
bladder cancer:  a meta-​ analysis Br J Cancer Linet MS, Gridley G, Nyrén O, Mellemkjaer L, Olsen
2007;96:1457–​61. JH, Keehn S et  al. Primary liver cancer, other
Larsson SC, Wolk A. Overweight, obesity and risk of malignancies, and mortality risks following por-
liver cancer: a meta-​analysis of cohort studies. Br J phyria:  a cohort study in Denmark and Sweden.
Cancer 2007;97:1005–​8. Am J Epidemiol 1999;149:1010–​15.
Lavanchy D, McMahon B. Worldwide prevalence and Lim MK, Ju YH, Franceschi S et al. Clonorchis sinen-
prevention of hepatitis C. In: Liang TJ, Hoofnagle sis infection and increasing risk of cholangiocar-
JH (eds.). Hepatitis C. San Diego, CA: Academic cinoma in the Republic of Korea. Am J Trop Med
Press, 2000: 185–​201. Hyg 2006;75:93–​6.
La Vecchia C, Lucchini F, Franceschi S, Negri E, Levi Lin MV, King LY, Chung RT. Hepatitis C virus–​
F. Trends in mortality from primary liver cancer in associated cancer. Annu Rev Pathol Mech Dis
Europe. Eur J Cancer 2000;36:909–​15. 2015;10:345–​70.
La Vecchia C, Negri E, Decarli A, D’Avanzo B, Lindbohm ML, Sallmen M, Kyyronen P, Kauppinen
Franceschi S. Risk factors for hepatocellu- T, Pukkala E. Risk of liver cancer and exposure
lar carcinoma in northern Italy. Int J Cancer to organic solvents and gasoline vapors among
1988;42:872–​6. Finnish workers. Int J Cancer 2009;124:2954–​9.
302

302 Textbook of Cancer Epidemiology

Liu K, Zhang L, Lin X, Chen L, Shi H, Magaye R et al. workers:  synergistic effect of occupational expo-
Association of GST genetic polymorphisms with sure with alcohol intake. Environ Health Perspec
the susceptibility to hepatocellular carcinoma 2004;112:1188–​92.
(HCC) in Chinese population evaluated by an McGlynn KA, London WT. Epidemiology and natural
updated systematic meta-​ analysis. PLoS One history of hepatocellular carcinoma. Best Prac &
2013;8:e57043. Res Clin Gastroenterol 2005;19:3–​23.
Liu S, Zhang H, Gu C et al. Associations between hep- McGlynn KA, Rosvold EA, Lustbader ED, Hu Y,
atitis B virus mutations and the risk of hepatocel- Clapper ML, Zhou T et al. Susceptibility to hepa-
lular carcinoma:  a meta-​analysis. J Natl Cancer tocellular carcinoma is associated with genetic
Inst 2009;101:1066–​82. variation in the enzymatic detoxification of afla-
Liu Y, Chang CC, Marsh GM, Wu F. Population attrib- toxin B1. Proc Natl Acad Sci USA 1995;92:2384–​7.
utable risk of aflatoxin-​related liver cancer:  sys- McGlynn KA, Petrick JL, London WT. Global epide-
tematic review and meta-​analysis. Eur J Cancer miology of hepatocellular carcinoma: an empha-
2012;48:2125e36. sis on demographic and regional variability. Clin
Locarnini S, Hatzakis A, Chen D-​J, Lok A. Strategies Liver Dis 2015;19:223–​38.
to control hepatitis B: public policy, epidemiology, McGlynn KA, Sahasrabuddhe VV, Campbell
vaccine and drugs. J Hepatol 2015;62:S76–​S86. PT, Graubard BI, Chen J, Schwartz LM et  al.
Lööf L, Adami HO, Sparen P, Danielsson A, Eriksson Reproductive factors, exogenous hormone use
LS, Hultcrantz R et al. Cancer risk in primary bili- and risk of hepatocellular carcinoma among US
ary cirrhosis: a population-​based study in Sweden. women:  results from the Liver Cancer Pooling
Hepatology 1994;20:101–​4. Project. Br J Cancer 2015;112:1266–​72.
Loomis D, Guyton KZ, Grosse Y, Lauby-​Secretan B, McGlynn KA, Tsao L, Hsing AW, Devesa SS, Fraumeni
Ghissassi F El, Bouvard V et al. on behalf of the Jr JF. International trends and patterns of primary
International Agency for Research on Cancer liver cancer. Int J Cancer 2001;94:290–​6.
Monograph Working Group. Carcinogenicity of Mohd Hanafiah K, Groeger J, Flaxman AD, Wiersma
drinking coffee, mate, and very hot beverages. ST. Global epidemiology of hepatitis c virus
Lancet Oncol 2016;17:877–​8. infection; new estimates of age-​ specific anti-
Lunn RM, Zhang YJ, Wang LY, Chen CJ, Lee PH, Lee body to HCV and seroprevalence. Hepatology
CS et al. p53 mutations, chronic hepatitis B virus 2013;57:1333–​42.
infection, and aflatoxin exposure in hepatocellular Momosaki S, Hsia CC, Nakashima Y, Kojiro M, Tabor
carcinoma in Taiwan. Cancer Res 1997;57:3471–​7. E. Integration of hepatitis B virus containing
Luo FJ, Luo ZD, Ma L. A study on the relation- mutations in the core promoter/​X gene in patients
ship between drinking water with high arsenic with hepatocellular carcinoma. Dig Liver Dis
content and incidence of malignant tumour in 2003;35:795–​800.
Heihe Village, western part of Huhehot, Inner Momosaki S, Nakashima Y, Kojiro M, Tabor E.
Mongolia. Zhonghua Liu Xing Bing Xue Za Zhi HBsAg-​ negative hepatitis B virus infections in
1995;16:289–​91. hepatitis C virus-​associated hepatocellular carci-
Lv L, Wang P, Zhou X, Sun B. Association between noma. J Viral Hepatitis 2005;12:325–​9.
the p53 codon 72 Arg/​Pro polymorphism and Montella M, D’Arena G, Crispo A, Capunzo M,
hepatocellular carcinoma risk. Tumour Biol Nocerino F, Grimaldi M et  al. Role of sex hor-
2013;34:1451–​9. mones in the development and progression of
Maheshwari S, Sarraj A, Kramer J, El-​Serag HB. Oral hepatitis B virus-​associated hepatocellular carci-
contraception and the risk of hepatocellular carci- noma. Int J Endocrinol 2015;2015:854530.
noma. J Hepatol 2007;47:506e13. Montesano R, Hainaut P, Wild CP. Hepatocellular car-
Manwong M, Songserm N, Promthet S, Matsuo K. cinoma: from gene to public health. J Natl Cancer
Risk factors for cholangiocarcinoma in the lower Inst 1997;89:1844–​51.
part of Northeast Thailand:  a hospital-​ based Morgan RL, Baack B, Smith BD, Yartel A, Pitasi M,
case-​control study. Asian Pac J Cancer Prev Falck-​ Ytter Y. Eradication of hepatitis C virus
2013;1410:5953–​6. infection and the development of hepatocellular
Marrero JA, Fontana RJ, Fu S, Conjeevaram HS, Su carcinoma: a meta-​analysis of observational stud-
GL, Lok AS. Alcohol, tobacco and obesity are syn- ies. Ann Intern Med 2013;158:329–​37.
ergistic risk factors for hepatocellular carcinoma. J Morgan TR, Mandayam S, Jamal MM. Alcohol and
Hepatology 2005;42:218–​24. hepatocellular carcinoma. Gastroenterology
Mastrangelo G, Fedell U, Fadda E, Valentini F, Agnesi 2004;127:S87–​96.
R, Magarotto G et al. Increased risk of hepatocellu- Mori M, Hara M, Wada I, Hara T, Yamamoto K,
lar carcinoma and liver cirrhosis in vinyl chloride Honda M et  al. Prospective study of hepatitis B
30

Cancer of the Liver and Biliary Tract 303

and C viral infections, cigarette smoking, alco- Pandey M, Shukla VK. Diet and gallbladder cancer: a
hol consumption, and other factors associated case control study. Eur J Cancer Prev 2002;11:1–​4.
with hepatocellular carcinoma risk in Japan. Am Parkin DM. The global health burden of infection-​
J Epidemiol 2000;151:131–​9. associated cancers in the year 2002. Int J Cancer
Mori T, Kido C, Fukutomi K, Kato Y, Hatakeyama S, 2006;118:3030–​44.
Machinami R et  al. Summary of entire Japanese Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer
Thorotrast follow-​ up study:  updated 1998. statistics, 2002. CA Cancer J Clin 2005;55:74–​108.
Radiation Res 1999;152:S84–​7. Parkin DM, Srivatanakul P, Khlat M, Chenvidhya D,
Mucci LA, Kuper HE, Tamimi R, Lagiou P, Spanos E, Chotiwan P, Insiripong S et  al. Liver cancer in
Trichopoulos D. Age at menarche and age at men- Thailand. I. A case-​control study of cholangiocar-
opause in relation to hepatocellular carcinoma in cinoma. Int J Cancer 1991;48:323–​8.
women. Br J Obstet Gynaec 2001;108:291–​4. Parr CL, Batty GD, Lam TH, Barzi F, Fang X, Ho SC
Muñoz N, Lingao A, Lao J, Esteve J, Viterbo G, et al. Body mass index and cancer mortality in the
Domingo EO et al. Patterns of familial transmis- Asia-​Pacific Cohort Studies Collaboration: pooled
sion of HBV and the risk of developing liver can- analyses of 424,519 participants. Lancet Oncol
cer: a case-​control study in the Philippines. Int J 2010;11:741–​52.
Cancer 1989;44:981–​4. Persson EC, Schwartz LM, Park Y, Trabert B,
Murakami Y, Saigo K, Takashima H, Minami M, Hollenbeck AR, Graubard BI et  al. Alcohol con-
Okanoue T, Brechot C et  al. Large scaled analy- sumption, folate intake, hepatocellular carcinoma,
sis of hepatitis B virus (HBV) DNA integration and liver disease mortality. Cancer Epidemiol
in HBV related hepatocellular carcinomas. Gut Biomarkers Prev 2013;22:415–​21.
2005;54:1162–​8. Petrick JL, Freedman ND, Graubard BI, Sahasrabuddhe
Nakatani T, Roy G, Fujimoto N, Asahara T, Ito A. VV, Lai GY, Alavanja MC et al. Coffee consump-
Sex hormone dependency of diethylnitrosamine-​ tion and risk of hepatocellular carcinoma and
induced liver tumors in mice and chemo- intrahepatic cholangiocarcinoma by sex:  the
prevention by leuprorelin. Jpn J Cancer Res Liver Cancer Pooling Project. Cancer Epidemiol
2001;92:249–​56. Biomarkers Prev 2015;24:1398–​406.
Namasivayam S, Martin DR, Saini S. Imaging of liver Pham T-​ M, Fujino Y, Ide R, Kubo T, Shirane K,
metastases: MRI. Cancer Imaging 2007;7:2–​9. Tokui N et al. Prospective study of vegetable con-
Negri E, La Vecchia C, Franceschi S, D’Avanzo B, sumption and liver cancer in Japan. Int J Cancer
Parazzini F. Vegetable and fruit consumption and 2006;119:2408–​11.
cancer risk Int J Cancer 1991:48:350–​4. Pollicino T, Squadrito G, Cerenzia G, Cacciola I, Raffa
Nelson DR, Gonzalez-​ Peralta RP, Qian K, Xu G, Craxi A et al. Hepatitis B virus maintains its pro-​
Y, Marousis CG et  al. Transforming growth oncogenic properties in the case of occult HBV
factor-​β1 in chronic hepatitis C. J Viral Hepat infection. Gastroenterology 2004;126:102–​10.
1997;4:29–​35. Poomphakwaen K, Promthet S, Kamsa-​ Ard S,
Neuvet C, Wei Y, Buendia MA. Mechanisms of Vatanasapt P, Chaveepojnkamjorn W, Klaewkla J
HBV-​ related hepatocarcinogenesis. J Hepatol et al. Risk factors for cholangiocarcinoma in Khon
2010;52:594–​604. Kaen, Thailand: a nested case-​control study. Asian
Ohishi W, Fujiwara S, Cologne JB, Suzuki G, Akahoshi Pac J Cancer Prev 2009;10:251–​58.
M, Nishi N et  al. Risk factors for hepatocellular Poynard T, Bedossa P, Opolon P. Natural history of
carcinoma in a Japanese population: a nested case-​ liver fibrosis progression in patients with chronic
control study. Cancer Epidemiol Biomarkers Prev hepatitis C. Lancet 1997;349:825–​32.
2008;17:846–​54. Prospective Studies Collaboration, Whitlock G,
Okuda K. Hepatitis C virus and hepatocellular carci- Lewington S, Sherliker P, Clarke R, Emberson
noma. In: Okuda K, Tabor E (eds.). Liver Cancer. J, Halsey J et  al. Body mass index and cause-​
New York, Churchill Livingstone, 1997: 39–​50. specific mortality in 900 000 adults:  collabora-
Ong JP, Pitts A, Younossi ZM. Increased overall mor- tive analyses of 57 prospective studies. Lancet
tality and liver-​related mortality in non-​alcoholic 2009;373:1083–​96.
fatty liver disease. J Hepatol 2008;49:608–​12. Qian GS, Ross RK, Yu MC, Yuan JM, Gao YT,
Ott JJ, Stevens GA, Groeger J, Wiersma ST. Global epi- Henderson BE et al. A follow-​up study of urinary
demiology of hepatitis B virus infection: new esti- markers of aflatoxin exposure and liver cancer risk
mates of age-​specific HBsAg seroprevalence and in Shanghai, People’s Republic of China. Cancer
endemicity. Vaccine 2012;30:2212–​19. Epidemiol Biomarkers Prev 1994;3:3–​10.
Ozturk M. Genetic aspects of hepatocellular carcino- Qin H, Liu B, Shi T, Liu Y, Sun Y, Ma Y. Tumour
genesis. Semin Liver Dis 1999;19:235–​42. necrosis factor-​ alpha polymorphisms and
304

304 Textbook of Cancer Epidemiology

hepatocellular carcinoma:  a meta-​analysis. J Int among white and black United States veterans.
Med Res 2010;38:760–​8. Cancer Causes Control 2004;15:35–​43.
Qu C, Chen T, Fan C, Zhan Q, Wang Y, Lu J et  al. Sardh E, Wahlin S, Björnstedt M, Harper P, Andersson
Efficacy of neonatal HBV vaccination on liver DE. High risk of primary liver cancer in a cohort
cancer and other liver diseases over 30-​ year of 179 patients with Acute Hepatic Porphyria. J
follow-​up of the Qidong hepatitis B intervention Inherit Metab Dis 2013;36:1063–​71.
study:  a cluster randomized control trial. PLoS Sauvaget C, Nagano J, Hayashi M, Spencer E, Shimizu
Med 2014;11:e1001774. Y, Allen N. Vegetables and fruit intake and cancer
Qu LS, Jin F, Guo YM, Liu TT, Xue RY, Huang XW mortality in the Hiroshima ⁄ Nagasaki Life Span
et al. Nine susceptibility loci for hepatitis B virus-​ Study. Br J Cancer 2003;88:689–​94.
related hepatocellular carcinoma identified by a Sawada N, Inoue M, Iwasaki M, Sasazuki S, Shimazu
pilot two-​stage genome-​wide association study. T, Yamaji T et al. Consumption of n-​3 fatty acids
Oncol Lett 2016;11:624–​32. and fish reduces risk of hepatocellular carcinoma.
Raimondi S, Bruno S, Mondelli MU, Maisonneuve P. Gastroenterology 2012;142:1468–​75.
Hepatitis C virus genotype 1b as a risk factor for Schneider-​Yin X, Harms J, Minder EI. Porphyria in
hepatocellular carcinoma development:  a meta-​ Switzerland, 15 years experience. Swiss Med Week
analysis. J Hepatol 2009;50:1142–​54. 2009;139:198–​206.
Randi G, Franceschi S, La Vecchia C. Gallbladder can- Schütze M, Boeing H, Pischon T, Rehm J, Kehoe T,
cer worldwide: geographical distribution and risk Gmel G et al. Alcohol attributable burden of inci-
factors. Int J Cancer 2006;118:1591–​602. dence of cancer in eight European countries based
Ratziu V, Bonyhay L, Di Martino V, Charlotte F, on results from prospective cohort study. BMJ
Cavallaro L, Sayegh-​Tainturier MH et al. Survival, 2011;342:d1584.
liver failure, and hepatocellular carcinoma in Setiawan VW, Hernandez BY, Lu SC, Stram DO,
obesity-​related cryptogenic cirrhosis. Hepatology Wilkens LR, Le Marchand L et  al. Diabetes and
2002;35:1485–​93. racial/​ethnic differences in hepatocellular carci-
Razumilava N, Gores GJ. Cholangiocarcinoma. noma risk: the multiethnic cohort. J Natl Cancer
Lancet 2014;383:2168–​79. Inst 2014;106:12.
Ray RB, Steele R, Meyer K, Ray R. Transcriptional Setiawan VW, Wilkens LR, Lu SC, Hernandez BY, Le
repression of p53 promoter by hepatitis C virus Marchand L, Henderson BE. Association of coffee
core protein. J Biol Chem 1997;272:10983–​6. intake with reduced incidence of liver cancer and
Ross RK, Yuan JM, Yu MC, Wogan GN, Qian death from chronic liver disease in the US multi-
GS, Tu JT et  al. Urinary aflatoxin biomarkers ethnic cohort. Gastroenterology 2015;148:118–​25.
and risk of hepatocellular carcinoma. Lancet Shaffer EA. Gallbladder cancer:  the basics.
1992;339:943–​6. Gastroenterol Hepatol 2008;4:737–​41.
Rui R, Lou J, Zou L, Zhong R, Wang J, Xia D et  al. Sherman M. Vinyl chloride and the liver. J Hepatol
Excess body mass index and risk of liver cancer: a 2009;51:1074–​81.
nonlinear dose-​ response meta-​ analysis of pro- Shi J, Zhu L, Liu S, Xie W-​F. A meta-​analysis of
spective studies. PLoS One 2012;7:e44522. case–​control studies on the combined effect of
Saab S, Mallam D, Cox GA 2nd, Tong MJ. Impact of hepatitis B and C virus infections in causing
coffee on liver diseases: a systematic review. Liver hepatocellular carcinoma in China. Br J Cancer
Int 2014;34:495–​504. 2005;92:607–​12.
Sakoda LC, Graubard BI, Evans AA, London WT, Lin Shi L, Feng Y, Lin H, Ma R, Cai X. Role of estrogen
W-​Y, Shen F-​M et  al. Toenail selenium and risk in hepatocellular carcinoma: is inflammation the
of hepatocellular carcinoma mortality in Haimen key? J Transl Med 2014;12:93.
City, China. Int J Cancer 2005;115:618–​24. Shimada M, Hashimoto E, Taniai M, Hasegawa K,
Sanyal A, Poklepovic A, Moyneur E, Barghout V. Okuda H, Hayashi N et  al. Hepatocellular carci-
Population-​based risk factors and resource utiliza- noma in patients with non-​alcoholic steatohepati-
tion for HCC: US perspective. Curr Med Res Opin tis. J Hepatol 2002;37:154–​60.
2010;26:2183–​91. Shimazu T, Saazuki S, Wakai K, Tamakoshi A, Tsuji I,
Sanyal AJ, Banas C, Sargeant C, Luketic VA, Sterling Sugawara Y et  al. Alcohol drinking and primary
RK, Stravitz RT et  al. Similarities and differ- liver cancer:  a pooled analysis of four Japanese
ences in outcomes of cirrhosis due to nonalco- cohort studies. Int J Cancer 2012;130:2645–​53.
holic steatohepatitis and hepatitis C. Hepatology Shin HR, Lee CU, Park HJ et  al. Hepatitis B and C
2006;43:682–​9. virus, Clonorchis sinensis for the risk of liver can-
Samanic C, Gridley G, Chow W-​H, Lubin J, Hoover cer:  a case-​control study in Pusan, Korea. Int J
RN, Fraumeni Jr JF. Obesity and cancer risk Epidemiol 1996;25:933–​40.
305

Cancer of the Liver and Biliary Tract 305

Shiraha H, Yamamoto K, Namba M. Human hep- Sun Z, Chen T, Thorgeirsson SS, Zhan Q, Chen J, Park
atocyte carcinogenesis (review). Int J Oncol JH et al. Dramatic reduction of liver cancer inci-
2013;42:1133–​8. dence in young adults: 28 year follow-​up of etio-
Shiratori Y, Shiina S, Imamura M, Kato N, Kanai F, logical interventions in an endemic area of China.
Okudaira T et  al. Characteristic difference of Carcinogenesis 2013;34:1800–​05.
hepatocellular carcinoma between hepatitis Sun Z, Lu P, Gail MH, Pee D, Zhang Q, Ming L et al.
B-​and C-​ viral infection in Japan. Hepatology Increased risk of hepatocellular carcinoma in male
1995;22:1027–​33. hepatitis B surface antigen carriers with chronic
Shirota Y, Kaneko S, Honda M, Kawai HF, Kobayashi hepatitis who have detectable urinary aflatoxin
K. Identification of differentially expressed genes metabolite M1. Hepatology 1999;30:379–​83.
in hepatocellular carcinoma with cDNA microar- Sung J-​ L. Prevention of hepatitis B and C virus
rays. Hepatology 2001;33:832–​40. infection for prevention of cirrhosis and hepa-
Sorensen HT, Mellemkjaer L, Jepsen P, Thulstrup AM, tocellular carcinoma. J Gastroenterol Hepatol
Baron J, Olsen JH et al. Risk of cancer in patients 1997;12(Suppl):S370–​6.
hospitalized with fatty liver:  a Danish cohort Tabor E. Tumor suppressor genes, growth fac-
study. J Clin Gastroenterol 2003;36:356–​9. tor genes, and oncogenes in hepatitis B virus–​
Sparos L, Tountas Y, Chapuis-​Cellier C, Theodoropoulos associated hepatocellular carcinoma. J Med Virol
G, Trichopoulos D. Alpha1-​antitrypsin levels and 1994;42:357–​65.
phenotypes and hepatitis B serology in liver can- Tabor E. The role of tumor suppressor genes in
cer. Br J Cancer 1984;49:567–​70. the development of hepatocellular carcinoma.
Smith-​Palmer J, Cerri K, Valentine W. Achieving sus- In:  Okuda K, Tabor E (eds.). Liver Cancer.
tained virologic response in hepatitis C: a system- New York, Churchill Livingstone, 1997: 89–​95.
atic review of the clinical, economic and quality of Tabor E. Viral hepatitis and liver cancer. In:  Goldin
life benefits. BMC Infect Dis 2015;15:1–​19. RD, Thomas HC, Gerber MA (eds.) Pathology of
Soliman AS, Hung CW, Tsodikov A, Seifeldin IA, Viral Hepatitis. London: Arnold, 1998: 161–​77.
Ramadan M, Al-​Gamal D et  al. Epidemiologic Tabor E. Global epidemiology of hepatocellular car-
risk factors of hepatocellular carcinoma in a rural cinoma associated with hepatitis B and C virus
region of Egypt. Hepatol Int 2010;4:681–​90. infections. In:  Margolis HS, Alter MJ, Liang
Sriamporn S, Pisani P, Pipitgool V et  al. Prevalence TJ, Dienstag JL (eds.) Viral Hepatitis Liver
of Opisthorchis viverrini infection and incidence Disease. Atlanta:  International Medical Press,
of cholangiocarcinoma in Khon Kaen, Northeast 2002: 400–​03.
Thailand. Trop Med Int Health 2004;9:588–​94. Tagger A, Donato F, Ribero ML, Chiesa R, Portera G,
Srivatanakul P, Parkin DM, Khlat M, Chenvidhya D, Gelatti U et al. Case-​control study on hepatitis C
Chotiwan P, Insiripong S et  al. Liver cancer in virus (HCV) as a risk factor for hepatocellular car-
Thailand. II. A case-​control study of hepatocellu- cinoma: the role of HCV genotypes and the syn-
lar carcinoma. Int J Cancer 1991;48:329–​32. ergism with hepatitis B virus and alcohol. Brescia
Stanford J, Thomas D, the WHO Collaborative Study of HCC Study. Int J Cancer 1999;81:695–​9.
Neoplasia and Steroid Contraceptives. Reproductive Takahashi M, Yamada G, Miyamoto R, Doi, T, Endo
factors in the etiology of hepatocellular carcinoma. H, Tsuji T. Natural course of chronic hepatitis C.
Cancer Causes Control 1992;2:37–​42. Am J Gastroenterol 1993;88:240–​3.
Stern MC, Umbach DM, Yu MC, London SJ, Zhang Talamini R, Polesel J, Montella M, Dal Maso L, Crispo
Z-​Q, Taylor JA. Hepatitis B, aflatoxin B1, and p53 A, Tommasi LG et al. Food groups and risk of hepa-
codon 249 mutation in hepatocellular carcino- tocellular carcinoma:  a multicenter case-​control
mas from Guangxi, People’s Republic of China, study in Italy. Int J Cancer 2006;119:2916–​21.
and a meta-​analysis of existing studies. Cancer Tanaka K, Ikematsu H, Hirohata T, Kashiwagi S.
Epidemiol Biomarkers Prev 2001;10:617–​25. Hepatitis C virus infection and risk of hepato-
Stuver SO, Trichopoulos D. Liver cancer. In:  Doll cellular carcinoma among Japanese:  possible
R, Fraumeni J, Muir C (Eds.) Trends in Cancer role of type 1b (II) infection. J Natl Cancer Inst
Incidence and Mortality. Cancer Surveys, Vol. 1996;88:742–​6.
19/​20. London:  Imperial Cancer Research Fund, Tanaka K, Stuver SO, Ikematsu H, Okayama A,
1994: 99–​124. Tachibana N, Hirohata T et  al. Heterosexual
Sun C-​A, Wu D-​M, Lin C-​C, Lu S-​N, You S-​L, Wang transmission of hepatitis C virus among mar-
L-​Y et  al. Incidence and cofactors in hepatitis C ried couples in southwestern Japan. Int J Cancer
virus-​related hepatocellular carcinoma:  a pro- 1997;72:50–​5.
spective study of 12,008 men in Taiwan. Am J Tanaka M, Tanaka H, Tsukuma H, Ioka A, Oshima
Epidemiol 2003;157:674–​82. A, Nakahara T. Risk factors for intrahepatic
306

306 Textbook of Cancer Epidemiology

cholangiocarcinoma: a possible role of hepatitis B etiology of hepatocellular carcinoma. Int J Cancer

virus. J Viral Hepat 2010;17:742–​8. 1987;39:45–​9.
Tanaka S, Takenaka K, Matsumata T, Mori R, Trichopoulos D, MacMahon B, Sparros L, Merikas
Sugimachi K. Hepatitis C virus replication is asso- G. Smoking and hepatitis B–​ negative primary
ciated with expression of transforming growth hepatocellular carcinoma. J Natl Cancer Inst
factor-​alpha and insulin-​like growth factor-​II in 1980;65:111–​14.
cirrhotic livers. Dig Dis Science 1996;41:208–​15. Tsai SM, Wang TN, Ko YC. Mortality for certain dis-
Tang TC-​M, Poon RT-​P, Fan S-​T. The significance of eases in areas with high levels of arsenic in drink-
cyclooxygenase-​2 expression in human hepatocel- ing water. Arch Environ Health 1999;54:186–​93.
lular carcinoma. Biomed Pharmacothera 2005;59 Tsuda T, Babazono A, Yamamoto E, Kurumatani N,
Suppl 2:S311–​6. Mino Y, Ogawa T et al. Ingested arsenic and inter-
Tavakolpour S, Sali S. Tumor Necrosis Factor-​α-​308 nal cancer: a historical cohort study followed for
G/​A Polymorphisms and Risk of Hepatocellular 33 years. Am J Epidemiol 1995;141:198–​209.
Carcinoma:  A Meta-​ Analysis. Hepat Mon Turati F, Galeone C, Rota M, Pelucchi C, Negri E,
2016;16:e33537. Published online 2016 Apr 12. Bagnardi V et al. Alcohol and liver cancer: a sys-
doi:10.5812/​hepatmon.33537 tematic review and meta-​analysis of prospective
Tennant BC, Toshkov IA, Peek SF, Jacob JR, Menne S, studies. Ann Oncol 2014;25:1526–​35.
Hornbuckle WE et  al. Hepatocellular carcinoma Tzonou A, Zavitsanos X, Hsieh CC, Trichopoulos D.
in the woodchuck model of hepatitis B virus infec- Liveborn children and risk of hepatic-​cellular car-
tion. Gastroenterol 2004;127:S283–​93. cinoma. Cancer Causes Control 1992;3:171–​4.
The Collaborative MILTS Project Team. Oral contracep- Uccello M, Malaguarnera G, Corriere T, Biondi A,
tives and liver cancer. Contraception 1997;56:275–​84. Basile F, Malaguarnera M. Risk of hepatocellu-
Thomas DL. Hepatitis C and human immunodefi- lar carcinoma in workers exposed to chemicals.
ciency virus infection. Hepatology 2002;36 Suppl Hepat Mon 2012;12:e5943.
1:S201–​9. Vall Mayans M, Calvet X, Bruix J, Bruguera M,
Thompson DE, Mabuchi K, Ron E, Soda M, Tokunaga Costa J, Esteve J et al. Risk factors for hepatocel-
M, Ochikubo S et al. Cancer incidence in atomic lular carcinoma in Catalonia, Spain. Int J Cancer
bomb survivors. Part II. Solid tumors, 1958–​1987. 1990;46:378–​81.
Radiation Res 1994;137:S17–​67. Vinceti M, Dennert G, Crespi CM, Zwahlen M,
Thorgeirsson SS, Grisham JW. Molecular pathogen- Brinkman M, Zeegers MP et  al. Selenium for
esis of human hepatocellular carcinoma. Nature preventing cancer. Cochrane Database Syst Rev
Genetics 2002;31:339–​46. 2014;3:CD005195
Tobler LH, Busch MP. History of posttransfusion hep- Vineis P, Alavanja M, Buffler P, Fontham E, Franceschi
atitis. Clin Chem 1997;43:1487–​93. S, Gao YT et  al. Tobacco and cancer:  recent
Tornesello ML, Buonaguro L, Tatangelo F, Botti epidemiological evidence. J Natl Cancer Inst
G, Izzo F, Buonaguro FM. Mutations in TP53, 2004;96:99–​106.
CTNNB1 and PIK3CA genes in hepatocellular Wang JT, Wang TH, Lin JT, Lee CZ, Sheu JC, Chen DS.
carcinoma associated with hepatitis B and hepa- Effect of hepatitis C antibody screening in blood
titis C virus infections. Genomics 2013;102:74–​83. donors on post-​transfusion hepatitis in Taiwan. J
Travis LB, Hauptmann M, Gaul LK, Storm HH, Gastroenterol Hepatol 1995;10:454–​8.
Goldman MB, Nyberg U et  al. Site-​specific can- Wang L-​Y, Hatch M, Chen C-​J, Levin B, You SL, Lu SN
cer incidence and mortality after cerebral angiog- et al. Aflatoxin exposure and risk of hepatocellular
raphy with radioactive Thorotrast. Radiation Res carcinoma in Taiwan. Int J Cancer 1996;67:620–​5.
2003;160:691–​706. Wang L-​Y, You S-​L, Lu S-​N, Ho H-​C, Wu M-​H, Sun
Trichopoulos D. Etiology of primary liver cancer and C-​A et  al. Risk of hepatocellular carcinoma and
the role of steroidal hormones. Cancer Causes habits of alcohol drinking, betel quid chewing
Control 1992;3:3–​5. and cigarette smoking; a cohort of 2416 HBsAg-​
Trichopoulos D, Bamia C, Lagiou P, Fedirko V, seropositive and 9421 HBsAg-​seronegative male
Trepo E, Jenab M et al. Hepatocellular carcinoma residents in Taiwan. Cancer Causes and Control
risk factors and disease burden in a European 2003;14:241–​50.
cohort: a nested case control study. J Natl Cancer Wang Y, Wang B, Shen F, Fan J, Cao H. Body mass index
Inst 2011;103:1686–​95. and risk of primary liver cancer: a meta-​analysis of
Trichopoulos D, Day NE, Kaklamani E, Tzonou A, prospective studies. Oncologist 2012;17:1461–​8.
Munoz N, Zavitsanos X et  al. Hepatitis B virus, Wang Y, Wang P, Wang B, Fu Z, Zhao W, Yan S.
tobacco smoking and ethanol consumption in the Diabetes mellitus and poorer prognosis in
307

Cancer of the Liver and Biliary Tract 307

hepatocellular carcinoma: a systematic review and World Health Organization. Immunization coverage.
meta-​analysis. PLoS One 2014;9:e95485. Fact sheet N°378. http://​www.who.int/​mediacen-
Wartenberg D, Reyner D, Scott CS. Trichloroethylene tre/​factsheets/​fs378/​en/​. September 2015.
and cancer:  epidemiologic evidence. Environ Wu CH, Chan TF, Changchien CC, Yang CY. Parity,
Health Perspect 2000;108:161–​76. age at first birth, and risk of death from liver
Wartenberg D. Environmental factors in cancer:  tri- cancer:  evidence from a cohort in Taiwan. J
chloroethylene and related solvents:  science, Gastroenterol Hepatol 2011;26:334–​9.
regulation, and cancer prevention. Rev Environ. Xia J, Jiang S-​C, Peng H-​J. Association between liver
Health 2009;24 297–​302. fluke infection and hepatobiliary pathological
Weihrauch M, Lehnert G, Köckerling F, Wittekind C, changes:  a systematic review and meta-​analysis.
Tannapfel A. p53 mutation pattern in hepatocel- PLoS One 2015;10:1–​19.
lular carcinoma in workers exposed to vinyl chlor- Yang H-​ I, Lu S-​ N, Liaw Y-​ F, You S-​
L, Sun C-​ A,
ide. Cancer 2000;88:1030–​6. Wang L-​Y et  al. Hepatitis B e antigen and the
Weng CJ, Hsieh YH, Tsai CM, Chu YH, Ueng KC, risk of hepatocellular carcinoma. N Engl J Med
Liu YF et al. Relationship of insulin-​like growth 2002;347:168–​74.
factors system gene polymorphisms with the Yang HI, Yeh SH, Chen PJ et  al. REVEAL-​HBV
susceptibility and pathological development Study Group. Associations between hepati-
of hepatocellular carcinoma. Ann Surg Oncol tis B virus genotype and mutants and the risk
2010;17: 1808–​15. of hepatocellular carcinoma. J Natl Cancer Inst
White DL, Li D, Nurgalieva Z, El-​Serag HB. Genetic 2008;100:1134–​43.
variants of glutathione S-​ transferase as possi- Yang WS, Va P, Bray F, Gao S, Gao J, Li HL et al. The
ble risk factors for hepatocellular carcinoma:  a role of pre-​existing diabetes mellitus on hepato-
HuGE systematic review and meta-​analysis. Am J cellular carcinoma occurrence and prognosis:  a
Epidemiol 2008;167:377–​89. meta-​analysis of prospective cohort studies. PLoS
WHO Collaborative Study of Neoplasia and Steroid One 2011;6:e27326
Contraceptives. Combined oral contracep- Yatsuji S, Hashimoto E, Tobari M, Taniai M, Tokushige
tives and gall bladder cancer. Int J Epidemiol K, Shiratori K. Clinical features and outcomes
1989a;18:309–​14. of cirrhosis due to non-​alcoholic steatohepatitis
WHO Collaborative Study of Neoplasia and Steroid compared with cirrhosis caused by chronic hep-
Contraceptives. Combined oral contraceptives atitis C. J Gastroenterol Hepatol 2009;24:248–​54.
and liver cancer. Int J Cancer 1989b;43:254–​9. Yeh SH, Chen PJ. Gender disparity of hepatocellular
Williams JH, Phillips TD, Jolly PE, Stiles JK, Jolly CM, carcinoma: the roles of sex hormones. Oncology
Aggarwal D. Human aflatoxicosis in developing 2010;78 Suppl 1:172–​9.
countries: a review of toxicology, exposure, poten- Yu MC, Tong MJ, Govindarajan S, Henderson BE.
tial health consequences, and interventions Am J Nonviral risk factors for hepatocellular carcinoma
Clin Nutr 2004:80;1106–​22. in a low-​risk population, the non-​Asians of Los
Wolk A, Gridley G, Svensson M, Nyren O, McLaughlin Angeles County, California. J Natl Cancer Inst
JK, Fraumeni Jr, JF et  al. A prospective study of 1991;83:1820–​6.
obesity and cancer risk (Sweden). Cancer Causes Yu M-​W, Chang H-​C, Chang S-​C, Liaw Y-​F, Lin S-​M,
Control 2001;12:13–​21. Liu C-​J et al. Role of reproductive factors in hepa-
World Cancer Research Fund/​American Institute for tocellular carcinoma:  impact on hepatitis B-​and
Cancer Research (WCRF/​AICR). Food, Nutrition C-​related risk. Hepatology 2003;38:1393–​1400.
and the Prevention of Cancer: A Global Perspective. Yu M-​W, Chen C-​J. Elevated serum testosterone levels
Washington, DC:  American Institute for Cancer and risk of hepatocellular carcinoma. Cancer Res
Research. 1997. 1993;53:790–​4.
World Cancer Research Fund/​American Institute for Yu M-​W, Cheng S-​W, Lin M-​W, Yang S-​Y, Liaw, Y-​F,
Cancer Research (WCRF/​AICR). Full Continuous Chang H-​C et  al. Androgen-​receptor gene CAG
Update Project Liver Cancer SLR 2014. wcrf.org/​ repeats, plasma testosterone levels, and risk of
sites/​default/​files/​Liver-​Cancer-​SLR-​2014.pdf hepatitis B-​ related hepatocellular carcinoma. J
World Cancer Research Fund International/​ Natl Cancer Inst 2000;92:2023–​8.
American Institute for Cancer Research (WCRF/​ Yu M-​ W, Gladek-​ Yarborough A, Chiamprasert S,
AICR). Continuous Update Project Report: Diet, Santella RM, Liaw Y-​F, Chen C-​J. Cytochrome
Nutrition, Physical Activity and Liver Cancer. P450 2E1 and S-​transferase M1 polymorphisms
2015. Available at:  wcrf.org/​sites/​default/​files/​ and susceptibility to hepatocellular carcinoma.
Liver-​Cancer-​2015-​Report.pdf. Gastroenterology 1995a;109:1266–​73.
308

308 Textbook of Cancer Epidemiology

Yu M-​W, Homg I-​S, Hsu K-​H, Chiang Y-​C, Liaw cancer risk in men: national health insurance cor-
Y-​F, Chen C-​J. Plasma selenium levels and risk poration study. BMC Cancer 2008;8:366.
of hepatocellular carcinoma among men with Zatonski WA, Lowenfels AB, Boyle P, Maisonneuve
chronic hepatitis virus infection. Am J Epidemiol P, Bueno de Mesquita HB, Ghadirian P et  al.
1999b;150:367–​74. Epidemiologic aspects of gallbladder cancer:  a
Yu M-​W, Hsieh HH, Pan WH, Yang CS, Chen CJ. case-​control study of the SEARCH Program of the
Vegetable consumption, serum retinol level, and International Agency for Research on Cancer. J
risk of hepatocellular carcinoma. Cancer Res Natl Cancer Inst 1997;89:1132–​8.
1995b;55:1301–​5. Zhang H, Zhai Y, Hu Z, Wu C, Qian J, Jia W. Genome-​
Yu M-​W, Yeh S-​H, Chen P-​J, Liaw Y-​F, Lin C-​L, wide association study identifies 1p36.22 as a new
Liu C-​J et  al. Hepatitis B virus genotype and susceptibility locus for hepatocellular carcinoma
DNA level and hepatocellular carcinoma:  a in chronic hepatitis B virus carriers. Nat Genet
prospective study in men. J Natl Cancer Inst 2010;42:755–​758.
2005;97:265–​72. Zhang W, Xiang Y-​B, Li H-​L, Yang G, Cai H, Ji B-​T et al.
Yuan J-​M, Gao Y-​T, Ong C-​N, Ross RK, Yu MC. Vegetable-​based dietary pattern and liver cancer
Prediagnostic level of serum retinol in relation to risk: Results from the Shanghai Women’s and Men’s
reduced risk of hepatocellular carcinoma. J Natl Health Studies. Cancer Sci 2013;104:1353–​61.
Cancer Inst 2006;98:482–​90. Zhang Y, Ren JS, Shi JF, Li N, Wang YT, Qu C, Zhang
Yuan J-​M, Govindarajan S, Arakawa K, Yu MC. Y, Dai M. International trends in primary liver
Synergism of alcohol, diabetes, and viral hepatitis cancer incidence from 1973 to 2007. BMC Cancer
on the risk of hepatocellular carcinoma in blacks 2015;15:94.
and whites in the U.S. Cancer 2004;101:1009–​17. Zhong JH, Xiang BD, Ma L, You XM, Li LQ, Xie GS.
Yuan J-​M, Ross RK, Stanczyk FZ, Govindarajan S, Meta-​analysis of microsomal epoxide hydrolase
Gao Y-​T, Henderson BE et  al. A cohort study of gene polymorphism and risk of hepatocellular
serum testosterone and hepatocellular carcinoma carcinoma. PLoS One 2013;8:e57064.
in Shanghai, China. Int J Cancer 1995;63:491–​3. Zhou H-​B, Hu J-​Y, Hu H-​P. Hepatitis B virus infec-
Yun YH, Lim MK, Won YJ, Park SM, Chang YJ, Oh tion and intrahepatic cholangiocarcinoma. World
SW et al. Dietary preference, physical activity and J Gastroenterol 2014;20:5721–​9.
309

13
Pancreatic Cancer
VA S S I L I K I B E N E T O U , A N D E R S E K B O M , A N D L O R E L E I   M U C C I

P ancreatic cancer is one of the most dreaded

cancer diagnoses, given its high fatality. The
absence of long-​term survivors and the nature
loss, steatorrhea, onset of diabetes, or symptoms
associated with conjugated hyperbilirubinemia,
such as pruritus, are also common (McIntyre &
of the disease process are probably the major Winter, 2015; Ducreux et al, 2015).
reasons for the anonymity surrounding pancre- Diagnosis Multidetector computed tomography
atic cancer. Count Basie (a famous jazz pianist), (CT) scan with three-​dimensional (3D) recon-
Michael Landon and Patrick Swayze (popular struction is the preferred imaging examination,
actors), and the mother and siblings of former which has improved substantially the accuracy
US President Jimmy Carter are among the excep- of CT in both the diagnosis and staging of the
tional few who have given faces to this malig- disease (McIntyre & Winter, 2015; Singh &
nancy. There is a similar anonymity in the world Siddiqui, 2015). Endoscopic ultrasonography
literature, although the novel The Death of a Bee (EUS) usually follows the CT scan if the results of
Keeper by the Swedish author Lars Gustafsson the latter are negative or inconclusive. Pancreatic
is a notable exception. In spite of this, primary protocol magnetic resonance imaging (MRI) or
prevention could contribute substantially in magnetic resonance cholangiopancreatography
reducing pancreatic cancer incidence and thus (MRCP) are also used, especially in the case of
mortality, since two-​thirds of its known risk fac- cystic neoplasms and for evaluating the biliary
tors are potentially modifiable. Secondary pre- anatomy. Endoscopic retrograde cholangiopan-
vention for the general population is neither creatography (ERCP) has little diagnostic value
possible nor recommended yet, while efficacy and more often is used as a palliative measure to
of treatment, although improvements have been relieve biliary obstruction and jaundice by stent-
achieved, is still poor. ing. A histopathologic diagnosis can be estab-
lished by a real time EUS-​guided fine-​needle
CLINICAL SYNOPSIS aspiration.
Subgroups Ninety-​ five percent of all pancre- Treatment Treatment differs by stage of the cancer
atic cancers develop in the exocrine pancreas, (Ducreux et al, 2015). Stage II pancreatic cancers
and most of them are ductal adenocarcinomas. and I are treated by surgery and adjuvant chemo-
About two-​thirds are located in the head of the therapy (with or without chemoradiation), while
pancreas. The endocrine tumors (around 5%) more advanced stages are not resectable and
that arise from the acinar cells are not dealt with treated only by chemotherapy (with or without
in this chapter. chemoradiation). Improvements in surgical tech-
Symptoms and Signs Pancreatic cancer has an niques and peri-​and postoperative management
insidious onset and often remains undiag- during the last years have led to better postoper-
nosed until prominent clinical symptoms and ative mortality.
signs appear abruptly. An estimated 50% of the Progress and Prognosis Prognosis depends on
patients have evidence of distant disease at the the stage of the tumor and whether it can be
time of diagnosis (Siegel et  al, 2015). The pre- removed by surgery. Unfortunately, only 9%
senting symptom is often jaundice caused by of cases are localized at the time of diagnosis,
blockage of the common bile duct. Abdominal with a 5-​year survival at the localized stage of
pain, nausea and/​or vomiting, anorexia, weight 27%. The respective 5-​year survival is 11% when
310

310 Textbook of Cancer Epidemiology

the tumor has spread regionally and 2% when with the worst prognosis (mortality:incidence
a distant metastasis has occurred (SEER, 2015). ratio is 0.98). Pancreatic cancer is the seventh
The median survival following diagnosis is less most common cause of death from cancer in
than 6  months. Serum carbohydrate antigen both sexes worldwide, and in the United States
19-​9 (CA 19-​9) levels can be used as prognos- is the third most common cause of death from
tic factor for measuring the disease burden and cancer but only the 12th most common cancer
potentially guide treatment decisions. A preop- diagnosis (SEER, 2015).
erative serum CA 19-​9 level ≥500 UI/​mL indi- Although a decrease in incidence rates has
cates a worse prognosis after surgery (Ducreux been observed in the United States after 1994,
et al, 2015). especially among African American men, an
average increase of 0.8% annually has been
DESCRIPTIVE noted over the years 2000–​2012 (SEER, 2015).
EPIDEMIOLOGY Pancreatic cancer is a disease of the elderly,
Pancreatic cancer accounted for 337,872 new most commonly diagnosed between age 65 and
cases in 2012, and worldwide ranked 12th among 75 years. Considering the rise of life expectancy
incident cancers in both sexes combined (Ferlay and the growing numbers of elderly popula-
et al, 2015). Incidence of pancreatic cancer exhib- tion, especially in the high-​income countries, an
its modest geographic differences (Figs. 13-​ 1, increase in the total burden of disease is expected
13-​2), while most cases occur in high-​income in the future (Maisonneuve & Lowenfels, 2010;
countries. These differences, which may high- Yeo, 2015). The age-​specific incidence rates are
light the influence of environmental and lifestyle similar in different populations (Figs. 13-​3, 13-​4),
factors in pancreatic cancer etiology, could partly although in some the incidence is decreasing in
be attributed to underdiagnosis of the disease the highest age groups. This decrease has been
(particularly among the elderly) in some regions attributed to a lower prevalence of ever-​smokers
of the world. The male/​female incidence ratio is in the oldest age groups or lower diagnostic
close to 1, slightly more common in men than intensity among elderly persons who develop
in women, with incidence rates varying between a silent jaundice with weight loss (Warshaw &
7 to 9/​100,000 in men and 5 to 6.5/​100,000 in Fernandez del Castillo, 1992). As the median
women, with lower rates in the less economically survival is less than 6 months—​and even less in
developed regions (Ferlay et al, 2015). Incidence the age group over 80—​autopsy frequency is an
also varies by race, with African American indi- important determinant of the reported incidence
viduals having highest rates. It is the malignancy among the elderly (Lindstrom et al, 1997).

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 1 2 3 4 5 6 7 8
ASR (world) per 100,000 person-years

FIGURE 13-​1  Age-​standardized (to the 2012 world population) incidence rates of pancreatic cancer among women.
Source: Ferlay et al, 2013.
31

Pancreatic Cancer 311

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America

0 2 4 6 8 10 12
ASR (world) per 100,000 person-years

FIGURE 13-​2  Age-​standardized (to the 2012 world population) incidence rates of pancreatic cancer among men.
Source: Ferlay et al, 2013.

GENETIC AND MOLECULAR with a substantially increased risk of the disease

EPIDEMIOLOGY (Ghadirian et  al, 1991; Fernandez et  al, 1994),
More than 80% of pancreatic cancer cases are as have cancer susceptibility syndromes with
due to sporadic mutations, and only a small pro- known gene mutations such as hereditary pan-
portion are due to inherited germline mutations. creatitis, hereditary nonpolyposis colon cancer
BRCA2 mutations may be the most common in association with microsatellite instability
inherited genetic alteration in familial pancreatic (Lynch et al, 1991), familial breast cancer with
cancer. Approximately 75%–​90% of pancreatic BRCA2 alterations (Ozcelik et  al, 1997), famil-
cancer cases involve an activating point mutation ial melanoma (Schenk et al, 1998), and ataxia-​
in the KRAS oncogene (Tersmette et  al, 2001; telangiectasia (Swift et al, 1991). A family history
Vincent et al, 2011). of pancreatic cancer, defined as at least two
Several instances of family clusters have first-​degree relatives with pancreatic cancer, is
been reported, the best known being the family associated with an almost 10-​fold increased risk
of former US President Jimmy Carter. A family compared to the general population and is even
history of pancreatic cancer has been associated more pronounced—​ a 30-​ fold excess risk—​ if
New cases per 100,000

120
100
person-years

80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 13-​3  Age-​specific incidence rates of pancreatic cancer among women.

Source: Ferlay et al, 2013.
312

312 Textbook of Cancer Epidemiology

New cases per 100,000

140
120

person-years
100
80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 13-​4  Age-​specific incidence rates of pancreatic cancer among men.

Source: Ferlay et al, 2013.

there are three or more affected first-​degree rel- (Maisonneuve & Lowenfels, 2015) Furthermore,
atives (Klein at al, 2004). Moreover, those with a the fraction of pancreatic cancers attributable to
positive family history are frequently diagnosed tobacco smoking was estimated between 11%
at younger ages, a finding which due to the nat- and 32% (Maisonneuve & Lowenfels, 2015).
ural history of the disease cannot be explained Although the underlying biological mechanisms
by surveillance bias (Petersen et al, 2006). remain elusive, several hypotheses have been
Significant interactions have been reported proposed to explain the association (Duell, 2012;
between cigarette smoking and different geno- Pandol et  al, 2012). Carcinogenic compounds
types, such as polymorphism of DNA repair and metabolites in cigarette smoking, such as
gene XRCC1. This supports the hypothesis that nitrosamines and polycyclic aromatic hydrocar-
individuals who have a deficient carcinogen bons, may act synergistically and probably exert
detoxification and DNA repair capacities are at both somatic genetic and epigenetic effects at
increased risk of pancreatic cancer (Duell et  al, early and later stages of pancreatic carcinogen-
2002). Inflammatory gene polymorphism in esis. For example, 4-​(methylnitrosamino)-​1-​(3-​
combination with inflammatory conditions and pyridyl)-​1-​butanone (NNK) has been implicated
possibly tobacco smoking also seem to increase in smoking-​related pancreatic cancer through the
the risk of pancreatic cancer (Duell et al, 2006). formation of DNA adducts and point mutations
An increased risk of pancreatic cancer has been in genes such as KRAS (Schuller, 2002; Duell,
reported for individuals with A, B, and AB blood 2012). Furthermore, tobacco smoking contrib-
types (non-​ O blood groups) which has been utes to inflammation of the pancreas, leading to
attributed to an ABO single nucleotide polymor- adverse morphological tissue changes that could
phism (rs505922) (Amundadottir et al, 2009). exacerbate already existing tumor cells, as well as
lipid peroxidation oxidative damage through the
influence of free radicals.
RISK FACTORS Smoking cessation seems to lower pancreatic
Tobacco cancer risk relatively rapidly. The risk in former
Smoking is the strongest established risk fac- smokers is similar to that of nonsmokers within
tor for pancreatic cancer (Iodice et  al, 2008; 10–​15  years after cessation (Lin Y et  al, 2002;
Lynch et  al, 2009; Duell, 2012; Maisonneuve & Lynch et al, 2009), although in some reports the
Lowenfels, 2015). In a review of 117 studies in risk reduction was less pronounced (Silverman
meta-​or pooled analyses investigating specific et  al, 1994). The reduction in risk of pancreatic
risk factors and pancreatic cancer, the associa- cancer after smoking cessation suggests that
tion between tobacco smoking and pancreatic tobacco smoke is also a late-​stage component in
cancer was characterized as strong and consistent the carcinogenic process.
with a summary relative risk (RR) between 1.6 Cigars and pipes (noncigarette tobacco smok-
and 2.2 for current smokers and between 1.1 and ing) and smokeless tobacco (e.g., snus, chew,
1.2 for former smokers compared to nonsmokers etc.) have also been investigated as risk factors
31

Pancreatic Cancer 313

for pancreatic cancer. In a pooled analysis of vegetables, such as intake of vitamin C (Fan et al,
11 case-​control studies from the International 2015) and folate (Lin at al, 2013), prediagnostic
Pancreatic Cancer Case-​ control Consortium, plasma concentrations of beta-​ carotene, zeax-
only cigar smoking was associated with pancre- anthin, and alpha-​ tocopherol (Jeurnink et  al,
atic cancer risk with similar magnitude effects as 2015)  and risk of pancreatic cancer have also
cigarette smoking (Bertuccio et al, 2011). On the been reported. A  significant inverse association
other hand, a meta-​analysis of 82 studies found between dietary fiber intake and pancreatic can-
a positive significant association for current pipecer risk was reported in a meta-​analysis of obser-
smoking only (Iodice et al, 2008). Evidence with vational studies among which only one cohort
respect to smokeless tobacco and passive smoking study was available (Wang et al, 2015).
is inconclusive (Bertuccio et al, 2011; Zhou et al, High fructose intake was associated with
2012). Compelling evidence that passive smoke higher risk of pancreatic cancer in a meta-​
exposure in utero or in early life may be associ- analysis (Aune et al, 2012a). Possible mechanisms
ated with pancreatic cancer remains to be con- that could explain this finding are fructose’s
firmed (Bao et al, 2009). involvement in pathways essential for DNA syn-
In conclusion, based on the strength and thesis and cell proliferation and its contribution
consistency of the association as well as on the in increasing postprandial plasma glucose levels
existence of plausible biological mechanisms, the and oxidative stress (WCRF/​AICR, 2012). The
association between tobacco use and pancreatic fourth edition of the European Code against
cancer risk is likely to be causal. Cancer noted some evidence that sugary drinks
may be related to an increased risk of pancreatic
Diet cancer (Norat et al, 2015).
In 2012, the World Cancer Research Fund/​ With respect to specific dietary patterns, a
American Institute for Cancer Research (WCRF/​ pooled analysis of 124,000 participants yielding
AICR) in its Continuous Update Project report 366 cases of pancreatic cancer found no associ-
(CUP) implemented the most extensive review ation between two major dietary patterns, pru-
of existing evidence (until September 2011)  on dent and Western, identified by factor analysis
diet and pancreatic cancer. Findings from this and risk (Michaud et al, 2005). On the contrary,
analysis suggested that there is limited evidence a meta-​analysis of 56 observational studies found
that consumption of red meat, processed meat, that adherence to a Mediterranean diet pattern
foods and beverages containing fructose, and was associated with lower risk of pancreatic can-
foods containing saturated fatty acids are causes cer (Schwingshackl & Hoffmann, 2015).
of pancreatic cancer. Moreover, evidence for a Of note, case-​ control studies investigating
protective effect of fruits and foods containing the association between diet and pancreatic can-
folate is even less consistent and too limited to cer tend to report stronger estimates of effect
draw conclusions (WCRF/​AICR, 2012). The evi- compared to cohort studies and pooled analyses
dence was also limited and not conclusive for (Paluszkiewicz et al, 2012). This general lack of
total fat, polyunsaturated and monounsaturated consistency between different types of nutri-
fats, sucrose, carbohydrates, glycemic index and tional epidemiological studies can be explained,
glycemic load, dietary cholesterol, vitamin C, at least in part, by the potential for methodologic
fish, eggs, vegetables, tea, and soft drinks. biases using the case-​control approach, includ-
Following the WCRF/​AICR report, a pooled ing recall bias and selection bias. As for cohort
analysis of 14 prospective studies from North studies, data are hampered by low statistical
America, Europe and Australia found no associa- power because of the relatively low incidence of
tion between fruit and vegetable intake (Koushik pancreatic cancer.
et al, 2012) or between dairy foods, calcium and
vitamin D intake (Genkinger et  al, 2014), and Coffee
pancreatic cancer risk. Nevertheless, with respect Coffee is probably the dietary compound most
to fruits and vegetables, a subsequent meta-​ widely assessed and discussed as a potential risk
analysis suggested that fruit and vegetable intake factor for pancreatic cancer. In spite of some
is inversely associated with the risk of pancre- residual concerns (Kuper et al, 2000), the consen-
atic cancer (Wu et al, 2016). Inverse associations sus today is that there is no association between
between micronutrients abundant in fruits and coffee consumption and pancreatic cancer risk
314

314 Textbook of Cancer Epidemiology

(Gordis, 1990; Michaud et  al, 2001; WCRF/​ Lowenfels, 2015). Heavy alcohol consumption
AICR, 2012; Maisonneuve & Lowenfels, 2015). may be related to increased pancreatic cancer
A positive association between coffee and risk by promoting the effects of other risk factors
pancreatic cancer risk was first suggested in the such as tobacco smoking and poor nutrition, but
1970s and 1980s based on results from hospital-​ also through independent genetic and epigenetic
based case-​ control studies (MacMahon et  al, effects (Duell, 2012).
1981). Subsequent studies, however, found that
the initial positive association could be explained Reproductive Factors
by bias and cohort studies (Hsieh et al, 1986) have Although a link between reproductive factors
subsequently found no association (Michaud and pancreatic cancer risk has been suggested in
et al, 2001). A Spanish study added a new twist epidemiological studies, and is biologically plau-
to the convoluted story of coffee in relation to sible, limited direct evidence is available.
pancreatic cancer by suggesting an association Among all reproductive factors, parity is the
only with respect to KRAS positive cases (Porta most extensively studied, with studies suggesting
et al, 1999). both inverse (Kalapothaki et al, 1993; Fernandez
These divergent results are a good illustra- et al, 1995; Teras et al, 2005) and positive (Ji et al,
tion of the problems that investigators face when 1996; Kvale et al, 1994; Karlson et al, 1998) asso-
studying etiologic factors for pancreatic cancer. ciations with pancreatic cancer risk. In a meta-​
Because of the nonspecific abdominal symptoms analysis of 10 cohort and 10 case-​control studies,
in the early stages of disease and the potential parous women had a lower pancreatic cancer
role of smoking as an important confounder, risk compared with nulliparous women, while
high-​quality instruments to assess exposures and the lower risk was among women who gave birth
associated latencies are of special importance, to two children (Zhu et al, 2014). The presence
as is careful selection of controls in case-​control of estrogen receptors in the exocrine pancreas
studies. The rapid clinical course after diagnosis (Andren-​Sandberg, 1986) and high serum levels
also introduces problems with the representative- of estrogens in patients with pancreatic cancer
ness of cases and raises the question of whether compared to controls (Greenway et al, 1981) sug-
the extreme short-​term survivors who would die gest the role of hormonal influences during preg-
before they could be recruited as cases are simi- nancy as underlying the association.
lar to or different from the longer-​term survivors. Other reproductive factors such as age at
menarche, age at menopause, hysterectomy,
Alcohol oophorectomy, hormone replacement therapy,
Various methodological issues have hampered and oral contraceptives have mostly shown no
the study of alcohol consumption with respect associations with pancreatic cancer risk (Duell
to pancreatic cancer risk, such as exposure meas- et al, 2013; Tang et al, 2015).
urement error, residual confounding due to inad-
equate adjustment for important confounders Hormones
(e.g., smoking, history of pancreatitis) and com- Gastrointestinal hormones, which among others
peting risks (Duell, 2012). regulate cellular growth and secretion of the gas-
Epidemiological studies have shown con- trointestinal track, have been associated with the
sistently that moderate and heavy alcohol con- promotion of carcinogenesis in these locations.
sumption is associated with an increased risk Cholecystokinin (CCK) and its selective recep-
of pancreatic cancer (Tramacere et  al, 2010; tors have been shown to play a role in pancreatic
Lucenteforte et la, 2012; Bagnardi et  al, 2015). carcinogenesis and stimulation of pancreatic can-
While the WCRF/​AICR found no linear asso- cer growth (Smith & Solomon, 2014). More spe-
ciation between alcohol and risk of pancreatic cifically, receptors for CCK have been detected in
cancer, there was evidence of an increased risk in human pancreatic cancer, whereas CCK seems to
heavier drinkers who consume more than about stimulate the growth of some pancreatic cancer
three drinks per day (WCRF/​AICR, 2012). In a cell lines (Edwards et  al, 1989). Nevertheless, a
meta-​analysis, daily consumption of ≥30 grams question remains whether CCK acts as an early
of ethanol (equivalent to three or more glasses of promoter of cancer or as a product of cancer
alcoholic beverages) was associated with a 20% (Matters et  al, 2011; Smith & Solomon, 2014).
increase in pancreatic cancer risk translating to a Gastrin, another gastrointestinal hormone with
population attributable risk of 9% (Maisonneuve & a trophic effect on the gastrointestinal mucosa,
315

Pancreatic Cancer 315

has also been proposed to play a role in pancre- early life and childhood, and specifically greater
atic carcinogenesis (Orlando et al, 2007; Maddalo childhood growth, in pancreatic cancer etiology
et al, 2014). The increased risk found in patients (Aune et al, 2012c; WCRF/​AICR, 2012).
with pernicious anemia (Borch et al, 1988), which
is associated with high gastrin levels, has been an Infections
argument for a causal relationship (Hsing et  al, There is growing evidence that some infections
1993). However, pernicious anemia is an autoim- may be associated with pancreatic cancer risk
mune disorder linked to diabetes mellitus, itself (Michaud, 2013; Yeo, 2015; Maisonneuve &
a risk factor for pancreatic cancer. Moreover, Lowenfels, 2015).
considering that in a clinical setting pernicious Several studies have investigated Helicobacter
anemia is difficult to distinguish from chronic pylori infection and pancreatic cancer. In a meta-​
gastritis (Karlson et al, 2000), which is associated
analysis of nine studies, there was no overall
with smoking, the evidence for gastrin as a causal association between H. pylori seropositivity and
factor in pancreatic cancer is equivocal. pancreatic cancer risk, although strain-​specific
associations, and in particular, increased risk
Anthropometric Measures associated with CagA-​ negative H.  pylori was
The evidence that obesity, reflected by higher reported (Schulte et al, 2015). In contrast, CagA-​
body mass index (BMI), is positively associated positive strains of H.  pylori were associated
with pancreatic cancer incidence and mortal- with a lower risk of pancreatic cancer in Asian
ity is convincing (Genkinger et al, 2011; WCRF/​ populations (Wang et  al, 2014a). Overall, cur-
AICR, 2012; Maisonneuve & Lowenfels, 2015). rent evidence on H.  pylori can be characterized
Overweight and obese individuals have a 10% as moderate and more studies are needed to elu-
and 20% increased risk of pancreatic cancer com- cidate this association further (Maisonneuve &
pared to normal weight individuals, respectively Lowenfels, 2015).
(Maisonneuve & Lowenfels, 2015). The posi- Data from the Health Professionals Follow-​
tive association with BMI seems to be stronger Up Study found that men who had periodontal
among women and among nonsmokers (Fryzek disease had an increased risk of pancreatic cancer
et  al, 2005). Interestingly, among nonsmokers, (RR 1.64, 95% confidence interval [CI] 1.19–​2.26)
the risk associated with higher BMI is appar- (Michaud et  al, 2007; Michaud et  al, 2008). In
ent even among persons within the normal BMI nested case-​ control studies to look at plasma
range (Aune et al, 2012b). antibodies to oral bacteria, a pathogen associated
Abdominal obesity, measured by waist cir- with periodontal disease, Porphyromonas gingi-
cumference or waist-​to-​hip ratio, has been also valis, was positively associated with pancreatic
associated with pancreatic cancer risk (Aune cancer risk (Michaud, 2013; Chang et al, 2016).
et al, 2012b; Genkinger et al, 2015). Being over- A positive association between chronic hep-
weight or obese during early adulthood may be atitis B virus infection and pancreatic cancer
important in influencing pancreatic cancer mor- has recently been reported, but results are not
tality risk later in life (Genkinger et al, 2015). consistent (Hassan et  al, 2008; Andersen et  al,
Biologically plausible mechanisms have 2015). A positive but weaker association has also
been described to support a causal associa- been reported for hepatitis C infection (Fiorino
tion between body fatness and pancreatic can- et  al, 2013). Finally, pancreatic cancer is one of
cer. Body fatness is linked with insulin, insulin the non-​ AIDS-​ defining malignancies reported
resistance, and hyperinsulinemia, with hor- among HIV-​ positive patients successfully
mones such as insulin-​like growth factors and treated with highly active antiretroviral therapy
estrogens, as well as with the stimulation of a (HAART). Several studies have found a threefold
low-​grade chronic inflammation state (WCRF/​ increased risk of pancreatic cancer death in this
AICR, 2012). Because the prevalence of obesity group of patients (Yeo, 2015).
varies considerably between the countries, the
proportion of pancreatic cancer cases attributed Physical Activity
to obesity ranges from 3% to 16% (Maisonneuve Physical activity seems to protect against pan-
& Lowenfels, 2015). creatic cancer, although the association is rather
Taller attained height is positively associ- weak and more evidence is needed to support a
ated with pancreatic cancer risk, implicating causal relationship. A meta-​analysis including 30
the influence of developmental factors during studies with a total of 10,501 pancreatic cancer
316

316 Textbook of Cancer Epidemiology

cases found an inverse association between phys- the high-​risk occupational groups for the devel-
ical activity and pancreatic cancer risk (Behrens opment of pancreatic cancer (Yeo, 2015).
et  al, 2015). Findings were somewhat weaker
from cohort (RR 0.93, 95% CI 0.88–​0.98) than Medical Conditions
case-​control (RR 0.78, 95% CI 0.66-​0.94) studies. and Procedures
In cohort studies, the association was stronger Associated morbidity and changes in the inci-
for being consistently active compared to more dence pattern following different medical con-
recent activity. Leisure time physical activity was ditions or surgical interventions might help to
inversely associated with pancreatic cancer risk, explain the etiology of pancreatic cancer.
with stronger associations among those 50 years
and younger (Farris et al, 2015). Diabetes Mellitus
Researchers have worked intensively to clarify
Ionizing Radiation whether diabetes mellitus is a causal risk factor
Reports, notably from an early study on British or a result of pancreatic cancer. The relationship
radiologists (Matanoski et al, 1975), have linked seems to be complex and bidirectional (Chaudhry
occupational exposure to ionizing radiation with et  al, 2013; Salvatore et  al, 2015). Diabetes mel-
an increased risk of pancreatic cancer, although litus, mostly type 2, is considered a risk factor
these reports were not confirmed after prolonged for pancreatic cancer (Giovannucci et  al, 2010;
follow-​up or in other settings (Smith & Doll, Maisonneuve & Lowenfels, 2015). Several meta-​
1981; Tolley et al, 1983). However, a study based and pooled analyses have demonstrated that
on the National Dose Registry of Canada, the long-​ term diabetes is associated with 50% or
largest national registry of occupational radiation higher increased risk for pancreatic cancer (Elena
exposures in the world incorporating data from et al, 2013; Batabyal et al, 2014; Song et al, 2015;
nuclear, industrial, medical, and dental work- Maisonneuve & Lowenfels, 2015). The biological
ers, reported a statistically significant increase in mechanism by which diabetes increases the pan-
pancreatic cancer risk among men with increas- creatic cancer risk probably operates through the
ing radiation dose (Zielinski et al, 2008). increase of insulin-​like growth factor 1 (IGF-​1)
levels, hyperglycemia, insulin resistance, and com-
Occupation pensatory hyperinsulinemia (Song et  al, 2015).
Various occupational exposures have been asso- Hyperinsulinemia increases local blood flow and
ciated with increased risk of pancreatic can- cell divisions within the pancreas and promotes
cer. With the possible exception of chlorinated growth in human pancreatic cell lines (Takeda
hydrocarbon compounds, used in a variety of and Escribano, 1991). Type 1 diabetes mellitus has
applications and products such as solvents, plas- not been associated with an excess risk of pancre-
tics, and pesticides, evidence to support a con- atic cancer (Szablewski, 2014). The percentage of
vincing association is lacking. pancreatic cancer cases that could be attributed
Ojajarvi and colleagues (2000) found occu- to diabetes mellitus is estimated to range from
pational exposure to chlorinated hydrocarbon 1% to 16% (Maisonneuve & Lowenfels, 2015).
compounds was associated with an 80% higher Interestingly, metformin, used to treat diabetes,
risk of pancreatic cancer, and found the higher has been associated with lower pancreatic cancer
the quality of the study, the higher the gradient risk among diabetics (Wang et al, 2014b).
of risk. Most studies have confirmed this associ- In many instances, the risk of pancreatic
ation (Romundstad et al, 2000; Santibañez et al, cancer is greatest soon after the diagnosis of
2010; Andreotti & Silverman, 2012; Antwi et al, diabetes, suggesting that a new-​onset diabetes
2015). Occupational exposures to chromium that precedes the diagnosis of pancreatic cancer
and chromic compounds, as well as to nickel and in an older person may be more likely related
nickel compounds, have also been considered as to the underlying cancer process (Chaudhry
possible risk factors, but for these, as well as for et al, 2013).
asbestos, formaldehyde, gasoline, diesel exhaust, However, diabetes mellitus is often the first
wood dust, and pesticides, no firm conclusions clinical manifestation of an emerging pancreatic
have been drawn. Nevertheless, people working cancer due to alterations in the islet cell function
in dry-​cleaning, sawmills, electrical equipment and the loss of beta-​cells induced by the can-
manufacturing, chemical plants, and uranium cer (Warshaw and Fernandez-​del Castillo, 1992;
mines and metal workers are considered among Yeo, 2015).
317

Pancreatic Cancer 317

Cholecystectomy and Gallbladder often beginning during childhood. Hereditary

Disease pancreatitis markedly increases the risk for pan-
Gallstones, as well as cholecystectomy, may creatic cancer (Weiss, 2014). In a meta-​analysis
increase the risk of pancreatic cancer (Lin et al, of 22 studies, the summary RR for hereditary
2012). Possible biological mechanisms could be pancreatitis was 69.0 (Raimondi et al, 2010). The
the increase in secretion of CCK, described ear- cumulative risk of developing pancreatic cancer in
lier, which has trophic effects on, and stimulates patients with hereditary pancreatitis is estimated
the growth of, the exocrine pancreas, the reduc- up to 40% (Lowenfels et al, 1997). Data from the
tion of primary bile salts, and increase of second- European Registry of Hereditary Pancreatitis and
ary bile salts (Roda et al, 1978). Also, the surgical Pancreatic Cancer suggests the cumulative risk is
procedure or the presence of choledocholithiasis rather negligible (3.4%) until the age of 50, but
frequently causes injury to the sphincter of Oddi, increases considerably thereafter reaching 19%
resulting in a continuous flow of bile and/​ or at age 70 and 33% at age 80 (Howes et al, 2004).
duodenal contents (Karlson et al, 1997a), which The earlier belief that the increased risk was con-
exposes the ductal epithelium of the pancreas to fined to those who inherited the disease from
potentially carcinogenic products. their fathers was contradicted in a study showing
However, because early signs of pancreatic maternal inheritance (Lerch et al, 1999).
cancer include nonspecific abdominal symp- Chronic pancreatitis has been associated
toms that lead to higher diagnostic intensity and with pancreatic cancer in numerous studies
thereby perhaps to an increased frequency of (Lowenfels et  al, 1993; Raimondi et  al, 2010).
cholecystectomies in patients with “silent” gall- In a meta-​ analysis including 22 studies, the
stones, a confounded association can emerge. summary RR for chronic pancreatitis was 13.3,
In cohort studies of patients subjected to chole- whereas the authors estimated that over a 20-​
cystectomy, the incidence of pancreatic cancer year period, pancreatic cancer developed in 5%
compared to that of the background population or less of all patients with chronic pancreatitis
was increased up to 5  years after the operation, (Raimondi et  al, 2010). Chronic inflammation
but not thereafter (Ekbom et al, 1996; Chow et al, appears to play a role in the malignant trans-
1999). This illustrates the problems encountered formation of the pancreas as for several other
in epidemiologic studies of pancreatic cancer and organs, such as the liver in patients with hepati-
the need to disregard “exposures” consequent to tis (Colombo, 1999; Kuper et al, 2001), the colon
early symptoms from the tumor. following inflammatory bowel disease (Ekbom
et  al, 1990), the bile duct among patients with
Pernicious Anemia sclerosing cholingitis (Broome et  al, 1996),
Pernicious anemia is an autoimmune disor- and the lung among patients with lung fibrosis
der characterized by vitamin B12 deficiency. (Askling et  al, 1999). The malignant transfor-
Antibodies to parietal cells in the stomach result mation in the pancreatic tissue from the onset
in gastric atrophy, subsequent achloryhydria, and of benign pancreatitis to the onset of cancer,
chronic hypergastrinemia, which could explain requires approximately 20 years (Raimondi et al,
the reported association with pancreatic cancer 2010). K-​ras oncogene activity, which is pres-
(Hsing et al, 1993). Nevertheless, this association ent in nearly all pancreatic adenocarcinomas,
is not supported by subsequent studies (Karlson may act as an important link between benign
et al, 2000; Shah et al, 2014). Smoking, a reported and malignant state (Logsdon et  al, 2009). If
correlate of vitamin B12 deficiency, may act as a there is a causal relationship, the population-​
confounder. attributable fraction of chronic pancreatitis for
pancreatic cancer is low, estimated at less than
Pancreatitis 3% (Maisonneuve & Lowenfels, 2015).
Three different forms of pancreatitis have been Other forms of pancreatitis, including a single epi-
investigated in relation to pancreatic cancer. sode of acute pancreatitis (Ekbom et al, 1993), as
Hereditary pancreatitis is a rare autosomal dom- well as one of tropical pancreatitis (Thomas et al,
inant disorder with incomplete penetrance asso- 1990; Raimondi et  al, 2010), have been associ-
ciated with a mutation in the trypsinogen gene ated with an increased risk of pancreatic cancer.
on chromosome 7q35 (Lowenfels et  al, 1997). Because some of these associations were confined
Patients have recurrent episodes of pancreatitis to the first few years of follow-​up, some episodes
318

318 Textbook of Cancer Epidemiology

of pancreatitis might be an early clinical manifes- enhanced tumor immune surveillance, mean-
tation of pancreatic cancer. ing that their immune system identifies, through
tumor-​specific antigens or the release of stress
Partial Gastrectomy molecules, the cancer cells and destroys them
A history of peptic ulcer surgery, especially partial (Olson et al, 2013).
gastrectomy, has been reported to increase the risk
of pancreatic cancer (Offerhaus et al, 1988; Gong Primary Sclerosing Cholangitis
et  al, 2012). A  meta-​ analysis of observational Patients with primary sclerosing cholangitis have
studies found that gastrectomy is associated with a substantial excess risk for cancer in the bile
a 54% excess risk of pancreatic cancer, whereas ducts but also have a more than 10-​fold excess
the risk increased markedly 20–​30 years after sur- risk of pancreatic cancer (Bergquist et al, 2002).
gery (Gong et  al, 2012). Several underlying bio-
logical mechanisms have been implicated for the Other Risk Factors
observed association such as the effect of nitrosa- Common drugs such as aspirin (Zhang et  al,
mines produced by nitrate reductase–​producing 2015), other nonsteroidal anti-​ inflammatory
bacteria that proliferate in the hypoacidic stom- drugs, and statins (Cui et  al, 2012)  have been
ach, reflux, pancreatitis due to afferent loop syn- associated with a lower risk of pancreatic can-
drome following Billroth II partial gastrectomy, cer, although most of the evidence suggests
and, more recently, H.  pylori infection. Of note no association. Chlorinated municipal water
that for peptic ulcer disease, smoking is also one (Ijsselmuiden et  al, 1992), cadmium (Schwartz
of the main risk factors, and surgical patients have and Reis, 2000), and organochlorines (Hoppin
an increased risk over time of smoking-​related et al, 2000) have occasionally been brought up as
cancers in general (Ekbom et al, 1998). It appears risk factors. However, the evidence is fragmen-
that taking up smoking again after surgery could tary and inconsistent.
generate a confounded association between par-
tial gastrectomy and pancreatic cancer. CONCLUSION
Cancer of the pancreas is a major cause of death
Cystic Fibrosis from cancer worldwide and is the third most com-
Patients with cystic fibrosis have, among other mon cause of cancer death in the United States.
things, substantial pancreatic disturbance lead- In the vast majority of the cases, the diagnosis
ing to insufficiency of the pancreas. Long-​ occurs at an advanced stage when effective therapy
standing cystic fibrosis has been associated with is no longer feasible. Pancreatic cancer is slightly
a fivefold increased risk of pancreatic cancer more common among men than among women,
(Maissonneuve et al, 2007). Currently, pancreatic and most frequently diagnosed among the eld-
cancer could be characterized as a rare complica- erly. Thus, due to the aging of the world popula-
tion of cystic fibrosis, but with increasing longev- tion, the burden of the disease is expected to rise
ity of cystic fibrosis patients, it could become a in the future, especially in high-​income countries.
more common problem. The disease shows evidence of a familial occur-
rence, but most of the cases occur sporadically.
Allergies Two-​thirds of its known risk factors are potentially
There is growing body of evidence that certain modifiable (Table 13-​1). Tobacco smoking is the
allergies are inversely associated with the risk of best-​established exogenous cause of the disease,
pancreatic cancer (Olson et al, 2013; Cotterchio and smoking cessation is followed by a reduction
et al, 2014). Hay fever, animal/​pet allergies, and in excess risk. Obesity has emerged as an important
dust/​mold allergies have been associated with a risk factor for the disease, whereas higher attained
reduced risk of pancreatic cancer, whereas asthma adult height is also positively associated, impli-
has not. Results from studies using information cating the influence of early growth rate in pan-
on allergy ascertained by medical records or creatic cancer etiology. Consumption of red and
from prospectively designed studies are less clear processed meat, foods and beverages containing
compared to studies using self-​reported history fructose, and foods containing saturated fatty acids
of allergy (Turner, 2012). Although some pos- may increase pancreatic cancer incidence, whereas
sible biologic mechanisms have been proposed, evidence for a protective effect of fruits and foods
none is yet established. It has been suggested containing folate is still limited. Alcohol consump-
that individuals with allergies may demonstrate tion more than about three drinks/​day increases
319

Pancreatic Cancer 319

TABLE 13-​1   EPIDEMIOLOGY OF PANCREATIC CANCER: RISK

FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

History of chronic pancreatitis ↑↑
Hereditary pancreatitis ↑↑
Family history in first-​degree relatives ↑
Other familial cancer syndromes*1 ↑
Tobacco smoking ↑
Obesity ↑

Probable Relationship Exists, Based on Substantial Data

Increased attained height ↑
Diabetes mellitus ↑
H. Pylori infection ↑
Non-​O blood groups ↑
History of gastrectomy ↑

Suggestive Relationship Exists*2

Red meat ↑
Processed meat ↑
Foods and beverages containing fructose ↑
Foods containing saturated fatty acids ↑
Moderate to heavy alcohol consumption*3 ↑
Physical activity ↓
Allergies ↓
History of cholecystectomy ↑
Occupational exposure to chlorinated hydrocarbons ↑

Null Relationship Exists, Based on Substantial Data

Coffee consumption —​
Aspirin use —​
Statin use —​

Inconsistent Findings or Limited Study to Date

Smokeless tobacco ↑
Passive Tobacco ↑
Fruits ↓
Vegetables ↓
Folate ↓
Fish ↓
Total fat ↑ or —​
Polyunsaturated fats —​
Monounsaturated fat —​
Sucrose ↑
Soft drinks ↑

(continued)
320

320 Textbook of Cancer Epidemiology

TABLE 13-​1  CONTINUED

Risk factor Direction of effect*

Carbohydrates ↑
Vitamin C ↓
Plasma 25(OH)D level ↓
Glycemic index and Glycemic load ↑
Hepatitis B ↑
Hepatitis C ↑
Ionizing radiation ↑

*1 Hereditary nonpolyposis colorectal cancer, hereditary breast cancer and ovarian cancer, Peutz-​Jeghers syn-
drome, ataxia-​telangiectasia, Li-​Fraumeni syndrome, cystic fibrosis, Fanconi anemia, familial atypical multiple
mole melanoma syndrome
*2 Data for the suggestive relationship may be limited or may be substantial.
*3 Defined as consumption of about more than three alcoholic drinks/​day
*Arrows indicate the approximate magnitude of the relationship:  ↑, slight to moderate increase in risk; ↑↑,
moderate to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; 
—​, no association.

the risk of pancreatic cancer. The role of infec- Antwi SO, Eckert EC, Sabaque CV, Leof ER,
tions was highlighted during the last years since H. Hawthorne KM, Bamlet WR et  al. Exposure to
Pylori infection has been associated with increased environmental chemicals and heavy metals, and
risk, as well as chronic hepatitis B and C infections, risk of pancreatic cancer. Cancer Causes Control.
although the latter less consistently. Adult-​onset 2015;26:1583–​91.
diabetes mellitus increases the risk of pancreatic Askling J, Grunewald J, Eklund A, Hillerdal G, Ekbom
cancer although the relationship is complex and A. Increased risk for cancer following sarcoidosis.
bidirectional. Hereditary and chronic pancreatitis Am J Respir Crit Care Med 1999;160:1668–​72.
Aune D, Chan DS, Vieira AR, Navarro Rosenblatt DA,
increases substantially the risk of pancreatic cancer.
Vieira R, Greenwood DC et al. Dietary fructose,
Several reports indicate that some allergies may
carbohydrates, glycemic indices and pancreatic
be inversely related to the risk of pancreatic can-
cancer risk: a systematic review and meta-​analysis
cer. Occupational exposures may also play a role. of cohort studies. Ann Oncol 2012a;23:2536–​46.
Finally, a substantial progress in the understanding Aune D, Greenwood DC, Chan DS, Vieira R, Vieira
of the genomics of pancreatic cancer holds promise AR, Navarro Rosenblatt DA et  al. Body mass
for the future management of the disease. index, abdominal fatness and pancreatic cancer
risk:  a systematic review and non-​linear dose-​
REFERENCES response meta-​ analysis of prospective studies.
Amundadottir L, Kraft P, Stolzenberg-​Solomon RZ, Ann Oncol 2012b;23:843–​52.
Fuchs CS, Petersen GM et al. Genome-​wide asso- Aune D, Vieira AR, Chan DS, Navarro Rosenblatt DA,
ciation study identifies variants in the ABO locus Vieira R, Greenwood DC et  al. Height and pan-
associated with susceptibility to pancreatic cancer. creatic cancer risk: a systematic review and meta-​
Nat Genet 2009;41:986–​90. analysis of cohort studies. Cancer Causes Control
Andersen ES, Omland LH, Jepsen P, Krarup H, 2012c;23:1213–​22.
Christensen PB, Obel N, Weis N, DANVIR Bagnardi V, Rota M, Botteri E, Tramacere I,
Cohort Study. Risk of all-​type cancer, hepatocellu- Islami F, Fedirko V et  al. Alcohol consump-
lar carcinoma, non-​Hodgkin lymphoma and pan- tion and site-​specific cancer risk:  a comprehen-
creatic cancer in patients infected with hepatitis B sive dose-​ response meta-​ analysis. Br J Cancer
virus. J Viral Hepat 2015;22:828–​34. 2015;112:580–​93.
Andren-​Sandberg A. Estrogens and pancreatic can- Bao Y, Giovannucci E, Fuchs CS, Michaud DS. Passive
cer:  some recent aspects. Scand J Gastroenterol smoking and pancreatic cancer in women:  a
1986;21:129–​33. prospective cohort study. Cancer Epidemiol
Andreotti G, Silverman DT. Occupational risk factors Biomarkers Prev 2009;18:2292–​6.
and pancreatic cancer: a review of recent findings. Batabyal P, Vander Hoorn S, Christophi C, Nikfarjam
Mol Carcinog 2012;51:98–​108. M. Association of diabetes mellitus and pancreatic
321

Pancreatic Cancer 321

adenocarcinoma:  a meta-​analysis of 88 studies. and risk of pancreatic adenocarcinoma. Cancer

Ann Surg Oncol 2014;21:2453–​62. Epidemiol Biomarkers Prev 2006;15:726–​31.
Behrens G, Jochem C, Schmid D, Keimling M, Ricci Duell EJ, Holly EA, Bracci PM, Wiencke JK, Kelsey KT.
C, Leitzmann MF. Physical activity and risk of A population-​based study of the Arg399Gln pol-
pancreatic cancer: a systematic review and meta-​ ymorphism in X-​ray repair cross-​complementing
analysis. Eur J Epidemiol 2015;30:279–​98. group 1 (XRCC1) and risk of pancreatic adeno-
Bergquist A, Ekbom A, Olsson R, Kornfeldt D, Loof carcinoma. Cancer Res 2002;15:4630–​6.
L, Danielsson A, et  al. Hepatic and extrahepatic Duell EJ, Travier N, Lujan-​ Barroso L, Dossus L,
malignancies in primary sclerosing cholangitis. J Boutron-​ Ruault MC, Clavel-​ Chapelon F et  al.
Hepatol 2002;36:321–​7. Menstrual and reproductive factors in women,
Bertuccio P, La Vecchia C, Silverman DT, Petersen GM, genetic variation in CYP17A1, and pancreatic
Bracci PM, Negri E et al. Cigar and pipe smoking, cancer risk in the European prospective investiga-
smokeless tobacco use and pancreatic cancer: an tion into cancer and nutrition (EPIC) cohort. Int J
analysis from the International Pancreatic Cancer Cancer. 2013;132:2164–​75.
Case-​Control Consortium (PanC4). Ann Oncol Edwards BF, Redding TW, Schally AV. The effect of
2011;22:1420–​6. gastrointestinal hormones on the incorporation
Borch K, Kullman E, Hallhagen S, Ledin T, Ihse of tritiated thymidine in the pancreatic adeno-
I. Increased incidence of pancreatic neopla- carcinoma cell line (WD PaCa). Int J Pancreatol
sia in pernicious anemia. World J Surg 1988 1989;5:191–​201.
Dec;12(6):866–​70. Ekbom A, Helmick C, Zack M, Adami HO. Ulcerative
Broome U, Olsson R, Lööf L, Bodemar G, colitis and colorectal cancer. A population-​based
Hultcrantz R, Danielsson A et  al. Natural his- study. N Engl J Med 1990;323:1228–​33.
tory and prognostic factors in 305 Swedish Ekbom A, McLaughlin JK, Nyrén O. Pancreatitis and
patients with primary sclerosing cholangitis. the risk of pancreaticcancer. N Engl J Med 1993
Gut 1996;38:610–​15. Nov 11;329(20):1502–​3.
Chaudhry ZW, Hall E, Kalyani RR, Cosgrove DP, Yeh Ekbom A, Lundegardh G, McLaughlin JK, Nyrén O.
HC. Diabetes and pancreatic cancer. Curr Probl Relation of vagotomy to subsequent risk of lung
Cancer 2013;37:287–​92. cancer: population based cohort study. Br Med J
Chang JS, Tsai CR, Chen LT, Shan YS. Investigating the 1998;316:518–​19.
association between periodontal disease and risk Ekbom A, Yuen J, Karlsson BM, McLaughlin JK,
of pancreatic cancer. Pancreas 2016;45:134–​41. Adami HO. Risk of pancreatic and periampullar
Chow WH, Johansen C, Gridley G, Mellemkjaer L, cancer following cholecystectomy:  a population-​
Olsen JH, Fraumeni JF Jr. Gallstones, cholecystec- based cohort study. Dig Dis Sci 1996;41:387–​91.
tomy and risk of cancers of the liver, biliary tract Elena JW, Steplowski E, Yu K, Hartge P, Tobias GS,
and pancreas. Br J Cancer 1999;79:640–​4. Brotzman MJ et  al. Diabetes and risk of pancre-
Colombo M. Natural history and pathogenesis of hep- atic cancer: a pooled analysis from the pancreatic
atitis C virus related hepatocellular carcinoma. J cancer cohort consortium. Cancer Causes Control
Hepatol 1999;31 Suppl 1:25–​30. 2013;24:13–​25.
Cotterchio M, Lowcock E, Hudson TJ, Greenwood Fan H, Kou J, Han D, Li P, Zhang D, Wu Q et  al.
C, Gallinger S. Association between allergies Association between vitamin C intake and the risk
and risk of pancreatic cancer. Cancer Epidemiol of pancreatic cancer: a meta-​analysis of observa-
Biomarkers Prev 2014;23:469–​80. tional studies. Sci Rep 2015;5:13973.
Cui X, Xie Y, Chen M, Li J, Liao X, Shen J et al. Statin Farris MS, Mosli MH, McFadden AA, Friedenreich
use and risk of pancreatic cancer: a meta-​analysis. CM, Brenner DR. The association between lei-
Cancer Causes Control 2012;23:1099–​111. sure time physical activity and pancreatic can-
Ducreux M, Cuhna AS, Caramella C, Hollebecque cer risk in adults:  a systematic review and
A, Burtin P, Goéré D et  al. Cancer of the pan- meta-​analysis. Cancer Epidemiol Biomarkers Prev
creas:  ESMO Clinical Practice Guidelines for 2015;240:1462–​73.
diagnosis, treatment and follow-​up. Ann Oncol Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers
2015;Suppl 5:v56–​68. C, Rebelo M, et  al. Cancer incidence and mor-
Duell EJ. Epidemiology and potential mechanisms of tality worldwide:  sources, methods and major
tobacco smoking and heavy alcohol consumption patterns in GLOBOCAN 2012. Int J Cancer
in pancreatic cancer. Mol Carcinog 2012;51:40–​52. 2015;136:E359–​86.
Duell EJ, Casella DP, Burk RD, Kelsey KT, Holly EA. Fernandez E, La Vecchia C, D’Avanzo B, Negri E.
Inflammation, genetic polymorphisms in proin- Menstrual and reproductive factors and pancreatic
flammatory genes TNF-​A, RANTES, and CCR5, cancer risk in women. Int J Cancer 1995;62:11–​14.
32

322 Textbook of Cancer Epidemiology

Fernandez E, La Vecchia C, D’Avanzo B, Negri E, serum organochlorine levels. Cancer Epidemiol

Franceschi S. Family history and the risk of Biomarkers Prev 2000;9:199–​205.
liver, gallbladder, and pancreatic cancer. Cancer Howes N, Lerch MM, Greenhalf W, Stocken DD, Ellis
Epidemiol Biomarkers Prev. 1994;3(3):209–​12. I, Simon P et al. European Registry of Hereditary
Fiorino S, Chili E, Bacchi-​Reggiani L, Masetti M, Pancreatitis and Pancreatic Cancer (EUROPAC).
Deleonardi G, Grondona AG, et  al. Association Clinical and genetic characteristics of heredi-
between hepatitis B or hepatitis C virus infection tary pancreatitis in Europe. Clin Gastroenterol
and risk of pancreatic adenocarcinoma devel- Hepatol 2004;2:252–​61.
opment:  a systematic review and metaanalysis. Hsing AW, Hansson LE, McLaughlin JK, Nyrén O,
Pancreatology 2013;13:147–​160. Blot WJ, Ekbom A et  al. Pernicious anemia and
Fryzek JP, Schenk M, Kinnard M, Greenson JK, subsequent cancer:  a population-​ based cohort
Garabrant DH. The association of body mass study. Cancer 1993;71:745–​50.
index and pancreatic cancer in residents Hsieh CC, MacMahon B, Yen S, Trichopoulos D,
of southeastern Michigan. Am J Epidemiol Warren K, Nardi G. Coffee and pancreatic cancer
2005;162:222–​8. (­chapter 2). N Engl J Med 1986;315:587–​9.
Genkinger JM, Spiegelman D, Anderson KE, Iodice S, Gandini S, Maisonneuve P, Lowenfels AB.
Bernstein L, van den Brandt PA, Calle EE et  al. Tobacco and the risk of pancreatic cancer:  a
A pooled analysis of 14 cohort studies of anthro- review and meta-​analysis. Langenbecks Arch Surg
pometric factors and pancreatic cancer risk. Int J 2008;393:535–​45.
Cancer 2011;129:1708–​17. Ijsselmuiden CB, Gaydos C, Feighner B, Novakoski
Genkinger JM, Wang M, Li R, Albanes D, Anderson WL, Serwadda D, Caris LH et  al. Cancer of the
KE, Bernstein L, et  al. Dairy products and pan- pancreas and drinking water:  a population-​
creatic cancer risk: a pooled analysis of 14 cohort based case-​control study in Washington County,
studies. Ann Oncol 2014;25:1106–​15. Maryland. Am J Epidemiol 1992;136:836–​42.
Genkinger JM, Kitahara CM, Bernstein L, Berrington Jeurnink SM, Ros MM, Leenders M, van Duijnhoven
de Gonzalez A, Brotzman M, Elena JW et  al. FJ, Siersema PD, Jansen EH et  al. Plasma carot-
Central adiposity, obesity during early adulthood, enoids, vitamin C, retinol and tocopherols levels
and pancreatic cancer mortality in a pooled analy- and pancreatic cancer risk within the European
sis of cohort studies. Ann Oncol 2015;26:2257–​66. Prospective Investigation into Cancer and
Ghadirian P, Boyle P, Simard A, Baillargeon J, Nutrition:  a nested case-​ control study:  plasma
Maisonneuve P, Perret C. Reported family aggre- micronutrients and pancreatic cancer risk. Int J
gation of pancreatic cancer within a population-​ Cancer 2015;136:E665–​76.
based case-​ control study in the Francophone Ji BT, Hatch MC, Chow WH, McLaughlin JK, Dai Q,
community in Montreal, Canada. Int J Pancreatol Howe GR et al. Anthropometric and reproductive
1991;10:183–​96. factors and the risk of pancreatic cancer: a case-​
Giovannucci E, Harlan DM, Archer MC, Bergenstal control study in Shanghai, China. Int J Cancer
RM, Gapstur SM, Habel LA, Pollak M, 1996;66:432–​7.
Regensteiner JG, Yee D. Diabetes and can- Kalapothaki V, Tzonou A, Hsieh CC, Toupadaki N,
cer:  a consensus report. Diabetes Care 2010 Karakatsani A, Trichopoulos D. Tobacco, ethanol,
Jul;33(7):1674–​85. coffee, pancreatitis, diabetes mellitus, and chole-
Gong Y, Zhou Q, Zhou Y, Lin Q, Zeng B, Chen R, Li lithiasis as risk factors for pancreatic carcinoma.
Z. Gastrectomy and risk of pancreatic cancer: sys- Cancer Causes Control 1993 Jul;4(4):375–​82.
tematic review and meta-​analysis of observational Karlson BM, Ekbom A, Arvidsson D, Yuen J,
studies. Cancer Causes Control 2012;23:1279–​88. Krusemo UB. Population-​based study of cancer
Gordis L. Consumption of methylxanthine-​containing risk and relative survival following sphincterot-
beverages and risk of pancreatic cancer. Cancer omy for stones in the common bile duct. Br J Surg
Lett 1990;52:1–​12. 1997a;84:1235–​8.
Greenway B, Iqbal MJ, Johnson PJ, Williams R. Karlson BM, Hsing AW, Ekbom A, Wacholder S,
Oestrogen receptor proteins in malignant and fetal McLaughlin JK. Cancer of the upper gastrointesti-
pancreas. Br Med J (Clin Res Ed) 1981;283:751–​3. nal tract among patients with pernicious anemia.
Hassan MM, Li D, El-​Deeb AS, Wolff RA, Bondy ML, Scand J Gastroenterol 2000;35:847–​51.
Davila M, Abbruzzese JL. Association between Karlson BM, Wuu J, Hsieh CC, Lambe M, Ekbom A.
hepatitis B virus and pancreatic cancer. J Clin Parity and the risk of pancreatic cancer: a nested
Oncol. 2008;26:4557–​62. case-​control study. Int J Cancer 1998;77:224–​7.
Hoppin JA, Tolbert PE, Holly EA, Brock JW, Korrick Karlson BM, Ekbom A, Wacholder S, McLaughlin JK,
SA, Altshul LM et  al. Pancreatic cancer and Hsing AW. Cancer of the upper gastrointestinal
32

Pancreatic Cancer 323

tract among patients with pernicious anemia:  a Hereditary Pancreatitis Study Group. J Natl
case-​cohort study. Scand J Gastroenterol Cancer Inst 1997;89:442–​6.
2000;35:847–​51. Lucenteforte E, La Vecchia C, Silverman D, Petersen
Klein AP, Brune KA, Petersen GM, Goggins M, GM, Bracci PM, Ji BT et  al. Alcohol consump-
Tersmette AC, Offerhaus GJ et al. Prospective risk tion and pancreatic cancer:  a pooled analysis in
of pancreatic cancer in familial pancreatic cancer the International Pancreatic Cancer Case-​Control
kindreds. Cancer Res 2004;64:2634–​8. Consortium (PanC4). Ann Oncol 2012;23:374–​82.
Koushik A, Spiegelman D, Albanes D, Anderson KE, Lynch HT, Richardson JD, Amin M, Lynch JF,
Bernstein L, van den Brandt PA et  al. Intake of Cavalieri RJ, Bronson E et al. Variable gastrointes-
fruits and vegetables and risk of pancreatic can- tinal and urologic cancers in a Lynch syndrome II
cer in a pooled analysis of 14 cohort studies. Am J kindred. Dis Colon Rectum 1991;34:891–​5.
Epidemiol 2012;176:373–​86. Lynch SM, Vrieling A, Lubin JH, Kraft P, Mendelsohn
Kuper HE, Mucci LA, Trichopoulos D. Coffee, pan- JB, Hartge P et al. Cigarette smoking and pancre-
creatic cancer and the question of causation. J atic cancer:  a pooled analysis from the pancre-
Epidemiol Community Health 2000;54:650–​651. atic cancer cohort consortium. Am J Epidemiol
Kuper H, Ye W, Broome U, Romelsjö A, Mucci LA, 2009;170:403–​13.
Ekbom A et al. The risk of liver and bile duct can- MacMahon B, Yen S, Trichopoulos D, Warren K,
cers in patients with chronic viral hepatitis, alco- Nardi G. Coffee and cancer of the pancreas. N
holism, or cirrhosis. Hepatology 2001;34:714–​18. Engl J Med 1981;304:630–​3.
Kvale G, Heuch I, Nilssen S. Parity in relation to Maddalo G, Spolverato Y, Rugge M, Farinati F. Gastrin:
mortality and cancer incidence:  a prospective from pathophysiology to cancer prevention and
study of Norwegian women. Int J Epidemiol treatment. Eur J Cancer Prev 2014;23:258–​63.
1994;23:691–​9. Maisonneuve P, Lowenfels AB. Epidemiology of pan-
Lerch MM, Ellis I, Whitcomb DC, Keim V, Simon P, creatic cancer: an update. Dig Dis 2010;28:645–​56.
Howes N et  al. Maternal inheritance pattern of Maisonneuve P, Lowenfels AB. Risk factors for pancre-
hereditary pancreatitis in patients with pancreatic atic cancer: a summary review of meta-​analytical
carcinoma. J Natl Cancer Inst 1999;91:723–​4. studies. Int J Epidemiol 2015;44:186–​98.
Lin Y, Tamakoshi A, Kawamura T, Inaba Y, Kikuchi Maisonneuve P, Marshall BC, Lowenfels AB. Risk of
S, Motohashi Y et  al. JACC Study Group. Japan pancreatic cancer in patients with cystic fibrosis.
Collaborative cohort:  a prospective cohort study Gut 2007;56:1327–​8.
of cigarette smoking and pancreatic cancer in Matanoski GM, Seltser R, Sartwell PE, Diamond EL,
Japan. Cancer Cases Control 2002;13:249–​54. Elliott EA. The current mortality rates of radiolo-
Lin HL, An QZ, Wang QZ, Liu CX. Folate intake gists and other physician specialists: specific causes
and pancreatic cancer risk:  an overall and of death. Am J Epidemiol 1975;101:199–​210.
dose-​response meta-​ analysis. Public Health Matters GL, McGovern C, Harms JF, Markovic K,
2013;127:607–​13. Anson K, Jayakumar C et al. Role of endogenous
Lin G, Zeng Z, Wang X, Wu Z, Wang J, Wang C et al. cholecystokinin on growth of human pancreatic
Cholecystectomy and risk of pancreatic cancer: a cancer. Int J Oncol 2011;38:593–​601.
meta-​ analysis of observational studies. Cancer McIntyre CA, Winter JM. Diagnostic evaluation and
Causes Control 2012;23:59–​67. staging of pancreatic ductal adenocarcinoma.
Lindstrom P, Janzon L, Sternby NH. Declining Semin Oncol 2015;42:19–​27.
autopsy rate in Sweden:  a study of causes and Michaud DS. Role of bacterial infections in pancreatic
consequences in Malmo, Sweden. J Intern Med cancer. Carcinogenesis 2013;34:2193–​7.
1997;242:157–​65. Michaud DS, Giovannucci E, Willett WC, Colditz
Logsdon CD, Ji B. Ras activity in acinar cells links GA, Fuchs CS. Coffee and alcohol consumption
chronic pancreatitis and pancreatic cancer. Clin and the risk of pancreatic cancer in two prospec-
Gastroenterol Hepatol 2009;7:S40–​3. tive United States cohorts. Cancer Epidemiol
Lowenfels AB, Maisonneuve P, Cavallini G, Ammann Biomarkers Prev 2001;10:429–​37.
RW, Lankisch PG, Andersen JR, Dimagno EP, Michaud DS, Joshipura K, Giovannucci E, Fuchs CS.
Andrén-​ Sandberg A, Domellöf L. Pancreatitis A prospective study of periodontal disease and
and the risk of pancreatic cancer. International pancreatic cancer in US male health professionals.
Pancreatitis Study Group. N Engl J Med 1993 May J Natl Cancer Inst 2007;99:171–​5.
20;328(20):1433–​7. Michaud DS, Liu Y, Meyer M, Giovannucci E,
Lowenfels AB, Maisonneuve P, DiMagno EP, Elitsur Y, Joshipura K. Periodontal disease, tooth loss, and
Gates LK Jr, Perrault J, et al. Hereditary pancreati- cancer risk in male health professionals:  a pro-
tis and the risk of pancreatic cancer. International spective cohort study. Lancet Oncol 2008;9:550–​8.
324

324 Textbook of Cancer Epidemiology

Michaud DS, Skinner HG, Wu K, Hu F, Giovannucci Romundstad P, Andersen A, Haldorsen T. Cancer

E, Willett W et al. Dietary patterns and pancreatic incidence among workers in six Norwegian alu-
cancer risk in men and women. J Natl Cancer Inst minum plants. Scand J work Environ Health
2005;97:518–​24. 2000;26:461–​9.
Norat T, Scoccianti C, Boutron-​Ruault MC, Anderson Salvatore T, Marfella R, Rizzo MR, Sasso FC.
A, Berrino F, Cecchini M et  al. European Code Pancreatic cancer and diabetes:  a two-​way rela-
against Cancer 4th edition:  diet and cancer. tionship in the perspective of diabetologist. Int J
Cancer Epidemiol 2015;39 Suppl 1:S56–​66. Surg 2015;21 Suppl 1:S72–​7.
Offerhaus GJ, Tersmette AC, Tersmette KW, Tytgat Santibañez M, Vioque J, Alguacil J, de la Hera MG,
GN, Hoedemaeker PJ, Vandenbroucke JP. Moreno-​Osset E, Carrato A et  al. Occupational
Gastric, pancreatic, and colorectal carcinogenesis exposures and risk of pancreatic cancer. Eur J
following remote peptic ulcer surgery: review of Epidemiol 2010;25:721–​30.
the literature with the emphasis on risk assess- Schenk M, Severson RK, Pawlish KS. The risk of sub-
ment and underlying mechanism. Mod Pathol sequent primary carcinoma of the pancreas in
1988;1:352–​6. patients with cutaneous malignant melanoma.
Olson SH, Hsu M, Satagopan JM, Maisonneuve P, Cancer 1998;82:1672–​6.
Silverman DT, Lucenteforte E et al. Allergies and Schuller HM. Mechanisms of smoking-​related lung
risk of pancreatic cancer: a pooled analysis from and pancreatic adenocarcinoma development.
the Pancreatic Cancer Case-​Control Consortium. Nat Rev Cancer 2002;2:455–​63.
Am J Epidemiol 2013;178:691–​700. Schulte A, Pandeya N, Fawcett J, Fritschi L, Risch HA,
Ojajarvi IA, Partanen TJ, Ahlbom A, Boffetta P, Webb PM, et al. Association between Helicobacter
Hakulinen T, Jourenkova N et  al. Occupational pylori and pancreatic cancer risk: a meta-​analysis.
exposures and pancreatic cancer: a meta-​analysis. Cancer Causes Control 2015;26:1027–​35.
Occup Environ Med 2000;57:316–​24. Schwartz GG, Reis IM. Is cadmium a cause of human
Orlando LA, Lenard L, Orlando RC. Chronic hyper- pancreatic cancer? Cancer Epidemiol Biomarkers
gastrinemia: causes and consequences. Dig Dis Sci Prev 2000;9:139–​45.
2007;52:2482–​9. Schwingshackl L, Hoffmann G. Adherence to
Ozcelik H, Schmocker B, Di Nicola N, Shi XH, Langer Mediterranean diet and risk of cancer: an updated
B, Moore M et  al. Germline BRCA2 6174delT systematic review and meta-​analysis of observa-
mutations in Ashkenazi Jewish pancreatic cancer tional studies. Cancer Med 2015;4(12):1933–​47.
patients. Nat Genet 1997;16:17–​18. Scjeml M, Schwartz AG, O’Neal E, Kinnard
Paluszkiewicz P, Smolińska K, Dębińska I, Turski WA. M, Greenson JK, Gryzek JP et  al. Familial
Main dietary compounds and pancreatic cancer risk of pancreatic cancer. J Natl Cancer Inst
risk: the quantitative analysis of case-​control and 2001;93:640–​4.
cohort studies. Cancer Epidemiol 2012;36:60–​7. SEER Stat Fact Sheets:  Pancreas Cancer. http://​
Pandol SJ, Apte MV, Wilson JS, Gukovskaya AS, seer.cancer.gov/​ s tatfacts/​ html/​ p ancreas.html.
Edderkaoui M. The burning question:  why is Assessed November 12, 2015
smoking a risk factor for pancreatic cancer? Shah P, Rhim AD, Haynes K, Hwang WT, Yang YX.
Pancreatology 2012;12:344–​9. Diagnosis of pernicious anemia and the risk of
Petersen GM, de Andrade M, Goggins M, Hruban pancreatic cancer. Pancreas 2014;43:422–​6.
RH, Bondy M, Korczak JF et  al. Pancreatic can- Siegel RL, Miller KD, Jemal A. Cancer statistics 2015.
cer genetic epidemiology consortium. Cancer CA Cancer J Clin 2015;65:5–​29.
Epidemiol Biomarkers Prev 2006;15:704–​10. Silverman DT, Dunn JA, Hoover RN, Schiffman M,
Porta M, Malats N, Guarner L, Carrato A, Rifa J, Lillemoe KD, Schoenberg JB et al. Cigarette smok-
Salas A et  al. Association between coffee drink- ing and pancreas cancer:  a case-​ control study
ing and K-​ras mutations in exocrine pancreatic based on direct interviews. J Natl Cancer Inst
cancer. PANKRAS II Study Group. J Epidemiol 1994;86:1510–​16.
Community Health 1999;53:702–​9. Singh H, Siddiqui AA. Endosonographic workup and
Raimondi S, Lowenfels AB, Morselli-​ Labate AM, preoperative biliary drainage for pancreatic can-
Maisonneuve P, Pezzilli R. Pancreatic cancer in cer. Semin Oncol 2015;42:59–​69.
chronic pancreatitis; aetiology, incidence, and Smith JP, Solomon TE. Cholecystokinin and pancre-
early detection. Best Pract Res Clin Gastroenterol atic cancer: the chicken or the egg? Am J Physiol
2010;24:349–​58. Gastrointest Liver Physiol 2014;306:G91–​G101.
Roda E, Aldini R, Mazzella G, Roda A, Sama C, Festi Smith PG, Doll R. Mortality from cancer and all
D et al. Enterohepatic circulation of bile acids after causes among British radiologists. Br J Radiol
cholecystectomy. Gut 1978;19:640–​9. 1981;54:187–​94.
325

Pancreatic Cancer 325

Song S, Wang B, Zhang X, Hao L, Hu X, Li Z, Sun S. Wang Y, Zhang FC, Wang YJ. Helicobacter pylori and
Long-​Term Diabetes Mellitus Is Associated with pancreatic cancer risk:  a meta-​analysis based on
an Increased Risk of Pancreatic Cancer: A Meta-​ 2,049 cases and 2,861 controls. Asian Pac J Cancer
Analysis. PLoS One 2015 Jul 29;10(7):e0134321. Prev 2014a;15:4449–​54.
Swift M, Morrell D, Massey RB, Chase CL. Incidence Wang Z, Lai ST, Xie L, Zhao JD, Ma NY, Zhu J et al.
of cancer in 161 families affected by ataxia-​ Metformin is associated with reduced risk of pan-
telangiectasia. N Engl J Med 1991;325:1831–​6. creatic cancer in patients with type 2 diabetes
Szablewski L. Diabetes mellitus: influences on cancer mellitus:  a systematic review and meta-​analysis.
risk. Diabetes Metab Res Rev 2014;30:543–​53. Diabetes Res Clin Pract 2014b;106:19–​26.
Takeda Y, Escribano MJ. Effects of insulin and somat- Wang CH, Qiao C, Wang RC, Zhou WP. Dietary fiber
ostatin on the growth and the colony formation of intake and pancreatic cancer risk: a meta-​analysis
two human pancreatic cancer cell lines. J Cancer of epidemiologic studies. Sci Rep 2015;5:10834.
Res Clin Oncol 1991;117(5):416–​20. Warshaw AL, Fernandez-​del Castillo C. Pancreatic
Tang B, Lv J, Li Y, Yuan S, Wang Z, He S. Relationship carcinoma. N Engl J Med 1992;326:455–​65.
between female hormonal and menstrual fac- Weiss FU. Pancreatic cancer risk in hereditary pancre-
tors and pancreatic cancer:  a meta-​analysis of atitis. Front Physiol 2014;5:70.
observational studies. Medicine (Baltimore) World Cancer Research Fund/​American Institute for
2015;94:e177. Cancer Research (WCRF/​ AICR). Continuous
Teras LR, Patel AV, Rodriguez C, Thun MJ, Calle EE. Update Project Report. Food, Nutrition, Physical
Parity, other reproductive factors, and risk of pan- Activity, and the Prevention of Pancreatic Cancer
creatic cancer mortality in a large cohort of U.S. 2012. Available at http://​www.dietandcancerre-
women (United States). Cancer Causes Control port.org.
2005;16:1035–​40. Wu QJ, Wu L, Zheng LQ, Xu X, Ji C, Gong TT.
Tersmette AC, Petersen GM, Offerhaus GJ, Falatko Consumption of fruit and vegetables reduces risk
FC, Brune KA, Goggins M et  al. Increased risk of pancreatic cancer: evidence from epidemiolog-
of incident pancreatic cancer among first-​degree ical studies. Eur J Cancer Prev 2016;25:196–​205.
relatives of patients with familial pancreatic can- Yeo TP. Demographics, epidemiology, and inheritance
cer. Clin Cancer Res 2001;7:738–​44. of pancreatic ductal abdocarcinoma. Semin Oncol
Thomas PG, Augustine P, Ramesh H, Rangabashyam 2015;42:8–​18.
N. Observations and surgical management of Zielinski JM, Shilnikova NS, Krewski D. Canadian
tropical pancreatitis in Kerala and southern India. National Dose Registry of radiation work-
World J Surg 1990;14:32–​42. ers: overview of research from 1951 through 2007.
Tolley HD, Marks S, Buchanan JA, Gilbert ES. Int J Occup Med Environ Health 2008;21:269–​75.
A further update of the analysis of mortal- Zhang YP, Wan YD, Sun YL, Li J, Zhu RT. Aspirin
ity of workers in a nuclear facility. Radiat Res might reduce the incidence of pancreatic cancer: a
1983;95:211–​13. meta-​analysis of observational studies. Sci Rep
Tramacere I, Scotti L, Jenab M, Bagnardi V, Bellocco 2015;5:15460.
R, Rota M et al. Alcohol drinking and pancreatic Zhou J, Wellenius GA, Michaud DS. Environmental
cancer risk: a meta-​analysis of the dose-​risk rela- tobacco smoke and the risk of pancreatic can-
tion. Int J Cancer 2010;126:1474–​86. cer among non-​smokers: a meta-​analysis. Occup
Turner MC. Epidemiology:  allergy history, IgE, Environ Med 2012;69:853–​7.
and cancer. Cancer Immunol Immunother Zhu B, Zou L, Han J, Chen W, Shen N, Zhong R et al.
2012;61:1493–​510. Parity and pancreatic cancer risk: evidence from a
Vincent A, Herman J, Schulick R, Hruban R, Goggins meta-​analysis of twenty epidemiologic studies. Sci
M. Pancreatic cancer. Lancet 2011;378:607–​20. Rep 2014;4:5313.
326
327

14
Cancer of the Lung, Larynx, and Pleura
J Y O T I M A L H O T R A , P A O L O B O F F E T TA , A N D L O R E L E I   M U C C I

I n the 19th century, cases of lung cancer were

found among miners and some other occupa-
tional groups, but otherwise the disease was rare.
Symptoms Cough, sometimes with bleeding
(hemoptysis), and recurring pneumonia and
bronchitis are the predominant symptoms.
An epidemic increase in lung cancer began in the Locally advanced and metastatic disease may
first half of the 20th century, and there were much cause a multitude of symptoms besides those
speculation and controversy about its possible typical of an advanced malignancy, namely,
environmental causes. Later, lung cancer became breathlessness, obstruction of the airway and
a milestone in epidemiology research when its esophagus, enlarged lymph nodes, bone pain,
predominant cause, tobacco smoking, was estab- and loss of appetite or weight.
lished in a series of landmark studies beginning Diagnosis Sputum cytology is a simple test that
in 1950. Indeed, the challenge to demonstrate a is used when lung cancer is suspected. For stag-
causal association between smoking and lung ing, imaging by pulmonary x-​ ray, computed
cancer was a driving force in the development of tomography, and magnetic resonance imag-
epidemiologic methods during the second half of ing is required. Bronchoscopy is used to find
the century. Despite the success in defining lung tumors in the larger airways. A  biopsy or fine
cancer’s etiology, this highly preventable disease needle aspirate is use to confirm lung cancer
remains among the most common and most histopathologically. Whenever surgical treat-
lethal cancers globally. ment is considered, a mediastinoscopy may
Methodologically, epidemiologic studies of clarify whether the tumor has metastasized into
lung cancer have been straightforward because regional lymph nodes.
the site of origin is well defined, progressive
Treatment Cure by means of radical surgical
symptoms prompt diagnostic activity, and the
removal of a lobe or of one entire lung can be
predominant causes are comparatively easy to
achieved only rarely. Five years after surgery
ascertain. Novel approaches on the classification
with a curative intent, approximately one-​third
of lung cancer based on histology and molecular
of all patients are still alive. Radiation therapy
techniques have started to bring new insights in
can produce cure in a minority and palliation in
its etiology, in particular among nonsmokers.
a majority of patients. In advanced tumors, che-
motherapy produces a modest effect, with little
CLINICAL SYNOPSIS
influence on overall survival. There have been
Subgroups There are two main histologic types
recent advances in the use of immunotherapy as
of lung cancer: non-​small-​cell, which constitutes
well as targeted therapies for lung cancers with
85%, and small cell lung cancer, which consti-
specific types of mutations (e.g., ALK rearrange-
tutes 10%–​ 15%. Within non-​ small-​
cell lung
ments and EGFR mutations).
cancer are further subtypes of adenocarcinoma
(40%), squamous cell carcinoma (25%–​ 30%), Prognosis Lung cancer remains a largely lethal
and the more rare subtype, large cell carcinoma disease, although recent treatment advances
(5%). Malignant neoplasms of the lung origi- have increased the 5-​year relative survival rates
nate mainly from the bronchial, bronchiolar, to 15%–​18% in the United States and in Europe.
and alveolar epithelium, more rarely from the Progress. Considerable progress has been made
tracheal epithelium or from interstitial tissue of in the use of immunotherapy as well as tumor
the lung. genomics to identify actionable mutations.
328

328 Textbook of Cancer Epidemiology

Low-​dose computed tomography is now recom- incidence. Other factors, such as genetic suscep-
mended for lung cancer screening in the United tibility, poor diet, occupational exposures, and
States for current or recent former smokers air pollution, may act in concert with tobacco
with a long smoking history based on results smoking in shaping the descriptive epidemiol-
from randomized controlled trials. Primary ogy of lung cancer.
prevention through tobacco smoking cessation The geographic pattern we observe today in
remains the best strategy to control this globally men (Figs. 14-​3, 14-​4,) is composed of popu-
dominant cancer. lations at high risk, in which consumption of
tobacco has been persistently high for decades,
and populations at low risk, either because
CANCER OF THE LUNG tobacco consumption has not been increasing for
Descriptive Epidemiology long (e.g., China, Africa) or because a decrease
Lung cancer is the most frequently diagnosed in consumption has been present for several
cancer among men in most countries, and the decades (e.g., Sweden).
main neoplastic cause of death in both sexes. In Differences in lung cancer rates are frequently
2012, lung cancer accounted for an estimated observed within racial and ethnic groups. In the
1,242,000 new cases among men, which is 17% United States, the rates are higher among male
of all cancers excluding nonmelanoma skin African Americans than among other ethnic
cancer, and 583,000, that is, 9%, of new cancer groups (Table 14-​1).
cases among women. Approximately 58% of all Over the last 25 years, the distribution of his-
cases occur in low-​income countries (Foreman tologic types has been changing, where squamous
et al, 2014). cell carcinoma, which was formerly the predom-
Among both women and men, the inci- inant type in the United States, is decreasing,
dence of lung cancer is low under age 40 and while adenocarcinoma has increased in both
increases up to age 70 or 75 (Figs. 14-​1, 14-​2,). sexes (Meza et  al, 2015; Gabrielson, 2006). In
The decline in incidence in the older age groups Europe, similar changes in histologic types are
can be explained, at least in part, by incomplete occurring in men, while in women, both squa-
diagnosis or by a generation (birth-​ cohort) mous cell carcinoma and adenocarcinoma are
effect. The geographic and temporal patterns increasing (Tyczynski et al, 2003). Although the
of lung cancer incidence are determined chiefly increase in the incidence of adenocarcinoma may
by consumption of tobacco. The increase in be due, at least in part, to improved diagnostic
tobacco use across populations is paralleled a techniques, changes in composition and patterns
few decades later by an increase in the incidence of tobacco consumption (deeper inhalation of
of lung cancer; decreased consumption noted in low-​nicotine and tar tobacco smoke) are addi-
some populations is followed by a decrease in tional explanations (Health & Services, 2004).

400
New cases per 100,000

350
person-years

300
250
200
150
100
50
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 14-​1  Age-​specific incidence rates of lung cancer among women.

Source: Ferlay et al, 2013.
329

900

New cases per 100,000

800
700

person-years
600
500
400
300
200
100
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 14-​2  Age-​specific incidence rates of lung cancer among men.

Source: Ferlay et al, 2013.

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 5 10 15 20 25 30 35 40
ASR (world) per 100,000 person-years

FIGURE 14-​3  Age-​standardized (to the 2012 world population) incidence rates of lung cancer among women.
Source: Ferlay et al, 2013.

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 10 20 30 40 50 60 70
ASR (world) per 100,000 person-years

FIGURE 14-​4  Age-​standardized (to the 2012 world population) incidence rates of lung cancer among men.
Source: Ferlay et al, 2013.
30

330 Textbook of Cancer Epidemiology

In 2008, three separate GWAS of lung can-

TABLE 14-​1  AGE-STANDARDIZED
cer were published (Amos et  al, 2008; McKay
INCIDENCE RATES OF LUNG CANCER et  al, 2008; Thorgeirsson et  al, 2008)  and all
PER 100,000 BY SEX AND ETHNIC GROUP, three provided strong evidence for a suscepti-
SURVEILLANCE, EPIDEMIOLOGY AND bility region in 15q25.1, including two single
END RESULTS (SEER) PROGRAM, UNITED nucleotide polymorphisms (SNP)s:  rs8034191
STATES, 2003–2007 and rs1051730 (Amos et  al, 2008; McKay et  al,
2008; Thorgeirsson et  al, 2008). rs1051730 and
Ethnic group Men Women
rs8034191 map to a 100-​kb region of strong link-
Asian and Pacific Islander 31.6 17.5 age disequilibrium (LD) on chromosome 15. The
15q25 susceptibility region contains six identi-
Black 66.8 35.5
fied coding regions, including three cholinergic
Hispanic White 25.0 16.5 nicotine receptor genes (CHRNA3, CHRNA5,
Non-​Hispanic White 51.2 38.1 and CHRNB4), encoding nicotinic acetylcholine
receptors (nAChRs) in neuronal and other tis-
sues (Amos et  al, 2008). Variants on the 15q25
locus are also associated with increased vulnera-
Genetic and Molecular bility to tobacco addiction and changed smoking
Epidemiology behavior including increasing cigarettes smoked
Family History per day (Liu et  al, 2010; Saccone et  al, 2010;
Findings from registry-​based studies support Thorgeirsson et  al, 2008). Since nAChRs medi-
the finding that a positive family history of lung ate sensitivity to nicotine, it has been proposed
cancer is a risk factor, particularly for early-​ that variant receptors might increase addiction
onset lung cancer (Bailey-​Wilson et  al, 1993). to tobacco and, therefore, exposure to tobacco
Although shared smoking patterns within fami- carcinogens. Two variants (rs16969968 and
lies may explain a large part of the association rs578776) in this region are also associated with
with family history, increased familial risks were cotinine levels in current smokers (Timofeeva
found even after adjustment for smoking status et al, 2011).
(Lorenzo Bermejo & Hemminki, 2005). A pooled analysis of 3,259 lung cancer cases
and 4,159 controls identified an additional
High-​Penetrance Genes susceptibility locus in 5p15.33, a region that
Segregation analyses suggested that inheritance includes TERT (human telomerase reverse tran-
of a high penetrant gene, in conjunction with scriptase gene) and CLPTM1L (cleft lip and pal-
tobacco smoking, might account for more than ate transmembrane-​1-​like gene) (McKay et  al,
50% of early-​onset cases diagnosed below age 60 2008). Two variants in this region, rs402710
(Gauderman et  al, 1997). A  linkage analysis of (OR 1.15; p-​value 7 × 10-​5) and rs2736100 (OR
high-​risk pedigrees identified a major susceptibil- 1.09; p-​value 0.016), were both associated with
ity locus to chromosome 6q23–​25 (Bailey-​Wilson lung cancer risk. TERT is the reverse transcrip-
et  al, 2004). Lung cancer risk is also increased tase component of telomerase that is essential for
within the framework of the Li-​Fraumeni syn- telomerase enzymatic activity and maintenance
drome, characterized by germline mutation in the of telomeres. Telomerase is responsible for tel-
tumor suppressor gene p53 (Malkin et al, 1990). omere regeneration, and up to 90% of human
tumors show telomerase activity (Fernandez-​
Genetic Polymorphisms Garcia et al, 2008).
Genome-​ wide association studies (GWAS) A third significant susceptibility locus was
have identified multiple genetic polymorphisms reported in 6p21.23 (rs4324798; OR 1.28; p-​
underlying lung cancer risk. Table 14-​2 sum- value: 4 × 10-​7) and validated in a GWAS from the
marizes the evidence of an association between United Kingdom (McKay et al, 2008; Wang et al,
genetic variants and lung cancer. The three main 2008). Subsequently, a pooled analysis of 13,300
susceptibility loci identified are in 15q25, 5p15, primary lung cancer cases and 19,666 controls
and 6p21 regions (Amos et al, 2008; McKay et al, reported the association in the 6p region to be
2008; Thorgeirsson et al, 2008; Wang et al, 2008), weak; this could have been due to differences in
but many other common variants have also been population genetic structure and substructure as
reported, as listed in Table 14-​1. the 6p region is close to MHC (Landi et al, 2009).
31

Cancer of the Lung, Larynx, and Pleura 331

TABLE 14-​2   SUMMARY OF THE EVIDENCE FROM GENETIC ASSOCIATION STUDIES OF

GENETIC POLYMORPHISMS AND LUNG CANCER RISK

Susceptibility loci Tagging SNPs Genes Main population with

polymorphism

15q25.1 rs1051730 CHRNA3, CHRNA5, and Caucasians, African

CHRNB4 Americans, Asians (?)
rs8034191
rs16969968
rs12914385

5p15.33 rs402710 intron 1 of TERT Caucasians, Asians, ever smok-

ers and never smokers
rs2736100 CLPTM1L
rs401681

6p21.33 rs4324798 APOM, BAG6 Caucasians

rs3117582

3q28 rs4488809 TP63 Caucasians, Asians

22q12.2 rs17728461 HORMAD2 Asians

rs36600 LOC105372988

12p13.33 rs6489769 RAD52 Caucasians

12q23.1 rs12296850 Asians, Never smokers

18p11.22 rs11080466 PIEZO2 Asians, Never smokers

rs11663246

9p21 rs1333040 CDKN2B-​AS1 Caucasians, Asians

17q24.3 Rs7216064 BPTF Caucasians, Asians

10q25.2 rs7086803 VTI1A Never smokers, Asians

8q21 rs2741354 Ever smokers

13q31.3 rs2352028 GPC5 Never smokers, Caucasians

The associations between the susceptibility did not replicate the 15q and 6p regions but
loci and lung cancer risk appear to differ by his- confirmed the previously identified loci in
tology, with the effect largely restricted to adeno- 5p (Hu et  al, 2011). Similarly, a large replica-
carcinoma for the 5p15 locus (Truong et al, 2010; tion study of 21 case-​control studies from the
Landi et  al, 2009). Another novel susceptibility International Lung Cancer Consortium did not
loci at 9p21 reported in Caucasians is restricted show an association for 15q25 in Asians, who
to squamous cell lung cancer (Timofeeva et  al, constituted 15% of subjects in the study (Landi
2012). The only locus that has been consistently et  al, 2009). The association with 9p21 and
replicated in all histologic types is the 15q25 lung cancer risk has also been reported in Han
region (Timofeeva et al, 2012). Chinese subjects (Timofeeva et  al, 2012). In
The association between these susceptibil- addition, several new loci have been reported
ity loci and lung cancer risk varies by race and in Asians including 3q28 (Miki et al, 2010; Hu
ethnicity. A  GWAS in Han Chinese subjects et  al, 2011)  and 22q12.2 (Hu et  al, 2011). In
32

332 Textbook of Cancer Epidemiology

African American subjects, the 15q25 locus was Somatic Cells and Genetic Events
associated with lung cancer risk in two separate There is strong evidence for stepwise accumula-
studies (Amos et  al, 2010; Walsh et  al, 2012), tion in lung carcinogenesis of genetic alterations,
with the association reported to be stronger for including allelic losses, chromosomal instability,
lung adenocarcinoma. mutations in proto-​oncogenes and tumor sup-
There is heterogeneity in genetic suscepti- pressor genes, epigenetic changes, and expres-
bility, and lung cancer risk based on smoking sion alterations (Gazdar et  al, 2004). There are
status as shown by results of GWAS focused quite distinct patterns of alterations in small
on never smokers. In a UK study, there was no cell lung cancer and non-​small-​cell lung can-
association between 6p21.33 or 15q25.1 varia- cer, likely reflecting the different origin of these
tion and risk of lung cancer in never smokers two categories of tumors (epithelial cell with
(Wang et al, 2010), but there was a statistically neuroendocrine characteristics versus bronchial
significant association with 5p15.33 for ade- and alveolar cell). In the latter group, squamous
nocarcinoma. The 5p15 (Hsiung et  al, 2010), cell carcinoma, originating from bronchial cells
3q28 (Hosgood et  al, 2012), and 17q24.3 loci undergoing squamous metaplasia and dysplasia,
(Lan et  al, 2012)  have also been replicated in shows distinct genetic alterations from adeno-
never-​ smoking Asians, which suggests that carcinoma, deriving from alveolar or bronchio-
these regions are associated with lung cancer lar cells. The data on the other, more rare, types
irrespective of smoking history. Several new of lung cancer are sparse. Table 14-​3 shows the
loci have also been identified in never smokers most important genetic alterations observed in
including 18p11.22 (Ahn et  al, 2012), 12q23.1 lung cancer.
(Dong et al, 2013), and 10q25.2 (Shi et al, 2012). Mutations in the tumor suppressor gene TP53
Studies focused on pathway-​based approaches are the most frequent events in lung carcinogen-
complement single SNP analysis by incorporat- esis. Adenocarcinoma cases present a lower prev-
ing biological knowledge into the analysis (Khatri alence of TP53 mutations than other histological
et  al, 2012; Zhao et  al, 2011). A  large pooled types (Pfeifer et  al, 2002). Among lung cancer
analysis of six studies to investigate associations cases, the proportion of TP53 mutations increases
between 7,650 genetic variants in 720 genes with the duration and amount of tobacco smok-
related to inflammation pathways and lung can- ing (Le Calvez et  al, 2005). Inactivation of the
cer risk identified one novel variant (rs2741354 in pathway controlling the retinoblastoma gene,
EPHX2 at 8q21.1 with p-​value: 7.4 × 10-​6 after cor- RB1, that acts as a gatekeeper for the G1 to S tran-
recting for multiple comparisons), and confirmed sition of the cell cycle is another common alter-
the associations between 5p and 6p regions with ation in all types of lung cancer. Alterations are
lung cancer risk (Brenner et  al, 2013). Another particularly frequent in three genes involved in
analyses used imputation to the 1,000 Genomes the pathway:  loss of RB1 expression, detectable
Project with pooled GWAS data in European sub- in almost all small cell lung cancers, inactivation
jects and identified large-​effect associations for of p16, and overexpression of CCND1, encod-
squamous cell lung cancer with the rare variants ing cyclin D1 (Brambilla et al, 1999). Activating
BRCA2 p.Lys3326X (rs11571833) and CHEK2 point mutations at codon 12 in the KRAS onco-
p.Ile157Thr (rs17879961) (Wang et al, 2014). This gene occurs in 30% to 40% of adenocarcinoma
analysis demonstrated that imputation can iden- of the lung. Presence of KRAS is associated with
tify rare variants associated with cancer risk using poor prognosis independent of therapy (Slebos
preexisting GWAS data. et  al, 1990). KRAS alteration is associated with
GWAS explain only a small proportion of smoking (Slebos et al, 1991), and a specific pat-
the overall genetic variance with lung cancer. tern of mutations in TP53 and KRAS (excess of
However, the fact that only a small fraction of G→T transversions) has been identified in lung
smokers develop cancer supports the hypoth- cancers of heavy smokers. These mutations often
esis that genetic susceptibility might contrib- occur at bases that are binding sites for adducts of
ute to carcinogenesis. Newer techniques such as polycyclic aromatic hydrocarbons (PAHs) (Bozic
next-​generation sequencing to study rare genetic et al, 2010).
variants, epigenetics, and gene–​ environmental The discovery of a number of driver muta-
interactions may help in better delineation of lung tions underlying lung cancer, mainly in lung ade-
cancer susceptibility in the future (Maher, 2008). nocarcinoma, has led to significant advances in
3

Cancer of the Lung, Larynx, and Pleura 333

TABLE 14-​3   ACQUIRED SOMATIC GENETIC ALTERATIONS IN HUMAN LUNG CANCER

AND THEIR TARGETED AGENT FOR TREATMENT (IF AVAILABLE)

Gene Type of alteration Percentage of tumor with alteration Targeted agent for
treatment

Adeno SQ SCC

EGFR mutation 10%–​14% <2% 2%–​4% Erlotinib,

Gefitinib, Afatinib

KRAS mutation 25%–​33% 1%

HER2 mutation 2%–​3% 1% 0.9%–​3.4% Trastuzumab,

Pertuzumab**

BRAF mutation 8.7%–​10% 9.9% 1% Vemurafenib

PIK3CA mutation 6.2%–​7%, 6.1% 2.7%–​4.8%

(amplification-​4%)

AKT1 mutation 1% 1% 2.4%

MEK1 mutation 0.6% 1%

EML4-​ALK rearrangement 2%–​7% Crizotinib,

Ceritinib

MET mutation 7%, (amplification-​3%) 2.2% 1.8%–​3.4% Crizotinib

ROS1 rearrangement 1 Crizotinib

Adeno = adenocarcinoma; SQ = Squamous cell carcinoma; SCC = Small cell carcinoma

** not FDA approved for lung cancer

prognostication and treatment of these cancers. The Cancer Genome Atlas (TCGA) project
Driver mutations are molecular alterations in has profiled lung cancers by three major histology
tumor that play a significant role in tumor progres- groups in order to provide a comprehensive land-
sion and growth (Bozic et al, 2010). These driver scape of genomic alterations. All three main lung
mutations are targets for therapies with inhibitor cancer subtypes showed high rates of mutations
drugs specifically targeted at these mutations. For (mean 5.5 to 8.1 mutations per megabase). These
instance, tyrosine kinase inhibitors (TKIs), which projects have also identified a number of additional
target the intracellular tyrosine kinase domain of mutations, copy number alterations, and amplifi-
EGFR, such as gefitinib and erlotinib, have shown cations that can be potential targets for therapy in
dramatic efficacy in advanced stage lung cancer. the future (Cancer Genome Atlas Research, 2012;
Greatest treatment benefit is derived in patients Peifer et al, 2012; Rudin et al, 2012).
with a “sensitizing” EGFR mutation (exon 19 or
exon 21 858R) with significant improvement in Risk Factors
progression-​free survivals with use of these drugs Tobacco
(Maemondo et  al, 2010; Mitsudomi et  al, 2010; There is very strong evidence that tobacco
Sequist et al, 2013). EGFR mutation is seen mostly smoking causes all major histologic types of
in adenocarcinoma (Toh & Lim, 2007)  and the lung cancer. A  carcinogenic effect of tobacco
prevalence of this mutation can be as high as smoke on the lung was first demonstrated in
40%–​50% in women, never or light smokers, and epidemiologic studies conducted in the early
Asians (Nordquist et  al, 2004; Riely et  al, 2006; 1950s and has been recognized by public health
Shigematsu & Gazdar, 2006; Toh et al, 2006). and regulatory authorities since the mid-​1960s.
34

334 Textbook of Cancer Epidemiology

Tobacco smoking is the main cause of lung can- blond tobacco are used, they are more frequently
cer in most populations, and the geographic and made of black tobacco.
temporal patterns of the disease largely reflect Although cigarettes are the main tobacco
tobacco consumption accumulated during the product smoked in Western countries, an
previous decades. Because of the strong car- exposure–​response relationship with lung can-
cinogenic potency of tobacco smoke, a major cer risk has also been shown for cigars, cigaril-
reduction in tobacco consumption would result los, and pipes, indicating a carcinogenic effect of
in the prevention of a large fraction of lung as these products (IARC, 2004c). An increased risk
well as other human cancers (Peto et  al, 1992; of lung cancer has also been shown following
International Agency for Research on Cancer consumption of local tobacco products, such as
[IARC], 2004c). bidi and hookah in India, khii yoo in Thailand,
The excess risk of lung cancer among regular and water pipe in China (IARC, 2004c). Limited
smokers relative to that among never smokers is data suggest an increased lung cancer risk follow-
on the order of 10-​to 20-​fold. The overall relative ing consumption of other tobacco products, such
risk (RR) reflects the contribution of the different as narghile in western Asia and northern Africa,
aspects of tobacco smoking:  average consump- and toombak in Sudan.
tion, duration of smoking, time since quitting,
age at start, type of tobacco product, and inhala- Differences in the Effect of Tobacco Smoking
tion pattern, as well as the absolute risk in never According to Histology, Sex, and Race
smokers. Although the evidence is abundant that tobacco
Several large cohort and case-​control stud- smoking causes all major histologic types of lung
ies have provided detailed information on the cancer, the associations appear to be stronger
relative contributions of duration and amount for squamous cell and small cell carcinoma and
of cigarette smoking on the excess lung cancer weaker for adenocarcinoma. The incidence of the
risk. In a large cohort of British doctors, Doll latter histologic type has greatly increased dur-
and Peto (Doll & Peto, 1978)  found that the ing the last decades. Some of the increase might
excess lung cancer risk rises in proportion to be attributable to improved diagnostic tech-
the square of the number of cigarettes smoked niques, but it is possible that changes in aspects
per day and to the fourth power of the duration of tobacco smoking have also played a role; it
of smoking. Therefore, duration of smoking is unclear however, which aspects of smoking
should be considered the strongest determinant might explain these changes.
of lung cancer risk in smokers. These results A few studies have suggested a difference in
have been confirmed by the analysis of the the risk of lung cancer between men and women
same cohort after 50  years of follow-​up (Doll who have smoked a comparable amount of
et al, 2004). tobacco (Zang & Wynder, 1996), but most of the
An important aspect of tobacco-​ related available evidence does not support this sex dif-
lung carcinogenesis is the effect of cessation of ference (IARC, 2004c).
smoking. The excess risk of lung cancer sharply The higher rate of lung cancer among African
decreases in ex-​smokers starting approximately Americans as compared to other ethnic groups in
5  years after quitting, and an effect is apparent the United States is probably explained by their
even for cessation late in life. However, a small higher tobacco consumption (Devesa et al, 1999).
excess risk throughout life likely persists even in The lower risk of lung cancer among smokers in
long-​term quitters (IARC, 2004c). China and Japan as compared to Europe and
The risk of lung cancer is lower among smok- North America might be due to relatively recent
ers of low-​tar cigarettes than among smokers of introduction of regular heavy smoking in Asia,
high-​tar cigarettes and among smokers of filtered although differences in the composition of tradi-
cigarettes than smokers of unfiltered cigarettes. tional smoking products and in genetic suscep-
Smokers of black (air-​cured) tobacco cigarettes tibility might also play a role (Yuan et al, 2009).
are at two-​to threefold higher risk of lung cancer
than smokers of blond (flue-​cured) tobacco ciga- Second-​Hand Tobacco Smoke
rettes (IARC, 2004c). Tar content, the presence of The epidemiologic evidence and biological plau-
a filter, and the type of tobacco are not independ- sibility support a causal association between
ent, however: high-​tar cigarettes tend to be unfil- second-​hand exposure to cigarette smoke and
tered and, in the countries where both black and lung cancer risk in nonsmokers (IARC, 2004b).
35

Cancer of the Lung, Larynx, and Pleura 335

The evidence of a high RR in the original stud- process, for example, induction of common
ies (Hirayama, 2000; Trichopoulos et  al,1981) metabolic enzymes or activation of common
has been challenged on the basis of possible con- molecular targets. The empirical evidence on
founding by active smoking or other factors, and interaction between tobacco smoking and
of possible reporting bias. When these factors other agents is scanty, mainly because of lack
were taken into account, the association of pas- of data among light smokers and nonsmok-
sive smoking was confirmed, although the excess ers (Boffetta, 1993). An interaction between
risk was on the order of 20% to 25% (Hackshaw additive and multiplicative has been suggested
et al, 1997; IARC, 2004b). for the combination of asbestos exposure and
The effect of involuntary smoking appears tobacco smoking (Liddell, 2001). The inter-
to be present for both household exposure, action between radon exposure and tobacco
mainly from the spouse, and workplace exposure smoking best fits a submultiplicative model;
(Boffetta, 2002; Stayner et al, 2007). On the other data for other agents are too sparse to allow
hand, there is little evidence of an effect of child- conclusions.
hood involuntary smoking exposure and future
lung cancer risk (Boffetta et al, 2000). Use of Smokeless Tobacco Products
Few studies have investigated the risk of lung
Confounding Effects of Tobacco Smoking cancer among users of smokeless tobacco prod-
The importance of tobacco smoking in the cau- ucts. In two US cohorts, the RRs for spit tobacco
sation of lung cancer complicates the investiga- use among nonsmokers were 1.08 (95% confi-
tion of the other causes of this disease because dence interval [CI] 0.64–​1.83) in one cohort and
tobacco smoking may act as a powerful con- 2.00 (95% CI 1.23–​3.24) in the other (Henley
founder. For example, a population of indus- et  al, 2005). In a Swedish cohort, the relative
trial workers exposed to a suspected carcinogen risk for ever-​use of the smokeless tobacco prod-
may smoke more than the unexposed compari- uct snus was 0.80 (95% CI 0.61–​1.05) (Boffetta
son population. An excessive lung cancer risk in et  al, 2005). In a large case-​control study from
the exposed group, especially if small, might be India, the relative of lung cancer for ever-​use of
due to the difference in smoking rather than to tobacco-​containing chewing products was 0.74
the effect of the occupational agent. One solu- (95% CI 0.57–​0.96) (Gajalakshmi et  al, 2003).
tion is to restrict the investigation to lifetime Overall, the evidence of an increased risk of lung
nonsmokers. However, they may represent a cancer from use of smokeless tobacco products
selected group, with low prevalence of exposure is weak; the apparent protective effect detected
to many agents of interest. An alternative is to in studies including smokers might be due to
collect detailed information on smoking hab- uncontrolled negative confounding.
its and to compare the effect of the suspected
carcinogens across different groups of smokers. Dietary Factors
This approach has shown that tobacco smok- Diet can also influence the risk for lung cancer
ing as a confounder rarely explains completely although the supporting evidence is not strong.
excess risks larger than about 50% (Siemiatycki The main studies that have explored the associ-
et al, 2006). ation of lung cancer risk with dietary factors are
listed below.
Interaction between Tobacco Smoke
and Other Lung Carcinogens Vegetables and Fruits
Other carcinogens might interact with tobacco There is some evidence that a diet rich in veg-
smoke in the determination of their carcino- etables and fruits probably exerts a protec-
genic action on the lung. In other words, the tive effect against lung cancer (World Cancer
absolute risk or RR from exposure to another Research Fund, American Institute for Cancer
agent might be greater (or smaller) among Research [WCRF/​AICR], 2007). Although most
heavy smokers compared to the corresponding case-​control studies reported a protective effect
risk among light smokers and nonsmokers. The of high vegetable and fruit intake, results of pro-
interaction might take place either at the stage spective studies with detailed information on
of exposure—​that is, the other agent has to be dietary intake are less consistent in showing a
absorbed on the tobacco particles to penetrate similar effect. Possible reasons for the inconsist-
the lung—​or at some stage of the carcinogenic ent results include bias from retrospective dietary
36

336 Textbook of Cancer Epidemiology

assessment, misclassification and limited het- Carotenoids

erogeneity of exposure in cohort studies, resid- Many studies have addressed the risk of lung can-
ual confounding by smoking, and variability in cer according to estimated intake of either beta-​
food composition. Among specific types of fruits carotene or total carotenoids (which in most
and vegetables, the evidence is stronger for cru- cases correspond to the sum of alpha-​and beta-​
ciferous vegetables (Lam et  al, 2009), but even carotene) (Albanes, 1999). Five cohort and 18
in this case it is unlikely that this group of foodscase-​control studies published up to 1994 (World
represents a strong protective factor against lung Cancer Research Fund, 1997)  provided 28 risk
cancer. estimates in different populations; with some
notable exceptions (Kalandidi et al, 1990), 25 of
Meat and Other Foods them indicated a protective effect of high beta-​
It has been suggested that high intake of meat, in carotene intake. The protective effect provided a
particular fried or well-​done red meat, increases 30% to 80% reduction in the risk of lung cancer
the risk of lung cancer (Sinha et  al, 1998), between the highest and lowest intake categories
although the available evidence does not support (World Cancer Research Fund, 1997). The risk
this hypothesis (World Cancer Research Fund, decreased for all major histologic types of lung
2007). If real, the association might be explained cancer in many countries, in both genders, and
by formation of nitrosamines during cooking of in both smokers and nonsmokers. Similar results
the meat (Sinha et  al, 2000), as well as by satu- have been obtained in studies based on measure-
rated fat content of meat (see the section on ment of beta-​carotene in prospectively collected
lipids below). Although some studies reported sera (Holick et al, 2002). The evidence of a pro-
risk estimates for intake of other foods, such as tective effect from most observational studies
cereals, pulses, eggs, milk, and dairy products, has been refuted by the results of randomized
these results are inadequate for a judgment of intervention trials based on beta-​carotene sup-
the evidence of an effect (World Cancer Research plementation (Table 14-​4). In two of them, which
Fund, 2007). included smokers or workers exposed to asbes-
tos, a significant increase in the incidence of lung
Coffee and Tea cancer was observed in the treated groups; in the
In a few studies, high consumption of coffee remaining studies, no effect was ascertained. The
has been associated with an increased risk of difference in results between observational stud-
lung cancer. Residual confounding by tobacco ies and preventive trials can be explained by con-
smoking, however, remains a distinct possibil- founding by factors in fruits and vegetables other
ity, and no evidence of an increased risk has been than beta-​carotene, or by the fact that high, non-
reported in studies of never-​smokers (WCRF/​ physiological doses, beta-​carotene might cause
AICR, 2007). There is some evidence of a che- oxidative damage, in particular among smokers
mopreventive effect of tea, notably green tea, in (Greenwald, 2003).
smokers (Clark & You, 2006). The overall evi-
dence, however, is not consistent. Other Micronutrients
For none of the antioxidant vitamins or the other
Lipids micronutrients is there conclusive evidence of a
Several ecologic studies show a positive associ- protective effect against lung cancer. In the case
ation between total lipid intake and lung cancer of selenium, vitamin A, lutein, and lycopene,
risk that appears to be independent of that of in particular, the data are inconclusive (Ruano-​
tobacco consumption (World Cancer Research Ravina et al, 2006; WCRF/​AICR, 2007). There is
Fund, 1997). The analytic studies that have evidence from observational studies that low lev-
addressed this association, however, have pro- els of vitamin D are associated with lung cancer
duced mixed results. Although no study pro- risk (Herr et al, 2011); results of randomized tri-
vided evidence of a protective effect of total als, however, do not provide supportive evidence,
lipid intake, an increased risk was shown only arguing for caution in drawing conclusions.
in case-​controls studies, while a pooled analysis
of eight cohort studies provided no evidence of Isothiocyanates
an increased risk of lung cancer for high intake Isothiocyanates are a group of chemicals with
of either total fat or saturated fat (Smith-​Warner cancer-​preventive activity in experimental sys-
et al, 2002). tems, and may be responsible for the possibly
37

Cancer of the Lung, Larynx, and Pleura 337

TABLE 14-​4   PREVENTION TRIALS ON B-CAROTENE SUPPLEMENTATION

AND LUNG CANCER RISK

Reference Setting, population, age Follow-​up Daily dose RR 95% CI

Kamangar et al. Linxian, China; 29,584, 40–​69 1986–​2001 15 mg* 0.98 0.71–​1.35
2006
ATBCCP Study Finland; 29,133 male smokers, 1985–​93** 20 mg 1.18 1.03–​1.36
Group 1994 50–​69
Hennekens et al. USA; 22,071 male physicians, 1982–​95 25 mg*** 0.93 NA
1996 40–​84
Omenn et al. USA; 18,314 smokers or asbes- 1985–​95 30 mg 1.28 1.04–​1.57
1994 tos workers, 45–​74

* combined with selenium (50 μg) and α-​tocopherol (30 mg)

** follow-​up for cancer incidence
*** 50 mg on alternate days
RR: relative risk; CI: confidence interval; NA: not available

reduced risk of lung cancer in relation to high while properly controlling for the potential
intake of cruciferous vegetables. The enzymes, confounding effect of tobacco. Meta-​analyses
glutathione-​S-​transferases M1 and T1 are have indicated that the increased risk of lung
involved in their metabolism: as indicated, these cancer observed among alcoholics is mainly
enzymes are polymorphic, with 5% to 10% of attributable to such residual confounding, but
Europeans and 30% to 40% of Asians being car- a smoking-​adjusted association was suggested
riers of a deletion in both. In four studies, it has for high alcohol consumption (Bandera et  al,
been shown that the protective effect of high 2001; Korte et  al, 2002). This conclusion was
intake of isothiocyanates is stronger in carriers of confirmed by a pooled analysis of seven cohort
both deletions than in other subjects (Fig. 14-​5). studies (Freudenheim et  al, 2005). Overall, it
No final conclusions can be drawn at present, but might be premature to conclude that an associ-
this is an example of possible gene–​environment ation between alcohol drinking and lung cancer
interaction in lung carcinogenesis. has been refuted by the available data (Bagnardi
et  al, 2015). If the association is causal, alco-
Alcohol hol might act as a solvent for carcinogens such
Given the strong correlation between alco- as those in tobacco smoke. In addition, it can
hol drinking and tobacco smoking in many induce metabolic enzymes or act through direct
populations, it is difficult to disentangle the DNA damage via the active metabolite acetalde-
contribution of alcohol to lung carcinogenesis hyde (Fang & Vaca, 1997).

1.25

1
Relative risk

0.75

0.5

0.25

0
London et al., Spitz et al., Zhao et al., Brennan et al.,
2000 2000 2001 2005

GSTM1 + or GSTT1 + GSTM1– and GSTT1–

FIGURE 14-​5  Interaction between high intake of isotiocyanates and polymorphism in GSTM1 and GSTT1 in four
case-​control studies of lung cancer.
38

338 Textbook of Cancer Epidemiology

Hormones (Aoki, 1993). A similar association was reported

Estrogen and progesterone receptors are from community-​based studies among smoking
expressed in the normal lung and in lung cancer and nonsmoking women (Alavanja et  al, 1992;
cell lines, and estradiol has a proliferative effect Gao et al, 1987; Ko et al, 1997; Wu et al, 1988).
on the latter type of cells. Although an effect of In a large cohort of tuberculosis patients from
estrogens on lung carcinogenesis has not been Shanghai, China (Zheng et  al, 1987), the RR of
demonstrated, they might act via formation of lung cancer in the whole cohort was 1.5, and was
DNA adducts and activation of growth factors 2.0 among individuals 20 years after the diagno-
(Thomas et al, 2005). Data on risk of lung cancer sis of tuberculosis. Moreover, there was correla-
following use of hormone replacement therapy tion with the location of the tuberculosis lesions
have been reported from five case-​control stud- and lung tumor. Whether the excess risk is
ies, two cohort studies, and one randomized trial caused by the chronic inflammatory status of the
(Adami et al, 1989; Blackman et al, 2002; Kreuzer lung parenchyma or by the specific action of the
et al, 2003; Lipson et al, 2012; Olsson et al, 2003; mycobacterium is not clear. A role of isoniazid,
Rossouw et al, 2002; Schabath et al, 2004; Taioli & a widely used tuberculosis drug that causes lung
Wynder,1994; Wu et al, 1988). A small increased tumors in experimental animals, was excluded in
risk of lung cancer has been reported in the early one large study (Boice & Fraumeni Jr, 1980).
studies, while a decreased risk was detected in the Chlamydia pneumoniae is a cause of acute
more recent studies. No effect was observed in respiratory infection. Six studies have been pub-
the only randomized trial (Rossouw et al, 2002). lished on risk of lung cancer among individuals
While the different results might be explained by with markers of C.  pneumoniae infection, with
changes in the formulations used for replacement consistent positive associations (Littman et  al,
therapy, the lack of an effect in the only study 2005). However, studies based on prediagnostic
with an experimental design argues against an blood samples had weaker risk estimates than
effect of this type of exposure on lung cancer. studies based on postdiagnostic samples. An
In a meta-​analysis of three cohorts and one association between infection with human pap-
case-​ control study of serum IGF-​ 1 level and illoma virus (HPV) and risk of lung cancer, par-
lung cancer risk, the overall RR was 1.01 (95% ticularly adenocarcinoma, has been suggested by
CI 0.49–​2.11) (Renehan et  al, 2004). Similarly, the results of analysis of series of cases (Brenner
there was no association with IGFBP-​3 levels and et al, 2011; Soini et al, 1996) and from the grow-
lung cancer risk (RR 0.83, 95% CI 0.38–​1.84), ing evidence of an increased risk among workers
although exclusion of a deviant study resulted in potentially exposed to this agent, such as butch-
a decreased risk of lung cancer for high IGFBP-​3 ers (Durusoy et al, 2006). However, the results are
levels (RR 0.53, 95% CI 0.34–​0.83). currently insufficient to conclude on the presence
or absence of a causal association between HPV
Anthropometric Measures and lung cancer. Other biological agents that
There is some evidence that a reduced body mass have been suggested to play a role in lung carcin-
index (BMI) is associated with an increased risk ogenesis include Simian virus 40 (Galateau-​Salle
of lung cancer. However, this inverse association et  al, 1997)  and the fungus Microsporum canis
can be explained, at least in part, by negative con- (Nakachi et al, 1999).
founding by smoking (Henley et al, 2002), and no
clear association has been demonstrated among Ionizing Radiation
never smokers. Subsequent studies supported There is conclusive evidence that high exposure
this explanation (Calle et al, 2003). There is evi- to ionizing radiation increases the risk of lung
dence suggesting a positive association between cancer (IARC, 2000). Atomic bomb survivors
height and lung cancer risk (Gunnell et al, 2001). and patients treated with radiotherapy for anky-
Subsequent studies supported this finding (Batty losing spondylitis or breast cancer are at mod-
et  al, 2006; Gunnell, et  al, 2003), although the erately increased risk of lung cancer, with RRs
evidence is not fully consistent (Rodriguez et al, of 1.5 to 2 for cumulative exposure in excess of
2001; Song et al, 2003). 100 rads (Boice, 1996). The association with high
doses of ionizing radiation was stronger for small
Infections cell carcinoma than for other histologic types of
Patients with pulmonary tuberculosis have been lung cancer. Studies of nuclear industry work-
found to be at increased risk of lung cancer ers exposed to relatively low levels, however,
39

Cancer of the Lung, Larynx, and Pleura 339

provided no evidence of an increased risk of lung and PAHs in determining the burden of occu-
cancer (IARC, 2000). pational cancer. The remaining occupational
Underground miners exposed to radioac- lung carcinogens are likely to play a lesser role
tive radon and its decay products, which emit in terms of disease burden.
alpha particles, have been consistently found to
be at increased risk of lung cancer (IARC, 2002). Asbestos
The risk increases with estimated cumulative The first evidence of increased risk of lung can-
exposure and decreased with attained age and cer following inhalation of asbestos fibers dates
time since cessation of exposure (Lubin, 1994). from the 1950s (Boffetta, 2004), and asbestos
A  pooled analysis of 11 cohorts estimated an is the agent responsible for a large number of
apparently linear association, with an approxi- occupationally related lung cancers. All different
mate 6% risk increase per working-​level year of forms of asbestos—​chrysotile and amphiboles,
exposure (Lubin, 1994). There was also evidence including crocidolite, amosite, and tremolite—​
that, for comparable cumulative exposure, the are carcinogenic to the human lung, although the
risk is greater for lower rates over a longer per- potency of chrysotile might be lower than that of
iod (Lubin et al, 1995) and that smoking modifies other types (Berman & Crump, 2008). Although
the carcinogenic effect of radon (Lubin, 1994). asbestos has been banned in many countries, a
Today, the main concern about lung cancer risk substantial number of workers are still exposed,
from radon and its decay products comes from mainly in the construction industry. In many
residential rather than occupational exposure. low-​and medium-​resource countries, occupa-
A  pooled analysis of 13 European case-​control tional exposure is widespread.
studies resulted in an RR of 1.08 (95% CI 1.03–​
1.16) per 100 Bq/​m3 increase in measured indoor Metals
radon (Darby et  al, 2005). After correction for Exposure to inorganic arsenic, a known lung
the dilution caused by measurement error, the carcinogen since the late 1960s, occurs mainly
RR was increased to 1.16 (95% CI 1.05–​1.31) and among workers employed in hot smelting.
the exposure–​response relationship was linear Other groups at increased risk include fur han-
with no evidence of a threshold. A similar anal- dlers, manufacturers of sheep-​dip compounds
ysis of North American studies found similar and pesticides, and vineyard workers (Hayes,
associations (Krewski et  al, 2006). These results 1997). Chromium [VI] compounds increase
suggest that indoor radon exposure might be the risk of lung cancer among chromate-​
an important cause of lung cancer, in particular production workers, chromate-​pigment
among nonsmokers unexposed to occupational manufacturers, chromium platers, and ferro-
carcinogens. chromium producers. No such risk has been
detected among workers exposed only to chro-
Occupation mium [III] compounds.
The important role of specific occupational Studies of nickel miners, smelters, electrol-
exposures in lung cancer etiology has been well ysis workers, and high-​nickel alloy manufac-
established since the 1950s. The risk of lung turers showed an increased risk of lung cancer
cancer is increased among workers employed (Hayes, 1997). There is debate on whether all
in a number of industries and occupations nickel compounds are carcinogenic for humans;
(Table 14-​ 5). The responsible agent(s) have the available evidence does not allow a clear
been identified for several, but not all, of these separation between different nickel salts to
high-​risk workplaces. Evidence for the carci- which workers are exposed. An increased risk
nogenicity of many occupational agents has of lung cancer has been demonstrated among
been reviewed (Siemiatycki, 2006). Two studies workers in cadmium-​based battery manufac-
reported an estimate of the proportion of lung ture, copper-​cadmium alloy workers, and cad-
cancer cases attributable to occupational agents mium smelters. The increased risk does not
in the United Kingdom (14.5% overall; Rushton seem to be attributable to concomitant expo-
et al, 2012) and France (12.5% in men, 6.5% in sure to nickel or arsenic. Studies from the
women; Boffetta et  al, 2010). While asbestos United States showed an excess risk of lung can-
remains the most important occupational lung cer among workers exposed to beryllium in the
carcinogen, there are uncertainties in defining early technological phase of the industry (IARC,
the precise role of silica, radon, heavy metals, 1993)  although the relevance of these results
340

340 Textbook of Cancer Epidemiology

TABLE 14-​5   OCCUPATIONAL AGENTS, GROUPS OF AGENTS, MIXTURES

AND OCCUPATIONS CLASSIFIED AS HUMAN CARCINOGENS (GROUP 1)
BY THE IARC MONOGRAPHS PROGRAM, VOLUMES 1-100, WHICH HAVE
THE LUNG AS TARGET ORGAN

Agents, mixtures, occupations Main industry, use

Agents and groups of agents

Arsenic and inorganic arsenic compounds Glass, metals, pesticides
Asbestos Insulation, filters, textiles
Beryllium and beryllium compounds Aerospace
Bis(chloromethyl)ether and chloromethyl methyl ether Chemical intermediate
Cadmium and cadmium compounds Dye/​pigment
Chromium[VI] compounds Metal plating, dye/​pigment
Involuntary tobacco smoking Hospitality
Nickel compounds Metallurgy, alloy, catalyst
Plutonium Defense
X-​ and γ-​radiation Medical
Radon-​222 and its decay products Mining
Silica, crystalline Stone cutting, mining, glass, paper

Mixtures
Coal-​tar pitch Construction, electrodes
Soot Pigments

Occupations
Aluminum production N/​A
Coal gasification N/​A
Coke production N/​A
Hematite mining (underground) N/​A
Iron and steel founding N/​A
Painting N/​A
Rubber production industry N/​A

* Since that publication, diesel engine exhausts (main uses:  mining and transportation) have been added to the list
(Benbrahim-​Tallaa et al, 2012).

for current exposure circumstances has been evidence of an exposure–​response relationship

debated (Boffetta et al, 2012). (Steenland et al, 2001).

Silica Polycyclic Aromatic Hydrocarbons

An increased risk of lung cancer has been con- PAHs are a complex and important group of
sistently reported in cohorts of silicotic patients chemicals formed during combustion of organic
(Steenland & Stayner, 1997). Many authors material. They are widespread in the human
investigated crystalline silica–​
exposed workers environment. For most people, diet and tobacco
in foundries, pottery making, ceramics, diato- smoke are the main sources of exposure of PAHs,
maceous earth mining, brick making, and stone although a number of occupational settings entail
cutting, some of whom might have developed exposure to high PAH levels. These chemicals,
silicosis. An increased risk of lung cancer was however, occur inevitably as complex mixtures
reported by some, but not all, studies, and in of variable composition; an assessment of the
the positive studies the increase was small, with risk from individual PAHs is therefore difficult.
341

Cancer of the Lung, Larynx, and Pleura 341

An increased risk of lung cancer has been dem- relationship has been shown, together with an
onstrated in several industries and occupations interactive effect of tobacco smoking.
entailing exposure to PAHs such as aluminum Several studies have assessed lung cancer risk
production, coal gasification, coke production, among regular users of aspirin and other nonste-
iron and steel founding, tar distillation, roofing, roidal anti-​inflammatory drugs. A meta-​analysis
and chimney sweeping (Bosetti et al, 2007). An of 15 studies resulted in a pooled RR of 0.86 (95%
increased risk has also been suggested in other CI 0.76–​0.98) (Oh et al, 2011). There was however
industries, including shale oil extraction, wood heterogeneity among the different studies, likely
impregnation, road paving, carbon black produc- due in part to differences in the definition of the
tion, and carbon electrode manufacture, with an exposure. The protective effect was stronger for
exposure–​response relationship in studies with case-​control studies (RR 0.74, 95% CI 0.57–​0.99)
detailed exposure information. Motor vehicle than cohort studies (RR 0.97, 95% CI 0.87–​1.08),
and other engine exhausts represent an impor- suggesting a role for recall bias. In particular, a
tant group of mixtures of PAHs, since they con- large cohort study of 1 million US volunteers did
tribute significantly to air pollution. The available not report a reduction in risk (Thun et al, 1991).
epidemiologic evidence shows an excess risk However, in a meta-​analysis of eight aspirin tri-
among workers with high occupational expo- als, the risk of lung cancer was reduced in the first
sure to diesel engine exhaust (Benbrahim-​Tallaa 10 years after the end of the trial (RR 0.68, 95%
et al, 2012). CI 0.50–​0.92) (Rothwell et al, 2011).

Medical Conditions and Treatment Indoor Air Pollution

In addition to tuberculosis and lung fibrosis from Indoor air pollution is thought to be the main
chronic exposure to high levels of fibers and dusts determinant of the elevated risk of lung cancer
(both discussed in previous sections), chronic experienced by nonsmoking women living in
respiratory diseases have been associated with several regions of China and other Asian coun-
lung cancer risk. Patients with chronic bronchitis tries. The evidence is stronger for coal burning
and emphysema are at moderately increased risk, in poorly ventilated houses, but also burning of
with greater risk for squamous cell carcinoma, wood and other solid fuels as well as fumes from
after adjustment for tobacco smoking (Gao et al, high-​temperature cooking using unrefined veg-
1987; Mayne et  al,1999; Wu et  al, 1995). The etable oils such as rapeseed oil (IARC, 2010).
roles of shared exposures, namely tobacco smok- A positive association between various indicators
ing, and that of chronic inflammation, have not of indoor air pollution and lung cancer risk has
been fully elucidated. A  meta-​analysis of stud- also been reported in populations exposed to less
ies of asthma and lung cancer in never smokers extreme conditions than those encountered by
resulted in a summary RR of 1.8 (95% CI 1.3–​2.3) some Chinese women, for example populations
(Santillan et  al, 2003); the results were similar in central and eastern Europe (Lissowska et  al,
when the analysis was restricted to studies that 2005) and other regions (Hosgood et al, 2010).
controlled for smoking. Since the evidence is
mainly based on case-​control studies, however, Outdoor Air Pollution
selection and recall bias cannot be fully excluded There is abundant evidence that lung cancer
(Rosenberger et al, 2012). rates are higher in cities than in rural settings
The risk of lung cancer is increased in patients (Katsouyanni & Pershagen, 1997). This pattern,
surviving other tobacco-​and lifestyle-​ related however, might result from confounding by other
cancers (Hemminki & Li, 2003). Commonality factors, notably tobacco smoking and occupa-
of risk factors, long-​term effects of radiother- tional exposures, rather than from air pollution.
apy, and increased susceptibility probably inter- Cohort and case-​control studies are limited by
act in the causation of second primary cancers. difficulties in assessing past exposure to the rele-
The effect of chemotherapy and radiotherapy vant air pollutants. The exposure to air pollution
on the risk of a second primary lung cancer has has been assessed either on the basis of proxy
been extensively investigated among long-​term indicators—​for example, the number of inhab-
survivors of breast cancer, 2% to 9% of whom itants in the community of residence, residence
develop lung cancer (Daly, 1999). The increased near a major pollution source—​or on the basis
risk is restricted to patients receiving radiother- of actual data on pollutant levels. These data,
apy. Among them, a clear exposure–​ response however, reflect mainly present levels or levels
342

342 Textbook of Cancer Epidemiology

TABLE 14-​6   RESULTS OF STUDIES OF EXPOSURE TO PM 2.5

AND RISK OF LUNG CANCER

Study; country, N cases RR 95% CI Exposure Basis for exposure

period (reference) contrast assessment

AHSMOG; USA, 17 2.23 0.56–​8.94 per 24.3 µg/​m3 Residential history and
nonsmokers, PM2.5 local pollutant estimates,
1977–​1992 1966–​1992
(McDonnell et al,
2000)

ASC CPS-​II; USA, NA 1.14 1.04–​1.23 per 10 µg/​m3 City of residence in 1982;
1982–​2000 (Krewski PM2.5 pollutant averages of
et al, 2009) 1979–​1983

Six cities; USA, 226 1.27 0.96–​1.69 per 10 µg/​m3 City of residence in
1975–​1998 (Laden PM2.5 1975; pollutant average
et al, 2006) 1979–​1985

ESCAPE; Europe, 2,095 1.18 0.96–​1.46 per 5 µg/​m3 Place of residence at

1990s (Raaschou-​ PM2.5 enrollment; pollutant
Nielsen et al, 2013) ­average 2008–​2011

in the recent past and refer to total suspended disease, caused by chronic arsenic poisoning—​
particulates, sulfur oxides, and nitrogen oxides, Japan, United States, and Chile (IARC, 2004a).
which are not likely to be the agents responsible An exposure–​response relationship was observed
for the carcinogenic effect, if any, of air pollution. in most of these studies. In particular, in a cohort
Furthermore, the sources of data might cover study from a contaminated area in Taiwan, the
quite a wide area, masking small-​scale differ- relative risk of lung cancer according to cumu-
ences in exposure levels. lative estimated exposure to arsenic from drink-
The combined evidence suggests that urban ing water was 4.0 for 20 or more milligrams per
air pollution might entail a small excess risk of liter of average drinking water contamination
lung cancer on the order of 50%, but residual compared to uncontaminated water (Chiou
confounding cannot be excluded. In four cohort et al, 1995).
studies, assessment of exposure to fine particles
was based on environmental measurements C A N C E R O F T H E   L A RY N X
(Table 14-​6). The results of these studies are sug-
More than 90% of cancers of the larynx are squa-
gestive of a small increase in risk among people mous cell carcinomas, and the majority originate
classified as most highly exposed to air pollu- from the supraglottic and glottic regions of the
tion. The International Agency for Research on organ. There are large sex differences, with inci-
Cancer (IARC) classifies outdoor air pollution dence in women is below 1 per 100,000 in most
as an established lung carcinogen in humans populations (Figs. 14-​6, 14-​7,), and an incidence
(Loomis et al, 2014). in men of 10 per 100,000 in southern and cen-
tral Europe, Brazil, Uruguay, and Argentina, as
Drinking Water Contamination well as among African Americans in the United
An increased risk of lung cancer has been consist- States (Figs. 14-​8, 14-​9,), while the lowest rates
ently reported among people exposed to arsenic (1/​100,000) are recorded in Southeast Asia and
in drinking water. Investigations include ecologic central Africa (Forman et  al, 2014). An esti-
studies from Argentina, Chile, and Taiwan, and mated 157,000 new cases occurred worldwide in
case-​control and cohort studies from Taiwan—​ 2012, of which 138,000 were among men (Ferlay
in particular, in areas endemic for black foot et  al, 2013). Incidence rates have not changed
34

Cancer of the Lung, Larynx, and Pleura 343

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 0,2 0,4 0,6 0,8 1 1,2

ASR (world) per 100,000 person-years

FIGURE 14-​6  Age-​standardized (to the 2012 world population) incidence rates of laryngeal cancer among women.
Source: Ferlay et al, 2013.

markedly during the last two decades. The esti- a beneficial effect of quitting smoking. Smoking
mated number of deaths due to laryngeal cancer black tobacco cigarettes entails a stronger risk
in 2012 was 83,000. than smoking blond tobacco cigarettes. Studies
Most cases of laryngeal cancer in Western from India have also reported an effect of chew-
countries are attributable to tobacco smoking, ing tobacco-​containing products. The effect of
alcohol drinking, and the interaction between alcohol is stronger for supraglottic tumors than
these two factors (Hashibe, 2010). The effect of for tumors in other sites. It is not clear, how-
tobacco, with risks in smokers on the order of 10 ever, whether different alcoholic beverages
relative to nonsmokers, seems to be stronger for exert a different carcinogenic effect (Boffetta &
supraglottic than for glottic neoplasms (Hashibe Hashibe, 2006).
et al, 2007). Studies in several populations have A protective effect is possibly exerted by
shown an exposure–​response relationship and high intake of fruits and vegetables, although

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 1 2 3 4 5 6 7 8 9
ASR (world) per 100,000 person-years

FIGURE 14-​7  Age-​standardized (to the 2012 world population) incidence rates of laryngeal cancer among men.
Source: Ferlay et al, 2013.
34

344 Textbook of Cancer Epidemiology

New cases per 100,000

7
6

person-years
5
4
3
2
1
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 14-​8  Age-​specific incidence rates of laryngeal cancer among women.

Source: Ferlay et al, 2013.

the evidence regarding specific micronutrients Papillomatosis patients have an increased risk
such as carotenoids and vitamin C is inade- of laryngeal cancer; however, studies aimed at
quate for drawing a conclusion (WCRF/​AICR, assessing the presence of HPV DNA have not
2007). Hot maté drinking has been suggested provided conclusive evidence for a higher prev-
to be a risk factor in studies from Brazil and alence of infection in cases of laryngeal cancer
Uruguay (Goldenberg et  al, 2003). Data con- than in controls (IARC, 2006). The evidence of an
cerning a possible effect of other food items are association between HPV types that are carcino-
not consistent. genic in the genital tract and in the oropharynx,
A possible effect has been suggested for some chiefly types 16 and 18, and laryngeal cancer is
occupational exposures, including nickel, asbes- not conclusive (Li et al, 2013).
tos, strong inorganic acid mists, and ionizing There is no evidence of strong genetic factors
radiation (Siemiatycki et al, 2006). in laryngeal carcinogenesis; however, polymor-
Laryngeal papillomatosis, a condition char- phisms of enzymes implicated in the metabolism
acterized by multiple benign tumors called of alcohol and tobacco, such as alcohol and alde-
papillomas, is caused by infection with human hyde dehydrogenases, may act as susceptibility
papilloma virus (HPV) types 6 and 11, the same factors (Hashibe et al, 2008).
types that cause genital condylomata acuminata.
Infection in children occurs in both sexes during CANCER OF THE PLEURA
delivery. In adults, infection is common among The most important malignant neoplasm
men and may occur via orogenital sexual contact. arising from the pleura is mesothelioma.
New cases per 100,000

50
person-years

40
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 14-​9  Age-​specific incidence rates of laryngeal cancer among men.

Source: Ferlay et al, 2013.
345

Cancer of the Lung, Larynx, and Pleura 345

Occurrence of mesothelioma has been linked suppressor gene on 3p21, are associated with a
conclusively to asbestos exposure, in particu- cancer syndrome characterized by mesothelioma
lar amphiboles such as crocidolite and amosite. as well as cutaneous and uveal melanomas (Testa
Past occupational exposure to asbestos is the et al, 2011; Carbone et al, 2012). Somatic muta-
main determinant of the descriptive epidemi- tions of BAP1 are also reported in more than
ology of pleural mesothelioma. High-​exposure 20% of mesothelioma (Bott et  al, 2011). Other
industries include mining, shipyard working, common somatic alterations are CDKN2A dele-
and asbestos cement manufacture (Boffetta & tion (72%) and NF2 loss (40%) (Illei et al, 2003;
Stayner, 2006). Thurneysen et al, 2009).
Despite a widespread reduction or ban of
asbestos use, the incidence of mesothelioma CONCLUSION
is still increasing in most populations, which Given the poor prognosis of lung cancer, pri-
reflects the long latency of the disease (Peto mary prevention remains the main weapon
et al, 1995). Time since first exposure to asbes- against this cancer, and tobacco control efforts
tos is the main determinant of mesothelioma are by far the most important preventive meas-
risk, and recent exposures have a limited effect ure. While the effects of tobacco control on
(La Vecchia & Boffetta, 2012). In the absence the incidence of the disease can be appreciated
of occupational exposure to asbestos, incidence in several populations, much remains to be
rates on the order of 0.1–​0.2 per 100,000 are done, in particular among women, and in low-​
estimated in both sexes. There is evidence of an income countries. Control of exposure to other
increased risk of pleural mesothelioma follow- lung carcinogens, in both the general and the
ing environmental exposure to asbestos:  rates occupational environment, is another meas-
in polluted areas have been reported to be 2-​to ure that has been taken and, at least in some
10-​fold higher than those in nonpolluted areas instances, has had substantial effects. Priorities
(Boffetta, 2004). Epidemics of mesothelioma for the prevention of lung cancer, in addition
have been reported from areas with exposure to tobacco control, include understanding the
to naturally occurring asbestos or asbestos-​like carcinogenic and preventive effects of dietary
fibers, such as some districts of central Turkey, and other lifestyle factors, control of occupa-
where erionite, a fibrous substance similar to tional exposures, avoidance of high exposure
amphibole asbestos, contaminated the materi- to outdoor and indoor pollution, and elucida-
als used for building construction (Artvinli & tion of conditions entailing increased genetic
Baris, 1979). Similar clusters of mesothelioma predisposition.
from naturally occurring asbestos or asbestos-​ Lung cancer in never smokers is not a rare
like fibers have been reported from other areas, disease. Occupational factors, passive smoking,
including New Caledonia and Corsica (France), and indoor exposure to radon explain a pro-
Montana (United States), and Yunnan province portion of these cases, and nutritional, infec-
(China). tious, and genetic factors are currently receiving
A role of Simian virus 40 has been sug- attention as additional risk factors. Lung cancer
gested based on detection of serum antibodies was the most important epidemic of the 20th
in mesothelioma cases (Mazzoni et  al, 2012). century, and it will likely remain a major pub-
However, a causal role has not been confirmed lic health problem in the 21st century also. It is
in analytical studies (Boffetta, 2007). Exposure ironic that this cancer causes more deaths than
to ionizing radiation entails an increased risk any other malignancy in the world, even though
of pleural mesothelioma, as it has been shown epidemiology has led to the identification of
in cohorts of patients treated with thorotrast, a more than 10 causes of the disease, including
radiological contrast medium that was used in the quantitatively dominant tobacco smoking.
the 1930s to 1950s (Boffetta & Stayner, 2006). Lung cancer is also a paradigm of the superior-
Tobacco smoking, alcohol drinking, and diet ity of prevention over treatment and a reminder
do not appear to be risk factors for pleural that scientific knowledge is not sufficient per
mesothelioma. se to ensure human health. A  summary of the
Genetic factors play a role in pathogene- current state of knowledge for lung cancer risk
sis of mesothelioma. Germline mutations in factors (by histopathologic type) is presented in
BRCA1 Associated Protein (BAP1) gene, a tumor Table 14-​7.
346

TABLE 14-​7   EPIDEMIOLOGY OF LUNG CANCER: RISK

FACTOR SUMMARY*

Risk factor Direction of effect

Well confirmed risk factors

Cigarette smoking ↑↑
Genetic susceptibility ↑
Family history ↑
Radon exposure ↑
Ionizing radiation ↑
Secondhand/​passive smoking ↑

Probable relationship exists, based on substantial data

Contaminated drinking water (arsenic) ↑
Occupational exposures: Asbestos, diesel engine exhaust, 4-​chloro-​ ↑
ortho-​toluidine, polycyclic aromatic hydrocarbons, coal tar pitch, soot,
­tetrachloroethylene, silica, metal working fluids
Occupations and industries: Fur handlers; pesticide, dye, chromate-​ ↑
pigment, and dye manufacturers; nickel miners, smelters, electrolysis
­workers, high-​nickel alloy manufacturers, cadmium-​related industries,
printing, textile
Pipes, cigars ↑

Weak, if any relationship exists, based on substantial data

Smokeless tobacco ↑

Inconsistent findings or limited evidence to date

Fruits and vegetables ↓
Height ↑
Infections (tuberculosis, Chlamydia pneumoniae, human papilloma virus, ↑
Simian virus 40, Microsporum canis)
Medications: NSAIDS ↓
Alcohol ↑
Lipid intake -​
Carotenoids -​
Isothiocyanates -​
Micronutrients (Selenium, vitamin A, vitamin D, lutein, lycopene) -​
Hormone replacement therapy -​
Obesity -​
Other diet (cereals, meat, poultry, eggs, tea, coffee) -​

*Two arrows indicate a consistent and strong direction of effect, one arrow indicates a consistent direction of effect, and a dash
indicates no consistent effect.
347

Cancer of the Lung, Larynx, and Pleura 347

REFERENCES Berman DW, Crump KS. Update of potency factors

Adami HO, Persson I, Hoover R, Schairer C, Bergkvist for asbestos-​related lung cancer and mesotheli-
L. Risk of cancer in women receiving hormone oma. Crit Rev Toxicol 2008;38 Suppl 1:1–​47.
replacement therapy. Int J Cancer 1989;44:833–​9. Blackman JA, Coogan PF, Rosenberg L, Strom
Ahn MJ, Won HH, Lee J, Lee ST, Sun JM, Park YH BL, Zauber AG, Palmer JR et  al. Estrogen
et al. The 18p11.22 locus is associated with never replacement therapy and risk of lung cancer.
smoker non-​small cell lung cancer susceptibility in Pharmacoepidemiol Drug Safety 2002;11:561–​7.
Korean populations. Hum Genet 2012;131:365–​72. Boffetta P. Occupational factors of lung cancer. In
Alavanja MC, Brownson RC, Boice JD, Hock Hirsch A, Martin JP, Masse R (eds.), Prevention
E. Preexisting lung disease and lung cancer of Respiratory Diseases (pp. 37–​64). New  York,
among nonsmoking women. Am J Epidemiol NY: Marcel Dekker, 1993.
1992;136:623–​32. Boffetta P. Involuntary smoking and lung cancer.
Albanes D. β-​Carotene and lung cancer: a case study. Scand J Work Environ Health 2002;30–​40.
Am J Clin Nutri 1999;69:1345s–​50s. Boffetta P. Epidemiology of environmental and occu-
Amos CI, Gorlov IP, Dong Q, Wu X, Zhang H Lu EY pational cancer. Oncogene 2004;23:6392–​403.
et  al. Nicotinic acetylcholine receptor region on Boffetta P. Epidemiology of peritoneal mesotheli-
chromosome 15q25 and lung cancer risk among oma: a review. Ann Oncol 2007;18:985–​90.
African Americans:  a case-​control study. J Natl Boffetta P, Aagnes B, Weiderpass E, Andersen A.
Cancer Inst 2010;102:1199–​1205. Smokeless tobacco use and risk of cancer of
Amos CI, Wu X, Broderick P, Gorlov IP, Gu J, Eisen T the pancreas and other organs. Int J Cancer
et al. Genome-​wide association scan of tag SNPs 2005;114:992–​5.
identifies a susceptibility locus for lung cancer at Boffetta P, Hashibe M. Alcohol and cancer. Lancet
15q25.1. Nat Genet 2008;40:616–​22. Oncol 2006;7:149–​56.
Aoki K. Excess incidence of lung cancer among pul- Boffetta P, Autier P, Boniol M, Boyle P, Hill C, Aurengo
monary tuberculosis patients. Jpn J Clin Oncol A et al. An estimate of cancers attributable to occu-
1993;23:205–​20. pational exposures in France. J Occup Environ
Artvinli M, Baris YI. Malignant mesotheliomas in a small Med 2010;52:399–​406.
village in the Anatolian region of Turkey: an epide- Boffetta P, Fryzek JP, Mandel JS. Occupational expo-
miologic study. J Natl Cancer Inst 1979;63:17–​22. sure to beryllium and cancer risk:  a review of
Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, the epidemiologic evidence. Crit Rev Toxicol
Fedirko V et  al. Alcohol consumption and site-​ 2012;42:107–​18.
specific cancer risk: a comprehensive dose-​response Boffetta P, Stayner LT. Pleural and Peritoneal Neoplasms.
meta-​analysis. Br J Cancer 2015;112:580–​93. In Schottenfeld D, Fraumeni JFJr (eds.), Cancer
Bailey-​Wilson JE, Amos CI, Pinney SM, Petersen GM, Epidemiology and Prevention, 3rd ed. New  York,
de Andrade M, Wiest JS et al. A major lung cancer Oxford University Press, 2006: 659–​73.
susceptibility locus maps to chromosome 6q23-​ Boffetta P, Trédaniel J, Greco A. Risk of childhood can-
25. Am J Hum Genet 2004;75:460–​74. cer and adult lung cancer after childhood expo-
Bailey-​Wilson JE, Sellers TA, Elston RC, Evens CC, sure to passive smoke:  a meta-​analysis. Environ
Rothschild H. Evidence for a major gene effect Health Persp 2000;108:73.
in early-​onset lung cancer. J La State Med Soc Boice JD, Fraumeni JF Jr. Late effects following isoni-
1993;145:157–​62. azid therapy. Am J Public Health 1980;70:987–​9.
Bandera EV, Freudenheim JL, Vena JE. Alcohol con- Boice J, Little, JB. Ionizing radiation. In F. J.  J.
sumption and lung cancer a review of the epide- Schottenfeld D (ed.), Cancer Epidemiology
miologic evidence. Cancer Epidemiol Biomarkers and Prevention, Vol. 2. New  York, NY:  Oxford
Prev 2001;10: 813–​21. University Press, 1996: 319–​354.
Batty G, Shipley M, Langenberg C, Marmot M, Smith Bosetti C, Boffetta P, La Vecchia C. Occupational
GD. Adult height in relation to mortality from 14 exposures to polycyclic aromatic hydrocar-
cancer sites in men in London (UK):  evidence bons, and respiratory and urinary tract can-
from the original Whitehall study. Ann Oncol cers:  a quantitative review to 2005. Ann Oncol
2006;17:157–​66. 2007;18:431–​46.
Benbrahim-​ Tallaa L, Baan RA, Grosse Y, Lauby-​ Bott M, Brevet M, Taylor BS, Shimizu S, Ito T, Wang
Secretan B, El Ghissassi F, Bouvard V et  al. L et al. The nuclear deubiquitinase BAP1 is com-
Carcinogenicity of diesel-​ engine and gasoline-​ monly inactivated by somatic mutations and
engine exhausts and some nitroarenes. Lancet 3p21.1 losses in malignant pleural mesothelioma.
Oncol 2012;13:663–​4. Nat Genet 2011;43:668–​72.
348

348 Textbook of Cancer Epidemiology

Bozic I, Antal T, Ohtsuki H, Carter H, Kim D, Chen S Doll R, Peto R, Boreham J, Sutherland I. Mortality
et al. Accumulation of driver and passenger muta- in relation to smoking: 50 years’ observations on
tions during tumor progression. Proc Natl Acad male British doctors. BMJ 2004;328:1519.
Sci U S A 2010;107;18545–​50. Dong J, Jin G, Wu C, Guo H, Zhou B, Lv J et  al.
Brambilla E, Moro D, Gazzeri S, Brambilla C. Genome-​wide association study identifies a novel
Alterations of expression of Rb, p16(INK4A) susceptibility locus at 12q23.1 for lung squa-
and cyclin D1 in non-​ small cell lung carci- mous cell carcinoma in Han Chinese. PLoS Genet
noma and their clinical significance. J Pathol 2013;9:e1003190.
1999;188:351–​60. Durusoy R, Boffetta P, Mannetje AT, Zaridze D,
Brenner DR, Brennan P, Boffetta P, Amos CI, Spitz Szeszenia‐Dabrowska N, Rudnai P et al. Lung can-
MR, Chen C et al. Hierarchical modeling identi- cer risk and occupational exposure to meat and
fies novel lung cancer susceptibility variants in live animals. Int J Cancer 2006;18:2543–​47.
inflammation pathways among 10,140 cases and Fang JL, Vaca CE. Detection of DNA adducts of acet-
11,012 controls. Hum Genet 2013;132:579–​89. aldehyde in peripheral white blood cells of alcohol
Brenner DR, McLaughlin JR, Hung RJ. Previous lung abusers. Carcinogenesis 1997;18:627–​32.
diseases and lung cancer risk: a systematic review Ferlay J, Bray F, Pisani P, et  al. 2004. Globocan
and meta-​analysis. PLoS One 2011;6:e17479. 2002: Cancer Incidence, Mortality and Prevalence
Calle EE, Rodriguez C, Walker-​Thurmond K, Thun Worldwide, Version 2.0. IARC Cancer Base No.
MJ. Overweight, obesity, and mortality from can- 5. Lyon, IARC Press, 2004: WHO IARC.
cer in a prospectively studied cohort of US adults. Fernandez-​Garcia I, Ortiz-​de-​Solorzano C,
N Engl J Med 2003;348:1625–​38. Montuenga LM. Telomeres and telomerase in lung
Cancer Genome Atlas Research. Comprehensive cancer. J Thorac Oncol 2008;3:1085–​88.
genomic characterization of squamous cell lung Foreman D, Bray F, Steliarova-​ Founcher E, Ferlay
cancers. Nature 2012;489;519–​25. J, Brewster D. Cancer Incidence in Five
Cancer Genome Atlas Research. Comprehensive Continents. IARC Scientific Publication No.
molecular profiling of lung adenocarcinoma. 164. Lyon: International Agency for Research on
Nature 2014;511;543–​50. Cancer, 2014.
Carbone M, Ferris LK, Baumann F, Napolitano A, Freudenheim JL, Ritz J, Smith-​Warner SA, Albanes
Lum CA, Flores EG et  al. BAP1 cancer syn- D, Bandera EV, van den Brandt PA et al. Alcohol
drome: malignant mesothelioma, uveal and cuta- consumption and risk of lung cancer:  a pooled
neous melanoma, and MBAITs. J Transl Med analysis of cohort studies. Am J Clin Nutr
2012;10:179. 2005;82:657–​67.
Chiou HY, Hsueh YM, Liaw KF, Horng SF, Chiang Gabrielson E. Worldwide trends in lung cancer
MH, Pu YS et  al. Incidence of internal can- pathology. Respirology 2006;11:533–​8.
cers and ingested inorganic arsenic:  a seven-​ Gajalakshmi V, Hung RJ, Mathew A, Varghese C,
year follow-​ up study in Taiwan. Cancer Res Brennan P, Boffetta P. Tobacco smoking and
1995;55:1296–​300. chewing, alcohol drinking and lung cancer risk
Clark J, You M. Chemoprevention of lung cancer by among men in southern India. Int J Cancer
tea. Mol Nutr Food Res 2006;50:144–​51. 2003;107:441–​7.
Daly MB. Breast Cancer. In:  Meadows AT, Neugut Galateau-​ Salle F, Bidet P, Iwatsubo Y, Gennetay
AI, Robinson E (ed.) Multiple Primary Cancers. E, Renier A, Letourneux M et  al. Detection of
Philadelphia, PA: Lippincott Williams & Wilkins, SV40-​ like DNA sequences in pleural meso-
1999: 303–​17. thelioma, bronchopulmonary carcinoma and
Darby S, Hill D, Auvinen A, Barros-​Dios JM, Baysson other pulmonary diseases. Dev Biol Stand
H, Bochicchio F et  al. Radon in homes and risk 1997;94:147–​52.
of lung cancer: collaborative analysis of individual Gao YT, Blot WJ, Zheng W, Ersnow AG, Hsu CW,
data from 13 European case-​control studies. BMJ Levin LI et  al. Lung cancer among Chinese
2005;330:223. women. Int J Cancer 1987;40:604–​9.
Devesa SS, Grauman DJ, Blot WJ, Fraumeni JF. Cancer Gauderman WJ, Morrison J L, Carpenter CL, Thomas
surveillance series: changing geographic patterns DC. Analysis of gene-​smoking interaction in lung
of lung cancer mortality in the United States, 1950 cancer. Genet Epidemiol 1997;14:199–​214.
through 1994. J Natl Cancer Inst 1999;91:1040–​50. Gazdar A, Franklin W, Brambilla E, Hainaut P, Yokota
Doll R, Peto R. Cigarette smoking and bronchial carci- J, Harris C. Genetic and molecular alterations.
noma: dose and time relationships among regular In: Pathology and Genetics, Tumors of the Lung,
smokers and lifelong non-​smokers. J Epidemiol Pleura, Thymus and Heart. Lyon:  IARC Press,
Commun Health 1978;32:303–​13. 2004: 21–​23.
349

Cancer of the Lung, Larynx, and Pleura 349

Goldenberg D, Golz A, Joachims HZ. The beverage of the International Lung Cancer Consortium.
maté: a risk factor for cancer of the head and neck. Environ Health Perspect 2010;118:1743–​7.
Head Neck 2003;25:595–​601. Hosgood HD 3rd, Wang WC, Hong YC, Wang JC,
Greenwald P. β-​carotene and lung cancer:  a lesson Chen K, Chang IS et al. Genetic variant in TP63
for future chemoprevention investigations? J Natl on locus 3q28 is associated with risk of lung ade-
Cancer Inst 2003;95:E1. nocarcinoma among never-​smoking females in
Gunnell D, May M, Ben-​Shlomo Y, Yarnell J, Smith Asia. Hum Genet 2012;131:1197–​1203.
GD. Height, leg length, and cancer: the Caerphilly Hsiung CA, Lan Q, Hong YC, Chen CJ, Hosgood HD,
Study. Nutr Cancer 2003;47:34–​9. Chang IS et al. The 5p15.33 locus is associated with
Gunnell D, Okasha M, Smith GD, Oliver S, Sandhu J, risk of lung adenocarcinoma in never-​smoking
Holly J. Height, leg length, and cancer risk: a sys- females in Asia. PLoS Genet 2010;6:e1001051.
tematic review. Epidemiol Rev 2001;23:313–​42. Hu Z, Wu C, Shi Y, Guo H, Zhao X, Yin Z et  al. A
Hackshaw AK, Law MR, Wald NJ. The accumulated genome-​ wide association study identifies two
evidence on lung cancer and environmental new lung cancer susceptibility loci at 13q12.12
tobacco smoke. BMJ 1997;315:980–​8. and 22q12.2 in Han Chinese. Nat Genet
Hashibe M. Risk factors:  tobacco and alcohol. 2011;43:792–​6.
In:  Olshan AF (ed.) Epidemiology, Pathogenesis, Illei PB, Rusch VW, Zakowski MF, Ladanyi M.
and Prevention of Head and Neck Cancer. Homozygous deletion of CDKN2A and codele-
New York: Springer, 2010: 65–​86. tion of the methylthioadenosine phosphorylase
Hashibe M, Boffetta P, Zaridze D, Shangina gene in the majority of pleural mesotheliomas.
O, Szeszenia-​ Dabrowska N, Mates D et  al. Clin Cancer Res 2003;9:2108–​13.
Contribution of tobacco and alcohol to the high IARC. Cadmium. In: Beryllium, Cadmium, Mercury
rates of squamous cell carcinoma of the supraglot- and Exposures in the Glass Manufacturing
tis and glottis in Central Europe. Am J Epidemiol Industry. IARC Monographs on the Evaluation
2007;165:814–​20. of Carcinogenic Risks to Humans, Vol. 58. Lyon,
Hashibe M, McKay JD, Curado MP, Oliveira JC, France: IARC, 1993: 119–​238.
Koifman S, Koifman R et al. Multiple ADH genes IARC. Household Use of Solid Fuels and High-​
are associated with upper aerodigestive cancers. Temperature Frying. IARC Monographs on the
Nat Genet 2008;40:707–​9. Evaluation of Carcinogenic Risks to Humans. Vol.
Hayes RB. The carcinogenicity of metals in humans. 95. Lyon, France: IARC, 1994.
Cancer Causes Control 1997;8:371–​85. IARC. Lung. In:  Ionizing Radiation. Part  1. X-​and
Hemminki K, Li X. Familial and second primary pan- Gamma (y)-​ Radiation, and Neutrons. IARC
creatic cancers: a nationwide epidemiologic study Monographs on the Evaluation of Carcinogenic
from Sweden. Int J Cancer 2003;103:525–​30. Risks to Humans, Vol. 75. Lyon, France:  IARC,
Henley SJ, Flanders WD, Manatunga A, Thun MJ. 2000: 253–​254.
Leanness and lung cancer risk:  fact or artifact? IARC. Arsenic in drinking-​water. In: Some Drinking-​
Epidemiology 2002;13:268–​76. Water Disinfectants and Contaminants, Including
Henley SJ, Thun MJ, Connell C, Calle EE. Two large Arsenic. IARC Monographs on the Evaluation
prospective studies of mortality among men who of Carcinogenic Risks to Humans, Vol. 84. Lyon,
use snuff or chewing tobacco (United States). France: IARC, 2004a: 39–​267.
Cancer Causes Control 2005;16:347–​58. IARC. Involuntary Smoking. In: Tobacco Smoke and
Herr C, Greulich T, Koczulla RA, Meyer S, Zakharkina Involuntary Smoking. IARC Monographs on the
T, Branscheidt M et  al. The role of vitamin D in Evaluation of Carcinogenic Risks to Humans. Vol.
pulmonary disease: COPD, asthma, infection, and 83. Lyon, France: IARC, 2004b: 1191–​1413.
cancer. Respir Res 2011;12:31. IARC. Tobacco smoke. In:  Tobacco Smoke and
Hirayama T. Non-​smoking wives of heavy smokes Involuntary Smoking. IARC Monographs on the
have a higher risk of lung cancer:  a study from Evaluation of Carcinogenic Risks to Humans. Vol.
Japan. Bull World Health Org 2000;78:940–​2. 83. Lyon, France: IARC, 2004c: 51–​1187.
Holick CN, Michaud DS, Stolzenberg-​ Solomon R, International Agency for Research on Cancer.
Mayne ST, Pietinen P, Taylor PR et al. Dietary carot- Household use of solid fuels and high-​temperature
enoids, serum β-​carotene, and retinol and risk of frying. Vol. 95. Lyon, IARC, 2010;95:1–​430.
lung cancer in the alpha-​tocopherol, beta-​carotene Kalandidi A, Katsouyanni K, Voropoulou N, Bastas
cohort study. Am J Epidemiol 2002;56:536–​47. G, Saracci R, Trichopoulos D. Passive smok-
Hosgood HD 3rd, Boffetta P, Greenland S, Lee YC, ing and diet in the etiology of lung cancer
McLaughlin J, Seow A et  al. In-​home coal and among non-​ smokers. Cancer Causes Control
wood use and lung cancer risk: a pooled analysis 1990;1:15–​21.
350

350 Textbook of Cancer Epidemiology

Katsouyanni K, Pershagen G. Ambient air pollution smoke: distinct patterns in never, former, and cur-
exposure and cancer. Cancer Causes Control rent smokers. Cancer Res 2005;65:5076–​83.
1997;8:284–​91. Li X, Gao L, Li H, Gao J, Yang Y, Zhou F et al. Human
Khatri P, Sirota M, Butte AJ. Ten years of pathway papillomavirus infection and laryngeal cancer
analysis:  current approaches and outstanding risk:  a systematic review and meta-​ analysis. J
challenges. PLoS Comput Biol 2012;8:e1002375. Infect Dis 2013;207:479–​88.
Ko YC, Lee CH, Chen MJ, Huang CC, Chang WY, Liddell F. The interaction of asbestos and smoking in
Lin HJ et  al. Risk factors for primary lung can- lung cancer. Ann Occup Hyg 2001;45:341–​56.
cer among non-​smoking women in Taiwan. Int J Lipson D, Capelletti M, Yelensky R, Otto G, Parker
Epidemiol 1997;26: 24–​31. A, Jarosz M et al. Identification of new ALK and
Korte JE, Brennan P, Henley SJ, Boffetta P. Dose-​ RET gene fusions from colorectal and lung cancer
specific meta-​analysis and sensitivity analysis of biopsies. Nat Med 2012;18:382–​4.
the relation between alcohol consumption and Lissowska J, Bardin-​Mikolajczak A, Fletcher T, Zaridze
lung cancer risk. Am J Epidemiol 2002;55:496–​506. D, Szeszenia-​Dabrowska N, Rudnai P et al. Lung
Kreuzer M, Gerken M, Heinrich J, Kreienbrock L, cancer and indoor pollution from heating and
Wichmann HE. Hormonal factors and risk of cooking with solid fuels:  the IARC international
lung cancer among women? Int J Epidemiol multicentre case-​control study in Eastern/​Central
2003;32:263–​71. Europe and the United Kingdom. Am J Epidemiol
Krewski D, Lubin JH, Zielinski JM, Alavanja M, 2005;162:326–​33.
Catalan VS, Field RW et al. A combined analysis Littman AJ, Jackson LA, Vaughan TL. Chlamydia
of North American case-​control studies of resi- pneumoniae and lung cancer:  epidemiologic
dential radon and lung cancer. J Toxicol Environ evidence. Cancer Epidemiol Biomarkers Prev
Health A 2006;69:533–​97. 2005;14:773–​8.
Krewski D, Jerrett M, Burnett RT, Ma R, Hughes E, Liu JZ, Tozzi F, Waterworth DM, Pillai SG, Muglia P,
Shi Y et al. Extended follow-​up and spatial analy- Middleton L et al. Meta-​analysis and imputation
sis of the American Cancer Society study linking refines the association of 15q25 with smoking
particulate air pollution and mortality. Res Rep quantity. Nat Genet 2010;42:436–​40.
Health Eff Inst 2009 May;(140):5–​114; discussion Loomis D, Huang W, Chen G. The International
115–​36. Agency for Research on Cancer (IARC) eval-
Laden F, Schwartz J, Speizer FE, Dockery DW. uation of the carcinogenicity of outdoor air
Reduction in fine particulate air pollution and pollution:  focus on China. Chinese J Cancer
mortality:  Extended follow-​ up of the Harvard 2014;33:189–​96.
Six Cities study. Am J Respir Crit Care Med Lorenzo Bermejo J, Hemminki K. Familial lung can-
2006;173(6):667–​72. cer and aggregation of smoking habits: a simula-
Lam TK, Gallicchio L, Lindsley K, Shiels M, tion of the effect of shared environmental factors
Hammond E, Tao XG et  al. Cruciferous vegeta- on the familial risk of cancer. Cancer Epidemiol
ble consumption and lung cancer risk: a system- Biomarkers Prev 2005;14:1738–​40.
atic review. Cancer Epidemiol Biomarkers Prev Lubin JH, Boice JD Jr, Edling C, Hornung RW, Howe G,
2009;8:184–​95. Kunz E et al. Radon-​exposed underground miners
Lan Q, Hsiung CA, Matsuo K, Hong YC, Seow A, and inverse dose-​rate (protraction enhancement)
Wang Z et  al. Genome-​wide association analy- effects. Health Phys 1995;69:494–​500.
sis identifies new lung cancer susceptibility loci Maemondo M, Inoue A, Kobayashi K, Sugawara S,
in never-​ smoking women in Asia. Nat Genet Oizumi S, Isobe H et al and the North-​East Japan
2012;44:1330–​5. Study Group. Gefitinib or chemotherapy for non-​
Landi MT, Chatterjee N, Yu K, Goldin LR, Goldstein small-​cell lung cancer with mutated EGFR. N
AM, Rotunno M et al. A genome-​wide association Engl J Med 2010;362:2380–​88.
study of lung cancer identifies a region of chro- Maher B. Personal genomes:  the case of the missing
mosome 5p15 associated with risk for adenocarci- heritability. Nature 2008;456:18–​21.
noma. Am J Hum Genet 2009;85:679–​91. Malkin D, Li FP, Strong LC, Fraumeni JF Jr, Nelson
La Vecchia C, Boffetta P. Role of stopping expo- CE, Kim DH et al. Germ line p53 mutations in a
sure and recent exposure to asbestos in the familial syndrome of breast cancer, sarcomas, and
risk of mesothelioma. Eur J Cancer Prev other neoplasms. Science 1990;250:1233–​38.
2012;21:227–​30. Mazzoni E, Corallini A, Cristaudo A, Taronna A,
Le Calvez F, Mukeria A, Hunt JD, Kelm O, Hung RJ, Tassi G, Manfrini M et  al. High prevalence of
Taniere P et  al. TP53 and KRAS mutation load serum antibodies reacting with simian virus 40
and types in lung cancers in relation to tobacco capsid protein mimotopes in patients affected by
351

Cancer of the Lung, Larynx, and Pleura 351

malignant pleural mesothelioma. Proc Natl Acad Peto J, Hodgson JT, Matthews FE, Jones JR. Continuing
Sci U S A 2012;109:18066–​71. increase in mesothelioma mortality in Britain.
Mayne ST, Buenconsejo J, Janerich DT. Previous Lancet 1995;345:535–​39.
lung disease and risk of lung cancer among Pfeifer GP, Denissenko MF, Olivier M, Tretyakova N,
men and women nonsmokers. Am J Epidemiol Hecht SS, Hainaut P. Tobacco smoke carcinogens,
1999;149:13–​20. DNA damage and p53 mutations in smoking-​
McDonnell WF, Nishino-​ Ishikawa N, Petersen FF, associated cancers. Oncogene 2002;21:7435–​51.
Chen LH, Abbey DE. Relationships of mortality Raaschou-​Nielsen O, Andersen ZJ, Beelen R, Samoli
with the fine and coarse fractions of long-​term E, Stafoggia M, Weinmayr G, et  al. Air pollu-
ambient PM10 concentrations in nonsmokers. tion and lung cancer incidence in 17 European
J Expo Anal Environ Epidemiol 2000;10:427–​36. cohorts: prospective analyses from the European
McKay JD, Hung RJ, Gaborieau V, Boffetta P, Chabrier Study of Cohorts for Air Pollution Effects
A, Byrnes G et al. Lung cancer susceptibility locus (ESCAPE). Lancet Oncol 2013Aug;14(9):813–​22.
at 5p15.33. Nat Genet 2008;40:1404–​6. Renehan AG, Zwahlen M, Minder CT, O’Dwyer S,
Meza R, Meernik C, Jeon J, Cote ML. Lung can- Shalet SM, Egger M. Insulin-​ like growth fac-
cer incidence trends by gender, race and histol- tor (IGF)-​I, IGF binding protein-​3, and cancer
ogy in the United States, 1973–​2010. PLoS One risk: systematic review and meta-​regression anal-
2015;10:e0121323. ysis. Lancet 2004;363:1346–​53.
Miki D, Kubo M, Takahashi A, Yoon KA, Kim J, Riely GJ, Politi KA, Miller VA, Pao W. Update on
Lee GK et  al. Variation in TP63 is associated epidermal growth factor receptor mutations
with lung adenocarcinoma susceptibility in in non-​small cell lung cancer. Clin Cancer Res
Japanese and Korean populations. Nat Genet 2006;12:7232–​41.
2010;42:893–​6. Rodriguez C, Patel AV, Calle EE, Jacobs EJ, Chao A,
Mitsudomi T, Morita S, Yatabe Y, Negoro S, Thun MJ. Body mass index, height, and prostate
Okamoto I, Tsurutani J et  al, West Japan cancer mortality in two large cohorts of adult
Oncology Group. Gefitinib versus cisplatin plus men in the United States. Cancer Epidemiol
docetaxel in patients with non-​small-​cell lung Biomarkers Prev 2001;10:345–​53.
cancer harbouring mutations of the epidermal Rosenberger A, Bickeboller H, McCormack V,
growth factor receptor (WJTOG3405): an open Brenner DR, Duell EJ, Tjonneland A et al. Asthma
label, randomised phase 3 trial. Lancet Oncol and lung cancer risk:  a systematic investigation
2010;11:121–​8. by the International Lung Cancer Consortium.
Nakachi K, Limtrakul PN, Sonklin P, Jarern CT, Carcinogenesis 2012;33:587–​97.
Lipigorngoson S, Arai K et al. Risk factors for lung Rossouw JE, Anderson GL, Prentice RL, LaCroix AZ,
cancer among northern Thai women:  epidemio- Kooperberg C, Stefanick ML et al. Risks and ben-
logical, nutritional, serological, and bacteriologi- efits of estrogen plus progestin in healthy post-
cal surveys of residents in high‐and low‐incidence menopausal women:  principal results From the
areas. Cancer Sci 1999;90:1187–​95. Women’s Health Initiative randomized controlled
Nordquist LT, Simon GR, Cantor A, Alberts WM, trial. JAMA 2002;288:321–​33.
Bepler G. Improved survival in never-​smokers vs Rothwell PM, Fowkes FGR, Belch JFF, Ogawa H,
current smokers with primary adenocarcinoma of Warlow CP, Meade TW. Effect of daily aspirin on
the lung. Chest 2004;126:347–​51. long-​term risk of death due to cancer: analysis of
Oh SW, Myung SK, Park JY, Lee CM, Kwon HT. individual patient data from randomised trials.
Aspirin use and risk for lung cancer:  a meta-​ Lancet 2011;377:31–​41.
analysis. Ann Oncol 2011;22:2456–​65. Ruano-​Ravina A, Figueiras A, Freire-​Garabal M, Barros-​
Olsson H, Bladstrom A, Ingvar C. Are smoking-​ Dios J. Antioxidant vitamins and risk of lung can-
associated cancers prevented or postponed in cer. Current Pharm Design 2006;12:599–​613.
women using hormone replacement therapy? Rudin CM, Durinck S, Stawiski EW, Poirier JT,
Obstet Gynecol 2003;102:565–​70. Modrusan Z, Shames DS, et  al. Comprehensive
Peifer M, Fernandez-​Cuesta L, Sos ML, George J, Seidel genomic analysis identifies SOX2 as a frequently
D, Kasper LH et  al. Integrative genome analyses amplified gene in small-​ cell lung cancer. Nat
identify key somatic driver mutations of small-​cell Genet 2012;44:1111–​6.
lung cancer. Nat Genet 2012;44:1104–​10. Rushton L, Hutchings SJ, Fortunato L, Young C, Evans
Peto R, Boreham J, Lopez AD, Thun M, Heath, C. GS, Brown T et al. Occupational cancer burden in
Mortality from tobacco in developed coun- Great Britain. Br J Cancer 2012;107 Suppl 1:3–​7.
tries:  indirect estimation from national vital sta- Saccone NL, Culverhouse RC, Schwantes-​ An TH,
tistics. Lancet 1992;339:1268–​78. Cannon DS, Chen X, Cichon S et  al. Multiple
352

352 Textbook of Cancer Epidemiology

independent loci at chromosome 15q25.1 affect Song YM, Smith GD, Sung J. Adult height and cause-​
smoking quantity:  a meta-​analysis and compar- specific mortality: a large prospective study of South
ison with lung cancer and COPD. PLoS Genet Korean men. Am J Epidemiol 2003;58:479–​85.
2010;6:e1001053. Stayner L, Bena J, Sasco AJ, Smith R, Steenland K,
Santillan AA, Camargo CA Jr, Colditz GA. A Kreuzer M, Straif K. Lung cancer risk and work-
meta-​analysis of asthma and risk of lung can- place exposure to environmental tobacco smoke.
cer (United States). Cancer Causes Control Am J Public Health 2007;97:545.
2003;14;327–​34. Steenland K, Mannetje A, Boffetta P, Stayner L,
Schabath MB, Wu X, Vassilopoulou-​ Sellin R, Attfield M, Chen J et  al. Pooled exposure-​
Vaporciyan AA, Spitz MR. Hormone replacement response analyses and risk assessment for lung
therapy and lung cancer risk: a case-​control analy- cancer in 10 cohorts of silica-​exposed workers: an
sis. Clin Cancer Res 2004;10:113–​23. IARC multicentre study. Cancer Causes Control
Sequist LV, Yang JC, Yamamoto N, O’Byrne K, Hirsh 2001;12:773–​84.
V, Mok T et al. Phase III study of afatinib or cis- Steenland K, Stayner L. Silica, asbestos, man-​made
platin plus pemetrexed in patients with metastatic mineral fibers, and cancer. Cancer Causes Control
lung adenocarcinoma with EGFR mutations. 1997;8:491–​503.
J Clin Oncol 2013;31:3327–​34. Taioli E, Wynder EL. Endocrine factors and adenocar-
Shi J, Chatterjee N, Rotunno M, Wang Y, Pesatori AC, cinoma of the lung in women. J Natl Cancer Inst
Consonni D et al. Inherited variation at chromo- 1994;86:869–​70.
some 12p13.33, including RAD52, influences the Testa JR, Cheung M, Pei J, Below JE, Tan Y, Sementino E
risk of squamous cell lung carcinoma. Cancer et al. Germline BAP1 mutations predispose to malig-
Discov 2012;2:131–​9. nant mesothelioma. Nat Genet 2011;43:1022–​5.
Shigematsu H, Gazdar AF. Somatic mutations of epi- Thomas L, Doyle LA, Edelman MJ. Lung cancer in
dermal growth factor receptor signaling pathway women: emerging differences in epidemiology, biol-
in lung cancers. Int J Cancer 2006;118:257–​62. ogy, and therapy. CHEST Journal 2005;128;370–​81.
Siemiatycki J, Richardson L, Boffetta P. Occupation. In Thorgeirsson TE, Geller F, Sulem P, Rafnar T, Wiste
F. J. J. Schottenfeld D (ed.), Cancer Epidemiology A, Magnusson KP et al. A variant associated with
and Prevention. 3 ed. New  York, NY:  Oxford nicotine dependence, lung cancer and peripheral
University Press, 2006: 322–​354. arterial disease. Nature 2008;452:638–​42.
Sinha R, Kulldorff M, Curtin J, Brown CC, Alavanja Thun MJ, Namboodiri MM, Heath CW Jr. Aspirin use
MC, Swanson CA. Fried, well-​done red meat and and reduced risk of fatal colon cancer. N Engl J
risk of lung cancer in women (United States). Med 1991;325:1593–​96.
Cancer Causes Control 1998;9:621–​30. Thurneysen C, Opitz I, Kurtz S, Weder W, Stahel RA,
Sinha R, Kulldorff M, Swanson CA, Curtin J, Felley-​Bosco E. Functional inactivation of NF2/​
Brownson RC, Alavanja MC. Dietary heterocy- merlin in human mesothelioma. Lung Cancer
clic amines and the risk of lung cancer among 2009;64:140–​7.
Missouri women. Cancer Res 2000;60:3753–​6. Timofeeva MN, Hung RJ, Rafnar T, Christiani DC,
Slebos RJ, Hruban RH, Dalesio O, Mooi WJ, Offerhaus Field JK, Bickeboller H et  al., Transdisciplinary
G, Rodenhuis S. Relationship between K-​ ras Research in Cancer of the Lung Research,
oncogene activation and smoking in adenocar- T. Influence of common genetic variation on lung
cinoma of the human lung. J Natl Cancer Inst cancer risk: meta-​analysis of 14 900 cases and 29
1991;83:1024–​27. 485 controls. Hum Mol Genet 2012;21:4980–​95.
Slebos RJ, Kibbelaar RE, Dalesio O, Kooistra A, Stam Timofeeva MN, McKay JD, Smith GD, Johansson M,
J, Meijer CJ et al. K-​ras oncogene activation as a Byrnes GB, Chabrier A et  al. Genetic polymor-
prognostic marker in adenocarcinoma of the lung. phisms in 15q25 and 19q13 loci, cotinine levels,
N Engl J Med 1990;323:561–​5. and risk of lung cancer in EPIC. Cancer Epidemiol
Smith-​ Warner SA, Ritz J, Hunter DJ, Albanes D, Biomarkers Prev 2011;20:2250–​61.
Beeson WL, van den Brandt PA et al. Dietary fat Toh CK, Gao F, Lim WT, Leong SS, Fong KW, Yap SP
and risk of lung cancer in a pooled analysis of pro- et al. Never-​smokers with lung cancer: epidemio-
spective studies. Cancer Epidemiol Biomarkers logic evidence of a distinct disease entity. J Clin
Prev 2002;11:987–​92. Oncol 2006;24:2245–​51.
Soini Y, Nuorva K, Kamel D, Pollanen R, Vahakangas Toh CK, Lim WT. Lung cancer in never-​smokers. J
K, Lehto VP, Paakko P. Presence of human Clin Pathol 2007;60:337–​40.
papillomavirus DNA and abnormal p53 pro- Trichopoulos D, Kalandidi A, Sparros L, Macmahon
tein accumulation in lung carcinoma. Thorax B. Lung cancer and passive smoking. Int J Cancer
1996;51:887–​93. 1981;27:1–​4.
35

Cancer of the Lung, Larynx, and Pleura 353

Truong T, Hung RJ, Amos CI, Wu X, Bickeboller H, World Cancer Research Fund, American Institute for
Rosenberger A et  al. Replication of lung cancer Cancer Research. Lung. In:  Food, Nutrition and
susceptibility loci at chromosomes 15q25, 5p15, the Prevention of Cancer:  A Global Perspective.
and 6p21: a pooled analysis from the International Part II. Cancers, Nutrition and Food. Washington,
Lung Cancer Consortium. J Natl Cancer Inst DC: American Institute for Cancer Research,
2010;102:959–​71. 1997: 130–​147.
Tyczynski JE, Bray F, Parkin DM. Lung cancer in World Cancer Research Fund, American Institute
Europe in 2000:  epidemiology, prevention, and for Cancer Research. Food, Nutrition, Physical
early detection. Lancet Oncol 2003;4:45–​55. Activity and the Prevention of Cancer: A Global
US Department of Health and Human Services. The Perspective. Washington, DC: American Institute
Health Consequences of Smoking: A Report of the for Cancer Research, 2007.
Surgeon General. Atlanta, GA:  US Department Wu AH, Fontham ET, Reynolds P, Greenberg RS,
of Health and Human Services, Centers for Buffler P, Liff J et  al. Previous lung disease and
Disease Control and Prevention, National Center risk of lung cancer among lifetime nonsmok-
for Chronic Disease Prevention and Health ing women in the United States. Am J Epidemiol
Promotion, Office on Smoking and Health, 1995;141:1023–​32.
62, 2004. Wu AH, Mimi CY, Thomas DC, Pike MC, Henderson
Walsh KM, Amos CI, Wenzlaff AS, Gorlov IP, Sison BE. Personal and family history of lung disease
JD, Wu X et  al. Association study of nicotinic as risk factors for adenocarcinoma of the lung.
acetylcholine receptor genes identifies a novel Cancer Res 1988;8:7279–​84.
lung cancer susceptibility locus near CHRNA1 in Yuan JM, Koh WP, Murphy SE, Fan Y, Wang R,
African-​Americans. Oncotarget 2012;3:1428–​38. Carmella SG et  al. Urinary levels of tobacco-​
Wang Y, Broderick P, Matakidou A, Eisen T, Houlston specific nitrosamine metabolites in relation to lung
RS. Role of 5p15.33 (TERT-​CLPTM1L), 6p21.33 cancer development in two prospective cohorts of
and 15q25.1 (CHRNA5-​CHRNA3) variation and cigarette smokers. Cancer Res 2009;69:2990–​5.
lung cancer risk in never-​smokers. Carcinogenesis Zang EA, Wynder EL. Differences in lung cancer risk
2010;31:234–​8. between men and women: examination of the evi-
Wang Y, Broderick P, Webb E, Wu X, Vijayakrishnan J, dence. J Natl Cancer Inst 1996;88:183–​92.
Matakidou A et al. Common 5p15.33 and 6p21.33 Zhao J, Gupta S, Seielstad M, Liu J, Thalamuthu A.
variants influence lung cancer risk. Nat Genet Pathway-​based analysis using reduced gene sub-
2008;40:1407–​9. sets in genome-​ wide association studies. BMC
Wang Y, McKay JD, Rafnar T, Wang Z, Timofeeva Bioinformatics 2011;12:17.
MN, Broderick P et al. Rare variants of large effect Zheng W, Blot W, Liao M, Wang Z, Levin L, Zhao J
in BRCA2 and CHEK2 affect risk of lung cancer. et al. Lung cancer and prior tuberculosis infection
Nat Genet 2014;46:736–​41. in Shanghai. Br J Cancer 1987;6:501.
354
35

15
Skin Cancer
A D È L E C . G R E E N , C AT H E R I N E M . O L S E N , A N D D AV I D J . H U N T E R

S kin cancer is one of the few types of cancer for

which exposure to the major carcinogen, solar
ultraviolet (UV) radiation, is strongly implicated
Diagnosis Often visual inspection allows prelim-
inary diagnosis, which is confirmed histopatho-
logically by biopsy or excision. A dermatoscope,
on the basis of descriptive epidemiologic data a magnifying video camera, is used frequently as
alone. There are three major forms of skin can- an adjunct to improve diagnostic accuracy.
cer considered in this chapter—​melanoma, basal Treatment Adequate surgical excision is the pri-
cell carcinoma (BCC), and squamous cell carci- mary method of treatment of primary skin can-
noma (SCC)—​and each appears to have different cer, though topical agents and photodynamic
causal relations to the pattern and total amount therapy are used, especially for superficial BCC.
of sun exposure. Although conceptually sun While there is still no curative treatment for mel-
exposure may appear straightforward to meas- anoma that has spread beyond the regional lymph
ure, past sun exposure is a complex variable, and nodes, immunotherapy and several molecularly
questionnaire or interview-​based methods have targeted agents have recently extended the life
limited ability to precisely capture patterns and span of those with advanced melanoma.
duration of exposure that vary with age, region
Prognosis Although fatality from BCC and SCC is
of residence, and calendar period. High-​intensity
low, patients with BCC and SCC require follow-​
UV exposure and long-​term UV exposure appear
up to detect recurrence, regional spread, or sub-
to be involved differentially in the various skin
sequent new primary tumors. The prognosis of
cancers and their subtypes. Underlying molecu-
melanoma is strongly dependent on the thickness
lar mechanisms are becoming better understood,
of the primary tumor, and, since it is increasingly
though many aspects like the cells of origin and
diagnosed at an early stage in high-​and middle-​
the exact roles of intermediate lesions like actinic
income countries, overall 5-​year relative survival
keratoses and nevi remain unclear. Because
rates exceed 80%. Relative survival is higher in
exposure of skin to UV radiation is modifiable,
women than men.
skin cancers are substantially preventable.
Progress Evidence suggests that BCC, SCC, and
CLINICAL SYNOPSIS melanoma are preventable through sun protec-
Subgroups Although subgroups of BCC and mel- tion measures. The substantial improvement in
anoma can be readily identified, descriptive and prognosis of melanoma patients overall is due
analytic epidemiologic studies mostly consider to diagnosis at early, curable stages through
each of them as single entities. Actinic keratoses increased general awareness, and to recent thera-
(AKs or solar keratoses) are benign markers of peutic advances for advanced melanoma patients.
increased risk of all three main forms of skin can-
cer, but especially SCC, for which they may be a DESCRIPTIVE
premalignant lesion. EPIDEMIOLOGY
Symptoms A  nonhealing skin lesion or nodule A major problem in the descriptive epidemiol-
is the predominant symptom of BCC and SCC, ogy of keratinocyte cancers (KCs) (namely, BCC
while melanoma usually appears as a new or and SCC) is that they are not registered by most
changing pigmented mole on the skin. Rarely cancer surveillance systems. This is because the
melanomas are first detected as metastases to numbers of primary KCs newly diagnosed each
lymph nodes or other organs. year would place a major burden on registries
356

356 Textbook of Cancer Epidemiology

and because they are perceived as less important from 1978 to 2007 significantly higher in those
than other cancers with higher fatality rates. Thus under 40  years, especially women, compared to
incidence estimates are rarely based on regional older people, against steep increases in BCC and
cancer registry data and mostly are derived from SCC overall (Birch-​Johansen et al, 2010). In the
ad hoc surveys or treatment records of geo- United States the age-​adjusted skin cancer pro-
graphically defined populations. For example in cedure rate in Medicare beneficiaries has also
the United States, some 5  million cases of BCC increased from 2006 to 2012 for BCC and SCC
and SCC were treated in over 3  million people (Rogers et al, 2015), and indeed overall, incidence
in 2012 based on Medicare fee-​for-​service pop- of both BCC and SCC appears to be increasing in
ulation and National Ambulatory Medical Care populations of European heritage (Lomas et  al,
Survey data (Rogers et  al, 2015). Studies show 2012). Incidence rates of AK are much higher
incidence rates of KC worldwide are highest at than for skin cancer but difficult to estimate due
more than 1,000 ⁄100,000 person-​years for BCC to their high rate of regression (Frost et al, 2000).
in white-​skinned populations of European her- AK prevalence increases with age and decreases
itage who live close to the Equator such as in with latitude. A  very small proportion of AKs
Australia, and rarest in dark-​ skinned popula- transform into SCC (Green, 2015).
tions of African and Asian heritage (Lomas et al,
2012). Specifically, the most recent national sur- Melanoma
vey in Australia in 2002 showed a cumulative risk Melanoma rates are also much higher in the
of at least one KC of 70% for men to age 70 years predominantly white-​ skinned populations of
and 58% for women (Staples et al, 2006). While Europe and the United States than in Africa or
rates had increased since 1985 for persons aged Asia (Figs. 15-​1, 15-​2), with the highest rates seen
60+ years, rates in younger age groups had stabi- in Australia and New Zealand (Erdmann et  al,
lized (Staples et al, 2006). Based on a more recent 2013; Whiteman et  al, 2016). In many popula-
analysis of Medicare Australia treatment data tions, rates among women are comparable to or
between 2000 and 2011, the differences in age-​ higher than in men (Figs. 15-​3, 15-​4). Incidence
specific trends in KC have continued with annual of cutaneous melanoma increased steadily in the
increases in excision rates highest for men aged second half of the 20th century worldwide, and in
75 to 84  years (8.6% per annum), but declin- an analysis of incidence data from 39 population-​
ing significantly in men and women aged under based cancer registries for 2000–​2002, incidence
45 (Olsen et  al, 2014a). Similar trends for BCC rates of melanoma have continued to rise in
have been seen in other populations such as in most European countries (particularly southern
British Columbia, Canada, over a 40-​year period and eastern Europe), whereas in Australia, New
to 2003 (McLean et al, 2012) while in Denmark Zealand, North America, Israel, and Norway
the opposite has been reported with the aver- rates appear to be stabilizing in adults under
age annual percentage change in BCC incidence 50 (Erdmann et  al, 2013). A  detailed analysis

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 5 10 15 20 25
ASR (world) per 100,000 person-years

FIGURE 15-​1  Age-​standardized (to the 2012 world population) incidence rates of melanoma among women.
Source: Ferlay et al, 2013.
357

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 5 10 15 20
ASR (world) per 100,000 person-years

FIGURE 15-​2  Age-​standardized (to the 2012 world population) incidence rates of melanoma among men.
Source: Ferlay et al, 2013.

80
New cases per 100,000

70
person-years

60
50
40
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 15-​3  Age-​specific incidence rates of melanoma among women.

Source: Ferlay et al, 2013.
New cases per 100,000

160
140
person-years

120
100
80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 15-​4  Age-​specific incidence rates of melanoma among men.

Source: Ferlay et al, 2013.
358

358 Textbook of Cancer Epidemiology

of melanoma incidence trends and projections cell growth and differentiation (Bale & Yu, 2001),
of rates to 2031 in six populations based on including the Gli family of transcription factors
long-​term registry data (Whiteman et al, 2016), (High & Zedan, 2005).
showed marked increases in US whites, the
United Kingdom, Sweden, and Norway between DNA Repair Defects
1982 and 2011, with rates projected to continue Xeroderma pigmentosum is a rare autosomal
rising. Incidence in Australia peaked in 2005 and recessive disorder that is part of an expanding
has since been declining, while melanoma inci- family of nucleotide excision repair defect dis-
dence in New Zealand was still increasing but eases (Cleaver, 2005). Manifested as an extreme
projected to soon decline. A  US study similarly photosensitivity to UV radiation resulting from
showed that incidence rates in whites were pro- a defect in any one of at least eight genes that
jected to increase to 2019 while death rates were regulate excisional repair of UV-​damaged DNA,
projected to remain stable (Guy et al, 2015). xeroderma pigmentosum incidence varies from 1
in 20,000 in Japan to 1 in 250,000 people in the
GENETIC AND MOLECULAR United States and Europe (Lehmann et al, 2011).
EPIDEMIOLOGY Although common in these patients, skin cancer
is not common in some other diseases with nucle-
Inherited Susceptibility otide excision repair defects such as Cockayne
Basal Cell Carcinoma syndrome, despite the sun sensitivity in both
Much insight into the pathogenesis of BCC has disorders. Hence additional chromosome insta-
been obtained from the study of patients with bility as seen in xeroderma pigmentosum may
Gorlin syndrome (nevoid BCC syndrome), be required for cancer development (Cleaver,
an autosomal dominant disorder character- 2005). Skin tumors in xeroderma pigmento-
ized by the development of multiple BCCs at sum patients show high levels of RAS oncogene
an early age, as well as diverse internal tumors activation, Ink4a-​Arf and TP53 tumor suppres-
and other phenotypic abnormalities including sor gene modifications, and aberrations of the
pitting of the palms and soles and dental and sonic hedgehog pathway (Daya-​ Grosjean &
brain malformations (Gorlin, 2004). Gorlin Sarasin, 2005), and UV-​specific mutations of the
syndrome patients develop BCCs as early as smoothened gene in BCCs are three times higher
2  years of age, with a clear increase in tumor in xeroderma pigmentosum patients than in
numbers between puberty and 35 years of age. those with sporadic BCCs (Couve-​Privat et  al,
The most common site of their initial appear- 2002). Although mutations in the sonic hedgehog
ance is the nape of the neck; tumors are rarely gene are relatively rare in sporadic BCCs, they are
found below the waist. Among white-​skinned found in 15% of BCCs from xeroderma pigmen-
Gorlin syndrome patients 90% develop BCC, tosum patients (Couve-​Privat et al, 2004).
but BCC develops in only approximately 40% of Two genome-​ wide association studies
black patients (Goldstein et al, 1994), thus sun (GWAS) have identified other susceptibility loci,
exposure is likely to play a role in addition to including a variant at 9q21 (containing both
the genetic component in BCCs associated with the CDKN2A and CDKN2B genes); variants in
this syndrome. the pigmentation genes ASIP, TYR, SLC45A2I
Gorlin syndrome patients carry mutations and MC1R, 1p36, 1q42, and TERT-​CLPTM1L
in the patched gene, which acts as a tumor sup- (Gudbjartsson et  al, 2008; Stacey et  al, 2008;
pressor gene. As predicted by the classic two-​hit Rafnar et al, 2009; Stacey et al, 2009; Nan et al,
model, one defective copy of this gene is inher- 2011a); and two novel susceptibility loci at TGM3
ited but tumors arise after inactivation of the and RGS22 (Stacey et al, 2014).
remaining allele (Tilli et  al, 2005). The patched
gene product is part of a receptor for the sonic Detoxifying Proteins
hedgehog protein involved in embryonic devel- The enzyme family glutathione-​ S-​
transferase
opment. When sonic hedgehog binds to patched is part of the skin’s defense mechanism against
it releases smoothened, a transmembrane sig- UV-​ induced oxidative stress. Polymorphisms
naling protein. Mutations of the patched and in GSTM1, GSTM3, GSTT1, and GSTP1 in par-
smoothened genes result in upregulation of the ticular appear to be associated with increased
hedgehog signaling pathway and activation of occurrence of BCC in immunosuppressed peo-
downstream target genes that are associated with ple, those with Gorlin syndrome and those
359

Skin Cancer 359

with multiple sporadic BCC (Lear et  al, 1996; approximately 2% of families (Goldstein et  al,
Ramachandran et al, 2000; Fryer et al, 2005; Yang 2006). Germline mutations in the BAP1 tumor
et  al, 2006), although no single gene has been suppressor gene occur in <1% of cutaneous
consistently associated with increased risk. and uveal melanoma families (Wiesner et  al,
2011). Other genes implicated in familial mel-
Squamous Cell Carcinoma anoma include MITF, TERT, POT1, ACD, and
Less is known about genetic risk factors for SCC, TERF2IP (Aoude et  al, 2015), each occurring
although some associated genes have been iden- in <1% of melanoma families. TERF2IP, ACD,
tified from hereditary disorders such as xero- and POT1 are genes involved in telomere func-
derma pigmentosum (XPA-​XPG and XPV), tion (Harland et  al, 2016). MITF encodes the
Ferguson-​Smith syndrome (TGFBR1), oculocu- micropthalmia transcription factor, a major reg-
taneous albinism (TYR, OCA2, MATP/​ OCA4, ulator of melanocyte-​ specific gene expression
TYRP1), and epidermodysplasia verruciformis (Steingrimsson et al, 2004). A missense mutation
(EVER1, EVER2) (Nikolaou et al, 2012). Gene in MITF increases protein stability and is asso-
mapping has excluded patched as a causative ciated with an approximately twofold increased
gene but has shown loss of heterozygosity in the risk of melanoma, that is, this is a medium-​
9q22-​q31 region, suggesting that the gene for this penetrance mutation (Yokoyama et  al, 2012;
syndrome is likely to be a tumor suppressor gene Bertolotto et al, 2012).
(Bose et al, 2006). Three loci associated with BCC There is no evidence that people with Li-​
have also been associated with SCC—​two loci at Fraumeni syndrome, characterized by germline
6p25 (near EXOC2) and 13q32 (near UBAC2) mutations in the TP53 gene, have an increased
(Nan et al, 2011a). In a GWAS for SCC, 11 loci risk of melanoma (Platz et al, 1998). Several other
reached genome-​ wide significance; seven pig- inherited cancer syndromes are associated with
mentation-​ associated loci (MC1R, ASIP, TYR, an increased risk of melanoma, including xero-
SLC45A2, OCA2, IRF4, and BNC2) as well as derma pigmentosum patients, who develop cuta-
1q23.3 CADM1, a metastasis suppressor gene; neous melanoma at more than 1,000 times the
2p22.3; 7p21.1 AHR, the dioxin receptor; and rate of the normal population, though a large pro-
9q34.3 SEC16A, a putative oncogene (Chahal et portion are lentigo maligna melanoma (Kraemer
al, 2016). et al, 1994; Spatz et al, 2001). Several other germ-
line defects underlie syndromes whose sufferers
Melanoma have secondarily been reported to be at increased
Approximately 5% to 10% of melanomas occur risk for melanoma. These include mutations
among people with a family history of the disease. in the Werner syndrome (WRN) gene that also
Several high-​risk melanoma susceptibility genes encodes a DNA helicase involved in DNA repair;
have been identified:  CDKN2A, CDK4, TERT, the NF1 gene that causes neurofibromatosis type
POT1, ACD, TERF2IP, and BAP1 (reviewed in I  and affects cells originating from the neural
Read & Hayward, 2016). CDKN2A encodes two crest including melanocytes; and the BRCA2
proteins, p16INK4A (commonly referred to as gene (de Snoo & Hayward, 2005).
p16) and p14ARF (Alternative Reading Frame), Candidate gene studies identified
both of which are tumor suppressors involved intermediate-​and low-​risk melanoma suscepti-
in regulation of the cell cycle (Goldstein et  al, bility genes associated with pigmentation, includ-
2007). Germline CDKN2A mutations have been ing melanocortin-​1 receptor (MC1R) and OCA2
identified in patients with familial melanoma gene variants (Palmer et  al, 2000; Jannot et  al,
in Australia, Europe, North America, and the 2005). Large-​scale GWAS have identified ASIP,
Middle East, however the prevalence varies (20% TYR, SLC45A2, and TYRP1 (Brown et al, 2008;
in Australia, 45% in North America, and 57% in Gudbjartsson et  al, 2008)  also associated with
Europe) (Goldstein et al, 2007). pigmentary traits, and MTAP, PLA2G6, IRF4, and
Other highly penetrant mutations occur in TERT-​CLPTMIL associated with the develop-
the CDK4 gene (Zuo et  al, 1996)  and BRCA1-​ ment of nevi (Falchi et al, 2009; Law et al, 2012).
associated protein 1 (BAP1) (Wiesner et  al, Several other loci identified through GWAS are
2011), although these mutations are rare (de not associated with nevus or pigmentation phe-
Snoo & Hayward, 2005). CDK4 mutations acts notypes (PARP1, CASP8, ATM, CCND1, VDR,
as oncogenes and prevent binding of CDK4 pro- and MX2 (Barrett et al, 2011; Chatzinasiou et al,
tein to p16, its negative regulator; it is found in 2011; Macgregor et  al, 2011). PARP1 and ATM
360

360 Textbook of Cancer Epidemiology

suggest a role for DNA maintenance pathways, abundance of C-​ to-​

T nucleotide substitutions
but the functional mechanisms for the other loci (Brash, 2015). Mutations in the p53 gene can
are currently unknown (Law et  al, 2015). There lead to uncontrolled cell proliferation and loss
have been six population-​based GWAS of mel- of apoptosis, thus promoting cancerous growth.
anoma published, collectively identifying 13 Immunohistochemically detectable clusters of
risk loci (Brown et  al, 2008; Bishop et  al, 2009; epidermal cells with accumulated nuclear p53
Amos et al, 2011; Barrett et al, 2011; Macgregor protein (“p53 patches”) are found in normal skin
et  al, 2011; Iles et  al, 2013). A  meta-​analysis of before tumors arise. These patches are thought to
11 GWAS confirmed all 13 previously reported be an early step in the development of AKs and
genome-​ wide significant loci, confirmed two subsequent SCCs, though the progression rate is
previously reported loci with associations low (Rebel et al, 2005). RAS mutations have been
below genome-​wide significance (CCND1 and reported in up to 50% of SCCs (Pierceall et  al,
OCA2) and identified a further five loci reach- 1991; Spencer et al, 1995) and are also associated
ing genome-​wide significance at 2p22.2, 6p22.3, with AKs (Spencer et al, 1995). Mutations of p16
7p21.1, 9q31.2, and 10q24.33 (Law et  al, 2015). have also been reported in up to 24% of SCCs
Law et al. estimated that the 20 loci collectively (Kubo et al, 1997; Soufir et al, 1999).
explained 19% of the familial relative risk for
cutaneous melanoma (Law et al, 2015). Melanoma
Melanomas have a higher mutation load than
Somatic Events other cancers (Hill et al, 2013), and this has been
Basal Cell Carcinoma attributed to the mutagenic effects of UV radi-
Aberrant upregulation of the sonic hedgehog sig- ation (Pleasance et  al, 2010). UVB signature
naling pathway is commonly found in sporadic mutations, characterized by C-​to-​T nucleotide
BCCs, and either inactivating mutations in the substitutions, account for most of the elevated
patched gene or activating mutations in smooth- mutational burden (i.e. 46% of melanoma driver
ened genes (or both) appear to be the cause. mutations), while UVA signature mutations,
Mutations and allelic loss in the patched gene are characterized by G-​to-​T nucleotide substitutions
found in up to 95% of sporadic BCCs (Teh et al, as a result of oxidative damage, account for 9%
2005)  and activating mutations in the smooth- (Hodis et al, 2012).
ened gene in around 10% of sporadic BCCs Melanomas arise from the activation, or inac-
(Reifenberger et  al, 2005). Increased expression tivation, of genes that regulate the important
levels of Gli1 and Gli2, members of the Gli fam- cellular functions of cell-​cycle regulation, prolif-
ily of zinc-​finger transcription factors and down- eration, apoptosis, angiogenesis, and cell migra-
stream targets of sonic hedgehog, are also known tion (Ho & Tsao, 2015). Somatic mutations in
to play a role in BCC carcinogenesis (Regl et al, melanoma identified to date are linked to senes-
2002; Kasper et al, 2006). cence and apoptosis resistance (tumor suppressor
Mutations in the p53 tumor suppressor gene genes) and also sustained proliferative signaling
are detected in around half of all BCC tumors. (oncogenes). The two most frequently mutated
However, people suffering from the Li-​Fraumeni genes in cutaneous melanoma are the oncogenes
syndrome, characterized by germline mutations BRAF and NRAS, first identified by candidate
in the p53 gene, do not show increased incidence screening of the RAS-​RAF-​MEK-​ERK-​MAPK
of BCC, and thus p53 mutations are presumed to pathway (Kolch, 2005). BRAF mutations occur
be secondary events in BCC pathogenesis (Tilli in approximately 50%–​ 60% of all cutaneous
et al, 2005). Some studies, but not all, suggest that melanomas (Davies et al, 2002; Maldonado et al,
up to 30% of BCCs harbor RAS mutations, which 2003; Thomas et al, 2004), and NRAS mutations
lead to an activated protein that promotes cell in 19%–​28% (Hill et al, 2013).
growth and survival (van der Schroeff et al, 1990; Prevalence of mutations in the BRAF proto-​
Pierceall et al, 1991; Campbell et al, 1993). oncogene in melanoma also varies according to
the histological subtype and site of melanoma,
Squamous Cell Carcinoma with a lower incidence reported in tumors on
Mutations of the p53 tumor-​ suppressor gene chronically sun-​exposed skin (Maldonado et al,
are found in the majority of SCCs (Ahmadian 2003). The frequency of BRAF mutations in
et  al, 1998). These mutations are often “UVB-​ benign melanocytic nevi appears even higher
signature mutations,” characterized by an than in melanomas, but like melanoma, varies
361

Skin Cancer 361

markedly by histological and anatomic region PPP6C, and ARID2, but also several previously
subtypes of tumor (de Snoo & Hayward, 2005; implicated melanoma genes (NF1, IDH1, and
Shain et al, 2015). In support of a precursor role RB1) (Draper et  al, 1986; Andersen et  al, 1993;
of some benign nevi, common benign nevi har- Lopez et al, 2010) and one novel candidate mel-
bor BRAF V600E mutations exclusively, while anoma gene, DDX3X (Cancer Genome Atlas
intermediate lesions carry NRAS and additional Network, 2015). The clinical importance of these
driver mutations (Shain et al, 2015). BRAF inhibi- somatic mutations may be that metastatic tumors
tors have been developed and are currently in use with higher mutational loads may be more
for treating patients with metastatic disease that responsive to immunotherapy than those with
carry the BRAF 600E mutation (Flaherty, 2011), lower mutation loads (Van Allen et al, 2015).
and combining these treatments with immuno- Telomere length measured in peripheral
therapies is showing even greater survival advan- blood lymphocytes has been proposed as a gen-
tages (Menzies & Long, 2014). eral marker for aging and cancer risk, as telomeres
Other genetic alterations include amplifica- shorten with age. In contrast to the hypothe-
tion of AKT3 and loss of PTEN (Stahl et al, 2004; sis that shorter telomeres would be associated
Stark & Hayward, 2007), which cause constitutive with cancer risk, several studies have found that
PI3K pathway signaling. In some melanomas, shorter telomeres are associated with decreased
MAPK and PI3K pathway dysregulation results risk of melanoma, perhaps because longer telo-
from activating mutations in KIT, MET, and meres lead to a survival advantage in precancer-
EGFR (Curtin et al, 2005), or inactivating muta- ous skin cells (Nan et al, 2011b; Rode et al, 2016).
tions in neurofibromin 1 (NF1), a negative regu-
lator of RAS (Maertens et al, 2013). Mutations in RISK FACTORS
NF1 have been reported to be the third most com- Available evidence regarding risk factors for skin
mon mutation in sun-​exposed melanomas after cancer is summarized in Table 15-​1.
BRAF and NRAS (Krauthammer et al, 2015).
Mutations and deletions affecting the Tobacco
CDKN2A/​ARF locus are found in melanoma cell Basal Cell Carcinoma
lines but rarely in primary sporadic melanomas Two recent systematic reviews and meta-​analyses
(Pollock et al, 2001). This is believed to occur due of the associations between smoking and BCC
to p16INK4A’s central role in senescence:  cells (Leonardi-​ Bee et  al, 2012; Song et  al, 2012a)
in vitro lacking p16 escape senescence and can reported different pooled estimates of risk of BCC
readily become immortalized (Bennett, 2003). in relation to ever smoking with the larger meta-​
The same appears to be true for mutations in the analysis (11 case-​control and 3 cohort studies)
tumor suppressor gene PTEN in cell lines and reporting no increased risk (Leonardi-​Bee et  al,
tumors (de Snoo & Hayward, 2005). In primary 2012) and the other (6 cohort studies) a slightly
melanoma, the CDKN2A/​ARF locus shows loss increased risk (Song et al, 2012a). The larger meta-​
of heterozygosity (LOH) in around 50% of mela- analysis (Leonardi-​Bee et al, 2012) also examined
nomas, but mutations in the CDKN2A or ARF the association for current and former smoking
genes are found only in a small proportion of and found a nonsignificant inverse association for
tumors (Pollock et al, 2001). When deletions are current smoking and no association for former
accounted for however, the overall proportion of smoking. There is no consistent evidence from
tumors showing inactivation of either CDKN2A cohort studies of a dose-​response relationship.
or ARF increases (de Snoo & Hayward, 2005).
Next-​generation sequencing technologies Squamous Cell Carcinoma
have identified other genes involved in cellular The two recent systematic reviews and meta-​
pathways such as RB, TP53, and AURKA, includ- analyses both reported an increased risk of SCC
ing ARID2, PPP6C, RAC1, SNX31, STK19, and associated with ever smoking (Leonardi-​Bee et al,
TACC1 (Hodis et  al, 2012; Krauthammer et  al, 2012; Song et al, 2012a) although the magnitude
2012). The largest integrative analysis of cuta- of the reported pooled estimates differed (50%
neous melanoma conducted to date (n  =  331 increase in risk [Leonardi-​Bee et  al,  2012] and
patients) identified significantly mutated genes 10% increased risk [Song et al, 2012a]). The larger
including previously identified oncogenes and meta-​analysis (Leonardi-​Bee et al, 2012) reported
tumor suppressor genes (BRAF, NRAS, CDKN2A, a stronger association for current smoking (50%
TP53, and PTEN) as well as RAC1, MAP2K1, increased risk) than former smoking (20%
362

362 Textbook of Cancer Epidemiology

TABLE 15-​1   EPIDEMIOLOGY OF SKIN CANCER: RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Ultraviolet radiation exposure ↑↑
Numbers of nevi on skin (melanoma) ↑↑
Sun-​sensitive skin (including inherited syndromes) ↑↑
Family history in first-​degree relative ↑
Immunosuppression ↑

Probable Relationship Exists, Based on Substantial Data

Ionizing radiation exposure (BCC) ↑
Height ↑
High body mass index (melanoma in men) ↑
Cigarette smoking (SCC) ↑
Cigarette smoking (BCC, melanoma) ↓
Caffeine (BCC) ↓
NSAID use (BCC, SCC) ↓

Weak, if Any, Relationship Exists, Based on Substantial Data

Current oral contraceptive use —​
Postmenopausal hormone use —​
Reproductive factors (melanoma) —​
High body mass index (BCC, SCC, melanoma in women) —​
Physical activity —​
Total dietary fat intake —​
Folate —​
Carotenoids —​
Infectious agents (apart from Merkel cell carcinoma) —​

Inconsistent Findings or Limited Study to Date

Burn scars (SCC) ↑
Alcohol use (>1 drink/​day) ↑
Parkinson’s Disease ↑
Type I and Type II diabetes ↑

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate
to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association

increased risk). There is no consistent evidence Melanoma

from cohort studies of a dose-​ response rela- There is evidence that melanoma is inversely asso-
tionship. Of note, studies included in these ciated with tobacco smoking, with a recent meta-​
meta-​analyses may have included SCC of the lip analysis of 23 studies reporting a significant 30%
together with cutaneous SCC, and since lip SCCs reduced risk among current smokers and a 10%
are strongly associated with smoking (Perea-​Milla reduced risk among former smokers (Li et  al,
Lopez et al, 2003), their results are difficult to eval- 2015); the pooled findings from cohort and case-​
uate. On balance however, it appears that cigarette control studies were similar. Of the three large pro-
smoking increases risk of SCC. spective studies that have examined the association
36

Skin Cancer 363

(Freedman et al, 2003a; DeLancey et al, 2011; Song to incident BCC, only baseline serum selenium
et  al, 2012a), two reported a significant inverse concentration showed any association, namely a
dose-​response relationship: one overall across men 60% decrease in BCC tumor incidence (van der
and women (Freedman et al, 2003a) and the other Pols et al, 2009). Some prospective studies have
for men only (Song et al, 2012a). also found a positive association between dietary
folate intake, erythrocyte folate concentration,
Diet and BCC (Donnenfeld et al, 2015).
Although BCC, SCC, and melanoma of the skin With respect to dietary fats, no associations
are distinct cancers, dietary factors have been with BCC and total dietary fat; saturated, poly-
hypothesized to modify their risks through a unsaturated, and monounsaturated fatty acids; or
number of pathways in common. These include percent energy from fat were seen in a prospec-
protective roles for antioxidant nutrients by direct tive study (Ibiebele et al, 2009), and a systematic
neutralization of UV-​ induced reactive oxygen review and meta-​analysis of all randomized con-
species that cause damage to target-​cell DNA, or trolled trials (RCTs) and observational studies
for omega-​3 polyunsaturated fatty acids (PUFAs) of PUFAs and skin cancer found no association
through anti-​ inflammatory effects modifying with BCC (Noel et al, 2014).
UV-​induced prostaglandin production. However,
the influence of nutrition on UV carcinogenesis Squamous Cell Carcinoma
is more complex in reality, since antioxidant sup- Again, animal studies of SCC have suggested a
plementation can also promote carcinogenesis protective effect of antioxidant intake, but human
(Black, 2015)  and indeed experimental research studies have been few and have shown inconsist-
suggests that melanoma metastases may be pro- ent results (McNaughton et  al, 2005). Several
moted by antioxidants (Le Gal et  al, 2015)  and studies (Asgari et  al, 2011; Miura et  al, 2015a)
inhibited by oxidative stress (Piskounova et  al, suggest that consumption of tea as a source of
2015). Similarly omega-​6 PUFAs may promote antioxidants does not modify SCC development,
tumors via lipoxygenase activation (Black, 2015). while associations with caffeine intake have been
Overall, little of the evidence for the effects of die- null in three cohort studies (Song et  al, 2012b;
tary antioxidants, PUFAs, and of other nutrients Miura et  al, 2014). A  prospective study of fat
and foods on individual skin cancer types as sum- intake and SCC with adjustment for sun expo-
marized in what follows is conclusive. sure showed no significant linear trends except
in people with a history of skin cancer (Ibiebele
Basal Cell Carcinoma et  al, 2009). Regarding omega-​3 PUFAs specifi-
Among the few antioxidant-​rich foods studied cally, a systematic review suggested an inverse
in relation to BCC is tea consumption, but often but nonsignificant association with SCC (Noel
without distinction between black tea (one pro- et  al, 2014). There has been little evaluation of
spective study found no association [Miura et al, the role of diet in the development of AKs, but
2015a]) and the even more antioxidant-​rich green one prospective community-​based study showed
tea in epidemiologic studies, though experimen- that AK acquisition was significantly decreased
tal studies support promising protective effects of among those with the highest consumption of
green tea (Afaq & Katiyar, 2011). Tea also contains wine (average of half a glass a day in this study
caffeine, and when total caffeine intake from tea, population) and in the highest consumers of oily
coffee, and other sources has been studied in rela- fish (average of one serving every 5 days) com-
tion to BCC adjusting for sun exposure, inverse pared with those with minimal intakes (Hughes
associations have been observed in various pop- et al, 2009), the latter consistent with a possible
ulation subgroups (Song et  al, 2012b; Ferrucci protective effect of dietary omega-​3 PUFAs on
et al, 2014; Miura et al, 2014). Studies of intakes AK and SCC.
of specific antioxidant nutrients, namely retinol,
beta-​carotene, vitamin E, vitamin C, and sele- Melanoma
nium, have shown weak and inconsistent results A recent review of prospective studies and RCTs
regarding BCC risk, while evidence regarding of antioxidant vitamin and/​or mineral intakes
other carotenoids is lacking (McNaughton et al, in relation to cutaneous melanoma found no
2005). Further, when baseline serum concentra- strong and significant associations of antioxidant
tions of carotenoids, alpha-​tocopherol, and sele- nutrients, individually or combined, nor of fruit
nium were investigated prospectively in relation and vegetable intake with melanoma (Miura &
364

364 Textbook of Cancer Epidemiology

Green, 2015). However, current available evi- alcohol had a small increase in risk of melanoma
dence is inadequate to assess possible beneficial compared with those never consuming alcohol
effects of antioxidant intake on melanoma risk, (Rota et  al, 2014), but the association was lost
given the small number of relevant studies and when restricted to studies that adjusted for sun
since melanoma was not a primary disease out- exposure. A  subsequent pooled analysis of data
come in any of the six salient RCTs reviewed from eight case-​control studies of melanoma in
whose results were potentially confounded by women, with detailed information about sun
sun exposure (Miura & Green, 2015). Although exposure and alcohol consumption, found a pos-
experimental studies suggest antimelanoma itive association with ever consuming alcohol but
effects of other nutrients like grape-​seed pro- no evidence of a trend with increasing amount of
anthocyanidins, epigallocatechin-​3-​gallate and total or beverage-​specific alcohol intake (Miura
resveratrol, again human evidence is limited et al, 2015b).
(Tong & Young, 2014). This dearth of evidence In summary, the effect of alcohol on skin can-
applied also in regard to n-​3 PUFAs and mela- cer, if any, remains uncertain.
noma, where high consumption was inversely
associated based on one meta-​ analysis (Noel Reproductive Factors
et  al, 2014). Interpretation of inverse associa- Basal Cell and Squamous Cell Carcinoma
tions with long-​ term coffee consumption and No studies have examined reproductive factors
melanoma seen in some large prospective stud- in relation to the development of SCC, and only
ies has been difficult due to inconsistency by one has considered BCC and reported an inverse
dose (Wu et al, 2015a) and lack of adjustment for association with very high parity (10 or more
sun exposure (Loftfield et al, 2015), though two births) and BCC incidence (Hognas et al, 2014).
meta-​analyses of all available studies suggest that
consumption of coffee reduces the risk of mela- Melanoma
noma (Wang et al, 2016; Liu et al, 2016). Finally a A meta-​analysis reported that higher parity was
small increase in melanoma risk with high citrus associated inversely with melanoma while late
consumption seen after pooling two large cohort age at first pregnancy significantly increased risk
studies of health professionals (Wu et al, 2015b) (Gandini et al, 2011), although the authors con-
had weaknesses in study design that limit causal cluded that confounding by socioeconomic fac-
interpretation (Berwick, 2015). tors may explain the observed associations. They
found no association between age at menarche,
Alcohol age at menopause, or menopausal status and mel-
Basal Cell Carcinoma anoma (Gandini et al, 2011).
The results of cohort studies of BCC have been It appears unlikely that reproductive factors
very mixed, with some studies finding asso- substantially affect risk of skin cancer.
ciations with total alcohol intake, others find-
ing various beverage-​specific associations (e.g., Exogenous Hormones
Fung et  al, 2002; Soleas et  al, 2002; Freedman Basal Cell and Squamous Cell Carcinoma
et  al, 2003b; Jensen et  al, 2012), and yet others Few studies have examined exogenous hormone
finding no associations (e.g., Ansems et al, 2008). use in relation to the development of KCs (i.e.,
The degree of adjustment for BCC risk factors, BCC and SCC), and the findings are mixed. An
especially sun exposure, has varied widely across RCT (the Women’s Health Initiative) reported
studies. no association between menopausal hormone
therapy (MHT) and risk of any KC (Tang et  al,
Squamous Cell Carcinoma 2011). A  prospective study found no increased
Only two prospective studies have examined the risk of either BCC or SCC with ever use of oral
relationship between alcohol and SCC, and they contraceptives (OCs) or MHT (Birch-​Johansen
reported no association with first primary SCC et al, 2012), although a longer duration of MHT
(Ansems et al, 2008; Jensen et al, 2012). use (>5 years) was associated with a significantly
increased risk of SCC. An increased risk of BCC
Melanoma was associated with MHT use in the US Radiologic
A systematic review and meta-​analysis of stud- Technologists cohort, but no association with OC
ies of alcohol consumption and melanoma risk use or reproductive factors (Cahoon et al, 2015).
showed that people who had consumed any Two additional case-​control studies (one nested)
365

Skin Cancer 365

reported different associations between OC use Wiren et al, 2014; Lahmann et al, 2016) and with
and SCC, namely, an increased risk (Applebaum higher nevus counts (Ribero et al, 2015).
et al, 2009) and, in the nested study, no associa- Overall, among the various anthropometric
tion (Asgari et al, 2010). measures, only increasing height appears to be
associated consistently with increasing risk of
Melanoma skin cancer.
The meta-​analysis by Gandini et  al. (2011) did
not report an increased risk of melanoma with Infections
use of OCs or MHT. Based on available evidence, Basal Cell Carcinoma
exogenous hormones appear to have little effect To date, epidemiologic evidence shows no associ-
on skin cancer risk. ations between infectious agents and BCC apart
from potential associations with some types of
Anthropometric Measures human papilloma viruses (Iannacone et al, 2013).
Anthropometric indicators such as height,
weight, and body surface are thought to affect Squamous Cell Carcinoma
skin cancer risk through modification of metab- Originally identified in patients with epidermo-
olism and/​ or immune function, physiological dysplasia verruciformis, specific cutaneous HPV
programming in early life, or simply though types classified in the beta genus of papilloma
the association of body size with number of viruses (de Villiers, 2013)  have been postulated
target cells. to be linked with SCC. Infection is common not
only among immunosuppressed organ trans-
Basal Cell Carcinoma plant recipients but also among the general pop-
Several studies have examined overweight/​obe- ulation. A number of studies that have measured
sity in relation to BCC, with two cohort studies beta HPV antibodies, as well as those that have
reporting an inverse association (Gerstenblith assessed presence of beta HPV DNA in tumors,
et al, 2012; Pothiawala et al, 2012) and two oth- have shown inconsistent results regarding associ-
ers no association (Olsen et  al, 2006; Lahmann ations with SCC and AK (Hall et al, 2006; Karagas
et al, 2016). Greater attained adult height has also et  al, 2006; Andersson et  al, 2008; Plasmeijer
been inconsistently associated with BCC (Olsen et al, 2011). It is likely that beta HPV, if etiolog-
et al, 2006; Pelucchi et al, 2008; Gerstenblith et al, ically involved, acts to potentiate the effect of
2012; Lahmann et al, 2016). UV radiation via viral inhibition of DNA repair
and apoptosis following UV radiation (Hall et al,
Squamous Cell Carcinoma 2006). Elsewhere HPV-​38-​mediated downregu-
Overweight and obesity have been inversely lation of TLR9 expression has been linked to
associated with SCC in a pooled analysis of two cell-​cycle deregulation as well as its impact on
large US cohort studies (Pothiawala et al, 2012), immune function (McLaughlin-​ Drubin, 2015;
but no association was found in a third cohort Pacini et al, 2015).
(Lahmann et  al, 2016). The latter study is the
only study that has examined the relationship Melanoma
with height, reporting a significant association There is no evidence available that supports a
for men only. role for infectious agents in the causation of
melanoma.
Melanoma
Two meta-​ analyses have reported that over- Merkel Cell Carcinoma
weight and obesity are associated with a 30% A relatively rare type of skin cancer, Merkel cell
increased risk of melanoma among men but not carcinoma is additionally notable here because
women (Renehan et  al, 2008; Sergentanis et  al, 80% appear to be caused by Merkel cell polyoma
2013). High body mass index (BMI) has also virus (MCV) (Feng et al, 2008). Like cutaneous
been associated with greater Breslow thickness at HPV, MCV has been found to be shed from the
diagnosis (de Giorgi et al, 2013). Greater attained surface of the skin of the majority of healthy
adult height has been consistently shown to be adults, and thus evidence for Merkel cell carci-
positively associated with melanoma (Shors et al, noma causation by MCV is based on nucleic acid
2001; Gallus et al, 2006; Dennis et al, 2008; Olsen identification methods to detect viral integration
et al, 2008; Kabat et al, 2014; Kvaskoff et al, 2014; in tumor DNA rather than classical infectious
36

366 Textbook of Cancer Epidemiology

disease epidemiological models comparing cases vacations spent in sunny countries, and sun-
and controls (Moore & Chang, 2014). burns) seen among indoor workers compared
with outdoor workers (Nelemans et  al, 1993).
Physical Activity Overall, specific occupational associations have
Few studies have examined the possible rela- not been consistent and, in many studies, host
tionship between skin cancer and physical activ- characteristics and sun exposure have not been
ity. One cohort study reported that recreational adequately taken into account. An increased
activity in men and occupational activity in risk of melanoma reported for airline pilots and
women were related to the occurrence of SCC cabin crew (Sanlorenzo et  al, 2015)  most likely
(Lahmann et al, 2011), but physical activity was reflects sun-​related behavior rather than cosmic
not a predictor of skin cancer risk among organ radiation exposure (dos Santos Silva et al, 2013).
transplant recipients (Whalen et  al, 2010); both The famous historical linking of SCC of the scro-
studies had accounted for sun exposure in their tum to soot exposure among chimney sweeps by
analyses. Percival Pott in the late 18th century was one of
the first examples of the environmental causation
Ionizing Radiation of a cancer (Gawkrodger, 2004).
Risk of BCC has been shown consistently to be
increased after exposure to ionizing radiation Medical Conditions
from a wide variety of sources, including the and Treatment
atomic bomb after long latent periods (Sugiyama Immunosuppression is a risk factor for all three
et al, 2014), occupationally (Yoshinaga et al, 2005; common types of skin cancer and certain pop-
Lee et  al, 2015), and therapeutically (Levi et  al, ulations with compromised immunity are at
2006; Schwartz et al, 2009; Watt et al, 2012). On increased risk, including organ transplant recipi-
the other hand, cutaneous SCC and melanoma ents (Grulich et al, 2007), those diagnosed with
have rarely been associated with ionizing radia- HIV/​AIDS (Silverberg et  al, 2013; Olsen et  al,
tion exposure (Sugiyama et  al, 2014). In regard 2014b; Zhao et  al, 2016), and those treated for
to increased melanoma rates in certain occupa- rheumatoid arthritis (Wolfe & Michaud, 2007;
tionally exposed groups, causal attribution to sun Amari et  al, 2011), inflammatory bowel disease
exposure cannot be ruled out. (Long et  al, 2012; McKenna et  al, 2014; Singh
et  al, 2014), and some lymphoproliferative dis-
Occupation orders including non-​Hodgkin lymphoma and
Many outdoor occupations are linked to skin chronic lymphocytic leukemia (Morton et  al,
cancer but are also associated with high levels 2010; Pirani et  al, 2011; Brewer et  al, 2015;
of UV exposure that at least partially explain the Famenini et al, 2015; Olsen et al, 2016).
occupational associations. In some patient populations the magnitude
A recent systematic review and a European of the increased risk differs between skin can-
multicenter case-​ control study noted that cer types. Notably in transplant recipients, the
increased risks of melanoma have been reported increased risk is 2-​to 3-​fold for melanoma (Green
in employees in the agricultural, airline, elec- & Olsen, 2015), 10-​fold for BCC (Hartevelt et al,
trical, nuclear, oil, pulp, tannery, and teaching 1990; Tessari & Girolomoni, 2012), and as high as
industries (Suarez et al, 2007; Pukkala et al, 2014; 250-​fold for SCC (Hartevelt et al, 1990; Euvrard
Jiang et al, 2015), although these increased risks et  al, 2003; Moloney et  al, 2006; Jensen et  al,
have not been clearly related to specific expo- 2010). Higher risks of melanoma are observed
sures with the possible exception of arsenic, pol- in the immediate post-​transplant period, when
ychlorinated biphenyls, and some other organic the state of immunosuppression is greatest
chemicals. The systematic review also collated (Vajdic et al, 2009; Bilmon et al, 2014). For SCC,
data on occupational UV exposure and mela- time of onset after transplantation is inversely
noma incidence, finding a significant associa- related to intensity of ambient sun exposure. In
tion with outdoor occupational UVR exposure Queensland, Australia, cumulative incidence
in five studies, and with indoor occupation in of SCCs is approximately 30% at 10  years and
two studies (Jiang et  al, 2015), with the latter 60% at 20  years, whereas comparable rates in
association most likely explained by confound- European transplant recipients are later in onset
ing by socioeconomic status and/​ or a higher by some 10  years (London et  al, 1995; Bouwes
level of recreational sun exposure (sunbathing, Bavinck et  al, 1996). For HIV/​ AIDS patients
367

Skin Cancer 367

there is evidence that the degree of immunosup- damage to skin cell DNA (Shain et al, 2015), there
pression is important, with a statistically signifi- is a diversity of epidemiologic evidence for this
cant trend of higher rate ratios of SCC with lower causal link, ranging from ecological to migrant
recent CD4 counts (Silverberg et al, 2013). to clinical studies. The lowest skin cancer rates
Parkinson’s disease has been associated with occur in nonwhite populations whose epidermal
an increased risk of melanoma (Bajaj et al, 2010; melanin pigmentation shields target cells in the
Liu et  al, 2011; Huang et  al, 2015)  and of KCs deeper dermo-​epidermis from solar UV damage
(without differentiating between BCC and SCC) (IARC), 2012). Among white populations, per-
(Elbaz et  al, 2005; Rugbjerg et  al, 2012; Huang manent residents of high-​latitude regions with
et al, 2015). low sunlight exposure throughout their lives have
An increased risk of SCC (Shu et al, 2010) but low risk of skin cancer, but those who migrate
not BCC has been reported in patients with type from high to low latitudes early in life or regularly
I diabetes and of melanoma in patients with type holiday in sunny countries materially increase
II diabetes (Qi et  al, 2014). Two studies have their lifetime sun exposure and their risk of skin
found an increased risk of BCC in patients with cancer (Whiteman & Green, 1999; Chang et  al,
diabetes, although neither type nor severity of 2009). All skin cancer types are most common
the associated disease was reported (Pelucchi on the most highly sun-​exposed body site, the
et al, 2008; Hemminki et al, 2010). face (Franceschi et al, 1996). Analytic epidemio-
Autoimmune hepatitis has been positively logic studies show that skin cancer patients have
associated with KCs (without differentiating skin that tans poorly and sunburns easily because
between BCC and SCC) (Leung et  al, 2010). of lack of melanin protection (IARC, 2012).
Exposure to various medicines containing inor- Moreover, different skin tumor types occur in
ganic arsenic, including Chinese proprietary the same patients, confirming common etiologic
medicines (Wong et al, 1998) and drugs used in processes (Green & O’Rourke, 1985; Dika et  al,
the past to treat various dermatologic conditions 2016). Finally skin cancer rates are extremely
(Reymann et  al, 1978), as well as some asthma high in xeroderma pigmentosum patients, who
medications (Boonchai et  al, 2000), has been are unable to repair UV-​induced lesions in DNA
associated with the development of BCC after of skin cells (Kraemer et al, 1994).
long latent periods. Use of glucocorticoids—​a Despite their overall associations with high
group of drugs with immunosuppressive proper- sun exposure, BCC, SCC, and melanoma all show
ties and widely used to treat acute and chronic somewhat different patterns of dependence on
inflammatory disease—​has been associated with solar UV, perhaps reflecting their different cells
increased risk of both BCC and SCC (Karagas of origin in the dermo-​epidermis. Heterogeneity
et  al, 2001; Sorensen et  al, 2004; Jensen et  al, among histological subtypes of BCC and mela-
2009). There is also an increase in the incidence noma adds to the complexity. Since accurate ret-
of cutaneous SCC in patients with skin disorders rospective measurement of a person’s lifetime sun
such as psoriasis who are treated with psoralen exposure is not possible, subjective recall of time
and ultraviolet A (PUVA) (Lindelof et al, 1999). spent outdoors or lifetime sunburns have been
Recent meta-​analyses have shown that use of used as proxies for solar UV exposure; questions
nonsteroidal anti-​ inflammatory medications to elicit the necessary sun exposure information
(NSAIDs) is inversely associated with the devel- and definitions of exposure variables in analyses
opment of both BCC (Muranushi et al, 2016) and have varied widely, however. Further details such
SCC (Muranushi et al, 2015) but not melanoma as geographic location, season and time of day
(Hu et al, 2014). when sun exposure occurred, and anatomic site-​
specific information about protective clothing or
Ultraviolet Radiation sunscreen use all have a bearing on the level of
Solar Ultraviolet Radiation UV exposure received at the skin surface, while
UV radiation in sunlight is the principal envi- skin color influences the final UV dose at the tar-
ronmental cause of the three major types of skin get cell level beneath the surface epidermis. No
cancer: BCC, SCC, and melanoma (IARC, 2012). epidemiologic studies are able to gather details of
The UV radiation spectral regions reaching the this scope and complexity, and investigators have
earth—​wavelengths of 280–​320  nm (UVB) and differed in the degree to which they attempted
320–​400 nm (UVA)—​are specifically implicated. to take such details into account. Thus the pos-
Apart from evidence at the molecular level of UV sibility of misclassification of cumulated hours
368

368 Textbook of Cancer Epidemiology

of self-​
reported sun exposure or cumulated Europe, the United States, and New Zealand
sunburns is likely to be substantial. Also when (Whiteman & Green, 1999); it was estimated that
assessing observational results of sun exposure the incidence of melanoma increased at a rate of
and skin cancer among those with high occupa- about 5% per degree of latitude toward the equa-
tional sun exposure, self-​selection for innate low tor (Bulliard et al, 1994).
sun-​sensitivity (Green & Williams, 2007)  needs The importance of fair skin is illustrated in
to be considered. Despite these limitations, there Australia, where melanoma is exceedingly rare
are clear differences between relations between in Indigenous Australians, while Australians of
UV and BCC, SCC, and melanoma. predominantly European ancestry have among
the world’s highest incidence rates (Erdmann
Basal Cell Carcinoma et al, 2013). In migrants, the risk of melanoma is
Theories proposed to explain the overall lack higher with earlier age at arrival in or later depar-
of a monotonic association between BCC and ture from sunny countries. Thus it is thought
chronic sun exposure (Kricker et  al, 1995; that UV induction of melanoma probably often
Zanetti et al, 2006) cannot satisfactorily explain occurs early in life (Whiteman et al, 2001).
all the observed epidemiological features of Numerous case-​ control studies have con-
BCC. Strong positive dose-​ response relation- firmed that exposure to the UV component of
ships with childhood sun-​exposure suggest that sunlight is the major environmental determi-
the UV dose received early in life is an impor- nant of melanoma, but the results are compli-
tant predictor of risk. Because BCCs appear to be cated (Whiteman et  al, 2003). The intermittent
highly mitotic histopathologically, the threshold UV exposure theory—​whereby a short, intense
of total solar radiation for malignant transforma- UV exposure is more important than a chronic
tion may be low. Susceptibility to UV in sporadic one—​was first suggested to explain the lack of
BCC may be enhanced by genetic variations acti- association between melanoma and chronic sun
vating the hedgehog signaling pathway (Heitzer exposure when subjective assessment (recall) of
et al, 2007) and reprogramming of interfollicular outdoor exposure is used to assess the level of
epidermal stem cells in the basal layer of the skin sun exposure (Elwood et  al, 1985; Armstrong,
(Youssef et al, 2012) as target cells. Sporadic BCCs 1988). As for BCC, however, this suggestion
appear to cluster in a substantial proportion of that the pattern rather than the total amount
affected people (Keim et al, 2015), which would of UV exposure to the target cell causes mela-
be consistent with an innately lower threshold for noma appears somewhat inconsistent with other
transformation by UV in these people compared evidence.
with the high UV doses needed for SCC (Ramos A meta-​analysis of 57 studies reported signif-
et al, 2004) and with experimental evidence that icant associations between both intermittent sun
multiple genetic hits are required for invasive exposure and sunburn history and melanoma,
SCC to develop (Lapouge et al, 2011). but an inverse association with high occupational
sun exposure (Gandini et al, 2005). Multiple sun-
Squamous Cell Carcinoma burns, a key marker of intermittent sun exposure,
In contrast to BCC and melanoma, there is a mon- represent repeated episodes of acute UV-​induced
otonically increasing risk of SCC with increasing burn injuries to the dermo-​ epidermis. Their
sun exposure (Ramos et al, 2004; IARC, 2012). effects appear cumulative because a history of
multiple sunburns strongly predicts the risk of
Melanoma solar keratoses and SCCs (English et  al, 1998;
A causal link specifically between sun exposure Iannacone et  al, 2012), diseases that are UV
and melanoma was first suggested because the dose-​dependent. Melanoma’s predilection for the
disease was more common on sun-​exposed skin face and the male ear (Franceschi et  al, 1996),
(McGovern, 1952). The sunlight hypothesis was and its strong association with a history of other
supported by the first published evidence that as skin cancers or the presence of solar keratoses
latitude decreased (and by inference exposure to (Green & O’Rourke, 1985), also point to a role
ambient solar radiation rose), melanoma mortal- for chronic sun exposure.
ity rates rose among white-​skinned populations When cutaneous melanomas are analyzed
(Lancaster, 1956). Latitude gradients in both by histologic subtype, lentigo maligna mela-
incidence and mortality rates were described noma is found to have the strongest relationship
subsequently in white populations in northern with cumulative sun exposure, but nodular and
369

Skin Cancer 369

superficial spreading melanoma have also been nonexposed skin that are consistent with mela-
strongly associated with signs of chronic solar nomas arising through different causal pathways.
damage to the skin (Olsen et al, 2011). Moreover, BRAF mutations are more likely to occur on non-​
most case-​ control studies of cutaneous mela- sun-​exposed body sites (Maldonado et al, 2003;
noma have excluded lentigo maligna melanoma Curtin et  al, 2005; Liu et  al, 2007; Hacker et  al,
from analyses because of its distinctive UV dose 2013) in people with high nevus counts (Hacker
dependence. Thus, the apparently paradoxi- et al, 2010) and are less likely to be associated with
cal findings of case-​control studies are not fully solar keratoses (Lee et  al, 2011; Garcia-​Casado
explained by the histologic heterogeneity of et  al, 2015)  or high cumulative sun exposure
melanomas. (Hacker et al, 2010). The second most commonly
A “divergent pathway” model has been pro- mutated oncogene, the mutually exclusive NRAS
posed to explain the associations of melanoma mutation, occurs more frequently in melano-
with both chronic and intermittent patterns of mas arising in chronic sun-​damaged skin (Lee
sun exposure (Whiteman et al, 1998; Whiteman et al, 2011).
et  al, 2003; Whiteman et  al, 2006). This model There is also emerging evidence of hetero-
predicts that people with many nevi who have geneity in genotypes within BRAF-​mutant mel-
an inherently high propensity for nevus devel- anoma, with BRAF V600K mutant melanomas
opment require a lower threshold dose of solar more likely be associated with higher levels of
UV for malignant transformation to occur prob- sun exposure (Jewell et al, 2012), occuring more
ably early in life (Whiteman et al, 2001). This is commonly in older persons and on the head and
because their melanocytes are manifestly prone neck (Menzies et al, 2012). Conversely, BRAFV600E
to proliferation—​hence they develop melanoma mutant melanomas occur more commonly in
on relatively sun-​ protected anatomic sites at people under the age of 50  years with higher
young or middle ages. In contrast, melanomas in nevus counts and are more common in mela-
people with few nevi will tend to be associated nomas arising on intermittently sun-​ exposed
with cumulative sun exposure and will thus arise body sites (i.e., the trunk) (Liu et al, 2007; Viros
on the head and neck at older ages. et al, 2008).
In support of this divergent etiologic path- The role of UV in the development of acral
way, laboratory studies have reported consist- melanomas, which occur on the non-​ sun-​
ent differences in genetic or molecular profiles exposed skin of the soles, palms, and digits
of melanomas by anatomical site. Melanomas (including subungually) is less clear. Acral mel-
of the head and neck are more likely to over- anoma is the predominant subtype in non-​
express the p53 protein (Whiteman et al, 1998; Caucasians (Chang, 2013), however its incidence
Purdue et al, 2005), whereas melanomas on the is similar among races (Liu et  al, 2015). An
trunk are more likely to arise from a preexisting absence or low frequency of UV-​induced genetic
nevus (Skender-​Kalnenas et al, 1995; Carli et al, markers has been reported in these melanomas
1999; Hacker et  al, 2010). In accordance with (Beadling et al, 2008; Krauthammer et al, 2012),
this model, melanomas with and without neval and a higher incidence of KIT mutations (Curtin
remnants differ in their risk factor profiles (Lee et al, 2005; Handolias et al, 2010), analogous with
et al, 2006). People with head and neck melano- mucosal melanomas (Bello et al, 2013), suggests
mas are less likely to have many nevi than peo- that UV may not be involved in the etiology of
ple with trunk melanomas but are more likely these melanomas. Plantar, palmar, and digital nevi
to have many solar keratoses and a past history were a risk factor for melanomas on these sites in
of excised skin cancers (Whiteman et  al, 2003; Caucasian (Green et  al, 1999)  but not Japanese
Kvaskoff et al, 2013). Patients with head and neck (Rokuhara et al, 2004) study populations.
melanoma are more likely than those with trunk
melanoma to report high levels of sun exposure Artificial Ultraviolet Radiation
in adulthood and specifically, higher levels of Although a wide variety of sources of artificial
occupational (cumulative) exposure, but lower UV radiation exist in medicine and industry,
levels of recreational (intermittent) sun exposure use of sunbeds and sunlamps for cosmetic tan-
(Whiteman et al, 2006; Olsen et al, 2009; Olsen ning purposes has become a predominant source
et al, 2011). of exposure to artificial UV radiation, especially
There are also somatic mutational differ- among young women in temperate climates
ences in melanomas arising on sun-​exposed and (IARC, 2007). Recent meta-​ analyses confirm
370

370 Textbook of Cancer Epidemiology

that use of indoor tanning facilities is associated and Slop (on sunscreen)” campaign on Australia.
with the development of BCC, SCC (Wehner In the coming decades, the extent to which these
et al, 2012), and melanoma (Boniol et al, 2012). programs translate into a lower incidence of skin
Jurisdictions in Australia (Sinclair et al, 2014) and cancer will become apparent.
Brazil (Schalka et al, 2014) have banned sunbed
use for cosmetic tanning. In other countries like REFERENCES
the United Kingdom, legislation restricts sunbed Afaq F, Katiyar SK. Polyphenols: skin photoprotection
use by minors, while in others like the United and inhibition of photocarcinogenesis. Mini Rev
States, federal regulation is limited and regu- in Med Chem2011;11:1200–​15.
lation of indoor tanning in minors varies from Ahmadian A, Ren ZP, Williams C, Ponten F,
state to state with uneven enforcement (Pan & Odeberg J, Ponten J, et al. Genetic instability in
Geller, 2015). the 9q22.3 region is a late event in the develop-
ment of squamous cell carcinoma. Oncogene
Trauma 1998;17:1837–​43.
Amari W, Zeringue AL, McDonald JR, Caplan L,
A systematic review of epidemiologic studies on
Eisen SA, Ranganathan P. Risk of non-​melanoma
skin cancer in scar tissue from surgery, ulcers,
skin cancer in a national cohort of veterans with
or burns suggests that burn patients are not at
rheumatoid arthritis. Rheumatology (Oxford)
higher risk of skin cancers in general, although 2011;50:1431–​9.
a modest excess of SCC in burn scars could not Amos CI, Wang LE, Lee JE, Gershenwald JE, Chen
be excluded (Wallingford et al, 2011). Although WV, Fang S, et al. Genome-​wide association study
there is little evidence that acute trauma is a risk identifies novel loci predisposing to cutaneous
factor for BCC, SCC, or the majority of cutaneous melanoma. Hum Mol Genet 2011;20:5012–​23.
melanomas, two case-​control studies of melano- Andersen LB, Fountain JW, Gutmann DH, Tarle SA,
mas on the extremities have both implicated local Glover TW, Dracopoli NC et al. Mutations in the
trauma, specifically penetrating injuries, as a risk neurofibromatosis 1 gene in sporadic malignant
factor (Rolon et al, 1997; Green et al, 1999). In a melanoma cell lines. Nat Genet 1993;3:118–​21.
cross-​sectional survey, however, it was observed Andersson K, Waterboer T, Kirnbauer R, Slupetzky K,
that 9% of patients with melanoma of any histo- Iftner T, de Villiers EM et al. Seroreactivity to cuta-
logic type reported a prior history of trauma at neous human papillomaviruses among patients with
the site of their lesion (these were mostly acral nonmelanoma skin cancer or benign skin lesions.
melanomas, however, or melanoma on the Cancer Epidemiol Biomarkers Prev 2008;17:189–​95.
extremities). Many of the injuries in that series Ansems TM, van der Pols JC, Hughes MC, Ibiebele T,
(Kaskel et  al, 2000)  occurred in the 12  months Marks GC, Green AC. Alcohol intake and risk of
before diagnosis, constituting weak evidence of a skin cancer: a prospective study. Eur J Clin Nutr
causal association. 2008;62:162–​70.
Aoude LG, Wadt KA, Pritchard AL, Hayward NK.
Genetics of familial melanoma:  20  years after
CONCLUSION
CDKN2A. Pigment Cell and Melanoma Research
Although the predominant carcinogen for all
2015;28:148–​60.
three forms of skin cancer is known, substantial
Applebaum KM, Nelson HH, Zens MS, Stukel TA,
research questions remain that pose a challenge Spencer SK, Karagas MR. Oral contraceptives:  a
to epidemiologic methods. The relationship of risk factor for squamous cell carcinoma? J Invest
age at first high level sun exposure, pattern and Dermatol 2009;129:2760–​5.
total amount of sun exposure, and the ways these Armstrong BK. Epidemiology of malignant mela-
interact with the host phenotype still require elu- noma: intermittent or total accumulated exposure
cidation for the three major types. Molecular to the sun? J Dermatol Surg Oncol 1988;14:835–​49.
techniques of UV dosimetry may be useful in Asgari MM, Efird JT, Warton EM, Friedman GD.
helping to answer these questions, given the lim- Potential risk factors for cutaneous squamous cell
itations of conventional epidemiologic methods carcinoma include oral contraceptives: results of a
in measuring the key exposure. In theory, skin nested case-​control study. Int J Environ Res Public
cancers in fair-​skinned people are substantially Health 2010;7:427–​42.
preventable by avoiding UV radiation expo- Asgari MM, White E, Warton EM, Hararah MK,
sure. Behavior modification programs have been Friedman GD, Chren MM. Association of tea
remarkably successful in some countries, for consumption and cutaneous squamous cell carci-
instance the “Slip (on a shirt), Slap (on a hat), noma. Nutr Cancer 2011;63:314–​8.
371

Skin Cancer 371

Bajaj A, Driver JA, Schernhammer ES. Parkinson’s dis- self-​

healing squamous epithelioma (MSSE)
ease and cancer risk: a systematic review and meta-​ gene: further mapping, analysis of candidates, and
analysis. Cancer Causes Control 2010;21:697–​707. loss of heterozygosity. Oncogene 2006;25:806–​12.
Bale AE, Yu KP. The hedgehog pathway and basal cell Bouwes Bavinck JN, Hardie DR, Green A, Cutmore
carcinomas. Hum Mol Genet 2001;10:757–​62. S, MacNaught A, O’Sullivan B et  al. The risk
Barrett JH, Iles MM, Harland M, Taylor JC, Aitken of skin cancer in renal transplant recipients
JF, Andresen PA et  al. Genome-​wide association in Queensland, Australia:  a follow-​ up study.
study identifies three new melanoma susceptibil- Transplantation 1996;61:715–​21.
ity loci. Nat Genet 2011;43:1108–​13. Brash DE. UV signature mutations. Photochem
Beadling C, Jacobson-​ Dunlop E, Hodi FS, Le C, Photobiol 2015;91:15–​26.
Warrick A, Patterson J et al. KIT gene mutations Brewer JD, Shanafelt TD, Khezri F, Sosa Seda IM,
and copy number in melanoma subtypes. Clin Zubair AS, Baum CL et  al. Increased incidence
Cancer Res 2008;14:6821–​8. and recurrence rates of nonmelanoma skin can-
Bello DM, Ariyan CE, Carvajal RD. Melanoma cer in patients with non-​Hodgkin lymphoma:  a
mutagenesis and aberrant cell signaling. Cancer Rochester Epidemiology Project population-​
Control 2013;20:261–​81. based study in Minnesota. J Am Acad Dermatol
Bennett DC. Human melanocyte senescence and 2015;72:302–​9.
melanoma susceptibility genes. Oncogene Brown KM, Macgregor S, Montgomery GW, Craig
2003;22:3063–​9. DW, Zhao ZZ, Iyadurai K et al. Common sequence
Bertolotto C, Lesueur F, Bressac de Paillerets B. variants on 20q11.22 confer melanoma suscepti-
MITF:  a genetic key to melanoma and renal cell bility. Nat Genet 2008;40:838–​40.
carcinoma? Med Sci (Paris). 2012;28:258–​61. Bulliard JL, Cox B, Elwood JM. Latitude gradients in
Berwick M. Dietary Advice for melanoma: not ready melanoma incidence and mortality in the non-​
for prime time. J Clin Oncol 2015;33:2487–​8. Maori population of New Zealand. Cancer Causes
Bilmon IA, Ashton LJ, Le Marsney RE, Dodds AJ, Control 1994;5:234–​40.
O’Brien TA, Wilcox L et al. Second cancer risk in Cahoon EK, Kitahara CM, Ntowe E, Bowen EM,
adults receiving autologous haematopoietic SCT Doody MM, Alexander BH et al. Female estrogen-​
for cancer: a population-​based cohort study. Bone related factors and incidence of basal cell carci-
Marrow Transplant 2014;49:691–​8. noma in a nationwide US cohort. J Clin Oncol
Birch-​Johansen F, Jensen A, Mortensen L, Olesen AB, 2015;33:4058–​65.
Kjaer SK. Trends in the incidence of nonmela- Campbell C, Quinn AG, Rees JL. Codon 12
noma skin cancer in Denmark 1978–​2007: Rapid Harvey-​ras mutations are rare events in non-​
incidence increase among young Danish women. melanoma human skin cancer. Br J Dermatol
Int J Cancer 2010;127:2190–​8. 1993;128:111–​4.
Birch-​Johansen F, Jensen A, Olesen AB, Christensen J, Cancer Genome Atlas Network. Genomic classification
Tjonneland A, Kjaer SK. Does hormone replace- of cutaneous melanoma. Cell 2015;161:1681–​96.
ment therapy and use of oral contraceptives Carli P, Massi D, Santucci M, Biggeri A, Giannotti
increase the risk of non-​melanoma skin cancer? B. Cutaneous melanoma histologically asso-
Cancer Causes Control 2012;23:379–​88. ciated with a nevus and melanoma de novo
Bishop DT, Demenais F, Iles MM, Harland M, Taylor have a different profile of risk:  results from
JC, Corda E et al. Genome-​wide association study a case-​ control study. J Am Acad Dermatol
identifies three loci associated with melanoma 1999;40:549–​57.
risk. Nat Genet 2009;41:920–​5. Chahal HS, Lin Y, Ransohoff KJ, Hinds DA, Wu W, Dai
Black HS. The role of nutritional lipids and antioxi- HJ et al. Genome-​wide association study identifies
dants in UV-​induced skin cancer. Front Biosci novel susceptibility loci for cutaneous squamous
(Schol Ed) 2015;7:30–​9. cell carcinoma. Nat Commun. 2016;18:12048.
Boniol M, Autier P, Boyle P, Gandini S. Cutaneous Chang JW. Acral melanoma: a unique disease in Asia.
melanoma attributable to sunbed use: systematic JAMA Dermatol 2013;149:1272–​3.
review and meta-​analysis. BMJ 2012;345:e4757. Chang YM, Barrett JH, Bishop DT, Armstrong BK,
Boonchai W, Green A, Ng J, Dicker A, Chenevix-​ Bataille V, Bergman W et  al. Sun exposure and
Trench G. Basal cell carcinoma in chronic arseni- melanoma risk at different latitudes:  a pooled
cism occurring in Queensland, Australia, after analysis of 5700 cases and 7216 controls. Int J
ingestion of an asthma medication. J Am Acad Epidemiol 2009;38:814–​30.
Dermatol 2000;43:664–​9. Chatzinasiou F, Lill CM, Kypreou K, Stefanaki I,
Bose S, Morgan LJ, Booth DR, Goudie DR, Ferguson-​ Nicolaou V, Spyrou G et al. Comprehensive field
Smith MA, Richards FM. The elusive multiple synopsis and systematic meta-​analyses of genetic
372

372 Textbook of Cancer Epidemiology

association studies in cutaneous melanoma. J Natl crew and air traffic control officers: disentangling
Cancer Inst 2011;103:1227–​35. the effect of occupational versus lifestyle expo-
Cleaver JE. Cancer in xeroderma pigmentosum and sures. Int J Cancer 2013;132:374–​84.
related disorders of DNA repair. Nat Rev Cancer Draper GJ, Sanders BM, Kingston JE. Second primary
2005;5:564–​73. neoplasms in patients with retinoblastoma. Br J
Couve-​ Privat S, Bouadjar B, Avril MF, Sarasin A, Cancer 1986;53:661–​71.
Daya-​ Grosjean L. Significantly high levels of Elbaz A, Peterson BJ, Bower JH, Yang P, Maraganore
ultraviolet-​specific mutations in the smoothened DM, McDonnell SK et  al. Risk of cancer after
gene in basal cell carcinomas from DNA repair-​ the diagnosis of Parkinson’s disease:  a historical
deficient xeroderma pigmentosum patients. cohort study. Mov Disord 2005;20:719–​25.
Cancer Res 2002;62:7186–​9. Elwood JM, Gallagher RP, Hill GB, Pearson JC.
Couve-​Privat S, Le Bret M, Traiffort E, Queille S, Cutaneous melanoma in relation to intermittent
Coulombe J, Bouadjar B et al. Functional analysis and constant sun exposure—​the Western Canada
of novel sonic hedgehog gene mutations identified Melanoma Study. Int J Cancer 1985;35:427–​33.
in basal cell carcinomas from xeroderma pigmen- English DR, Armstrong BK, Kricker A, Winter MG,
tosum patients. Cancer Res 2004;64:3559–​65. Heenan PJ, Randell PL. Case-​control study of sun
Curtin JA, Fridlyand J, Kageshita T, Patel HN, exposure and squamous cell carcinoma of the
Busam KJ, Kutzner H et  al. Distinct sets of skin. Int J Cancer 1998;77:347–​53.
genetic alterations in melanoma. N Engl J Med Erdmann F, Lortet-​ Tieulent J, Schuz J, Zeeb H,
2005;353:2135–​47. Greinert R, Breitbart EW et  al. International
Davies H, Bignell GR, Cox C, Stephens P, Edkins trends in the incidence of malignant melanoma
S, Clegg S et  al. Mutations of the BRAF gene in 1953–​2008—​are recent generations at higher or
human cancer. Nature 2002;417:949–​54. lower risk? Int J Cancer 2013;132:385–​400.
Daya-​Grosjean L, Sarasin A. The role of UV induced Euvrard S, Kanitakis J, Claudy A. Skin cancers
lesions in skin carcinogenesis:  an overview of after organ transplantation. N Engl J Med
oncogene and tumor suppressor gene modifica- 2003;348:1681–​91.
tions in xeroderma pigmentosum skin tumors. Falchi M, Bataille V, Hayward NK, Duffy DL, Bishop
Mutat Res 2005;571:43–​56. JA, Pastinen T et  al. Genome-​wide association
de Giorgi V, Gori A, Papi F, Grazzini M, Rossari S, study identifies variants at 9p21 and 22q13 asso-
Verdelli A et al. Excess body weight and increased ciated with development of cutaneous nevi. Nat
Breslow thickness in melanoma patients: a retro- Genet 2009;41:915–​9.
spective study. Eur J Cancer Prev 2013;22:480–​5. Famenini S, Martires KJ, Zhou H, Xavier MF, Wu JJ.
de Snoo FA, Hayward NK. Cutaneous melanoma sus- Melanoma in patients with chronic lymphocytic
ceptibility and progression genes. Cancer Lett leukemia and non-​ Hodgkin lymphoma. J Am
2005;230:153–​86. Acad Dermatol 2015;72:78–​84.
DeLancey JO, Hannan LM, Gapstur SM, Thun MJ. Feng H, Shuda M, Chang Y, Moore PS. Clonal integra-
Cigarette smoking and the risk of incident and tion of a polyomavirus in human Merkel cell carci-
fatal melanoma in a large prospective cohort noma. Science (New York, NY) 2008;319:1096–​100.
study. Cancer Causes Control 2011;22:937–​42. Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S,
Dennis LK, Lowe JB, Lynch CF, Alavanja MC. Cutaneous Mathers C, Rebelo M, Parkin DM, Forman D, Bray,
melanoma and obesity in the Agricultural Health F. GLOBOCAN 2012 v1.0, Cancer Incidence and
Study. Ann Epidemiol 2008;18:214–​21. Mortality Worldwide: IARC CancerBase No. 11
de Villiers EM. Cross-​roads in the classification of [Internet]. Lyon, France: International Agency for
papillomaviruses. Virology 2013;445:2–​10. Research on Cancer; 2013. Available from: http://
Dika E, Fanti PA, Misciali C, Vaccari S, Crisman G, globocan.iarc.fr, accessed 2017.
Barisani A et al. Risk of skin cancer development Ferrucci LM, Cartmel B, Molinaro AM, Leffell DJ,
in 672 patients affected by Actinic Keratoses. G Bale AE, Mayne ST. Tea, coffee, and caffeine and
Ital Dermatol Venereol 2016;151:628-​633. early-​onset basal cell carcinoma in a case-​control
Donnenfeld M, Deschasaux M, Latino-​ Martel P, study. Eur J Cancer Prev 2014;23:296–​302.
Diallo A, Galan P, Hercberg S et  al. Prospective Flaherty KT. BRAF inhibitors and melanoma. Cancer
association between dietary folate intake and skin J 2011;17:505–​11.
cancer risk: results from the Supplementation en Franceschi S, Levi F, Randimbison L, La Vecchia C. Site
Vitamines et Mineraux Antioxydants cohort. Am distribution of different types of skin cancer: new
J Clin Nutr 2015;102:471–​8. aetiological clues. Int J Cancer 1996;67:24–​8.
dos Santos Silva I, De Stavola B, Pizzi C, Evans AD, Freedman DM, Sigurdson A, Doody MM, Mabuchi K,
Evans SA. Cancer incidence in professional flight Linet MS. Risk of basal cell carcinoma in relation
37

Skin Cancer 373

to alcohol intake and smoking. Cancer Epidemiol African-​American families with the nevoid basal
Biomarkers Prev 2003b;12:1540–​3. cell carcinoma syndrome (NBCC). Am J Med
Freedman DM, Sigurdson A, Doody MM, Rao RS, Genet 1994;50:272–​81.
Linet MS. Risk of melanoma in relation to smok- Gorlin RJ. Nevoid basal cell carcinoma (Gorlin) syn-
ing, alcohol intake, and other factors in a large drome. Genet Med 2004;6:530–​9.
occupational cohort. Cancer Causes Control Green A, McCredie M, MacKie R, Giles G, Young P,
2003a;14:847–​57. Morton C, et  al. A case-​control study of mela-
Frost C, Williams G, Green A. High incidence and nomas of the soles and palms (Australia and
regression rates of solar keratoses in a Queensland Scotland). Cancer Causes Control 1999;10:21–​5.
community. J Invest Dermatol 2000;115:273–​7. Green AC. Epidemiology of actinic keratoses. Curr
Fryer AA, Ramsay HM, Lovatt TJ, Jones PW, Hawley Probl Dermatol 2015;46:1–​7.
CM, Nicol DL et  al. Polymorphisms in glutathi- Green AC, O’Rourke MG. Cutaneous malignant mel-
one S-​transferases and non-​melanoma skin can- anoma in association with other skin cancers. J
cer risk in Australian renal transplant recipients. Natl Cancer Inst 1985;74:977–​80.
Carcinogenesis 2005;26:185–​91. Green AC, Olsen CM. Increased risk of melanoma
Fung TT, Hunter DJ, Spiegelman D, Colditz GA, in organ transplant recipients:  systematic review
Rimm EB, Willett WC. Intake of alcohol and alco- and meta-​analysis of cohort studies. Acta Derm
holic beverages and the risk of basal cell carci- Venereol 2015;95:923–​7.
noma of the skin. Cancer Epidemiol Biomarkers Green AC, Williams GM. Point:  sunscreen use is a
Prev 2002;11:1119–​22. safe and effective approach to skin cancer pre-
Gallus S, Naldi L, Martin L, Martinelli M, La Vecchia vention. Cancer Epidemiol Biomarkers Prev
C. Anthropometric measures and risk of cutane- 2007;16:1921–​2.
ous malignant melanoma:  a case-​control study Grulich AE, van Leeuwen MT, Falster MO, Vajdic
from Italy. Melanoma Res 2006;16:83–​7. CM. Incidence of cancers in people with HIV/​
Gandini S, Iodice S, Koomen E, Di Pietro A, Sera F, AIDS compared with immunosuppressed
Caini S. Hormonal and reproductive factors in transplant recipients:  a meta-​ analysis. Lancet
relation to melanoma in women:  current review 2007;370:59–​67.
and meta-​analysis. Eur J Cancer 2011;47:2607–​17. Gudbjartsson DF, Sulem P, Stacey SN, Goldstein AM,
Gandini S, Sera F, Cattaruzza MS, Pasquini P, Picconi Rafnar T, Sigurgeirsson B et  al. ASIP and TYR
O, Boyle P et al. Meta-​analysis of risk factors for pigmentation variants associate with cutaneous
cutaneous melanoma. II. Sun exposure. Eur J melanoma and basal cell carcinoma. Nat Genet
Cancer 2005;41:45–​60. 2008;40:886–​91.
Garcia-​Casado Z, Traves V, Banuls J, Niveiro M, Guy GP Jr, Thomas CC, Thompson T, Watson M,
Gimeno-​ Carpio E, Jimenez-​ Sanchez AI, et  al. Massetti GM, Richardson LC. Vital signs:  mel-
BRAF, NRAS and MC1R status in a prospec- anoma incidence and mortality trends and
tive series of primary cutaneous melanoma. Br J projections—​United States, 1982–​2030. MMWR
Dermatol 2015;172:1128–​31. 2015;64:591–​6.
Gawkrodger DJ. Occupational skin cancers. Occup Hacker E, Hayward NK, Dumenil T, James MR,
Med (Lond) 2004;54:458–​63. Whiteman DC. The association between MC1R
Gerstenblith MR, Rajaraman P, Khaykin E, Doody genotype and BRAF mutation status in cutaneous
MM, Alexander BH, Linet MS et  al. Basal cell melanoma:  findings from an Australian popula-
carcinoma and anthropometric factors in the tion. J Invest Dermatol 2010;130:241–​8.
U.S.  radiologic technologists cohort study. Int J Hacker E, Nagore E, Cerroni L, Woods SL, Hayward
Cancer 2012;131:E149–​55. NK, Chapman B et  al. NRAS and BRAF muta-
Goldstein AM, Chan M, Harland M, Gillanders EM, tions in cutaneous melanoma and the associa-
Hayward NK, Avril MF et al. High-​risk melanoma tion with MC1R genotype: findings from Spanish
susceptibility genes and pancreatic cancer, neu- and Austrian populations. J Invest Dermatol
ral system tumors, and uveal melanoma across 2013;133:1027–​33.
GenoMEL. Cancer Res 2006;66:9818–​28. Hall L, Struijk L, Neale RE, Feltkamp MC. Re: Human
Goldstein AM, Chan M, Harland M, Hayward NK, papillomavirus infection and incidence of squa-
Demenais F, Bishop DT et al. Features associated mous cell and basal cell carcinomas of the skin. J
with germline CDKN2A mutations: a GenoMEL Natl Cancer Inst 2006;98:1425–​6.
study of melanoma-​ prone families from three Handolias D, Salemi R, Murray W, Tan A, Liu W, Viros
continents. J Med Genet 2007;44:99–​106. A et al. Mutations in KIT occur at low frequency
Goldstein AM, Pastakia B, DiGiovanna JJ, Poliak S, in melanomas arising from anatomical sites
Santucci S, Kase R et al. Clinical findings in two
374

374 Textbook of Cancer Epidemiology

associated with chronic and intermittent sun expo- cancer:  a prospective study in Australian adults.
sure. Pigment Cell Melanoma Res 2010;23:210–​5. Int J Cancer 2009;125:1678–​84.
Harland M, Petljak M, Robles-​Espinoza CD, Ding Z, Iles MM, Law MH, Stacey SN, Han J, Fang S, Pfeiffer
Gruis NA, van Doorn R et  al. Germline TERT R et  al. A variant in FTO shows association
promoter mutations are rare in familial mela- with melanoma risk not due to BMI. Nat Genet
noma. Fam Cancer 2016;15:139–​44. 2013;45:428–​32, 32e1.
Hartevelt MM, Bavinck JN, Kootte AM, Vermeer International Agency for Research on Cancer
BJ, Vandenbroucke JP. Incidence of skin cancer (IARC). The association of use of sunbeds with
after renal transplantation in The Netherlands. cutaneous malignant melanoma and other
Transplantation 1990;49:506–​9. skin cancers:  a systematic review. Int J Cancer
Heitzer E, Lassacher A, Quehenberger F, Kerl H, Wolf 2007;120:1116–​22.
P. UV fingerprints predominate in the PTCH International Agency for Research on Cancer (IARC).
mutation spectra of basal cell carcinomas inde- Solar and UV radiation. In:  Radiation. IARC
pendent of clinical phenotype. J Invest Dermatol Monographs on the Evaluation of Carcinogenic
2007;127:2872–​81. Risks to Humans. Vol. 100D. Lyon:  IARC,
Hemminki K, Li X, Sundquist J, Sundquist K. Risk of 2012: 35–​101.
cancer following hospitalization for type 2 diabe- Jannot AS, Meziani R, Bertrand G, Gerard B,
tes. Oncologist 2010;15:548–​55. Descamps V, Archimbaud A et al. Allele variations
High A, Zedan W. Basal cell nevus syndrome. Curr in the OCA2 gene (pink-​eyed-​dilution locus) are
Opin Oncol 2005;17:160–​6. associated with genetic susceptibility to mela-
Hill VK, Gartner JJ, Samuels Y, Goldstein AM. The noma. Eur J Hum Genet 2005;13:913–​20.
genetics of melanoma: recent advances. Ann Rev Jensen A, Birch-​Johansen F, Olesen AB, Christensen
Genomics Hum Genet 2013;14:257–​79. J, Tjonneland A, Kjaer SK. Intake of alcohol may
Ho AW, Tsao H. Targeted therapies in mela- modify the risk for non-​ melanoma skin can-
noma: translational research at its finest. J Invest cer:  results of a large Danish prospective cohort
Dermatol 2015;135:1929–​33. study. J Invest Dermatol 2012;132:2718–​26.
Hodis E, Watson IR, Kryukov GV, Arold ST, Imielinski Jensen AO, Svaerke C, Farkas D, Pedersen L, Kragballe
M, Theurillat JP et al. A landscape of driver muta- K, Sorensen HT. Skin cancer risk among solid
tions in melanoma. Cell 2012;150:251–​63. organ recipients:  a nationwide cohort study in
Hognas E, Kauppila A, Pukkala E, Tapanainen JS. Denmark. Acta Derm Venereol 2010;90:474–​9.
Cancer risk in women with 10 or more deliveries. Jensen AO, Thomsen HF, Engebjerg MC, Olesen AB,
Obstet Gynecol 2014;123:811–​6. Friis S, Karagas MR et  al. Use of oral glucocorti-
Hu H, Xie Y, Yang G, Jian C, Deng Y. Nonsteroidal anti-​ coids and risk of skin cancer and non-​Hodgkin’s
inflammatory drug use and the risk of melanoma: a lymphoma: a population-​based case-​control study.
meta-​analysis. Eur J Cancer Prev 2014;23:62–​8. Br J Cancer 2009;100:200–​5.
Huang P, Yang XD, Chen SD, Xiao Q. The association Jewell R, Chambers P, Harland M, Laye J, Conway C,
between Parkinson’s disease and melanoma:  a Mitra A et al. Clinicopathologic features of V600E
systematic review and meta-​ analysis. Transl and V600K melanoma—​letter. Clin Cancer Res
Neurodegener 2015;4:21. 2012;18:6792; author’s reply p. 3.
Hughes MC, Williams GM, Fourtanier A, Green AC. Jiang AJ, Rambhatla PV, Eide MJ. Socioeconomic
Food intake, dietary patterns, and actinic kerato- and lifestyle factors and melanoma:  a systematic
ses of the skin:  a longitudinal study. American J review. Br J Dermatol 2015;172:885–​915.
Clin Nutr 2009;89:1246–​55. Kabat GC, Kim MY, Hollenbeck AR, Rohan TE.
Iannacone MR, Gheit T, Waterboer T, Giuliano AR, Attained height, sex, and risk of cancer at different
Messina JL, Fenske NA et  al. Case-​control study anatomic sites in the NIH-​AARP diet and health
of cutaneous human papillomavirus infection in study. Cancer Causes Control 2014;25:1697–​706.
basal cell carcinoma of the skin. J Invest Dermatol Karagas MR, Cushing GL Jr, Greenberg ER, Mott LA,
2013;133:1512–​20. Spencer SK, Nierenberg DW. Non-​melanoma skin
Iannacone MR, Wang W, Stockwell HG, O’Rourke K, cancers and glucocorticoid therapy. Br J Cancer
Giuliano AR, Sondak VK et al. Patterns and tim- 2001;85:683–​6.
ing of sunlight exposure and risk of basal cell and Karagas MR, Nelson HH, Sehr P, Waterboer T, Stukel
squamous cell carcinomas of the skin—​a case-​ TA, Andrew A et al. Human papillomavirus infec-
control study. BMC Cancer 2012;12:417. tion and incidence of squamous cell and basal
Ibiebele TI, van der Pols JC, Hughes MC, Marks GC, cell carcinomas of the skin. J Natl Cancer Inst
Green AC. Dietary fat intake and risk of skin 2006;98:389–​95.
375

Skin Cancer 375

Kaskel P, Kind P, Sander S, Peter RU, Krähn G. Trauma Lancaster HO. Some geographical aspects of the mor-
and melanoma formation: a true association? Br J tality from melanoma in Europeans. Med J Aust
Dermatol 2000;143:749–​53. 1956;43:1082–​7.
Kasper M, Regl G, Frischauf AM, Aberger F. GLI tran- Lapouge G, Youssef KK, Vokaer B, Achouri Y,
scription factors: mediators of oncogenic Hedgehog Michaux C, Sotiropoulou PA et al. Identifying the
signalling. Eur J Cancer 2006;42:437–​45. cellular origin of squamous skin tumors. Proc Natl
Keim U, van der Pols JC, Williams GM, Green AC. Acad Sci USA 2011;108:7431–​6.
Exclusive development of a single type of keratin- Law MH, Bishop DT, Lee JE, Brossard M, Martin
ocyte skin cancer:  evidence from an Australian NG, Moses EK et  al. Genome-​ wide meta-​
population-​based cohort study. J Invest Dermatol analysis identifies five new susceptibility loci
2015;135:728–​33. for cutaneous malignant melanoma. Nat Genet
Kolch W. Coordinating ERK/​ MAPK signalling 2015;47(9):987–​95.
through scaffolds and inhibitors. Nat Rev Mol Cell Law MH, Macgregor S, Hayward NK. Melanoma genet-
Biol 2005;6:827–​37. ics:  recent findings take us beyond well-​traveled
Kraemer KH, Lee MM, Andrews AD, Lambert WC. pathways. J Invest Dermatol 2012;132:1763–​74.
The role of sunlight and DNA repair in mela- Le Gal K, Ibrahim MX, Wiel C, Sayin VI, Akula MK,
noma and nonmelanoma skin cancer:  the xero- Karlsson C et al. Antioxidants can increase mela-
derma pigmentosum paradigm. Arch Dermatol noma metastasis in mice. Sci Transl Med 2015;7:8.
1994;130:1018–​21. Lear JT, Heagerty AH, Smith A, Bowers B, Payne
Krauthammer M, Kong Y, Bacchiocchi A, Evans P, CR, Smith CA et al. Multiple cutaneous basal cell
Pornputtapong N, Wu C et  al. Exome sequenc- carcinomas:  glutathione S-​ transferase (GSTM1,
ing identifies recurrent mutations in NF1 and GSTT1) and cytochrome P450 (CYP2D6,
RASopathy genes in sun-​exposed melanomas. Nat CYP1A1) polymorphisms influence tumour num-
Genet 2015;47:996–​1002. bers and accrual. Carcinogenesis 1996;17:1891–​6.
Krauthammer M, Kong Y, Ha BH, Evans P, Bacchiocchi Lee EY, Williamson R, Watt P, Hughes MC, Green AC,
A, McCusker JP et al. Exome sequencing identifies Whiteman DC. Sun exposure and host phenotype
recurrent somatic RAC1 mutations in melanoma. as predictors of cutaneous melanoma associated
Nat Genet 2012;44:1006–​14. with neval remnants or dermal elastosis. Int J
Kricker A, Armstrong BK, English DR, Heenan PJ. Cancer 2006;119:636–​42.
Does intermittent sun exposure cause basal cell Lee JH, Choi JW, Kim YS. Frequencies of BRAF and
carcinoma? A  case-​ control study in Western NRAS mutations are different in histological types
Australia. Int J Cancer 1995;60:489–​94. and sites of origin of cutaneous melanoma:  a
Kubo Y, Urano Y, Matsumoto K, Ahsan K, Arase S. meta-​analysis. Br J Dermatol 2011;164:776–​84.
Mutations of the INK4a locus in squamous cell Lee T, Sigurdson AJ, Preston DL, Cahoon EK,
carcinomas of human skin. Biochem Biophys Res Freedman DM, Simon SL et al. Occupational ion-
Commun 1997;232:38–​41. ising radiation and risk of basal cell carcinoma in
Kvaskoff M, Bijon A, Mesrine S, Vilier A, Clavel-​ US radiologic technologists (1983–​2005). Occup
Chapelon F, Boutron-​Ruault MC. Anthropometric Environ Med 2015;72:862–​9.
features and cutaneous melanoma risk:  a pro- Lehmann AR, McGibbon D, Stefanini M. Xeroderma
spective cohort study in French women. Cancer pigmentosum. Orphanet J Rare Dis 2011;6:70.
Epidemiol 2014;38:357–​63. Leonardi-​Bee J, Ellison T, Bath-​Hextall F. Smoking
Kvaskoff M, Pandeya N, Green AC, Perry S, Baxter and the risk of nonmelanoma skin cancer:  sys-
C, Davis MB et  al. Site-​ specific determinants tematic review and meta-​analysis. Arch Dermatol
of cutaneous melanoma:  a case-​ case compari- 2012;148:939–​46.
son of patients with tumors arising on the head Leung J, Dowling L, Obadan I, Davis J, Bonis PA,
or trunk. Cancer Epidemiol Biomarkers Prev Kaplan MM et  al. Risk of non-​melanoma skin
2013;22:2222–​31. cancer in autoimmune hepatitis. Dig Dis Sci
Lahmann PH, Hughes MCB, Williams GM, Green 2010;55:3218–​23.
AC. A prospective study of measured body size Levi F, Moeckli R, Randimbison L, Te VC, Maspoli M,
and height and risk of keratinocyte cancers and La Vecchia C. Skin cancer in survivors of childhood
melanoma. Cancer Epidemiol 2016;40:119–​25. and adolescent cancer. Eur J Cancer 2006;42:656–​9.
Lahmann PH, Russell A, Green AC. Prospective Li Z, Wang Z, Yu Y, Zhang H, Chen L. Smoking is
study of physical activity and risk of squa- inversely related to cutaneous malignant mela-
mous cell carcinoma of the skin. BMC Cancer noma:  results of a meta-​analysis. Br J Dermatol
2011;11:516. 2015;173:1540–​3.
376

376 Textbook of Cancer Epidemiology

Lindelof B, Sigurgeirsson B, Tegner E, Larko O, factor-​alpha inhibitor monotherapy and in combi-

Johannesson A, Berne B et al. PUVA and cancer nation with thiopurines: analysis of the Food and
risk: the Swedish follow-​up study. Br J Dermatol Drug Administration Adverse Event Reporting
1999;141:108–​12. System. J Gastrointestin Liver Dis 2014;23:267–​71.
Liu J, Shen B, Shi M, Cai J. Higher caffeinated cof- McLaughlin-​ Drubin ME. Human papillomaviruses
fee intake is associated with reduced malignant and non-​ melanoma skin cancer. Semin Oncol
melanoma risk: a meta-​analysis study. PLoS One 2015;42:284–​90.
2016;11:e0147056. McLean DI, Phillips N, Zhou Y, Gallagher R, Lee
Liu L, Zhang W, Gao T, Li C. Is UV an etiological TK. 40-​ year trends in skin cancer in British
factor of acral melanoma? J Expo Sci Environ Columbia, Canada, 1973 to 2003. J Cutan Med
Epidemiol 2015. Surg 2012;16:83–​91.
Liu R, Gao X, Lu Y, Chen H. Meta-​analysis of the rela- McNaughton SA, Marks GC, Green AC. Role of die-
tionship between Parkinson disease and mela- tary factors in the development of basal cell can-
noma. Neurology 2011;76:2002–​9. cer and squamous cell cancer of the skin. Cancer
Liu W, Kelly JW, Trivett M, Murray WK, Dowling Epidemiol Biomarkers Prev 2005;14:1596–​607.
JP, Wolfe R et  al. Distinct clinical and path- Menzies AM, Haydu LE, Visintin L, Carlino MS,
ological features are associated with the Howle JR, Thompson JF et al. Distinguishing clin-
BRAF(T1799A(V600E)) mutation in primary icopathologic features of patients with V600E and
melanoma. J Invest Dermatol 2007;127:900–​5. V600K BRAF-​mutant metastatic melanoma. Clin
Loftfield E, Freedman ND, Graubard BI, Hollenbeck Cancer Res 2012;18:3242–​9.
AR, Shebl FM, Mayne ST et  al. Coffee drinking Menzies AM, Long GV. Systemic treatment for BRAF-​
and cutaneous melanoma risk in the NIH-​AARP mutant melanoma: where do we go next? Lancet
diet and health study. J Natl Cancer Inst 2015;107. Oncol 2014;15:e371–​81.
Lomas A, Leonardi-​Bee J, Bath-​Hextall F. A systematic Miura K, Hughes MC, Arovah NI, van der Pols JC,
review of worldwide incidence of non-​melanoma Green AC. Black tea consumption and risk of skin
skin cancer. Br J Dermatol 2012:166:1069–​80. cancer: an 11-​year prospective study. Nutr Cancer
London NJ, Farmery SM, Will EJ, Davison AM, Lodge 2015a;67:1049–​55.
JP. Risk of neoplasia in renal transplant patients. Miura K, Hughes MC, Green AC, van der Pols JC.
Lancet 1995;346:403–​6. Caffeine intake and risk of basal cell and squa-
Long MD, Martin CF, Pipkin CA, Herfarth HH, mous cell carcinomas of the skin in an 11-​year
Sandler RS, Kappelman MD. Risk of mela- prospective study. Eur J Nutr 2014;53:511–​20.
noma and nonmelanoma skin cancer among Miura K, Green AC. Dietary antioxidants and mel-
patients with inflammatory bowel disease. anoma:  evidence from cohort and intervention
Gastroenterology 2012;143:390–​9.e1. studies. Nutr Cancer 2015:1–​10.
Lopez GY, Reitman ZJ, Solomon D, Waldman T, Bigner Miura K, Zens MS, Peart T, Holly EA, Berwick M, Gallagher
DD, McLendon RE et  al. IDH1(R132) mutation RP et al. Alcohol consumption and risk of melanoma
identified in one human melanoma metastasis, among women: pooled analysis of eight case-​control
but not correlated with metastases to the brain. studies. Arch Dermatol Res 2015b;307:819–​28.
Biochem Biophys Res Commun 2010;398:585–​7. Moloney FJ, Comber H, O’Lorcain P, O’Kelly P,
Macgregor S, Montgomery GW, Liu JZ, Zhao ZZ, Conlon PJ, Murphy GM. A population-​ based
Henders AK, Stark M et  al. Genome-​wide asso- study of skin cancer incidence and prevalence
ciation study identifies a new melanoma suscepti- in renal transplant recipients. Br J Dermatol
bility locus at 1q21.3. Nat Genet 2011;43:1114–​8. 2006;154:498–​504.
Maertens O, Johnson B, Hollstein P, Frederick DT, Moore PS, Chang Y. The conundrum of causality in
Cooper ZA, Messiaen L et al. Elucidating distinct tumor virology:  the cases of KSHV and MCV.
roles for NF1 in melanomagenesis. Cancer Discov Semin Cancer Biol 2014;26:4–​12.
2013;3:338–​49. Morton LM, Curtis RE, Linet MS, Bluhm EC, Tucker
Maldonado JL, Fridlyand J, Patel H, Jain AN, Busam MA, Caporaso N et  al. Second malignancy risks
K, Kageshita T et al. Determinants of BRAF muta- after non-​Hodgkin’s lymphoma and chronic lym-
tions in primary melanomas. J Natl Cancer Inst phocytic leukemia: differences by lymphoma sub-
2003;95:1878–​90. type. J Clin Oncol 2010;28:4935–​44.
McGovern VJ. Melanoblastoma. Med J Aust Muranushi C, Olsen CM, Green AC, Pandeya N. Can
1952;1:139–​42. oral nonsteroidal antiinflammatory drugs play
McKenna MR, Stobaugh DJ, Deepak P. Melanoma a role in the prevention of basal cell carcinoma?
and non-​melanoma skin cancer in inflammatory A systematic review and metaanalysis. J Am Acad
bowel disease patients following tumor necrosis Dermatol 2016;74:108–​119.
37

Skin Cancer 377

Muranushi C, Olsen CM, Pandeya N, Green AC. receptor 9 expression by beta human papilloma-
Aspirin and nonsteroidal anti-​ inflammatory virus 38 and implications for cell cycle control. J
drugs can prevent cutaneous squamous cell car- Virol 2015;89:11396–​405.
cinoma: a systematic review and meta-​analysis. J Palmer JS, Duffy DL, Box NF, Aitken JF, O’Gorman
Invest Dermatol 2015;135:975–​83. LE, Green AC et al. Melanocortin-​1 receptor poly-
Nan H, Xu M, Kraft P, Qureshi AA, Chen C, Guo Q morphisms and risk of melanoma: is the associa-
et  al. Genome-​wide association study identifies tion explained solely by pigmentation phenotype?
novel alleles associated with risk of cutaneous Am J Hum Genet 2000;66:176–​86.
basal cell carcinoma and squamous cell carci- Pan M, Geller L. Update on indoor tanning leg-
noma. Hum Mol Genet 2011a;20:3718–​24. islation in the United States. Clin Dermatol
Nan H, Du M, De Vivo I, Manson JE, Liu S, McTiernan 2015;33:387–​92.
A et al. Shorter telomeres associate with a reduced Pelucchi C, Naldi L, Di Landro A, La Vecchia C.
risk of melanoma development. Cancer Res Anthropometric measures, medical history and
2011b;71:6758–​63. risk of basal cell carcinoma in an Italian case-​con-
Nelemans PJ, Groenendal H, Kiemeney LA, Rampen trol study. Dermatology 2008;216:271–​6.
FH, Ruiter DJ, Verbeek AL. Effect of intermittent Perea-​Milla Lopez E, Minarro-​Del Moral RM, Martinez-​
exposure to sunlight on melanoma risk among Garcia C, Zanetti R, Rosso S, Serrano S et  al.
indoor workers and sun-​ sensitive individuals. Lifestyles, environmental and phenotypic factors
Environ Health Perspect 1993;101:252–​5. associated with lip cancer: a case-​control study in
Nikolaou V, Stratigos AJ, Tsao H. Hereditary non- southern Spain. Br J Cancer 2003;88:1702–​7.
melanoma skin cancer. Semin Cutan Med Surg Pierceall WE, Goldberg LH, Tainsky MA,
2012;31:204–​10. Mukhopadhyay T, Ananthaswamy HN. Ras gene
Noel SE, Stoneham AC, Olsen CM, Rhodes LE, Green mutation and amplification in human nonmela-
AC. Consumption of omega-​3 fatty acids and the noma skin cancers. Mol Carcinog 1991;4:196–​202.
risk of skin cancers: a systematic review and meta-​ Pirani M, Marcheselli R, Marcheselli L, Bari A,
analysis. Int J Cancer 2014;135:149–​56. Federico M, Sacchi S. Risk for second malignan-
Olsen CM, Green AC, Zens MS, Stukel TA, Bataille V, cies in non-​ Hodgkin’s lymphoma survivors:  a
Berwick M et al. Anthropometric factors and risk meta-​analysis. Ann Oncol 2011;22:1845–​58.
of melanoma in women:  a pooled analysis. Int J Piskounova E, Agathocleous M, Murphy MM, Hu
Cancer 2008;122:1100–​8. Z, Huddlestun SE, Zhao Z et al. Oxidative stress
Olsen CM, Hughes MC, Pandeya N, Green AC. inhibits distant metastasis by human melanoma
Anthropometric measures in relation to basal cells. Nature 2015;527:186–​91.
cell carcinoma: a longitudinal study. BMC Cancer Plasmeijer EI, Pandeya N, O’Rourke P, Pawlita M,
2006;6:82. Waterboer T, Feltkamp MC et  al. The association
Olsen CM, Knight LL, Green AC. Risk of melanoma between cutaneous squamous cell carcinoma and
in people with HIV/​AIDS in the pre-​and post-​ betapapillomavirus seropositivity:  a cohort study.
HAART eras: a systematic review and meta-​anal- Cancer Epidemiol Biomarkers Prev 2011;20:1171–​7.
ysis of cohort studies. PLoS One 2014b;9:e95096. Platz A, Hansson J, Ringborg U. Screening of germ-
Olsen CM, Lane SW, Green AC. Increased risk of line mutations in the CDK4, CDKN2C and TP53
melanoma in patients with chronic lymphocytic genes in familial melanoma: a clinic-​based popu-
leukaemia:  systematic review and meta-​analysis lation study. Int J Cancer 1998;78:13–​5.
of cohort studies. Melanoma Res 2016;26:188–​94. Pleasance ED, Cheetham RK, Stephens PJ, McBride
Olsen CM, Williams PF, Whiteman DC. Turning the DJ, Humphray SJ, Greenman CD et  al. A com-
tide? Changes in treatment rates for keratinocyte prehensive catalogue of somatic mutations from
cancers in Australia 2000 through 2011. J Am a human cancer genome. Nature 2010;463:191–​6.
Acad Dermatol 2014a;71:21–​6.e1. Pollock PM, Welch J, Hayward NK. Evidence for
Olsen CM, Zens MS, Green AC, Stukel TA, Holman three tumor suppressor loci on chromosome 9p
CD, Mack T et al. Biologic markers of sun expo- involved in melanoma development. Cancer Res
sure and melanoma risk in women: pooled case-​ 2001;61:1154–​61.
control analysis. Int J Cancer 2011;129:713–​23. Pothiawala S, Qureshi AA, Li Y, Han J. Obesity and the
Olsen CM, Zens MS, Stukel TA, Sacerdote C, Chang incidence of skin cancer in US Caucasians. Cancer
YM, Armstrong BK et al. Nevus density and mel- Causes Control 2012;23:717–​26.
anoma risk in women:  a pooled analysis to test Pukkala E, Martinsen JI, Weiderpass E, Kjaerheim
the divergent pathway hypothesis. Int J Cancer K, Lynge E, Tryggvadottir L et  al. Cancer inci-
2009;124:937–​44. dence among firefighters:  45  years of follow-​up
Pacini L, Savini C, Ghittoni R, Saidj D, Lamartine in five Nordic countries. Occup Environ Med
J, Hasan UA et  al. Downregulation of toll-​like 2014;71:398–​404.
378

378 Textbook of Cancer Epidemiology

Purdue MP, From L, Kahn HJ, Armstrong BK, Kricker Rogers HW, Weinstock MA, Feldman SR, Coldiron
A, Gallagher RP et al. Etiologic factors associated BM. Incidence estimate of nonmelanoma skin
with p53 immunostaining in cutaneous malignant cancer (keratinocyte carcinomas) in the US pop-
melanoma. Int J Cancer 2005;117:486–​93. ulation, 2012. JAMA Dermatol 2015;151:1081–​6.
Qi L, Qi X, Xiong H, Liu Q, Li J, Zhang Y, et  al. Rokuhara S, Saida T, Oguchi M, Matsumoto K, Murase
Type 2 diabetes mellitus and risk of malig- S, Oguchi S. Number of acquired melanocytic nevi
nant melanoma:  a systematic review and meta-​ in patients with melanoma and control subjects in
analysis of cohort studies. Iran J Public Health Japan: nevus count is a significant risk factor for
2014;43:857–​66. nonacral melanoma but not for acral melanoma. J
Rafnar T, Sulem P, Stacey SN, Geller F, Gudmundsson Am Acad Dermatol 2004;50:695–​700.
J, Sigurdsson A et  al. Sequence variants at the Rolón PA, Kramárová E, Rolón HI, Khlat M,
TERT-​CLPTM1L locus associate with many can- Parkin DM. Plantar melanoma:  a case-​ con-
cer types. Nat Genet 2009;41:221–​7. trol study in Paraguay. Cancer Causes Control
Ramachandran S, Hoban PR, Ichii-​Jones F, Pleasants 1997;8:850–​6.
L, Ali-​Osman F, Lear JT et al. Glutathione S-​trans- Rota M, Pasquali E, Bellocco R, Bagnardi V, Scotti
ferase GSTP1 and cyclin D1 genotypes:  associa- L, Islami F et  al. Alcohol drinking and cuta-
tion with numbers of basal cell carcinomas in a neous melanoma risk:  a systematic review
patient subgroup at high-​risk of multiple tumours. and dose-​ risk meta-​analysis. Br J Dermatol
Pharmacogenetics 2000;10:545–​56. 2014;170:1021–​8.
Ramos J, Villa J, Ruiz A, Armstrong R, Matta J. UV Rugbjerg K, Friis S, Lassen CF, Ritz B, Olsen JH.
dose determines key characteristics of nonmela- Malignant melanoma, breast cancer and other
noma skin cancer. Cancer Epidemiol Biomarkers cancers in patients with Parkinson’s disease. Int J
Prev 2004;13:2006–​11. Cancer 2012;131:1904–​11.
Read J, Wadt KA, Hayward NK. Melanoma genetics. J Sanlorenzo M, Wehner MR, Linos E, Kornak J, Kainz
Med Genet. 2016;53:1–​14. W, Posch C et al. The risk of melanoma in airline
Rebel H, Kram N, Westerman A, Banus S, van Kranen pilots and cabin crew:  a meta-​ analysis. JAMA
HJ, de Gruijl FR. Relationship between UV-​ Dermatology 2015;151:51–​8.
induced mutant p53 patches and skin tumours, Schalka S, Steiner D, Ravelli FN, Steiner T, Terena AC,
analysed by mutation spectra and by induction Marcon CR et  al. Brazilian consensus on photo-
kinetics in various DNA-​ repair-​deficient mice. protection. An Bras Dermatol 2014;89:1–​74.
Carcinogenesis 2005;26:2123–​30. Schwartz JL, Kopecky KJ, Mathes RW, Leisenring
Regl G, Neill GW, Eichberger T, Kasper M, Ikram MS, WM, Friedman DL, Deeg HJ. Basal cell skin
Koller J et al. Human GLI2 and GLI1 are part of a cancer after total-​body irradiation and hema-
positive feedback mechanism in basal cell carci- topoietic cell transplantation. Radiat Res
noma. Oncogene 2002;21:5529–​39. 2009;171:155–​63.
Reifenberger J, Wolter M, Knobbe CB, Kohler B, Sergentanis TN, Antoniadis AG, Gogas HJ,
Schonicke A, Scharwachter C et al. Somatic muta- Antonopoulos CN, Adami HO, Ekbom A et  al.
tions in the PTCH, SMOH, SUFUH and TP53 Obesity and risk of malignant melanoma: a meta-​
genes in sporadic basal cell carcinomas. Br J analysis of cohort and case-​control studies. Eur J
Dermatol 2005;152:43–​51. Cancer 2013;49:642–​57.
Renehan AG, Tyson M, Egger M, Heller RF, Shain AH, Yeh I, Kovalyshyn I, Sriharan A, Talevich
Zwahlen M. Body-​mass index and incidence E, Gagnon A et al. The genetic evolution of mel-
of cancer: a systematic review and meta-​analy- anoma from precursor lesions. N Engl J Med
sis of prospective observational studies. Lancet 2015;373:1926–​36.
2008;371:569–​78. Shors AR, Solomon C, McTiernan A, White E.
Reymann F, Moller R, Nielsen A. Relationship between Melanoma risk in relation to height, weight, and
arsenic intake and internal malignant neoplasms. exercise (United States). Cancer Causes Control
Arch Dermatol 1978;114:378–​81. 2001;12:599–​606.
Ribero S, Glass D, Aviv A, Spector TD, Bataille V. Height Shu X, Ji J, Li X, Sundquist J, Sundquist K, Hemminki
and bone mineral density are associated with naevus K. Cancer risk among patients hospitalized
count supporting the importance of growth in mel- for type 1 diabetes mellitus:  a population-​
anoma susceptibility. PLoS One 2015;10:e0116863. based cohort study in Sweden. Diabet Med
Rode L, Nordestgaard BG, Bojesen SE. Long telo- 2010;27:791–​7.
meres and cancer risk among 95 568 individu- Silverberg MJ, Leyden W, Warton EM, Quesenberry
als from the general population. Int J Epidemiol CP Jr, Engels EA, Asgari MM. HIV infection sta-
2016;45:1634-​1643. tus, immunodeficiency, and the incidence of
379

Skin Cancer 379

non-​ melanoma skin cancer. J Natl Cancer Inst activity promotes development of malignant mel-
2013;105:350–​60. anoma. Cancer Res 2004;64:7002–​10.
Sinclair CA, Makin JK, Tang A, Brozek I, Rock V. Staples MP, Elwood M, Burton RC, Williams JL,
The role of public health advocacy in achieving Marks R, Giles GG. Non-​melanoma skin cancer
an outright ban on commercial tanning beds in in Australia: the 2002 national survey and trends
Australia. Am J Public Health 2014;104:e7–​9. since 1985. Med J Aust 2006;184:6–​10.
Singh S, Nagpal SJ, Murad MH, Yadav S, Kane SV, Pardi Stark M, Hayward N. Genome-​wide loss of hetero-
DS et al. Inflammatory bowel disease is associated zygosity and copy number analysis in melanoma
with an increased risk of melanoma:  a system- using high-​ density single-​
nucleotide polymor-
atic review and meta-​analysis. Clin Gastroenterol phism arrays. Cancer Res 2007;67:2632–​42.
Hepatol 2014;12:210–​8. Steingrimsson E, Copeland NG, Jenkins NA.
Skender-​Kalnenas TM, English DR, Heenan PJ. Benign Melanocytes and the microphthalmia tran-
melanocytic lesions:  risk markers or precursors scription factor network. Annu Rev Genet
of cutaneous melanoma? J Am Acad Dermatol 2004;38:365–​411.
1995;33:1000–​7. Suarez B, Lopez-​Abente G, Martinez C, Navarro C,
Soleas GJ, Grass L, Josephy PD, Goldberg DM, Tormo MJ, Rosso S et al. Occupation and skin can-
Diamandis EP. A comparison of the anticarcino- cer: the results of the HELIOS-​I multicenter case-​
genic properties of four red wine polyphenols. control study. BMC Public Health 2007;7:180.
Clin Biochem 2002;35:119–​24. Sugiyama H, Misumi M, Kishikawa M, Iseki M,
Song F, Qureshi AA, Gao X, Li T, Han J. Smoking and Yonehara S, Hayashi T et al. Skin cancer incidence
risk of skin cancer:  a prospective analysis and a among atomic bomb survivors from 1958 to 1996.
meta-​analysis. Int J Epidemiol 2012a;41:1694–​705. Radiat Res 2014;181:531–​9.
Song F, Qureshi AA, Han J Increased caffeine intake Tang JY, Spaunhurst KM, Chlebowski RT, Wactawski-​
is associated with reduced risk of basal cell car- Wende J, Keiser E, Thomas F et  al. Menopausal
cinoma of the skin. Cancer Res 2012b;72:3282–​9. hormone therapy and risks of melanoma and
Sorensen HT, Mellemkjaer L, Nielsen GL, Baron JA, nonmelanoma skin cancers:  women’s health ini-
Olsen JH, Karagas MR. Skin cancers and non-​ tiative randomized trials. J Natl Cancer Inst
Hodgkin lymphoma among users of systemic glu- 2011;103:1469–​75.
cocorticoids:  a population-​based cohort study. J Teh MT, Blaydon D, Chaplin T, Foot NJ, Skoulakis S,
Natl Cancer Inst 2004;96:709–​11. Raghavan M et al. Genomewide single nucleotide
Soufir N, Moles JP, Vilmer C, Moch C, Verola O, Rivet polymorphism microarray mapping in basal cell
J et al. P16 UV mutations in human skin epithelial carcinomas unveils uniparental disomy as a key
tumors. Oncogene 1999;18:5477–​81. somatic event. Cancer Res 2005;65:8597–​603.
Spatz A, Giglia-​ Mari G, Benhamou S, Sarasin A. Tessari G, Girolomoni G. Nonmelanoma skin can-
Association between DNA repair-​deficiency and cer in solid organ transplant recipients:  update
high level of p53 mutations in melanoma of xero- on epidemiology, risk factors, and management.
derma pigmentosum. Cancer Res 2001;61:2480–​6. Dermatol Surg 2012;38:1622–​30.
Spencer JM, Kahn SM, Jiang W, DeLeo VA, Weinstein Thomas NE, Alexander A, Edmiston SN, Parrish E,
IB. Activated ras genes occur in human actinic ker- Millikan RC, Berwick M et  al. Tandem BRAF
atoses, premalignant precursors to squamous cell mutations in primary invasive melanomas. J
carcinomas. Arch Dermatol 1995;131:796–​800. Invest Dermatol 2004;122:1245–​50.
Stacey SN, Gudbjartsson DF, Sulem P, Bergthorsson Tilli CM, Van Steensel MA, Krekels GA, Neumann
JT, Kumar R, Thorleifsson G et al. Common vari- HA, Ramaekers FC. Molecular aetiology and path-
ants on 1p36 and 1q42 are associated with cutane- ogenesis of basal cell carcinoma. Br J Dermatol
ous basal cell carcinoma but not with melanoma 2005;152:1108–​24.
or pigmentation traits. Nat Genet 2008;40:1313–​8. Tong LX, Young LC. Nutrition:  the future of mel-
Stacey SN, Sulem P, Gudbjartsson DF, Jonasdottir A, anoma prevention? J Am Acad Dermatol
Thorleifsson G, Gudjonsson SA et  al. Germline 2014;71:151–​60.
sequence variants in TGM3 and RGS22 con- Vajdic CM, van Leeuwen MT, Webster AC, McCredie
fer risk of basal cell carcinoma. Hum Mol Genet MR, Stewart JH, Chapman JR et  al. Cutaneous
2014;23:3045–​53. melanoma is related to immune suppression in
Stacey SN, Sulem P, Masson G, Gudjonsson SA, kidney transplant recipients. Cancer Epidemiol
Thorleifsson G, Jakobsdottir M et  al. New com- Biomarkers Prev 2009;18:2297–​303.
mon variants affecting susceptibility to basal cell Van Allen EM, Miao D, Schilling B, Shukla SA, Blank
carcinoma. Nat Genet 2009;41:909–​14. C, Zimmer L et al. Genomic correlates of response
Stahl JM, Sharma A, Cheung M, Zimmerman M, to CTLA-​ 4 blockade in metastatic melanoma.
Cheng JQ, Bosenberg MW et al. Deregulated Akt3 Science. 2015;350:207–​11.
380

380 Textbook of Cancer Epidemiology

van der Pols JC, Heinen MM, Hughes MC, Ibiebele Wiesner T, Obenauf AC, Murali R, Fried I, Griewank
TI, Marks GC, Green AC. Serum antioxidants and KG, Ulz P et  al. Germline mutations in BAP1
skin cancer risk: an 8-​year community-​based fol- predispose to melanocytic tumors. Nat Genet
low-​up study. Cancer Epidemiol Biomarkers Prev 2011;43:1018–​21.
2009;18:1167–​73. Wiren S, Haggstrom C, Ulmer H, Manjer J, Bjorge T,
van der Schroeff JG, Evers LM, Boot AJ, Bos JL. Ras Nagel G et al. Pooled cohort study on height and
oncogene mutations in basal cell carcinomas and risk of cancer and cancer death. Cancer Causes
squamous cell carcinomas of human skin. J Invest Control 2014;25:151–​9.
Dermatol 1990;94:423–​5. Wolfe F, Michaud K. Biologic treatment of rheuma-
Viros A, Fridlyand J, Bauer J, Lasithiotakis K, Garbe C, toid arthritis and the risk of malignancy: analyses
Pinkel D et al. Improving melanoma classification from a large US observational study. Arthritis
by integrating genetic and morphologic features. Rheum 2007;56:2886–​95.
PLoS Med 2008;5:e120. Wong SS, Tan KC, Goh CL. Cutaneous manifestations
Wallingford SC, Olsen CM, Plasmeijer E, Green AC. of chronic arsenicism: review of seventeen cases. J
Skin cancer arising in scars:  a systematic review. Am Acad Dermatol 1998;38:179–​85.
Dermatol Surg 2011;37:1239–​44. Wu H, Reeves KW, Qian J, Sturgeon SR. Coffee, tea,
Wang J, Li X, Zhang D. Coffee consumption and the and melanoma risk among postmenopausal
risk of cutaneous melanoma: a meta-​analysis. Eur women. Eur J Cancer Prev 2015a;24:347–​52.
J Nutr 2016;55:1317–​29. Wu S, Han J, Feskanich D, Cho E, Stampfer MJ,
Watt TC, Inskip PD, Stratton K, Smith SA, Kry SF, Willett WC et  al. Citrus consumption and risk
Sigurdson AJ et  al. Radiation-​related risk of basal of cutaneous malignant melanoma. J Clin Oncol
cell carcinoma: a report from the Childhood Cancer 2015b;33:2500–​8.
Survivor Study. J Natl Cancer Inst 2012;104:1240–​50. Yang RX, Pfeiffer RM, Goldstein AM. Influence of glu-
Wehner MR, Shive ML, Chren MM, Han J, Qureshi tathione-​S-​transferase (GSTM1, GSTP1, GSTT1)
AA, Linos E. Indoor tanning and non-​melanoma and cytochrome p450 (CYP1A1, CYP2D6) poly-
skin cancer: systematic review and meta-​analysis. morphisms on numbers of basal cell carcinomas
BMJ 2012;345:e5909. (BCCs) in families with the naevoid basal cell car-
Whalen FM, Jambusaria-​ Pahlajani A, Speck RM, cinoma syndrome. J Med Genet 2006;43:e16.
Schmitz K, Schmults CD. Effect of physical activ- Yokoyama S, Woods SL, Boyle GM, Aoude LG,
ity on nonmelanoma skin cancer risk in kidney, MacGregor S, Zismann V et al. A novel recurrent
liver, and pancreatic transplant patients. Dermatol mutation in MITF predisposes to familial and
Surg 2010;36:1510–​3. sporadic melanoma. Nature 2011;480:99–​103.
Whiteman DC, Green AC. Melanoma and sun Yoshinaga S, Hauptmann M, Sigurdson AJ, Doody
exposure:  where are we now? Int J Dermatol MM, Freedman DM, Alexander BH et  al.
1999;38:481–​9. Nonmelanoma skin cancer in relation to ionizing
Whiteman DC, Green AC, Olsen CM. The grow- radiation exposure among U.S.  radiologic tech-
ing burden of invasive melanoma: Projections of nologists. Int J Cancer 2005;115:828–​34.
incidence rates and numbers of new cases in six Youssef KK, Lapouge G, Bouvree K, Rorive S,
susceptible populations through 2031. J Invest Brohee S, Appelstein O et al. Adult interfollicular
Dermatol 2016;136:829–​37. tumour-​initiating cells are reprogrammed into an
Whiteman DC, Parsons PG, Green AC. p53 expres- embryonic hair follicle progenitor-​like fate dur-
sion and risk factors for cutaneous melanoma:  a ing basal cell carcinoma initiation. Nat Cell Biol
case-​control study. Int J Cancer 1998;77:843–​8. 2012;14:1282–​94.
Whiteman DC, Stickley M, Watt P, Hughes MC, Davis Zanetti R, Rosso S, Martinez C, Nieto A, Miranda A,
MB, Green AC. Anatomic site, sun exposure, Mercier M et  al. Comparison of risk patterns in
and risk of cutaneous melanoma. J Clin Oncol carcinoma and melanoma of the skin in men:  a
2006;24:3172–​7. multi-​centre case-​case-​control study. Br J Cancer
Whiteman DC, Watt P, Purdie DM, Hughes MC, 2006;94:743–​51.
Hayward NK, Green AC. Melanocytic nevi, solar Zhao H, Shu G, Wang S. The risk of non-​melanoma
keratoses, and divergent pathways to cutaneous skin cancer in HIV-​infected patients: new data and
melanoma. J Natl Cancer Inst 2003;95:806–​12. meta-​analysis. Int J STD AIDS 2016;27:568–​75.
Whiteman DC, Whiteman CA, Green AC. Childhood Zuo L, Weger J, Yang Q, Goldstein AM, Tucker MA,
sun exposure as a risk factor for melanoma: a sys- Walker GJ et  al. Germline mutations in the
tematic review of epidemiologic studies. Cancer p16INK4a binding domain of CDK4 in familial
Causes Control 2001;12:69–​82. melanoma. Nat Genet 1996;12:97–​9.
381

16
Breast Cancer
R U L L A TA M I M I , S U S A N H A N K I N S O N , A N D P A G O N A   L A G I O U

B reast cancer has probably attracted more sci-

entific interest among epidemiologists than
any other malignancy. The profound impor-
intrinsic subtypes including, but not limited to
luminal A, luminal B, HER2-​overexpressing, and
triple negative cancers (Goldhirsch et al, 2013).
tance of reproductive factors was first revealed Symptoms Prior to mammographic screening,
in the 18th century, when the Italian researcher a palpable tumor was the most common ini-
Ramazzini noted that breast cancer was more tial symptom. Occasionally, however, changes
common among nuns than among other women. in the nipple or spread to axillary lymph nodes
During the last decades of the 20th century, thou- or distant metastases, often in the lung or bone,
sands of research papers described in increasing prompted a consultation. Today, many breast
detail the associations between a woman’s repro- cancers are diagnosed before symptoms occur by
ductive life and her risk of developing breast means of mammography.
cancer. In addition, large natural experiments
Diagnostic Methods When breast cancer is sus-
on breast cancer risk began when oral contracep-
pected on the basis of clinical examination or
tives came into wide use in the 1960s, and later
mammography, pathologic confirmation is a
on when an increasing proportion of women
necessary prelude to definitive primary treat-
began using postmenopausal hormone therapy.
ment. Fine needle aspiration, core biopsy, or sur-
During the 1990s, accumulating evidence that
gical excision are used to establish the diagnosis.
a woman’s risk may be influenced even in utero
further complicated our view of breast cancer Treatment Until the early 1980s, extensive sur-
etiology. More recently we have come to appre- gical removal of the breast and regional lymph
ciate that breast cancer is not a single disease but nodes was believed to improve both local tumor
rather a heterogeneous group of subtypes with control and long-​term cure. Accumulating sci-
different etiologies. Yet the wealth of available entific data, mostly from clinical trials, forced a
epidemiologic information can be synthesized radical rethinking. Breast-​ conserving surgery,
into a consistent and testable, albeit still hypo- with or without preoperative chemotherapy, has
thetical, causal model. been used increasingly in combination with post-
operative radiation therapy, which limits local
CLINICAL SYNOPSIS recurrences. Moreover, adjuvant treatment with
Subgroups Virtually all breast cancers are ade- postoperative chemotherapy or tamoxifen, an
nocarcinomas. In general clinical practice, the antiestrogen, has become part of routine treat-
distinction between different histopathologic ment. Current clinical guidelines also recommend
types has limited implications for either diag- the inclusion of aromatase inhibitors, either alone
nosis or treatment. Cancer in situ of the breast, or after tamoxifen therapy, for postmenopausal
notably the ductal (but not the lobular) type, is women with estrogen receptor positive (ER+)
now detected more often due to widespread use breast cancer. Additions to treatment options are
of mammography. The etiology and natural his- targeted adjuvant therapies, which include mon-
tory of in situ lesions and their relation to inva- oclonal antibodies and tyrosine kinase inhibitors,
sive cancer are largely unknown. Therefore, this both used against for HER2-​positive tumors, as
chapter focuses on invasive cancer only. Current well as cyclin-​dependent kinase inhibitors.
methods of breast cancer classification group Prognosis In the United States, overall 5-​ year
tumors into genetically and molecularly defined relative survival among women diagnosed in
382

382 Textbook of Cancer Epidemiology

1986–​93 was 84.2%, and this increased to 91% in 40  years of age. In the United States, incidence
2007. This reflects both earlier detection through rates per 100,000 women are 25 in the 30–​34
mammography and improved treatment. There age group, 190 in the 40–​45 age group, and 455
continues to be a difference in survival rates in the 70–​74 age group (SEER, 2014b). Of note,
by race; 5-​ year relative survival is substan- the rapid rise in rates with increasing age slows
tially higher among white compared with black somewhat around age 50, near the time of men-
women. In Europe, 5-​year relative survival rates opause, which strongly suggests a role for repro-
increased from about 77% in 1997–​2002 to 82.5% ductive hormones in the etiology of the disease.
in 2004–​2009 and have generally been higher in In 2008, breast cancer incidence rates var-
western compared to eastern European countries. ied by more than fivefold internationally (ACS,
Progress Randomized controlled trials have 2011). The highest rates are observed in Europe
documented that breast cancer mortality can and North America and the lowest rates in Asia
be reduced by early detection through mam- (Fig 16-​2). This likely reflects a combination of
mography screening and by adjuvant systemic difference in breast cancer screening, reporting,
treatment following primary surgery. Increasing and risk factor exposures. Migrant studies, in
public awareness may also have contributed to which changes in breast cancer rates are evalu-
earlier clinical diagnosis. Thanks to an improved ated in women who move from low-​to high-​
prognosis, although breast cancer incidence has risk countries—​or vice versa—​have shown that
increased in several countries, mortality has the rates of the host country are assumed over
remained fairly stable and has even been declin- time, frequently one or two generations later
ing in high-​income countries since the 1990s. (Tominaga, 1985; Ziegler et al, 1993). These data
indicate that international differences in breast
DESCRIPTIVE cancer rates may be due, at least in part, to envi-
EPIDEMIOLOGY ronmental and lifestyle differences.
Breast cancer is the most commonly occurring Data from a cancer registry in the state of
cancer in women, with an estimated 1.67  mil- Connecticut show that breast cancer rates have
lion cases newly diagnosed worldwide in 2012. been steadily increasing, by approximately 1.4%
Overall, breast cancer accounts for 25% of all per year, since the 1950s, with a more marked
cancer diagnoses in women (IARC, 2014). increase in the 1980s (Ravdin et al, 2007). The
The shape of the age incidence curves is gen- reasons for this increase are unclear, although
erally similar across countries, but with big differ- changes in reproductive patterns and other life-
ences in the absolute rates at every age (Fig. 16-​1). style factors, for example, increasing obesity in
Overall, rates increase substantially with increas- postmenopausal women and use of postmeno-
ing age. The diagnosis is rare in women under pausal hormones, almost certainly have played

500
New cases per 100,000

450
400
person-years

350
300
250
200
150
100
50
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States

FIGURE 16-​1  Age-​specific incidence rates of breast cancer among women.

Source: Ferlay et al, 2013.
38

Breast Cancer 383

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States
0 20 40 60 80 100 120
ASR (world) per 100,000 person-years

FIGURE 16-​2  Age-​standardized (to the 2012 world population) incidence rates of breast cancer among women.
Source: Ferlay et al, 2013.

an important role. Much of the increase in rates women have a 3.4% (30.8 per 100,000) risk of
in the 1980s was attributable to increased diag- dying from breast cancer compared to a 3.1%
nosis of small (≤2 cm), localized breast tumors (22.7 per 100,000) for white women (Carey et al,
attributed in part to widespread uptake of 2006; SEER, 2014a).
screening mammography. Additionally, from
2001 to 2004, a decrease in incidence rates
(3.5%) occurred, likely due, at least in part, to GENETIC AND MOLECULAR
cessation of hormone therapy following the EPIDEMIOLOGY
2002 Women’s Health Initiative (WHI) report Inherited Susceptibility
that estrogen plus progestin menopausal hor- A family history of breast cancer in a first-​degree
mone therapy increased breast cancer risk for relative is an established risk factor. A  relative
postmenopausal women (Ravdin et  al, 2007; risk (RR) of 1.5–​3.0 has generally been found
Rossouw et  al, 2002). Since the 1950s, breast when women whose mother or sister had breast
cancer rates have been increasing in a number cancer are compared to those whose first-​degree
of other countries as well, including low-​risk female relatives did not have breast cancer
countries such as China (Jin et al, 1993), Japan (Greene, 1997). Among women whose mother
(Nagata et al, 1997), and Singapore (Chia et al, and sister had breast cancer, particularly with an
2005). Changes in reproductive patterns, along early age at onset, the risk is substantially higher.
with dramatic changes in other aspects of life- Below we describe associations with breast can-
style such as diet and physical activity, likely cer for specific genetic variants according to
account for much of this increase. penetrance.
Within the United States, Hispanic, Asian,
and American Indian women have the lowest High-​Penetrance Gene Mutations
incidence rates of breast cancer, while white and It is estimated that 5% to 10% of all breast cancers
African American women have the highest inci- can be attributed to highly penetrant germline
dence rates. Overall, white US women have the mutations (Bennett et al, 1999). This proportion
highest lifetime risk (12.8%) of breast cancer, varies by age, with about one-​third of breast can-
while African American women have a slightly cer cases in women younger than 30 years of age
lower lifetime risk of 10.1% (SEER, 2014a). attributed to inherited factors (Ellisen & Haber,
African American women have higher incidence 1998). Breast cancer gene 1 (BRCA1), located
rates before age 40 and also have higher rates on chromosome 17q21, was identified in 1994
of more aggressive breast cancer subtypes such (Ellisen & Haber, 1998). Two years later BRCA2,
as estrogen receptor negative (ER-​) breast can- located on chromosome 13q12-​13, was identi-
cers compared with white women (ACS, 2013). fied. Accumulating evidence suggests that BRCA1
Although white women have higher lifetime risk functions in cell-​ cycle control, genetic stabil-
and age-​adjusted incidence rates, they have lower ity, and repair of DNA damage (Deng, 2006).
mortality rates than African American women A mutation in at least one of these genes is esti-
(ACS, 2013; SEER, 2014a). African American mated to cause 2% to 5% of all breast cancers,
384

384 Textbook of Cancer Epidemiology

with the highest frequency in those with multi- individuals that carry this mutation (Consortium,
ply affected first-​degree relatives and early onset 2004; Meijers-​Heijboer et al, 2002). CHEK2 is a
of disease. Up to age 40, BRCA1 mutation car- tumor suppressor gene. The 1100delC mutation
riers are estimated to have a 20-​fold greater risk leads to a truncated protein, in which kinase
of breast cancer compared to the general popu- activity of the protein is abolished.
lation and a lifetime risk of 60% to 85% (Chen & Ataxia-​telangiectasia (AT) is inherited in an
Parmigiani, 2007; Ellisen & Haber, 1998), whereas autosomal recessive pattern; homozygotes expe-
the lifetime risk of breast cancer among women rience a range of severe, progressive neurologic
who are carriers of BRCA2 mutations is estimated and immunologic effects. An estimated 1.4% of
to be approximately 50% (Chen & Parmigiani, the general population are heterozygotes (Bennett
2007). Factors that modify penetrance most et  al, 1999; Greene, 1997; Warmuth et  al, 1997),
strongly appear to be reproductive factors and with a markedly increased susceptibility to DNA
exogenous hormones (Narod, 2006). Estimates damage through ionizing radiation, chemother-
of the prevalence of BRCA1 or BRCA2 mutations apy, or hydroxy radicals. Additional moderate pen-
in the general population range from 1/​200 to 1/​ etrance gene mutations include BRIP1(BACH1)
1,000 (Ellisen & Haber, 1998). Tumors associated (Seal et al, 2006) and PALB2 (Rahman et al, 2007),
with BRCA1 mutations tend to have a basal-​like both of which are associated with a 2-​to 2.3-​fold
phenotype (ER-​/​PR-​/​HER2-​) (Rakha et al, 2008), increased risk of breast cancer.
while those associated with BRCA2 are more sim-
ilar to sporadic tumors (Lakhani et al, 2002). Low-​Penetrance Polymorphisms
The p53 is a tumor suppressor gene associ- Hundreds of papers have now been published on
ated with hereditary breast cancer. Breast can- associations between common (>5% allele fre-
cer is one of several cancers that can occur in quency) polymorphisms and breast cancer risk.
women with the rare Li-​Fraumeni familial can- Initial candidate gene studies focused on varia-
cer syndrome, which is linked to mutations in tion in carcinogen-​metabolizing genes, genes in
p53 (Ellisen & Haber, 1998). Germline muta- the steroid metabolism pathways, growth factor,
tions in p53 have low prevalence in the general cell cycle, apoptosis, and inflammation pathways.
population (<1%) and are thus responsible for Due to the small to modest predicted relative
few breast cancers (Bennett et al, 1999; Greene, risks, and the comparatively small sample sizes
1997). However, somatic p53 mutations may also of most studies, most initial reports were not rep-
play a role in tumor progression, independent of licated (Dunning et  al, 1999). The development
predisposition, and have been found in 50% of of genome-​wide association studies (GWAS) and
sporadic breast cancers (Ellisen & Haber, 1998). improvements in imputation have provided an
By age 50, 30% to 50% of women with unbiased test of association at multiple loci, and
Cowden’s disease—​also called multiple hamar- introduce some order into the search for low-​
toma syndrome—​develop breast cancer, includ- penetrance alleles. The challenges of replication
ing bilateral cancer (Bennett et  al, 1999). The due to the unprecedented multiple comparisons
PTEN gene (phosphatase and tensin homologue problem inherent in this approach have been
deleted on chromosome 10), which is responsible addressed by the formation of large international
for this disease, is located on chromosome 10q23 consortia that have included thousands of breast
(Greene, 1997) and has an autosomal dominant cancer cases and controls. GWAS in women of
inheritance pattern. PTEN functions as a tumor European ancestry have reproducibly identified
suppressor gene, and participates in intracell sig- over 90 common (minor allele frequency > 0.05)
naling and control of cell division. Women with risk loci for breast cancer; the variants at these
a PTEN mutation are estimated to have an 85% loci explain approximately 16% of the familial
lifetime risk of breast cancer (Tan et  al, 2012). risk of breast cancer (Michailidou et al, 2015).
Additional rare, highly penetrant gene muta-
tions include CDH1 (E-​cadherin) (Pharoah et al,
2001) and STK11 (Boardman et al, 1998). RISK FACTORS
Tobacco
Moderate-​Penetrance Gene Mutations The relationship between cigarette smoking and
A rare variant (~1.1%) identified in the CHEK2 breast cancer risk has been evaluated in many
gene has been reported to confer a 1.5-​ fold studies. Initial reports, overall, did not support
to 2.0-​
fold increase in breast cancer risk in any important association (Laden & Hunter,
385

Breast Cancer 385

1998; Palmer & Rosenberg, 1993). However, sev- and hundreds of epidemiologic and laboratory
eral studies suggest that there may be an asso- studies have been conducted to evaluate the rela-
ciation particularly among women with early tionship. The interpretation of the results of these
exposure prior to first pregnancy, long durations studies has been notably controversial, and these
of smoking, or women with specific genetic data are therefore reviewed in some detail.
mutations. Large cohort studies have reported
increased risks for breast cancer among women Animal Studies and Biological Mechanisms
with the longest durations of smoking; those The interpretation of animal studies of high-​fat
with significant findings had RRs ranging from diets in relation to mammary tumors has been
1.2 to 1.5 comparing long-​term smokers with controversial, in large part due to the difficulty
never smokers (Al-​Delaimy et  al, 2004; Calle of separating the effect of energy from that of
et  al, 1994; Cui et  al, 2006; Gram et  al, 2005). fat intake per se. It is well established that high-​
A meta-​analysis of 15 cohort studies found that calorie diets result in higher mammary tumor
current (RR 1.1, 95% CI 1.1–​1.2) and former rates in animals (Hunter & Willett, 1993).
smoking (RR 1.1, 95% CI 1.0–​1.2) were weakly Whether an independent effect of fat intake
associated with breast cancer risk (Gaudet et al, exists, after holding energy intake constant, is
2013a). A  stronger association (RR 1.2, 95% less clear (Stephen & Wald, 1990; Willett, 1998).
CI 1.1–​1.3) was reported in women who initiated In terms of biological mechanisms for an effect
smoking before first birth. The Canadian Expert of fat intake on breast cancer risk, a number of
Panel on Tobacco Smoke and Breast Cancer Risk studies have evaluated the influence of fat intake
(2009) (Johnson et al, 2011) concluded that the on circulating hormone levels. However, impor-
relationship between active smoking and breast tantly, the majority did not have a concurrent
cancer was consistent with causality. They also control group, did not adequately account for
considered the association between second- changes in weight associated with the change in
hand smoke and premenopausal breast cancer diet—​important because lower weight is known
as consistent with causality, however there was to be associated with lower hormone levels, and
insufficient evidence for postmenopausal breast were generally short-​term. Thus, any effect of fat
cancer. These results are in line with the hypoth- intake on hormone levels remains uncertain.
esis that breast tissue is particularly susceptible
to carcinogens between early puberty and the Descriptive Data
first full term pregnancy (Russo et al, 1982). The hypothesis that fat intake increases the risk of
Because secondhand smoke may contain breast cancer was initially motivated by the strong
more carcinogens than direct exposure due to international correlation between per capita fat
the lack of filtration (Laden & Hunter, 1998), intake and breast cancer mortality (Willett, 1998).
there has been some interest in evaluating passive Although this type of analysis is important as a
smoking exposure as a cause of breast cancer. In first approach to evaluate differences in disease
several case-​control studies and one large cohort occurrence among countries, it is limited both by
study of cancer mortality, either statistically sig- the quality of the data and by the inability to con-
nificant yet modest or nonsignificant increases in trol for other possible risk factors. For example, the
risk have been observed; again, no dose-​response countries characterized by the lowest and highest
relationship was seen. Although additional study breast cancer rates also tend to be very low-​and
is needed to resolve this issue, it is of limited pub- very high-​income countries, respectively.
lic health importance because of the overwhelm-
ing evidence of the deleterious effects of cigarette Analytic Epidemiologic Studies
smoking overall. In a meta-​ analysis of 12 case-​ control stud-
ies including a total of 4,312 cases and 5,978
Diet controls, the RR of breast cancer for a 100-​g
Fat Intake increase in total fat intake was 1.35 overall and
Overall, observational studies do not support any 1.48 among postmenopausal women (Howe et al,
important relationship between total dietary fat 1990). However, most women with a typical daily
intake in adulthood and the risk of breast can- intake of 1,600 to 1,800 calories consume 50 to 80
cer. However, the potential role of dietary fat g of fat daily, and a reduction of 100 g/​day would
intake in breast carcinogenesis has been of con- not be possible. Based on a more realistic change
siderable interest over the last several decades, in intake of 25 g of fat per day, the RR from the
386

386 Textbook of Cancer Epidemiology

meta-​analysis would be 1.08 overall and 1.10 for In an updated pooled analysis of 7,329 incident
postmenopausal women. Relative risks of this invasive breast cancer cases occurring among
magnitude could easily be due to bias, either 351,821 women, the pooled RRs (95% CI) for an
related to differences between cases and controls increment of 5% of energy were 1.09 (1.00–​1.19)
in recalling past diet, or selection bias, or both. for saturated fat, 0.93 (0.84–​1.03) for monoun-
In a pooled analysis of six prospective stud- saturated fat, and 1.05 (0.96–​1.16) for polyun-
ies including 4,980 cases among 337,819 women saturated fat compared with equivalent energy
(Hunter et al, 1996), no association was noted for intake from carbohydrates (Smith-​Warner et al,
total fat intake; the RR associated with a 25-​g/​ 2001). Intake of olive oil specifically could not be
day increment in intake was 1.02. Long-​standing addressed in the pooled analysis. A  small ran-
concerns related to observational studies of this domized controlled trial of Mediterranean diet
association were that fat intake could not be ade- supplemented with extra-​virgin olive oil showed
quately assessed by means of a questionnaire—​ a significantly reduced risk of breast cancer (RR
substantial measurement error would result in 0.32, 95% CI 0.13–​0.79) in the intervention group
missing a true-​ positive association—​ and that compared with the control group after a median
the range of fat intake in the populations stud- follow-​up of 4.8 years (Toledo et al, 2015).
ied was too narrow and did not overlap the low The relationship between total and specific
intakes found in countries with the lowest breast subtypes of fat and fatty acids was evaluated in
cancer mortality rates. In each of the studies in a combined study of Nurses’ Health Study and
the pooled analysis, a validation study was con- Nurses’ Health Study II cohorts (Boeke et  al,
ducted among a subset of participants to allow 2014). This large study of 9,979 breast cancer
an assessment of the validity of the reported diet. cases benefited from having repeated dietary
After using statistical methods to correct for the data collected over more than 20 years of follow-​
measurement error (using the validation data), up. Overall, intakes of total fat and other specific
the comparable relative risk for total fat was 1.07, types of fat were not associated with risk of total
suggesting, again, little if any association between or lethal breast cancer in multivariable-​adjusted
fat intake and risk. models. For example, the RR for the highest ver-
In a large scale US randomized controlled sus lowest quintile of total fat intake and total
trial, the Women’s Health Initiative (WHI), the breast cancer was 1.02, (95% CI 0.96–​1.10; p-​
effect of a low-​fat dietary pattern on risk of breast trend = 0.88). Additionally, there was no associa-
cancer was evaluated. The dietary intervention tion between individual subtypes of fat with total
consisted of a low fat (20% of calories), high fruit or lethal breast cancer in this study.
and vegetable (≥5 servings/​day), and high grain
(≥6 servings/​day) diet. After 8  years of follow-​ Fiber Intake
up, the dietary intervention group experienced Fiber intake has been hypothesized to decrease
a nonsignificant 9% lower incidence of breast the risk of breast cancer by inhibiting the intes-
cancer compared to the comparison group (RR, tinal reabsorption of estrogens secreted via the
0.91, 95% CI 0.83–​1.01) (Prentice et  al, 2006). biliary tract, resulting in lower circulating estro-
These results suggest that short-​term reductions gen levels. In a meta-​analysis of 16 prospective
in fat intake during the postmenopausal years studies including 26,523 breast cancer cases
do not reduce the risk of breast cancer. However, among 999,271 participants, there is evidence
this does not preclude an association between of a decreased risk of breast cancer with increas-
fat intake earlier in life and subsequent breast ing fiber intake. The summary RR for a 10-​g/​day
cancer risk. increase of total dietary fiber was 0.95, (95% CI
Fat subtypes have also been examined in a 0.91–​0.98) (Aune et al, 2012). It is important to
number of studies. In several case-​control stud- recognize that “fiber” can be further divided into
ies, intake of olive oil specifically was inversely a number of subtypes, for example, soluble and
related to breast cancer risk (Trichopoulou, 1995; insoluble, as well as the food sources of fiber (e.g.,
Trichopoulou et  al, 1995; Willett, 1998). In one vegetable fiber, cereal fiber). In the same meta-​
large prospective analysis, monounsaturated fats analysis, there was evidence that intake of soluble
were inversely related to risk (Wolk et al, 1998), fiber, but not other fiber subtypes (i.e., insoluble
although this was not observed in the pooled fiber, fruit fiber, vegetable fiber, or cereal fiber)
analysis described earlier (Hunter et  al, 1996). was inversely associated with breast cancer risk.
387

Breast Cancer 387

It is important to note that fewer studies were Epidemiologic data from over 50 case-​control
able to contribute to analyses of fiber subtype and 15 prospective cohort studies have reported
and power was more limited. Only three stud- inconsistent results on the association between
ies could be included in the analysis of fiber and intake of carotenoids and overall breast cancer
breast cancer subtype; however, these data sug- risk (AICR, 2007); however, this may reflect dif-
gest that fiber intake is more strongly associated ferences in association with breast cancer sub-
with ER-​ /​
PR-​breast cancers (RR comparing types. In a large pooled analysis of 18 cohort
high vs. low fiber intake 0.76, 95% CI 0.52–​1.11) studies, there was evidence that intakes of alpha-​
(Aune et al, 2012). Given the limited number of carotene, beta-​ carotene, and lutein/​zeaxanthin
modifiable risk factors especially for ER-​breast were inversely associated with risk of the less
cancer, further research in this area is warranted. common ER-​breast cancer, but not ER+ breast
cancer. For example, the multivariable adjusted
Micronutrient Intake RR comparing the highest quintile of beta-​
Of the micronutrients, the epidemiologic data carotene intake with the lowest was 0.84, (95% CI
most strongly support a protective role for vita- 0.77–​0.93) for ER-​breast cancer and 1.04, (95%
min A and the carotenoids. However, the protec- CI 0.98–​1.10) for ER+ breast cancer.
tive effect, if any, may be limited to specific breast Micronutrients, specifically carotenoids,
cancer subtypes. exhibit a great deal of interindividual variation
in their absorption, metabolism, and excre-
Carotenoids and Vitamin A tion (Mathews-​Roth, 1990; Stahl & Sies, 1992);
Vitamin A consists of both preformed vitamin A, therefore, blood levels may give a more accurate
that is, retinol, primarily from animal products, measurement of the bioavailable amount of these
and several carotenoids, such as beta-​carotene, nutrients than dietary intake estimates. A pooled
which are found primarily in fruits and vegeta- analysis of circulating carotenoids (Eliassen et al,
bles and can be converted to retinol. The carot- 2012a) found that total carotenoids were associ-
enoids are well-​established antioxidants and, as ated with a significant 19% reduced breast cancer
such, may protect against cellular and DNA dam- risk (Fig. 16-​3). For many of the carotenoids, the
age caused by reactive oxygen species. Vitamin inverse association was stronger for ER-​cancer
A  is important in maintaining cellular differen- than for ER+ cancer. For example, circulating
tiation and thus may help prevent the emergence beta-​carotene had a 48% reduced risk of ER-​can-
of cells with a malignant phenotype. cer (RR comparing highest vs. lowest quintile 0.5,

2
Overall ER+Breast Cancer ER-Breast Cancer
1.8

1.6

1.4
Relative Risk

1.2

0.8

0.6

0.4
p-trend = 0.01 p-trend = 0.21 p-trend = 0.08
0.2

0
Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5
Quintile of Plasma Total Carotenoids

FIGURE 16-​3  Relative risk of breast cancer according to quintile of plasma carotenoids by overall and ER status.
Source: Data from Eliassen et al, 2012a.
38

388 Textbook of Cancer Epidemiology

95% CI 0.4–​0.8); in contrast, only a 17% reduced

risk of ER + breast cancer (RR 0.8, 95% CI 0.7–​1.0; there was no evidence of an association between
p-​heterogeneity  =  0.01). Additionally, because folate intake and breast cancer overall or by
there is a high degree of collinearity between the tumor subtype (Chen et  al, 2014). Alcohol is a
different plasma carotenoids, it is difficult to con- known folate antagonist and thus could plausi-
clude that any specific carotenoid is responsible bly increase the requirement for folate intake.
for the observed inverse associations, but these A number of studies have evaluated whether the
findings do suggest that carotenoids in general association between folate intake and breast can-
are associated with a reduced risk of breast can- cer risk is modified by alcohol consumption. The
cer specifically for ER-​breast cancer (Zhang et al, majority of these studies, but not all, reported
2012b). an inverse association between folate intake and
breast cancer risk among women with high levels
Vitamin C of alcohol consumption. The summary RR com-
Vitamin C is an effective water-​soluble antiox- paring high folate intake to low from six stud-
idant. This nutrient was noted to be inversely ies was 0.60, (95% CI 0.45–​0.82) among women
related to breast cancer risk in the pooled anal- with high alcohol consumption, and 0.92, (95%
ysis of nine case-​control studies (Howe et  al, CI 0.79–​1.07) for women with low alcohol con-
1990). However, this relationship has not been sumption (Chen et al, 2014). These data support
confirmed in a meta-​analysis of prospective stud- the hypothesis that alcohol may modify the asso-
ies (Fulan et al, 2011). ciation between folate and breast cancer risk. In
the United States, flour and grains have been for-
Vitamin E tified with folic acid since 1998. It remains to be
A meta-​analysis including 38 studies on vitamin seen if this pattern of association remains post-​
E and breast cancer reported an inverse associ- fortification, when most of the population will be
ation among case-​control studies, but there was obtaining adequate folate levels.
no association when data were pooled from none
cohort studies (pooled RR comparing highest to Vitamin D
lowest categories: 0.98, 95% CI 0.92–​1.05) (Fulan Biologic, ecologic, and epidemiologic studies
et al, 2011). suggest that vitamin D may be involved in reduc-
ing the risk of breast cancer. Sunlight and die-
Selenium tary intake of fortified dairy products, cereals,
Several studies have evaluated the relationship and supplements are the major contributors to
between selenium status (as measured in blood circulating levels of vitamin D (Salamone et  al,
or toenails) and breast cancer risk because of sele- 1994; Webb et  al, 1988; Webb et  al, 1990). In
nium’s important role in the function of the anti- vitro and in vivo data suggest that vitamin D and
oxidant enzyme glutathione peroxidase (Willett, its receptor function in inhibiting proliferation
1998). In the largest study, one of about six pro- and inducing differentiation (Zinser et al, 2002).
spective studies of this relationship, Hunter and Circulating levels of vitamin D are likely a more
colleagues (Hunter et al, 1990) found no associ- integrated and better measure of vitamin D expo-
ation between selenium and breast cancer risk sure than dietary intake. A meta-​analysis of nine
(RR 1.1, 95% CI 0.70–​1.72 for top versus bottom prospective studies found a modest nonlinear
quintile comparison in toenail selenium levels). inverse association between prediagnostic levels
Most other case-​control and cohort studies sim- of 25-​hydroxyvitamin D (25(OH)D) and breast
ilarly found little evidence for an important pro- cancer among postmenopausal women (Bauer
tective effect of selenium. et al, 2013). In contrast, no association was found
among premenopausal women. Because there
Folate is strong mechanistic evidence and 25(OH)D
Inadequate folate levels may result in abnormal levels can easily be raised through supplemen-
DNA synthesis and disrupted DNA repair and tation, there still remains a great deal of interest
hence may influence breast cancer risk. Several in resolving the association between vitamin D
epidemiologic studies have examined the asso- and breast cancer. A large, ongoing, randomized
ciation between folate intake and circulating controlled trial of vitamin D supplementation
folate levels in relation to breast cancer risk. In may help to better understand this association
a pooled meta-​analysis of prospective studies, (Manson et al, 2012; Pradhan & Manson, 2015).
389

Breast Cancer 389

Phytoestrogens been apparent when all cancers were considered

There is considerable interest in the relation- together. The Pooling Project of Prospective
ship between breast cancer and phytoestrogens, Studies of Diet and Cancer, examined this asso-
compounds with estrogenic activity found in ciation with 993,466 women in 20 cohort studies
plants. Soy intake has been of particular inter- (Jung et al, 2013). Overall, there was no associ-
est because it is consumed in high quantities in ation between fruit and vegetable consumption
Asian countries, where breast cancer rates are and breast cancer risk. However, there was an
low. Both genistein and daidzein, the primary inverse association between vegetable consump-
isoflavones in soy, can bind to estrogen receptors tion and ER-​breast cancer (RR comparing high-
but are much less potent than estradiol—​and est vs. lowest quintile 0.82, 95% CI 0.74–​0.90),
thus could potentially serve as estrogen antago- but no association with ER+ breast cancer (RR
nists (Fournier et al, 1998). In addition, several, comparing highest vs. lowest quintile 1.04, 95%
but not all, small studies have reported lower CI 0.97–​1.11). In addition to the micronutrients
estrogen levels or lengthening of the follicular described above, fruits and vegetables contain
phase in the menstrual cycle in women con- many other phytochemicals, including indoles,
suming soy supplements. These mechanisms are flavonoids, and protease inhibitors, and one or
not well established, however, and more work is more of these substances could play a role in risk
needed. reduction.
Only a few studies have evaluated the asso-
ciation between biomarkers of phytoestrogens Alcohol
and breast cancer risk. In a nested case-​control Alcohol intake is the dietary factor most consist-
study within the Dutch component of European ently associated with breast cancer risk, although
Prospective Investigation into Cancer and the relationship observed has generally been
Nutrition study, plasma levels of isoflavones and modest. Cumulatively, existing evidence sup-
lignans were evaluated in 383 cases and matched ports alcohol as being causally related to breast
controls (Verheus et  al, 2007). The RR for the cancer risk. In a pooled analysis of 53 case-​
highest versus the lowest tertile of genistein was control and cohort studies, a positive association
0.68, (95% CI 0.47–​0.98). Inverse but nonsignifi- was confirmed, with no significant heterogene-
cant associations were observed for daidzein, gly- ity noted across studies (Hamajima et al, 2002).
citein, O-​desmethylangolensin, and equol. There Overall, for every 10-​g increase in alcohol intake,
was no association between circulating lignan the relative risk of breast cancer increased by
levels and breast cancer risk. Similarly, a Japanese 7.1% (95% CI 5.5%–​8.7%); relative to nondrink-
nested case-​control study also reported an inverse ers, women who drank 35–​44 g of alcohol (3–​4
association between plasma genistein, but not drinks) per day had a 46% increase in risk (95%
daidzein, and breast cancer risk (Iwasaki et  al, CI 1.33–​1.61). These studies were conducted in
2008). Three cohort studies in high-​consuming diverse populations, and in most of them, other
countries have examined dietary intake of soy well-​established breast cancer risk factors were
and isoflavones in relation to risk of breast cancer accounted for in the analysis. Additionally, in the
(Lee et al, 2009; Wu et al, 2008; Yamamoto et al, studies in which various types of alcohol were
2003). Of these, two reported statistically signifi- evaluated, beer, wine, and spirits were each posi-
cant modest inverse associations (Wu et al, 2008; tively associated with breast cancer risk, suggest-
Yamamoto et  al, 2003). Together, these studies ing that the association is due to ethanol per se.
suggest that there may be an inverse associa- Although data are limited, a similar association
tion between isoflavones and breast cancer risk. has been observed in women of varying ethnici-
Additional research is warranted. ties (Park et al, 2014).
In a large pooled analysis of 20 prospective
Food Intake studies (Jung et al, 2015) including 37,191 breast
Intake of specific food items has also been evalu- cancer cases identified during cohort follow-​up,
ated in a number of studies. Fruit and vegetable the RR for each 10-​g increase in alcohol intake
consumption has been hypothesized to reduce (compared to nonuse) was similar (RR 1.08,
risk of breast cancer, and has been examined in 95% CI 1.07–​1.09); this association did not vary
a number of epidemiologic studies with incon- significantly by estrogen receptor status of the
clusive results. Again, this may reflect differ- tumor (per 10-​ g increase in intake:  RR 1.10,
ences by breast cancer subtype that may not have 95% CI 1.08–​1.12 for ER+ [21,624 cases], and
390

390 Textbook of Cancer Epidemiology

2
ER+ Breast Cancer ER-Breast Cancer
1.8
p-trend <0.001 p-trend <0.001
1.6
1.4
Relative Risk

1.2
1
0.8
0.6
0.4
0.2
0
5 5 5 5
–< <1 –< <1
0

0
rs

rs
0

0
3

3
≥3

≥3
>0 5– >0 5–
ke

ke
–<

–<
rin

in
15

15
dr
-d

on
on

N
N

Total Alcohol Intake (g/day)

FIGURE 16-​4  Pooled multivariable relative risks for categories of total alcohol intake by ER status.
Source: Data from Jung et al, 2015.

RR 1.06, 95% CI 1.02–​1.10 for ER-​[5,113 cases]) intake with both ER+ and ER-​breast tumors
(Fig. 16-​4). suggests that additional biologic mechanisms
A number of important details regard- are likely involved.
ing this relation remain unresolved, however.
Whether timing of the exposure in a woman’s Reproductive Factors
life is important is not known. Alcohol intake Age at Menarche
during adulthood is well confirmed to increased Age at menarche has been consistently associ-
risk (Hamajima et al, 2002), while several stud- ated with the risk of both premenopausal and
ies suggest that exposure earlier in life, partic- postmenopausal breast cancer. In a combined
ularly before first pregnancy when breast tissue analysis of 117 epidemiologic studies (includ-
may be particularly susceptible, also is independ- ing 118,964 women with breast cancer), risk was
ently associated with increased risk (Chen et al, increased by 5% for each year earlier that men-
2011; Liu et al, 2013). Relatively few studies have arche occurred (RR per year increase in men-
evaluated this aspect of the alcohol–​breast can- arche:1.05, 95% CI 1.04–​ 1.06) (Collaborative
cer association. Other important details, such as Group on Hormonal Factors in Breast Cancer,
the possible influence of drinking patterns (e.g., 2012). The inverse association was similar by
binge drinking), also still need to be resolved. ethnicity (Ambrosone et  al, 2015, Collaborative
Several biological mechanisms have been Group on Hormonal Factors in Breast Cancer,
proposed for this association, including an 2012), and was observed for both ER+ and ER-​
increase in circulating hormone levels, a direct breast tumors, but was stronger among lobular
carcinogenic effect of alcohol metabolites (e.g., cancers compared to the more common duc-
acetaldehyde, a known mutagen), increases in tal cancers (Collaborative Group on Hormonal
oxidative stress, and an antagonistic effect on Factors in Breast Cancer, 2012). Several mecha-
folate absorption and metabolism (Seitz et  al, nisms have been proposed. Menarche marks the
2012). The best-​supported mechanism to date onset of the mature hormonal milieu, that is,
is that related to circulating hormone levels—​ cyclic hormonal changes that result in ovulation,
which has been observed in both premenopausal menstruation, and cellular proliferation in the
and postmenopausal women and documented breast (Willett et  al, 2000a). The earlier the age
in several controlled human feeding studies. at menarche, the earlier a young woman starts
However, the positive association of alcohol experiencing increased steroid hormone levels.
391

Breast Cancer 391

An earlier age at menarche also has been related takes place gradually, culminating in terminally
to an earlier onset of regular ovulatory cycles. differentiated tissue (termed lobule types 3 and
In addition, although data are limited, women 4)  whose epithelial cells have longer cell cycles
with earlier menarche may have higher circulat- (lower proliferation rates) and spend more time
ing estrogen levels for a number of years after- in G1, the resting phase that allows for DNA
ward (Henderson et  al, 1996; MacMahon et  al, repair (Russo & Russo, 2012), both factors that
1982). Although age at menarche is a relatively decrease the risk of malignant transformation.
weak risk factor overall, substantial international Other mechanisms, such as long-​term alteration
variation—​for example, mean age in China of in hormones such as prolactin, also may play a
16–​17  years versus 12–​13  years in the United mechanistic role.
States—​makes it a contributor to the observed A woman’s age at the time of a first full-​term
international variation in breast cancer rates. pregnancy is also important to her risk of breast
cancer (MacMahon et  al, 1970). Independent
Age at First Birth and Parity of parity, the younger a woman is at her first
Overall, parous women have a lower risk of breast full-​term birth, the lower her subsequent risk.
cancer than nulliparous women, with about a 7% Although the magnitude of the effect has var-
decrease in risk per birth (Collaborative Group ied among studies, the reduction in risk from
on Hormonal Factors in Breast Cancer, 2002). having had a first birth by age 20 compared to
However, the relationship is complex and var- age 35 or older is about 30% (Kelsey et al, 1993).
ies with both time since childbirth and the total Mechanistically, when the first pregnancy occurs
number of births (Adami et al, 1998; Kelsey et al, at an early age, fewer cells are likely to have been
1993). For 10–​20  years after delivery, a wom- initiated and the period of protection, afforded
an’s risk of breast cancer is higher than that of by the terminal differentiation of the breast glan-
a comparable nulliparous woman (Lambe et al, dular epithelium, covers a larger fraction of the
1994). This is thought to be due to the promo- woman’s remaining lifetime.
tional effects of high pregnancy hormone levels
on a preexisting malignancy. After this time, Breastfeeding
however, a parous woman experiences a long-​ The data relating breastfeeding practices to sub-
lasting reduction in risk compared to that of a sequent breast cancer risk have not been as con-
nulliparous woman. This risk is further reduced, sistent as those for other reproductive factors.
although more modestly, with each subsequent A  reduction has been seen most consistently
birth (Rosner et  al, 1994; Trichopoulos et  al, among premenopausal women who breastfed
1983). Because the lower risk occurs when a for an extended period, but even here the mag-
woman is older and her baseline risk of breast nitude of the observed effect has varied substan-
cancer is relatively high, the long-​term reduc- tially (Lipworth et al, 2000). In a pooled analysis
tion in risk substantially outweighs the short-​ of 47 epidemiologic studies in 30 countries, life-
term postpregnancy increase in risk. Adding time duration of breastfeeding was associated
to this complexity, these associations vary by with a significant reduction in breast cancer risk
characteristics of the breast tumor. The inverse among parous women. Although variation was
association with parity is seen only among the observed across studies in terms of the magni-
more common ER+ breast cancers, while a sig- tude of the risk reduction, no significant varia-
nificant positive association with parity has been tion was noted by menopausal status. The RR of
observed for ER-​tumors (Palmer et  al, 2014; breast cancer decreased by 4.3% (95% CI 2.9–​5.8;
Yang et al, 2011). p < 0.0001) for every 12 months of breastfeed-
The biological mechanism behind this over- ing (Collaborative Group on Hormonal Factors
all long-​term reduction in risk associated with in Breast, 2002). A  number of factors may be
pregnancy has been studied extensively (Russo & responsible for these inconsistent results. For
Russo, 2012), and continues to evolve. The ductal example, the pattern of breastfeeding has gen-
system of the breast undergoes profound changes erally not been taken into account in these stud-
from birth through adulthood. After menarche ies, and whether the breastfeeding was exclusive
but prior to a first pregnancy, the breast con- or supplemental may influence the ultimate
tains relatively undifferentiated ducts and asso- reduction in risk. The relationship between lac-
ciated alveolar buds (termed lobule types 1 and tation and breast cancer risk also may vary by
2). Differentiation of the glandular epithelial cells tumor subtype. In several studies, the inverse
392

392 Textbook of Cancer Epidemiology

associations appeared stronger among ER-​, tri- the cessation of ovarian function and the result-
ple negative (ER-​/​PR-​/​HER2-​), or basal-​like ant dramatic reduction in circulating steroid
breast tumors (Anderson et  al, 2014; Tamimi hormone levels. An artificial menopause through
et  al, 2012; Work et  al, 2014). Importantly, bilateral oophorectomy confers more protection
breastfeeding appeared to negate the increased than a natural menopause at the same age because
risk of ER-​tumors observed in parous (vs. nul- of the abrupt decrease in both ovarian estrogens
liparous) women (Palmer et  al, 2014; Work and androgens. Although estrogen levels decline
et al, 2014). within 1–​2 years of menopause, androgens ordi-
Several mechanisms may be responsible for narily continue to be produced for a number of
any protective effect of lactation (Faupel-​Badger years. Women with a bilateral oophorectomy
et al, 2013). Breastfeeding may result in further before age 45 have about half the risk of breast
terminal differentiation of the breast epithelium, cancer of those with a natural menopause at age
thus making it more resistant to carcinogenic 55 or older (Trichopoulos et al, 1972).
change. Additionally, breastfeeding delays the The consistent associations observed with age
postpregnancy reestablishment of the ovulatory at menarche and menopause, as well as parity
menstrual cycle and hence may reduce the risk. and age at first birth are all likely causal associa-
tions with breast cancer; data are less clear for the
Spontaneous and Induced Abortions other reproductive factors.
Based on data from animal studies, it has been
hypothesized that the incomplete differentia- Hormones and Hormone
tion of breast cells during the first trimester of a Receptors
pregnancy could result in increased susceptibil- Endogenous Hormones
ity of the tissue to malignant transformation. In a Several lines of evidence have long suggested
meta-​analysis that included data from 28 studies, that sex hormones play a central role in the eti-
a positive association was indeed noted between ology of breast cancer. As just reviewed, several
induced abortion and breast cancer (Brind et al, reproductive factors are consistently associated
1996). However, most of these investigations with breast cancer risk. In addition, after men-
were case-​control studies and thus were poten- opause, adipose tissue is the major source of
tially biased by differential recall of past induced estrogen, and obese postmenopausal women
abortions. In fact, in a subsequent pooled analy- have both higher levels of endogenous estrogen
sis where 13 studies with prospectively collected and a higher risk of breast cancer (Harris et al,
information were analyzed (Beral et  al, 2004), 2014; Huang et al, 1997). Further, estrogens and
no association was observed between induced progesterone promote mammary tumors in ani-
abortion and breast cancer risk (RR 0.98, 95% mals (Briand, 1983). Also, tamoxifen, which
CI 0.92–​1.04), and when results from studies acts as an antiestrogen in the breast (by block-
using prospectively collected abortion informa- ing binding of estrogen to the estrogen recep-
tion were compared to results from studies using tor), decreases the incidence of breast cancer
retrospectively collected exposure data, signifi- in high-​risk women (Cuzick et  al, 2013; Fisher
cant heterogeneity was observed. In addition, no et al, 1998).
association was observed between spontaneous Given that diverse lines of evidence support
abortion and risk. Hence overall evidence sug- a role for endogenous hormones in breast cancer
gests that abortion does not alter subsequent risk etiology, multiple studies have been conducted to
of breast cancer. evaluate plasma and urinary hormone levels in
relation to subsequent breast cancer risk. A sum-
Age at Menopause mary, which relies predominantly on the larger
The positive relationship between age at meno- prospective studies where possible, is provided in
pause and subsequent breast cancer risk is also what follows.
well established. Based on a pooled analysis of
117 epidemiologic studies, the risk of breast can- Estrogens
cer increases by about 3% per year that meno- Estradiol, considered the most biologically active
pause is delayed (RR 1.029, 95% CI 1.025–​1.032) endogenous estrogen, circulates in blood either
(Collaborative Group on Hormonal Factors in unbound (“free”) or bound to sex hormone-​
Breast Cancer, 2012). The reduction in risk asso- binding globulin or albumin. Free or bioavailable
ciated with an earlier menopause is likely due to (free plus albumin-​bound) estradiol is thought to
39

Breast Cancer 393

be readily available to breast tissue and thus may positive association between circulating estra-
be more strongly related to risk than total estra- diol and risk was observed (highest vs. lowest
diol. Postmenopausally, estrone is the source of quintile comparison RR 1.41, 95% CI 1.02–​1.95;
most circulating estradiol, and estrone sulfate is p-​trend = 0.004); this association was nonsignifi-
the most abundant circulating estrogen (Roberts cantly higher when hormone levels were meas-
et  al, 1980). Both normal and malignant breast ured in the follicular versus luteal phase of the
cells have sulfatase and 17-​beta-​dehydrogenase menstrual cycle.
activity (Pasqualini et  al, 1996). Thus, estrone
and estrone sulfate could serve as ready sources Estrogen Metabolites
of intracellular estradiol. A woman’s pattern of estrogen metabolism
In 2002, a pooled analysis was published con- also has been hypothesized to influence her
sisting of all prospective studies of endogenous breast cancer risk. Estradiol and estrone can be
estrogens and androgens in postmenopausal metabolized through several pathways, includ-
women that had been available at that time (Key ing the 16-​a and 2 (and 4-​)-​hydroxy pathways
et  al, 2002). Data were from nine prospective (Fig. 16-​5) (Yager & Liehr, 1996). Products of
studies with a total of 663 breast cancer cases and these pathways have markedly different biolog-
1,765 healthy controls; none of the women were ical properties, and opposing hypotheses have
using exogenous hormones at blood collection. been proposed concerning their influence on risk
The risk of breast cancer increased with increas- (Yager & Liehr, 1996). This has been a particu-
ing estrogen levels. For example, the RRs (95% larly challenging hypothesis to address because
CI) for increasing quintiles of estradiol level, all of complexities in estrogen metabolite measure-
relative to the lowest quintile, were 1.4 (1.0–​2.0), ment (Ziegler et  al, 2015). Nonetheless, about
1.2 (0.9–​1.7), 1.8 (1.3–​2.4), and 2.0 (1.5–​2.7). a dozen studies have been conducted to date,
Estrone, estrone sulfate, and free estradiol were and although not entirely consistent, overall,
similarly related to risk. Subsequent to the pooled enhanced 2-​hydroxylation of estrogens has been
analysis, a number of additional prospective associated with lower breast cancer risk in both
studies have been published and essentially all postmenopausal and premenopausal women
have supported these findings (Kaaks et al, 2005; (Eliassen et al, 2012b; Ziegler et al, 2015).
Manjer et  al, 2003; Missmer et  al, 2004; Zhang
et al, 2013). Estrogen Receptor
Several studies have assessed whether the Given the role of ER in estrogen response, and
positive associations of circulating estrogens and the consistently observed associations between
risk of breast cancer varied by tumor estrogen plasma estrogen levels and breast cancer risk,
receptor (ER) and/​or progesterone receptor (PR) the expression of ER in normal breast tissue is
status. Stronger associations in ER+ (or ER+/​ of interest. The proportion of ER+ breast tumors
PR+) have been reported in two of the studies has been reported to be higher among Caucasian
(Gunter et al, 2009; Zhang et al, 2013) but not in than Asian patients (Adami et  al, 1998)  and
a third (James et al, 2011). One would hypothe- Caucasians were found to have greater expres-
size associations to be stronger among ER+ (or sion of ER-​alpha in normal breast epithelial tissue
ER+/​PR+) tumors, given tamoxifen use prevents than Japanese women of a similar age (Lawson
only ER+ tumors (Cuzick et al, 2013), and, in epi- et al, 1999). However, whether ER expression in
demiologic studies, postmenopausal BMI is most normal epithelium is related to subsequent breast
robustly associated with ER+ disease. Reasons cancer risk is unclear (Khan et  al, 1998; Lagiou
for this lack of consistency across studies are et al, 2009).
not clear.
Data on premenopausal estrogen levels and Androgens
breast cancer risk are more limited, in large Androgens have been hypothesized to increase
part because of the complexities related to sam- the breast cancer risk either directly, by increas-
pling during the menstrual cycle. In 2003, a ing the growth and proliferation of breast cancer
pooled analysis was conducted among seven cells, or indirectly, by their conversion to estro-
prospective studies, including 767 cases who gen (Bernstein & Ross, 1993; Liao & Dickson,
were premenopausal at diagnosis and 1,699 2002). Dehydroepiandrosterone (DHEA) admin-
controls (Endogenous Hormones Breast Cancer istered to rodents can decrease tumor formation.
Collaborative Group, 2013). A significant linear In humans, DHEA may act like an antiestrogen
394

394 Textbook of Cancer Epidemiology

1.4

1.2

1
Relative Risk

0.8

0.6

0.4

0.2

0
1 1.5–2 2.5–3 3.5–4 ≥ 4.5
Average Adolescent Body Fatness
(Ages 10–20 Years)

ER+/PR+ ER-/PR-

FIGURE 16-​5  Relative risk of breast cancer according to average adolescent body fatness by breast cancer subtype.
Source: Data from Baer et al, 2010.

premenopausally but an estrogen postmeno- independent effect of circulating androgens on

pausally in stimulating cell growth. This is in part breast cancer risk. However, interpretation of
because of the estrogenic effect of its metabolite, these data is complicated because of possible dif-
5-​androstene-​3b,17b-​diol, which can bind to the ferences between estradiol and the androgens
ER (Liao & Dickson, 2002; Seymour-​Munn & in terms of assay precision, hormone stability
Adams, 1983). within woman over time, and intracellular con-
In postmenopausal women, the best sum- version of androgens to estrogens, which can-
mary of evidence on circulating androgens and not be accounted for in epidemiologic analyses.
breast cancer risk is from the pooled analysis of These results were confirmed in several subse-
nine prospective studies described above (Key quent studies (Baglietto et  al, 2010; James et  al,
et  al, 2002), along with published reports from 2011; Zhang et al, 2013).
the European EPIC study (Kaaks et al, 2005) and Among premenopausal women, data are
US Nurses’ Health Study (Zhang et al, 2013). In much more limited. In a pooled analysis, sta-
the pooled analysis, testosterone was positively tistically significant positive associations with
associated with breast cancer risk: the RRs (95% breast cancer risk were seen for testosterone,
CI) for increasing quintile category (all relative androstenedione, and DHEA, the RRs compar-
to the lowest quintile of levels) were 1.3 (1.0–​ ing extreme quintile categories being 1.3, 1.7, and
1.9), 1.6 (1.2–​2.2), 1.6 (1.1–​2.2) and 2.2 (1.6–​ 1.5, respectively (Endogenous Hormones Breast
3.1). Findings were generally similar for several Cancer Collaborative Group, 2013).
other androgens measured. In EPIC and NHS,
similar positive associations also were observed Progesterone
for each of the androgens measured. In each of Progesterone exerts powerful influences on
these analyses, when estradiol was added to the breast physiology and can influence tumor
statistical models, relative risks for the androgens development in rodents (Kelsey, 1979). Based
were only modestly attenuated, suggesting some largely on indirect evidence, progesterone has
395

Breast Cancer 395

been hypothesized both to decrease breast can- monkeys treated with growth hormone or IGF-​
cer risk by opposing estrogenic stimulation of 1 show histologic evidence of mammary gland
the breast (Kelsey, 1979)  and to increase this hyperplasia. In addition, positive associations
risk because breast mitotic rates are highest in have been observed between breast cancer and
the luteal (high-​progesterone) phase of the men- birthweight as well as height, which are both pos-
strual cycle (Bernstein & Ross, 1993). In the itively correlated with IGF-​1 levels (Pollak, 1998;
pooled analysis of seven prospective studies in Renehan et  al, 2004). Several initial prospective
premenopausal women (Endogenous Hormones studies reported a significant positive association,
Breast Cancer Collaborative Group, 2013), no particularly in premenopausal women (Renehan
association was observed, with a relative risk of et al, 2004); this was not seen consistently in sub-
0.98 (95% CI 0.68–​1.40) comparing the top to sequent studies. To draw a firmer conclusion
bottom quintile categories. In postmenopausal and evaluate possible reasons for these inconsis-
women, only a single prospective study has tencies, a pooled analysis was conducted using
been conducted and no association was found data from 17 prospective studies, and including
(Missmer et al, 2004). 4,790 cases and 4,928 matched controls. Overall
a modest but significant positive association was
Prolactin observed (top vs. bottom quintile category RR
Prolactin receptors have been found on more 1.28, 95% CI 1.14–​1.44; p-​trend < 0.001) that
than 50% of breast tumors, and prolactin can did not vary significantly by menopausal sta-
increase the growth of both normal and malig- tus and was stronger for ER+ tumors (compa-
nant breast cells in vitro (Clevenger et al, 2003). rable RR 1.38, 95% CI 1.14–​1.68) (Endogenous
Prolactin administration also is well documented Hormones Breast Cancer Collaborative Group,
to increase mammary tumor development rates 2010). In this same pooled analysis, high levels
in mice. Because prolactin is influenced by both of IGF binding protein-​3 (IGFBP-​3) were not
physical and emotional stress (Herman et  al, associated with breast cancer risk. Although
1981; Yen & Jaffe, 1991), levels in women with there was considerable variation in absolute
breast cancer may not reflect their predisease IGF-​1 levels between studies, relatively little het-
levels. Thus, evaluation of this association in pro- erogeneity of results between individual stud-
spective studies is particularly important. ies was observed. Further, in a study conducted
Only two large prospective studies have been among women with benign breast disease, 75
conducted to date. In the Nurses’ Health Studies, of whom went on to be diagnosed with breast
2,468 cases and 4,021 controls were included, cancer, cytoplasmic IGF-​1 receptor staining in
and associations were assessed both by time normal terminal duct lobular units (TDLU) was
from hormone measurement to disease diagno- associated with a twofold higher risk of breast
sis, as well as by menopausal status and tumor cancer; women with high cytoplasmic to mem-
ER status at diagnosis (Tworoger et  al, 2013). branous staining (perhaps reflecting active IGF-​1
Among premenopausal women, no association pathway signaling) had a 15-​fold higher risk of
was observed. Among postmenopausal women, breast cancer (Tamimi et  al, 2011). Higher lev-
a statistically significant positive association was els of IGF-​1 among incident breast cancer cases
seen in the first 10  years after blood collection (n = 466) compared to women with benign breast
(top to bottom quartile RR = 1.4), particularly for disease (BBD) (n = 704) were also reported in a
ER+ tumors (comparable RR = 1.5). In the large case-​ control study; the differences were more
EPIC cohort with 2,271 cases, generally similar pronounced among postmenopausal women
associations were observed:  no significant asso- and when breast cancer patients were contrasted
ciation was observed in premenopausal women, to women with proliferative BBD (Lagiou et  al,
and a significant positive association was seen in 2013). Cumulatively these data support a role for
postmenopausal women (Tikk et al, 2014). the IGF axis in breast carcinogenesis, although
the magnitude of the association based on stud-
Insulin-​Like Growth Factors and Insulin ies of circulating IGF-​1 levels is modest.
Insulin-​like growth factor 1 (IGF-​1) is a protein The insulin pathway is important in breast
hormone with a structural homology to insulin. development (Rosenfeld, 2003), but overex-
The growth hormone–​IGF-​1 axis can stimulate pression of insulin receptor (IR) in breast epi-
proliferation of both breast cancer and normal thelium can induce malignant transformation
breast epithelial cells (Pollak, 1998). Rhesus (Frittitta et  al, 1995); in breast cancer cell lines,
396

396 Textbook of Cancer Epidemiology

insulin promotes growth and decreases apoptosis among either postmenopausal women or women
(Belardi et al, 2013; Pollak, 2008). Multiple pro- over the age of 45 years. In contrast however, cur-
spective studies have evaluated blood insulin (or rent and recent users of oral contraceptives had
c-​peptide, a marker of insulin secretion) levels an increased risk of breast cancer (for current vs.
and risk of breast cancer. No overall association never-​users, RR  =  1.24) (Collaborative Group
was seen in a meta-​analysis including 11 prospec- on Hormonal Factors in Breast Cancer, 1996).
tive studies (Autier et  al, 2013), although non- This increased risk disappeared within 10  years
significant positive associations were observed of stopping oral contraceptive use (RR by years
in postmenopausal women (RRs 1.6 and 1.3 since stopping use vs. never use: 1–​4 years, 1.16;
comparing highest vs. lowest levels of insulin 5–​9  years, 1.07; 10–​14  years, 0.98; more than
and c-​peptide, respectively). In two subsequent 15 years, 1.03). When the investigators evaluated
prospective studies (with 796 and 302 cases), a both time since last use and duration of use, they
significant positive association was observed observed a modestly increased risk only among
(Ahern et  al, 2013; Gaudet et  al, 2013b). In the current and recent users, and no independent
two studies to address these issues the association effect of long duration of use on the risk of breast
was independent of endogenous estradiol levels cancer even among very young women. Thus, the
and BMI, and was similar, or somewhat stronger, increased risk of breast cancer observed among
for ER-​versus ER+ tumors Overall, the weight young, long-​term users of oral contraceptives in
of evidence from basic and epidemiologic stud- prior individual studies appears due primarily to
ies supports a role for insulin resistance in post- recency of use rather than to duration. These data
menopausal breast cancer risk. suggest that oral contraceptives may act as late-​
stage promoters. Importantly, current and recent
Oral Contraceptives users, the women who appear to have a mod-
Since oral contraceptives were first introduced est increase in risk, are generally young (under
in the 1960s, they have been used by millions of 45  years of age) and thus have a low absolute
women (Committee on the Relationship Between risk of breast cancer. Hence, a modest increase
Oral Contraceptives and Breast Cancer Institute in their risk will result in few additional cases
of Medicine, 1991; Mosher & Jones, 2010). Most of breast cancer. Nevertheless, this apparently
combined oral contraceptives contain ethinyl increased risk among current and recent users
estradiol (or mestranol, which is metabolized to should be considered in deciding whether to use
ethinyl estradiol) and a progestin. The estrogen oral contraceptives.
dose in oral contraceptives has ranged from at Because any influence of oral contraceptives
least 100 micrograms in 1960 to 20–​30 micro- on the breast has been hypothesized to be great-
grams, the doses most commonly used today; est prior to the cellular differentiation that occurs
during this same time period, at least nine differ- with a full-​term pregnancy (Russo et al, 1990), a
ent progestins have been used (Bassuk & Manson, number of investigators have evaluated the effect
2015). Patterns of use also have changed consid- of oral contraceptive use prior to a first full-​term
erably over time, with both increasing durations pregnancy. In the pooled analysis (Collaborative
of use and a trend toward earlier age at first use. Group on Hormonal Factors in Breast Cancer,
Over 75 epidemiologic studies have evaluated the 1996), a significant trend of increasing risk
relationship between oral contraceptive use and with first use before age 20 years was observed.
breast cancer risk. Among women ages 30–​34 years, the RR associ-
Most studies have observed no significant ated with recent oral contraceptive use was 1.54
increase in breast cancer risk even with long if use began before age 20  years and 1.13 if use
durations of use. Individual data from 54 epi- began at age 20 years or older.
demiologic studies were collected and analyzed Whether the increased risk of breast cancer
centrally (Collaborative Group on Hormonal with recent oral contraceptive use varies accord-
Factors in Breast Cancer, 1996). In this large ing to the type or dose of either the estrogen or
pooled analysis, in which data from 53,297 progestin is of considerable interest, yet data
women with and 100,239 women without breast addressing this issue remain limited and firm
cancer were evaluated, no overall relationship conclusions cannot be drawn. The Women’s
was observed between duration of use and risk CARE Study, a large population-​ based case-​
of breast cancer. Similar findings were gener- control study with 4,575 breast cancer cases
ally observed when long-​term use was evaluated and 4,682 controls, examined the risk of breast
397

Breast Cancer 397

cancer associated with oral contraceptives by long-​term use of Norplant, a long-​acting contra-
formulation (Marchbanks et  al, 2002). In this ceptive that is implanted subdermally, as it was
study, there was no increased risk of breast can- introduced in the United States only in 1990.
cer among current users (RR 1.0, 95% CI 0.8–​1.0) Further epidemiologic research is needed for
or former users (RR 0.9, 95% CI 0.8–​1.0) over- each of these drugs.
all. In addition, the risk of breast cancer did not
appear to vary by dose or type of progestin used Postmenopausal Hormone Use
(Marchbanks et al, 2002). In the only prospective Postmenopausal estrogens have been used for
cohort to examine newer generation oral contra- over half a century. By the mid-​1970s, almost
ceptives as well as specific formulations in detail, 30 million prescriptions were being filled annu-
current use of any oral contraceptives was associ- ally in the United States (Kennedy et  al, 1985).
ated with a modest increase in risk (RR 1.33, 95% A challenge in studying the relationship between
CI 1.03–​1.73), similar to associations seen in the postmenopausal hormones and breast cancer
large collaborative analysis. However, this asso- is the substantial variation in formulations and
ciation was largely accounted for by use of a sin- patterns of use that has occurred over time. By
gle type of preparation, triphasic contraceptives the time sufficient use of one type of hormone
with levonorgestrel (comparable RR 3.05, 95% CI has occurred to allow a detailed epidemiologic
2.00–​4.66) (Hunter et al, 2010). evaluation, new formulations are already being
Ever use of oral contraceptives also has been introduced.
assessed by hormone receptor status of the breast Following publication of at least six meta-​
tumor and, across eight case-​control studies, no analyses during the 1990s, relationships between
consistent differences were observed (Althuis postmenopausal hormone use and breast can-
et al, 2004). In the Black Women’s Health Study cer risk were evaluated in considerable detail in
cohort, ever use of oral contraceptives was more the pooled analysis that combined the results of
strongly associated with risk of ER-​/P ​ R-​breast 51 epidemiologic studies (Collaborative Group
cancers (ever versus never use RR 1.7, 95% CI on Hormonal Factors in Breast Cancer, 1997).
1.2–​2.3) than ER+/​PR+ breast cancers (compa- Importantly, in these analyses, women with an
rable RR 1.1, 95% CI 0.9–​1.4) (Rosenberg et al, uncertain age at menopause were excluded (e.g.,
2010). In case-​case analyses, compared to lumi- women with simple hysterectomies), as inade-
nal A  breast cancers, luminal B cases were less quate accounting for age at menopause in the
likely to ever use oral contraceptives in one study analysis can lead to substantial attenuation of the
(Kwan et  al, 2009), but not in a second study observed relationships between postmenopausal
(Millikan et al, 2008); although statistical power hormone use and breast cancer risk.
was limited, no differences were observed with The investigators observed a statistically sig-
triple negative, basal-​ like, or HER-​ 2-​
enriched nificant association between current or recent
breast cancers. More data are needed to deter- use of postmenopausal hormones and the risk
mine whether oral contraceptives differentially of breast cancer; the positive association was
influence breast cancer molecular subtypes. strongest among those with the longest dura-
Progestin-​only contraceptives include tion of use. For example, among women who
progestin-only pills (“mini-​ pill”), depot- had used postmenopausal hormones within the
medroxyprogesterone (DMPA, an injectable previous 5  years compared to never users, the
contraceptive), and implantable levonorgestrel RRs for duration of use were 1.08 for 1–​4 years
(Norplant); few epidemiologic studies have eval- of use, 1.31 for 5–​9  years, 1.24 for 10–​14  years,
uated their association with breast cancer risk. and 1.56 for 15 years or more of use. No signifi-
To date, longer-​term users of the progestin-​only cant increase in breast cancer risk was noted for
pill have been observed to have either a similar women who had stopped using postmenopausal
or lower risk of breast cancer than never users hormones 5 or more years in the past, regardless
(Stanford & Thomas, 1993). Recent use of DMPA of their duration of use.
was associated with an increased risk of breast Subsequent to this meta-​ analysis, findings
cancer (RR~1.5–​2.0 for most vs. least exposed) from the Nurses’ Health Study and the Million
in several but not all studies to date (Li et  al, Women Study support the findings that short-​
2012; Strom et  al, 2004; WHO Collaborative term use of unopposed estrogen is not associ-
Study of Neoplasia and Steroid Contraceptives, ated with an increased risk of breast cancer, while
1991). Only very sparse data are available on the longer-​term use is (Beral & Million Women
398

398 Textbook of Cancer Epidemiology

Study, 2003; Chen et al, 2006). In addition, Chen is biologically plausible since lean women tend
et  al reported that 15  years or more of use was to have the lowest estrogen levels. Hence, tak-
significantly associated specifically with ER+/​ ing postmenopausal hormones will result in a
PR+ breast cancers (Chen et al, 2006). greater proportional increase in estrogen levels
One concern has been whether the increased compared to that in heavier women, who tend to
risk observed is an artifact of increased surveil- have higher estrogen levels.
lance for breast cancer among women taking The Women’s Health Initiative (WHI) ran-
hormones. Consistent with that possibility, a domized controlled trial in which 10,739 women
higher RR has been associated with in situ dis- with hysterectomy were randomized to estrogen-​
ease than with invasive disease (Colditz et  al, only therapy reported a nonsignificant inverse
1995; Schairer et  al, 1994). However, in both of association between therapy and breast cancer
these studies, a significant positive, although after an average of 7 years of follow-​up (RR 0.77,
weaker, association was noted when only inva- 95% CI 0.59–​1.01) (Anderson et  al, 2004). At
sive breast cancer cases were considered. Also, enrollment, the average age of WHI participants
in the Nurses’ Health Study, when differences in was 64 years and approximately 45% of the pop-
mammographic screening rates between expo- ulation had a body mass index ≥ 30 kg/​m2. Given
sure groups were carefully accounted for, some that these women were randomized to hormone
attenuation of the association was observed, use many years after menopause and that a large
although positive associations remained (Cook proportion are considered obese, the difference
et al, 2009). Finally, an elevated death rate from between results from WHI and observational
breast cancer was observed among current users studies may reflect differences in study popula-
of 10 or more years’ duration at the time of diag- tions. Indeed, laboratory studies have shown that
nosis. These analyses indicate that the associa- breast cancer cells that are estrogen deprived for
tion cannot be explained simply as an artifact of extended periods, but then exposed to estrogen,
screening. are more likely to undergo apoptosis, poten-
Limited data are available regarding the effects tially leading to lower risk of a clinically diag-
of the dose or type of estrogen on breast cancer nosed tumor (Jordan, 2015). Further, in the large
risk. Again, the best data come from the pooled Million Women’s Study, current estrogen use was
analysis (Collaborative Group on Hormonal associated with a significant increase in risk when
Factors in Breast Cancer, 1997). No significant begun near the time of menopause (RR 1.4, 95%
differences in relative risk were observed accord- CI, 1.4–​1.5), but not associated with risk when
ing to either the type of estrogen used (conjugated begun 5 or more years after menopause (RR 1.0,
estrogen vs. other) or the estrogen dose (<0.625 95% CI  =  0.9–​1.2) (Beral et  al, 2011). Because
mg vs. ≥1.25 mg), although some modest differ- postmenopausal hormones are typically begun
ences in estimates suggested that further evalua- near the time of menopause, few observational
tion is warranted. Although the effect of estrogen studies have been able to address in detail the use
use on breast cancer risk could be reasonably pattern evaluated in the randomized WHI.
hypothesized to vary by mode of estrogen deliv-
ery, for example, by avoiding the first-​pass effect Estrogen Plus Progestin Use
in the liver (patch estrogen does not increase sex The addition of a progestin to estrogen regimens
hormone-​binding globulin to the extent that oral has become increasingly common, as it mini-
preparations do), insufficient data are available to mizes or eliminates the increased risk of endo-
evaluate these potential differences. metrial hyperplasia and cancer associated with
The risk associated with postmenopausal hor- using unopposed estrogens. In the United States,
mone use was assessed in a number of specific by the mid-​ 1980s, almost 30% of postmeno-
subgroups in the pooled analysis (Collaborative pausal hormone prescriptions included a pre-
Group on Hormonal Factors in Breast Cancer, scription for progestin (Hemminki et  al, 1988).
1997). Risk did not appear to vary according to There are data suggesting that the addition of a
reproductive history, alcohol or smoking history, progestin may increase the risk beyond that asso-
or by family history of breast cancer. However, ciated with estrogen alone (Bergkvist et al, 1989;
the RRs associated with 5 or more years of post- Colditz & Rosner, 2000). In the PEPI double-​
menopausal hormone use were highest among blind, randomized trial of placebo, estrogens, and
the leanest women; this interaction has been three estrogen-​progestin regimens, the estrogen-​
consistently observed (Huang et  al, 1997)  and progestin combinations were associated with a
39

Breast Cancer 399

much larger increase in mammographic density currently using estrogen plus testosterone thera-
than the estrogen-​only formulation (Greendale pies were at a significant increased risk of inva-
et al, 1999). In the pooled analysis (Collaborative sive breast cancer (RR 2.48, 95% CI 1.53–​4.04)
Group on Hormonal Factors in Breast Cancer, (Tamimi et al, 2006).
1997), data on postmenopausal hormone for- It has now become well accepted that cur-
mulation were available from only 39% of the rent use of combined E + P therapies increase the
women, and only 12% of these reported use of risk of breast cancer and that this increase in risk
estrogen plus a progestin. The RR associated with is greater than that observed for estrogen-​only
5 or more years of recent use, relative to never therapies. The difference between the breast and
use, was 1.53. The comparable relative risk for the endometrium in this regard is striking. Given
estrogen use alone was 1.34. the substantial evidence implicating combined E
Since the time of the pooled analysis, multi- + P therapy in breast cancer and other postmeno-
ple large observational studies—​including sev- pausal hormone therapies, it has been recom-
eral cohort studies (Beral & Million Women mended that women and their physicians should
Study, 2003; Colditz & Rosner, 2000; Saxena reconsider use and, more specifically, long-​term
et  al, 2010; Schairer et  al, 2000)—​investigated use of these therapies (Willett et al, 2000a).
the influence of combined estrogen-​ progestin
(E + P) treatment on breast cancer risk. Overall, Hormonal Chemoprevention
results have been quite consistent, showing sig- A number of agents are approved for use, being
nificant positive associations between current E developed or undergoing testing as breast can-
+ P use, that increase with increasing duration cer chemopreventives. One of the most widely
and that are stronger than associations observed studied and a US Food and Drug Administration
with E use only. For example, the Million Women (FDA)–​ approved breast cancer chemopreven-
Study, a prospective cohort of over 1 million UK tive, tamoxifen, is a member of an increasing
women, was designed to specifically examine family of agents called selective estrogen receptor
the effects of different postmenopausal hormone modulators (SERMs), which have tissue-​specific
regimens. Current users of E + P preparations estrogen agonist/​antagonist effects (MacGregor
were at increased risk of breast cancer compared & Jordan, 1998). To date, four randomized pre-
to never users (RR 2.00, 95% CI 1.88–​2.12) and vention trials have been conducted to evaluate
risk increased with increasing duration (Beral & tamoxifen, and cumulatively a 33% reduction in
Million Women Study, 2003). The comparable breast cancer incidence overall was observed over
RR for use of E-​only formulations was 1.30, (95% a median follow-​up of 5.5  years; the decreased
CI 1.21–​1.40). The data are more limited and less risk occurred only among ER+ tumors, where
clear as to whether associations vary by proges- a 44% reduction was seen (Cuzick et  al, 2013).
tin type or regimen (Bakken et al, 2011; Saxena Trials of a related SERM, raloxifene, found a sim-
et al, 2010), although several of the studies have ilar reduction in breast cancer risk (Cuzick et al,
suggested stronger associations for continuous 2013). However, for all SERMs, there was also
combined, versus sequential progestin use. an increase in the number of thromboembolic
In 2002, the Women’s Health Initiative trial, events in the treatment arm, and for tamoxi-
in which 16,608 postmenopausal women were fen, the risk of endometrial cancer also was ele-
randomized to either a combined E + P therapy vated. The breast cancer prevention trials of these
or placebo, demonstrated that this combination SERMs have clearly demonstrated their effective-
therapy increased the risk of invasive breast can- ness in reducing the incidence of breast cancer,
cer and was stopped early. Over 7 years of follow-​ and specifically ER+ breast cancer, however con-
up there was a 24% increased risk of breast cern over side effects remain. In fact, uptake of
cancer (RR 1.24, 95% CI 1.01–​1.54) (Chlebowski tamoxifen and raloxifene for use in breast cancer
et  al, 2003). Since the publication of the WHI prevention has been very limited, likely due, at
results, the prescription patterns for E + P thera- least in part, to concerns about toxicity and the
pies have declined (Majumdar et  al, 2004). In overall risk-​benefit ratio (Ravdin, 2010; Waters
contrast, there is evidence that use of estrogen et al, 2012).
plus testosterone therapies are increasing (Davis, Aromatase inhibitors, such as anastrazole,
1999; Tamimi et  al, 2006). Consistent with the letrozole, and exemestane, suppress estro-
elevation in risk for endogenous testosterone lev- gen production by blocking the aromatase
els, one prospective study reported that women enzyme, which converts androgens to estrogens
40

400 Textbook of Cancer Epidemiology

(Gradishar, 2006). Several trials have demon- using a height-​associated genetic risk score in a
strated that aromatase inhibitors have better effi- Mendelian randomization study design (Zhang
cacy than tamoxifen in treating ER+ breast cancer et  al, 2015). Height may indirectly reflect early
(Dowsett et  al, 2010). The International Breast energy intake, such that (for a given genetic
Cancer Intervention Study (IBIS) II and MAP.3 makeup) short women may have been energy-​
(National Cancer Institute, Canada) are two ran- restricted in childhood. It is well established
domized control trials that compared aromatase that energy restriction in rodents substantially
inhibitors to placebo among postmenopausal reduces mammary tumor rates. Adult height and
women at high risk for breast cancer. In IBIS-​II mammary gland mass are likely to be positively
(Cuzick et  al, 2014), a 53% reduction in breast associated, albeit weakly, since both reflect, to
cancer risk was observed (58% reduction in ER+ a certain extent, overall growth. Of note, IGF-​1
breast tumors; RR 0.42, 95% CI 0.32–​0.68), while levels in childhood are strongly correlated with
in MAP.3 (Goss et al, 2011), a 65% reduction in height (Juul et al, 1994; Juul et al, 1995).
breast cancer was seen (73% decrease in ER+
tumors; RR 0.27, 95% CI 0.12–​0.60). Other drugs, Body Mass Index
such as somatostatin analogues, which decrease The relationship between body mass index (BMI)
levels of IGF-​1, or metformin, a commonly used and breast cancer risk is complex, and the asso-
diabetes drug that decreases hyperglycemia and ciation varies according to when in life body
insulin resistance, could be evaluated in the size is assessed, for example, in childhood or
future (Pollak, 1998; Pollak, 2012). Thus, chemo- before versus after menopause. This complexity
prevention for breast cancer is now becoming an is not unexpected, given the large differences in
option; however, the risk-​benefit and cost-​benefitthe hormonal milieu at these life stages as well
ratios, including optimal criteria for use of these as the known impact of adipose tissue on vari-
agents, have yet to be established. ous hormones. Most data exist for BMI during
Several breast cancer risk prediction models adulthood. In premenopausal women, an inverse
(i.e., statistical models that provide information association between BMI and risk is seen. In a
on an individual woman’s 5-​or 10-​year absolute combined analysis of seven cohorts, the RR was
risk of developing breast cancer) were devel- 0.89, (95% CI 0.81–​0.97) per 4-​kg/​m2 increment
oped to estimate a woman’s risk of breast cancer in BMI (van den Brandt et  al, 2000). Several
based on her family history and to set the entry mechanisms may explain this consistently
criteria for randomized chemoprevention tri- observed association (Friedenreich, 2001). For
als (Anothaisintawee et al, 2012; Freedman et al, example, obese women are less likely to ovulate,
2005; Meads et  al, 2012). More recently, addi- which reduces their lifetime number of ovula-
tional risk factors, including biomarkers (e.g., tions and alters their circulating hormone levels
endogenous hormones; Tworoger et  al, 2014), and thus may decrease their breast cancer risk.
genetic risk alleles (Mavaddat et  al, 2015)  and Body mass index is positively related to
mammographic density (Vachon et al, 2015) have postmenopausal breast cancer, although the
been added to improve model discrimination. magnitude of the relationship overall has fre-
These models may prove useful to women and quently been modest (Friedenreich, 2001; van
their healthcare providers in deciding whether or den Brandt et al, 2000). The mechanism behind
not to use chemoprevention. this increase in risk is likely related largely to
the increase in circulating hormone levels. After
Anthropometric Measures menopause, plasma estrogens are derived prima-
Height rily from adipose tissue, such that a linear rela-
Height has been consistently associated with an tionship exists between BMI and estrogen levels.
increase in breast cancer risk, although the mag- In addition, overweight women have low levels of
nitude of the effect is modest. In a meta-​analysis sex hormone-​binding globulin and thus a higher
of 159 prospective studies (Zhang et  al, 2015), biologically available estrogen level; again, this
the RR of breast cancer was 1.17 (95% CI 1.15–​ may increase their risk of breast cancer. Further
1.19) per 10-​cm increase in height. Comparable supporting the role of estrogens in this associa-
RRs were noted for both premenopausal and tion is the stronger BMI–​breast cancer associa-
postmenopausal women, and the association tion consistently observed in postmenopausal
was seen primarily among ER+ breast can- women who have not used postmenopausal
cers. This association was further confirmed estrogens; in these women, their adiposity would
401

Breast Cancer 401

be the primary determinant of their estrogen has also been of interest. Greater central or upper
exposure. body fat distribution is associated with multi-
A stronger positive association has generally ple hormonal and metabolic changes, including
been observed between weight gain and risk of insulin resistance, decreases in sex hormone-​
postmenopausal breast cancer. This likely occurs binding globulin, and increases in androgen lev-
because the reduction in breast cancer risk due els and the conversion of androgens to estrogens
to overweight in early adulthood appears to con- in peripheral tissues, and thus could increase the
tinue into a woman’s postmenopausal years—​ breast cancer risk. Several different measures of
hence women who are heavy premenopausally fat distribution have been used: waist circumfer-
and postmenopausally will not be at as high a ence, waist/​hip ratio and various skinfold mea-
risk as women who were lean premenopaus- surements. A positive association has been noted
ally and then gained weight (Huang et al, 1997). in most studies of postmenopausal women. For
In the Nurses’ Health Study, postmenopausal example, in one large prospective study (Huang
women who were able to lose weight and keep it et al, 1999) among postmenopausal women who
off realized a reduction in their breast cancer risk never used postmenopausal hormone therapy,
(Eliassen et al, 2006). and controlling for BMI, women in the highest
Given the known importance of adult adi- versus lowest quintile of waist circumference had
posity to breast cancer risk, as well as early life an 83% increased risk of breast cancer (95% CI
factors (e.g., age at menarche) to risk, a number 1.12–​2.99). Associations among premenopausal
of studies also have addressed whether adiposity women have been weaker and less consistent.
earlier in life may influence adult breast cancer
risk. A  weak but consistent positive association Infections
exists between birthweight and breast cancer In 1936, a virus that became known as the
risk (Silva Idos et al, 2008). The majority of now mouse mammary tumor virus (MMTV) was
several dozen studies examining childhood or first reported to substantially increase breast
adolescent fatness and breast cancer have found cancer rates in certain strains of mice (Bittner,
inverse associations (Fuemmeler et al, 2009). For 1936). Since that time, speculation regarding the
example, in one of the largest prospective stud- relationship between viral infection and human
ies (7,582 cases) (Baer et al, 2010), women who breast cancer has persisted. DNA sequences
were heaviest during adolescence (compared to from several viruses have been found in human
the leanest) had a 43% lower breast cancer risk breast cancer cells, breast milk, sera, and cyst
(95% CI 0.37–​0.87); this relationship did not vary fluid (Wang et  al, 1995). However, the relation-
by menopausal status and was largely unchanged ship between the presence of these particles
after accounting for current adult BMI. Further, and breast cancer development and prognosis
the association was observed for both ER+ and remains unknown.
ER-​breast cancer (Fig. 16-​5), which is in contrast In 1998, French researchers discovered the
to the postmenopausal BMI and breast cancer human mammary tumor virus (HMTV), which
association, which is observed primarily in ER+ is genetically similar to MMTV, in 37% of 387
tumors (Vrieling et al, 2010). A number of mech- breast cancer tissue samples (Wang et al, 1995).
anisms may explain the association between Previously, most (albeit limited) attention had
early life body size and adult breast cancer risk, focused on several common viruses in humans
although none have been confirmed (Colditz including Epstein Barr virus (EBV) and human
et al, 2014). Girls who are overweight at this time papilloma virus (HPV) (Amarante & Watanabe,
may experience earlier differentiation of breast 2009). Despite a moderate prevalence of expres-
tissue resulting in tissue that is less susceptible to sion of these viruses in breast tumors, there is lit-
breast cancer initiation. Also, early life body size tle epidemiologic data to support a role of viral
may alter adult hormone levels, such as IGF-​1, a infections in breast cancer etiology.
growth hormone thought to have an important
role in breast carcinogenesis. Physical Activity
Physical activity has been hypothesized to reduce
Body Fat Distribution the risk of breast cancer, but the epidemiologic
As described, both height and overall adiposity data have not been entirely consistent. This may
are consistent breast cancer risk factors. Whether be due to a number of factors including imper-
body fat distribution contributes further to risk fect assessment methods for physical activity,
402

402 Textbook of Cancer Epidemiology

timing of physical activity during the life course, breast cancer risk (RR 0.85, 95% CI 0.78–​0.93; p-​
and heterogeneity by breast cancer subtypes. trend ≤ 0.0001 for ≥27 MET vs <3 MET).
Strenuous activity in girls can delay both men- The modest inverse association between
arche and the onset of regular menstrual cycles. physical activity and breast cancer appears to
In women, strenuous activity, such as competi- be limited to hormone receptor–​positive can-
tive running, can alter circulating hormone lev- cers in most but not all studies. In the pooled
els and increase the frequency of anovulation case-​control study described earlier (Enger et al,
(Willett et al, 2000a). The influence of more mod- 2000), leisure-​time since age 50 was inversely
erate activity on hormone levels and menstrual associated with ER+/​PR+ cancers, but not other
patterns is less well defined. An indirect result subtypes. Results from the NIH-​AARP Diet and
of increased physical activity in postmenopausal Health Study (Peters et  al, 2009)  also observed
women is weight reduction, which would likely similar results. Total weekly energy expenditure
result in a decrease in breast cancer risk (dis- from recreational physical activity (modest-​low
cussed previously). intensity) was inversely associated with ER+
Epidemiologic studies of this association breast cancer (p < 0.01). Already since 2002, the
have varied in their approach to defining phys- International Agency for Research on Cancer of
ical activity, and this has likely contributed to the World Health Organization had concluded
the observed inconsistency in results (Gammon that there is sufficient evidence for the role of
et  al, 1998; Willett et  al, 2000b). Studies have physical activity in preventing breast cancer
focused on activity in different periods of life (Vainio et al, 2002).
(e.g., early life only versus recent activity) or on
different types of activity (e.g., recreational ver- Ionizing Radiation
sus occupational). The frequency and intensity Exposure to ionizing radiation is a well-​
of activity have been assessed in some but not established cause of somatic DNA mutations. The
all studies. relationship with breast cancer specifically has
Overall, epidemiological evidence suggests been evaluated in a number of cohorts exposed
a probable inverse relationship between physi- to moderate to high levels of ionizing radiation,
cal activity and breast cancer risk, specifically in including women exposed to the atomic bomb
postmenopausal women. Results from one of the in Nagasaki and Hiroshima, Japan, and women
first pooled studies (Enger et al, 2000) (two case-​ who received repeated fluoroscopic exams for
control studies) on physical activity and breast tuberculosis (John & Kelsey, 1993). Follow-​up
cancer observed a nonsignificant inverse associ- of these cohorts consistently supports a signifi-
ation between leisure time physical activity since cantly increased risk of breast cancer. The magni-
age 30 and breast cancer risk. However, there tude of the effect increases as the radiation dose
was a significant inverse association between increases and as the age at exposure decreases,
total physical activity and breast cancer risk with the highest risk found in women exposed
(OR 0.90, 95% CI 0.79–​1.04; p-​trend  =  0.027) prepubertally (Davis et  al, 1998; Miller et  al,
and leisure time (OR 0.81, 95% CI 0.70–​0.93; p-​ 1989; Wolff et  al, 1996). As much as a ninefold
trend = 0.017) for the highest quintile versus the increase in risk has been reported within this
lowest quintile (Enger et  al, 2000). In the EPIC exposed subgroup.
study, total physical activity was associated with
a decreased risk of breast cancer (RRactive 0.87, Occupation
95% CI 0.79–​0.97; RRmoderately active 0.92, 95% CI Goldberg and Labreche (Goldberg & Labreche,
0.86–​0.99; p-​trend ≤ 0.01) (Steindorf et al, 2013). 1996)  reviewed 115 studies that addressed the
They also observed inverse associations with relationship between occupation and breast can-
household and recreational physical activity cer risk. They found that in most studies, expo-
(RRnonoccupational physical activity 0.87, 95% CI 0.81–​0.94; sure was categorized according to job title rather
p-​trend ≤ 0.0001 for >123 metabolic equivalent than by quantifying the known or hypothesized
of task [MET] vs ≤ 50.5 MET) (Steindorf et  al, carcinogens to which women in those occupa-
2013). The inverse associations appeared to be tions may be exposed. In addition, few analy-
limited to ER+/​PR+ breast cancer. In the Nurses’ ses controlled for potential confounders, such
Health Study (Eliassen et  al, 2010), postmeno- as reproductive factors or socioeconomic sta-
pausal women that engaged in higher amounts tus, nor did they evaluate past occupation and
of recent total physical activity also had a lower current occupation independently. Therefore,
403

Breast Cancer 403

conclusions regarding the relationship between which type II diabetes may increase breast cancer
specific occupations and breast cancer cannot risk are unknown. However, a number of mecha-
be made. nisms have been proposed including a mito-
More recent studies have focused on occupa- genic effect of insulin through hyperinsulinemia
tions with exposures related to shift work. There (Cannata et al, 2010; Kaaks, 1996), and chronic
has been accumulating evidence that occupa- hyperglycemia via the Warburg effect (Brown
tions involving night shift work may increase the & Simpson, 2010). Further, obesity is associated
risk of breast cancer. A meta-​analysis of 13 obser- with type II diabetes and is also associated with
vational studies found a 48% (RR 1.48, 95% CI postmenopausal breast cancer. Postmenopausal
1.36–​1.61) increased risk of breast cancer among obesity is associated with increased levels of
shift workers (Megdal et al, 2005). Soon after in endogenous estrogens. Thus, it has been difficult
2007, IARC concluded that shift work was prob- to disentangle whether the observed association
ably carcinogenic to humans. A  meta-​analysis between diabetes and breast cancer is causal or
reported that night work was associated with may be due to confounding by BMI.
an increase in breast cancer risk (RR 1.20, 95% Strong mechanistic evidence and some epide-
CI 1.08–​1.33), however the estimates are atten- miologic studies suggest that metformin, a drug
uated when limited to cohort studies (RR 1.08; widely used to treat type 2 diabetes, may reduce
95% CI 0.97–​1.21) (Jia et al, 2013). One under- breast cancer risk. In animal models metformin
lying mechanism for an association is proposed inhibits neoplasia, including transplanted breast
to be the suppression of melatonin levels and cancers (Martin & Marais, 2012). In a meta-​
disruption of circadian rhythm. Accumulating analysis (Col et  al, 2012)  metformin use was
epidemiologic and laboratory data suggest that associated with a significantly lower breast cancer
melatonin may influence breast cancer risk. In risk (RR 0.83, 95% CI 0.71–​0.97) among diabetic
vitro studies indicate that melatonin reduces the women. However, a 2012 methodologic review
growth of breast tumor cells and removal of the (Suissa & Azoulay, 2012) concluded that all pub-
pineal gland, the source of melatonin, boosts lished studies on metformin and cancer were
tumor growth in rodent models (Tamarkin affected by at least one source of time-​related
et  al, 1981). A  meta-​analysis of six prospective bias, such as immortal time bias (person-​time
studies supported a modest inverse association included for metformin users during which
between melatonin levels and breast cancer risk they could not have died). Thus, despite prom-
(Yang et  al, 2014). An increase in urinary 6-​ ising mechanistic data, the association between
sulfatoxymelatonin of 15 ng/​mg creatinine was metformin and breast cancer is not well estab-
associated with a 14% reduced risk of breast can- lished, and rigorous epidemiologic studies are
cer (RR 0.86, 95% CI: 0.78–​0.95). Two additional critical before investing in large prevention trials
prospective cohorts were published subsequently (Carlson, 2013; Pollak, 2012).
and neither observed an association between uri-
nary 6-​sulfatoxymelatonin and risk (Brown et al, Infertility
2015). Together, these studies suggest that shift- Given that nulliparity and later age at first birth are
work may play a role in breast carcinogenesis, but established breast cancer risk factors, the effect of
the mechanism remains unclear. infertility on risk has also been evaluated in epide-
miologic studies. Additionally, there is increasing
Medical Conditions use of fertility drugs, which stimulate ovulation
and Treatment and increase endogenous hormone levels. In the
Diabetes Nurses’ Health Study II cohort, a 60% decrease in
Adult-​onset (type II) diabetes is a common risk was found among women reporting ovulatory
chronic condition that has been increasing in infertility compared to fertile women (Garland
incidence globally. Recent research has examined et  al, 1998). Although a number of studies have
the association between diabetes and breast can- evaluated in vitro fertilization and fertility agents
cer risk. A meta-​analysis of 22 prospective studies on risk of breast cancer, results have been incon-
reported 23% (95% CI 12%–​35%) increased risk sistent (Sergentanis et al, 2014; Tomao et al, 2014).
of breast cancer comparing women with diabetes There are a number of limitations to these studies
to those without (Boyle et  al, 2012). Additional including small sample sizes, short follow-​up, and
adjustment for BMI attenuated the association lack of information on type of infertility, as well as
(RR 1.16, 95% CI 1.08–​1.24). The mechanisms by dose and duration of drug exposures.
40

404 Textbook of Cancer Epidemiology

Nonsteroidal Anti-​Inflammatory Drugs although was later found to provide no benefit

While there is substantial evidence that the use of against spontaneous abortion. Daughters who
nonsteroidal anti-​inflammatory drugs (NSAIDs) were exposed in utero to DES have an increased
decreases the risk of colon cancer, the associ- risk of developing breast cancer (Hoover et  al,
ation between NSAID use and breast cancer 2011)  and the association is stronger among
is less convincing. In a large prospective study women 40 years or older (RR 1.82, 95% CI 1.04–​
(Zhang et al, 2012a), there was a nonsignificant 3.18) and women with vaginal epithelial changes,
inverse association found with the use of two an indicator of higher in utero exposure to DES
or more aspirin tablets per week for more than (RR 2.21, 95% CI 1.11–​ 4.38) (Hoover et  al,
20 years compared to nonusers (RR 0.91, 95% CI 2011). Animal evidence also supports the poten-
0.81–​1.01). The Women’s Health Study evaluated tial importance of early-​life exposures for subse-
long-​term follow-​up of a randomized interven- quent breast cancer risk (Hilakivi-​Clarke & de
tion of alternate-​day low-​dose aspirin therapy Assis, 2006). To address this hypothesis, factors
and found no association with breast cancer likely to be associated with hormone levels dur-
risk (Cook et  al, 2013). More detailed informa- ing pregnancy have been evaluated in relation to
tion on indications for and patterns of NSAID breast cancer risk in the offspring.
use, as well as specific categorization of types of Some, but not all, data suggest that women
both over-​the-​counter and prescription prepara- who are first-​born are at increased risk of breast
tions, need to be evaluated before any conclu- cancer in adulthood; compared to subsequent
sions can be drawn. While overall there does not pregnancies, first pregnancies have been asso-
appear to be a strong protective effect of aspirin ciated with higher circulating estrogen levels
on breast cancer risk, there is accumulating data (Lipworth, 1995). Maternal preeclampsia has
the aspirin may improve survival for women with been inversely associated with breast cancer
breast cancer (Algra & Rothwell, 2012; Holmes in the offspring in one large population-​based
et al, 2010). study in Sweden (Ekbom et al, 1997), perhaps on
account of the lower circulating estrogen levels in
Statins such pregnancies. Dizygotic twins also have been
A growing body of literature suggests that statins observed to have a higher risk of breast cancer
have antitumor activity by interrupting cell-​cycle (Cerhan et al, 2000; Weiss et al, 1997), once again
progression and inducing apoptosis. Statins are because these pregnancies are associated with
a class of lipid-​lowering drugs prescribed for the higher estrogen levels.
prevention of cardiovascular disease. A  meta-​ In a study that compiled and reanalyzed
analysis of randomized trials (Bonovas et  al, individual-​participant birth-​record data from 32
2005)  and observational studies (Undela et  al, studies, comprising 22,058 cases of breast cancer,
2012)  suggest that statins as a group are not an increase of birthweight by 500 grams was asso-
associated with breast cancer risk. However, it is ciated with a statistically significant 6% increase
possible that specific classes of statins and long-​ in breast cancer risk (Silva Idos et al, 2008).
term use may be associated with a reduced risk of Birthweight has also been reported to be
breast cancer. positively associated with adult mammographic
density, a risk factor for breast cancer (discussed
Other Risk Factors later) (Tamimi et  al, 2010a) whereas neonatal
In Utero and Perinatal Exposures growth, a proxy of exposure to growth hormones
There is evidence that intrauterine exposures in utero and during the first days of life, has been
play an important role in the etiology of adult-​ positively associated with adult breast cancer
onset diseases. This, and the role of hormones in risk (Lagiou et  al, 2008). Interestingly, hema-
breast cancer etiology, led to the hypothesis that topoietic stem cell levels, which are positively
the intrauterine hormonal environment might correlated with levels of putative breast stem
influence the risk of breast cancer in adulthood and progenitor cells (Qiu et al, 2012), are posi-
(Trichopoulos, 1990). Direct evidence of a role of tively associated with cord blood levels of IGF-​1
in utero hormonal exposures influencing breast and estriol (Savarese et al, 2007), as well as with
cancer risk comes from diethylstilbestrol (DES)-​ birthweight (Strohsnitter et  al, 2008). Overall,
exposed cohorts. DES is a synthetic estrogen that the accumulated evidence suggests an influence
was prescribed to pregnant women in the 1940s of in utero exposures on the subsequent risk of
to 1960s with the hope of preventing miscarriage, breast cancer.
405

Breast Cancer 405

Benign Breast Disease (Boyd et  al, 2007; Brisson et  al, 1982; Byrne
A number of benign breast conditions have been et  al, 1995; Kato et  al, 1995; Kerlikowske, 2007;
evaluated in terms of their influence on subse- Pettersson et al, 2014; Saftlas et al, 1989; Wolfe,
quent breast cancer risk (Bodian, 1993). A clas- 1976a; Wolfe, 1976b). Women in the highest cat-
sification scheme originally proposed by Dupont egory of mammographic density have a three-​to
and Page (1985) divides benign breast lesions fivefold greater risk of developing breast cancer
into three primary categories:  nonproliferative compared to women in the lowest category (Boyd
breast disease (e.g., cysts, apocrine metaplasia), et al, 2007; Byrne et al, 1995), making mammo-
proliferative breast disease without atypia (e.g., graphic density one of the strongest risk factors
intraductal papilloma, sclerosing adenosis), and for breast cancer (Boyd et  al, 2007; Byrne et  al,
proliferative disease with atypia (atypical hyper- 1995; Kerlikowske, 2007). Importantly, dense tis-
plasia). Compared to women with nonprolif- sue can mask tumors on a mammogram (Bailey
erative breast disease, those with proliferative et al, 2010; Ma et al, 1992; Mandelson et al, 2000).
disease without atypia have approximately a 1.3-​ Although masking bias exists, it cannot explain
to 1.9-​fold higher risk of breast cancer (Bodian, away the strong effects of breast density on breast
1993; Marshall et  al, 1997). Women with atypi- cancer risk, as noted by associations evident even
cal hyperplasia generally are observed to have a 10 years after a mammogram (Boyd et al, 2007;
four-​to sixfold higher risk of breast cancer; this Byrne et  al, 1995; McCormack & dos Santos
appears to be somewhat more marked for women Silva, 2006). Breast density has become such an
with atypical lobular (versus ductal) hyperplasia important risk and masking factor that a number
(Marshall et al, 1997). The subsequent breast can- of US states now mandate women receive their
cer is as likely to be diagnosed in the contralateral individual breast density measures, and a federal
breast as in the ipsilateral breast, which suggests bill is pending.
that these benign conditions serve as general The mechanism by which mammographic
markers of increased risk rather than being pre- density increases breast cancer risk is unclear,
cursor lesions. Further support for this concept is although a number of hypotheses have been put
the observation that the increased risk of breast forward. The primary mechanisms suggested
cancer following benign breast biopsy can last are that mammographic density reflects:  (1)
for several decades. Breast cancer risk prediction the number of mammary stem cells “at risk”
models that consider the type of benign lesion (Trichopoulos et  al, 2008; Trichopoulos &
demonstrate modest improvements (Tamimi Lipman, 1992); (2)  the combined effects of cell
et al, 2010b; Tice et al, 2015) over those that con- proliferation and genetic damage (Martin &
sider benign lesions as a single group. Boyd, 2008); and (3) local estrogen production in
the breast (Vachon et al, 2011).
Mammographic Density
The radiographic appearance of the breast on a Environmental Pollution
mammogram varies depending on the composi- It has been hypothesized that environmental
tion of the breast. Fat is radiolucent and appears carcinogens, whether through occupational, die-
dark. In contrast, epithelial cells and connec- tary, or home exposure, act to increase the risk of
tive tissue are radiodense; they appear light on breast cancer through interaction with hormone
a mammogram and are considered to be “mam- receptors, frequently termed endocrine disrup-
mographically dense.” Thus, both epithelial and tors (Adami et  al, 1995; Colditz, 200;, Emery
stromal cells are responsible for the appearance et  al, 2009). Although many environmental
of dense tissue on a mammogram (Ingleby & pollutants have been identified as carcinogenic
Gerson-​Cohen, 1960). A number of approaches in animal studies, there are few epidemiologic
have been proposed to assess mammographic studies positively identifying these pollutants as
density (American College of Radiology, 2003; risk factors for breast cancer (Brody et al, 2007).
Wolfe, 1976a; Wolfe, 1976b), including quanti- Geographic variation, cluster identification, and
tative assessments using computer-​assisted tech- the increasing incidence of breast cancer over
niques (Byng et  al, 1997), and fully automated time have played a role in promoting the belief
measures (Cheddad et al, 2014; Eng et al, 2014; that environmental toxins are significant con-
Heine et al, 2011). tributors to breast cancer development. As noted
Extensive data support the association in the Institute of Medicine (IOM) report “Breast
between breast density and breast cancer risk Cancer and the Environment:  A Life Course
406

406 Textbook of Cancer Epidemiology

Approach,” part of the challenge has been the and integrated into a proposed etiologic model
lack of new methodologies for classifying envi- consisting of several components (Trichopoulos
ronmental exposures over a woman’s life course et  al, 2008). Breast cancer appears to increase
(IOM, 2012). with the number of susceptible cells in the breast;
To date, several prospective epidemiologic this number is, in turn, influenced by early-​life,
investigations have been relatively consistent in even prenatal, exposures. Both mammographic
showing no substantial increase in risk among density and mammary gland mass are likely to
those with higher dichlorodiphenyltrichloroethane reflect, at least in part, the total number of at-​risk
(DDT) and polychlorinated biphenyl (PCB) blood cells. Several lines of evidence, such as the asso-
levels (Davis et  al, 1998; Laden & Hunter, 1998). ciations with height and early exposure to radia-
Although a modest positive association is difficult tion, support an important influence of early-​life
to exclude, it is unlikely that exposures to these events on subsequent breast cancer risk.
compounds are important causes of breast cancer. In adult life, levels of hormones (whether
Air pollution has been hypothesized to be an resulting from obesity or use of exogenous estro-
environmental risk factor for breast cancer, due gens, for example) and their receptors may influ-
to the presence of potentially carcinogenic con- ence the number of target cells, the likelihood of
stituents. Some of these constituents, for example, retaining spontaneous somatic mutations, and
benzene and polycyclic aromatic hydrocarbons the rate of growth of malignant cells. The role of
(PAHs) have been shown to be associated with circulating high estrogen levels is now well estab-
incidence of breast cancer in occupational stud- lished to increase the risk; the role of other hor-
ies (Gammon et  al, 2002; Labreche et  al, 2010; mones such as prolactin or IGF is also likely to
Meyerhardt et  al, 2006). Two ecologic studies be important. Risk is further influenced by preg-
using SEER registry data demonstrated associa- nancy, which can stimulate the growth of already
tions of increased incidence of breast cancer in initiated cells but conveys long-​term protection,
areas of the country with higher air pollution lev- perhaps through permanent structural changes
els (Chen & Bina, 2012; Wei et al, 2012). A few to the tissue or other still unknown mechanisms.
observational studies in recent years have found Although a number of exposures are now well
suggestive, although not significant evidence that established to influence breast cancer risk, this
air pollution (Bonner et  al, 2005; Lewis-​Michl knowledge, unfortunately, does not readily trans-
et al, 1996; Rohan et al, 2006) may be associated late into a means for cancer prevention. Most of
with an increased risk of breast cancer. Other the established reproductive risk factors, such
environmental exposures such as endocrine dis- as age at menarche, age at first birth, and par-
ruptors and flame retardants have received more ity, are not readily altered or not appropriate for
limited evaluation in epidemiologic studies, and public health intervention. The best-​established
additional research is ongoing. modifiable risk factors include postmenopausal
hormone use, moderate alcohol intake, and
CONCLUSION adult weight gain. Individual decision-​making
As shown in Table 16-​1, at least 15 risk factors is complex, however, as several of these factors,
are now well established. In addition to the con- such as, hormone use, are probably beneficial
firmed risk factors, there are many that either in decreasing the risk of other chronic diseases
have been reported only inconsistently in the such as osteoporosis. Other factors that may
literature or have received limited study to date. decrease the risk of breast cancer include breast-
When interpreting this table, it is important to feeding, physical activity, and increased dietary
note that many factors are continuous variables; intake of carotenoid-​rich fruits and vegetables,
thus, the magnitude of the association is highly although the evidence here is not as strong and
dependent on what two groups are being com- consistent as for other factors. With our increas-
pared. For example, the relative risk associated ing knowledge of the relation between endoge-
with a first birth at age 35 (versus less than age nous hormones and breast cancer risk, and the
20)  is much higher than that associated with a development of SERMs such as tamoxifen and
first birth at age 25 (versus age 20). raloxifene as well as aromatase inhibitors, che-
The mechanisms underlying these confirmed moprevention will likely become more common
or possible risk factors are known with varying in the future. Substantial additional research,
levels of certainty. The epidemiologic literature including risk-​ benefit and cost-​ benefit assess-
related to breast cancer risk has been summarized ments, are needed, however.
407

TABLE 16-​1   EPIDEMIOLOGY OF BREAST CANCER: RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Family history in first-​degree relative ↑↑
Menarche at 12 years versus 14 ↑
Height ↑↑
Benign breast disease ↑↑
Mammographically dense breasts ↑↑
Age at first birth >30 years versus 20 ↑↑
Menopause at >54 years versus 45 ↑↑
High endogenous estrogen levels ↑↑
High endogenous androgen levels ↑↑
High body mass index (premenopausal) ↓
High body mass index (postmenopausal) ↑
Postmenopausal hormone use ↑
Alcohol use (>1 drink/​day) ↑
Tamoxifen ↓
Ionizing radiation exposure ↑↑
Probable Relationship Exists, Based on Substantial Data
In utero exposures** ↑
Childhood body fatness ↓↓
Physical activity ↓
Current oral contraceptive use ↑
Lactation (longer durations) ↓
Folate ↓
Carotenoids/​Carotenoid rich fruits and vegetables ↓
Fiber intake ↓
High endogenous prolactin levels (postmenopausal) ↑↑
High endogenous insulin levels (postmenopausal) ↑
High plasma insulin-​like growth factor levels ↑
Weak, if Any, Relationship Exists, Based on Substantial Data
Total dietary fat intake during adulthood —​
Past oral contraceptive use —​
Induced or spontaneous abortion —​
Cigarette smoking (heavy smoking early in life) ↑
Inconsistent Findings or Limited Study to Date
High endogenous progesterone levels ↑
High endogenous vitamin D levels ↓
Nonsteroidal anti-​inflammatory drug use —​
Adult-​onset diabetes ↑
Thyroid disease —​
Organochlorine exposure —​

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate to large
increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association
Note: The magnitude of the risk can vary substantially, depending on the exact comparison being made. For example, having a
family history of breast cancer in a first-​degree relative is a consistent breast cancer risk factor. However, the magnitude of the
association increases substantially the earlier the age at diagnosis in the relative(s) and with the number of relatives affected.
** associated with higher hormone levels.
408

408 Textbook of Cancer Epidemiology

REFERENCES Anothaisintawee T, Teerawattananon Y, Wiratkapun

ACS. Global Cancer Facts & Figures. 2 ed. Atlanta, C, Kasamesup V, Thakkinstian A. Risk predic-
GA: American Cancer Society, 2011. tion models of breast cancer: a systematic review
ACS. Breast Cancer Facts & Figures  2013–​ 2014. of model performances. Breast Cancer Res Treat
Atlanta, GA: American Cancer Society, Inc, 2013. 2012;133:1–​10.
Adami HO, Lipworth L, Titus-​Ernstoff L, Hsieh CC, Aune D, Chan DS, Greenwood DC, Vieira AR,
Hanberg A, Ahlborg U et al. Organochlorine com- Rosenblatt DA, Vieira R et  al. Dietary fiber
pounds and estrogen-​related cancers in women. and breast cancer risk:  a systematic review and
Cancer Causes Control 1995;6:551–​66. meta-​analysis of prospective studies. Ann Oncol
Adami HO, Signorello LB, Trichopoulos D. Towards 2012;23:1394–​402.
an understanding of breast cancer etiology. Semin Autier P, Koechlin A, Boniol M, Mullie P, Bolli G,
Cancer Biol 1998;8:255–​62. Rosenstock J et  al. Serum insulin and C-​peptide
Ahern TP, Hankinson SE, Willett WC, Pollak MN, concentration and breast cancer: a meta-​analysis.
Eliassen AH, Tamimi RM. Plasma C-​ peptide, Cancer Causes Control 2013;24(5):873–​83.
mammographic breast density, and risk of inva- Baer HJ, Tworoger SS, Hankinson SE, Willett
sive breast cancer. Cancer Epidemiol Biomarkers WC. Body fatness at young ages and risk of
Prev 2013;22:1786–​96. breast cancer throughout life. Am J Epidemiol
AICR. Food, Nutrition, Physical Activity, and the 2010;171:1183–​94.
Prevention of Cancer:  A Global Perspective. Baglietto L, Severi G, English DR, Krishnan K, Hopper
Washington DC: AICR, 2007. JL, McLean C et al. Circulating steroid hormone
Al-​Delaimy WK, Cho E, Chen WY, Colditz G, Willet levels and risk of breast cancer for postmeno-
WC. A prospective study of smoking and risk pausal women. Cancer Epidemiol Biomarkers
of breast cancer in young adult women. Cancer Prev 2010;19:492–​502.
Epidemiol Biomarkers Prev 2004;13(3):398–​404. Bailey SL, Sigal BM, Plevritis SK. A simulation model
Algra AM, Rothwell PM. Effects of regular aspirin investigating the impact of tumor volume dou-
on long-​term cancer incidence and metastasis:  a bling time and mammographic tumor detect-
systematic comparison of evidence from obser- ability on screening outcomes in women aged
vational studies versus randomised trials. Lancet 40-​49 years. J Natl Cancer Inst 2010;102:1263–​71.
Oncol 2012;13:518–​27. Bakken K, Fournier A, Lund E, Waaseth M, Dumeaux
Althuis MD, Fergenbaum JH, Garcia-​ Closas M, V, Clavel-​Chapelon F et  al. Menopausal hor-
Brinton LA, Madigan MP, Sherman ME. Etiology mone therapy and breast cancer risk:  impact of
of hormone receptor-​defined breast cancer: a sys- different treatments:  the European Prospective
tematic review of the literature. Cancer Epidemiol Investigation into Cancer and Nutrition. Int J
Biomarkers Prev 2004;13:1558–​68. Cancer 2011;128:144–​56.
Amarante MK, Watanabe MA. The possible involve- Bassuk SS, Manson JE. Oral contraceptives and men-
ment of virus in breast cancer. J Cancer Res Clin opausal hormone therapy:  relative and attrib-
Oncol 2009;135:329–​37. utable risks of cardiovascular disease, cancer,
Ambrosone CB, Zirpoli G, Hong CC, Yao S, Troester and other health outcomes. Ann Epidemiol
MA, Bandera EV et al. Important role of menar- 2015;25:193–​200.
che in development of estrogen receptor-​negative Bauer SR, Hankinson SE, Bertone-​Johnson ER, Ding
breast cancer in African American women. J Natl EL. Plasma vitamin D levels, menopause, and risk
Cancer Inst 2015;107(9):djv172. doi:10.1093/​jnci/​ of breast cancer:  dose-​ response meta-​ analysis
djv172. of prospective studies. Medicine (Baltimore)
American College of Radiology. Breast Imaging 2013;92:123–​31.
Reporting and Data System (BI-​RADS). Reston, Belardi V, Gallagher EJ, Novosyadlyy R, LeRoith D.
VA: American College of Radiology, 2003. Insulin and IGFs in obesity-​related breast cancer. J
Anderson GL, Limacher M, Assaf AR, Bassford Mammary Gland Biol Neoplasia 2013;18:277–​89.
T, Beresford SA, Black H et  al. Effects of con- Bennett IC, Gattas M, Teh BT. The genetic basis of
jugated equine estrogen in postmenopausal breast cancer and its clinical implications. Aust N
women with hysterectomy:  the Women’s Health Z J Surg 1999;69:95–​105.
Initiative randomized controlled trial. JAMA Beral V, Bull D, Doll R, Peto R, Reeves G, Collaborative
2004;291:1701–​12. Group on Hormonal Factors in Breast C. Breast
Anderson KN, Schwab RB, Martinez ME. cancer and abortion:  collaborative reanalysis of
Reproductive risk factors and breast cancer sub- data from 53 epidemiological studies, including
types: a review of the literature. Breast Cancer Res 83?000 women with breast cancer from 16 coun-
Treat 2014;144:1–​10. tries. Lancet 2004;363:1007–​16.
409

Breast Cancer 409

Beral V, Million Women Study C. Breast cancer and Brody JG, Moysich KB, Humblet O, Attfield KR,
hormone-​ replacement therapy in the Million Beehler GP, Rudel RA. Environmental pollutants
Women Study. Lancet 2003;362:419–​27. and breast cancer: epidemiologic studies. Cancer
Beral V, Reeves G, Bull D, Green J, Million Women 2007;109 Suppl 12:2667–​711.
Study C. Breast cancer risk in relation to the inter- Brown KA, Simpson ER. Obesity and breast can-
val between menopause and starting hormone cer:  progress to understanding the relationship.
therapy. J Natl Cancer Inst 2011;103:296–​305. Cancer Res 2010;70:4–​7.
Bergkvist L, Adami HO, Persson I, Hoover R, Schairer Brown SB, Hankinson SE, Eliassen AH, Reeves KW,
C. The risk of breast cancer after estrogen and Qian J, Arcaro KF et  al. Urinary melatonin con-
estrogen-​progestin replacement. N Engl J Med centration and the risk of breast cancer in Nurses’
1989;321:293–​7. Health Study II. Am J Epidemiol 2015;181:155–​62.
Bernstein L, Ross RK. Endogenous hormones and Byng JW, Yaffe MJ, Lockwood GA, Little LE, Tritchler
breast cancer risk. Epidemiol Rev 1993;15(1):48–​65. DL, Boyd NF. Automated analysis of mammo-
Bittner JJ. Some possible effects of nursing on the graphic densities and breast carcinoma risk.
mammary gland tumor incidence in mice. Science Cancer 1997;80:66–​74.
1936;84:162. Byrne C, Schairer C, Wolfe J, Parekh N, Salane M,
Boardman LA, Thibodeau SN, Schaid DJ, Lindor NM, Brinton LA et  al. Mammographic features and
McDonnell SK, Burgart LJ et  al. Increased risk breast cancer risk: effects with time, age, and men-
for cancer in patients with the Peutz-​Jeghers syn- opause status. J Natl Cancer Inst 1995;87:1622–​9.
drome. Ann Intern Med 1998;128:896–​9. Calle EE, Miracle-​McMahill HL, Thun MJ, Heath CW
Bodian CA. Benign breast diseases, carcinoma Jr. Cigarette smoking and risk of fatal breast can-
in situ, and breast cancer risk. Epidemiol Rev cer. Am J Epidemiol 1994;139:1001–​7.
1993;15:177–​87. Cannata D, Fierz Y, Vijayakumar A, LeRoith D. Type
Boeke CE, Eliassen AH, Chen WY, Cho E, Holmes 2 diabetes and cancer: what is the connection? Mt
MD, Rosner B et al. Dietary fat intake in relation to Sinai J Med 2010;77:197–​213.
lethal breast cancer in two large prospective cohort Carey LA, Perou CM, Livasy CA, Dressler LG, Cowan
studies. Breast Cancer Res Treat 2014;146:383–​92. D, Conway K et al. Race, breast cancer subtypes,
Bonner MR, Han D, Nie J, Rogerson P, Vena JE, Muti and survival in the Carolina Breast Cancer Study.
P et al. Breast cancer risk and exposure in early life JAMA 2006;295:2492–​502.
to polycyclic aromatic hydrocarbons using total Carlson RH. Metformin excitement slows. Oncology
suspended particulates as a proxy measure. Cancer Times 2013;35(SABCS Meeting Reporter Special
Epidemiol Biomarkers Prev 2005;14:53–​60. Edition Supplement for 2012 meeting):20.
Bonovas S, Filioussi K, Tsavaris N, Sitaras NM. Use of Cerhan JR, Kushi LH, Olson JE, Rich SS, Zheng W,
statins and breast cancer: a meta-​analysis of seven Folsom AR et  al. Twinship and risk of post-
randomized clinical trials and nine observational menopausal breast cancer. J Natl Cancer Inst
studies. J Clin Oncol 2005;23:8606–​12. 2000;92:261–​5.
Boyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell Cheddad A, Czene K, Eriksson M, Li J, Easton D,
E et  al. Mammographic density and the risk Hall P et al. Area and volumetric density estima-
and detection of breast cancer. N Engl J Med tion in processed full-​field digital mammograms
2007;356:227–​36. for risk assessment of breast cancer. PLoS One
Boyle P, Boniol M, Koechlin A, Robertson C, 2014;9:e110690.
Valentini F, Coppens K et  al. Diabetes and Chen F, Bina WF. Correlation of white female breast
breast cancer risk:  a meta-​analysis. Br J Cancer cancer incidence trends with nitrogen dioxide
2012;107:1608–​17. emission levels and motor vehicle density pat-
Briand P. Hormone-​dependent mammary tumors in terns. Breast Cancer Res Treat 2012;132:327–​33.
mice and rats as a model for human breast cancer Chen P, Li C, Li X, Li J, Chu R, Wang H. Higher die-
(review). Anticancer Res 1983;3:273–​81. tary folate intake reduces the breast cancer risk: a
Brind J, Chinchilli VM, Severs WB, Summy-​Long J. systematic review and meta-​analysis. Br J Cancer
Induced abortion as an independent risk factor 2014;110:2327–​38.
for breast cancer:  a comprehensive review and Chen S, Parmigiani G. Meta-​ analysis of BRCA1
meta-​ analysis. J Epidemiol Community Health and BRCA2 penetrance. J Clin Oncol
1996;50:481–​96. 2007;25:1329–​33.
Brisson J, Merletti F, Sadowsky NL, Twaddle JA, Chen WY, Manson JE, Hankinson SE, Rosner B,
Morrison AS, Cole P. Mammographic features of Holmes MD, Willett WC et al. Unopposed estro-
the breast and breast cancer risk. Am J Epidemiol gen therapy and the risk of invasive breast cancer.
1982;115:428–​37. Arch Intern Med 2006;166:1027–​32.
410

410 Textbook of Cancer Epidemiology

Chen WY, Rosner B, Hankinson SE, Colditz GA, women with breast cancer and 96973 women with-
Willett WC. Moderate alcohol consumption dur- out the disease. Lancet 2002;360:187–​95.
ing adult life, drinking patterns, and breast cancer Collaborative Group on Hormonal Factors in Breast
risk. JAMA 2011;306:1884–​90. Cancer. Menarche, menopause, and breast can-
Chia KS, Reilly M, Tan CS, Lee J, Pawitan Y, Adami HO cer risk:  individual participant meta-​ analysis,
et al. Profound changes in breast cancer incidence including 118 964 women with breast cancer
may reflect changes into a Westernized lifestyle: a from 117 epidemiological studies. Lancet Oncol
comparative population-​based study in Singapore 2012;13:1141–​51.
and Sweden. Int J Cancer 2005;113:302–​6. Committee on the Relationship Between Oral
Chlebowski RT, Hendrix SL, Langer RD, Stefanick Contraceptives and Breast Cancer Institute of
ML, Gass M, Lane D et al. Influence of estrogen Medicine. Oral contraceptives and Breast Cancer.
plus progestin on breast cancer and mammog- Washington, DC: Author, 1991.
raphy in healthy postmenopausal women:  the Consortium CBCC-​ C. CHEK2*1100delC and sus-
Women’s Health Initiative Randomized Trial. ceptibility to breast cancer:  a collaborative anal-
JAMA 2003;289:3243–​53. ysis involving 10,860 breast cancer cases and
Clevenger CV, Furth PA, Hankinson SE, Schuler LA. 9,065 controls from 10 studies. Am J Hum Genet
The role of prolactin in mammary carcinoma. 2004;74:1175–​82.
Endocr Rev 2003;24:1–​27. Cook NR, Lee IM, Zhang SM, Moorthy MV, Buring
Col NF, Ochs L, Springmann V, Aragaki AK, JE. Alternate-​ day, low-​dose aspirin and cancer
Chlebowski RT. Metformin and breast cancer risk: long-​term observational follow-​up of a ran-
risk: a meta-​analysis and critical literature review. domized trial. Ann Intern Med 2013;159:77–​85.
Breast Cancer Res Treat 2012;135:639–​46. Cook NR, Rosner BA, Hankinson SE, Colditz GA.
Colditz GA, Bohlke K, Berkey CS. Breast cancer risk Mammographic screening and risk factors for
accumulation starts early:  prevention must also. breast cancer. Am J Epidemiol 2009;170:1422–​32.
Breast Cancer Res Treat 2014;145:567–​79. Cui Y, Miller AB, Rohan TE. Cigarette smoking and
Colditz GA, Hankinson SE, Hunter DJ, Willett WC, breast cancer risk: update of a prospective cohort
Manson JE, Stampfer MJ et  al. The use of estro- study. Breast Cancer Res Treat 2006;100:293–​9.
gens and progestins and the risk of breast can- Cuzick J, Sestak I, Bonanni B, Costantino JP,
cer in postmenopausal women. N Engl J Med Cummings S, DeCensi A et al. Selective oestrogen
1995;332:1589–​93. receptor modulators in prevention of breast can-
Colditz GA, Rosner B. Cumulative risk of breast can- cer:  an updated meta-​analysis of individual par-
cer to age 70  years according to risk factor sta- ticipant data. Lancet 2013;381:1827–​34.
tus:  data from the Nurses’ Health Study. Am J Cuzick J, Sestak I, Forbes JF, Dowsett M, Knox J,
Epidemiol 2000;152:950–​64. Cawthorn S et  al. Anastrozole for prevention
Collaborative Group on Hormonal Factors in Breast of breast cancer in high-​ risk postmenopausal
C. Breast cancer and breastfeeding:  collaborative women (IBIS-​II):  an international, double-​blind,
reanalysis of individual data from 47 epidemio- randomised placebo-​ controlled trial. Lancet
logical studies in 30 countries, including 50302 2014;383:1041–​8.
women with breast cancer and 96973 women Davis DL, Axelrod D, Bailey L, Gaynor M, Sasco AJ.
without the disease. Lancet 2002;360:187–​95. Rethinking breast cancer risk and the environ-
Collaborative Group on Hormonal Factors in Breast ment:  the case for the precautionary principle.
Cancer. Breast cancer and hormonal contracep- Environ Health Perspect 1998;106:523–​9.
tives:  collaborative reanalysis of individual data Davis S. Androgen replacement in women: a commen-
on 53 297 women with breast cancer and 100 239 tary. J Clin Endocrinol Metab 1999;84:1886–​91.
women without breast cancer from 54 epidemio- Deng CX. BRCA1: cell cycle checkpoint, genetic insta-
logical studies. Lancet 1996;347:1713–​27. bility, DNA damage response and cancer evolu-
Collaborative Group on Hormonal Factors in Breast tion. Nucleic Acids Res 2006;34:1416–​26.
Cancer. Breast cancer and hormone replacement Dowsett M, Cuzick J, Ingle J, Coates A, Forbes J, Bliss
therapy:  collaborative reanalysis of data from 51 J et  al. Meta-​analysis of breast cancer outcomes
epidemiological studies of 52,705 women with in adjuvant trials of aromatase inhibitors versus
breast cancer and 108,411 women without breast tamoxifen. J Clin Oncol 2010;28:509–​18.
cancer. Lancet 1997;350:1047–​59. Dunning AM, Healey CS, Pharoah PD, Teare MD,
Collaborative Group on Hormonal Factors in Breast Ponder BA, Easton DF. A systematic review
Cancer. Breast cancer and breastfeeding:  collabo- of genetic polymorphisms and breast can-
rative reanalysis of individual data from 47 epide- cer risk. Cancer Epidemiol Biomarkers Prev
miological studies in 30 countries, including 50302 1999;8:843–​54.
41

Breast Cancer 411

Dupont WD, Page DL. Risk factors for breast cancer of a National Cancer Institute-​sponsored work-
in women with proliferative breast disease. N Engl shop. J Natl Cancer Inst 2013;105:166–​74.
J Med 1985;312:146–​51. Fisher B, Costantino JP, Wickerham DL, Redmond
Ekbom A, Hsieh CC, Lipworth L, Adami HQ, CK, Kavanah M, Cronin WM et al. Tamoxifen for
Trichopoulos D. Intrauterine environment and prevention of breast cancer: report of the National
breast cancer risk in women: a population-​based Surgical Adjuvant Breast and Bowel Project P-​1
study. J Natl Cancer Inst 1997;89:71–​6. Study. J Natl Cancer Inst 1998;90:1371–​88.
Eliassen AH, Colditz GA, Rosner B, Willett WC, Fournier DB, Erdman JW Jr, Gordon GB. Soy, its com-
Hankinson SE. Adult weight change and ponents, and cancer prevention: a review of the in
risk of postmenopausal breast cancer. JAMA vitro, animal, and human data. Cancer Epidemiol
2006;296:193–​201. Biomarkers Prev 1998;7:1055–​65.
Eliassen AH, Hankinson SE, Rosner B, Holmes MD, Freedman AN, Seminara D, Gail MH, Hartge P,
Willett WC. Physical activity and risk of breast Colditz GA, Ballard-​Barbash R et al. Cancer risk
cancer among postmenopausal women. Arch prediction models:  a workshop on development,
Intern Med 2010;170:1758–​64. evaluation, and application. J Natl Cancer Inst
Eliassen AH, Hendrickson SJ, Brinton LA, Buring 2005;97:715–​23.
JE, Campos H, Dai Q et  al. Circulating carot- Friedenreich CM. Review of anthropometric fac-
enoids and risk of breast cancer: pooled analysis tors and breast cancer risk. Eur J Cancer Prev
of eight prospective studies. J Natl Cancer Inst 2001;10:15–​32.
2012a;1041905–​16. Frittitta L, Vigneri R, Stampfer MR, Goldfine ID.
Eliassen AH, Spiegelman D, Xu X, Keefer LK, Veenstra Insulin receptor overexpression in 184B5 human
TD, Barbieri RL et al. Urinary estrogens and estro- mammary epithelial cells induces a ligand-​
gen metabolites and subsequent risk of breast can- dependent transformed phenotype. J Cell
cer among premenopausal women. Cancer Res Biochem 1995;57:666–​9.
2012b;72:696–​706. Fuemmeler BF, Pendzich MK, Tercyak KP. Weight,
Ellisen LW, Haber DA. Hereditary breast cancer. Annu dietary behavior, and physical activity in child-
Rev Med 1998;49:425–​36. hood and adolescence: implications for adult can-
Emery LA, Tripathi A, King C, Kavanah M, Mendez cer risk. Obes Facts 2009;2:179–​86.
J, Stone MD, et  al. Early dysregulation of cell Fulan H, Changxing J, Baina WY, Wencui Z,
adhesion and extracellular matrix pathways Chunqing L, Fan W et  al. Retinol, vitamins A,
in breast cancer progression. Am J Pathol C, and E and breast cancer risk:  a meta-​analysis
2009;175:1292–​302. and meta-​ regression. Cancer Causes Control
Endogenous Hormones Breast Cancer Collaborative 2011;22:1383–​96.
Group. Insulin-​like growth factor 1 (IGF1), IGF Gammon MD, John EM, Britton JA. Recreational and
binding protein 3 (IGFBP3), and breast cancer occupational physical activities and risk of breast
risk:  pooled individual data analysis of 17 pro- cancer. J Natl Cancer Inst 1998;90:100–​17.
spective studies. Lancet Oncol 2010;11:530–​42. Gammon MD, Santella RM, Neugut AI, Eng SM,
Endogenous Hormones Breast Cancer Collaborative Teitelbaum SL, Paykin A et  al. Environmental
Group. Sex hormones and risk of breast can- toxins and breast cancer on Long Island.
cer in premenopausal women:  a collabora- I.  Polycyclic aromatic hydrocarbon DNA
tive reanalysis of individual participant data adducts. Cancer Epidemiol Biomarkers Prev
from seven prospective studies. Lancet Oncol 2002;11:677–​85.
2013;14:1009–​19. Garland M, Hunter DJ, Colditz GA, Manson JE,
Eng A, Gallant Z, Shepherd J, McCormack V, Li J, Stampfer MJ, Spiegelman D et  al. Menstrual
Dowsett M et al. Digital mammographic density cycle characteristics and history of ovulatory
and breast cancer risk: a case-​control study of six infertility in relation to breast cancer risk in a
alternative density assessment methods. Breast large cohort of US women. Am J Epidemiol
Cancer Res 2014;16:439. 1998;147:636–​43.
Enger SM, Ross RK, Paganini-​Hill A, Carpenter CL, Gaudet MM, Gapstur SM, Sun J, Diver WR, Hannan
Bernstein L. Body size, physical activity, and LM, Thun MJ. Active smoking and breast cancer
breast cancer hormone receptor status:  results risk: original cohort data and meta-​analysis. J Natl
from two case-​control studies. Cancer Epidemiol Cancer Inst 2013a;105:515–​25.
Biomarkers Prev 2000;9:681–​7. Gaudet MM, Patel AV, Teras LR, Sun J, Campbell PT,
Faupel-​Badger JM, Arcaro KF, Balkam JJ, Eliassen AH, Stevens VL et  al. Obesity-​ related markers and
Hassiotou F, Lebrilla CB et al. Postpartum remod- breast cancer in CPS-​II Nutrition Cohort. Int J
eling, lactation, and breast cancer risk: summary Mol Epidemiol Genet 2013b;4:156–​66.
412

412 Textbook of Cancer Epidemiology

Goldberg MS, Labreche F. Occupational risk factors Herman V, Kalk WJ, de Moor NG, Levin J. Serum
for female breast cancer: a review. Occup Environ prolactin after chest wall surgery:  elevated lev-
Med 1996;53:145–​56. els after mastectomy. J Clin Endocrinol Metab
Goldhirsch A, Winer EP, Coates AS, Gelber 1981;52:148–​51.
RD, Piccart-​ Gebhart M, Thurlimann B et  al. Hilakivi-​Clarke L, de Assis S. Fetal origins of breast
Personalizing the treatment of women with cancer. Trends Endocrinol Metab 2006;17:340–​8.
early breast cancer:  highlights of the St Gallen Holmes MD, Chen WY, Li L, Hertzmark E, Spiegelman
International Expert Consensus on the Primary D, Hankinson SE. Aspirin intake and survival
Therapy of Early Breast Cancer 2013. Annals of after breast cancer. J Clin Oncol 2010;28:1467–​72.
Oncology 2013;24:2206–​23. Hoover RN, Hyer M, Pfeiffer RM, Adam E, Bond B,
Goss PE, Ingle JN, Ales-​ Martinez JE, Cheung Cheville AL et  al. Adverse health outcomes in
AM, Chlebowski RT, Wactawski-​ Wende J women exposed in utero to diethylstilbestrol. N
et  al. Exemestane for breast-​ cancer preven- Engl J Med 2011;365:1304–​14.
tion in postmenopausal women. N Engl J Med Howe GR, Hirohata T, Hislop TG, Iscovich JM, Yuan
2011;364:2381–​91. JM, Katsouyanni K et al. Dietary factors and risk
Gradishar W. Landmark trials in endocrine adju- of breast cancer:  combined analysis of 12 case-​
vant therapy for breast carcinoma. Cancer control studies. J Natl Cancer Inst 1990;82:561–​9.
2006;106:975–​81. Huang Z, Hankinson SE, Colditz GA, Stampfer MJ,
Gram IT, Braaten T, Terry PD, Sasco AJ, Adami HO, Hunter DJ, Manson JE et al. Dual effects of weight
Lund E et  al. Breast cancer risk among women and weight gain on breast cancer risk. JAMA
who start smoking as teenagers. Cancer Epidemiol 1997;278:1407–​11.
Biomarkers Prev 2005;14:61–​6. Huang Z, Willett WC, Colditz GA, Hunter DJ, Manson
Greendale GA, Reboussin BA, Sie A, Singh HR, Olson JE, Rosner B et al. Waist circumference, waist:hip
LK, Gatewood O et  al. Effects of estrogen and ratio, and risk of breast cancer in the Nurses’
estrogen-​progestin on mammographic parenchy- Health Study. Am J Epidemiol 1999;150:1316–​24.
mal density. Postmenopausal Estrogen/​Progestin Hunter DJ, Colditz GA, Hankinson SE, Malspeis S,
Interventions (PEPI) Investigators. Ann Intern Spiegelman D, Chen W et  al. Oral contracep-
Med 1999;130(4 Pt 1):262–​9. tive use and breast cancer:  a prospective study
Greene MH. Genetics of breast cancer. Mayo Clin of young women. Cancer Epidemiol Biomarkers
Proc 1997;72:54–​65. Prev 2010;19:2496–​502.
Gunter MJ, Hoover DR, Yu H, Wassertheil-​Smoller Hunter DJ, Morris JS, Stampfer MJ, Colditz GA,
S, Rohan TE, Manson JE et  al. Insulin, insulin-​ Speizer FE, Willett WC. A prospective study of
like growth factor-​ I, and risk of breast cancer selenium status and breast cancer risk. JAMA
in postmenopausal women. J Natl Cancer Inst 1990;264:1128–​31.
2009;101:48–​60. Hunter DJ, Spiegelman D, Adami HO, Beeson L, van
Hamajima N, Hirose K, Tajima K, Rohan T, Calle den Brandt PA, Folsom AR et al. Cohort studies of
EE, Heath CW Jr et  al. Alcohol, tobacco and fat intake and the risk of breast cancer—​a pooled
breast cancer—​collaborative reanalysis of indi- analysis. N Engl J Med 1996;334:356–​61.
vidual data from 53 epidemiological studies, Hunter DJ, Willett WC. Diet, body size, and breast
including 58,515 women with breast cancer and cancer. Epidemiol Rev 1993;15:110–​32.
95,067 women without the disease. Br J Cancer IARC. GLOBOCAN 2012:  Estimated Cancer
2002;87:1234–​45. Incidence, Mortality and Prevalence Worldwide
Harris J, Lippman M, Morrow M, Osborne C. Diseases in 2012. International Agency for Research on
of the Breast. 5 ed. Philadelphia: Wolters Kluwer/​ Cancer and World Health Organization, Lyon,
Lippincott Williams & Wilkins Health, 2014. France, 2014.
Heine JJ, Cao K, Rollison DE. Calibrated measures for Ingleby H, Gerson-​Cohen J. Comparative Anatomy,
breast density estimation. Academic Radiology Pathology, and Roentgenology of the Breast.
2011;18:547–​55. Philadelphia:  University of Philadelphia
Hemminki E, Kennedy DL, Baum C, McKinlay SM. Press, 1960.
Prescribing of noncontraceptive estrogens and Institute of Medicine (IOM). Breast Cancer and
progestins in the United States, 1974–​86. Am J the Environment:  A Life Course Approach.
Public Health 1988;78:1479–​81. Washington, DC: National Academies Press, 2012.
Henderson B, Pike M, Bernstein L, Ross R. Breast can- Iwasaki M, Inoue M, Otani T, Sasazuki S, Kurahashi
cer. In: Schottenfeld D, Fraumeni JFJr (eds.) Cancer N, Miura T et al. Plasma isoflavone level and sub-
Epidemiology and Prevention. New York: Oxford sequent risk of breast cancer among Japanese
University Press, 1996: 1022–​35. women:  a nested case-​ control study from the
413

Breast Cancer 413

Japan Public Health Center-​ based prospective oestrogens and breast cancer risk:  the European
study group. J Clin Oncol 2008;26:1677–​83. prospective investigation into cancer and nutri-
James RE, Lukanova A, Dossus L, Becker S, Rinaldi tion. Endocr Relat Cancer 2005;12:1071–​82.
S, Tjonneland A et  al. Postmenopausal serum Kato I, Beinart C, Bleich A, Su S, Kim M, Toniolo
sex steroids and risk of hormone receptor-​ PG. A nested case-​ control study of mammo-
positive and -​negative breast cancer:  a nested graphic patterns, breast volume, and breast can-
case-​control study. Cancer Prev Res (Phila) cer (New  York City, NY, United States). Cancer
2011;4:1626–​35. Causes Control 1995;6:431–​8.
Jia Y, Lu Y, Wu K, Lin Q, Shen W, Zhu M et  al. Kelsey JL. A review of the epidemiology of human
Does night work increase the risk of breast can- breast cancer. Epidemiol Rev 1979;1:74–​109.
cer? A  systematic review and meta-​ analysis Kelsey JL, Gammon MD, John EM. Reproductive
of epidemiological studies. Cancer Epidemiol factors and breast cancer. Epidemiol Rev
2013;37:197–​206. 1993;15:36–​47.
Jin F, Shu XO, Devesa SS, Zheng W, Blot WJ, Gao YT. Kennedy DL, Baum C, Forbes MB. Noncontraceptive
Incidence trends for cancers of the breast, ovary, estrogens and progestins: use patterns over time.
and corpus uteri in urban Shanghai, 1972–​ 89. Obstet Gynecol 1985;65:441–​6.
Cancer Causes Control 1993;4:355–​60. Kerlikowske K. The mammogram that cried Wolfe. N
John EM, Kelsey JL. Radiation and other environmen- Engl J Med 2007;356(3):297–​300.
tal exposures and breast cancer. Epidemiol Rev Key T, Appleby P, Barnes I, Reeves G, Endogenous H,
1993;15:157–​62. Breast Cancer Collaborative G. Endogenous sex
Johnson KC, Miller AB, Collishaw NE, Palmer JR, hormones and breast cancer in postmenopausal
Hammond SK, Salmon AG et  al. Active smok- women:  reanalysis of nine prospective studies. J
ing and secondhand smoke increase breast cancer Natl Cancer Inst 2002;94:606–​16.
risk: the report of the Canadian Expert Panel on Khan SA, Rogers MA, Khurana KK, Meguid MM,
Tobacco Smoke and Breast Cancer Risk (2009). Numann PJ. Estrogen receptor expression in
Tob Control 2011;20:e2. benign breast epithelium and breast cancer risk. J
Jordan VC. The new biology of estrogen-​ induced Natl Cancer Inst 1998;90:37–​42.
apoptosis applied to treat and prevent breast can- Kwan ML, Kushi LH, Weltzien E, Maring B, Kutner
cer. Endocr Relat Cancer 2015;22:R1–​31. SE, Fulton RS et  al. Epidemiology of breast can-
Jung S, Spiegelman D, Baglietto L, Bernstein L, Boggs cer subtypes in two prospective cohort studies
DA, van den Brandt PA et al. Fruit and vegetable of breast cancer survivors. Breast Cancer Res
intake and risk of breast cancer by hormone recep- 2009;11:R31.
tor status. J Natl Cancer Inst 2013;105:219–​36. Labreche F, Goldberg MS, Valois MF, Nadon L.
Jung S, Wang M, Anderson K, Baglietto L, Bergkvist L, Postmenopausal breast cancer and occupational
Bernstein L et al. Alcohol consumption and breast exposures. Occup Environ Med 2010;67:263–​9.
cancer risk by estrogen receptor status: in a pooled Laden F, Hunter DJ. Environmental risk factors and
analysis of 20 studies. Int J Epidemiol 2015 Aug female breast cancer. Annu Rev Public Health
28. pii: dyv156. [Epub ahead of print]. 1998;19:101–​23.
Juul A, Bang P, Hertel NT, Main K, Dalgaard P, Lagiou P, Georgila C, Samoli E, Lagiou A, Zourna
Jorgensen K et  al. Serum insulin-​ like growth P, Minaki P et  al. Estrogen alpha and progester-
factor-​I in 1030 healthy children, adolescents, and one receptor expression in the normal mammary
adults: relation to age, sex, stage of puberty, testic- epithelium in relation to breast cancer risk. Int J
ular size, and body mass index. J Clin Endocrinol Cancer 2009;124:440–​2.
Metab 1994;78:744–​52. Lagiou P, Hsieh CC, Trichopoulos D, Adami HO, Hall
Juul A, Dalgaard P, Blum WF, Bang P, Hall K, P, Chie L et al. Neonatal growth and breast cancer
Michaelsen KF et al. Serum levels of insulin-​like risk in adulthood. Br J Cancer 2008;99:1544–​8.
growth factor (IGF)-​binding protein-​3 (IGFBP-​3) Lagiou P, Samoli E, Lagiou A, Zourna P, Barbouni
in healthy infants, children, and adolescents:  the A, Georgila C et  al. A comparison of hormo-
relation to IGF-​I, IGF-​II, IGFBP-​1, IGFBP-​2, age, nal profiles between breast cancer and benign
sex, body mass index, and pubertal maturation. J breast disease:  a case-​control study. Ann Oncol
Clin Endocrinol Metab 1995;80:2534–​42. 2013;24:2527–​33.
Kaaks R. Nutrition, hormones, and breast cancer:  is Lakhani SR, Van De Vijver MJ, Jacquemier J, Anderson
insulin the missing link? Cancer Causes Control TJ, Osin PP, McGuffog L et  al. The pathology of
1996;7:605–​25. familial breast cancer: predictive value of immu-
Kaaks R, Rinaldi S, Key TJ, Berrino F, Peeters PH, nohistochemical markers estrogen receptor, pro-
Biessy C et al. Postmenopausal serum androgens, gesterone receptor, HER-​2, and p53 in patients
41

414 Textbook of Cancer Epidemiology

with mutations in BRCA1 and BRCA2. J Clin Mandelson MT, Oestreicher N, Porter PL, White D,
Oncol 2002;20:2310–​8. Finder CA, Taplin SH et al. Breast density as a pre-
Lambe M, Hsieh C, Trichopoulos D, Ekbom A, Pavia dictor of mammographic detection:  comparison
M, Adami HO. Transient increase in the risk of of interval-​and screen-​detected cancers. J Natl
breast cancer after giving birth. N Engl J Med Cancer Inst 2000;92:1081–​7.
1994;331:5–​9. Manjer J, Johansson R, Berglund G, Janzon L, Kaaks
Lawson JS, Field AS, Champion S, Tran D, Ishikura R, Agren A et  al. Postmenopausal breast cancer
H, Trichopoulos D. Low oestrogen receptor risk in relation to sex steroid hormones, prolac-
alpha expression in normal breast tissue under- tin and SHBG (Sweden). Cancer Causes Control
lies low breast cancer incidence in Japan. Lancet 2003;14:599–​607.
1999;354:1787–​8. Manson JE, Bassuk SS, Lee IM, Cook NR, Albert MA,
Lee SA, Shu XO, Li H, Yang G, Cai H, Wen W et al. Gordon D et  al. The VITamin D and OmegA-​3
Adolescent and adult soy food intake and breast TriaL (VITAL):  rationale and design of a large
cancer risk:  results from the Shanghai Women’s randomized controlled trial of vitamin D and
Health Study. Am J Clin Nutr 2009;89:1920–​6. marine omega-​3 fatty acid supplements for the
Lewis-​Michl EL, Melius JM, Kallenbach LR, Ju CL, primary prevention of cancer and cardiovascular
Talbot TO, Orr MF et  al. Breast cancer risk and disease. Contemp Clin Trials 2012;33:159–​71.
residence near industry or traffic in Nassau and Marchbanks PA, McDonald JA, Wilson HG, Folger
Suffolk Counties, Long Island, New  York. Arch SG, Mandel MG, Daling JR et al. Oral contracep-
Environ Health 1996;51:255–​65. tives and the risk of breast cancer. N Engl J Med
Li CI, Beaber EF, Tang MT, Porter PL, Daling JR, 2002;346:2025–​32.
Malone KE. Effect of depo-​medroxyprogesterone Marshall LM, Hunter DJ, Connolly JL, Schnitt SJ,
acetate on breast cancer risk among women 20 to Byrne C, London SJ et  al. Risk of breast cancer
44 years of age. Cancer Res 2012;72:2028–​35. associated with atypical hyperplasia of lobular and
Liao DJ, Dickson RB. Roles of androgens in the ductal types. Cancer Epidemiol Biomarkers Prev
development, growth, and carcinogenesis of the 1997;6:297–​301.
mammary gland. J Steroid Biochem Mol Biol Martin LJ, Boyd NF. Mammographic density.
2002;80:175–​89. Potential mechanisms of breast cancer risk asso-
Lipworth L. Epidemiology of breast cancer. Eur J ciated with mammographic density:  hypotheses
Cancer Prev 1995;4:7–​30. based on epidemiological evidence. Breast Cancer
Lipworth L, Bailey LR, Trichopoulos D. History of Res 2008;10:201.
breast-​feeding in relation to breast cancer risk: a Martin M, Marais R. Metformin:  a diabetes drug
review of the epidemiologic literature. J Natl for cancer, or a cancer drug for diabetics? J Clin
Cancer Inst 2000;92:302–​12. Oncol 2012;30:2698–​700.
Liu Y, Colditz GA, Rosner B, Berkey CS, Collins LC, Mathews-​Roth MM. Plasma concentrations of carot-
Schnitt SJ et al. Alcohol intake between menarche enoids after large doses of beta-​carotene. Am J
and first pregnancy: a prospective study of breast Clin Nutr 1990;52:500–​1.
cancer risk. J Natl Cancer Inst 2013;105:1571–​8. Mavaddat N, Pharoah PD, Michailidou K, Tyrer J,
Ma L, Fishell E, Wright B, Hanna W, Allan S, Boyd Brook MN, Bolla MK et  al. Prediction of breast
NF. Case-​control study of factors associated with cancer risk based on profiling with common
failure to detect breast cancer by mammography. J genetic variants. J Natl Cancer Inst 2015;107(5).
Natl Cancer Inst 1992;84:781–​5. McCormack VA, dos Santos Silva I. Breast den-
MacGregor JI, Jordan VC. Basic guide to the mecha- sity and parenchymal patterns as markers of
nisms of antiestrogen action. Pharmacol Rev breast cancer risk:  a meta-​ analysis. Cancer
1998;50:151–​96. Epidemiology, Biomarkers and Prevention
MacMahon B, Cole P, Lin TM, Lowe CR, Mirra AP, 2006;15:1159–​69.
Ravnihar B et al. Age at first birth and breast can- Meads C, Ahmed I, Riley RD. A systematic review of
cer risk. Bull World Health Organ 1970;43:209–​21. breast cancer incidence risk prediction models
MacMahon B, Trichopoulos D, Brown J, Andersen AP, with meta-​analysis of their performance. Breast
Cole P, deWaard F et  al. Age at menarche, urine Cancer Res Treat 2012;132:365–​77.
estrogens and breast cancer risk. Int J Cancer Megdal SP, Kroenke CH, Laden F, Pukkala E,
1982;30:427–​31. Schernhammer ES. Night work and breast cancer
Majumdar SR, Almasi EA, Stafford RS. Promotion risk: a systematic review and meta-​analysis. Eur J
and prescribing of hormone therapy after report Cancer 2005;41:2023–​32.
of harm by the Women’s Health Initiative. JAMA Meijers-​Heijboer H, van den Ouweland A, Klijn J,
2004;292:1983–​8. Wasielewski M, de Snoo A, Oldenburg R et  al.
415

Breast Cancer 415

Low-​penetrance susceptibility to breast cancer due postmenopausal breast cancer risk in the NIH-​
to CHEK2(*)1100delC in noncarriers of BRCA1 AARP diet and health study. Cancer Epidemiol
or BRCA2 mutations. Nat Genet 2002;31:55–​9. Biomarkers Prev 2009;18:289–​96.
Meyerhardt JA, Giovannucci EL, Holmes MD, Chan Pettersson A, Graff RE, Ursin G, Santos Silva ID,
AT, Chan JA, Colditz GA et  al. Physical activity McCormack V, Baglietto L et al. Mammographic
and survival after colorectal cancer diagnosis. J density phenotypes and risk of breast cancer:  a
Clin Oncol 2006;24:3527–​34. meta-​analysis. J Natl Cancer Inst 2014;106(5).
Michailidou K, Beesley J, Lindstrom S, Canisius S, Pharoah PD, Guilford P, Caldas C, International Gastric
Dennis J, Lush MJ et al. Genome-​wide association Cancer Linkage C. Incidence of gastric cancer and
analysis of more than 120,000 individuals identi- breast cancer in CDH1 (E-​cadherin) mutation car-
fies 15 new susceptibility loci for breast cancer. riers from hereditary diffuse gastric cancer fami-
Nat Genet 2015;47:373–​80. lies. Gastroenterology 2001;121:1348–​53.
Miller AB, Howe GR, Sherman GJ, Lindsay JP, Yaffe Pollak M. IGF-​I physiology and breast cancer. Recent
MJ, Dinner PJ et al. Mortality from breast cancer Results Cancer Res 1998;152:63–​70.
after irradiation during fluoroscopic examinations Pollak M. Insulin and insulin-​like growth factor signal-
in patients being treated for tuberculosis. N Engl J ling in neoplasia. Nat Rev Cancer 2008;8:915–​28.
Med 1989;321:1285–​9. Pollak MN. Investigating metformin for cancer pre-
Millikan RC, Newman B, Tse CK, Moorman PG, vention and treatment: the end of the beginning.
Conway K, Dressler LG et  al. Epidemiology of Cancer Discov 2012;2:778–​90.
basal-​like breast cancer. Breast Cancer Res Treat Pradhan AD, Manson JE. Update on the Vitamin D
2008;109:123–​39. and OmegA-​3 trial (VITAL). J Steroid Biochem
Missmer SA, Eliassen AH, Barbieri RL, Hankinson Mol Biol 2016;155(Pt B):252–​6.
SE. Endogenous estrogen, androgen, and pro- Prentice RL, Caan B, Chlebowski RT, Patterson R,
gesterone concentrations and breast cancer risk Kuller LH, Ockene JK et al. Low-​fat dietary pattern
among postmenopausal women. J Natl Cancer and risk of invasive breast cancer:  the Women’s
Inst 2004;96:1856–​65. Health Initiative Randomized Controlled Dietary
Mosher WD, Jones J. Use of contraception in the Modification Trial. JAMA 2006;295:629–​42.
United States:  1982–​2008. Vital Health Stat 23 Qiu L, Low HP, Chang CI, Strohsnitter WC, Anderson
2010:1–​44. M, Edmiston K et al. Novel measurements of mam-
Nagata C, Kawakami N, Shimizu H. Trends in the mary stem cells in human umbilical cord blood as
incidence rate and risk factors for breast cancer prospective predictors of breast cancer suscepti-
in Japan. Breast Cancer Res Treat 1997;44:75–​82. bility in later life. Ann Oncol 2012;23:245–​50.
Narod SA. Modifiers of risk of hereditary breast can- Rahman N, Seal S, Thompson D, Kelly P, Renwick A,
cer. Oncogene 2006;25:5832–​6. Elliott A et al. PALB2, which encodes a BRCA2-​
Palmer JR, Rosenberg L. Cigarette smoking and risk interacting protein, is a breast cancer susceptibil-
of breast cancer. In: Kelsey JL (ed.), Epidemiologic ity gene. Nat Genet 2007;39:165–​7.
Reviews. Baltimore:  John Hopkins University of Rakha EA, Reis-​ Filho JS, Ellis IO. Basal-​ like
Hygiene and Public Health, 1993: 145–​56. breast cancer:  a critical review. J Clin Oncol
Palmer JR, Viscidi E, Troester MA, Hong CC, Schedin 2008;26:2568–​81.
P, Bethea TN et al. Parity, lactation, and breast can- Ravdin PM. The lack, need, and opportunities for
cer subtypes in African American women: results decision-​making and informational tools to edu-
from the AMBER Consortium. J Natl Cancer Inst cate primary-​care physicians and women about
2014;5:106(10). breast cancer chemoprevention. Cancer Prev Res
Park SY, Kolonel LN, Lim U, White KK, Henderson (Phila) 2010;3:686–​8.
BE, Wilkens LR. Alcohol consumption and breast Ravdin PM, Cronin KA, Howlader N, Berg CD,
cancer risk among women from five ethnic groups Chlebowski RT, Feuer EJ et  al. The decrease in
with light to moderate intakes:  the Multiethnic breast-​cancer incidence in 2003 in the United
Cohort Study. Int J Cancer 2014;134:1504–​10. States. N Engl J Med 2007;356:1670–​4.
Pasqualini JR, Chetrite G, Blacker C, Feinstein MC, Renehan AG, Zwahlen M, Minder C, O’Dwyer ST,
Delalonde L, Talbi M et  al. Concentrations of Shalet SM, Egger M. Insulin-​ like growth fac-
estrone, estradiol, and estrone sulfate and evalu- tor (IGF)-​I, IGF binding protein-​3, and cancer
ation of sulfatase and aromatase activities in pre-​ risk: systematic review and meta-​regression anal-
and postmenopausal breast cancer patients. J Clin ysis. Lancet 2004;363:1346–​53.
Endocrinol Metab 1996;81:1460–​4. Roberts KD, Rochefort JG, Bleau G, Chapdelaine A.
Peters TM, Schatzkin A, Gierach GL, Moore SC, Lacey Plasma estrone sulfate levels in postmenopausal
JV Jr, Wareham NJ et  al. Physical activity and women. Steroids 1980;35:179–​87.
416

416 Textbook of Cancer Epidemiology

Rohan TE, Li SQ, Hartwick R, Kandel RA. p53 of breast cancer (United States). Cancer Causes
Alterations and protein accumulation in benign Control 1994;5:491–​500.
breast tissue and breast cancer risk:  a cohort Schairer C, Lubin J, Troisi R, Sturgeon S, Brinton L,
study. Cancer Epidemiol Biomarkers Prev Hoover R. Menopausal estrogen and estrogen-​
2006;15:1316–​23. progestin replacement therapy and breast cancer
Rosenberg L, Boggs DA, Wise LA, Adams-​Campbell risk. JAMA 2000;283:485–​91.
LL, Palmer JR. Oral contraceptive use and estro- Seal S, Thompson D, Renwick A, Elliott A, Kelly
gen/​ progesterone receptor-​ negative breast can- P, Barfoot R et  al. Truncating mutations in the
cer among African American women. Cancer Fanconi anemia J gene BRIP1 are low-​penetrance
Epidemiol Biomarkers Prev 2010;19:2073–​9. breast cancer susceptibility alleles. Nat Genet
Rosenfeld RG. Insulin-​like growth factors and the 2006;38:1239–​41.
basis of growth. N Engl J Med 2003;349:2184–​6. SEER. Cancer Statistics Review, 1975–​2011. 2014a.
Rosner B, Colditz GA, Willett WC. Reproductive SEER. Stat Fact Sheets: Breast Cancer. 2014b. https://​
risk factors in a prospective study of breast can- seer.cancer.gov/​statfacts/​html/​breast.html
cer:  the Nurses’ Health Study. Am J Epidemiol Seitz HK, Pelucchi C, Bagnardi V, La Vecchia C.
1994;139:819–​35. Epidemiology and pathophysiology of alcohol
Rossouw JE, Anderson GL, Prentice RL, LaCroix AZ, and breast cancer: Update 2012. Alcohol Alcohol
Kooperberg C, Stefanick ML et al. Risks and ben- 2012;47:204–​12.
efits of estrogen plus progestin in healthy post- Sergentanis TN, Diamantaras AA, Perlepe C,
menopausal women:  principal results From the Kanavidis P, Skalkidou A, Petridou ET. IVF and
Women’s Health Initiative randomized controlled breast cancer:  a systematic review and meta-​
trial. JAMA 2002;288:321–​33. analysis. Hum Reprod Update 2014;20:106–​23.
Russo J, Gusterson BA, Rogers AE, Russo IH, Wellings Seymour-​ Munn K, Adams J. Estrogenic effects of
SR, van Zwieten MJ. Comparative study of human 5-​androstene-​3 beta, 17 beta-​diol at physiolog-
and rat mammary tumorigenesis. Lab Invest ical concentrations and its possible implication
1990;62:244–​78. in the etiology of breast cancer. Endocrinology
Russo J, Russo IH. Molecular basis of pregnancy-​ 1983;112:486–​91.
induced breast cancer prevention. Horm Mol Biol Silva Idos S, De Stavola B, McCormack V, Collaborative
Clin Investig 2012;9:3–​10. Group on Pre-​Natal Risk F, Subsequent Risk of
Russo J, Tay LK, Russo IH. Differentiation of the Breast C. Birth size and breast cancer risk:  re-​
mammary gland and susceptibility to carcinogen- analysis of individual participant data from 32
esis. Breast Cancer Res Treat 1982;2:5–​73. studies. PLoS Med 2008;5:e193.
Saftlas AF, Wolfe JN, Hoover RN, Brinton LA, Schairer Smith-​Warner SA, Spiegelman D, Adami HO, Beeson
C, Salane M et  al. Mammographic parenchymal WL, van den Brandt PA, Folsom AR et al. Types of
patterns as indicators of breast cancer risk. Am J dietary fat and breast cancer: a pooled analysis of
Epidemiol 1989;129:518–​26. cohort studies. Int J Cancer 2001;92:767–​74.
Salamone LM, Dallal GE, Zantos D, Makrauer F, Stahl W, Sies H. Uptake of lycopene and its geomet-
Dawson-​Hughes B. Contributions of vitamin D rical isomers is greater from heat-​processed than
intake and seasonal sunlight exposure to plasma from unprocessed tomato juice in humans. J Nutr
25-​hydroxyvitamin D concentration in elderly 1992;122:2161–​6.
women. Am J Clin Nutr 1994;59:80–​6. Stanford JL, Thomas DB. Exogenous progestins and
Savarese TM, Strohsnitter WC, Low HP, Liu Q, Baik breast cancer. Epidemiol Rev 1993;15:98–​107.
I, Okulicz W et al. Correlation of umbilical cord Steindorf K, Ritte R, Eomois PP, Lukanova A, Tjonneland
blood hormones and growth factors with stem A, Johnsen NF et  al. Physical activity and risk of
cell potential:  implications for the prenatal ori- breast cancer overall and by hormone receptor sta-
gin of breast cancer hypothesis. Breast Cancer Res tus: the European prospective investigation into can-
2007;9:R29. cer and nutrition. Int J Cancer 2013;132:1667–​78.
Saxena T, Lee E, Henderson KD, Clarke CA, West Stephen AM, Wald NJ. Trends in individual consump-
D, Marshall SF et al. Menopausal hormone ther- tion of dietary fat in the United States, 1920–​1984.
apy and subsequent risk of specific invasive Am J Clin Nutr 1990;52:457–​69.
breast cancer subtypes in the California Teachers Strohsnitter WC, Savarese TM, Low HP, Chelmow DP,
Study. Cancer Epidemiol Biomarkers Prev Lagiou P, Lambe M et  al. Correlation of umbili-
2010;19:2366–​78. cal cord blood haematopoietic stem and progen-
Schairer C, Byrne C, Keyl PM, Brinton LA, Sturgeon itor cell levels with birth weight: implications for
SR, Hoover RN. Menopausal estrogen and a prenatal influence on cancer risk. Br J Cancer
estrogen-​progestin replacement therapy and risk 2008;98:660–​3.
417

Breast Cancer 417

Strom BL, Berlin JA, Weber AL, Norman SA, recent update for an old problem. Curr Opin
Bernstein L, Burkman RT et  al. Absence of an Obstet Gynecol 2014;26:130–​7.
effect of injectable and implantable progestin-​only Tominaga S. Cancer incidence in Japanese in Japan,
contraceptives on subsequent risk of breast can- Hawaii, and western United States. Natl Cancer
cer. Contraception 2004;69:353–​60. Inst Monogr 1985;69:83–​92.
Suissa S, Azoulay L. Metformin and the risk of can- Trichopoulos D. Hypothesis: does breast cancer origi-
cer:  time-​related biases in observational studies. nate in utero? Lancet 1990;335(8695):939–​40.
Diabetes Care 2012;35:2665–​73. Trichopoulos D, Adami HO, Ekbom A, Hsieh CC,
Tamarkin L, Cohen M, Roselle D, Reichert C, Lippman Lagiou P. Early life events and conditions and
M, Chabner B. Melatonin inhibition and pine- breast cancer risk: from epidemiology to etiology.
alectomy enhancement of 7,12-​dimethylbenz(a) Int J Cancer 2008;122:481–​5.
anthracene-​induced mammary tumors in the rat. Trichopoulos D, Hsieh CC, MacMahon B, Lin TM,
Cancer Res 1981;41(11 Pt 1):4432–​6. Lowe CR, Mirra AP et  al. Age at any birth and
Tamimi RM, Colditz GA, Hazra A, Baer HJ, breast cancer risk. Int J Cancer 1983;31:701–​4.
Hankinson SE, Rosner B et al. Traditional breast Trichopoulos D, Lipman RD. Mammary gland
cancer risk factors in relation to molecular sub- mass and breast cancer risk. Epidemiology
types of breast cancer. Breast Cancer Res Treat 1992;3:523–​6.
2012;131:159–​67. Trichopoulos D, MacMahon B, Cole P. Menopause
Tamimi RM, Colditz GA, Wang Y, Collins LC, Hu R, and breast cancer risk. J Natl Cancer Inst
Rosner B et  al. Expression of IGF1R in normal 1972;48:605–​13.
breast tissue and subsequent risk of breast cancer. Trichopoulou A. Olive oil and breast cancer. Cancer
Breast Cancer Res Treat 2011;128:243–​50. Causes Control 1995;6:475–​6.
Tamimi RM, Eriksson L, Lagiou P, Czene K, Ekbom Trichopoulou A, Katsouyanni K, Stuver S, Tzala L,
A, Hsieh CC et  al. Birth weight and mammo- Gnardellis C, Rimm E et  al. Consumption of
graphic density among postmenopausal women olive oil and specific food groups in relation to
in Sweden. Int J Cancer 2010a;126:985–​91. breast cancer risk in Greece. J Natl Cancer Inst
Tamimi RM, Hankinson SE, Chen WY, Rosner B, 1995;87:110–​6.
Colditz GA. Combined estrogen and testosterone Tworoger SS, Eliassen AH, Zhang X, Qian J, Sluss PM,
use and risk of breast cancer in postmenopausal Rosner BA et  al. A 20-​year prospective study of
women. Arch Intern Med 2006;166:1483–​9. plasma prolactin as a risk marker of breast cancer
Tamimi RM, Rosner B, Colditz GA. Evaluation of a development. Cancer Res 2013;73:4810–​9.
breast cancer risk prediction model expanded to Tworoger SS, Zhang X, Eliassen AH, Qian J, Colditz
include category of prior benign breast disease GA, Willett WC et  al. Inclusion of endoge-
lesion. Cancer 2010b;116:4944–​53. nous hormone levels in risk prediction models
Tan MH, Mester JL, Ngeow J, Rybicki LA, Orloff MS, of postmenopausal breast cancer. J Clin Oncol
Eng C. Lifetime cancer risks in individuals with 2014;32:3111–​7.
germline PTEN mutations. Clin Cancer Res Undela K, Srikanth V, Bansal D. Statin use and
2012;18:400–​7. risk of breast cancer: a meta-​analysis of obser-
Tice JA, Miglioretti DL, Li CS, Vachon CM, Gard vational studies. Breast Cancer Res Treat
CC, Kerlikowske K. Breast density and benign 2012;135:261–​9.
breast disease:  risk assessment to identify Vachon CM, Pankratz VS, Scott CG, Haeberle L, Ziv
women at high risk of breast cancer. J Clin Oncol E, Jensen MR et  al. The contributions of breast
2015;33:3137–​43. density and common genetic variation to breast
Tikk K, Sookthai D, Johnson T, Rinaldi S, Romieu cancer risk. J Natl Cancer Inst 2015;4;107(5).
I, Tjonneland A et  al. Circulating prolactin and Vachon CM, Sasano H, Ghosh K, Brandt KR, Watson
breast cancer risk among pre-​and postmeno- DA, Reynolds C et  al. Aromatase immunoreac-
pausal women in the EPIC cohort. Ann Oncol tivity is increased in mammographically dense
2014;25:1422–​8. regions of the breast. Breast Cancer Res Treat
Toledo E, Salas-​ Salvado J, Donat-​ Vargas C, Buil-​ 2011;125:243–​52.
Cosiales P, Estruch R, Ros E et al. Mediterranean Vainio H, Kaaks R, Bianchini F. Weight control and
diet and invasive breast cancer risk among women physical activity in cancer prevention:  interna-
at high cardiovascular risk in the PREDIMED tional evaluation of the evidence. Eur J Cancer
trial:  a randomized clinical trial. JAMA Intern Prev 2002;11 Suppl 2:S94–​100.
Med 2015;175:1752–​60. van den Brandt PA, Spiegelman D, Yaun SS, Adami
Tomao F, Lo Russo G, Spinelli GP, Tomao S. Clinical HO, Beeson L, Folsom AR et  al. Pooled anal-
use of fertility agents and risk of breast cancer: a ysis of prospective cohort studies on height,
418

418 Textbook of Cancer Epidemiology

weight, and breast cancer risk. Am J Epidemiol Morrow M, Osborne C (eds.), Diseases of the
2000;152:514–​27. Breast. Philadelphia:  Lippincott-​ Raven Press,
Verheus M, van Gils CH, Keinan-​Boker L, Grace PB, 2000b: 175–​220.
Bingham SA, Peeters PH. Plasma phytoestrogens Wolfe JN. Breast patterns as an index of risk for
and subsequent breast cancer risk. J Clin Oncol developing breast cancer. Am J Roentgenol
2007;25:648–​55. 1976a;126:1130–​7.
Vrieling A, Buck K, Kaaks R, Chang-​ Claude J. Wolfe JN. Risk for breast cancer development deter-
Adult weight gain in relation to breast cancer mined by mammographic parenchymal pattern.
risk by estrogen and progesterone receptor sta- Cancer 1976b;37:2486–​92.
tus:  a meta-​analysis. Breast Cancer Res Treat Wolff MS, Collman GW, Barrett JC, Huff J. Breast
2010;123:641–​9. cancer and environmental risk factors:  epidemi-
Wang Y, Holland JF, Bleiweiss IJ, Melana S, Liu X, ological and experimental findings. Annu Rev
Pelisson I et  al. Detection of mammary tumor Pharmacol Toxicol 1996;36:573–​96.
virus env gene-​like sequences in human breast Wolk A, Bergstrom R, Hunter D, Willett W, Ljung H,
cancer. Cancer Res 1995;55:5173–​9. Holmberg L et al. A prospective study of associ-
Warmuth MA, Sutton LM, Winer EP. A review of ation of monounsaturated fat and other types of
hereditary breast cancer: from screening to risk fat with risk of breast cancer. Arch Intern Med
factor modification. Am J Med 1997;102:407–​15. 1998;158:41–​5.
Waters EA, McNeel TS, Stevens WM, Freedman AN. Work ME, John EM, Andrulis IL, Knight JA, Liao Y,
Use of tamoxifen and raloxifene for breast can- Mulligan AM et al. Reproductive risk factors and
cer chemoprevention in 2010. Breast Cancer Res oestrogen/​progesterone receptor-​negative breast
Treat 2012;134:875–​80. cancer in the Breast Cancer Family Registry. Br J
Webb AR, Kline L, Holick MF. Influence of season Cancer 2014;110:1367–​77.
and latitude on the cutaneous synthesis of vita- Wu AH, Yu MC, Tseng CC, Pike MC. Epidemiology of
min D3:  exposure to winter sunlight in Boston soy exposures and breast cancer risk. Br J Cancer
and Edmonton will not promote vitamin D3 syn- 2008;98:9–​14.
thesis in human skin. J Clin Endocrinol Metab Yager JD, Liehr JG. Molecular mechanisms of estro-
1988;67:373–​8. gen carcinogenesis. Annu Rev Pharmacol Toxicol
Webb AR, Pilbeam C, Hanafin N, Holick MF. An eval- 1996;36:203–​32.
uation of the relative contributions of exposure to Yamamoto S, Sobue T, Kobayashi M, Sasaki S, Tsugane
sunlight and of diet to the circulating concentra- S, Japan Public Health Center-​Based Prospective
tions of 25-​hydroxyvitamin D in an elderly nurs- Study on Cancer Cardiovascular Diseases G. Soy,
ing home population in Boston. Am J Clin Nutr isoflavones, and breast cancer risk in Japan. J Natl
1990;51:1075–​81. Cancer Inst 2003;95:906–​13.
Wei Y, Davis J, Bina WF. Ambient air pollution Yang WS, Deng Q, Fan WY, Wang WY, Wang X.
is associated with the increased incidence of Light exposure at night, sleep duration, melato-
breast cancer in US. Int J Environ Health Res nin, and breast cancer:  a dose-​response analy-
2012;22:12–​21. sis of observational studies. Eur J Cancer Prev
Weiss HA, Potischman NA, Brinton LA, Brogan 2014;23:269–​76.
D, Coates RJ, Gammon MD et  al. Prenatal and Yang XR, Chang-​ Claude J, Goode EL, Couch FJ,
perinatal risk factors for breast cancer in young Nevanlinna H, Milne RL et  al. Associations
women. Epidemiology 1997;8:181–​7. of breast cancer risk factors with tumor sub-
WHO Collaborative Study of Neoplasia and Steroid types:  a pooled analysis from the Breast Cancer
Contraceptives. Breast cancer and depot-​ Association Consortium studies. J Natl Cancer
medroxyprogesterone acetate:  a multinational Inst 2011;103:250–​63.
study. Lancet 1991;338:833–​8. Yen SSC, Jaffe RB. Reproductive Endocrinology.
Willett WC. Nutritional Epidemiology. 2 ed. New Philadelphia: WB Saunders, 1991.
York: Oxford University Press, 1998. Zhang B, Shu XO, Delahanty RJ, Zeng C, Michailidou
Willett WC, Colditz G, Stampfer M. Postmenopausal K, Bolla MK et  al. Height and breast cancer
estrogens—​opposed, unopposed, or none of the risk:  evidence from prospective studies and
above. JAMA 2000a;283:534–​5. Mendelian randomization. J Natl Cancer Inst
Willett WC, Rockhill B, Hankinson SE, Hunter D, 2015; 20;107(11).
Colditz G. Non-​ genetic factors in the causa- Zhang X, Smith-​ Warner SA, Collins LC,
tion of breast cancer. In:  Harris JP, Lippman M, Rosner B, Willett WC, Hankinson SE. Use of
419

Breast Cancer 419

aspirin, other nonsteroidal anti-​ inflammatory breast cancer risk over 20  years of follow-​ up.
drugs, and ­ acetaminophen and postmeno- Breast Cancer Res Treat 2013;137:883–​92.
pausal breast ­ cancer incidence. J Clin Oncol Ziegler RG, Fuhrman BJ, Moore SC, Matthews CE.
2012a;30:3468–​77. Epidemiologic studies of estrogen metabolism
Zhang X, Spiegelman D, Baglietto L, Bernstein L, and breast cancer. Steroids 2015;99(Pt A):67–​75.
Boggs DA, van den Brandt PA et  al. Carotenoid Ziegler RG, Hoover RN, Pike MC, Hildesheim A,
intakes and risk of breast cancer defined by estro- Nomura AM, West DW et al. Migration patterns
gen receptor and progesterone receptor status:  a and breast cancer risk in Asian-​American women.
pooled analysis of 18 prospective cohort studies. J Natl Cancer Inst 1993;85(22):1819–​27.
Am J Clin Nutr 2012b;95:713–​25. Zinser G, Packman K, Welsh J. Vitamin D(3) receptor
Zhang X, Tworoger SS, Eliassen AH, Hankinson SE. ablation alters mammary gland morphogenesis.
Postmenopausal plasma sex hormone levels and Development 2002;129:3067–​76.
420
421

17
Cervical Cancer
K A R I N S U N D S T R Ö M , J O A K I M D I L L N E R , A N D H A N S - ​O L O V   A D A M I

T here has been a long prelude to the current

realization that infection has an important
role in human cancer. Observations already in
increasingly common. Squamous cell cancer
has a distinct precursor detectable by cytologic
screening, namely, carcinoma in situ (CIS).
the 18th and 19th centuries indicated that sex- Likewise, the immediate precursor to adenocar-
ual behavior was associated with cancer of the cinoma is named adenocarcinoma in situ (AIS),
cervix, because the disease was rare in nuns but but is generally more difficult to detect by the
common in prostitutes. However, it was not until hitherto standard method of cytological screen-
the late 1970s that the sexually transmitted agent ing. Tumors of rare histological types can also
human papilloma virus (HPV) that explains the occur in the cervix, including, but not limited to,
association between sexual activity and cervi- clear cell carcinoma and sarcomas.
cal cancer was proposed (zur Hausen, 1977). Dysplastic, noninvasive precancerous lesions
Since then, numerous studies have confirmed with varying grades of severity can also be
the association, and today persistent infection detected in the cervix. The principal systems
with certain oncogenic types of HPV is consid- used to classify cervical abnormalities include
ered the main, though not sufficient, cause of one proposed by the World Health Organization
cervical cancer (Bosch et  al, 2002; Walboomers (WHO), one advanced by Richart, and the
et al, 1999). Needless to say, this understanding Bethesda System developed in conjunction with
opened concrete prospects for primary preven- the US National Cancer Institute (Kiviat et  al,
tion through HPV vaccination, which is now 1992; IARC, 2007a; Mitchell et  al, 1996). The
available throughout most of the world. original WHO system involved histologic classi-
Methodologically, studies of the viral etiology fication of cervical dysplasia as mild, moderate,
of cervical cancer have offered numerous chal- or severe as well as a separate category for CIS.
lenges both in designing epidemiologic studies Richart introduced the term “cervical intraepi-
and in measuring exposure to oncogenic viruses. thelial neoplasia” (CIN) to classify squamous
Molecular biology allowed research to go beyond cervical lesions into three stages:  CIN I  repre-
crude surrogate measures of sexually transmit- sents mild dysplasia; CIN II, moderate dyspla-
ted infection, such as age at first intercourse sia; and CIN III, severe dysplasia, or CIS. In
and number of sexual partners. Much work was the Bethesda System, which is based on cytol-
required, however, before HPV infection became ogy from Papanicolaou (Pap) smear readings,
the established cause of cervical cancer. This a low-​ grade squamous intraepithelial lesion
work has now led to the successful introduc- (LSIL) corresponds to CIN I/​mild dysplasia and
tion of HPV-​DNA-​based screening technologies, koilocytotic atypia and a high-​grade SIL (HSIL)
which are replacing Pap-​smear based screening encompasses moderate and severe dysplasia/​
because they are anticipated to further reduce the CIN II and CIN III/​CIS.
cervical cancer burden worldwide. Symptoms CIS is typically symptomless, whereas
invasive cancer of the cervix causes postcoital
CLINICAL SYNOPSIS and spontaneous bleeding, discharge, and dis-
Subgroups Although squamous cell cancer comfort during intercourse. In advanced disease,
accounts for 80% to 90% of all malignancies of which is predominant in low-​income countries,
the cervix in most settings, adenocarcinoma, the a wide variety of symptoms arise due to exten-
other main histopathologic type, is becoming sive local tumor growth and/​or metastasis.
42

422 Textbook of Cancer Epidemiology

Diagnosis A  combination of visual inspection, DESCRIPTIVE

aided by colposcopy, acetic acid application, EPIDEMIOLOGY
and palpation are highly informative. Cytologic In excess of half a million women develop cer-
examination of a Pap smear reveals whether vical cancer every year, which makes this cancer
abnormal cells are present. Biopsy with histo- the most common in eastern and middle Africa,
pathologic confirmation is necessary for definite and the fourth most frequently occurring malig-
diagnosis. nancy among women worldwide. Mortality var-
Treatment As with most other malignancies, ies 18-​fold between different areas of the world,
treatment planning depends profoundly on with women in low-​income countries account-
the clinical stage. In localized disease, cure can ing for almost 90% of these deaths (Ferlay
be achieved by surgical removal, irradiation et  al, 2012). There is only limited variation in
(internal or external), or a combination of these age-​standardized incidence between the high-​
modalities. For very early stage cancers, fertility-​ income countries shown in Fig. 17-​1, except for
preserving surgery through cervical conization, Poland and Japan. In several Western countries
or, the partial hysterectomy procedure trachelec- there have been substantial declines in incidence
tomy, is available for women whose childbear- rates following cytologic screening during the
ing is not complete. For more advanced stages, latter 20th century (Gustafsson et al, 1997). On a
radiochemotherapy with a combination of global scale, however, incidence rates of invasive
weekly cisplatinum-​based chemotherapy, exter- cervical cancer vary substantially between differ-
nal radiation, and cervical brachytherapy, is rec- ent regions, from less than 5 per 100,000 in west-
ommended to improve survival. ern Asia to more than 40 per 100,000 in parts of
eastern Africa (Ferlay et al, 2012). In general, the
Prognosis Cure rates depend strongly on the
highest incidence rates are found in sub-​Saharan
stage at diagnosis. In the United States and most
Africa, Melanesia, South-​central and Southeast
European countries, the 5-​year relative survival
Asia, Latin America, and the Caribbean; whereas
rate is approximately 65% but ranges between
the lowest rates are observed in North America,
15% and 90%, depending on the extent of the
Australia/​New Zealand, and western Asia (Ferlay
disease.
et al, 2012).
Progress During the 20th century, earlier clin- Unlike most other solid tumors, the inci-
ical diagnosis greatly reduced mortality from dence of cervix cancer becomes high during
cervix cancer in many high-​income countries. middle age, with limited further increase at older
However, since the 1980s, there has been little, if ages in many settings (Fig. 17-​2). However, these
any, improvement in the age-​standardized mor- age-​specific incidence figures have been distorted
tality incidence ratio worldwide. markedly by screening for cervical cancer, which

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 2 4 6 8 10 12 14
ASR (world) per 100,000 person-years

FIGURE 17-​1  Age-​standardized (to the 2012 world population) incidence rates of cervical carcinoma among women.
Source: Ferlay et al, 2013.
423

Cervical Cancer 423

New cases per 100,000

40
35

person-years
30
25
20
15
10
5
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 17-​2  Age-​specific incidence rates of cervical carcinoma.

Source: Ferlay et al, 2013.

provides proportionally greater protection in and trials of HPV-​based screening have demon-
younger than in older women (Gustafsson et al, strated an increased effectiveness of HPV screen-
1997), and where screening is absent, cervical ing over cytology-​based screening against this
cancer is still a malignancy of the young. histological type (Ronco et al, 2014).
The mortality incidence ratio is about 50% Furthermore, some reports have indicated
worldwide (Ferlay et  al, 2012)  but tends to be possible increasing rates of cervical cancer in
lower in low-​ risk areas due chiefly to earlier regions such as eastern Europe, and among
clinical diagnosis followed by adequate treat- younger women in Central Asia, Japan, and
ment (Pontén et  al, 1995), so that 5-​year sur- China, believed chiefly to be due to increasing
vival is around 20% in low human development prevalence of sexually transmitted HPV (Torre
index (HDI) countries and >65% in very high et al, 2012). If robust, these trends further illus-
HDI countries (Forman et  al, 2012), with large trate the need for continued vigilance in cervical
variations even within Europe, depending on cancer prevention.
screening performance (von Karsa et  al, 2015).
However, because women are diagnosed with GENETIC AND MOLECULAR
and die from this disease, on average, at younger EPIDEMIOLOGY
ages than women with other types of cancer,
the total years of life expectancy lost due to this Inherited Susceptibility
major female cancer form are substantial (Bosch Heritable factors have been proposed to play
et al, 2012). a role in cervical cancer development, as sug-
In countries where organized cytological gested from a large study on siblings (Magnusson
screening was implemented during the 20th cen- et  al, 2000), although this has not been con-
tury, the incidence of and mortality from cervi- firmed in twin studies, perhaps because power
cal carcinoma have declined steadily (Gustafsson may have been limited due to the rarity of the
et  al, 1997; Torre et  al, 2012). However, the disease (Mucci et al, 2016). Candidate gene and
observed effectiveness of cytological screening subsequent genome-​ wide association studies
is mainly attributed to the predominant squa- (GWAS) have however identified many genetic
mous cell type of cervical carcinoma. For cervical variants associated with risk of cervical neo-
adenocarcinomas, which constitute about 10%–​ plasia and cancer (Chen & Gyllensten, 2015).
20% of all cervical cancers, increases in the age-​ The first hypothesis-​ driven studies frequently
adjusted incidence rates, primarily among young focused on classic human leukocyte antigen
women, have been reported in several well-​ (HLA)-​alleles, where several different DRB1 and
screened populations (Dahlström et  al, 2010; DQB1 haplotypes were implicated in suscepti-
Bergström et  al, 1999; Castellsagué et  al, 2006). bility (odds ratios (ORs) of around 2) to as well
These upward trends have raised the concern that as protection from (ORs of 0.3–​0.4) cervical car-
routine cytological screening might not be an cinogenesis (Hildesheim & Wang, 2002). Two
efficient way to prevent against adenocarcinoma, population-​ based studies in Scandinavia both
42

424 Textbook of Cancer Epidemiology

identified HLA-​DQ6 and HLA-​DR15 as suscep- cancer patients (Tornesello et  al, 2014; Wright
tibility markers for high-​ grade CIN (Helland et al, 2013). One study found that 60% of cervical
et  al, 1998; Sanjeevi et  al, 1996). These associa- tumors examined harbored validated mutations
tions were further confirmed in a meta-​analysis and 9% harbored concurrent mutations in two
on the relation between different HLA class  II or more genes. Again, molecularly distinct pro-
haplotypes and cervical cancer (Konya & Dillner, files in squamous cell–​versus adenocarcinoma-​
2001). Of interest, some HLA-​associations were derived tissue were found (Wright et al, 2013).
only evident among carriers of HPV type 16, the
oncogenic HPV type that is most important for RISK FACTORS
cervical cancer development. Thus, it has been
proposed that the role of HLA type in cervical Tobacco
cancer occurrence involves the host’s immune Along with HPV infection and oral contracep-
response to HPV infection and the ability of the tives (OCs), tobacco smoking is considered
virus to persist and induce malignant transfor- one of the major agents where the literature has
mation. Other studies have implicated several shown sufficient evidence for cervical carcino-
non-​ HLA loci involved in immune response, genicity in humans (IARC, 2004; World Cancer
DNA repair, cell cycle, and other processes (Chen Research Fund, 2007). In addition to numerous
& Gyllensten, 2015)  in cervical cancer risk, but previous studies (Dillner et al, 1997; Chichareon
results are so far not conclusive. et  al, 1998; Ngelangel et  al, 1998; Ylitalo et  al,
GWAS of cervical CIS confirmed several 1999), two large collaborative studies lent further
previously reported HLA associations, as well support for the association between smoking and
as reported three novel susceptibility loci with cervical carcinoma. One study used pooled data
ORs of 0.67 for protective and 1.25–​1.42 for risk restricted to HPV-​positive women only within
alleles (Chen et  al, 2013). These findings have the IARC’s multicenter case-​control studies; the
provided novel leads but have not been consist- other pooled and reanalyzed data from 23 dif-
ent across all GWAS or ethnicities, and the SNPs ferent epidemiological studies (International
identified so far explain only a small fraction of Collaborative Study of Cervical Cancer, 2006a;
cervical cancer heritability (Chen & Gyllensten, Plummer et  al, 2003). Both studies reported an
2015). Improved understanding of genetic inter- approximate doubling in risk among current
action with persistence and/​ or progression of smokers compared to never smokers in models
HPV infection would be essential. adjusted for other life-​style factors such as OCs
and parity. In the latter study, risk was further
Somatic Events increased with number of cigarettes smoked
Cervical cancers may harbor a number of per day and younger age at starting smoking.
somatic mutations. Aneuploidy with loss of het- In both studies the effect of smoking appeared
erozygosity in the short arm at chromosome 3p to be limited to squamous cell carcinoma of the
and gain in the long arm at 3q are among the cervix, a finding supported by previous work
most frequently reported. Specific mutations in (International Collaborative Study of Cervical
proto-​oncogenes such as RAS, MYC, and ERB Cancer, 2006a; Plummer et  al, 2003; Schiffman
have also been reported, although their relevance and Brinton, 1995; Winkelstein, 1990).
is not entirely clear (IARC, 2012a). A  sequenc- Cigarette smoking may promote carcino-
ing analysis of 115 cervical carcinoma-​normal genicity both by affecting local cell-​ mediated
paired samples reported previously unknown immune response and by inducing genetic dam-
somatic mutations in MAPK1, HLA-​B, ERBB2 age (Barton et  al, 1988; Palefsky & Holly, 1995;
and several other genes. Squamous cell carcino- IARC, 2012a). Both nicotine and cotinine have
mas had higher frequency of Tp*C sites than ade- been found in the cervical secretions of smok-
nocarcinomas and gene expression levels at HPV ers and passive smokers (Winkelstein, 1990;
integration sites varied depending on HPV inte- McCann et  al, 1992; Palefsky and Holly, 1995).
gration status (Ojesina et al, 2014). Other somatic Moreover, polycyclic aromatic hydrocarbons
mutations that have been implicated include epi- from cigarette smoke can inhibit cell proliferation
dermal growth factor receptor (EGFR), TP53, of normal cervical cell lines in vitro (Southern &
STK11, KRAS, and PIK3CA genes. Activation of Herrington, 1998). In addition to its known
the latter pathway has been associated with poor mutagenic effects, tobacco smoking may produce
outcomes and decreased survival in cervical local immune suppression that enhances HPV
425

Cervical Cancer 425

persistence (Brinton, 1992; Burger et  al, 1993; women compared to typically nulliparous nuns
IARC, 2012a; zur Hausen, 1982). This immuno- (International Collaboration of Epidemiological
logic effect may be mediated through a reduc- Studies on Cervical Cancer, 2006b). It is a com-
tion in the number of Langerhans cells, which plex challenge in cervical cancer epidemiology to
are important antigen-​presenting cells (Barton disentangle the relationship between HPV, sexual
et al, 1988). activity, hormonal environment, and reproduc-
Although an increased risk of cervical can- tive factors such as pregnancy and parity. Clearly,
cer has been reported among women exposed several or all could be confounded by, or modify
passively to tobacco smoke, an analysis of seven the effect of, HPV infection. A meta-​analysis of
IARC case-​ control studies could not confirm 25 published studies supported a modestly sized
passive smoking as an independent risk factor association (relative risk ~1.8) between multipar-
in absence of active smoking (Louie et al, 2011). ity and invasive cervical cancer, which appears to
Finally, a nonrandomized trial of smoking ces- be independent of HPV infection or other poten-
sation among women with minor-​grade cervical tial reproductive and sexual behavior variables.
lesions showed increased regression rates among A  relation between earlier age at first-​full time
women who stopped or reduced their smoking pregnancy (FFTP) and CIN3/​CIS/​invasive cancer
compared to nonquitters (Szarewski et al, 1996). was also observed (International Collaboration
of Epidemiological Studies on Cervical Cancer,
Diet 2006b). It should be noted that multiparity was
Dietary factors have long been proposed as plau- defined as ≥7 pregnancies compared to 1–​2 such,
sible cofactors in the course of HPV infection and meaning that this association will carry little
cervical neoplasia (Potischman & Brinton, 1996). impact in countries with low average number of
Research in this area has mainly focused on carot- children and higher age at FFTP. But, if the asso-
enoids; retinol; vitamins A, C, and E; and folate. ciation is causal, public health implications may
A  potentially protective association from eating be substantial in countries where multiparity
carrots was reported, based on limited evidence is common and cervical cancer burden is high.
from five case-​ control studies (World Cancer Of interest, a Danish cohort study found a sim-
Research Fund, 2007). Reviews have however ilarly sized association (hazard ratio ~1.8) with
concluded that the epidemiological evidence for risk for CIN3 or worse (CIN3+) comparing par-
an association between diet or nutritional status ous women to nulliparous, but only in women
and cervical cancer is not convincing, largely due who were persistent for HPV for ≥2 years, which
to absence of adjustment for HPV and cervical may further elucidate this relationship (Jensen
screening status, as well as discouraging results et al, 2013).
from existing chemoprevention trials (Garcia-​ Several hypotheses have been suggested to
Closas et al, 2005; World Cancer Research Fund, explain the risk associated with multiparity,
2007). Reports also differ on whether poten- including hormonal influence during pregnancy,
tial dietary effects could be expected at the immunological mechanisms/​immunodepres-
early-​stage—​namely, HPV persistence—​or late-​ sion triggered at pregnancy, or local tissue injury
stage—​namely, invasive cervical cancer—​disease during birth (International Collaboration of
end of the spectrum (Castle & Giuliano, 2003; Epidemiological Studies on Cervical Cancer,
World Cancer Research Fund, 2007; Gonzalez 2006b; Jensen et al, 2013).
et al, 2011). A large prospective cohort study of
300,000 women did show an inverse association Hormones
between the eating of fruits, as well as possibly Similarly to the study of endogenous hormonal
other dietary factors, and invasive cervical can- changes induced by pregnancy, the use of OCs
cer, but again, results should be interpreted with has been of interest since it represents long peri-
caution, since HPV and screening status in par- ods of exposure to exogenous hormones. The use
ticipants was unknown (Gonzalez et al, 2011). of OCs is so closely associated with sexual behav-
ior and, consequently, with HPV infection, that
Reproductive Factors it has been difficult to establish whether these
Parity was proposed as a risk factor for cer- have an independent effect. Nonetheless, a mod-
vical cancer already in the 1840s by Rigoni-​ erate association of long-​term OC use with cer-
Stern, who noted that cervical cancer was more vical cancer was reported in a number of studies
common in married, typically multiparous, that have controlled for potential confounding
426

426 Textbook of Cancer Epidemiology

factors. And as part of collaborative efforts on adenocarcinoma of the cervix (Lacey et al, 2003),
studying cofactors in cervical cancer develop- and cervical cancer mortality is higher in obese
ment, in 2007 the reanalysis of 24 epidemiologi- than nonobese US women (Calle et al, 2003). It
cal studies found a 90% increased risk for current is however consistently observed in the literature
users of more than 5  years, compared to never that cytological screening participation is lower
users. The risk declined after use ceased and by in obese than nonobese women (Maruthur et al,
≥10 years’ nonuse returned to that of never users. 2009). Adequate control of the potential con-
Associations were broadly similar for both in situ founding effect of HPV, sexual behavior, hormo-
and invasive cancers and robust when restrict- nal factors, and screening participation is thus of
ing to only HPV-​positive women. No differences paramount importance and well-​controlled evi-
were identified whether the underlying histol- dence on this risk factor in cervical carcinogen-
ogy was squamous cell or adenocarcinoma, or esis is limited.
whether women had been ever or never cervically
screened. A slight risk association was observed Sexual Behavior
for injectable progestogen-​ only contraceptives As discussed previously, it was recognized early
(International Collaboration of Epidemiological on that sexual behavior played an important
Studies on Cervical Cancer, 2007). role in cervical cancer (Beral, 1974); low rates of
Estrogen-​ progesterone contraceptives are disease were observed in nuns and virgins and
now classified as carcinogens in the human cer- high rates in women who married at young ages
vix by IARC (IARC, 2007b). Proposed mecha- (Boyd & Doll, 1964; Brinton, 1992; Palefsky and
nisms on how estrogen and progesterone could Holly, 1995; Schiffman & Brinton, 1995). Sexual
act on the cervix include hormonally mediated behavior factors in the woman shown to be asso-
changes in immune response to HPV infection, ciated with increased risk for cervical cancer and
stimulation of HPV gene expression and cellu- its precursor lesions include a higher number of
lar proliferation via hormone-​response elements, lifetime sexual partners and an early age at first
and cervical epithelial differentiation and matu- intercourse—​the latter an accepted proxy for age
ration (Harris et al, 2009). at, and likely long-​term duration of, first HPV
Similar to the collaborative study on OCs, a infection (Plummer et al, 2012). Some investiga-
pooled analysis investigated the potential for an tions found that rates of cervical cancer and neo-
association between intrauterine devices (IUDs) plasia were higher in women whose husbands had
and risk for cervical cancer. This was based on cancer of the penis or who were married to men
previous findings that IUDs confer reduced risk whose former wives were diagnosed with cervi-
for endometrial carcinoma. Indeed, a risk reduc- cal cancer. The woman’s male partner’s number of
tion of 50% for cervical cancer was observed in previous and extramarital partners also has been
previous compared to never users of IUDs when associated with the risk of cervical cancer (Bosch
adjusting for HPV infection and previous screen- et al, 1996). This so-​called male factor was among
ing status. It is proposed that IUDs might act in the first clues that cervical cancer likely had an
the progression from HPV to cancer, possibly by infectious etiology (Brinton, 1992).
improving cellular immunity through infliction
of a low-​grade inflammatory response in the cer- Infections
vix (Castellsagué et al, 2011). Ever since the connection between sexual activ-
On a separate note, there is strong evidence ity and cervical carcinoma became evident,
that daughters whose mothers were exposed to researchers have speculated on involvement
the long-​discontinued practice of administering of various sexually transmitted agents (Beral,
the synthetic estrogen compound diethylstilbes- 1974). Before the strong association between
trol during pregnancies in the 1940s–​1970s are HPV and cervical cancer was revealed, herpes
at increased risk for high-​ grade cervical dys- simplex virus type 2 (HSV-​2) was considered a
plasia and clear cell adenocarcinoma (Hoover likely etiologic infectious agent involved in cer-
et al, 2011), a rare histological variant of cervical vical cancer; this hypothesis was dismissed due
cancer. to insufficient supporting data (Brinton, 1992).
And although HSV-​2 has been proposed to act
Anthropometric Measures as a cofactor with HPV in the causation of cervi-
Obesity has been implicated as a risk fac- cal cancer (zur Hausen, 1982), null associations
tor for cervical carcinoma, particularly for with the disease have been reported in several
427

Cervical Cancer 427

prospective studies (Arnheim Dahlström et  al, which more than 40 infect the genital mucosa
2011; Lehtinen et al, 2002). (de Villiers et  al, 2004). A  distinct HPV “type”
A robust link between Chlamydia tracho- is established when the DNA sequence of the L1
matis infection and cervical carcinoma has open reading frame of the cloned viral genome
been reported in several well-​ designed stud- differs from that of any other characterized type
ies; among HPV DNA–​ positive women who by at least 10%. If the genome differs by less
were also C.  trachomatis seropositive, risk was than 10%, isolates of the same HPV type are
increased almost twofold compared to seroneg- referred to as variants, or subtypes. Despite phy-
ative women (Arnheim Dahlström et  al, 2011; logenetic similarity, HPV variants—​especially of
Smith et  al, 2004). It is however still discussed HPV-​16—​are reported to differ in pathogenic-
whether C. trachomatis has an independent effect ity (Schiffman et al, 2010), and there is a need to
on cervical cancer risk, whether it interacts with characterize their association with cancer out-
HPV persistence and/​or progression, or whether comes (Burk et al, 2013).
residual confounding is at play. The suggestion HPVs are highly epitheliotrophic and estab-
that C. trachomatis would act through increased lish productive infections only within the strat-
progression of HPV infection to high-​grade CIN ified epithelia of the skin, the oral cavity, and
was recently contradicted by the null associa- the anogenital area. Infection occurs at the basal
tion between C.  trachomatis and incident CIN2 cell layer and is facilitated by injury or abrasion
or worse (CIN2+) reported when restricting to of the affected tissue (IARC, 2012a). The virus
baseline HPV-​positive women in a large cohort itself replicates productively only in the termi-
(Safaiean et al, 2010). nally differentiated keratinocytes at the surface
Last, in addition to its known deleterious layer of the epithelium. The genital HPVs are
effects on host cell–​mediated immune function further classified into high-​ risk and low-​ risk
and, thus, control of chronic virus infections, types, depending on their oncogenic potential
human immunodeficiency virus (HIV) may (Muñoz et  al, 2003). The “low-​risk” HPV types
directly interact with HPV in coinfected individ- (mainly types 6 and 11) are generally associated
uals, increasing the risk of cervical cancer by up with low-​grade dysplastic lesions and with gen-
to 22-​fold depending on population and screen- ital warts (i.e., condyloma). High-​grade lesions,
ing characteristics (Denny et al, 2012). the true precursors of invasive cervical cancer,
are related to infection with high-​risk HPV types
Human Papilloma Virus Infection where 12 are classified as established carcinogens
Almost forty years have passed since zur Hausen high-​risk (HPV types 16, 18, 31, 33, 35, 39, 45,
first suggested a possible role for HPV in the 51, 52, 56, 58, and 59), one as a probable carcino-
development of squamous cell carcinomas (zur gen (HPV 68), and seven as possible carcinogens
Hausen, 1977). Today, infection with oncogenic (HPV types 26, 53, 66, 67, 70, 73, 82). HPV-​16
types of HPV is generally accepted as the main is the most common HPV type across the spec-
cause of the development of cervical carcinoma trum from normal cytology to cervical cancer
and its immediate precursor lesions (IARC, and exhibits the strongest oncogenic potential in
2012a). This strong epidemiologic evidence has both the cervix and at other human sites (Bosch
gradually emerged through a successful multidis- et al, 2012; Sundström et al, 2010a).
ciplinary approach using molecular epidemiolog- Extensive experimental work has been done
ical methods to study HPV and other potential on the function of the HPV proteins, especially
risk factors during cervical carcinogenesis. with regard to the oncogenic potential of this
virus (IARC, 2012a, Woodman et  al, 2007). Of
Virology most interest are the E6 and E7 proteins, both of
Papilloma viruses are small nonenveloped DNA which have transforming properties. Although
viruses about 55 nm in diameter (IARC, 2007a). the HPV genome exists in episomal form in
Their circular, double-​stranded genome of close benign cervical lesions, HPV DNA detected in
to 8,000 base pairs contains eight known viral cervical cancer cells most often is found to be
genes; the “early” genes (E1, E2, E4, E6, and E7) integrated into the host chromosome. When the
encode regulatory proteins, and the “late” genes virus genome becomes integrated, the E6 and E7
(L1 and L2) code for the structural capsid pro- regions appear to be retained as functional genes.
teins. Currently, there are over 200 different In contrast, the E2 region, which exerts tran-
HPV types described (Bzhalava et  al, 2015), of scriptional control over these two genes, usually
428

428 Textbook of Cancer Epidemiology

is disrupted or lost. Thus, as with hepatitis B virus cell immortalization. Moreover, E7 may further
and hepatocellular carcinoma, integration of induce transformation by binding the AP1 tran-
HPV into the host genome is likely to represent scription factors and thus inhibiting differenti-
an important step in the progression to invasive ation of keratinocytes, as well as by interfering
disease. with other cell cycle regulators such as cyclin
Both E6 and E7 interact with host tumor A and cyclin-​dependent kinases (IARC, 2012a).
suppressor proteins (IARC, 2012a). In vitro, E6
abrogates the function of the p53 tumor sup- Detection
pressor protein, by binding p53 and inducing its A number of methods have been used to detect
degradation through the ubiquitin pathway. p53 HPV DNA in cervical specimens. The Hybrid
plays an important role in cell-​cycle control; it Capture 2 test is a hybridization-​based test. Using
prevents the transition from the checkpoint G1 a liquid-​based system with nonradioactive RNA
phase to the replication S phase, thus allowing probes to detect the DNA of 1 HPV types (the 13
for the repair of damaged DNA or the induction high-​risk HPV types 16, 18, 31, 33, 35, 39, 45, 51,
of apoptosis (Fig. 17-​3). The E7 protein of HPV 52, 56, 58, 59, and 68 plus HPV-​66) this was the
is able to bind and inactivate the Rb tumor sup- first HPV test to be approved by the Food and
pressor protein, which also participates in the Drug Administration (FDA) of the United States.
dynamics of cell cycle regulation. Of note, the E6 It has been used in a large number of scientific
and E7 proteins of high-​risk HPV types can bind studies and is still the most frequently used HPV-​
and degrade their respective tumor suppressors test in the world (Poljak et al, 2012).
more effectively than the same proteins of the The polymerase chain reaction (PCR) detec-
low-​risk HPVs. tion of HPV DNA provided researchers with
Cells infected with high-​risk HPV types do a highly sensitive assay that could identify just
not have a functional G1 checkpoint, which leads a few copies of HPV in a very small amount of
to the suppression of cell-​cycle arrest and apop- specimen. The PCR-​ based test using consen-
tosis and the accumulation of genetic damage sus GP5+/​6+ primers is the second most evalu-
(Fig. 17-​3). The E6 protein also has been shown ated test in the literature. Apart from HC2, other
to induce telomerase activity (IARC, 2012a); the FDA-​approved tests are the Cervista HR HPV
telomerase enzyme elongates chromosomal telo- test, the cobas 4800 test—​both including partial
meres, a process that appears to be necessary for genotyping to identify HPV-​16 and -​18—​and the

binds
p53 HPV E6

Apoptosis DNA repair

RB p
p21

E2F1 pRB
binds
G1 arrest
G1
S
HPV E7

Cell cycle
Cyclin dependent kinases

G2

FIGURE 17-​3  The cell-​cycle regulating mechanisms of HPV oncoproteins E6 and E7.

Source: Burd, 2003.
429

Cervical Cancer 429

cervical mRNA test APTIMA, which detects E6/​ The prevalence of detectable cervical HPV
E7 mRNA products (Poljak et al, 2012). All tar- infection averages around 10.5% worldwide in
get the 12 high-​risk HPV types defined by IARC, women with normal cervical cytology. The preva-
plus 66 and/​or 68. At least 120 other distinct lence ranges from >35% in eastern Africa and the
HPV tests are commercially available (Poljak et Caribbean to less than 5% in studies from north-
al, 2012). This opens opportunities for highly ern America and western Asia, and decreases
competitive purchasing of HPV tests for clinical after age 25 (Bosch et al, 2012; de Sanjosé et al,
use, but also requires strict quality control proce- 2007). The typically decreasing prevalence may
dures and test validations that are traceable to the be explained by decreased transmission during
International Standards of HPV DNA. later years of sexual activity or by the acquisition
As expected, the higher validity and relia- of a protective immune response as a result of
bility of PCR in detecting HPV DNA greatly previous exposure to the virus. In some popula-
increased the strength of the association of HPV tions there is also a second peak in HPV preva-
infection with cervical neoplasia and allowed lence, occurring around 45–​54  years of age. As
for improved epidemiological studies as well this peak is seen only in some populations, it is
as novel opportunities for clinical screening. likely to be attributable to the sexual habits and/​
Sampling methods are also important, and the or period/​cohort effects in these populations (de
introduction of liquid-​based cytology (LBC) has Sanjosé et al, 2007).
been a great advance compared to conventional Most of the epidemiologic evidence for an
cytology in the securing of adequate samples. association between certain HPV types and cer-
With LBC, cervical cells are suspended in a liq- vical cancer first came from case-​control studies
uid medium, and then prepared for cytology (IARC, 2012a). Due to the long latency involved
through centrifugation, rather than being imme- in cervical carcinogenesis as well as to the ethi-
diately physically fixated to a glass slide. This cal imperative to treat early lesions, cohort stud-
has substantially improved visual cytological ies are difficult to perform but have nevertheless
diagnostics and decreased the amount of drying yielded sufficient information to reasonably
artifacts (Strander et al, 2007). LBC also enables describe the natural history of HPV to develop-
reflex testing of both HPV and other biomark- ment of CIN3/​CIS (Moscicki et al, 2012).
ers using the same sample, as well as subsequent In brief, incident HPV infection is common
biobanking of the residual sample for quality among sexually active women. Among HPV-​
assurance as well as research and development infected women, the most important determi-
purposes. Primary HPV screening, as recom- nants of cancer risk are viral genotype—​where
mended in current European guidelines (von HPV type 16 has the highest cancer risk—​and
Karsa et  al, 2015), uses LBC for primary HPV the length of the persistence of the infection.
testing, with reflex cytology of the same sample The small proportion of women that do not
if the HPV test is positive. clear HPV but have a continuously detectable
persistence of HPV have a high risk to progress
Epidemiology to high-​grade CIN. This lesion may then prog-
Sexual transmission is the major route of geni- ress to invasive disease, typically at least 10 or
tal HPV infection. In fact, HPV is considered more years later (Moscicki et  al, 2012)  if it is
the most common sexually transmitted infec- not detected by screening. The time from HPV
tion in the world, with the lifetime risk of being infection to CIN3 is much shorter. Regarding
exposed at least once estimated to be >75% in the time to invasion and progression to CIN3
some countries (Tota et  al, 2011). In concord- there is room for substantial individual varia-
ance, an important study on condom use and risk tion (Moscicki et al, 2012). However, long-​term
of HPV infection confirmed a protective effect; persistence of high-​risk HPV infection without
the incidence rates of both genital HPV infection development of high-​grade lesions appears to be
and cervical intraepithelial lesions were reduced uncommon (Castle et al, 2011).
in condom users compared to nonusers (Winer The cohort studies of HPV and high-​grade
et  al, 2006). Most HPV infections resolve with- cervical lesions are complemented by evidence
out leading to detectable lesions, with the average of HPV and invasive cancer from global tumor-​
duration of new infections being about 6 months typing studies and from prospective nested case-​
and with some 90% of infections clearing within control studies. HPV-​16 and -​18 are the most
a few years (Moscicki et al, 2012). common types and contribute around 70% of
430

430 Textbook of Cancer Epidemiology

all invasive cervical cancers worldwide, a find- interactions between the two infections (Denny
ing that has been robust across all continents. et al, 2012).
Global HPV-​typing efforts of >25,000 cervical Another component of research in the natural
cancer tumors have found that the most com- history of HPV involves the study of the humoral
mon types in cervical cancer worldwide con- immune response to this infection, which also
sistently are HPV types 16, 18, 33, 45, 31, 58, 52, has implications for follow-​up of effects of HPV
and 35, although with some variation in which vaccination. Detection of HPV type-​ specific
order these types place in position 3–​8 (Bosch antibodies is typically performed using L1-​
et al, 2008). In adenocarcinoma, the spectrum of containing virus-​like particles (VLPs) or neutral-
main types appears somewhat narrower, with the ization assays (Palmroth et al, 2010; Porras et al,
largest contributions from HPV types 16, 18, 45, 2010; Castellsagué et al, 2014). Following a genital
59, and 35 (Dahlström et  al, 2010; Castellsagué HPV infection, seroconversion—​that is, detect-
et al, 2006). Risk factors for the development of ability of specific antibodies in serum—​occurs
persistent HPV infection (compared to transient within months in up to 50%–​70% of women and
infection) include older age and infection with the antibodies remain detectable for many years
a high-​risk HPV, particularly HPV-​16 and -​18 (Castellsagué et al, 2014). However, transient cer-
(Castle et al, 2011). vical infections (detectable by HPV DNA PCR,
A population-​based cohort study in Sweden but not persistently detectable) are less likely to
found a strong association of persistent high induce an antibody response (Palmroth et  al,
viral load of HPV-​16 and the development of 2010), probably because they are cleared by an
CIS (Ylitalo et al, 2000a; Ylitalo et al, 2000b). The innate immune response and do not result in an
association between high viral load and invasive adaptive immune response (Moscicki et al, 2012).
squamous cervical cancer has been confirmed, Factors that have been associated with HPV-​16
although the relationship was complex and not seroconversion include abnormal cytology, num-
so strong (Sundström et  al, 2013). High viral ber of sexual partners, OC use, and elevated viral
load associates to persistence and progression of load (Porras et al, 2010). It was initially difficult
HPV-​16 infection (Xi et al, 2011), although this to study whether the antibodies induced by nat-
does not appear to be valid also for other high-​ ural infection provide protection against future
risk types (Gravitt et  al, 2007). Another associ- HPV infection or not (Moscicki et  al, 2012).
ation of note is coinfections between high-​and However, a cohort study found protective trends
low-​risk HPV types, where several reports have against new infections, persistent infections, low-​
found a lower risk for CIN and cancer compared grade cytological abnormalities, and CIN2+ with
to women infected with high-​ risk HPV only increasing HPV-​ 16 antibody titers—​ although
(Luostarinen et  al, 1999; Arnheim Dahlström power was too low to assess other high-​risk HPV
et al, 2011; Sundström et al, 2015). types (Castellsagué et al, 2014).
Cell-​mediated immunity plays an important For many years, seroepidemiological studies
role in the clearance of HPV infection (Dillner have contributed valuable understanding of risks
et al, 2011). Thus, the host immune response, as associated with HPV infection, and numerous
well as the oncogenic potential of the infecting case-​control studies observed consistent positive
HPV, likely contributes to the persistence and associations for cervical cancer, with a 2-​to 40-​
progression of infection. Immune suppression fold relative risk associated with antibody detec-
frequently leads to the emergence of HPV-​related tion (Viscidi et al, 1993; Dillner et al, 1995; IARC,
lesions. Indeed, a high prevalence of HPV and 2007a). The fact that studies using HPV DNA
CIN has been observed among women infected testing typically find higher odds ratios than
with HIV (de Vuyst et  al, 2008). An important studies of HPV antibodies is likely due to the fact
question is whether HIV-​positive women have a that the presence of HPV antibodies represents
higher likelihood of HPV infection due to shared past as well as current infection. Currently, sero-
transmission routes and/​or whether the immu- epidemiology continues to serve an important
nosuppressive effects of HIV enhance the pro- function in HPV research (Arnheim Dahlström
gression of HPV infection to cervical cancer. et  al, 2011; Luostarinen et  al, 2013), and the
Several studies have suggested that HIV is inde- measurement of HPV antibodies has emerged as
pendently associated with HPV acquisition and a key methodology for HPV vaccine research.
morbidity. Taken together, behavioral, genetic, In summary, cervical cancer is a rare con-
and immunologic data suggest a complex web of sequence of an incident genital HPV infection.
431

Cervical Cancer 431

Highly oncogenic HPV types, in

particular type 16
High HPV16 viral load?
Increasing no
Smoking
of. sexual
partners Long-term OC use

Smoking? Increasing age

Coinfections, in particular HIV
Multiparity? Certain HLA types?

Young age? Dietary factors?

HPV HPV HPV Invasive

CIS
negative positive persistence cancer

FIGURE 17-​4 
Established or possible(?) cofactors during different stages of the natural history of HPV infection and
subsequent development of cervical cancer.

Several additional cofactors are known to further are negative by cytology in screening (von Karsa
increase the risk of cervical cancer. These may et al, 2015), due to the high risk for high-​grade
act at different points in the natural history of lesions. Conversely, for women who are HPV-​
HPV and the progression to invasive malignancy negative, substantial evidence indicates their
(Fig. 17-​4). screening interval could be prolonged to at least
5, possibly 10 years, depending on age and previ-
Prevention through Screening ous screening history (Elfström et al, 2014; von
Given the advanced understanding of the role Karsa et al, 2015).
of HPV infection in cervical cancer, this disease The mainstay of cervical screening, whether
has emerged as a uniquely preventable form of using HPV or cytology as the primary screening
cancer. The current recommendations to favor method, will continue to be the detection of CIN
HPV-​ based screening constitute a major new lesions for early treatment before invasion can
advance for population-​based screening with a occur. CIN3/​CIS is clearly a disease with poten-
clinical virology test (EU Guidelines; von Karsa tial to progress to invasive cancer if left untreated,
et  al). Pap smear screening has long been rec- although the extent of such progression has been
ognized as an effective means of reducing cervi- estimated at different levels, ranging chiefly from
cal cancer morbidity and mortality through the the 12% estimated in a Swedish modeling study
detection of preclinical lesions. However, ran- using nationwide data (Gustafsson et al, 1989) to
domized trials have shown that a switch to HPV-​ the 30% observed in a cohort study deriving from
testing as the primary screening method adds New Zealand (McCredie et al, 2008). These vary-
greater protection against invasive cervical and ing estimates are likely explained at least in part
a longer duration of protection for test-​negative by differing study populations and histopatho-
women, compared to cytology (Ronco et al, 2014; logical criteria. Around 50%–​60%, that is, a sub-
Elfström et al, 2014). stantial proportion of CIN3 cases, are observed to
The current evidence favors cytological remain stationary—​that is, will neither clear nor
screening until the age of 30–​35, after which cause invasive disease—​(Wright et al, 2003) and
a switch to primary HPV testing with cytol- the regressive potential is held to be around 20%–​
ogy triage of high-​risk HPV-​positive women is 30% (Mosckicki et  al, 2012). CIN1 may have a
appropriate. Women who are persistently HPV-​ regressive potential of close to 90%, and CIN2
positive in primary screening should be referred somewhere between 30% and 70% depending on
to colposcopy, even in the absence of abnormal age (Moscicki et al, 2012).
cytology. It is imperative to monitor (continue Overtreatment of women with ­cervical dys-
to screen) HPV-​persistent women, even if they plastic lesions, which would have been self-​limited
432

432 Textbook of Cancer Epidemiology

if left untreated, is thus substantial. Given the sig- shown in those up to age 25–​26  years and to
nificant extra cost for this overtreatment, as well some degree also in women up to middle ages
as the potential adverse effect on fertility and (Schiller et al, 2012). There is no effect of vacci-
pregnancy outcome, the magnitude of the prob- nation on ongoing/​prevalent infection or cervi-
lem is not trivial (Kyrgiou et al, 2006). It is key that cal HPV-​related disease. The nonavalent vaccine
HPV-​ based screening programs employ triage is the most recently approved and has so far suc-
methods such as cytology, and that double testing cessfully shown noninferiority to the quadriva-
is not used (i.e., concurrent use of both HPV test- lent vaccine’s clinical and antibody results (Joura
ing and cytology in all women) in order to keep et al, 2015).
the number of colposcopies and biopsies at a rea- By including the two most common HPV
sonable level. Also, the evidence of whether HPV-​ types, HPV-​16 and -​18, the bivalent and quad-
based screening should be performed in ages rivalent vaccines are anticipated to prevent the
25–​34 is unclear. Although there is definitely an over 70% of cervical cancers attributed to these
improved protection against cervical cancer in the two HPVs and the nonavalent vaccine (protect-
age-​group 30–​35 years (Ronco et al, 2014), there is ing against high-​risk HPV types 16, 18, 31, 33,
also a substantial risk of overdetection and subse- 45, 52, and 58)  may raise this figure to 85% to
quent overtreatment of regressive CIN (von Karsa 90% (Joura et  al, 2015). Because these vaccines
et al, 2015). are preventive and thus not intended for treat-
In many low-​income countries, visual inspec- ing infections already present, the primary group
tion of the cervix using acetic acid (VIA) or low-​ targeted for vaccination are young girls (around
cost methods for HPV DNA testing on one or 12  years old) just before they become sexually
two occasions per lifetime may be cost-​effective active. Efficacy studies have not been carried
alternatives in these settings (Goldie et al, 2005, out in this primary target group, who are diffi-
Sankaranarayanan et al, 2009). cult to evaluate specifically in clinical efficacy
trials, due to the long incubation period to dis-
Prevention through Vaccination ease. However, immunogenicity bridging stud-
Much interest and research have been directed ies have established safety and a strong immune
to the development of HPV vaccines, which responses in children, which in its turn have
have emerged as key preventive tools for reduc- allowed extension of vaccine recommendations
ing the global spread of the most oncogenic (Schiller et al, 2012).
HPV types. In 2016, three HPV vaccines were Real-​life population-​based effectiveness stud-
approved for clinical use:  the bivalent vaccine ies have indeed established radical declines in
Cervarix (GlaxoSmithKline), the quadrivalent HPV-​ related disease after the introduction of
vaccine Gardasil (Merck and Co.) and the non- HPV vaccination, so far shown both for infec-
avalent vaccine 9-​ Gardasil (Merck and Co.). tion prevalence, cervical lesions, and condyloma
The vaccines are based on the VLP technology, (Leval et  al, 2013; Ali et  al, 2013; Crowe et  al,
where viral genes encoding HPV surface pro- 2014; Söderlund-​Strand et al, 2013; Tabrizi et al,
teins are used to produce empty virus-​like shells 2014; Chen et al, 2015). Although many countries
that lack any infectious or malignant potential, have launched organized vaccination programs,
but are capable of inducing a potent and lasting access to these in low-​income countries has the
antibody-​mediated immune response. highest need. Currently, the GAVI Alliance is
Both the bivalent and quadrivalent vaccines supporting HPV vaccination in several low-​and
have shown virtually complete protection against middle-​income countries of the world (Herrero
vaccine-​type infection and related CIN3 lesions et al, 2015).
in randomized clinical trials of HPV-​negative Even in resource-​rich countries with imple-
girls and women age 15–​26  years, along with mented programs, controversy remains with
excellent safety profiles. The quadrivalent vac- regard to whom should be vaccinated. A number
cine also confers strong protection against con- of important questions have been raised, includ-
dyloma caused by HPV-​6 and -​11, in previously ing the cost-​benefit ratio of vaccination for adults
unexposed individuals. Even when including already infected with HPV, as well as the length
previously exposed subjects (i.e., seropositive of time that protection will last and the need for
indicating past HPV infection but cervical HPV potential boosters (Sundström et al, 2010b). The
DNA test–​negative, indicating no ongoing infec- issue of targeting males for vaccination against
tion), protective benefits against CIN have been HPV also is under debate. Some reports suggest
43

Cervical Cancer 433

that including male vaccination could accelerate mechanism has been confirmed in humans.
a drastic reduction in HPV prevalence, as well as Studies conducted since then have not strength-
increase the resilience of vaccination programs ened the evidence grade, and, regarding other
in the event of unforeseen reductions in coverage occupations, proposed associations would need
(Elfström et al, 2015). However, although in 2016, well-​designed confirmatory studies before asser-
62 countries have introduced vaccination in girls; tions could be made (IARC, 2014; Weiderpass &
there are only a few countries (notably Australia, Labrèche, 2012).
Austria, Canada, and the United States) that have
also introduced vaccination of boys (Elfström Medical Conditions
et al, 2015; Sankaranarayanan 2015). Female solid organ transplant recipients who are
It should be noted that vaccine programs immunosuppressed may be at increased risk for
will not replace traditional cytological or other cervical cancer, although null associations have
screening programs, because many women will also been reported, perhaps due to recommenda-
have already been exposed to HPV, and current tions for intense screening against cervical pre-
vaccines do not cover all potentially oncogenic cursor lesions in this group (Dugué et al, 2013).
HPV types. Therefore, despite the success of vac- Some reports have further suggested systemic
cination, there is no doubt that cervical screen- inflammatory diseases (SIDs) such as rheuma-
ing will remain as a key tool for the prevention toid arthritis (RA), systemic lupus erythema-
of cervical cancer (Franco et  al, 2006)  although tosus (SLE), and inflammatory bowel disease
screening programs are likely to be modified for as risk factors for HPV infection, cervical dys-
birth cohorts heavily vaccinated against HPV. plasia, and cancer (Feldman & Kim, 2014; Zard
et  al, 2014), perhaps due to immunosuppres-
Physical Activity sion. A  large cohort study using insurance data
Few studies have focused on an association from more than 650,000 women observed a 50%
between physical activity and cervical cancer increased risk of high-​grade cervical dysplasia
(Albanes et al, 1989; Dosemeci et al, 1993; Hirose in women with RA or SLE, but not in those with
et al, 1996), and there is insufficient evidence to psoriasis, compared to women without SID (Kim
draw a conclusion on a direct modifying role et al, 2014).
(World Cancer Research Fund, 2007; Cust, 2011).
CONCLUSION
Ionizing Radiation Persistent infection with an oncogenic HPV type
Little evidence exists for an association between is identified as the main causal factor of cervi-
exposure to radiation and development of cervi- cal cancer. The oncogenic potential of the virus
cal cancer (IARC, 2012b). appears to be mediated through the action of
particular viral proteins that interact with host
Occupation tumor suppressor proteins, thereby suppress-
A number of occupations have been associated ing cell cycle control and apoptosis. A  number
with increased risk for cervical cancer; prima- of cofactors summarized in Table 17-​1 may fur-
rily service occupations (e.g., maids, cleaners, ther increase the risk of cervical cancer, includ-
waitresses, cooks) and manufacturing and tex- ing host factors such as immunity, and lifestyle
tile industry occupations (e.g., machine opera- factors such as tobacco smoking, multiparity, and
tors, textile workers, printers, dry cleaners) as OC use, although, compared to the risk associ-
well as agricultural work. However, most stud- ated with HPV, these contribute modestly sized
ies have been exploratory in nature and lacked effects.
adjustment for main potential confounders such One of the leading causes of cancer among
as HPV infection, smoking habits, and socioec- women worldwide, cervical cancer remains an
onomic status (Weiderpass & Labrèche, 2012). important public health problem, particularly
Several of these occupations involve potential in low-​income countries. Although Pap smear
exposure to hazardous chemicals such as organic screening has been highly effective in decreasing
solvents. In 1995, the IARC found limited evi- cancer incidence and mortality, the development
dence for a carcinogenic effect of tetrachloroeth- of a vaccine against HPV infection probably
ylene, a solvent used in dry cleaning, in textile offers the best means of reducing the burden of
processing, and in degreasing agents, on the cer- this malignancy. Despite the advent of HPV vac-
vix, although no possibly explanatory oncogenic cination, early detection of oncogenic HPV and
43

434 Textbook of Cancer Epidemiology

TABLE 17-​1   EPIDEMIOLOGY OF CERVICAL CANCER:

RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed risk factors

HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 ↑↑
Tobacco smoking ↑
Oral contraceptive use ≥5 years ↑
Multiparity ↑
HIV-​1 infection ↑↑
Diethylstilbestrol ↑↑

Probable Relationship Exists, Based on Substantial Data

HPV types 26, 53, 66, 67, 68, 70, 73, 82 ↑↑
Tetracholoroethylene ↑
Chlamydia trachomatis infection ↑
Intrauterine device use ever ↓
Carrots ↓

Inconsistent Findings or Limited Study to Date

Systemic inflammatory disease (SID) ↑
Solid organ transplantation ↑
Obesity ↑
Fruits ↓
Physical activity −
Ionizing radiation −
Occupation (several proposed) ↑

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑,
moderate to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in
risk; —​, no association

cytological abnormalities within screening pro- REFERENCES

grams will nonetheless continue to be a key ele- Albanes D, Blair A, Taylor PR. Physical activity and
ment in the prevention of cervical cancer. The use risk of cancer in the NHANES I population. Am J
of HPV testing to characterize virus status (i.e., Public Health 1989;79:744–​50.
HPV genotype, persistence of infection) also may Ali H, Donovan B, Wand H, Read TR, Regan
enhance the identification of those women most DG, Grulich AE et  al. Genital warts in young
at risk for progression to cervical cancer. Australians five years into national human papil-
Through a combination of public education lomavirus vaccination programme:  national sur-
about the existence and prevention of HPV, use veillance data. BMJ 2013;346:f2032.
of condoms and benefit of smoking cessation, Arnheim Dahlström L, Andersson K, Luostarinen
regular screening with the incorporation of HPV T, Thoresen S, Ögmundsdottír H, Tryggvadottír
L et  al. Prospective seroepidemiologic study of
testing, and vaccination against high-​risk types
human papillomavirus and other risk factors in
of HPV, carcinoma of the cervix should become
cervical cancer. Cancer Epidemiol Biomarkers
preventable on a global scale. The necessary
Prev 2011;20:2541–​50.
tools for primary and secondary cancer preven- Barton SE, Maddox PH, Jenkins D, Edwards R, Cuzick
tion efforts are now available; the challenge is to J, Singer A. Effect of smoking on cervical epithelial
implement timely and equitable worldwide deliv- immunity:  a mechanism for neoplastic change?
ery of these. Lancet 1988;2:652–​4.
435

Cervical Cancer 435

Beral V. Cancer of the cervix: a sexually transmitted detected infections and cervical abnormalities in
infection? Lancet 1974;1:1037–​40. women seropositive for naturally acquired human
Bergström R, Sparen P, Adami HO. Trends in cancer papillomavirus type 16/​ 18 antibodies:  analy-
of the cervix uteri in Sweden following cytological sis of the control arm of PATRICIA. J Infect Dis
screening. Br J Cancer 1999;81:159–​66. 2014;210:517–​34.
Bosch FX, Castellsague X, Muñoz N, de Sanjosé S, Castle PE, Giuliano AR. Chapter  4:  Genital tract
Ghaffari AM, Gonzalez LC et  al. Male sexual infections, cervical inflammation, and antioxidant
behavior and human papillomavirus DNA:  key nutrients—​assessing their roles as human papil-
risk factors for cervical cancer in Spain. J Natl lomavirus cofactors. J Natl Cancer Inst Monogr
Cancer Inst 1996;88:1060–​7. 2003;30:29–​34.
Bosch FX, Lorincz A, Munoz N, Meijer CJ, Shah Castle PE, Rodríguez AC, Burk RD, Herrero R,
KV. The causal relation between human pap- Wacholder S, Hildesheim A et  al. Long-​ term
illomavirus and cervical cancer. J Clin Pathol persistence of prevalently detected human pap-
2002;55:244–​65. illomavirus infections in the absence of detect-
Bosch FX, Burchell AN, Schiffman M, Giuliano AR, able cervical precancer and cancer. J Infect Dis
de Sanjose S, Bruni L, et al. Epidemiology and nat- 2011;203:814–​22.
ural history of human papillomavirus infections Chen D, Juko-​ Pecirep I, Hammer J, Ivansson E,
and type-​specific implications in cervical neopla- Enroth S, Gustavsson I et al. Genome-​wide asso-
sia. Vaccine 2008;26 Suppl 10:K1–​16. ciation study of susceptibility loci for cervical can-
Bosch FX, Broker TR, Forman D, Moscicki AB, cer. J Natl Cancer Inst 2013;105:624–​33.
Gillison ML, Doorbar J et  al. Comprehensive Chen D, Gyllensten U. Lessons and implications from
control of human papillomavirus infections and association studies and post-​GWAS analyses of
related diseases. Vaccine 2012;30 Suppl 5:H1–​31. cervical cancer. Trends Genet 2015;31:41–​54.
Boyd JT, Doll R. A study of the aetiology of carcinoma Chichareon S, Herrero R, Muñoz N, Bosch FX, Jacobs
of the cervix uteri. Br J Cancer 1964;18:419–​34. MV, Deacon J et al. Risk factors for cervical can-
Brinton LA. Epidemiology of cervical cancer— cer in Thailand: a case-​control study. J Natl Cancer
overview. In: Muñoz N, Bosch FX, Shah KV, Meheus Inst 1998;90:50–​7.
A (eds.) The Epidemiology of Cervical Cancer Crowe E, Pandeya N, Brotherton JM, Dobson AJ,
and Human Papillomavirus. Lyon:  International Kisely S, Lambert SB et al. Effectiveness of quad-
Agency for Research on Cancer, 1992: 3–​23. rivalent human papillomavirus vaccine for the
Burger MPM, Hollema H, Gouw ASH, Pieters WJ, prevention of cervical abnormalities: case-​control
Quint WG. Cigarette smoking and human papil- study nested within a population based screening
lomavirus in patients with reported cervical cyto- programme in Australia. BMJ 2014;348:g1458.
logical abnormality. Br Med J 1993;306:749–​52. Cust AE. Physical activity and gynecologic can-
Burk RD, Harari A, Chen Z. Human papillomavirus cer prevention. Recent results. Cancer Res
genome variants. Virology. 2013;445:232–​43. 2011;186:159–​85.
Bzhalava D, Eklund C, Dillner J. International stand- Dahlström LA, Ylitalo N, Sundström K, Palmgren J,
ardization and classification of human papilloma- Ploner A, Eloranta S et  al. Prospective study of
virus types. Virology 2015;476:341–​4. human papillomavirus and risk of cervical adeno-
Calle EE, Rodriguez C, Walker-​Thurmond K, Thun carcinoma. Int J Cancer 2010;127:1923–​30.
MJ. Overweight, obesity, and mortality from can- Denny LA, Franceschi S, de Sanjosé S, Heard I,
cer in a prospectively studied cohort of U.S. adults. Moscicki AB, Palefsky J. Human papillomavirus,
N Engl J Med 2003;348:1625–​38. human immunodeficiency virus and immunosup-
Castellsagué X, Díaz M, de Sanjosé S, Muñoz N, pression. Vaccine 2012;30 Suppl 5:F168–​74.
Herrero R, Franceschi S, et al. Worldwide human de Sanjosé S, Diaz M, Castellsagué X, Clifford G, Bruni
papillomavirus etiology of cervical adenocarci- L, Muñoz N et al. Worldwide prevalence and gen-
noma and its cofactors: implications for screening otype distribution of cervical human papilloma-
and prevention. J Natl Cancer Inst 2006;98:303–​15. virus DNA in women with normal cytology:  a
Castellsagué X, Díaz M, Vaccarella S, de Sanjosé S, meta-​analysis. Lancet Infect Dis 2007;7:453–​9.
Muñoz N, Herrero R et  al. Intrauterine device de Villiers EM, Fauquet C, Broker TR, Bernard HU,
use, cervical infection with human papillomavi- zur Hausen H. Classification of papillomaviruses.
rus, and risk of cervical cancer:  a pooled analy- Virology 2004;324:17–​27.
sis of 26 epidemiological studies. Lancet Oncol De Vuyst H, Lillo F, Broutet N, Smith JS. HIV, human
2011;12:1023–​31. papillomavirus, and cervical neoplasia and cancer
Castellsagué X, Naud P, Chow SN, Wheeler CM, in the era of highly active antiretroviral therapy.
Germar MJ, Lehtinen M et  al. Risk of newly Eur J Cancer Prev 2008;17:545–​54.
436

436 Textbook of Cancer Epidemiology

Dillner J, Wiklund F, Lenner P, Eklund C, Frederiksson-​ HPV vaccine against HPV Infection and disease
Shanazarian V, Schiller JT et al. Antibodies against in males. N Engl J Med 2011;364:401–​11.
linear and conformational epitopes of human Goldie S, Gaffkin L, Goldhaber-​Fiebert JD, Gordillo-​
papillomavirus type 16 that independently asso- Tobar A, Levin C, Mahe C et al. Cost-​effectiveness
ciate with incident cervical cancer. Int J Cancer of cervical-​
cancer screening in five developing
1995;60:377–​82. countries. N Engl J Med 2005;353:2158–​68.
Dillner J, Lehtinen M, Bjorge T, Luostarinen T, González CA, Travier N, Luján-​Barroso L, Castellsagué
Youngman L, Jellum E et al. Prospective seroepi- X, Bosch FX, Roura E et  al. Dietary factors and
demiologic study of human papillomavirus infec- in situ and invasive cervical cancer risk in the
tion as a risk factor for invasive cervical cancer. J European prospective investigation into cancer
Natl Cancer Inst 1997;89:1293–​9. and nutrition study. Int J Cancer 2011;129:449–​59.
Dillner J, Arbyn M, Unger E, Dillner L. Monitoring Gravitt PE, Kovacic MB, Herrero R, Schiffman M,
of human papillomavirus vaccination. Clin Exp Bratti C, Hildesheim A et al. High load for most
Immunol 2011;163:17–​25. high risk human papillomavirus genotypes is asso-
Dosemeci M, Hayes RB, Vetter R, Hoover RN, Tucker ciated with prevalent cervical cancer precursors
M, Engin K et al. Occupational physical activity, but only HPV16 load predicts the development of
socioeconomic status, and risks of 15 cancer sites incident disease. Int J Cancer 2007;121:2787–​93.
in Turkey. Cancer Causes Control 1993;4:313–​21. Gustafsson L, Pontén J, Zack M, Adami HO.
Dugué PA, Rebolj M, Garred P, Lynge E. International incidence rates of invasive cervical
Immunosuppression and risk of cervical cancer. cancer after introduction of cytological screening.
Expert Rev Anticancer Ther 2013;13:29–​42. Cancer Causes Control 1997;8:755–​63.
Elfström KM, Smelov V, Johansson AL, Eklund C, Harris TG, Miller L, Kulasingam SL, Feng Q, Kiviat NB,
Nauclér P, Arnheim-​ Dahlström L et  al. Long Schwartz SM, et  al. Depot-​medroxyprogesterone
term duration of protective effect for HPV nega- acetate and combined oral contraceptive use and
tive women: follow-​up of primary HPV screening cervical neoplasia among women with oncogenic
randomised controlled trial. BMJ 2014;348:g130. human papillomavirus infection. Am J Obstet
Elfström KM, Lazzarato F, Franceschi S, Dillner J, Gynecol 2009;200:489.e1–​8.
Baussano I. Human papillomavirus vaccina- Helland A, Olsen AO, Gjoen K, Akselsen HE, Sauer
tion of boys and extended catch-​ up vaccina- T, Magnus P et  al. An increased risk of cervical
tion: effects on the resilience of programs. J Infect intra-​
epithelial neoplasia grade II–​ III among
Dis 2016;213:199–​205. human papillomavirus positive patients with
Feldman CH, Kim SC. Should we target patients the HLA-​DQA1*0102-​DQB1*0602 haplo-
with autoimmune diseases for human papillo- type:  a population-​ based case-​control study of
mavirus vaccine uptake? Expert Rev Vaccines Norwegian women. Int J Cancer 1998;76:19–​24.
2014;13:931–​4. Herrero R, González P, Markowitz LE. Present status of
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, human papillomavirus vaccine development and
Mathers C, Rebelo M, Parkin DM, Forman D, Bray, implementation. Lancet Oncol 2015;16:e206–​16.
F. GLOBOCAN 2012 v1.0, Cancer Incidence and Hildesheim A, Wang SS. Host and viral genetics
Mortality Worldwide: IARC CancerBase No. 11 and risk of cervical cancer:  a review. Virus Res
[Internet]. Lyon, France: International Agency for 2002;89:229–​40.
Research on Cancer; 2013. Available from: http:// Hirose K, Tajima K, Hamajima N, Takezaki T, Inoue
globocan.iarc.fr, accessed 2017. M, Kuroishi T et  al. Subsite (cervix/​ endome-
Forman D, de Martel C, Lacey CJ, Soerjomataram I, trium)-​specific risk and protective factors in
Lortet-​Tieulent J, Bruni L et  al. Global burden uterus cancer. Jpn J Cancer Res 1996;87:1001–​9.
of human papillomavirus and related diseases. Hoover RN, Hyer M, Pfeiffer RM, Adam E, Bond B,
Vaccine 2012;30 Suppl 5:F12–​23. Cheville AL et  al. Adverse health outcomes in
Franco EL, Cuzick J, Hildesheim A, de Sanjosé S. women exposed in utero to diethylstilbestrol. N
Chapter  20:  Issues in planning cervical cancer Engl J Med 2011;365:1304–​14.
screening in the era of HPV vaccination. Vaccine IARC. Tobacco Smoke and Involuntary Smoking. IARC
2006;24:171–​7. Monographs on the Evaluation of Carcinogenic
Garcia-​Closas R, Castellsague X, Bosch X, Gonzalez Risks to Humans, Vol. 83. Lyon: IARC, 2004.
CA. The role of diet and nutrition in cervical car- IARC. Human papillomaviruses. IARC Monographs
cinogenesis:  a review of recent evidence. Int J on the Evaluation of Carcinogenic Risks to
Cancer 2005;117:629–​37. Humans, Vol. 90. Lyon: IARC, 2007a.
Giuliano AR, Palefsky JM, Goldstone S, Moreira ED Jr, IARC. Combined Estrogen−Progestogen Contraceptives
Penny ME, Aranda C et al. Efficacy of quadrivalent and Combined Estrogen−Progestogen Menopausal
437

Cervical Cancer 437

Therapy. IARC Monographs on the Evaluation Papillomavirus. Lyon:  International Agency for
of Carcinogenic Risks to Humans, Vol. 91. Lyon: Research on Cancer, 1992: 59–​66.
IARC, 2007b. Konya J, Dillner J. Immunity to oncogenic human pap-
IARC. Biological Agents:  Human papillomavi- illomaviruses. Adv Cancer Res 2001;82:205–​38.
ruses. IARC Monographs on the Evaluation Kyrgiou M, Kolipoulos P, Martin-​Hirsch P, Arbyn
of Carcinogenic Risks to Humans, Vol. 100B. M, Prendiville W, Paraskevaidis E. Obstetric
Lyon: IARC, 2012a. outcomes after conservative treatment for
IARC. Radiation. IARC Monographs on the Evaluation intraepithelial or early invasive cervical
of Carcinogenic Risks to Humans, Vol. 100D. Lyon: lesions:  systematic review and meta-​ analysis.
IARC, 2012b. Lancet 2006;367:489–​98.
IARC. Trichloroethylene, Tetrachloroethylene, and Lacey JV, Swanson CA, Brinton LA, Altekruse SF,
Some Other Chlorinated Agents. IARC Mono­ Barnes WA, Grawitt PE et al. Obesity as a poten-
graphs on the Evaluation of Carcinogenic Risks to tial risk factor for adenocarcinomas and squa-
Humans, Vol. 106. Lyon: IARC, 2014. mous cell carcinomas of the uterine cervix. Cancer
International Collaboration of Epidemiological 2003;98:814–​21.
Studies of Cervical Cancer. Carcinoma of the cer- Lehtinen M, Koskela P, Jellum E, Bloigu A, Anttila
vix and tobacco smoking: Collaborative reanalysis T, Hallmans G et  al. Herpes simplex virus and
of individual data on 13,541 women with carci- risk of cervical cancer:  a longitudinal, nested
noma of the cervix and 23,017 women without case-​control study in the Nordic countries. Am J
carcinoma of the cervix from 23 epidemiological Epidemiol 2002;156:687–​92.
studies. Int J Cancer 2006a;118:1481–​95. Leval A, Herweijer E, Ploner A, Eloranta S, Fridman
International Collaboration of Epidemiological Simard J, Dillner J et  al. Quadrivalent human
Studies of Cervical Cancer. Cervical carcinoma papillomavirus vaccine effectiveness:  a Swedish
and reproductive factors: collaborative reanalysis national cohort study. J Natl Cancer Inst
of individual data on 16,563 women with cervi- 2013;105:469–​74.
cal carcinoma and 33,542 women without cervical Luostarinen T, af Geijersstam V, Bjørge T, Eklund
carcinoma from 25 epidemiological studies. Int J C, Hakama M, Hakulinen T et al. No excess risk
Cancer 2006b;119:1108–​24. of cervical carcinoma among women seroposi-
International Collaboration of Epidemiological tive for both HPV16 and HPV6/​11. Int J Cancer
Studies of Cervical Cancer. Cervical cancer and 1999;80:818–​22.
hormonal contraceptives: collaborative reanalysis Luostarinen T, Namujju PB, Merikukka M, Dillner
of individual data for 16,573 women with cervi- J, Hakulinen T, Koskela P et  al. Order of HPV/​
cal cancer and 35,509 women without cervical chlamydia infections and cervical high-​ grade
cancer from 24 epidemiological studies. Lancet precancer risk:  a case-​cohort study. Int J Cancer
2007;370:1609–​21. 2013;133:1756–​9.
Jensen KE, Schmiedel S, Norrild B, Frederiksen K, Louie KS, Castellsague X, de Sanjose S, Herrero R,
Iftner T, Kjaer SK. Parity as a cofactor for high-​ Meijer CJ, Shah K et  al. Smoking and passive
grade cervical disease among women with per- smoking in cervical cancer risk: pooled analysis of
sistent human papillomavirus infection: a 13-​year couples from the IARC multicentric case-​control
follow-​up. Br J Cancer 2013;108:234–​9. studies. Cancer Epidemiol Biomarkers Prev
Joura EA, Giuliano AR, Iversen OE, Bouchard C, Mao 2011;20:1379–​90.
C, Mehlsen J et al. A 9-​valent HPV vaccine against Magnusson PK, Lichtenstein P, Gyllensten U.
infection and intraepithelial neoplasia in women. Heritability of cervical tumours. Int J Cancer
N Engl J Med 2015;372:711–​23. 2000;88:698–​701.
Kim SC, Glynn RJ, Giovannucci E, Hernández-​ Maruthur NM, Bolen SD, Brancati FL, Clark JM.
Díaz S, Liu J, Feldman S et  al. Risk of high-​ The association of obesity and cervical cancer
grade cervical dysplasia and cervical cancer in screening: a systematic review and meta-​analysis.
women with systemic inflammatory diseases:  a Obesity (Silver Spring) 2009;17:375–​81.
population-​based cohort study. Ann Rheum Dis McCann MF, Irwin DE, Walton LA, Hulka BS, Morton
2015;74:1360–​7. JL, Axelrad CM. Nicotine and cotinine in the cer-
Kiviat NB, Critchcow CW, Kurman RJ. Reassessment vical mucus of smokers, passive smokers, and
of the morphological continuum of cervical nonsmokers. Cancer Epidemiol Biomarkers Prev
intraepithelial lesions:  does it reflect different 1992;1:125–​9.
stages in the progression to cervical carcinoma? McCredie MR, Sharples KJ, Paul C, Baranyai J, Medley
In: Muñoz N, Bosch FX, Shah KV, Meheus A (eds.) G, Jones RW et al. Natural history of cervical neo-
The Epidemiology of Cervical Cancer and Human plasia and risk of invasive cancer in women with
438

438 Textbook of Cancer Epidemiology

cervical intraepithelial neoplasia 3: a retrospective Porras C, Bennett C, Safaeian M, Coseo S, Rodríguez
cohort study. Lancet Oncol 2008;9:425–​34. AC, González P et  al. Determinants of seroposi-
Mitchell MF, Tortolero-​Luna G, Wright T, Sarkar A, tivity among HPV-​ 16/​18 DNA positive young
Richards-​ Kortum R, Hong WK et  al. Cervical women. BMC Infect Dis 2010;10:238.
human papillomavirus infection and intraepithe- Ronco G, Dillner J, Elfström KM, Tunesi S, Snijders
lial neoplasia: a review. J Natl Cancer Inst Monogr PJ, Arbyn M et  al. International HPV screening
1996;21:17–​25. working group. Efficacy of HPV-​based screening
Moscicki AB, Schiffman M, Burchell A, Albero G, for prevention of invasive cervical cancer: follow-​
Giuliano AR, Goodman MT et  al. Updating the up of four European randomised controlled trials.
natural history of human papillomavirus and ano- Lancet 2014;383:524–​32.
genital cancers. Vaccine 2012;30 Suppl 5:F24–​33. Safaeian M, Quint K, Schiffman M, Rodriguez AC,
Mucci LA, Hjelmborg J, Harris J, Czene K, Havelick Wacholder S, Herrero R et al. Chlamydia tracho-
D, Scheike T et  al. Familial risk and heritabil- matis and risk of prevalent and incident cervical
ity of cancer in the Nordic Twin Cohorts. JAMA premalignancy in a population-​based cohort. J
2016;315:68–​76. Natl Cancer Inst 2010;102:1794–​804.
Muñoz N, Bosch FX, de Sanjosé S, Herrero R, Sanjeevi CB, Hjelmstrom P, Hallmans G, Wiklund
Castellsagué X, Shah KV et  al. Epidemiologic F, Lenner P, Angstrom T et  al. Different HLA-​
classification of human papillomavirus types DR-​DQ haplotypes are associated with cervical
associated with cervical cancer. N Engl J Med intraepithelial neoplasia among human papillo-
2003;348:518–​27. mavirus type-​ 16 seropositive and seronegative
Ngelangel C, Muñoz N, Bosch FX, Limson GM, Festin Swedish women. Int J Cancer 1996;68:409–​14.
MR, Deacon J et al. Causes of cervical cancer in Sankaranarayanan R, Nene BM, Shastri SS, Jayant K,
the Philippines: a case-​control study. J Natl Cancer Muwonge R, Budukh AM et  al. HPV screening
Inst 1998;90:43–​9. for cervical cancer in rural India. N Engl J Med
Ojesina AI, Lichtenstein L, Freeman SS, Pedamallu 2009;360:1385–​94.
CS, Imaz-​Rosshandler I, Pugh TJ et al. Landscape Sankaranarayanan R. HPV vaccination:  The most
of genomic alterations in cervical carcinomas. pragmatic cervical cancer primary prevention
Nature 2014;506:371–​5. strategy. Int J Gynaecol Obstet 2015;131:Suppl
Palefsky JM, Holly EA. Molecular virology and epi- 1:S33–​5.
demiology of human papillomavirus and cervi- Schiffman MH, Brinton LA. The epidemi-
cal cancer. Cancer Epidemiol Biomarkers Prev ology of cervical carcinogenesis. Cancer
1995;4:415–​28. 1995;76:Suppl10:1888–​901.
Palmroth J, Namujju P, Simen-​Kapeu A, Kataja V, Schiffman M, Rodriguez AC, Chen Z, Wacholder S,
Surcel HM, Tuppurainen M et  al. Natural sero- Herrero R, Hildesheim A et al. A population-​based
conversion to high-​risk human papillomaviruses prospective study of carcinogenic human papillo-
(hrHPVs) is not protective against related HPV mavirus variant lineages, viral persistence, and
genotypes. Scand J Infect Dis 2010;42:379–​84. cervical neoplasia. Cancer Res 2010;70:3159–​69.
Plummer M, Herrero R, Franceschi S, Meijer CJ, Schiller JT, Castellsagué X, Garland SM. A review of
Snijder P, Bosch FX et  al. Smoking and cervi- clinical trials of human papillomavirus prophy-
cal cancer:  pooled analysis of the IARC multi-​ lactic vaccines. Vaccine 2012;30 Suppl 5:F123–​38.
centric case-​control study. Cancer Causes Control Smith JS, Bosetti C, Muñoz N, Herrero R, Bosch FX,
2003;14:805–​14. Eluf-​Neto J et  al. Chlamydia trachomatis and
Plummer M, Peto J, Franceschi S; International invasive cervical cancer: a pooled analysis of the
Collaboration of Epidemiological Studies of IARC multicentric case-​control study. Int J Cancer
Cervical Cancer. Time since first sexual inter- 2004;111:431–​9.
course and the risk of cervical cancer. Int J Cancer Southern SA, Herrington CS. Molecular events in uter-
2012;130:2638–​44. ine cervical cancer. Sex Transm Inf 1998;74:101–​9.
Poljak M, Cuzick J, Kocjan BJ, Iftner T, Dillner J, Strander B, Andersson-​ Ellström A, Milsom I,
Arbyn M. Nucleic acid tests for the detection of Rådberg T, Ryd W. Liquid-​based cytology versus
alpha human papillomaviruses. Vaccine 2012;30 conventional Papanicolaou smear in an organized
Suppl 5:F100–​6. screening program:  a prospective randomized
Pontén J, Adami HO, Bergstrom R, Dillner J, Friberg study. Cancer 2007;111:285–​91.
LG, Gustafsson L et al. Strategies for global control Sundström K, Eloranta S, Sparén P, Arnheim Dahlström
of cervical cancer. Int J Cancer 1995;60:1–​26. L, Gunnell A, Lindgren A et al. Prospective study
Potischman N, Brinton LA. Nutrition and cervical of human papillomavirus (HPV) types, HPV per-
neoplasia. Cancer Causes Control 1996;7:113–​26. sistence, and risk of squamous cell carcinoma of
439

Cervical Cancer 439

the cervix. Cancer Epidemiol Biomarkers Prev Von Karsa L, Arbyn M, De Vuyst H, Dillner J,
2010a;19:2469–​78. Dillner L, Franceschi S et  al. European guide-
Sundström K, Ploner A, Arnheim-​ Dahlström lines for quality assurance in cervical cancer
L, Eloranta S, Palmgren J, Adami HO et  al. screening. Summary of the supplements on HPV
Interactions between high-​and low-​ risk HPV screening and vaccination. Papillomavirus Res
types reduce the risk of squamous cervical cancer. 2015;1:22–​31.
J Natl Cancer Inst 2015;107(10): pii: djv185. Woodman CB, Collins SI, Young LS. The natural his-
Sundström K, Ploner A, Dahlström LA, Palmgren J, tory of cervical HPV infection: unresolved issues.
Dillner J, Adami HO et  al. Prospective study of Nat Rev Cancer 2007;7:11–​22.
HPV16 viral load and risk of in situ and inva- World Cancer Research Fund/​American Institute for
sive squamous cervical cancer. Cancer Epidemiol Cancer Research. Food, nutrition, physical activ-
Biomarkers Prev 2013;22:150–​8. ity and the prevention of cancer: a global perspec-
Sundström K, Tran TN, Lundholm C, Young C, Sparén tive. Washington DC: AICR, 2007.
P, Dahlström LA. Acceptability of HPV vacci- Wright TC Jr, Cox JT, Massad LS, Carlson J, Twiggs LB,
nation among young adults aged 18–​30  years:  a Wilkinson EJ; American Society for Colposcopy
population based survey in Sweden. Vaccine and Cervical Pathology. 2001 consensus guide-
2010b;28:7492–​500. lines for the management of women with cervi-
Szarewski A, Jarvis MJ, Sasieni P, Anderson M, cal intraepithelial neoplasia. Am J Obstet Gynecol
Edwards R, Steele SJ et al. Effect of smoking cessa- 2003;189:295–​304.
tion on cervical lesion size. Lancet 1996;347:941–​3. Wright AA, Howitt BE, Myers AP, Dahlberg
Söderlund-​Strand A, Uhnoo I, Dillner J. Change in SE, Palescandolo E, Van Hummelen P et  al.
population prevalences of human papilloma- Oncogenic mutations in cervical cancer:  geno-
virus after initiation of vaccination:  the high-​ mic differences between adenocarcinomas and
throughput HPV monitoring study. Cancer squamous cell carcinomas of the cervix. Cancer
Epidemiol Biomarkers Prev 2014;23:2757–​64. 2013;119:3776–​83.
Tornesello ML, Annunziata C, Buonaguro L, Losito S, Xi LF, Hughes JP, Castle PE, Edelstein ZR, Wang C,
Greggi S, Buonaguro FM. TP53 and PIK3CA gene Galloway DA et al. Viral load in the natural his-
mutations in adenocarcinoma, squamous cell car- tory of human papillomavirus type 16 infec-
cinoma and high-​grade intraepithelial neoplasia tion:  a nested case-​ control study. J Infect Dis
of the cervix. J Transl Med 2014;12:255. 2011;203:1425–​33.
Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-​Tieulent J, Ylitalo N, Sørensen P, Josefsson A, Frisch M, Sparen
Jemal A. Global cancer statistics, 2012. CA Cancer P, Pontén J et al. Smoking and oral contraceptives
J Clin 2015;65:87–​108. as risk factors for cervical carcinoma in situ. Int J
Tota JE, Chevarie-​Davis M, Richardson LA, Devries M, Cancer 1999;81:357–​65.
Franco EL. Epidemiology and burden of HPV infec- Ylitalo N, Josefsson A, Melbye M, Sörensen P, Frisch
tion and related diseases:  implications for preven- M, Kragh Andersen P et  al. A prospective study
tion strategies. Prev Med 2011;53 Suppl 1:S12–​21. showing long-​term infection with human papil-
Walboomers JM, Jacobs MV, Manos MM, Bosch FX, lomavirus 26 before the development of cervical
Kummer JA, Shah KV et al. Human papillomavi- carcinoma in situ. Cancer Res 2000a;60:6027–​32.
rus is a necessary cause of invasive cervical cancer Ylitalo N, Sørensen P, Josefsson AM, Magnusson
worldwide. J Pathol 1999;189:12–​9. PKE, Andersen PK, Pontén J et al. Consistent high
Viscidi R, Sun Y, Tsuzaki B, Bosch FX, Muñoz N, Shah viral load of human papillomavirus 16 and risk of
KV. Serologic response in human papillomavirus–​ cervical carcinoma in situ:  a nested case-​control
associated invasive cervical cancer. Int J Cancer study. Lancet 2000b;355:2194–​8.
1993;55:780–​4. Zard E, Arnaud L, Mathian A, Chakhtoura Z, Hie M,
Weiderpass E, Labrèche F. Malignant tumors of the Touraine P et al. Increased risk of high grade cer-
female reproductive system. Saf Health Work vical squamous intraepithelial lesions in systemic
2012;3:166–​80. lupus erythematosus: A meta-​analysis of the liter-
Winer RL, Hughes JP, Feng Q, O’Reilly S, Kiviat NB, ature. Autoimmun Rev 2014;13:730–​5.
Holmes KK et  al. Condom use and the risk of zur Hausen H. Human papillomaviruses and their
genital human papillomavirus infection in young possible role in squamous cell carcinomas. Curr
women. N Engl J Med 2006;354:2645–​54. Top Microbiol 1977;78:1–​30.
Winkelstein W Jr. Smoking and cervical cancer—​current zur Hausen H. Human genital cancer:  synergism
status: a review. Am J Epidemiol 1990;131:945–​57. between two virus infections or synergism
Winkelstein W Jr. Smoking and cancer of the cer- between a virus infection and initiating events?
vix: hypothesis. Am J Epidemiol 1977;10:257–​9. Lancet 1982;2:1370–​2.
40
41

18
Endometrial Cancer
M A R TA C R O U S - ​B O U , I M M A C U L ATA D E V I V O , A N D P A G O N A   L A G I O U

T wo of the most common gynecological can-

cers originate from the uterus:  cervical and
endometrial cancer. These malignancies differ
uterine body). Endometrial adenocarcinomas
are a heterogeneous group of tumors tradition-
ally classified in two major types: endometrioid
dramatically with respect to both etiology and carcinomas (also known as type 1) and their vari-
clinical characteristics. Endometrial cancer ema- ants, which account for 80%–​90% of the tumors;
nates from the corpus uteri, and is the fifth most and nonendometrioid carcinomas (type 2),
common gynecological malignancy worldwide which is much less common (10%) and includes
and the fourth most common cancer among clear cell and serous carcinomas and carcino-
women in high-​income countries (Bray et  al, sarcomas (Tavassoli & Devilee, 2003). The term
2013; Ferlay et al, 2013; Ferlay et al, 2015). When “endometrioid” derives from the tumor’s pre-
diagnosed at advanced stages, which is still com- dominant glandular pattern, which resembles
mon in low-​ income countries, cancers of the proliferative-​phase endometrium. Endometrioid
corpus and cervix uteri can be difficult to distin- tumors include many histologic variants (muci-
guish, and this may distort incidence and mor- nous, secretory, squamous, ciliated, or sex cord-​
tality statistics. like pattern), but it is tumor grade that is most
Lifestyle factors contribute appreciably to important for prognosis (Grades I, II, and III).
the etiology of endometrial cancer, with obesity Nonendometrioid tumors are ungraded, but
accounting for over one-​third of incident endo- histotypes should be diagnosed separately, as
metrial cancer in high-​income societies (Kaaks each has a distinctive natural history (Mutter &
et  al, 2002). Unlike cervical cancer, which is a Prat, 2014).
model of viral carcinogenesis, endometrial can- Symptoms In postmenopausal women—​80% of
cer is considered a model of hormonal carcin- patients—​menstruation-​like bleedings or scanty,
ogenesis, as use of postmenopausal estrogen irregular bleedings are the key early symptoms,
unopposed by progesterone and obesity are the present in 90% of all patients. Premenopausal
best-​ established risk factors (Akhmedkhanov patients typically present with bleedings between
et  al, 2001; Henderson & Feigelson, 2000; menstruations or heavier periodic bleedings. The
Henderson et  al, 1982; Key & Pike, 1988). The fact that they bleed, however, simply means that
effect of estrogens was so substantial that wide- the tumor is large or in advanced stage. Smaller
spread use of hormone therapy for menopausal carcinomas might be asymptomatic. Patients
symptoms increased the incidence of endometrial with advanced disease may complain of pelvic
cancer in the United States to almost epidemic pain, which reflects tumor spread.
proportions in the 1970s. Endometrial cancer is
Diagnosis The diagnosis is established through
also the only known malignancy for which cig-
histopathologic examination of specimens
arette smoking has been convincingly shown to
obtained through endometrial biopsy or curet-
confer protection (Viswanathan et al, 2005).
tage of the endometrium. Imaging techniques
and hysteroscopy are being used more and can
CLINICAL SYNOPSIS
play an important role in diagnosis and staging.
Subgroups Most tumors in the uterine body orig-
inate in the glands of the endometrium, which Treatment Established treatments include sur-
means they are endometrial adenocarcinomas gery, radiation therapy, endocrine therapy, and
(also denoted cancers of the corpus uteri or chemotherapy. A  patient with low-​ risk stage
42

442 Textbook of Cancer Epidemiology

I  cancer would typically receive hysterectomy receptors, markers of proliferation, proteins

and bilateral salpingo-​ oophorectomy, with or related to p53 mutations, oncogene mutations as
without postoperative irradiation; a patient with prognostic factors, and molecular classification
a high-​risk cancer of the same stage would receive of tumors to better define subtypes.
extensive surgery, irradiation, and adjuvant cyto-
static treatment. Stage II patients and those with DESCRIPTIVE
more extensive tumors would get individualized EPIDEMIOLOGY
combinations of surgery, adjuvant chemother- In terms of incidence, endometrial can-
apy, and irradiation. Lymph node dissections are cer is a common gynecological malignancy.
commonly carried out. Patients with the most Approximately 167,859 women in high-​income
advanced disease would receive cytotoxic drugs countries and 151,746 in low-​income countries
or endocrine therapy, the most common hormo- (2012) are diagnosed with the disease annually.
nal treatment being progestational agents. In terms of mortality, endometrial cancer ranks
Prognosis The prognosis in a patient with endo- much lower as a cause of cancer death. In 2012,
metrial carcinoma depends on multiple factors, 34,720 women died from endometrial cancer
including histologic type and grade, depth of in high-​income countries, and 41,440 in low-​
myometrial invasion, vascular channel involve- income countries. In Europe the cumulative
ment, age at diagnosis, and tumor stage. The 5-​ mortality rates have generally been around four
year survival rate is about 70%–​80% overall in times lower than the incidence rates, whereas in
Western countries, ranging from 90% for stage Canada and the United States, they have been
I patients to 15% for stage IV patients (American about seven times lower (Ferlay et al, 2013; Ferlay
Cancer Society, 2015). et al, 2015).
Endometrial cancer accounts for 4.8% of all
Since 70% of endometrial cancer survivors are
cancers in women worldwide (Bray et  al, 2013;
obese, work on quality of life has focused on
Ferlay et  al, 2013). Rates are highest in North
the related issues of physical activity and diet.
America and Europe (Fig. 18-​ 1). In Asian
Physical activity interventions in cancer survi-
countries, they are lower than in other parts of
vors have positive effects on health.
the world, although rising in successive birth
Progress To improve clinical management, a cohorts, especially in Japan. A marked decrease
number of procedures are being evaluated: rad- in incidence was seen in European and North
ical surgery and hormonal treatment, timing American women born around 1925, and some-
and mode of radiation therapy, new virulence time thereafter (Ferlay et  al, 2014; Waterhouse
factors and methods to optimize staging, proto- & Correa, 1992)  this was attributed partly to
cols for individualized treatments, assays of DNA the change from the risk-​increasing estrogen-​
ploidy and of nuclear morphology of hormone only hormone replacement therapy introduced

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America

0 5 10 15 20 25
ASR (world) per 100,000 person-years

FIGURE 18-​1  Age-​standardized (to the 2012 world population) incidence rates of endometrial cancer.
Source: Ferlay et al, 2013.
43

Endometrial Cancer 443

120

New cases per 100,000

100

person-years
80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 18-​2  Age-​specific incidence rates of endometrial cancer.

Source: Ferlay et al, 2013.

in the 1960s to the less harmful use of com- the latter probably reflecting also differences in
bined estrogen-​progestogen regimens, initiated access to optimal care (Cote et al, 2015; Maxwell
about 10  years later, as well as to the protective et al, 2006; Wright et al, 2009).
exposure of a large proportion of premenopau- Survival rates for endometrial cancer differ
sal women to oral contraceptives (Persson et al, by stage:  the 5-​year survival rate ranges from
1990). Data published in the mid-​2000 suggested 90% for stage I to 15% for stage IV (American
that incidence rates among postmenopausal Cancer Society, 2015). Data on mortality show
women were rising again in many, though not all, consistent trends; rates are falling in most
western and northern European countries, cor- countries and age groups (Ferlay et  al, 2014;
relating with increasing obesity rates; a decrease Waterhouse & Correa, 1992). One exception is
in incidence was noted for women younger than Japan, where mortality has doubled since the
55 years of age and it was attributed to combined 1970s, with a significant increase in the young-
oral contraceptive use (Bray et al, 2005). The inci- est age groups. In North America, the decline
dence rates rise steadily 5 to 10 years before men- has been steady since the 1960s. The  overall
opause and peak at ages 65–​70. This has been reduction in mortality suggests that endome-
observed in many countries, albeit at different trial cancers are detected earlier and/​or treated
levels (Fig. 18-​2). more efficiently. Further, risk factors that have
When interpreting incidence and mortality become more prevalent, such as hormone
statistics, changes in characterization (i.e., fewer replacement therapy or obesity, may be asso-
cases included in the subgroup “not otherwise ciated predominantly with biologically less
specified”) and in therapy (i.e., increasing hyster- aggressive and, hence, more curable tumors
ectomy rates) must be considered. (Grady et al, 1995; Weiderpass et al, 1999).
In the US population, the age-​specific inci-
dence rates are higher among whites than among GENETIC AND MOLECULAR
blacks and native Asian women (Devesa et  al, EPIDEMIOLOGY
1995a; Devesa et  al, 1995b; Ferlay et  al, 2014;
Maxwell et  al, 2006; Sherman et  al, 2005). In Genetic Susceptibility
contrast, mortality rates are lower, about 60%, Having a family history of endometrial cancer is
among white women—​ and still lower among an established risk factor. Heritability is the pro-
Asian women (Devesa et al, 1995b; Maxwell et al, portion of total variance in cancer risk that can
2006) compared to black women. Differences in be attributed to genetic variation. The heritabil-
recurrence and disease-​free survival are strongly ity estimates for endometrial cancer have a wide
correlated with genetic admixture, with higher range, between 0% and 52% (Mucci et al, 2016;
proportion of African ancestry related to poorer Lu et al, 2014; Schildkraut et al, 1989). A minority
outcome. Black women are more likely to have of the endometrial cancer cases can be attributed
serous, clear cell, or undifferentiated tumors and to family cancer syndromes, and the remaining
more likely to have more advanced stage disease, majority to nonsyndromic familial inheritance.
4

444 Textbook of Cancer Epidemiology

Endometrial cancer is part of a family of hered- identify rare variants associated with endome-
itary cancers known as hereditary nonpolyposis trial cancer susceptibility (Chen et  al, 2014).
colorectal cancer syndrome (HNPCC), attribut- Recently, a meta-​ analysis of all GWAS stud-
able to the inheritance of rare, highly penetrant, ies, including a total of 4,907 endometrial can-
mutated DNA mismatch repair genes (Aaltonen cer cases and 11,945 controls, was conducted to
et al, 1993; Nicolaides et al, 1994; Papadopoulos identify additional common genetic risk vari-
et  al, 1994; Peltomaki et  al, 1993). HNPCC is ants that are associated with endometrial cancer
characterized by a familial predisposition to (Chen et al, 2016). One new susceptibility locus
colorectal carcinoma and extracolonic cancers; reaching genome-​wide significance (P < 5 × 10–​8)
endometrial cancer is the second most common at 6p22.3 (rs1740828; P = 2.29 × 10–​8, OR = 1.20)
malignancy in this syndrome (Risinger et  al, has been identified, and confirmed previously
1993). Women with this inherited predisposi- discovered loci at 6q22.31 and 13q22.1. The new
tion to endometrial neoplasia tend to develop the locus at 6p22.3, represented by rs1740828, lies
disease 15 years earlier than the general popula- between two genes, SOX4 and CASC15, provid-
tion (Vasen et al, 1994). Mutations in these genes ing evidence of an additional region of interest
are thought to be carried by considerably less for genetic susceptibility to endometrial cancer.
than 1% of the general population, and to date,
extensive studies have not identified common Somatic Alterations
low-​penetrance polymorphisms in these genes as The Cancer Genome Atlas Research Network
associated with endometrial cancer. (TCGA) is a comprehensive and coordinated
Given the small percentage (2%–​ 5%) of effort to accelerate our understanding of the
endometrial cancers attributable to family cancer molecular basis of cancer through the applica-
syndromes, other genetic factors may play a role tion of genome analysis technologies, including
in their development. A population-​based case-​ large-​scale genome sequencing. As reported by
control study has shown that a family history of TCGA (Kandoth et  al, 2013), endometrial car-
endometrial cancer in a first-​degree female relative cinomas display more diversity than captured by
increased the risk by nearly threefold (Gruber & the traditional classification scheme (endome-
Thompson, 1996). A  meta-​ analysis including trioid versus nonendometrioid). After perform-
3,871 endometrial cancer cases and 49,475 con- ing an integrated genomic, transcriptomic, and
trols, concluded that women with a first-​degree proteomic characterization of 373 endometrial
family history of endometrial or colorectal can- carcinomas using array-​and sequencing-​based
cer have a higher risk of developing endometrial technologies, TCGA identified four groups of
cancer than those without a family history:  the endometrial tumors:  (1) an ultramutated group
pooled relative risks (RRs) for endometrial can- caused by mutations in the exonuclease domain
cer associated with a first-​degree family history of the POLE DNA polymerase; (2)  a hypermu-
of endometrial and colorectal cancer were 1.82, tated group characterized by microsatellite insta-
(95% CI 1.65–​1.98) and 1.17, (95% CI 1.03–​1.31), bility and defects in mismatch repair factors;
respectively (Win et al, 2015). (3) a low-​copy-​number group that exhibits mic-
Genome-​ wide association studies (GWAS) rosatellite stability, and (4) a group characterized
are offering an agnostic approach to discover pre- by chromosomal instability and high copy num-
dictive genetic markers. To date, three endome- ber variations. Approximately 25% of high-​grade
trial cancer GWAS have been published (De Vivo endometrioid tumors had extensive copy number
et al, 2014; Long et al, 2012; Spurdle et al, 2011). alterations, few DNA methylation changes, low
The first GWAS reported one genome-​wide sig- estrogen receptor/​progesterone receptor levels,
nificant association, at the 17q21 (HNF1B) locus and frequent TP53 mutations, similar to serous
(Spurdle et al, 2011), which has been reported to tumors. A  subset of women with endometrioid
be also associated with type 2 diabetes and pros- tumors who had a markedly increased transver-
tate cancer susceptibility (Machiela et  al, 2012; sion mutation frequency and newly identified
Zhao et al, 2015). This genetic susceptibility var- hotspot mutations in POLE, had considerably
iant was confirmed in the two other GWAS (De worse outcomes. TCGA also identified several
Vivo et al, 2014; Long et al, 2012); no additional novel oncogenes and tumor suppressors (Kandoth
loci have been identified, probably because even et al, 2013). TCGA results demonstrated that the
larger sample sizes are needed. An exome-​wide genomic features of endometrial carcinomas
association study of endometrial cancer did not require tumor reclassification. Genomic-​ based
45

Endometrial Cancer 445

classification may be useful in developing treat- with excess weight influences both the synthesis
ment strategies to improve patient outcomes. and bioavailability of sex steroids through several
biologic pathways. Adipose tissue expresses sex
RISK FACTORS steroid–​ metabolizing enzymes, which convert
androgenic precursors to estrogens. After men-
Anthropometric opause, adipose tissue becomes the main site of
Measures—​Obesity estrogen synthesis through the aromatization of
Obesity is a well-​established risk factor for endo- adrenal androgens. Greater overall and abdomi-
metrial cancer in both pre-​and postmenopausal nal adiposity after menopause increases the levels
women, and approximately 40% of endome- of insulin and insulin-​like growth factor 1 (IGF-​1),
trial cancer cases are attributed to excess body which reduce hepatic synthesis and blood con-
weight (Bergstrom et al, 2001). High body mass centrations of sex hormone–​ binding globulin
index (BMI) has been consistently reported as (SHBG), and lead to higher levels of free estro-
a risk factor for endometrial cancer (Aune et al, gens (Calle & Kaaks, 2004). Higher BMI is line-
2015). It has been estimated that endometrial arly associated to increased serum concentrations
cancer risk increases by approximately 60% for of estrogens, which in turn increase endometrial
every 5-​kg/​m2 increase in BMI (World Cancer cancer risk (Allen et al, 2008; Kaaks et al, 2002;
Research Fund/​ American Institute for Cancer Kaye et al, 1991; Key et al, 2003; MacDonald et al,
Research [WCRF/​ AICR], 2013; Calle et  al, 1978; Potischman et al, 1996a; Potischman et al,
2003; Lindemann et al, 2008; Reeves et al, 2007; 1996b). Furthermore, in premenopausal women,
Renehan et al, 2008). The association is particu- obesity has been associated with an increase in
larly evident among currently obese postmeno- the frequency of anovulatory (without oppos-
pausal women, with a two-​to threefold excess ing progesterone) menstrual cycles (Key &
risk for women in the highest versus the low- Pike, 1988).
est BMI categories. Aune and colleagues (2015) Other potential carcinogenic consequences
meta-​analyzed 30 prospective studies of BMI and of obesity include its effects on glucose metabo-
endometrial cancer risk with 22,320 cases among lism and the wide range of adipocytokines and
6,445,402 participants. The summary RR for a inflammatory mediators that are produced by
5-​unit increment was 1.54, (95% CI 1.47–​1.61). adipose tissue (Parekh et  al, 2009; Potischman
Although the curve was steeper within the over- et al, 1996a; Soliman et al, 2006; Troisi et al, 1997).
weight and obese BMI ranges, there was evidence
of increased risk even within the higher values in Hormones
the normal BMI range (Aune et  al, 2015). The Most of the risk factors that have been identified
strongest associations have been reported with for endometrial cancer manifest their effects by
adult weight gain and current weight, but BMI influencing the levels of and/​or length of exposure
at young ages (18–​25 years) has also been shown to estrogens. Endometrial cancer is considered a
to increase endometrial cancer risk (Aune et al, disease related, in large part, to excessive expo-
2015). Encouragingly, weight loss has been asso- sure to unopposed estrogens. Exposure to exces-
ciated with risk reduction (Nagle et  al, 2013; sive estrogen, endogenous and/​ or exogenous,
Trentham-​Dietz et al, 2006). Endometrial cancer results in continued stimulation of the endome-
risk has also been positively associated to waist trium. This appears to be the key mechanism in
circumference and hip circumference, as well endometrial carcinogenesis, perhaps a common
as to the waist-​to-​hip ratio (Aune et  al, 2015; denominator for most of the established risk fac-
Swanson et al, 1993). However, the associations tors. Cellular proliferation, decreased apoptosis,
are attenuated after adjustment for BMI (Aune and DNA damage all are mechanisms by which
et al, 2015; Ju et al, 2015). Some studies suggest estrogens may increase endometrial cancer risk.
that, although most strongly associated with the
more common endometrioid cancer subtypes, Endogenous Estrogens
obesity also increases the risk for nonendometri- Convincing evidence indicates that a crucial
oid cancers, albeit to a lesser extend (Bjorge et al, pathway for most endometrial cancers is exces-
2007; Setiawan et al, 2013; Yang et al, 2013). sive exposure to unopposed estrogens. In post-
A causal link between greater adiposity and menopausal women, the main mechanism is
increased endometrial cancer risk is biologically the conversion of androstenedione to estrone.
plausible. Greater amount of fat tissue in women Other mechanisms, such as hyperinsulinemia,
46

446 Textbook of Cancer Epidemiology

have also been suggested. Benign granulosa Since 1975, numerous cohort studies and
and theca cell tumors of the ovary cause exces- case-​control studies have analyzed the associa-
sive production of estrogens in postmenopausal tions between unopposed estrogens and endome-
patients; these women are at a substantially trial cancer risk (IARC, 1999; Grady et al, 1995).
increased risk of endometrial cancer (Bjorkholm Epidemiological studies have established a strong
& Pettersson, 1980). The same is true for young link between long-​term use of replacement estro-
women with polycystic ovary syndrome (Grady gens alone (i.e., without the concomitant use of a
& Ernster, 1996), which involves chronic hyper- progestin, denoted “unopposed” treatment) and
stimulation of the ovary through persistently an excess risk of endometrial cancer. The overall
elevated levels of luteinizing hormone, result- RR for endometrial cancer among women tak-
ing in anovulation and increased production of ing versus those not taking unopposed estrogens
androstenedione, which is converted to estrone was two-​to sixfold (Brinton & Hoover, 1993;
through aromatization. Karageorgi et al, 2010; Persson et al, 1989), rising
To date, only three prospective studies have to over sevenfold when used for 5 years or longer
examined the association between circulat- (Grady et  al, 1995; Karageorgi et  al, 2010). The
ing estrogen levels and endometrial cancer, as increased risk was higher for current or recent
reviewed by Brown and Hankinson (2015). All users, but a twofold increased risk remained for
studies have shown statistically significant pos- several years after cessation of treatment (Pike
itive associations between estradiol and post- et  al, 1997). No clear differences in risk eleva-
menopausal endometrial cancer (Allen et  al, tion were seen between treatment regimens—​
2008; Gunter et al, 2008; Lukanova et al, 2004). intermittent or cyclic versus daily—​or types of
The magnitude of the association comparing estrogen—​ conjugated equine estrogens versus
high versus low hormone levels ranged from estradiol (Grady et al, 1995; Pike et al, 1997).
2.1, (95% CI 1.2–​3.6) to 4.1, (95% CI 1.8–​9.7). In summary, these data demonstrate the defi-
These results further underscore the critically nite increase in endometrial cancer risk for ever-​
important role of estrogens in the etiology of users of estrogens, the possible dose-​response
endometrial cancer. However, the relation- relationship, and the obvious duration-​risk rela-
ship between estradiol levels and endometrial tionship, with the risk declining after cessation of
cancer risk in premenopausal women has not use but remaining for at least 10 years.
been clearly elucidated and requires further
investigation. Postmenopausal Hormone Use: Replacement
with Combined Estrogen-​Progestin Regimens
Epidemiologic data have convincingly shown
Exogenous Hormones that unopposed estrogens, even orally adminis-
Postmenopausal Hormone Use: Replacement tered low-​potency estrogens, cause an excess risk
with Estrogens Alone of endometrial cancer. To lower the risk, proges-
The first evidence that exogenous estrogens may tins were added to hormone therapy to prevent
cause endometrial cancer comes from studies in the endometrium from developing premalignant
the 1970s. They showed that young women with and malignant lesions during estrogen treatment.
inherited gonadal insufficiency who received The use of these opposed regimens has become
synthetic estrogens for replacement (MacMahon, standard. In a cyclic regimen, progestins are usu-
1974)  and breast cancer patients treated with ally added during 10 to 12 days of the monthly
high-​dose estrogens (Hoover et  al, 1976)  devel- cycle, producing withdrawal bleeding. In a con-
oped endometrial cancer more often than tinuous combined regimen, progesterone is
expected. Following a fourfold increase in the added during all days of the cycle; this regimen
prescriptions for estrogen replacement for the is commonly recommended to women who are a
alleviation of perimenopausal and postmeno- few years beyond menopause to avoid bleeding.
pausal symptoms in women during the 1960s Overall, the addition of progestins reduced
and early 1970s in the United States, an epidemic endometrial cancer risk close to baseline pop-
rise in endometrial cancer incidence was noted ulation levels. A  large cohort study with 778
(Kennedy et  al, 1985). Later on, the additional endometrial cancer cases and over 28  years of
administration of progestins reduced endome- follow-​up reported a slightly increased risk in
trial cancer risk close to baseline population ever users of estrogen plus progesterone, com-
levels. pared to never users. The association was seen
47

Endometrial Cancer 447

only among long-​term users (5 years or longer) 36 epidemiological studies, the authors found
(Karageorgi et  al, 2010). Other observational that every 5  years of use is associated with a
studies report little or no association with use of RR of 0.76, (95% CI 0.73–​ 0.78; p < 0.0001)
sequential preparations of at least 10 days of pro- (Collaborative Group on Epidemiological Studies
gesterone per month, and reduced risks among on Endometrial Cancer, 2015). This protective
women who have used continuous combined effect persisted for more than 30 years after cessa-
preparations (Anderson et  al, 2003; Beral et  al, tion of OC use and was independent of OC com-
2005; Brinton & Felix, 2014; Karageorgi et  al, position, although other studies have reported
2010; Persson et al, 1989; Persson et al, 1999). preparations with higher progestogen potency
A protective effect of added progestins is not to be more protective (Mueck et  al, 2010). In
surprising. It is well known that progesterone and high-​income countries, 10 years use of OCs was
synthetic progestins counteract the estrogenic estimated to reduce the absolute risk of endo-
proliferative effect of estrogens by downregulat- metrial cancer arising before age 75  years from
ing estrogen receptor levels, enhancing the met- 2.3 to 1.3 per 100 women (Collaborative Group
abolic inactivation of estradiol, and decreasing on Epidemiological Studies on Endometrial
DNA synthesis (Graham & Clarke, 1997). Cancer, 2015).

Oral Contraceptives Reproductive Factors

In the early 1970s, epidemiologic studies revealed It is well known that reproductive history plays
an increased risk of endometrial cancer in pre- a major role in determining a woman’s risk of
menopausal women who used sequential oral developing endometrial cancer. It is clear from
contraceptives characterized by a high dose of epidemiologic data that the hormonal exposures
synthetic estrogen and then low-​dose progestin associated with childbearing and with proges-
for part of the treatment cycle (Henderson et al. terone throughout the regular menstrual cycles
1983). This adverse effect led to the removal of confer substantial protection against endome-
sequential contraceptive compounds from the trial cancer. Under the hypothesis that estrogens
market. The introduction and use of combined increase the incidence of endometrial cancer,
(estrogen-progestin) oral contraceptives (COCs), all factors reflecting ovarian, placental, or adre-
with formulations containing a high-​ potency nal sex hormone production may influence the
synthetic estrogen (ethinyl estradiol or mestra- risk of endometrial cancer. Thus, aspects of men-
nol) for 3 of the cycle’s 4 weeks, always combined struation and reproduction have been explored
with a potent synthetic progestin, was consist- extensively in numerous epidemiologic studies.
ently found to markedly reduce endometrial can-
cer risk (IARC, 1999). Menstruation, Age at Menarche
Overall, the epidemiologic literature provides and Menopause
convincing evidence that COCs confer substantial A history of regular menstruations has been
and long-​lasting protection against endometrial reported to be more common in endometrial
cancer (Collaborative Group on Epidemiological cancer cases than in controls (Henderson et  al,
Studies on Endometrial Cancer, 2015; Mueck 1983). Many cycles without opposing progester-
et al, 2010). Results on COC use and endometrial one (anovulatory cycles in fertile women or in
cancer suggest that, in high-​income countries, those with irregular periods) or cycles without
about 400,000 cases of endometrial cancer before interrupting pregnancies (nulliparity, late meno-
the age of 75 years have been prevented over the pause) increase endometrial cancer risk (Key &
past 50 years (1965–​2014), including 200,000 in Pike, 1988).
the past decade (2005–​14) (Collaborative Group Risk of endometrial cancer is associated with
on Epidemiological Studies on Endometrial the length of a woman’s reproductive life, such
Cancer, 2015). that it is reduced by 6%–​9% for every year that
Endometrial cancer risk decreases by about menarche is delayed or menopause is advanced
50% with ever use of combined oral contracep- (Dossus et  al, 2010). Endometrial cancer risk
tives versus never use, and the protective effect increases with early menarche:  women who
increases with longer duration of use accord- started menstruating at age 15 or older have an
ing to most studies. In a meta-​analysis includ- approximately 30% decreased risk of endome-
ing 27,276 women with endometrial cancer trial cancer compared with women who started
and 115,743 without endometrial cancer from menstruating at age 11 or younger (Brinton
48

448 Textbook of Cancer Epidemiology

et  al, 1992; Karageorgi et  al, 2010). Some stud- each delivery, or elimination of hyperplasia
ies reported this association to be limited among through atrophy of the endometrium due to high
premenopausal (La Vecchia et al, 1984) or obese levels of progesterone produced by the placenta.
women (Henderson et al, 1983). In most studies,
late age at menopause is a risk factor for endo- Breastfeeding
metrial cancer (Karageorgi et al, 2010; Kvale et al, Although earlier studies did not observe asso-
1988). Women who reach menopause at 55 years ciations between endometrial cancer risk and
or older are at approximately 50% increased risk duration of breastfeeding (Dossus et  al, 2010),
compared to women who reached menopause two meta-​analyses have reported that breastfeed-
between the ages of 45 and 49 years (Karageorgi ing significantly reduced the risk of endometrial
et al, 2010). cancer:  summary RR 0.77, 95% CI 0.62–​0.96.
A  linear dose-​response relationship was found
Parity in both studies, with the risk of endometrial can-
Nulliparity is a strong independent risk factor cer decreasing by 2% per 1-​month increase in the
for endometrial cancer. A 30% increased risk of duration of breastfeeding (Wang et al, 2015; Zhan
endometrial cancer has been observed for nul- et al, 2015).
liparous versus parous women (Brinton et  al,
1992; Karageorgi et  al, 2010; McPherson et  al, Diet
1996; Wu et  al, 2015). The parity-​ dependent Although many studies have investigated the role
reduction of risk is mostly related to the number of dietary factors on endometrial cancer risk,
of births:  among parous women, there is a sig- their results have been inconsistent and no firm
nificant decrease in risk with increasing number conclusions can be drawn about the association
of births (Brinton et  al, 1992; Karageorgi et  al, between diet and endometrial cancer risk.
2010). The protective effects of pregnancy appear The 2013 update on endometrial cancer of
to persist for several decades, although the great- the report from the WCRF/​AICR, concluded that
est reductions in risk are seen for women who there was sufficient evidence to suggest a prob-
were older at the time of their last pregnancy able association of endometrial cancer risk with
(Setiawan et al, 2012). glycemic load (positive) and coffee consumption
Infertility, defined as unsuccessful attempts (inverse) (WCRF/​AICR, 2013).
to become pregnant during a 3-​year period, or A “nutrient-​wide association study” approach
seeking medical advice because of inability to has been applied to systematically evaluate die-
become pregnant, has been associated with an up tary risk associations for endometrial cancer
to threefold increased risk of endometrial cancer (Merritt et  al, 2015). Authors evaluated endo-
(Brinton et al, 1992; Henderson et al, 1983), par- metrial cancer risk associations with dietary
ticularly when it is linked to anovulation and pro- intake of 84 foods and nutrients based on dietary
gesterone insufficiency (Escobedo et al, 1991). questionnaires in a total of 2,834 endometrial
cancer cases from three prospective studies, the
Age at First and Last Birth European Prospective Investigation into Cancer
Evidence suggests that age at first birth appears not and nutrition (EPIC; n = 1,303 cases) followed by
to be related to endometrial cancer risk (Brinton validation in the Nurses’ Health Studies (NHS/​
et al, 1992; Karageorgi et al, 2010). Instead, stud- NHSII; n = 1,531 cases). In multivariate adjusted
ies report that pregnancy at a relatively advanced comparisons, only coffee intake was inversely
age is linked to a lower risk:  women who have associated with endometrial cancer risk in both
their last child at 40  years or older have a 44% discovery and validation steps (EPIC, HR: 0.81,
decreased risk of endometrial cancer compared 95% CI 0.68–​ 0.97; NHS/​ NHSII, HR:  0.82,
to women who had their last birth under the age 95% CI:  0.70–​ 0.96). The inverse association
of 25 years (Setiawan et al, 2012). In a pooled anal- between coffee intake and endometrial cancer
ysis of 8,671 endometrial cancer cases and 16,562 risk is consistent with the results of two meta-​
controls in the Epidemiology of Endometrial analyses(WCRF/​AICR, 2013; Je & Giovannucci,
Cancer Consortium, the RR for endometrial 2012). Evidence for red meat and nonstarchy veg-
cancer per 5-​year increase in age at last birth was etables is less consistent to suggest an association
0.87, (95% CI 0.85–​0.90) (Setiawan et al, 2012). of either exposure with endometrial cancer risk.
The proposed mechanism is both shedding and Consumption of red meat was classified as
removal of malignant or premalignant cells after “probably carcinogenic to humans” (Group 2A),
49

Endometrial Cancer 449

while consumption of processed meat as “carcin- fact that smokers tend to have lower circulating
ogenic to humans” (Group 1). These associations estrogen levels and earlier menopause than non-
were observed mainly for colorectal, and less so smokers (Tanko & Christiansen, 2004).
for pancreatic and prostate cancer (Bouvard et al, An inverse association between cigarette
2015), whereas evidence for endometrial cancer smoking and endometrial cancer risk was first
was again inconsistent. reported by Weiss and colleagues (1980). Since
then, several case-​ control and cohort studies
Alcohol have evaluated the association. In general, pro-
Although most studies report a nonsignificant spective studies show at least a 20% reduced risk
decrease in endometrial cancer risk with alcohol for current smokers and a weak reduction or no
consumption—​ especially light consumption (Je reduction in risk for former smokers (Terry et al,
et al, 2014)—​there is also some evidence suggest- 2002b; Zhou et al, 2008). It has been reported that
ing that alcohol intake may increase the risk of the “protective effect” of smoking occurs early
endometrial cancer (Fedirko et al, 2013). A study in adulthood and is long-​lasting (Viswanathan
of 37,000 alcoholics showed a 24% reduction in et al, 2005). In studies stratifying by menopausal
endometrial cancer risk compared to the general status, the effect is stronger in postmenopausal
population (Weiderpass et al, 2001b). These results women, while the association is less consistent
are contrary to what would be expected, since among premenopausal women (Al-​ Zoughool
alcohol is associated with increased estrogen levels et al, 2007; Zhou et al, 2008). Some studies that
(Hankinson et al, 1995; Reichman et al, 1993). analyzed intensity (cigarettes/​day) or duration
(years of smoking) have shown greater endome-
Physical Activity trial cancer risk reduction among current smok-
There are indications that persistent, vigorous ers; however, no dose effect was observed among
physical exercise may reduce endometrial can- former smokers (Terry et  al, 2002a; Terry et  al,
cer risk. The hypothesis is that it reduces the risk 2002b; Viswanathan et al, 2005).
by acting indirectly, through other risk factors
such as BMI, age at menarche or menopause, or Medical Conditions
prevalence of anovulation, or directly by lower- Diabetes
ing the levels of ovarian hormones. Evidence Most studies have identified diabetes mellitus
that supports this hypothesis suggests that phys- and abnormal glucose tolerance as risk factors
ically active women have lower levels of circu- for endometrial cancer (Friberg et al, 2007; Grady
lating estrogens compared to sedentary women & Ernster, 1996; La Vecchia et al, 1994). Women
(Sturgeon et al, 1993). with type 2 diabetes have an approximately two-
A number of prospective studies have now fold increased risk of developing endometrial
reported that increased physical activity is asso- cancer. The mechanism was assumed to be obe-
ciated with a decreased risk of endometrial can- sity, however, the specific magnitude of the asso-
cer independent of obesity, with estimated risk ciation has been difficult to establish because of
reductions of 20%–​40% comparing active versus the high frequency of coexistence of diabetes with
inactive women (Du et  al, 2014; Moradi et  al, obesity. Improved quantitative markers (such as
1998; Moradi et  al, 2000). The 2013 update on serum adiponectin, a surrogate marker for insu-
endometrial cancer of the WCRF/​AICR report lin resistance) and better-​designed studies have
found that there was sufficient evidence to con- confirmed a positive and independent associa-
clude that physical activity of all types (including tion between diabetes and endometrial cancer,
occupational, household, transport, and recrea- even after adjusting for BMI (Brinton et al, 1992;
tional) probably does protect against endome- Cust et  al, 2007; Friberg et  al, 2007; Schmandt
trial cancer (WCRF/​AICR, 2013). et al, 2011). These observations led to the hypoth-
esis that the increased risk in women with non-​
Smoking insulin-​dependent diabetes mellitus is actually
The preponderance of evidence consistently sug- caused by the increase in the levels of insulin.
gests that cigarette smoking reduces the risk of
endometrial cancer; the greatest risk reductions Hypertension
have been observed in postmenopausal women Hypertension has been reported to be a risk fac-
who are current smokers. This somewhat sur- tor for endometrial cancer (Grady & Ernster,
prising observation has been attributed to the 1996; MacMahon, 1974). Confounding by other
450

450 Textbook of Cancer Epidemiology

risk factors, such as estrogen use or high BMI, affecting cancer risk. Large studies, however,
cannot be excluded, since most studies show reveal no associations of blood levels of orga-
that after adjustment for these and other poten- nochlorine compounds with endometrial can-
tial confounding factors, there is no increase in cer risk (Adami et al, 1995; Sturgeon et al, 1998;
risk (Brinton et al, 1992). In some women, hyper- Weiderpass et al, 2000).
tension may be part of a metabolic syndrome
involving hyperinsulinemia and other endocrine Tamoxifen
abnormalities (Kaaks, 1996). Tamoxifen is a nonsteroidal compound with
both a partially estrogenic and an antiestrogenic
Polycystic Ovary Syndrome effect, depending on the target tissues. Tamoxifen
Women with polycystic ovary syndrome have has been widely used as neoadjuvant therapy
an approximately 2.5fold increased risk of devel- for women with breast cancer. Extensive clini-
oping endometrial cancer, even after adjusting cal trials have established its efficacy as adjuvant
for BMI (Chittenden et  al, 2009; Fearnley et  al, therapy, both in reducing the risk of recurrence
2010; Park et al, 2011). A history of fibroids also or contralateral breast cancer and in improving
increases a woman’s risk of developing endome- survival in postmenopausal women with early-​
trial cancer, possibly because both conditions stage, receptor-​ positive tumors (Henderson
are associated with increased estrogen exposure et al, 1988).
(Rowlands et al, 2011). A causal association with endometrial cancer
is plausible due to tamoxifen’s potential to stimu-
Endometriosis late tumor growth, in part because of its partially
Literature suggests that history of endometri- agonistic estrogenic property and its ability to
osis does not increase a woman’s risk of endo- form DNA adducts (Rutqvist, 1998). In women
metrial cancer (Kvaskoff et  al, 2015; Rowlands of childbearing age it antagonizes endogenous
et al, 2011). estrogens and induces endometrial inactivity
and atrophy, but in postmenopausal women,
Other Risk Factors who are normally hypoestrogenic, it may have
Infections a weak estrogenic effect (Mutter & Prat, 2014).
A possible role for oncogenic human papilloma Tamoxifen administration is associated with an
viruses (HPV), an established cause of cervi- overall two-​to threefold increased risk of endo-
cal cancer (Leminen et  al, 1991), has been sug- metrial adenocarcinoma (Henderson et al, 1988;
gested in endometrial cancer etiology, but data Cohen, 2004; Rutqvist et  al, 1995). Because of
are sparse and controversial and HPV appears to this association, the IARC has classified tamox-
be unrelated to the endometrial neoplastic trans- ifen as a carcinogen (IARC, 1996). The adverse
formation process (Brewster et al, 1999). effects of tamoxifen need to be considered for
chemoprevention of breast cancer in healthy or
Occupation and Socioeconomic Status high-​risk women with an intact uterus.
Results of a large Finnish study using census
data, showed no association between job titles Risk Prediction Models
and endometrial cancer risk (Weiderpass et  al, The goal of a risk prediction model is to com-
2001a). However, women with sedentary jobs bine multiple risk factors into a single, personal-
and women exposed to animal dust (i.e., furriers) ized predictive risk estimate of developing some
had a significantly increased risk of endometrial condition in a defined period of time based on
cancer risk (Weiderpass et al, 2001a). As for other an individual’s values for the different risk fac-
possible correlates of socioeconomic status, level tors (Gail et  al, 1989). Risk prediction models
of education has been inversely associated to have been successfully applied in clinical practice
endometrial cancer risk (La Vecchia et al, 1992; to estimate an individual’s risk of a chronic dis-
Mouw et al, 2008). ease. Such models have helped guide clinicians
in making prevention and treatment decisions.
Environmental Xenoestrogens Additionally, cancer risk prediction models have
It has been argued that persistent environmen- been useful in identifying high-​risk individuals
tal pollutants, such as dichlorodiphenyltrichlo- for intervention and screening trials.
roethane (DDT) and polychlorinated biphenyls Two risk prediction models have been devel-
(PCBs), may exert hormonal effects, thereby oped for endometrial cancer (Husing et al, 2016;
451

Endometrial Cancer 451

TABLE 18-​1   EPIDEMIOLOGY OF ENDOMETRIAL CANCER:

RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Family history in first-​degree relative ↑
High body mass index (postmenopausal) ↑↑
High body mass index (premenopausal) ↑
High endogenous estrogen levels ↑↑
Postmenopausal hormone use (estrogen-​only) ↑↑
Oral contraceptive use (estrogen-​progestin combined) ↓↓
Nulliparity ↑
Age at last birth ↓
Infertility ↑
Cigarette smoking ↓
Diabetes ↑
Polycystic ovary syndrome ↑
Tamoxifen ↑

Probable Relationship Exists, Based on Substantial Data

Early menarche ↑
Late menopause ↑
Physical activity ↓
High glycemic load ↑
Coffee consumption ↓
Hypertension ↑

Weak, if Any, Relationship Exists, Based on Substantial Data

Postmenopausal hormone use (estrogen-​progestin combined) —​, ↑**
Age at first birth —​
Vegetables consumption —​
Endometriosis —​

Inconsistent Findings or Limited Study to Date

Breastfeeding (long duration) ↓
Red meat consumption —​
Alcohol consumption ↑
Organochlorine exposure —​

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑,
moderate to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in
risk; —​, no association
**
A weak association has been reported only among long-​term users (5 years or longer).

Pfeiffer et  al, 2013). Both have been based on a The first model (Pfeiffer et  al, 2013)  was
combination of reproductive risk factors such as developed in a combined dataset from two
parity and age at menopause, as well as environ- United States cohort studies (NIH-​AARP and
mental risk factors such as smoking. However, nei- PLCO) consisting of 146,679 non-​ Hispanic
ther model incorporated any genetic risk factors. white American women, including a total of
452

452 Textbook of Cancer Epidemiology

1,559 cases of endometrial cancer. The model was Moreover, risk prediction models should take
externally validated in a third cohort, the Nurses’ into account differences across all ethnic groups.
Health Study. The risk factors included were Pooling data across multiple studies within large
BMI, hormone therapy use, parity, premenopau- international consortia will greatly improve our
sal status, age at menopause, smoking status, oral ability to evaluate risk factors by subtype. Larger
contraceptive use, and a BMI–​hormone therapy sample sizes will help clarify patterns and mag-
interaction term. Husing et al (2016) developed nitudes of risk associated with less well under-
a model based on 201,811 European women, stood or less common exposures, and will allow
including 855 cases of endometrial cancer, from a more robust evaluation of gene–​environment
the European Prospective Investigation into interactions.
Cancer and Nutrition (EPIC) cohort. This model
was not externally validated but instead was vali- REFERENCES
dated internally using fivefold cross-​validation. Aaltonen LA, Peltomaki P, Leach FS, Sistonen P,
In addition to the risk factors included in the Pylkkanen L, Mecklin JP et al. Clues to the path-
Pfeiffer model, Husing’s model included age at ogenesis of familial colorectal cancer. Science
menarche and age at first full-​term pregnancy. 1993;260:812–​6.
The Husing model had better statistical discrim- Adami HO, Lipworth L, Titus-​Ernstoff L, Hsieh CC,
ination (AUC  =  0.77) than the Pfeiffer model Hanberg A, Ahlborg U et al. Organochlorine com-
(AUC = 0.68), and both models have better dis- pounds and estrogen-​related cancers in women.
Cancer Causes Control 1995;6:551–​66.
crimination than many risk prediction models
Akhmedkhanov A, Zeleniuch-​Jacquotte A, Toniolo
developed for other tumor sites.
P. Role of exogenous and endogenous hormones
in endometrial cancer:  review of the evidence
CONCLUSION and research perspectives. Ann N Y Acad Sci
In Western countries, endometrial cancer is a 2001;943:296–​315.
common malignancy that occurs mostly after the Allen NE, Key TJ, Dossus L, Rinaldi S, Cust A,
menopause. The variation in age-​standardized Lukanova A et  al. Endogenous sex hormones
incidence across countries suggests the strong and endometrial cancer risk in women in the
influence of modifiable risk factors (Table European Prospective Investigation into Cancer
18-​1). Among these, estrogens emerge as cen- and Nutrition (EPIC). Endocr Relat Cancer
tral in malignant transformation of the endo- 2008;15:485–​497.
metrium. Therefore, it is conceivable that risk Al-​Zoughool M, Dossus L, Kaaks R, Clavel-​Chapelon
reduction conferred by current smoking, past F, Tjonneland A, Olsen A et al. Risk of endome-
OC use, childbearing, and physical activity is trial cancer in relationship to cigarette smok-
mediated by hormones. This may also apply to ing:  results from the EPIC study. Int J Cancer
the increase in risk associated with obesity, which 2007;121:2741–​7.
increases peripheral production of estrogens, and American Cancer Society. What Are the Key Statistics
with diabetes mellitus. Hence, it should be possi- for Endometrial Cancer? 2015. www.cancer.org.
ble to prevent a substantial fraction of all endo- Accessed November 2015.
metrial cancers through lifestyle modification, Anderson GL, Judd HL, Kaunitz AM, Barad DH,
maintaining normal weight, and optimal use of Beresford SA, Pettinger M et  al. Effects of estro-
OCs and postmenopausal hormones. gen plus progestin on gynecologic cancers and
associated diagnostic procedures:  the Women’s
Recent research has shown that cancer eti-
Health Initiative randomized trial. JAMA
ology may differ across histological subtypes.
2003;290:1739–​48.
Pathological classification of endometrial cancer is
Aune D, Navarro Rosenblatt DA, Chan DS, Vingeliene
currently evolving and molecular studies are reveal-
S, Abar L, Vieira AR et al. Anthropometric factors
ing different molecular subtypes within the same and endometrial cancer risk: a systematic review
histological groups. Future studies should take this and dose-​response meta-​analysis of prospective
heterogeneity into account to better understand studies. Ann Oncol 2015; 26:1635–​48.
molecular subtypes of endometrial cancer. Beral V, Bull D, Reeves G. Endometrial cancer and
Given the push toward personalized med- hormone-​ replacement therapy in the Million
icine, investment in accurate risk prediction Women Study. Lancet 2005;365:1543–​51.
models that includes refining and adding data Bergstrom A, Pisani P, Tenet V, Wolk A, Adami HO.
on exposures, modeling by molecular subtypes Overweight as an avoidable cause of cancer in
with validation in other cohorts, is warranted. Europe. Int J Cancer 2001;91:421–​30.
453

Endometrial Cancer 453

Bjorge T, Engeland A, Tretli S, Weiderpass E. Body size in Chen MM, O’Mara TA, Thompson DJ, Painter JN;
relation to cancer of the uterine corpus in 1 million Attia J, Black A, et  al. GWAS meta-​analysis of
Norwegian women. Int J Cancer 2007;120:378–​83. 16 852 women identifies new susceptibility
Bjorkholm E, Pettersson F. Granulosa-​cell and theca-​ locus for endometrial cancer. Hum Mol Genet
cell tumors. The clinical picture and long-​term 2016;15;25(12):2612–​20.
outcome for the Radiumhemmet series. Acta Chittenden BG, Fullerton G, Maheshwari A,
Obstet Gynecol Scand 1980;59:361–​5. Bhattacharya S. Polycystic ovary syndrome and
Bouvard V, Loomis D, Guyton KZ, Grosse Y, Ghissassi the risk of gynaecological cancer:  a systematic
FE, Benbrahim-​Tallaa L et al. Carcinogenicity of review. Repr Biomed Online 2009;19:398–​405.
consumption of red and processed meat. Lancet Cohen I. Endometrial pathologies associated with
Oncol 2015;16:1599–​600. postmenopausal tamoxifen treatment. Gynecol
Bray F, Dos Santos Silva I, Moller H, Weiderpass Oncol 2004;94:256–​66.
E. Endometrial cancer incidence trends in Collaborative Group on Epidemiological Studies on
Europe:  underlying determinants and prospects Endometrial Cancer. Endometrial cancer and
for prevention. Cancer Epidemiol Biomarkers oral contraceptives:  an individual participant
Prev 2005;14:1132–​42. meta-​analysis of 27 276 women with endometrial
Bray F, Ren JS, Masuyer E, Ferlay J. Global estimates of cancer from 36 epidemiological studies. Lancet
cancer prevalence for 27 sites in the adult popula- Oncology 2015;16:1061–​70.
tion in 2008. Int J Cancer 2013;132:1133–​45. Cote ML, Alhajj T, Ruterbusch JJ, Bernstein L, Brinton
Brewster WR, Monk BJ, Burger RA, Bergen S, LA, Blot WJ et al. Risk factors for endometrial cancer
Wilczynski SP. Does human papillomavirus have in black and white women: a pooled analysis from the
a role in cancers of the uterine corpus? Gynecol Epidemiology of Endometrial Cancer Consortium
Oncol 1999;75: 51–​4. (E2C2). Cancer Causes Control 2015;26:287–​96.
Brinton LA, Berman ML, Mortel R, Twiggs LB, Barrett Cust AE, Kaaks R, Friedenreich C, Bonnet F, Laville
RJ, Wilbanks GD et al. Reproductive, menstrual, M, Lukanova A et  al. Plasma adiponectin levels
and medical risk factors for endometrial can- and endometrial cancer risk in pre-​and post-
cer: results from a case-​control study. Am J Obstet menopausal women. J Clin Endocrinol Metabol
Gynecol 1992;167:1317–​25. 2007;92:255–​63.
Brinton LA, Felix AS. Menopausal hormone therapy De Vivo I, Prescott J, Setiawan VW, Olson SH,
and risk of endometrial cancer. J Steroid Biochem Wentzensen N, Attia J et al. Genome-​wide associ-
Mol Biol 2014;142:83–​9. ation study of endometrial cancer in E2C2. Hum
Brinton LA, Hoover RN. Estrogen replacement ther- Gen 2014;133: 211–​24.
apy and endometrial cancer risk:  unresolved Devesa SS, Blot WJ, Stone BJ, Miller BA, Tarone RE,
issues. The Endometrial Cancer Collaborative Fraumeni JF Jr. Recent cancer trends in the United
Group. Obstet Gynecol 1993;81:265–​71. States. J Natl Cancer Inst 1995a;87:175–​82.
Brown SB, Hankinson SE. Endogenous estrogens and Devesa SS, Donaldson J, Fears T. Graphical pres-
the risk of breast, endometrial, and ovarian can- entation of trends in rates. Am J Epidemiol
cers. Steroids 2015;99:8–​10. 1995b;141:300–​4.
Calle EE, Kaaks R. Overweight, obesity and can- Dossus L, Allen N, Kaaks R, Bakken K, Lund E,
cer:  epidemiological evidence and proposed Tjonneland A et al. Reproductive risk factors and
mechanisms. Nat Rev Cancer 2004;4:579–​91. endometrial cancer:  the European Prospective
Calle EE, Rodriguez C, Walker-​Thurmond K, Thun Investigation into Cancer and Nutrition. Int J
MJ. Overweight, obesity, and mortality from can- Cancer 2010;127:442–​51.
cer in a prospectively studied cohort of U.S. adults. Du M, Kraft P, Eliassen AH, Giovannucci E, Hankinson
New Engl J Med 2003;348:1625–​38. SE, De Vivo I. Physical activity and risk of endo-
IARC. Some Pharmaceutical Drugs. Monographs on metrial adenocarcinoma in the Nurses’ Health
the Evaluation of Carcinogenic Risk to Humans, Study. Int J Cancer 2014;134:2707–​16.
Vol. 66, Lyon: IARC, 1996. Escobedo LG, Lee NC, Peterson HB, Wingo PA.
IARC. Postmenopausal Hormone Therapy and Infertility-​
associated endometrial cancer risk
Hormonal Contraception. IARC Monographs on may be limited to specific subgroups of infertile
the Evaluation of Carcinogenic Risk to Humans, women. Obstet Gynecol 1991;77:124–​8.
Vol 72. Lyon: IARC, 1999. Fearnley EJ, Marquart L, Spurdle AB, Weinstein P,
Chen MM, Crous-​Bou M, Setiawan VW, Prescott J, Webb PM. Polycystic ovary syndrome increases
Olson SH, Wentzensen N et al. Exome-​wide asso- the risk of endometrial cancer in women aged less
ciation study of endometrial cancer in a multieth- than 50  years:  an Australian case-​control study.
nic population. PLoS One 2014;9:e97045. Cancer Causes Control 2010;21:2303–​8.
45

454 Textbook of Cancer Epidemiology

Fedirko V, Jenab M, Rinaldi S, Biessy C, Allen NE, endometrial cancer in young women. Br J Cancer
Dossus L et  al. Alcohol drinking and endome- 1983;47:749–​56.
trial cancer risk in the European Prospective Henderson BE, Feigelson HS. Hormonal carcinogene-
Investigation into Cancer and Nutrition (EPIC) sis. Carcinogenesis 2000;21:427–​33.
study. Ann Epidemiol 2013;23:93–​8. Henderson BE, Ross RK, Pike MC, Casagrande JT.
Ferlay J Bray F, Steliarova-​ Foucher E, Forman D. Endogenous hormones as a major factor in human
Cancer Incidence in Five Continents, CI5plus. cancer. Cancer Res 1982;42:3232–​9.
IARC CancerBase No. 9.  2014. http://​ci5.iarc. Henderson IC, Mouridsen H, Abe O, Abeloff M,
fr. 2015. Ahmann D, Andersen K et al. Effects of adjuvant
Ferlay J, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo tamoxifen and of cytotoxic therapy on mortality
M, et al. GLOBOCAN 2012 v1.0. Cancer Incidence in early breast cancer: an overview of 61 random-
and Mortality Worldwide: IARC CancerBase No. ized trials among 28,896 women. Early Breast
11 [Internet]. Lyon, France: International Agency Cancer Trialists’ Collaborative Group. N Engl J
for Research on Cancer (2013). Available at http://​ Med 1988;319:1681–​92.
globocan.iarc.fr (accessed November 2015). Hoover R, Fraumeni JF, Everson R, Myers MH. Cancer
Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers of the uterine corpus after hormonal treatment for
C, Rebelo M et al. Cancer incidence and mortality breast cancer. Lancet 1976;1:885–​7.
worldwide: sources, methods and major patterns in Husing A, Dossus L, Ferrari P, Tjonneland A, Hansen
GLOBOCAN 2012. Int J Cancer 2015;136:E359–​86. L, Fagherazzi G et al. An epidemiological model
Friberg E, Orsini N, Mantzoros CS, Wolk A. Diabetes for prediction of endometrial cancer risk in
mellitus and risk of endometrial cancer:  a meta-​ Europe. Eur J Epidemiol 2016;31:51–​60.
analysis. Diabetologia 2007;50:1365–​74. Je Y, DeVivo I, Giovannucci E. Long-​ term alco-
Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, hol intake and risk of endometrial cancer in the
Schairer C et  al. Projecting individualized prob- Nurses’ Health Study, 1980–​ 2010. Br J Cancer
abilities of developing breast cancer for white 2014;111:186–​94.
females who are being examined annually. J Natl Je Y, Giovannucci E. Coffee consumption and risk of
Cancer Inst 1989;81:1879–​86. endometrial cancer:  findings from a large up-​to-​
Grady D, Ernster V. Endometrial cancer. date meta-​analysis. Int J Cancer 2012;131:1700–​10.
In:  Schottenfeld D, Fraumeni JFJr (eds.) Ju W, Kim HJ, Hankinson SE, De Vivo I, Cho E.
Cancer Epidemiology and Prevention, 2 ed. Prospective study of body fat distribution and
New York: Oxford University Press, 1996: 1058–​89. the risk of endometrial cancer. Cancer Epidemiol
Grady D, Gebretsadik T, Kerlikowske K, Ernster V, 2015;39:567–​70.
Petitti D. Hormone replacement therapy and Kaaks R. Nutrition, hormones, and breast cancer:  is
endometrial cancer risk:  a meta-​analysis. Obstet insulin the missing link? Cancer Causes Control
Gynecol 1995;85:304–​13. 1996;7:605–​25.
Graham JD, Clarke CL. Physiological action of pro- Kaaks R, Lukanova A, Kurzer MS. Obesity, endoge-
gesterone in target tissues. Endocrine Reviews nous hormones, and endometrial cancer risk:  a
1997;18:502–​19. synthetic review. Cancer Epidemiol Biomarkers
Gruber SB, Thompson WD. A population-​ based Prev 2002;11:1531–​43.
study of endometrial cancer and familial risk in Kandoth C, Schultz N, Cherniack AD, Akbani R,
younger women. Cancer and Steroid Hormone Liu Y, Shen H et  al. Integrated genomic char-
Study Group. Cancer Epidemiol Biomarkers Prev acterization of endometrial carcinoma. Nature
1996;5:411–​7. 2013;497:67–​73.
Gunter MJ, Hoover DR, Yu H, Wassertheil-​Smoller Karageorgi S, Hankinson SE, Kraft P, De Vivo I.
S, Manson JE, Li J et al. A prospective evaluation Reproductive factors and postmenopausal hor-
of insulin and insulin-​like growth factor-​I as risk mone use in relation to endometrial cancer risk in
factors for endometrial cancer. Cancer Epidemiol the Nurses’ Health Study cohort 1976-​2004. Int J
Biomarkers Prev 2008;17:921–​9. Cancer 2010;126: 208–​16.
Hankinson SE, Willett WC, Manson JE, Hunter DJ, Kaye SA, Folsom AR, Soler JT, Prineas RJ, Potter JD.
Colditz GA, Stampfer MJ et  al. Alcohol, height, Associations of body mass and fat distribution
and adiposity in relation to estrogen and prolactin with sex hormone concentrations in postmeno-
levels in postmenopausal women. J Natl Cancer pausal women. Int J Epidemiol 1991;20:151–​6.
Inst 1995;87:1297–​302. Kennedy DL, Baum C, Forbes MB. Noncontraceptive
Henderson BE, Casagrande JT, Pike MC, Mack estrogens and progestins: use patterns over time.
T, Rosario I, Duke A. The epidemiology of Obstet Gynecol 1985;65:441–​6.
45

Endometrial Cancer 455

Key TJ, Appleby PN, Reeves GK, Roddam A, Cancer Cohort Consortium. Am J Epidemiol
Dorgan JF, Longcope C et  al. Body mass index, 2012; 176:1121–​9.
serum sex hormones, and breast cancer risk MacMahon B. Risk factors for endometrial cancer.
in postmenopausal women. J Natl Cancer Inst Gyn Oncol 1974;2:122–​9.
2003;95:1218–​26. Maxwell GL, Tian C, Risinger J, Brown CL, Rose
Key TJ, Pike MC. The dose-​ effect relationship GS, Thigpen JT et al. Racial disparity in survival
between “unopposed” oestrogens and endome- among patients with advanced/​recurrent endo-
trial mitotic rate: its central role in explaining and metrial adenocarcinoma: a Gynecologic Oncology
predicting endometrial cancer risk. Br J Cancer Group study. Cancer 2006;107:2197–​205.
1988;57:205–​12. McPherson CP, Sellers TA, Potter JD, Bostick RM,
Kvale G, Heuch I, Ursin G. Reproductive factors and Folsom AR. Reproductive factors and risk of
risk of cancer of the uterine corpus: a prospective endometrial cancer. The Iowa Women’s Health
study. Cancer Res 1988;48:6217–​21. Study. Am J Epidemiol 1996;143:1195–​202.
Kvaskoff M, Mu F, Terry KL, Harris HR, Poole EM, Merritt MA, Tzoulaki I, Tworoger SS, De Vivo I,
Farland L et  al. Endometriosis:  a high-​risk pop- Hankinson SE, Fernandes J, et  al. Investigation
ulation for major chronic diseases? Hum Reprod of dietary factors and endometrial cancer risk
Update 2015;21:500–​16. using a nutrient-​wide association study approach
La Vecchia C, Franceschi S, Decarli A, Gallus G, in the EPIC and Nurses’ Health Study (NHS)
Tognoni G. Risk factors for endometrial cancer at and NHSII. Cancer Epidemiol Biomarkers Prev
different ages. J Natl Cancer Inst 1984;73:667–​71. 2015;24:466–​71.
La Vecchia C, Negri E, Franceschi S. Education and Moradi T, Nyren O, Bergstrom R, Gridley G, Linet
cancer risk. Cancer 1992;70: 2935–​41. M, Wolk A et  al. Risk for endometrial cancer
La Vecchia C, Negri E, Franceschi S, D’Avanzo B, in relation to occupational physical activity:  a
Boyle P. A case-​control study of diabetes mellitus nationwide cohort study in Sweden. Int J Cancer
and cancer risk. Br J Cancer 1994;70:950–​3. 1998;76:665–​70.
Leminen A, Paavonen J, Vesterinen E, Wahlstrom T, Moradi T, Weiderpass E, Signorello LB, Persson I,
Rantala I, Lehtinen M. Human papillomavirus Nyren O, Adami HO. Physical activity and post-
types 16 and 18 in adenocarcinoma of the uterine menopausal endometrial cancer risk (Sweden).
cervix. Am J Clin Pathol 1991;95:647–​52. Cancer Causes Control 2000;11:829–​37.
Lindemann K, Vatten LJ, Ellstrom-​Engh M, Eskild A. Mouw T, Koster A, Wright ME, Blank MM, Moore SC,
Body mass, diabetes and smoking, and endome- Hollenbeck A et al. Education and risk of cancer
trial cancer risk:  a follow-​up study. Br J Cancer in a large cohort of men and women in the United
2008;98:1582–​5. States. PLoS One 2008;3:e3639.
Long J, Zheng W, Xiang YB, Lose F, Thompson D, Mucci LA, Hjelmborg J, Harris J, Czene K, Havelick
Tomlinson I et al. Genome-​wide association study D, Scheike T et  al. Familial risk and heritabil-
identifies a possible susceptibility locus for endo- ity of cancer in the Nordic Twin Cohorts. JAMA
metrial cancer. Cancer Epidemiol Biomarkers 2016;315:68–​76.
Prev 2012;21:980–​7. Mueck AO, Seeger H, Rabe T. Hormonal contracep-
Lu Y, Ek WE, Whiteman D, Vaughan TL, Spurdle AB, tion and risk of endometrial cancer: a systematic
Easton DF et al. Most common “sporadic” cancers review. Endocrine Relat Cancer 2010;17:R263–​71.
have a significant germline genetic component. Mutter GL, Prat J. Pathology of the Female Reproductive
Hum Mol Gen 2014;23:6112–​8. Tract. 3 ed. New York: Elsevier, 2014.
Lukanova A, Lundin E, Micheli A, Arslan A, Ferrari Nagle CM, Marquart L, Bain CJ, O’Brien S, Lahmann
P, Rinaldi S et al. Circulating levels of sex steroid PH, Quinn M et al. Impact of weight change and
hormones and risk of endometrial cancer in post- weight cycling on risk of different subtypes of endo-
menopausal women. International journal of can- metrial cancer. Eur J Cancer 2013;49:2717–​26.
cer. Int J Cancer 2004;108:425–​32. Nicolaides NC, Papadopoulos N, Liu B, Wei YF,
MacDonald PC, Edman CD, Hemsell DL, Porter Carter KC, Ruben SM et al. Mutations of two PMS
JC, Siiteri PK. Effect of obesity on conversion of homologues in hereditary nonpolyposis colon
plasma androstenedione to estrone in postmeno- cancer. Nature 1994;371:75–​80.
pausal women with and without endometrial can- Papadopoulos N, Nicolaides NC, Wei YF, Ruben SM,
cer. Am J Obstet Gynecol 1978;130:448–​55. Carter KC, Rosen CA et  al. Mutation of a mutL
Machiela MJ, Lindstrom S, Allen NE, Haiman CA, homolog in hereditary colon cancer. Science
Albanes D, Barricarte A et al. Association of type 1994;263:1625–​9.
2 diabetes susceptibility variants with advanced Parekh N, Okada T, Lu-​ Yao GL. Obesity, insulin
prostate cancer risk in the Breast and Prostate resistance, and cancer prognosis: implications for
456

456 Textbook of Cancer Epidemiology

practice for providing care among cancer survi- Risinger JI, Berchuck A, Kohler MF, Watson P,
vors. J Am Diet Ass 2009;109:1346–​53. Lynch HT, Boyd J. Genetic instability of micro-
Park JC, Lim SY, Jang TK, Bae JG, Kim JI, Rhee JH. satellites in endometrial carcinoma. Cancer Res
Endometrial histology and predictable clinical 1993;53:5100–​3.
factors for endometrial disease in women with Rowlands IJ, Nagle CM, Spurdle AB, Webb PM.
polycystic ovary syndrome. Clin Experiment Gynecological conditions and the risk of endome-
Reprod Med 2011;38:42–​6. trial cancer. Gynecol Oncol 2011;123:537–​41.
Peltomaki P, Aaltonen LA, Sistonen P, Pylkkanen L, Rutqvist LE. Controversial issues in adjuvant sys-
Mecklin JP, Jarvinen H et al. Genetic mapping of temic therapy of early breast cancer. Acta Oncol
a locus predisposing to human colorectal cancer. 1998;37:421–​30.
Science 1993; 260:810–​2. Rutqvist LE, Johansson H, Signomklao T, Johansson
Persson I, Adami HO, Bergkvist L, Lindgren A, U, Fornander T, Wilking N. Adjuvant tamoxifen
Pettersson B, Hoover R et al. Risk of endometrial therapy for early stage breast cancer and second
cancer after treatment with oestrogens alone or in primary malignancies. Stockholm Breast Cancer
conjunction with progestogens:  results of a pro- Study Group. J Natl Cancer Inst 1995;87:645–​51.
spective study. BMJ 1989;298:147–​51. Schildkraut JM, Risch N, Thompson WD. Evaluating
Persson I, Schmidt M, Adami HO, Bergstrom R, genetic association among ovarian, breast, and
Pettersson B, Sparen P. Trends in endometrial can- endometrial cancer:  evidence for a breast/​ovarian
cer incidence and mortality in Sweden, 1960–​84. cancer relationship. Am J Hum Gen 1989;45:521–​9.
Cancer Causes Control 1990;1:201–​8. Schmandt RE, Iglesias DA, Co NN, Lu KH.
Persson I, Weiderpass E, Bergkvist L, Bergstrom R, Understanding obesity and endometrial cancer
Schairer C. Risks of breast and endometrial can- risk:  opportunities for prevention. Am J Obstet
cer after estrogen and estrogen-​progestin replace- Gynecol 2011;205:518–​25.
ment. Cancer Causes Control 1999;10:253–​60. Setiawan VW, Pike MC, Karageorgi S, Deming
Pfeiffer RM, Park Y, Kreimer AR, Lacey JV Jr, Pee SL, Anderson K, Bernstein L et  al. Age at last
D, Greenlee RT et  al. Risk prediction for breast, birth in relation to risk of endometrial can-
endometrial, and ovarian cancer in white women cer:  pooled analysis in the epidemiology of
aged 50 y or older:  derivation and validation endometrial cancer consortium. Am J Epidemiol
from population-​based cohort studies. PLoS Med 2012;176:269–​78.
2013;10:e1001492. Setiawan VW, Yang HP, Pike MC, McCann SE, Yu H,
Pike MC, Peters RK, Cozen W, Probst-​Hensch NM, Xiang YB et  al. Type I  and II endometrial can-
Felix JC, Wan PC et  al. Estrogen-​ progestin cers: have they different risk factors? J Clin Oncol
replacement therapy and endometrial cancer. J 2013;31:2607–​18.
Natl Cancer Inst 1997;89:1110–​6. Sherman ME, Carreon JD, Lacey JV Jr, Devesa SS.
Potischman N, Hoover RN, Brinton LA, Siiteri P, Impact of hysterectomy on endometrial carci-
Dorgan JF, Swanson CA et al. Case-​control study noma rates in the United States. J Natl Cancer Inst
of endogenous steroid hormones and endometrial 2005;97:1700–​2.
cancer. J Natl Cancer Inst 1996a;88:1127–​35. Soliman PT, Wu D, Tortolero-​Luna G, Schmeler KM,
Potischman N, Swanson CA, Siiteri P, Hoover RN. Slomovitz BM, Bray MS et al. Association between
Reversal of relation between body mass and adiponectin, insulin resistance, and endometrial
endogenous estrogen concentrations with meno- cancer. Cancer 2006;106:2376–​81.
pausal status. J Natl Cancer Inst 1996b;88:756–​8. Spurdle AB, Thompson DJ, Ahmed S, Ferguson K,
Reeves GK, Pirie K, Beral V, Green J, Spencer E, Healey CS, O’Mara T et  al. Genome-​wide asso-
Bull D. Cancer incidence and mortality in rela- ciation study identifies a common variant asso-
tion to body mass index in the Million Women ciated with risk of endometrial cancer. Nat Gen
Study: cohort study. BMJ 2007;335:1134. 2011;43:451–​4.
Reichman ME, Judd JT, Longcope C, Schatzkin A, Sturgeon SR, Brinton LA, Berman ML, Mortel R,
Clevidence BA, Nair PP et  al. Effects of alcohol Twiggs LB, Barrett RJ et al. Past and present phys-
consumption on plasma and urinary hormone ical activity and endometrial cancer risk. Br J
concentrations in premenopausal women. J Natl Cancer 1993;68:584–​9.
Cancer Inst 1993;85:722–​7. Sturgeon SR, Brock JW, Potischman N, Needham LL,
Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen Rothman N, Brinton LA et al. Serum concentra-
M. Body-​mass index and incidence of cancer:  a tions of organochlorine compounds and endome-
systematic review and meta-​analysis of prospective trial cancer risk (United States). Cancer Causes
observational studies. Lancet 2008;371:569–​78. Control 1998;9:417–​24.
457

Endometrial Cancer 457

Swanson CA, Potischman N, Wilbanks GD, Twiggs Weiderpass E, Adami HO, Baron JA, Wicklund-​Glynn
LB, Mortel R, Berman ML et al. Relation of endo- A, Aune M, Atuma S et al. Organochlorines and
metrial cancer risk to past and contemporary body endometrial cancer risk. Cancer Epidemiol
size and body fat distribution. Cancer Epidemiol Biomarkers Prev 2000;9:487–​93.
Biomarkers Prev 1993;2:321–​7. Weiderpass E, Pukkala E, Vasama-​ Neuvonen K,
Tanko LB, Christiansen C. An update on the anties- Kauppinen T, Vainio H, Paakkulainen H et  al.
trogenic effect of smoking: a literature review with Occupational exposures and cancers of the endo-
implications for researchers and practitioners. metrium and cervix uteri in Finland. Am J Indust
Menopause 2004;11:104–​9. Med 2001a;39:572–​80.
Tavassoli FA, Devilee P. Pathology and Genetics Weiderpass E, Ye W, Mucci LA, Nyren O, Trichopoulos
of Tumours of the Breast and Female Genital D, Vainio H et al. Alcoholism and risk for endo-
Organs. Chapter  4:  Tumours of the uterine cor- metrial cancer. Int J Cancer 2001b;93:299–​301.
pus. World Health Organization Classification of Weiss NS, Farewall VT, Szekely DR, English DR, Kiviat
Tumours, 2003. N. Oestrogens and endometrial cancer:  effect of
Terry PD, Miller AB, Rohan TE. A prospective cohort other risk factors on the association. Maturitas
study of cigarette smoking and the risk of endo- 1980;2:185–​90.
metrial cancer. Br J Cancer 2002a;86:1430–​5. Win AK, Reece JC, Ryan S. Family history and risk
Terry PD, Rohan TE, Franceschi S, Weiderpass E. of endometrial cancer:  a systematic review and
Cigarette smoking and the risk of endometrial meta-​analysis. Obstet Gynecol 2015;125:89–​98.
cancer. Lancet Oncol 2002b;3:470–​80. World Cancer Research Fund/​American Institute for
Trentham-​ Dietz A, Nichols HB, Hampton JM, Cancer Research (WCRF/​ AICR). Endometrial
Newcomb PA. Weight change and risk of endome- Cancer 2013 Report. Food, Nutrition, Physical
trial cancer. Int J Epidemiol 2006;35:151–​8. Activity, and the Prevention of Cancer. 2013.
Troisi R, Potischman N, Hoover RN, Siiteri P, Brinton http://​www.dietandcancerreport.org.
LA. Insulin and endometrial cancer. Am J Wright JD, Fiorelli J, Schiff PB, Burke WM, Kansler
Epidemiol 1997;146:476–​82. AL, Cohen CJ et  al. Racial disparities for uter-
Vasen HF, Watson P, Mecklin JP, Jass JR, Green JS, ine corpus tumors:  changes in clinical char-
Nomizu T et al. The epidemiology of endometrial acteristics and treatment over time. Cancer
cancer in hereditary nonpolyposis colorectal can- 2009;115:1276–​85.
cer. Anticancer Res 1994;14:1675–​8. Wu QJ, Li YY, Tu C, Zhu J, Qian KQ, Feng TB et  al.
Viswanathan AN, Feskanich D, De Vivo I, Hunter Parity and endometrial cancer risk: a meta-​analysis
DJ, Barbieri RL, Rosner B et  al. Smoking of epidemiological studies. Sci Rep 2015;5:14243.
and the risk of endometrial cancer:  results Yang HP, Wentzensen N, Trabert B, Gierach GL, Felix
from the Nurses’ Health Study. Int J Cancer AS, Gunter MJ et al. Endometrial cancer risk factors
2005;114:996–​1001. by 2 main histologic subtypes: the NIH-​AARP Diet
Wang L, Li J, Shi Z. Association between breastfeed- and Health Study. Am J Epidemiol 2013;177:142–​51.
ing and endometrial cancer risk:  evidence from Zhan B, Liu X, Li F, Zhang D. Breastfeeding and the
a systematic review and meta-​analysis. Nutrients incidence of endometrial cancer: a meta-​analysis.
2015;7:5697–​711. Oncotarget 2015;6:38398–​409.
Waterhouse J, Muir C, Correa P. Cancer Incidence Zhao Y, Liang J, Qi JG, Yang N, Wu G, Lin YL et al.
in Five Continents. IARC Scientific Publications. Meta-​ analysis of the association between the
Lyon: International Agency for Research on HNF1B rs4430796 (A>G) polymorphism and risk
Cancer, 1992. of prostate cancer based on case-​control studies.
Weiderpass E, Adami HO, Baron JA, Magnusson Gen Mol Res 2015;14:7426–​35.
C, Bergstrom R, Lindgren A et al. Risk of endo- Zhou B, Yang L, Sun Q, Cong R, Gu H, Tang N et al.
metrial cancer following estrogen replacement Cigarette smoking and the risk of endometrial can-
with and without progestins. J Natl Cancer Inst cer: a meta-​analysis. Am J Med 2008;121:501–​508.
1999:91:1131–​7.
458
459

19
Ovarian Cancer
P E N E L O P E M . W E B B , S U S A N J . J O R D A N , A N D D AV I D J . H U N T E R

A mong the malignancies originating in the

female reproductive organs, cancer of the
ovary presents a particular challenge as there
serous (<5%) and mucinous (<5%) cancers,
which can resemble the endocervix or gastro-
intestinal epithelium, are often associated with
are no effective tools for screening and, because areas of borderline tumor (Kurman et al, 2014).
women often present after the cancer has spread Symptoms Ovarian cancer is usually insidious,
beyond the ovaries, the prognosis is poor. While with most patients presenting when their disease
the etiology of cancers of the cervix uteri and has reached an advanced stage. Symptoms include
endometrium is quite well understood, less abdominal pain or discomfort, abdominal swell-
is known about the causes of ovarian cancer. ing, or other nonspecific symptoms due to pelvic
Furthermore, a high proportion of what were pre- pressure such as urinary frequency. An abdominal
viously called ovarian cancers are now thought to mass is frequently present on examination.
originate from the fallopian tube. The term “ovar-
Diagnosis A definitive diagnosis requires his-
ian cancer” as used here therefore includes both
topathological examination of the tumor tissue
these and the histologically similar primary per-
following surgery or biopsy. Cytological exami-
itoneal cancers. Epidemiologists have identified
nation (of ascites or pleural effusions) can iden-
reproductive factors, notably higher parity, and
tify a likely ovarian cancer but cannot confirm
interventions, such as oral contraceptive use and
the primary site.
tubal ligation, as protecting against these cancers,
but these cannot be widely promoted at the pop- Treatment Optimal treatment involves surgical
ulation level for cancer prevention. Although removal of as much of the tumor as possible and,
several plausible biologic mechanisms have been in most cases, neoadjuvant and/​or adjuvant com-
proposed to explain the relationship between bination chemotherapy.
these factors and the carcinogenic process, con- Prognosis Overall 5-​year relative survival rates
vincing molecular evidence for these mecha- are 40% to 45%, compared to 35%–​40% two to
nisms is lacking. three decades ago. Most women respond well to
primary treatment however, with the exception
CLINICAL SYNOPSIS of cancers confined to the ovaries at diagnosis
Subgroups Ovarian tumors are classified accord- (only about a quarter of cases), most will subse-
ing to their cellular origins as epithelial (90%), quently experience a recurrence that will eventu-
sex-​cord stromal (6%), or germ-​cell (3%) tumors. ally prove fatal.
Malignant epithelial tumors may be classified as
either borderline (low malignant potential) or DESCRIPTIVE
invasive. Five major histotypes are now recog- EPIDEMIOLOGY
nized. High-​grade serous cancers, which are his- The incidence of ovarian cancer varies widely
tologically similar to fallopian tube epithelium among countries. In general, incidence rates
and, in many cases, appear to arise from the fal- are higher in higher-​ income than in lower-​
lopian tube, constitute about 65% of invasive can- income countries. Northern and central/​eastern
cers. Endometrioid cancers (10%–​15%), which European countries like Poland and the United
resemble endometrial tissue, and clear cell can- Kingdom have the highest age-​ standardized
cers (5%–​10%) both commonly arise in areas of rates (12–​14 cases/​100,000 women), the United
ovarian endometriosis, while the rarer low-​grade States, Japan, and western European countries
460

460 Textbook of Cancer Epidemiology

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 2 4 6 8 10 12 14 16
ASR (world) per 100,000 person-years

FIGURE 19-​1  Age-​standardized (to the 2012 world population) incidence rates of ovarian cancer.
Source: Ferlay et al, 2013.

like the Netherlands have intermediate rates low-​risk countries, incidence rates of ovarian can-
(7–​9/​100,000), and other Asian countries such cer have increased modestly over the past 25 years.
as China have the lowest rates (<7/​ 100,000) Studies of women emigrating from low-​risk
(Fig. 19-​ 1). Differences in completeness of countries to the United States, and studies of the
cancer surveillance may explain some of this descendants of these women, suggest that women
international variation, as may variation in hys- born in Asia generally maintain a lower inci-
terectomy and oophorectomy rates, since fewer dence of ovarian cancer than women born in the
women are at risk of ovarian cancer in countries United States. However, rates are higher among
with higher rates of oophorectomy. Other life- younger Chinese and Japanese women born in
style and reproductive factors are also likely to the United States than among their Asian-​born
play a major role. counterparts and are more similar to rates among
In most high-​ income countries, age-​white women in the United States (Herrinton
standardized incidence rates have declined slightly et al, 1994).
over the past 25  years. This may be explained in Ovarian cancer is rare in women under
part by the protective effects of oral contraceptives 40 years of age, and most cancers in this age group
(discussed later). In the United States, white women are germ cell tumors. The risk increases sharply
have approximately 40% higher rates of ovarian can- after about age 40, peaking between the ages of
cer than African American and Hispanic women, 65 and 79 years (Fig. 19-​2), and more than 90% of
and this difference has persisted in both pre-​and cancers at these ages are epithelial tumors. Most
postmenopausal women (Ferlay et  al, 2013). In epidemiologic studies have therefore focused on
New cases per 100,000

70
60
person-years

50
40
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 19-​2  Age-​specific incidence rates of ovarian cancer among women.

Source: Ferlay et al, 2013.
461

Ovarian Cancer 461

epithelial cancers and the following discussion associated with an increased risk of colorectal
relates primarily to this group. and endometrial cancers, but also increases risk
of ovarian cancer, particularly nonserous cancer,
GENETIC AND MOLECULAR which occurs in about 5% to 10% of women with
EPIDEMIOLOGY Lynch syndrome (Lynch et al, 1998).
Other rare hereditary disorders predispose to
Inherited Susceptibility the development of ovarian tumors other than
Family history has consistently emerged as a risk epithelial tumors. Peutz-​Jeghers syndrome is an
factor for ovarian cancer with a relative risk of about autosomal dominant condition characterized by
three for women with one affected first-​degree rela- intestinal hamartomas and mucocutaneous pig-
tive (Stratton et al, 1998) and higher risks for those mentation, as well as an increase in the frequency
whose relative was diagnosed below the age of of sex-​cord stromal tumors and epithelial ovarian
50 (Jervis et al, 2014). The data are less consistent tumors. Germ-​cell tumors, in particular gonado-
regarding a family history of breast cancer, however blastomas, occur with increased frequency in
familial aggregation of ovarian and breast cancer phenotypic females with an XY karyotype and
has been described in numerous kindreds. dysgenetic gonads.
In addition to these high-​ risk syndromes,
High-​Penetrance Germline Mutations rare germline mutations in other genes includ-
Two hereditary syndromes account for at least ing BRIP1 (Ramus et al, 2015) and RAD51 (Song
10% of all ovarian cancers (Boyd, 2003) although et  al, 2015)  have now been shown to confer a
a variety of other less common syndromes also moderately increased risk of ovarian cancer.
increase risk of ovarian cancer. Breast-​ovarian
cancer syndrome accounts for about 90% of Low-​Penetrance Gene Polymorphisms
hereditary ovarian cancers. Affected families While many studies have suggested that com-
usually have an excess of both breast and ovar- mon variation in genes involved in metabolism
ian cancer but in 10%–​15% of families the ova- of carcinogens (for example, the glutathione S-​
ries are the only site involved. A wide variety of transferases GST-​M1, GST-​T1, GST-​P1, micro-
germline mutations in BRCA1, located on chro- somal epoxide hydrolase), hormonal pathways
mosome 17q, and BRCA2, located on chromo- (the PROGINS variant of the progesterone
some 13q, account for the majority of hereditary receptor gene, the CAG repeat in the androgen
breast-​ovarian cancers, and the exact location receptor gene, variants in estrogen-​metabolizing
of the mutation within the BRCA1 and BRCA2 genes such as CYP1A1, CYP1B1, and CYP17),
genes may influence the development of breast and DNA repair (XRCC2, XRCC3, RAD51,
or ovarian cancer (Rebbeck et al, 2015). BRCA1 RAD52) may increase susceptibility to ovarian
and BRCA2 mutations are also common among cancer, these have rarely been confirmed in sub-
women with ovarian cancer who do not have a sequent studies. However, to date, genome-​wide
family history of either breast or ovarian cancer association studies have now identified more
(Alsop et al, 2012), and approximately 25% of the than 20 confirmed susceptibility loci for epi-
familial aggregation of epithelial ovarian cancer thelial ovarian cancer, although some appear
can be attributed to mutations in these genes specific for individual histotypes (Kelemen
(Jervis et  al, 2014). Women carrying mutations et  al, 2015)  or BRCA1 mutation carriers
in BRCA1 have an estimated 40% to 50% risk of (Kuchenbaecker et al, 2015). These loci include
developing ovarian cancer by age 70. For BRCA2 the CHEK2, WNT4, TERT, and ABO genes,
mutation carriers, the risk is lower, about 10% to which have also been associated with cancers at
20% (Boyd, 2003). The majority of ovarian can- other sites.
cers associated with germline BRCA1 or BRCA2
mutations are high-​ grade serous tumors, and Somatic Events
up to 20% of these cancers are associated with a The different histotypes of ovarian cancer are
germline BRCA1/​2 mutation (Alsop et al, 2012). characterized by distinct molecular changes.
Lynch syndrome, or hereditary nonpolypo- High-​grade serous cancers exhibit high levels of
sis colorectal cancer (HNPCC), is an autosomal chromosomal instability and widespread copy
dominant disorder caused by mutations in genes number changes with amplification common in
involved in DNA mismatch repair, including regions containing well known oncogenes such
MLH1, MSH2, MSH6, and PMS2. It is primarily as CCNE1 (which encodes Cyclin E1), MYC,
462

462 Textbook of Cancer Epidemiology

PIK3CA, and AKT, and deletions common in have been reported in some high-​grade endo-
regions with tumor suppressor genes including metrioid cancers, similarities between these and
PTEN, RB1, and NF1 (Cancer Genome Atlas high-​grade serous cancers suggest that these
Research Network, 2011). Almost all (>95%) high-​grade endometrioid cancers were actually
high-​grade serous cancers carry a mutation in the misclassified serous tumors.
TP53 gene, and a high proportion also have inacti-
vation (either via germline or somatic mutations, RISK FACTORS
or promoter methylation) of BRCA1 or BRCA2
(Cancer Genome Atlas Research Network, 2011; Tobacco
Prat, 2012). These cancers are now thought to Tobacco smoking appears to increase risk of
arise from intraepithelial carcinomas in the fal- the mucinous subtype of ovarian cancer only,
lopian tube (TIC) and the observation that TIC with little association seen for serous ovar-
and concomitant high-​grade serous cancers har- ian cancers and an inverse association with the
bor identical TP53 mutations suggests that this endometrioid and clear cell subtypes. The rela-
is an early event in the carcinogenic process. The tionship between tobacco use and reproductive
group of high-​grade serous cancers can be fur- cancers is potentially complex, as the effects of
ther subdivided based on their RNA expression cigarette smoking on ovarian function may be
profiles, with four subtypes that differ in clini- both direct and indirect. Cigarette smokers have
cal outcome now recognized (Tothill et al, 2008; reduced levels of urinary estradiol (Westhoff
Cancer Genome Atlas Research Network, 2011). et al, 1996), which may result in elevated gonad-
In contrast to high-​ grade serous cancers, otropin levels and, on average, smokers have an
low-​grade serous cancers tend to be genetically earlier onset of menopause than nonsmokers
stable and TP53 and BRCA mutations are rare. (McKinlay, 1996). Postmenopausal smokers also
Instead about two-​ thirds of low-​ grade serous have elevated levels of androgens compared to
cancers carry either a BRAF or a KRAS mutation nonsmokers (Khaw et al, 1988). Two large-​scale
with similar mutations seen in adjacent areas of pooled analyses, one including 51 cohort and
benign and borderline tumor, suggesting they are case-​control studies (median year of diagnosis
early changes that precede a progressive transfor- 1995) (Collaborative Group on Epidemiological
mation to cancer (Ho et al, 2004). KRAS but not Studies of Ovarian Cancer, 2012), and the other
BRAF mutations have also been observed in up 21 case-​control studies (including 12 newer stud-
to half of borderline and invasive mucinous ovar- ies not included in the previous analysis) (Faber
ian cancers (Gemignani et al, 2003). Again, iden- et al, 2013), have reported significant 30%–​50%
tical mutations have been reported in benign, increases in risk of invasive mucinous cancer
borderline, and invasive areas within the same and 80%–​130% increases for borderline muci-
tumor suggesting that the mutations are early nous tumors among current smokers. The risks
events in the development of mucinous tumors increased with increasing duration of smoking
via an adenoma-​ carcinoma sequence (Garrett and declined with time after smoking cessation.
et al, 2001). In contrast, smoking did not appear to increase
KRAS mutations are less common in clear cell risk of serous cancers and current smokers had
cancers and rare in endometrioid tumors. Both a 20% lower risk of developing endometrioid
clear cell and endometrioid tumors are associ- and clear cell cancers. This latter observation is
ated with microsatellite instability, although this consistent with the observed reduction in risk of
is rarely seen in serous and mucinous tumors endometrial cancer among smokers (Setiawan
(Shih & Kurman, 2004), and 30%–​50% carry et al, 2013).
mutations in the ARID1A gene. In two patients
with clear cell cancers, identical ARID1A muta- Diet
tions were identified in the cancer and contig- The initial suggestions of a potential link between
uous areas of endometriosis, but not in distant ovarian cancer and dietary factors came from
endometriotic lesions, suggesting ARID1A muta- international correlations between national ovar-
tion may be an early event in the transformation ian cancer incidence and mortality rates and per
of endometriosis to cancer (Wiegand et al, 2010). capita dietary consumption data. Positive correla-
Endometrioid cancers are also associated with tions were noted for intake of total fat, particularly
abnormalities in PTEN and CTNNB1, the gene animal fat; milk and meats; and, more weakly,
that codes beta-​catenin. While TP53 mutations eggs, while strong inverse correlations were
463

Ovarian Cancer 463

observed for vegetable availability (Armstrong In an early case-​control study GALT activity
& Doll, 1975; Rose et al, 1986). An association measured in erythrocytes was inversely related to
between increased westernization of the Japanese ovarian cancer (Cramer et al, 1989) however sub-
diet, that is, increases in the intake of fat-​rich sequent studies looking at genetic variants asso-
foods, and increased mortality from ovarian ciated with GALT activity have not confirmed
cancer has also been noted (Kato et al, 1987). the association (Fung et al, 2003).
Although many case-​control studies have since
evaluated the relation between diet and ovarian Calcium and Vitamin D
cancer risk, they are susceptible to both selec- Studies evaluating the relation between vitamin
tion and recall biases and may therefore overstate D and calcium intake and ovarian cancer risk
the magnitude of any associations with dietary have reported mixed results. A  pooled cohort
factors. The following discussion focuses on the analysis found no association with calcium
evidence from cohort studies, as this is likely to intake but reported a modest but statistically sig-
be more reliable. Overall, there is currently little nificant increase in risk with increasing dietary
evidence to suggest a major role for diet in the vitamin D intake (pooled RR = 1.06 per 100 IU
development of ovarian cancer (World Cancer vitamin D per day), although no association was
Research Fund/​ American Institute for Cancer seen for total vitamin D including supplements
Research, 2014). (Genkinger et  al, 2006a). In contrast, a system-
atic review found no evidence of an association
Dairy Foods and Lactose/​Galactose between either calcium or serum vitamin D and
There are insufficient data to draw firm conclu- ovarian cancer risk (World Cancer Research
sions regarding the relation between intake of Fund/​ American Institute for Cancer Research
lactose-​containing foods and risk of ovarian [WCRF/​AICR], 2013a) and a randomized con-
cancer. The strong correlation between ovarian trolled trial of calcium plus vitamin D supple-
cancer and per capita milk consumption led to ments found no difference in ovarian cancer
the hypothesis that consumption of lactose may incidence between the intervention and placebo
be a risk factor for ovarian cancer. Dietary lac- groups during 7  years of follow-​ up (Brunner
tose is hydrolyzed to glucose and galactose, and et al, 2011).
galactose is toxic to oocytes. Hypergonadotropic
hypogonadism and early ovarian failure occur Coffee and Tea
in women with reduced activity of the enzyme Evidence regarding the association between cof-
galactose-​1-​phosphate uridyltransferase (GALT), fee and tea drinking and ovarian cancer risk is
which converts galactose to glucose. Subsequent currently limited and inconsistent (WCRF/​
increases in gonadotropin-​ releasing hormones AICR, 2014). A  meta-​analysis of cohort studies
may stimulate neoplastic changes in the ovary did not show any association with coffee drink-
(Cramer et al, 1989). ing (WCRF/​AICR, 2013a). The mechanism for
A Swedish linkage study reported a signif- a potential association between coffee drinking
icantly reduced risk of ovarian cancer among and ovarian cancer is unclear. Caffeine may be
women with lactose intolerance compared to the associated with menstrual cycle abnormalities
general population (Ji et al, 2015). Recent cohort (Fenster et  al, 1999), and coffee drinking has
studies have not, however, shown a significant been reportedly linked with decreased mater-
association between lactose intake and ovar- nal pregnancy estrogen levels (Petridou et  al,
ian cancer (Merritt et  al, 2014), and an earlier 1992) and early follicular estrogen levels (Lucero
pooled analysis that included both published and et  al, 2001). However, there is little evidence
unpublished data from 12 cohort studies with a for a direct role of estrogens in ovarian cancer,
total of 2,132 cases reported only a modest asso- and, although coffee has been associated with
ciation with a nonsignificant 4% increase in risk a reduced risk of endometrial cancer (WCRF/​
per 10 g lactose (Genkinger et al, 2006a). Women AICR, 2013) which is strongly estrogen-​driven, it
who consumed more than 30 g lactose per day has not been associated with breast cancer, which
(equivalent to three glasses of milk) were, how- is also more strongly associated with increased
ever, at significantly higher risk than those con- estrogen levels than ovarian cancer.
suming less than 10 g (relative risk [RR] = 1.2). In 2002, Zhang and colleagues reported that
The results did not vary appreciably for the dif- tea drinking, particularly green tea, was asso-
ferent subtypes of ovarian cancer. ciated with a significant 70% reduction in risk
46

464 Textbook of Cancer Epidemiology

of ovarian cancer in a Chinese case-​ control neither a pooled analysis of 10 cohort studies nor
study (Zhang et  al, 2002a). In contrast, most a more recent meta-​analysis found an association
cohort studies conducted in non-​Asian popula- with any of the major carotenoids (Koushik et al,
tions where black tea is more common have not 2006; WCRF/​AICR, 2013a). Likewise, there is
reported any association between tea drinking no evidence to suggest that folate intake is asso-
and ovarian cancer risk (WCRF/​AICR, 2013b). ciated with risk overall (WCRF/​AICR, 2013a),
An association is biologically plausible as both although increasing folate intake has been
black and green tea contain abundant polyphe- associated with a reduced risk of ovarian can-
nols that have been shown to inhibit carcin- cer among women with higher alcohol intake,
ogenesis in in vitro and animal studies (Yang in some (Kelemen et  al, 2004; Larsson et  al,
et al, 2002). 2004) but not all (Tworoger et al, 2006) studies.
The weak inverse associations with phystoestro-
Dietary Fat, Eggs, and Red Meat gen intake reported in case-​control studies and
Fat intake has not been consistently linked with one cohort study (Chang et  al, 2007)  were not
ovarian cancer. Although case-​ control stud- confirmed in a second cohort study (Hedelin
ies in several countries have reported associa- et al, 2011).
tions between intake of different types of fat and
ovarian cancer risk, early cohort studies did Alcohol
not report an association. However two pooled Alcohol intake, at least at moderate levels, does
analyses, one of 12 cohort studies that included not appear to increase risk of ovarian cancer.
2,132 cases (Genkinger et  al, 2006b) and one Alcohol affects the reproductive system and
of two cohort studies that included 1,522 cases is, therefore, potentially a risk factor for ovar-
(Merritt et al, 2016), have now reported modest ian cancer. It could be a direct ovarian toxin
but significant increases in risk for women with and may also act on the hypothalamic-​pituitary
the highest saturated fat intake (RR  =  1.29 and axis. Alcoholism is associated with menstrual
1.21, respectively). abnormalities, reproductive problems, and
A number of case-​control and early cohort changes in secondary sex characteristics, while
studies reported a significant association more moderate alcohol consumption has been
between egg consumption and ovarian cancer associated with elevated levels of hormones
risk (Pirozzo et al, 2002), but more recent cohort including estradiol among postmenopausal
studies have not seen any association; there is women (Endogenous Hormones Breast Cancer
also no consistent evidence for an association Collaborative Group, 2011).
with red meat (WCRF/​AICR, 2013a). Results of epidemiologic studies have been
inconsistent, however large-​scale pooled analy-
Fruit and Vegetables ses of data from 10 cohort studies (Genkinger
Overall it seems unlikely that fruit and vegetable et  al, 2006c) and 12 case-​ control studies
intake or their components have any major effect (Kelemen et  al, 2013)  have reported no associ-
on ovarian cancer risk, although prospective ation between total alcohol intake and ovarian
data are currently limited. An inverse association cancer risk and no clear associations for different
between ovarian cancer and intake of green veg- types of alcohol.
etables, carrots, and certain micronutrients, for
example, vitamins A  and E, was noted in some Reproductive Factors
case-​control studies (Cramer et  al, 2001; Zhang Parity
et al, 2002b), while a meta-​analysis of five case-​ Nulliparity is a well-​established risk factor for
control studies found a significant inverse associ- ovarian cancer (Whittemore et  al, 1992; Adami
ation for beta-​carotene (Huncharek et al, 2001). et al, 1994; Risch et al, 1994). The protective effect
Recent cohort studies including a pooled analy- of parity is seen for all invasive ovarian cancers,
sis of 12 cohort studies have, however, generally including epithelial, germ cell, and stromal
found no association between either vegetable tumors, but is less pronounced for borderline
or fruit intake and ovarian cancer risk (Koushik tumors. Reduced risks have been observed for
et al, 2005; WCRF/​AICR, 2013a). all histological subtypes of epithelial ovarian can-
Cohort studies have also reported null results cers, although some studies have suggested that
for micronutrients including vitamins A, C, and the effect is weaker for mucinous cancers (Gates
E (Kushi et  al, 1999; Fairfield et  al, 2001), and et al, 2010).
465

Ovarian Cancer 465

Most studies have demonstrated significant pregnancies, and indeed the existing literature
trends of increasing protection with increasing supports this (Dick et al, 2009).
parity, with a decrease in risk of approximately
20% for each subsequent live birth (Negri et  al, Infertility and Fertility Drugs
1991; Whittemore et al, 1992; Adami et al, 1994; There is some evidence that women who report
Risch et  al, 1994). However, the potential role a history of infertility have an increased risk of
of difficulty in conceiving must be considered invasive ovarian cancer (Tworoger et  al, 2007),
in order to distinguish the effects of nulliparity although this is not consistent across studies,
from those of infertility (Weiss, 1988). In stud- perhaps in part because of varying definitions
ies addressing this issue, a significant inverse of infertility. While the evidence is very consist-
association between parity and ovarian cancer ent for some specific causes of infertility (e.g.,
was noted, even among women with no history endometriosis—​ see “Medical Conditions and
of infertility (Whittemore et al, 1992; Risch et al, Treatment” section), whether this effect is also
1994). The benefit for women with BRCA muta- true for other causes of infertility, beyond the
tions is also not so clear-​cut and may even differ effect of reduced parity, is unclear.
between women with BRCA1 (reduced risk) and It is also unclear whether the use of fertility
BRCA2 (no reduction) mutations (Kotsopoulos drugs increases risk. As their use tends to increase
et al, 2015). the number of ovulations per cycle and raise lev-
There is conflicting evidence regarding the els of endogenous reproductive hormones, there
effect of age at first or last pregnancy on ovar- is concern around their potential to increase
ian cancer risk. Some studies observed a lower ovarian cancer risk. The main drugs used for
risk among women who were older at the time infertility treatment, either separately or in com-
of their first (Mills, et  al, 2004)  or last birth bination, are clomiphene citrate, an estrogen
(Whiteman et  al, 2003), but others have not antagonist; gonadotropins; and gonadotropin-​
confirmed this. releasing hormone agonists.
Several mechanisms have been proposed to A considerable number of studies have inves-
explain the protective effect of parity. Ovulation tigated the relationship between use of infertil-
ceases for the duration of pregnancy, thus the ity drugs and ovarian cancer risk (Rizzuto et al,
protective effect of parity is consistent with the 2013), but the evidence is inconsistent. Studies
incessant ovulation hypothesis, which states have been limited by difficulties in distinguish-
that repeated trauma to ovarian surface epi- ing associations with treatment from confound-
thelium from recurrent ovulation increases the ing by indications for treatment, changing
risk of ovarian cancer. Alternatively, pregnancy treatment regimens over time, and small num-
may exert long-​lasting effects on gonadotropin bers of exposed cases. Some large cohort stud-
secretion, although differences in gonadotropin ies, carried out among women reviewed in
concentration according to parity or age at first fertility clinics have revealed no excess risk
pregnancy have not been consistently observed. among women treated with ovarian stimulation
Pregnancy is also associated with high levels of (Venn et al, 1999; Jensen et al, 2009). However,
progesterone, which may reduce ovarian can- an increased risk of ovarian cancer more than
cer risk (Risch, 1998). Evidence to support this 15  years after treatment and among subgroups
latter hypothesis comes from observations that, of women with prolonged use or who remain
although they have higher gonadotropin lev- nulliparous cannot be excluded (Whittemore
els and are more likely to double ovulate than et al, 1992; Ness et al, 2002; Trabert et al, 2013).
women with only singleton births, mothers of In a recent retrospective cohort study from
twins do not have an elevated risk ovarian cancer. Israel, parous women with a history of IVF (in
If anything women who have twins may have a vitro fertilization) had a fourfold increase in
slightly lower risk of ovarian cancer, and this has ovarian cancer risk compared with women with
been attributed to the higher progesterone lev- no history of infertility, but no increase was seen
els associated with twin pregnancies (Whiteman for ovulation induction. However this study was
et al, 2000). not able to control for parity or oral contra-
On the basis of these hypotheses, incomplete ceptive use, both potentially lower in the IVF
pregnancies (miscarriages, abortions, or ectopic group (Kessous et al, 2016). As the use of fertil-
pregnancies) would be predicted to confer less ity medications has increased substantially over
protection against ovarian cancer than full-​term the past 20 years, further research is required.
46

466 Textbook of Cancer Epidemiology

Breastfeeding retrograde menstruation, then tubal ligation

Lactation may reduce risk of ovarian cancer by would block this process. In contrast, cells from
suppressing ovulation, but the results of individ- the fimbria of the tubes could still seed onto the
ual studies have mostly not suggested a strong ovarian surface (Sieh et  al, 2013). It might be
association. Meta-​analyses have suggested that expected that tubal ligation undertaken at an ear-
women who have ever breastfed have around a lier age would block more retrograde menstru-
20%–​25% reduction in risk of ovarian cancer (Ip ation and thus provide greater risk reduction.
et  al, 2009; Luan et  al, 2013), with greater risk However, studies that have looked at the timing
reductions associated with longer durations of of tubal ligation show that although the reduced
breastfeeding. However, there was moderate het- risk appears to persist for up to 30 years follow-
erogeneity across studies and a recent pooled ing surgery, it does not vary appreciably by age at
analysis of prospective studies did not observe which the procedure was undertaken (Sieh et al,
a statistically significant association for ovarian 2013). Thus questions remain about the biologi-
cancer overall or for any subtype (Wentzensen cal mechanisms responsible for this association.
et al, 2016), so the findings are not definitive. It is Tubal ligation has also been associated with
possible that observed variation relates to differ- a reduced risk of ovarian cancer among BRCA1
ences in breastfeeding practices across women in mutation carriers, suggesting this may be a fea-
different studies and the effect these differences sible option for high-​risk women who have com-
may have on duration of ovulation suppression. pleted their family to reduce their risk of ovarian
Because the suppression of ovulation is stronger cancer (Narod et al, 2001).
in the early postnatal period than after a longer While earlier studies found an inverse associ-
duration of breastfeeding, the reduction in risk ation between hysterectomy and ovarian cancer
would be expected to be greatest during the first risk, studies including women diagnosed more
few months after birth. In keeping with this, recently have not (Jordan et  al, 2013). It is not
Jordan et al (2010) demonstrated that for individ- clear why the association may have changed over
ual episodes of breastfeeding, reduction in risk time, but possible explanations include chang-
appeared to plateau at between 6 and 12 months ing indications for hysterectomy, changing rates
duration. Further work is required to clarify the of concurrent oophorectomy, different patterns
nature of the association between breastfeeding of use of estrogen-​only menopausal therapy, or
and ovarian cancer. bias. This relationship requires further clarifica-
tion. Presumably hysterectomy would also have
Tubal Ligation and Hysterectomy the effect described above of blocking retrograde
The relationship between tubal ligation and passage of carcinogens to the ovaries. It is there-
reduced risk of ovarian cancer is well established fore not clear why the effect of hysterectomy
with the majority of epidemiologic studies show- should differ from tubal ligation, although the
ing an inverse association. A  pooled analysis of difference in the average age at which women
13 case-​control studies found tubal ligation was have these procedures may be one explanation.
associated with a 29% overall risk reduction, with
stronger effects for the endometrioid (OR = 0.48) Hormones
and clear cell (OR = 0.52) subtypes than invasive Oral Contraceptive Use
serous cancers (OR  =  0.81) (Sieh et  al, 2013); There is consistent evidence that oral contra-
findings were very similar in a large prospective ceptive use is inversely associated with the risk
cohort (Gaitskell et al, 2015), in which a similarly of ovarian cancer. A pooled analysis of 45 epide-
reduced risk of cancers recorded as fallopian miological studies from 21 countries showed that
tube or peritoneal was also observed (Gaitskell women who had ever used oral contraceptives
et al, 2016). had an almost 30% reduction in risk of ovarian
The main proposed biological rationale for cancer compared to women who had never used
tubal ligation preventing ovarian cancer is that them (Collaborative Group on Epidemiological
it blocks the ascent of potential carcinogens, via Studies of Ovarian Cancer, 2008).
the fallopian tubes, to the ovaries (Cramer et al, The protective effect of oral contraceptives
1982). Indeed, this may explain the stronger increases with the duration of use, with around
protective effect of tubal ligation for endometri- a 20% reduction in risk per 5 years of use, and it
oid and clear cell cancers. If these cancers orig- appears to persist for at least 30 years after use,
inate mainly from endometriosis derived from although the effect does attenuate somewhat over
467

Ovarian Cancer 467

time (Collaborative Group on Epidemiological 1.10, 95% CI 1.01–​ 1.2) (Collaborative Group
Studies of Ovarian Cancer, 2008). The risk reduc- on Epidemiological Studies of Ovarian Cancer,
tion does not appear to vary according to the age 2015). That study found that risk was elevated
of first use. Similar reductions have also been irrespective of whether estrogen was given alone
reported for women who carry BRCA mutations or in combination with a progestin; however
in most studies that have evaluated this (Iodice another meta-​ analysis of 14 population-​ based
et al, 2010). studies (Pearce et  al, 2009)  found that risk was
There may be differences among the associa- lower in women who used combined MHT
tions between oral contraceptive use and differ- compared with those who used estrogen-​only
ent histologic types of ovarian tumors. In the MHT. There are insufficient data to determine
pooled analysis, the risks of serous, endometri- whether risk varies according to whether women
oid, and clear cell cancers were all significantly use sequential or continuous preparations of
reduced by a similar amount. However, the combined MHT.
risk reduction was smaller for mucinous can- It is likely that the risk associated with MHT
cers (particularly borderline mucinous cancers) varies according to the histological subtype of
and not statistically significant (Collaborative ovarian cancer. Using only data from prospective
Group on Epidemiological Studies of Ovarian studies, the pooled analysis found that MHT was
Cancer, 2008). Similar results were seen in a large associated with an increased risk of serous and
pooled analysis of cohort studies (Wentzensen endometrioid cancers (the two most common
et al, 2016). subtypes) but was inversely associated with clear
The mechanism of protection by oral contra- cell cancers (RR  =  0.75, 95% CI 0.57–​0.98) and
ceptives may relate to suppression of pituitary not significantly associated with mucinous can-
gonadotropins or inhibition of ovulation. Higher-​ cers (Collaborative Group on Epidemiological
dose preparations cause greater suppression of Studies of Ovarian Cancer, 2015). Similar pat-
gonadotropins than more recent lower-​ dose terns have since been reported from a larger
preparations. However, few studies have spe- pooled cohort analysis (Wentzensen et al, 2016).
cifically considered the effect of hormone dose. Why MHT increases ovarian cancer risk
Because the amount of estrogen in pill prepara- remains incompletely understood. Estrogens
tions halved between the 1960s and 1980s, the administered during menopause decrease pitui-
Collaborative study investigated the issue of dose tary hormone levels through negative feedback,
by considering calendar year of use and found and, under the gonadotropin hypothesis, this
that risk reductions were similar for use across all effect would be expected to protect against ovar-
time periods, suggesting that dose does not have ian cancer (Table 19-​1). However, estrogen can
an appreciable effect on the degree of risk reduc- stimulate cellular proliferation in fallopian tubes
tion. Furthermore, use of oral progestin-​ only and the endometrium, and if these tissues are, in
contraceptives, which do not consistently inhibit fact, the origin of some ovarian cancers, MHT
ovulation, may be associated with a reduced risk may increase ovarian cancer risk by driving cel-
of ovarian cancer (Anonymous, 1987), as may lular proliferation.
progestin-​ releasing intrauterine devices (Soini
et al, 2016), although these associations warrant Lifetime Number of Ovulatory Cycles
further investigation. Women who have experienced higher numbers
of ovulations are at increased risk of ovarian can-
Menopausal Hormone Therapy cer, as would be predicted under the incessant
Use of menopausal hormone therapy (MHT) ovulation hypothesis. The number of ovulatory
is associated with an increase in risk of ovarian cycles a woman has over her lifetime is prima-
cancer (Collaborative Group on Epidemiological rily determined by parity and by duration of oral
Studies of Ovarian Cancer, 2015; Pearce et  al, contraceptive use, as well as the age at onset and
2009; Wentzensen et  al, 2016). An individual cessation of menses. In addition to the large body
participant meta-​ analysis of 52 studies found of evidence supporting the inverse association
that current users had a 40% increase in risk of parity and oral contraceptive use with ovar-
irrespective of duration of use. Risk decreased ian cancer, studies have specifically examined
after cessation but remained elevated for at least the association of calculated number of lifetime
5  years after stopping in those who had been ovulatory cycles and ovarian cancer. These stud-
long-​term (5 or more years) users of MHT (RR ies observed a reduction in the risk of ovarian
468

468 Textbook of Cancer Epidemiology

TABLE 19-​1   MECHANISMS PROPOSED FOR OVARIAN CANCER PATHOGENESIS

Hypothesis Mechanisms proposed for Epidemiologic evidence Epidemiologic evidence

ovarian cancer pathogenesis consistent with hypotheses inconsistent with
hypotheses

Incessant ovulation Damage to ovarian Reduced risk with parity, No clear protective effect
epithelium from repeated breastfeeding, and oral of late menarche or early
ovulation leads to contraceptive use. menopause.
neoplasia. Increased risk with Greater protection from
increased lifetime ovulatory pregnancy than equivalent
cycles. reduction in ovulation.

Excess Excess gonadotropin Reduced risk with parity Menopausal hormone

gonadotropin secretion leads to increased and oral contraceptive use. replacement therapy
stimulation estrogen stimulation of (MHT) has been associated
ovarian epithelium. with increased risk, but
MHT reduces gonadotropin
levels.

Pregnancy High levels of progesterone Reduced risk with increas-

clearance in pregnancy clear the ing parity and possibly with
ovaries of transformed cells. higher age at first/​lower age
at last pregnancy; this effect
diminishes with time.

Exogenous Environmental carcinogens Tubal ligation reduces risk.

carcinogens gain access to ovaries Perineal talc use associated
through reproductive tract. with risk of ovarian cancer
in some studies.

Inflammation Local inflammation and Factors associated with Anti-​inflammatory drugs

inflammatory cytokines inflammation such not clearly associated with
damage the ovarian as ovulation, talc use, reduced risk. PID and
epithelium leading to endometriosis associated other infections not clearly
neoplasia. with increased risk. associated with risk

cancer among women with fewer lifetime ovu- recent prospective studies and the pooled cohort
latory cycles compared to women with a greater analysis have reported higher risks associated with
number of cycles (Schildkraut et al, 1998; Webb later age at natural menopause (Braem et al, 2010;
et al, 1998). Gates et al, 2010; Tsilidis et al, 2011; Wentzensen
Age at menarche and age at menopause are et  al, 2016), although this may be restricted to
determining factors of reproductive duration, nonmucinous cancers (Wentzensen et  al, 2016).
however, it is not clear whether either of these A  possible explanation for the inconsistent find-
is associated with ovarian cancer risk. A  meta-​ ings comes from an examination of the timing of
analysis of 27 studies (Gong et al, 2013) found a ovulation and ovarian cancer. Menstrual cycles
reduced risk in those who were older at menarche, around the age of menarche and menopause,
but a number of studies that were not included in which are more likely to be anovulatory, were less
the meta-​analysis (Mills et  al. 2004; Braem et  al, strongly associated with ovarian cancer risk than
2010; Ness et al, 2011) and a recent pooled analysis those in the peak reproductive years (20–​29 years
of 21 cohort studies (Wentzensen et al, 2016) did of age) (Purdie et al, 2003).
not find an association. Results are also inconsist- Although the number of ovulations appears
ent for age at menopause, although several more to play a role in the etiology of ovarian cancer,
469

Ovarian Cancer 469

it is not sufficient to explain the pathogenesis Anthropometric Measures

of ovarian cancer. This is because the reduction Obesity and greater height appear to modestly
in risk associated with factors such as parity increase the risk of ovarian cancer, although the
and oral contraceptive use is much greater than relation with obesity may differ by histological
would be predicted simply from their suppres- subtype and/​ or use of menopausal hormones.
sion of ovulation (Risch, 1998). Height has been fairly consistently associated
with ovarian cancer risk, with prospective studies
Proposed Hormonal Mechanisms showing an 8% increase in risk per 5 cm height
A number of theories have been put forward to (WCRF/​AICR, 2013). Results for obesity were
explain the pathogenesis of ovarian cancer (Table less consistent, but two large pooled analyses
19-​1). Fathalla (1971) proposed that incessant have now reported very similar positive asso-
ovulation causes repeated minor trauma to the ciations between obesity, measured as body-​mass
epithelial surface of the ovary with resultant index, and risk of endometrioid, clear cell, and
proliferation of ovarian epithelium to repair the mucinous ovarian cancer but no association for
ovulatory wound. If the fallopian tube is indeed invasive serous cancer. The first concluded the dif-
the site of origin for many ovarian cancers, then ferences by histology were not statistically signif-
the breakdown and repair of the ovarian sur- icant but the association was restricted to women
face would be a less likely causal mechanism who had not used postmenopausal hormones
for induction of malignancy. However, recur- (Collaborative Group on Epidemiological Studies
rent ovulation results in the repeated exposure of Ovarian Cancer, 2012), while the second con-
of both the ovarian epithelium and the fimbrial cluded the association was restricted to nonserous
tubal epithelium to hormone-​and prostaglan- cancers but did not differ by use of MHT (Olsen
din-​rich follicular fluid, which may be a stimulus et  al, 2013). A  subsequent Mendelian random-
for carcinogenesis. ization analysis using genetic markers of obe-
An alternative hypothesis, originally pro- sity to reduce the potential for confounding also
posed by Stadel (1975), implicates exposure of reported an association restricted to non-​high-​
ovarian surface epithelium to high levels of cir- grade serous cancers (Dixon et al, 2016).
culating pituitary gonadotropins, particularly Obesity affects sex-​steroid hormone levels
follicle-​
stimulating hormone (FSH). Animal via several pathways: in postmenopausal women
studies have shown that high gonadotropin levels endogenous estrogen is produced via the con-
can induce ovarian cancers, although these are version of androgens to estrogens in adipose tis-
usually stromal in origin and thus less relevant sue. Obesity also increases free estradiol levels by
to epithelial cancers in humans. Gonadotropins reducing SHBG capacity, and it increases andro-
also stimulate growth of human ovarian can- gen production in the adrenal gland (Farrow
cer cell lines (Simon et  al, 1983). Furthermore, et al, 1989). Obesity has been associated with an
gonadotropin secretion increases with age, peak- increased risk of hormone-​related cancers, par-
ing after menopause. This pattern resembles that ticularly endometrial and postmenopausal breast
of ovarian cancer incidence, although the latter cancers, and bowel cancers. It might therefore be
peak occurs approximately 15  years after the expected that any association with ovarian can-
peak in FSH levels (Cramer et al, 1983a). cer would be strongest for the endometrioid and
Adami et al (1994) suggested that pregnancy mucinous subtypes, and there is some suggestion
may have the effect of clearing cells in various of this. A difference by menopausal status and/​or
stages of malignant transformation from the ova- menopausal hormone use is also plausible, as any
ries; others have proposed that this effect could effects of endogenous estrogen from adipose tis-
be mediated by high progesterone levels in preg- sue in the postmenopausal period may be masked
nancy. Data to support this hypothesis come from by the much higher levels of exogenous estrogens
a primate study in which oral contraceptives con- from postmenopausal hormone therapy.
taining progestin were found to induce apoptosis
in the ovarian epithelium (Rodriguez et al, 1998). Physical Activity
It has also been suggested that androgens may be Findings regarding physical activity, sedentary
relevant to the etiology of ovarian cancer (Risch, behavior, and risk of ovarian cancer are con-
1998); however, epidemiological data have not tradictory. Vigorous physical activity can lead
convincingly supported this suggestion (Olsen to anovulation, luteal-​
phase insufficiency, and
et al, 2008; Ose et al, 2015). amenorrhea (Ellison et  al, 1993); thus, during
470

470 Textbook of Cancer Epidemiology

the reproductive years, vigorous activity may et al, 1987). The increase was particularly prom-
decrease ovarian cancer risk due to an increase inent among women who were of reproductive
in anovulatory cycles. It may also decrease estro- age at the time of the detonation and was not
gen levels, suppress chronic inflammation, and seen until 20 years after exposure.
enhance immune function, which would also Whether radiation treatment for cancers or
potentially reduce risk; however lower estrogen other disorders is associated with risk is unclear.
levels may lead to an increase in gonadotropins Among 150,000 women who received radiation
via hypothalamic feedback, thereby potentially therapy for cervical cancer, the incidence of ovar-
increasing ovarian cancer risk. ian cancer was nonsignificantly increased after
While several case-​ control studies have 10 years (RR = 1.4) (Boice et al, 1988). Radiation
reported an inverse association between rec- therapy for benign pelvic conditions was asso-
reational physical activity and ovarian cancer ciated with an 80% increase in the risk of ovar-
risk, results from prospective studies have been ian cancer in two studies (Doll & Smith, 1968;
mixed or null (Cannioto & Moysich, 2015). Wagoner, 1984)  but not another (Ryberg et  al,
Results for sedentary behavior are also incon- 1990), and a British study found no elevation
sistent, with some prospective studies (Patel in risk among women irradiated for ankylosing
et al, 2006; Hildebrand et al, 2015) and a recent spondylitis (Darby et al, 1987).
pooled analysis of case-​control studies (Cannioto Few studies have investigated associations
et  al, 2016)  reporting increased ovarian cancer with radiation for diagnostic imaging. In one
risk among those who are sedentary while oth- small case-​control study, a history of pelvic x-​
ers have found no association (Xiao et al, 2013). rays was associated with a nonsignificant 30%
Interpretation of these data is, however, hindered increase in risk of ovarian cancer overall, with a
by the differing instruments used to assess activ- significant excess risk seen for women of Jewish
ity and varying classifications of active and sed- ancestry (Harlap et  al, 2002). In contrast, no
entary behavior (Cannioto & Moysich, 2015). excess of ovarian cancer was observed in a study
of Chinese diagnostic x-​ ray workers (Wang
Infections et al, 1988).
There are few data to support an association
between infection and ovarian cancer. Some Occupation
early studies indicated an inverse association A potential link between ovarian cancer and
between ovarian cancer and a history of mumps asbestos exposure was first reported in 1960 in
(Menczer et al, 1979; Cramer et al, 1983b); how- a case series of women with asbestosis who also
ever, findings of other studies did not support had a diagnosis of ovarian cancer (Keal, 1960).
this (Chen et al, 1992; Merritt et al, 2008). Late Follow-​up studies of female workers employed
infection with Epstein Barr virus (Littman et al, in asbestos-​related industries have documented
2003)  has also been associated with increased excess deaths from ovarian cancer (Wignall &
risk of ovarian cancer, but these results have yet Fox, 1982). An excess of mortality from ovar-
to be confirmed. Some studies have also observed ian cancer has been noted among employees in
an association between pelvic inflammatory dis- healthcare industries that may entail exposure to
ease and ovarian cancer (Lin et  al, 2011), but hazardous substances such as radiation and che-
others have found no association (Merritt et  al, motherapeutic drugs (Sala et al, 1998). Potential
2008; Wu et al, 2009) and still others have found problems with these studies are that peritoneal
an association only with borderline cancers mesotheliomas and ovarian cancer may be dif-
(Rasmussen et al, 2017). ficult to distinguish and that histologic material
was not available for all cases. Some studies have
Ionizing Radiation suggested an increased risk of ovarian cancer
Ionizing radiation, particularly at higher doses, among specific occupations such as hairdressers
may modestly increase the risk of ovarian can- and beauticians, and teachers, but overall the data
cer. Follow-​
up studies of female survivors of are limited and inconclusive (Shen et al, 1998; Le
the atomic bomb explosions in Hiroshima and et  al, 2014). While there is some evidence that
Nagasaki suggest that the incidence of ovarian shiftwork might increase risks of breast and colo-
cancer was increased twofold among women rectal cancer, there is currently little to suggest
exposed to more than 100 rads compared to an association with ovarian cancer (Poole et  al,
women who had not been exposed (Tokuoka 2011); and although one small prospective study
471

Ovarian Cancer 471

reported an inverse association with sleep dura- showed a significant risk reduction (RR  =  0.8)
tion (Weiderpass et  al, 2012), this has yet to be associated with regular aspirin use and a simi-
confirmed. lar, but nonsignificant, reduction associated with
use of nonsteroidal anti-​ inflammatory drugs
Medical Conditions (NSAIDs).
and Treatment The potential link between inflammatory
Associations between a number of other gyneco- medical conditions, use of anti-​ inflammatory
logical conditions and ovarian cancer have been medications, and ovarian cancer risk led Ness
evaluated due to their potential shared hormonal and Cottreau (1999) to propose a possible role
origins or the physical effects of these conditions. for inflammation in the development of ovarian
A history of endometriosis has consistently cancer (Table 19-​1).
been linked to a two-​to threefold increase in
risk of endometrioid and clear cell ovarian can- Other Risk Factors
cers in both case-​control (Pearce et al, 2012) and Talcum Powder
cohort studies (Brinton et  al, 2005; Wang et  al, There is lack of agreement regarding the rela-
2014; Wentzensen et  al, 2016). Polycystic ovar- tionship between perineal talc use and ovarian
ian syndrome (PCOS) affects about 1% to 5% cancer. Talc is a natural mineral fiber similar in
of women and is characterized by androgen composition to asbestos, which has been linked
excess and an elevated LH (luteinizing hormone) to ovarian cancer. Talc fibers have been detected
to FSH ratio. In one study, an elevated ovarian in benign and malignant ovarian tissue, dem-
cancer risk was seen among women diagnosed onstrating that they can migrate up the female
with PCOS (RR  =  2.5), but the study included genital tract and lodge in the ovaries. However,
only seven cases with PCOS (Schildkraut et  al, no relation between reported levels of talc expo-
1996). Subsequent studies have not replicated sure and ovarian talc counts has been discerned
these findings (Brinton et  al, 2010; Gottschau (Heller et al, 1996). Studies in rats have not dem-
et  al, 2015), although one study found a 2.5-​ onstrated a consistent increase in ovarian cancer
fold increase in risk of borderline serous cancers among animals chronically exposed to aerosol
among women who reported PCOS (Olsen et al, talc (Boorman & Seely, 1995). However, these
2008). Data reporting links with conditions such data should be interpreted cautiously, given the
as ovarian cysts or uterine fibroids are too limited anatomic and physiologic differences between
and inconsistent to draw conclusions (Brinton rodents and humans.
et al, 2005; Pearce et al, 2012; Wang et al, 2014). The first case-​control study linking talc and
Apart from gynecological conditions and ovarian cancer, published in 1982, showed an
infections, another condition that has frequently almost twofold increase in the risk of ovarian
been evaluated in relation to ovarian cancer is cancer with any perineal talc use (Cramer et al,
diabetes mellitus. Evidence for an association has 1982). Subsequent case-​control studies, includ-
been inconsistent, but a meta-​analysis of obser- ing one by the same authors (Cramer et al, 2016),
vational studies found a small statistically signif- have also found positive associations, although
icant increase in risk (RR = 1.17) among women not all were statistically significant, there has been
with a diagnosis of diabetes compared to those little evidence of a dose-​response, and the magni-
without (Lee et al, 2013). It is possible that some tude of the observed risks was modest, with an
of the inconsistencies relate to treatment for dia- estimate of 1.24 from a pooled analysis of eight
betes. In particular, there has been suggestion studies (Terry et  al, 2013). No association was
that use of the diabetes drug metformin may seen for nongenital talc use. Case-​control studies
reduce the risk of ovarian cancer while insulin are potentially limited by recall bias, which might
and possibly sulfonylureas have been associated account for the observed positive associations.
with increased risks (Bodmer et al, 2011). The only prospective studies to address this ques-
Use of simple analgesics has been considered tion have found no overall association between
as a possible protective factor for ovarian cancer, talc use and ovarian cancer (Gertig et  al, 2000;
but the results are inconsistent. Acetaminophen Houghton et al, 2014).
use has been associated with risk reductions The association between talc use and ovar-
in the order of 30%–​50% (Bonovas et  al, 2006; ian cancer may vary by histological subtype, but
Baandrup et al, 2014), although not consistently again the data are inconsistent, with one prospec-
(Trabert et  al, 2014). The latter pooled analysis tive study reporting an association only for serous
472

TABLE 19-​2   EPIDEMIOLOGY OF OVARIAN CANCER:

RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Age ↑
Current tobacco smoking (mucinous cancers only) ↑
Parity ↓↓
Infertility ↑
Tubal ligation ↓
Oral contraceptive use ↓↓
Postmenopausal hormone use ↑
Family history of ovarian cancer ↑↑
Endometriosis (endometrioid and clear cell cancers only) ↑

Probable Relationship Exists, Based on Substantial Data

Age at menopause ↑
Breastfeeding ↓
Height ↑
Obesity (possibly excluding high-​grade serous cancers) ↑

Weak, if Any, Relationship Exists, Based on Substantial Data

Age at menarche –​

Inconsistent Findings or Limited Study to Date

Dietary factors: Dairy foods, lactose, galactose
Calcium, vitamin D
Coffee, tea
Dietary fat, eggs, red meat
Fruit, vegetables
Alcohol intake
Hysterectomy
Infertility drugs
Age at menopause
Pelvic infections
Anti-​inflammatory medication use
Ionizing radiation
Diabetes mellitus
PCOS

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑ ↑, moderate
to large increase in risk; ↓, slight to moderate decrease in risk; ↓ ↓, moderate to large increase in risk; –​, no association

cancers (Gertig et  al, 2000)  and a case-​control cancers (Terry et al, 2013). It would be expected
study reporting associations for serous and endo- that any association between talc and ovarian
metrioid cancers (Cramer et al, 2016), while the cancer would be restricted to women with a pat-
pooled analysis of case-​control studies reported ent genital tract in whom talc could conceivably
associations for all subtypes except mucinous migrate up to the ovaries. However, most studies
473

Ovarian Cancer 473

have not reported stronger associations among tumors only. It is likely that obesity has a modest
women who have not undergone tubal steriliza- effect on risk, although this may be limited to
tion (Gertig et al, 2000; Terry et al, 2013) and the the nonserous subtypes and/​or women who do
most recent study reported a stronger association not use menopausal hormones, while the possi-
among women who had undergone hysterectomy ble risk-​reduction associated with use of aspirin
or tubal sterilization (Cramer et al, 2016). has yet to be confirmed in prospective studies.
It is unclear whether physical activity influences
CONCLUSION ovarian cancer risk. Together with the poor
Older age and family history have strong positive prognosis and the lack of procedures availa-
associations with risk of ovarian cancer, while ble for population-​based routine screening, the
oral contraceptive use, increased parity, breast- lack of modifiable risk factors underscores the
feeding, and tubal ligation are inversely associ- need for further research on preventable causes
ated with risk (Table 19-​2). These factors provide of ovarian cancer in addition to advances in
evidence for more than one proposed mechanism treatment.
of ovarian cancer pathogenesis (Table 19-​1). Oral
contraceptive use and parity provide support for REFERENCES
both the incessant ovulation hypothesis and the Adami HO, Hsieh CC, Lambe M, Trichopoulos
gonadotropin hypothesis. Both factors induce D, Leon D, Persson I et  al. Parity, age at first
anovulation and are associated with decreases in childbirth, and risk of ovarian cancer. Lancet
1994;344:1250–​4.
gonadotropin concentrations.
Alsop K, Fereday S, Meldrum C, deFazio A, Emmanuel
The association between breastfeeding and
C, George J et  al. BRCA mutation frequency
ovarian cancer is consistent with the incessant
and patterns of treatment response in BRCA
ovulation hypothesis, but this, and the associ-
mutation-​positive women with ovarian cancer: a
ation with postmenopausal hormone use, con- report from the Australian Ovarian Cancer Study
tradicts the gonadotropin hypothesis. Use of Group. J Clin Oncol 2012;30:2654–​63.
postmenopausal hormones decreases circulating Anonymous. The reduction in risk of ovarian can-
levels of LH and FSH and thus would be expected cer associated with oral-​ contraceptive use:  the
to decrease the ovarian cancer risk. In contrast, Cancer and Steroid Hormone Study of the Centers
FSH is often within the follicular phase range of for Disease Control and National Institutes of
the normal menstrual cycle during lactation, but Health and Human Development. N Engl J
studies have generally observed an inverse asso- Med;1987;316:650–​5.
ciation between lactation and ovarian cancer. The Armstrong B, Doll R. Environmental factors and can-
biological mechanism for the effects of tubal liga- cer incidence and mortality in different countries
tion is unclear and may involve decreased access with special reference to dietary practices. Int J
of environmental carcinogens, changes in the Cancer 1975;15:617–​31.
hormonal milieu, or changes in blood flow to the Baandrup L, Friis S, Dehlendorff C, Andersen KK,
ovaries or the distal fallopian tubes. Olsen JH, Kjaer SK. Prescription use of parace-
Ovarian tumors are histologically heteroge- tamol and risk for ovarian cancer in Denmark. J
neous, resembling different epithelial surfaces Natl Can Inst 2014;106:dju111.
of the reproductive tract, and the different his- Bodmer M, Becker C, Meier C, Jick SS, Meier CR. Use
tologic subtypes have somewhat different etiolo- of metformin and the risk of ovarian cancer: a case-​
control analysis. Gynecol Oncol 2011;123:200–​4.
gies. Although a number of risk factors appear to
Boice J, Engholm G, Kleiner R, Blettner M, Stovall M,
be common to all subtypes, others such as smok-
Lisco H et  al. Radiation dose and second cancer
ing and endometriosis are specific to one or more
risk in patients treated for cancer of the cervix.
subtypes. The inability of many studies to stratify
Radiation Res 1988;116:3–​55.
tumors by subtype due to insufficient numbers Bonovas S, Filioussi K, Sitaras NM. Paracetamol use
of cases of the less common subtypes may thus and risk of ovarian cancer:  a meta-​analysis. Br J
explain the lack of consistent findings for some Clin Pharmacol 2006;62:113–​21.
risk factors. Boorman GA, Seely JC. The lack of an ovarian effect
Apart from oral contraceptive use and, pos- of lifetime talc exposure in F344/​ N rats and
sibly, breastfeeding, there are few readily mod- B6C3F1 mice. Regulatory Toxicol Pharmacol
ifiable risk factors for ovarian cancer. There is 1995;21:242–​3.
little evidence for a role of diet, while smoking Boyd J. Specific keynote:  hereditary ovarian can-
appears to increase risk of mucinous ovarian cer: what we know. Gynecol Oncol 2003;88:S8–​10.
47

474 Textbook of Cancer Epidemiology

Braem MG, Onland-​Moret NC, van den Brandt PA, including 25157 women with ovarian cancer
Goldbohm RA, Peeters PH, Kruitwagen RF et al. from 47 epidemiological studies. PLoS Medicine
Reproductive and hormonal factors in association 2012;9:e1001200.
with ovarian cancer in the Netherlands cohort Collaborative Group on Epidemiological Studies of
study. Am J Epidemiol 2010;172:1181–​9. Ovarian Cancer. Menopausal hormone use and
Brinton LA, Moghissi KS, Westhoff CL, Lamb EJ, ovarian cancer risk:  individual participant meta-​
Scoccia B. Cancer risk among infertile women analysis of 52 epidemiological studies. Lancet
with androgen excess or menstrual disorders 2015;385:1835–​42.
(including polycystic ovary syndrome). Fertil Cramer DW, Harlow BL, Willett WC, Welch WR, Bell
Steril 2010;94:1787–​92. DA, Scully RE et al. Galactose consumption and
Brinton LA, Sakoda LC, Sherman ME, Frederiksen metabolism in relation to the risk of ovarian can-
K, Kjaer SK, Graubard BI et  al. Relationship of cer. Lancet 1989;2:66–​71.
benign gynecologic diseases to subsequent risk Cramer DW, Kuper H, Harlow BL, Titus-​Ernstoff L.
of ovarian and uterine tumors. Cancer Epidemiol Carotenoids, antioxidants and ovarian cancer risk
Biomarkers Prev 2005;14:2929–​35. in pre-​and postmenopausal women. Int J Cancer
Brunner RL, Wactawski-​Wende J, Caan BJ, Cochrane 2001;94:128–​34.
BB, Chlebowski RT, Gass ML et  al. The effect of Cramer DW, Welch WR. Determinants of ovarian
calcium plus vitamin D on risk for invasive can- cancer risk. II. Inference regarding pathogenesis.
cer:  results of the Women’s Health Initiative J Natl Cancer Inst 1983a;71:717–​21.
(WHI) calcium plus vitamin D randomized clini- Cramer DW, Welch WR, Cassells S, Scully RE.
cal trial. Nutr Cancer 2011;63:827–​41. Mumps, menarche, menopause, and ovarian can-
Cancer Genome Atlas Research Network. Integrated cer. Am J Obstet Gynecol 1983b;147:1–​6.
genomic analyses of ovarian carcinoma. Nature Cramer DW, Welch WR, Scully RE, Wojciechowski
2011;474:609–​15. CA. Ovarian cancer and talc: a case-​control study.
Cannioto RA, LaMonte MJ, Risch HA, Hong CC, Cancer 1982;50:372–​6.
Sucheston-​Campbell LE, Eng KH et  al. Chronic Cramer DW, Vitonis AF, Terry KL, Welch WR, Titus
recreational physical inactivity and epithelial ovar- LJ. The association between talc use and ovarian
ian cancer risk: evidence from the Ovarian Cancer cancer:  a retrospective case-​control study in two
Association Consortium. Cancer Epidemiol US states. Epidemiology 2016;27:334–​46.
Biomarkers Prev 2016;25:1114–​24. Darby SC, Doll R, Gill SK, Smith PG. Long term mor-
Cannioto RA, Moysich KB. Epithelial ovarian can- tality after a single treatment course with X-​rays
cer and recreational physical activity:  a review in patients treated for ankylosing spondylitis. Br J
of the epidemiological literature and implica- Cancer 1987;55:179–​90.
tions for exercise prescription. Gynecol Oncol Dick ML, Siskind V, Purdie DM, Green AC.
2015;137:559–​73. Incomplete pregnancy and risk of ovarian can-
Chang ET, Lee VS, Canchola AJ, Clarke CA, Purdie cer: results from two Australian case-​control stud-
DM, Reynolds P et  al. Diet and risk of ovarian ies and systematic review. Cancer Causes Control
cancer in the California Teachers Study cohort. 2009;20:1571–​85.
Am J Epidemiol 2007;165:802–​13. Dixon SC, Nagle CM, Thrift AP, Pharoah PD,
Chen Y, Wu PC, Lang JH, Ge WJ, Hartge P, Brinton Pearce CL, Zheng W et  al. Adult body mass
LA. Risk factors for epithelial ovarian cancer in index and risk of ovarian cancer by subtype:  a
Beijing, China. Int J Epidemiol 1992;21:23–​9. Mendelian randomization study. Int J Epidemiol
Collaborative Group on Epidemiological Studies of 2016;45:884–​95.
Ovarian Cancer. Ovarian cancer and oral contra- Doll R, Smith PG. The long-​term effects of x irradia-
ceptives: collaborative reanalysis of data from 45 tion in patients treated for metropathia haemor-
epidemiological studies including 23,257 women rhagica. Br J Radiol 1968;41362–​8.
with ovarian cancer and 87,303 controls. Lancet Ellison P, Panter-​Brick C, Lipson SF, O’Rourke MT.
2008;371:303–​14. The ecological context of human ovarian function.
Collaborative Group on Epidemiological Studies Hum Reprod 1993;8:2248–​58.
of Ovarian Cancer. Ovarian cancer and smok- Endogenous Hormones Breast Cancer Collaborative
ing: individual participant meta-​analysis including Group. Circulating sex hormones and breast can-
28,114 women with ovarian cancer from 51 epide- cer risk factors in postmenopausal women: reanal-
miological studies. Lancet Oncol 2012;13:946–​56. ysis of 13 studies. Br J Cancer 2011;105:709–​22.
Collaborative Group on Epidemiological Studies Faber MT, Kjaer SK, Dehlendorff C, Chang-​Claude
of Ovarian Cancer. Ovarian cancer and body J, Andersen KK, Hogdall E et al. Cigarette smok-
size:  individual participant meta-​ analysis ing and risk of ovarian cancer: a pooled analysis
475

Ovarian Cancer 475

of 21 case-​control studies. Cancer Causes Control Genkinger JM, Hunter DJ, Spiegelman D, Anderson
2013;24:989–​1004. KE, Buring JE, Freudenheim JL et  al. Alcohol
Fairfield KM, Hankinson SE, Rosner BA, Hunter DJ, intake and ovarian cancer risk:  a pooled
Colditz GA, Willett WC. Risk of ovarian carci- analysis of 10 cohort studies. Br J Cancer
noma and consumption of vitamins A, C, and E 2006c;94:757–​62.
and specific carotenoids:  a prospective analysis. Gertig D, Hunter D, Cramer D, Colditz G, Speizer F,
Cancer 2001;92:2318–​26. Willett W et al. A prospective study of talc use and
Farrow DC, Weiss NS, Lyon JL, Daling JR. Association ovarian cancer. J Natl Cancer Inst 2000;92:249–​52.
of obesity and ovarian cancer in a case-​control Gong TT, Wu QJ, Vogtmann E, Lin B, Wang YL. Age
study. Am J Epidemiol 1989;129:1300–​4. at menarche and risk of ovarian cancer:  a meta-​
Fathalla MF. Incessant ovulation—​a factor in ovarian analysis of epidemiological studies. Int J Cancer
neoplasia? Lancet 1971;2:163. 2013;132:2894–​2900.
Fenster L, Quale C, Waller K, Windham G, Elkin E, Gottschau M, Kjaer SK, Jensen A, Munk C,
Benowitz N et al. Caffeine consumption and men- Mellemkjaer L. Risk of cancer among women
strual function. Am J Epidemiol 1999;149:550–​7. with polycystic ovary syndrome: a Danish cohort
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, study. Gynecol Oncol 2015;136:99–​103.
Mathers C, Rebelo M, Parkin DM, Forman D, Bray, Harlap S, Olson SH, Barakat RR, Caputo TA, Forment
F. GLOBOCAN 2012 v1.0, Cancer Incidence and S, Jacobs AJ et al. Diagnostic x-​rays and risk of epi-
Mortality Worldwide: IARC CancerBase No. 11 thelial ovarian carcinoma in Jews. Ann Epidemiol
[Internet]. Lyon, France: International Agency for 2002;12:426–​34.
Research on Cancer; 2013. Available from: http:// Hedelin M, Lof M, Andersson TM, Adlercreutz H,
globocan.iarc.fr, accessed 2017. Weiderpass E. Dietary phytoestrogens and the
Fung WL, Risch H, McLaughlin J, Rosen B, Cole D, risk of ovarian cancer in the women’s lifestyle and
Vesprini D et  al. The N314D polymorphism of health cohort study. Cancer Epidemiol Biomarkers
galactose-​1-​
phosphate uridyl transferase does Prev 2011;20:308–​17.
not modify the risk of ovarian cancer. Cancer Heller DS, Westhoff C, Gordon RE, Katz N. The rela-
Epidemiol Biomarkers Prev 2003;12:678–​80. tionship between perineal cosmetic talc usage and
Gaitskell K, Green J, Pirie K, Reeves G, Valerie B, on ovarian talc particle burden. Am J Obstet Gynecol
behalf of the Million Women Study C. Tubal liga- 1996;174:1507–​10.
tion and ovarian cancer risk in a large cohort: sub- Herrinton LJ, Stanford JL, Schwartz SM, Weiss NS.
stantial variation by histological type. Int J Cancer Ovarian cancer incidence among Asian migrants
2015;138:1076–​84. to the United States and their descendants. J Natl
Gaitskell K, Coffey K, Green J, Pirie K, Reeves GK, Cancer Inst 1994;86:1336–​9.
Ahmed AA, Barnes I, Beral V. Tubal ligation and Hildebrand JS, Gapstur SM, Gaudet MM, Campbell
incidence of 26 site-​specific cancers in the Million PT, Patel AV. Moderate-​ to-​
vigorous physical
Women Study. Br J Cancer 2016;114:1033–​7. activity and leisure-​time sitting in relation to ovar-
Garrett AP, Lee KR, Colitti CR, Muto MG, Berkowitz ian cancer risk in a large prospective US cohort.
RS, Mok SC. k-​ras mutation may be an early event Cancer Causes Control 2015;26:1691–​97.
in mucinous ovarian tumorigenesis. Int J Gynecol Ho CL, Kurman RJ, Dehari R, Wang TL, Shih Ie M.
Pathol 2001;20:244–​51. Mutations of BRAF and KRAS precede the devel-
Gates MA, Rosner BA, Hecht JL, Tworoger SS. Risk opment of ovarian serous borderline tumors.
factors for epithelial ovarian cancer by histologic Cancer Res 2004;64:6915–​8.
subtype. Am J Epidemiol 2010;171:45–​53. Houghton SC, Reeves KW, Hankinson SE, Crawford
Gemignani ML, Schlaerth AC, Bogomolniy F, Barakat L, Lane D, Wactawski-​Wende J et al. Perineal pow-
RR, Lin O, Soslow R et al. Role of KRAS and BRAF der use and risk of ovarian cancer. J Natl Cancer
gene mutations in mucinous ovarian carcinoma. Inst 2014;106(9).
Gynecol Oncol 2003;90:378–​81. Huncharek M, Klassen H, Kupelnick B. Dietary beta-​
Genkinger JM, Hunter DJ, Spiegelman D, Anderson carotene intake and the risk of epithelial ovarian
KE, Arslan A, Beeson WL et  al. Dairy products cancer: a meta-​analysis of 3,782 subjects from five
and ovarian cancer: a pooled analysis of 12 cohort observational studies. In Vivo 2001;15:339–​43.
studies. Cancer Epidemiol Biomarkers Prev Iodice S, Barile M, Rotmensz N, Feroce I, Bonanni B,
2006a;15:364–​72. Radice P et al. Oral contraceptive use and breast
Genkinger JM, Hunter DJ, Spiegelman D, Anderson or ovarian cancer risk in BRCA1/​2 carriers:  a
KE, Beeson WL, Buring JE et al. A pooled analy- meta-​analysis. Eur J Cancer 2010;46:2275–​84.
sis of 12 cohort studies of dietary fat, cholesterol Ip S, Chung M, Raman G, Trikalinos TA, Lau J. A
and egg intake and ovarian cancer. Cancer Causes summary of the Agency for Healthcare Research
Control 2006b;17:273–​85. and Quality’s evidence report on breastfeeding in
476

476 Textbook of Cancer Epidemiology

developed countries. Breastfeed Med 2009;4 Suppl Koushik A, Hunter DJ, Spiegelman D, Anderson KE,
1:S17–​30. Arslan AA, Beeson WL et al. Fruits and vegetables
Jensen A, Sharif H, Frederiksen K, Kjaer SK. Use of and ovarian cancer risk in a pooled analysis of
fertility drugs and risk of ovarian cancer: Danish 12 cohort studies. Cancer Epidemiol Biomarkers
Population Based Cohort Study. Br Med Journal Prev 2005;14:2160–​7.
2009;338:b249. Koushik A, Hunter DJ, Spiegelman D, Anderson KE,
Jervis S, Song H, Lee A, Dicks E, Tyrer J, Harrington P Buring JE, Freudenheim JL et  al. Intake of the
et al. Ovarian cancer familial relative risks by tumour major carotenoids and the risk of epithelial ovar-
subtypes and by known ovarian cancer genetic sus- ian cancer in a pooled analysis of 10 cohort stud-
ceptibility variants. J Med Genet 2014;51:108–​13. ies. Int J Cancer 2006;119:2148–​54.
Ji J, Sundquist J, Sundquist K. Lactose intolerance and Kuchenbaecker KB, Ramus SJ, Tyrer J, Lee A, Shen
risk of lung, breast and ovarian cancers:  aetio- HC, Beesley J et al. Identification of six new sus-
logical clues from a population-​based study in ceptibility loci for invasive epithelial ovarian can-
Sweden. Br J Cancer 2015;112:149–​52. cer. Nat Genet 2015;47:164–​71.
Jordan SJ, Nagle CM, Coory MD, Maresco D, Kurman R, Carcangiu M, Herrington C, Young R,
Protani MM, Pandeya NA et  al. Has the associ- eds. WHO Classification of Tumours of Female
ation between hysterectomy and ovarian can- Reproductive Organs. World Health Organization
cer changed over time? A  systematic review and Classification of Tumours. Lyon: IARC, 2014.
meta-​analysis. Eur J Cancer 2013;49:3638–​47. Kushi LH, Mink PJ, Folsom AR, Anderson KE,
Jordan SJ, Siskind V, A CG, Whiteman DC, Webb PM. Zheng W, Lazovich D et  al. Prospective study
Breastfeeding and risk of epithelial ovarian cancer. of diet and ovarian cancer. Am J Epidemiol
Cancer Causes Control 2010;21:109–​16. 1999;149:21–​31.
Kato I, Tominaga S, Kuroishi T. Relationship between Larsson SC, Giovannucci E, Wolk A. Dietary folate
westernization of dietary habits and mortality intake and incidence of ovarian cancer:  the
from breast and ovarian cancers in Japan. Jpn J Swedish Mammography Cohort. J Natl Cancer
Cancer Res 1987;78:349–​57. Inst 2004;96:396–​402.
Keal E. Asbestosis and abdominal neoplasms. Lancet Le ND, Leung A, Brooks-​Wilson A, Gallagher RP,
1960;2:1211–​16. Swenerton KD, Demers PA et  al. Occupational
Kelemen L, Sellers T, Vierkant R, Harnack L, Cerhan exposure and ovarian cancer risk. Cancer Cause
J. Association of folate and alcohol with risk of Control 2014;25:829–​41.
ovarian cancer in a prospective study of post- Lee J-​Y, Jeon I, Kim JW, Song Y-​S, Yoon J-​M, Park SM.
menopausal women. Cancer Causes and Control Diabetes mellitus and ovarian cancer risk:  a sys-
2004;15:1085–​93. tematic review and meta-​analysis of observational
Kelemen LE, Bandera EV, Terry KL, Rossing MA, studies. Int J Gynecol Cancer 2013;23:402–​12.
Brinton LA, Doherty JA et  al. Recent alcohol Lin HW, Tu YY, Lin SY, Su WJ, Lin WL, Lin WZ et al.
consumption and risk of incident ovarian carci- Risk of ovarian cancer in women with pelvic
noma: a pooled analysis of 5,342 cases and 10,358 inflammatory disease:  a population-​based study.
controls from the Ovarian Cancer Association Lancet Oncol 2011;12:900–​4.
Consortium. BMC Cancer 2013;13:28. Littman AJ, Rossing MA, Madeleine MM, Tang MT,
Kelemen LE, Lawrenson K, Tyrer J, Li Q, Lee JM, Seo Yasui Y. Association between late age at infec-
JH et  al. Genome-​wide significant risk associa- tious mononucleosis, Epstein-​Barr virus antibod-
tions for mucinous ovarian carcinoma. Nat Genet ies, and ovarian cancer risk. Scand J Infect Dis
2015;47:888–​97. 2003;35:728–​35.
Kessous R, Davidson E, Meirovitz M, Sergienko R, Luan NN, Wu QJ, Gong TT, Vogtmann E, Wang YL,
Sheiner E. The risk of female malignancies after fer- Lin B. Breastfeeding and ovarian cancer risk:  a
tility treatments: a cohort study with 25-​year follow-​ meta-​analysis of epidemiologic studies. Am J Clin
up. J Cancer Res Clin Oncol 2016;142:287–​93. Nutr 2013;98:1020–​31.
Khaw K, Tazuke S, Barrett-​Connor E. Cigarette smok- Lucero J, Harlow BL, Barbieri RL, Sluss P, Cramer DW.
ing and levels of adrenal androgens in postmeno- Early follicular phase hormone levels in relation to
pausal women. New Engl J Med 1988;318:1705–​9. patterns of alcohol, tobacco and coffee use. Fertil
Kotsopoulos J, Lubinski J, Gronwald J, Cybulski C, Steril 2001;76:723–​9.
Demsky R, Neuhausen SL et  al. Factors influ- Lynch HT, Casey M, Lynch J, Tacey EK. Genetics and
encing ovulation and the risk of ovarian cancer ovarian carcinoma. Semin Oncol 1998;25:265–​80.
in BRCA1 and BRCA2 mutation carriers. Int J McKinlay SM. The normal menopause transition: an
Cancer 2015;137:1136–​46. overview. Maturitas 1996;23:137–​45.
47

Ovarian Cancer 477

Menczer J, Modon M, Ranon L. Possible role of Investigation into Cancer and Nutrition. Int J
mumps virus in the etiology of ovarian cancer. Cancer 2015;136:399–​410.
Cancer 1979;43:1375. Patel AV, Rodriguez C, Pavluck AL, Thun MJ, Calle EE.
Merritt MA, Green AC, Nagle CM, Webb PM. Talcum Recreational physical activity and sedentary behav-
powder, chronic pelvic inflammation and NSAIDs ior in relation to ovarian cancer risk in a large cohort
in relation to risk of epithelial ovarian cancer. Int J of US women. Am J Epidemiol 2006;163:709–​16.
Cancer 2008;122:170–​6. Pearce CL, Chung K, Pike MC, Wu AH. Increased
Merritt MA, Poole EM, Hankinson SE, Willett WC, ovarian cancer risk associated with menopausal
Tworoger SS. Dairy food and nutrient intake in estrogen therapy is reduced by adding a progestin.
different life periods in relation to risk of ovarian Cancer 2009;115:531–​9.
cancer. Cancer Cause Control 2014;25:795–​808. Pearce CL, Templeman C, Rossing MA, Lee A, Near
Merritt MA, Tzoulaki I, van den Brandt PA, Schouten AM, Webb PM, et al. Association between endo-
LJ, Tsilidis KK, Weiderpass E et al. Nutrient-​wide metriosis and risk of histological subtypes of
association study of 57 foods/​nutrients and epi- ovarian cancer: a pooled analysis of case–​control
thelial ovarian cancer in the European Prospective studies. Lancet Oncol 2012;13:385–​94.
Investigation into Cancer and Nutrition study and Petridou E, Katsouyanni K, Spanos E, Skalkidis Y,
the Netherlands Cohort Study. Am J Clin Nutr Panagiotopoulou K, Trichopoulos D. Pregnancy
2016;103:161–​7. estrogens in relation to coffee and alcohol intake.
Mills PK, Riordan DG, Cress RD. Epithelial ovar- Ann Epidemiol 1992;2:241–​7.
ian cancer risk by invasiveness and cell type in Pirozzo S, Purdie D, Kuiper-​Linley M, Webb P, Harvey
the Central Valley of California. Gynecol Oncol P, Green A et al. Ovarian cancer, cholesterol, and
2004;95:215–​25. eggs:  a case-​control analysis. Cancer Epidemiol
Narod SA, Sun P, Ghadirian P, Lynch H, Isaacs C, Biomarkers Prev 2002;11:1112–​4.
Garber J et al. Tubal ligation and risk of ovarian can- Poole EM, Schernhammer ES, Tworoger SS. Rotating
cer in carriers of BRCA1 or BRCA2 mutations: a night shift work and risk of ovarian cancer. Cancer
case-​control study. Lancet 2001;357:1467–​70. Epidemiol Biomarkers Prev 2011;20:934–​8.
Negri E, Franceschi S, Tzonou A, Booth M, La Vecchia Prat J. New insights into ovarian cancer pathology.
C, Parazzini F et al. Pooled analysis of 3 European Ann Oncol 2012;23 Suppl 10:x111–​117.
case-​control studies. I.  Reproductive factors and Purdie DM, Bain CJ, Siskind V, Webb PM, Green AC.
risk of epithelial ovarian cancer. Int J Cancer Ovulation and risk of epithelial ovarian cancer. Int
1991;49:50–​6. J Cancer 2003;104:228–​32.
Ness R, Cottreau C. Possible role of ovarian epithelial Ramus SJ, Song H, Dicks E, Tyrer JP, Rosenthal AN,
inflammation in ovarian cancer. J Natl Cancer Inst Intermaggio MP et  al. Germline Mutations in
1999;91:1459–​67. the BRIP1, BARD1, PALB2, and NBN genes in
Ness RB, Cramer DW, Goodman MT, Kjaer SK, Mallin women with ovarian cancer. J Natl Cancer Inst
K, Mosgaard BJ et al. Infertility, fertility drugs, and 2015;107:djv214.
ovarian cancer: a pooled analysis of case-​control Rasmussen CB, Kjaer SK, Albieri V, Bandera EV,
studies. Am J Epidemiol 2002;155:217–​24. Doherty JA, Hogdall E et al. Pelvic inflammatory
Ness RB, Dodge RC, Edwards RP, Baker JA, Moysich disease and risk of ovarian cancer and border-
KB. Contraception methods, beyond oral con- line ovarian tumors: a pooled analysis of 13 case-​
traceptives and tubal ligation, and risk of ovarian control studies. Am J Epidemiol 2017;185:8–​20.
cancer. Ann Epidemiol 2011;21:188–​96. Rebbeck TR, Mitra N, Wan F, Sinilnikova OM, Healey
Olsen CM, Green AC, Nagle CM, Jordan SJ, Whiteman S, McGuffog L et al. Association of type and loca-
DC, Bain CJ et al. Epithelial ovarian cancer: test- tion of BRCA1 and BRCA2 mutations with risk of
ing the “androgens hypothesis.” Endocr Relat breast and ovarian cancer. J Am Med Association
Cancer 2008;15:1061–​68. 2015;313:1347–​61.
Olsen CM, Nagle CM, Whiteman DC, Ness R, Pearce Risch HA. Hormonal etiology of epithelial ovarian
CL, Pike MC et al. Obesity and risk of ovarian can- cancer, with a hypothesis concerning the role of
cer subtypes:  evidence from the Ovarian Cancer androgens and progesterone. J Natl Cancer Inst
Association Consortium. Endocr Relat Cancer 1998;90:1744–​86.
2013;20:251–​62. Risch HA, Marrett LD, Howe GR. Parity, contracep-
Ose J, Fortner RT, Rinaldi S, Schock H, Overvad tion, infertility, and the risk of epithelial ovarian
K, Tjonneland A et  al. Endogenous androgens cancer. Am J Epidemiol 1994;140:585–​97.
and risk of epithelial invasive ovarian cancer by Rizzuto I, Behrens RF, Smith LA. Risk of ovarian can-
tumor characteristics in the European Prospective cer in women treated with ovarian stimulating
478

478 Textbook of Cancer Epidemiology

drugs for infertility. Cochrane Database Syst Rev Stadel BV. The etiology and prevention of ovar-
2013;8:CD008215. ian cancer [letter]. Am J Obstet Gynecol
Rodriguez G, Walmer D, Cline M, Krigman H, Lessey B, 1975;123:772–​3.
Whitaker R et al. Effect of progestin on the ovarian Stratton J, Pharoah P, Smith SK, Easton DF, Ponder
epithelium of macaques: cancer prevention through BA. A systematic review and meta-​ analysis of
apoptosis? J Soc Gynecol Invest 1998;5:271–​6. family history and risk of ovarian cancer. Br J
Rose DP, Boyar AP, Wynder EL. International com- Obstet Gynaecol 1998;105:493–​9.
parisons of mortality rates for cancer of the breast, Terry KL, Karageorgi S, Shvetsov YB, Merritt MA,
ovary, prostate, and colon, and per capita food Lurie G, Thompson PJ et al. Genital powder use
consumption. Cancer 1986;58:2363–​71. and risk of ovarian cancer:  a pooled analysis of
Ryberg M, Lundell M, Nilsson B, Pettersson F. 8,525 cases and 9,859 controls. Cancer Prev Res
Malignant disease after radiation treatment of 2013;6:811–​21.
benign gynaecological disorders. Acta Oncologica Tokuoka S, Kawai K, Shimizu Y, Inai K, Ohe K,
1990;29:563–​7. Fujikura T et  al. Malignant and benign ovar-
Sala M, Dosemeci M, Zahm S. A death certificate–​ ian neo-​plasms among atomic bomb survivors,
based study of occupation and mortality from Hiroshima and Nagasaki, 1950–​80. J Natl Cancer
reproductive cancers among women in 24 US Inst 1987;79:47–​57.
states. J Occup Environ Med 1998;40:632–​9. Tothill RW, Tinker AV, George J, Brown R, Fox SB,
Schildkraut J, Bastos E, Berchuck A. Relationship Lade S et al. Novel molecular subtypes of serous
between lifetime ovulatory cycles and overexpres- and endometrioid ovarian cancer linked to clini-
sion of mutant p53 in epithelial ovarian cancer. J cal outcome. Clin Cancer Res 2008;14:5198–​5208.
Natl Cancer Inst 1998;90:1729–​34. Trabert B, Lamb EJ, Scoccia B, Moghissi KS, Westhoff
Schildkraut JM, Schwingl PJ, Bastos E, Evanoff A, CL, Niwa S et  al. Ovulation-​inducing drugs and
Hughes C. Epithelial ovarian cancer risk among ovarian cancer risk:  results from an extended
women with polycystic ovary syndrome. Obstet follow-​ up of a large United States infertility
Gynecol 1996;88:554–​9. cohort. Fertil Steril 2013;100:1660–​6.
Setiawan VW, Yang HP, Pike MC, McCann SE, Yu H, Trabert B, Ness RB, Lo-​Ciganic WH, Murphy MA,
Xiang YB et  al. Type I  and II endometrial can- Goode EL, Poole EM et  al. Aspirin, nonaspirin
cers: have they different risk factors? J Clin Oncol nonsteroidal anti-​ inflammatory drug, and ace-
2013;31:2607–​18. taminophen use and risk of invasive epithelial
Shen N, Weiderpass E, Antilla A, Goldberg MS, ovarian cancer: a pooled analysis in the Ovarian
Vasama-​ Neuronenk KM, Boffetta P et  al. Cancer Association Consortium. J Natl Cancer
Epidemiology of occupational and environmen- Inst 2014;106:djt431.
tal risk factors related to ovarian cancer. Scand J Tsilidis KK, Allen NE, Key TJ, Dossus L, Kaaks R,
Work, Environ Health 1998;24:175–​82. Bakken K et al. Menopausal hormone therapy and
Shih I-​M, Kurman RJ. Ovarian tumorigenesis: a pro- risk of ovarian cancer in the European prospective
posed model based on morphological and molecu- investigation into cancer and nutrition. Cancer
lar genetic analysis. Am J Pathol 2004;164:1511–​8. Causes Control 2011;22:1075-​84.
Sieh W, Salvador S, McGuire V, Palmieri R, Terry Tworoger SS, Fairfield KM, Colditz GA, Rosner BA,
K, Rossing M et  al. Tubal ligation and the risk Hankinson SE. Association of oral contracep-
of ovarian cancer subtypes. Int J Epidemiol tive use, other contraceptive methods, and infer-
2013;42:579–​89. tility with ovarian cancer risk. Am J Epidemiol
Simon WE, Albrecht M, Hansel M, Dietel D, Hotzel F. 2007;166:894–​901.
Cell lines derived from human ovarian carcinomas: Tworoger SS, Hecht JL, Giovannucci E, Hankinson SE.
growth stimulation by gonadotrophic and steroid Intake of folate and related nutrients in relation to
hormones. J Natl Cancer Inst 1983;70:839–​45. risk of epithelial ovarian cancer. Am J Epidemiol
Soini T, Hurskainen R, Grénman S, Mäenpää J, 2006;163:1101–​11.
Paavonen J, Pukkala E. Impact of levonorgestrel-​ Venn A, Watson L, Bruinsma F, Giles G, Healy D. Risk
releasing intrauterine system use on the cancer of cancer after use of fertility drugs with in-​vitro
risk of the ovary and fallopian tube. Acta Oncol fertilisation. Lancet 1999;354:1586–​90.
2016;55:1281–​4. Wagoner J. Leukaemia and Other Malignancies
Song H, Dicks E, Ramus SJ, Tyrer JP, Intermaggio MP, Following Radiation Therapy for Gynecologic
Hayward J et al. Contribution of germline muta- Disorders. New York: Raven Press, 1984.
tions in the RAD51B, RAD51C, and RAD51D Wang J-​X, Boice J, Li B-​X, Zhang J-​Y, Fraumeni JF.
genes to ovarian cancer in the population. J Clin Cancer among medical diagnostic X-​ray workers
Oncol 2015;33:2901–​7. in China. J Natl Cancer Inst 1988;80:344–​50.
479

Ovarian Cancer 479

Wang KC, Chang WH, Lee WL, Huang N, Huang endometriosis-​associated ovarian carcinomas. N
HY, Yen MS et al. An increased risk of epithelial Engl J Med 2010;363:1532–​43.
ovarian cancer in Taiwanese women with a new Wignall BK, Fox AJ. Mortality of female gas mask
surgico-​pathological diagnosis of endometriosis. assemblers. Br J Indust Med 1982;39:34–​8.
BMC Cancer 2014;14:831. World Cancer Research Fund/​ American Institute
Webb PM, Green A, Cummings MC, Purdie DM, for Cancer Research (WCRF/​AICR). Systematic
Walsh MD, Chevenix-​ Trench G. Relationship Literature Review Continuous Update Project
between number of ovulatory cycles and accumu- Report: The Associations Between Food, Nutrition
lation of mutant p53 in epithelial ovarian cancer. J and Physical Activity and the Risk of Ovarian
Natl Cancer Inst 1998;90:1729–​34. Cancer. 2013a. http://​www.wcrf.org/​sites/​default/​
Weiderpass E, Sandin S, Inoue M, Shimazu T, Iwasaki files/​Ovarian-​Cancer-​SLR-​2013.
M, Sasazuki S et al. Risk factors for epithelial ovar- World Cancer Research Fund/​American Institute for
ian cancer in Japan—​results from the Japan Public Cancer Research (WCRF/​ AICR). Continuous
Health Center-​Based Prospective Study cohort. Update Project Report:  Food, Nutrition, Physical
Int J Oncol 2012;40:21–​30. Activity and the Prevention of Endometrial Cancer.
Weiss NS. Measuring the separate effects of low par- 2013b. http://​www.dietandcancerreport.org.
ity and its antecedents on the incidence of ovarian World Cancer Research Fund/​American Institute for
cancer. Am J Epidemiol 1988;128:451–​5. Cancer Research (WCRF/​ AICR). Continuous
Wentzensen N, Poole EM, Trabert B, White E, Arslan Update Project Report: Food, Nutrition, Physical
AA, Patel AV et  al. Ovarian cancer risk fac- Activity, and the Prevention of Ovarian Cancer.
tors by histologic subtype:  an analysis from the 2014. http://​www.dietandcancerreport.org/​cup/​
Ovarian Cancer Cohort Consortium. J Clin Oncol cup_​resources.php.
2016;34:2888–​98. Wu AH, Pearce CL, Tseng CC, Templeman C, Pike
Westhoff C, Gentile G, Lee J, Zacur H, Helbig D. MC. Markers of inflammation and risk of ovar-
Predictors of ovarian steroid secretion in reproduc- ian cancer in Los Angeles County. Int J Cancer
tive age women. Am J Epidemiol 1996;144:381–​8. 2009;124:1409–​15.
Whiteman DC, Murphy MF, Cook LS, Cramer DW, Xiao Q, Yang HP, Wentzensen N, Hollenbeck A,
Hartge P, Marchbanks PA et  al. Multiple births Matthews CE. Physical activity in different peri-
and risk of epithelial ovarian cancer. J Natl Cancer ods of life, sedentary behavior, and the risk
Inst 2000;92:1172–​7. of ovarian cancer in the NIH-​ AARP diet and
Whiteman DC, Siskind V, Purdie DM, Green AC. health study. Cancer Epidemiol Biomarkers Prev
Timing of pregnancy and the risk of epithelial 2013;22:2000–​8.
ovarian cancer. Cancer Epidemiol Biomarkers Yang CS, Maliakal P, Meng X. Inhibition of carcin-
Prev 2003;12:42–​6. ogenesis by tea. Annu Rev Pharmacol Toxicol
Whittemore AS, Harris R, Itnyre J. Characteristics 2002;42:25–​54.
relating to ovarian cancer risk: collaborative anal- Zhang M, Binns CW, Lee AH. Tea consumption
ysis of 12 U.S.  case-​control studies. II. Invasive and ovarian cancer risk:  a case-​ control study
epithelial ovarian cancers in white women. in China. Cancer Epidemiol Biomarkers Prev
Collaborative Ovarian Cancer Group [see com- 2002a;11:713–​8.
ments]. Am J Epidemiol 1992;136:1184–​203. Zhang M, Yang ZY, Binns CW, Lee AH. Diet and ovar-
Wiegand KC, Shah SP, Al-​ Agha OM, Zhao Y, ian cancer risk: a case-​control study in China. Br J
Tse K, Zeng T et  al. ARID1A mutations in Cancer 2002b;86:712–​7.
480
481

20
Prostate Cancer
K AT H R Y N M . W I L S O N A N D L O R E L E I   M U C C I

T here are two giants in the cancer landscape

of women and men in Western countries,
each of them accounting for approximately one-​
generally greater than 4.0 ng/​mL, or an abnor-
mal digital rectal exam, would indicate follow-​up
core biopsy to confirm cancer. Staging is done
fourth of all incident malignancies: breast cancer by digital rectal examination (DRE), transrectal
in women and prostate cancer in men. Breast ultrasound, and skeletal scintigraphy. Grading
cancer has been investigated in an overwhelming of the degree of tumor differentiation is assessed
number of epidemiologic studies, beginning in most commonly using the Gleason score. A con-
the 1920s. Until rather recently, however, pros- troversial 2012 recommendation against routine
tate cancer remained a hidden disease. Its causes PSA screening by the US Preventive Services
were enigmatic, little epidemiologic research Task Force seems to have resulted in a decrease
was being done, and no curative treatment could in screening and in diagnosis of early-​stage pros-
be offered. Starting in the 1980s, this situation tate cancer (Jemal et al, 2015).
changed dramatically. The number of epidemi- Treatment Treatment of prostate cancer is com-
ological studies is now astounding, attempts to plex and differs greatly among doctors, hospitals,
achieve early diagnosis and treatment are wide- and countries. Options vary from active surveil-
spread, and several renowned men with prostate lance, to radical local treatment in patients with
cancer have given this common malignancy an localized prostate cancer, to systemic androgen
identity. A  fascinating story of the increasing deprivation therapy (ADT) in advanced disease.
complexity of the etiology and progression of this Treatment with a curative intent—​radical pros-
disease is emerging. tatectomy or local irradiation—​was introduced
first during the 1980s without evidence of sur-
CLINICAL SYNOPSIS vival benefit from randomized trials; however,
Subgroups Almost all (99%) prostate cancers are 18  years of follow-​up from a Swedish trial of
adenocarcinomas derived from the glandular watchful waiting versus radical prostatectomy
epithelial cells. Data has begun to emerge sug- has demonstrated a significant reduction in mor-
gesting unique molecular subtypes of prostate tality associated with treatment, especially in
cancer, the most common of which is the gene men less than 65 years of age. Local irradiation
fusion TMPRSS2:ERG. can effectively alleviate pain from skeletal metas-
Symptoms The general clinical symptoms—​ tases, whereas cytotoxic treatment plays a limited
which cannot be distinguished from those role. Survival in patients with metastatic prostate
caused by benign prostatic hyperplasia—​are uri- cancer has also been significantly prolonged with
nary frequency, nocturia, and urgency caused by newer agents (abiraterone, enzalutamide) and
obstruction of the urethra. In rare cases, initial by using docetaxol chemotherapy earlier in men
symptoms come from painful skeletal metastases who are still responding to ADT.
or, more rarely, from anemia due to bone marrow Prognosis Prognosis varies greatly across time
replacement. The majority of cases in the current periods and countries. Survival has increased
clinical domain present without symptoms. in many countries since the mid-​1990s, but still
Diagnosis Since the late 1980s, most prostate varies a great deal, ranging from less than 60%
cancers are diagnosed in asymptomatic men 5-​year survival in Bulgaria and Thailand to 95%
through screening with prostate-​ specific anti- or more in Brazil, Puerto Rico, and the United
gen (PSA) in the blood. An elevated PSA level, States (Allemani et al, 2015). Very high survival
482

482 Textbook of Cancer Epidemiology

rates in countries with substantial PSA screening countries. A 30-​fold difference in incidence has
reflect both the lead time introduced by screen- been noted between population groups with the
ing and the detection of indolent cancers unlikely highest (African American men in the United
to progress to advanced lethal disease. States) and the lowest (Japanese and Chinese
Progress In the wake of the 2012 US Preventive men living in their native countries) prostate
Services Task Force recommendation against cancer rates (Muir et al, 1987).
routine PSA screening, efforts have increased to More than with any other cancer, prostate
develop “smarter” screening strategies that would cancer incidence must be interpreted in the con-
continue to allow for early diagnosis of disease text of diagnostic intensity and screening behav-
while reducing overdiagnosis and overtreatment. ior. Latent prostate cancer is quite common, and
Moreover, molecular signatures are being evalu- screening by PSA has allowed for the detection of
ated that could distinguish lethal and indolent many of these lesions. A review of 19 studies of
prostate cancer. prostate cancer prevalence upon autopsy found
that 29% of Caucasian men aged 60–​69 and 36%
aged 70–​79 had undiagnosed prostate cancer at
DESCRIPTIVE the time of death. Prevalence was notably higher
EPIDEMIOLOGY among African American men, with prevalence
of 47% at age 60–​69 and 51% at age 70–​79 (Jahn
Incidence et al, 2015).
Prostate cancer is among the most common PSA screening has dramatically influenced
cancers among men, with more than 1 million the incidence and presentation of prostate can-
cases diagnosed in 2012 (Ferlay et  al, 2013). cer. PSA was first introduced in 1986 to monitor
The burden of this disease shows remarkable disease progression. Widespread opportunistic
worldwide variation, and it is the most fre- screening by PSA has since been introduced in
quently diagnosed male cancer in high-​or several Western countries, allowing the detec-
middle-​income countries (Ferlay et  al, 2013). tion of a significant proportion of latent lesions.
Globally, it is the second most common cancer Consequently, incidence rates in some coun-
in men (world age-​standardized incidence rate tries, the United States being a prime example,
of 31 per 100,000) (Ferlay et  al, 2013). Figure reflect the sum of clinical disease and latent
20-​1 shows the striking international dispar- disease but in other countries only clinical dis-
ity in prostate cancer incidence rates. Among ease. Nevertheless, geographic variation in pros-
countries with reliable cancer statistics, pros- tate cancer incidence, although less extreme,
tate cancer incidence rates are highest in west- was apparent in the pre-​PSA era. Data from the
ernized countries including Australia and New 1970s and early 1980s reflect a country ranking
Zealand, North America, and western Europe of prostate cancer incidence similar to the cur-
and lowest in eastern and south-​central Asian rent standings.

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America

0 50 100 150
ASR (world) per 100,000 person-years

FIGURE 20-​1  Age-​standardized (to the 2012 world population) incidence rates of prostate cancer.
Source: Ferlay et al, 2013.
483

Prostate Cancer 483

140

ASR (world) per 100,000 person-years 120

100

80

60

40

20

0
70

72

74

76

78

80

82

84

86

88

90

92

94

96

98

00

02

04

06
19

19

19

19

19

19

19

19

19

19

19

19

19

19

19

20

20

20

20
Year

USA, Connecticut: White Sweden

UK, England, South and Western Regions Australia, New South Wales

FIGURE 20-​2  Trends in prostate cancer incidence over time in selected countries, IARC.

Source: Ferlay et al, 2013.

A secular trend of increasing worldwide inci- prostate cancer relative to life expectancy in this
dence, on the order of about 3% per year, had group (Vickers et al, 2014). Because undiagnosed
been evident for the past several decades (Quinn prostate cancer is highly prevalent and screening
& Babb, 2002). The United States first displayed results in substantial excess diagnosis, epidemi-
a dramatic departure from this trend, begin- ological studies of risk factors for prostate can-
ning in the late 1980s, with the introduction of cer must account for screening behavior to avoid
intensive PSA screening (Figure 20-​2). Similar identifying risk factors for screening rather than
patterns of increase have occurred in other coun- for prostate cancer itself. Thus a focus on identi-
tries after introduction of screening (Figure fying risk factors for “clinically important” pros-
20-​2). PSA screening has also led to a shift in tate cancer, or disease that represents the greatest
stage at diagnosis, with concomitant increase in potential for lethality, is critical both for biolog-
the ratio of localized to advanced disease cases ical relevance and to deal with confounding by
and a decrease in the age at diagnosis (Brawley screening behavior.
et  al, 1998; Penney et  al, 2013); estimated lead
time gained by PSA screening is approximately Mortality
10 years (Savage et al, 2010). In 2012, 307,000 men died of prostate can-
PSA screening leads to considerable overdi- cer worldwide (Ferlay et al, 2013). It is the fifth
agnosis, or diagnosis of disease that would never leading cause of cancer death in men globally,
have been diagnosed in the absence of screen- with an age-​standardized mortality rate of 7.8
ing. An estimated 25%–​50% of prostate cancer per 100,000. Age-​ standardized mortality rates
cases may be overdiagnosed by PSA screening range from 19.9 per 100,000 in Africa to 3.6 per
(Etzioni et  al, 2002; Ciatto et  al, 2005; Etzioni 100,000 in Southeast Asia. In the United States
et al, 2013). The number of overdiagnosed cases and other westernized countries, it is the second
increases with the age of the screened popula- or third most common cause of cancer death
tion; among men 70 years and older, the major- in men (age-​standardized mortality rate of 9.8
ity of cases detected through PSA screening are per 100,000 in the United States and Canada;
overdiagnosed, due to the long natural history of 11.5 per 100,000 in Europe) (Ferlay et al, 2013).
48

484 Textbook of Cancer Epidemiology

Geographic variation in mortality rates is nota- The ERSPC found a modest mortality benefit
bly lower than that in incidence rates, as PSA from screening (Schröder et  al, 2009; Schröder
screening has a greater impact on incidence than et  al, 2012), while the PLCO found no signifi-
mortality. cant effect (Andriole et al, 2009; Andriole et al,
Prostate cancer has a long and variable course. 2012). However, there were concerns about the
Natural history studies demonstrate that deaths high level of contamination (i.e., screening) in
from cancer can occur even 20 years after diag- the control arm in the PLCO. In addition, the
nosis (Popiolek et al, 2013; Albertsen et al, 2005; conclusions were based on 13  years median
Cuzick et al, 2006). On the other hand, most men follow-​up in the European study and 11  years
die with and not from their cancer, and many in the US study, which is probably inadequate
harbor tumors that remain indolent even in the to fully measure the impact of screening on
absence of therapy. A  trial of watchful waiting prostate cancer mortality, due to the long nat-
versus radical prostatectomy for the treatment of ural history of the disease. Based on the trial
localized prostate cancer found that among those data, the Task Force concluded, “there is ade-
in the watchful waiting group, 29% had died from quate evidence that the benefit of PSA screening
prostate cancer after 18 years, while 40% had died and early treatment ranges from 0 to 1 prostate
from other causes (Bill-​Axelson et al, 2014). cancer deaths avoided per 1,000 men screened”
Mortality rates for prostate cancer have (Moyer & the US Preventive Services Task Force,
declined since the mid-​to late-​1990s in most 2012). However, a review of modeling studies
European countries, the United States, Australia, combining long-​ term population trends with
and New Zealand. Trends in mortality rates are the trial results suggested a greater benefit of
variable in Central and South America and Asia, PSA screening, estimating 5.9 lives saved per
with increases in some countries and decreases in 1,000 men screened, with an overdiagnosis rate
others (Center et al, 2012). In the United States, of 28% of screen-​detected cases (Etzioni et  al,
there has been a notable  50% decrease in age-​ 2013). Etzioni et al concluded that screening can
adjusted mortality rates between 1990–​1999 and explain 45% of the decline in US prostate cancer
2000–​2009 (SEER, 2015). mortality rates in recent decades (Etzioni et  al,
The impact of PSA screening on prostate 2013). At this point, it appears that PSA screen-
cancer mortality has been controversial. In ing does significantly reduce mortality from
2012, the US Preventive Services Task Force prostate cancer, though at the cost of consider-
recommended against routine PSA screening able overdiagnosis.
for prostate cancer based on a concern that the
benefits of screening do not outweigh the harms Age
of overdiagnosis and overtreatment. The esti- Age is the most well documented risk factor for
mates of the benefits of screening were based prostate cancer. The age-​specific incidence curve
on two large trials:  the European Randomized provided in Figure 20-​3 shows the exponential
Study of Screening for Prostate Cancer (ERSPC) rise in prostate cancer incidence beginning at
and the US-​based Prostate, Lung, Colorectal, approximately age 50–​55. This pattern is less evi-
and Ovarian (PLCO) Cancer Screening Trial. dent in Asian countries, but in Western countries
New cases per 100,000

1200
1000
person-years

800
600
400
200
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 20-​3  Age-​specific incidence rates of prostate cancer.

Source: Ferlay et al, 2013.
485

Prostate Cancer 485

it represents the steepest age-​dependent incline 70 years) (Lesko et al, 1996; Cerhan et al, 1999;
of any cancer (Brawley et  al, 1998). US white Schuurman et al, 1999a), and the risk associated
men aged 75–​79 have roughly 130 times the risk with family history is stronger when the father or
of prostate cancer of men aged 45–​49 (Chan brother was diagnosed with the disease earlier in
et  al, 1998). PSA screening has led to a shift in life (Whittemore et  al, 1995; Lesko et  al, 1996).
age-​specific incidence curves to a younger age at This is compatible with our understanding of
presentation in populations where screening is the etiology of many cancer types—​that geneti-
common (Hemminki et al, 2005). cally predisposed individuals will exhibit earlier
onset of the disease. Finally, the larger the num-
Migrant Studies ber of affected first-​degree relatives, the higher
Migrant studies have demonstrated that emi- the prostate cancer risk (Whittemore et al, 1995;
grants from low-​risk countries manifest some Lesko et al, 1996; Rodriguez et al, 1997).
excess prostate cancer incidence by moving to a It has been suggested that men with affected
high-​risk country but that they typically do not family members are more likely to seek prostate
assume the full risk profile of high-​risk country cancer screening; this could create or inflate an
natives. Such evidence has been found in studies association between family history and prostate
of Japanese and Polish immigrants to the United cancer incidence. Evidence from the Prostate
States, whose prostate cancer risk is many times Cancer Prevention Trial (PCPT), a trial of the
higher than that of their native counterparts yet alpha-​reductase inhibitor finasteride versus pla-
still well below that of US white men (Lilienfeld cebo (Thompson et  al, 2003), suggests that this
et al, 1972; Shimizu et al, 1991). Although a shift may be the case. In PCPT, all men received
toward higher incidence rates following migra- annual testing by PSA and DRE, and all men
tion to a high-​risk country could reflect differ- were encouraged to undergo an end-​ of-​
study
ent or more widespread screening practices, the biopsy regardless of PSA and DRE results. In
observation of similar changes in prostate cancer this setting, family history was associated with
mortality does argue for an environmental cause. an adjusted relative risk (RR) of 1.31 (95% CI,
1.11–​1.55), which was statistically significant but
GENETIC AND MOLECULAR notably lower than other family history estimates
EPIDEMIOLOGY (Thompson et al, 2006).
Inherited Susceptibility Germ-​Line Mutations in DNA
Data from twin studies suggest that prostate Repair Genes
cancer has one of the strongest heritable com- A comprehensive molecular analysis of 333 pri-
ponents of any cancer, with an estimated 57% of mary prostate cancers in The Cancer Genome
cancer risk explained by genetic factors (Mucci Atlas (TCGA) found that 19% of tumors har-
et al, 2015; Hjelmborg et al, 2014). Epidemiologic bored mutations or deletions in DNA repair
studies have provided further strong evidence genes; these alterations are even more common
of a heritable component. Data from both case-​ in metastatic tumors (Cancer Genome Atlas
control (McLellan & Norman, 1995; Bratt et  al, Research Network, 2015). In line with this, there
1999; Ghadirian et al, 1997; Lesko et al, 1996) and is consistent evidence that men with mutations
cohort (Grönberg et  al, 1996; Cerhan et  al, in DNA repair genes including BRCA1, BRCA2,
1999; Schuurman et  al, 1999a; Thompson et  al, MLH1, MSH2, MSH6, and PMS2 are at increased
2006) studies show that men with a first-​degree risk of prostate cancer, and perhaps have more
relative with prostate cancer have a two-​to four- advanced disease and poorer clinical outcomes.
fold increased risk of being diagnosed with this Among the DNA repair mutations, those in
disease. the mismatch repair genes BRCA1 and BRCA2
A notable finding in some studies (Keetch have been the most studied. The RR of prostate
et al, 1995; Whittemore et al, 1995; Cerhan et al, cancer diagnosis by 65 years is estimated at 1.8-​
1999; Schuurman et  al, 1999a) is that having to 4.5-​fold for BRCA1 carriers (Leongamornlert
an affected brother confers a stronger elevation et al, 2012; Thompson et al, 2002) and 2.5-​to 8.6-​
in risk than does having an affected father. In fold for BRCA2 carriers (Breast Cancer Linkage
addition, family history appears to be a stronger Consortium, 1999; Kote-​Jarai et  al, 2011; van
risk factor for men who develop the disease Asperen et  al, 2005). In addition, BRCA1 and
earlier in life (usually defined as younger than BRCA2 mutation carriers are at increased risk
486

486 Textbook of Cancer Epidemiology

of high-​ grade and advanced-​ stage disease at prostate cancer-​ specific mortality (Shui et  al,
diagnosis (Castro et  al, 2013; Bancroft et  al, 2014). However, among 24,000 patients, there
2014)  and have worse survival (Edwards et  al, were no significant associations between com-
2010; Castro et al, 2015). The IMPACT study, an mon genetic variants and prostate cancer–​specific
international consortium evaluating the use of survival (Szulkin et  al, 2015). In a case-​control
targeted screening in men with BRCA1/​2 muta- analysis, 22 of 47 previously identified risk sin-
tions found that the positive predictive value of gle nucleotide polymorphisms (SNPs) were asso-
biopsy based on a PSA screening threshold of ciated with the risk of fatal prostate cancer, but
3 ng/​mL was 48% in BRCA2 mutation carriers, they did not distinguish between fatal and nonfa-
twice that reported in population-​based screen- tal prostate cancer risk (Shui et al, 2014).
ing studies, suggesting that targeted screening In addition to the GWAS approach, numer-
among carriers would be beneficial (Bancroft ous studies have examined low-​penetrance sus-
et al, 2014). ceptibility polymorphisms in candidate genes in
Lynch syndrome is an inherited disorder that relation to prostate cancer risk. While significant
increases risk of colorectal, ovarian, and endome- findings have been reported for genes including
trial cancers and comprises of mutations in four androgen receptor, vitamin D receptor, insulin-​
mismatch repair genes (MLH1, MSH2, MSH6, like growth factor 1 (IGF-​1), and inflammation-​
and PMS2). In a cohort of men with Lynch syn- related genes, many of these associations could
drome, mutation carriers had a cumulative risk not be replicated in subsequent studies, suggest-
of prostate cancer by age 70  years of 30% com- ing the findings of earlier studies may have been
pared to 8% in the general population, and were due to chance or bias.
at greater risk of high-​grade disease (Grindedal In summary, genetic predisposition plays an
et  al, 2009). A  study of 188 men with Lynch important role in prostate cancer risk and pos-
syndrome found a standardized rate ratio for sibly in progression. The highly penetrant genes
prostate cancer of 4.87 (95% CI 2.43–​8.71) com- identified to date explain only a small proportion
pared to the general population (Haraldsdottir of disease risk.
et al, 2014).
RISK FACTORS
Genome-​Wide Association Studies Epidemiological studies have an important role
Genome-​wide association studies (GWAS) and to shed light on potential risk factors for prostate
meta-​analyses have now identified >100 com- cancer, and toward understanding opportuni-
monly occuring, low-​ penetrance susceptibil- ties for prevention. Researchers are beginning to
ity loci for prostate cancer risk in populations make distinctions between prostate cancer that
of European ancestry (Eeles et  al, 2014)  and is of clinical significance versus indolent disease,
seven variants in multiancestry analysis (Al and these two entities may reflect different etiolo-
Olama et  al, 2014). The most recent meta-​ gies. Thus, it is critical in evaluating the epidemi-
analysis identified 57 additional risk loci, and ological literature to consider the associations of
together, identified variants now explain 41% of risk factors with respect to clinically advanced or
the familiar risk of prostate cancer in popula- fatal disease.
tions of European ancestry (Schumacher, ASHG The mix of indolent and aggressive disease
2016). The variants interact multiplicatively, so in a study population will depend on the time
the combined effect can be large in spite of low period of the study and the PSA screening prac-
odds ratios associated with individual variants tices in the population. This makes it difficult to
(on the order of 1.10). The top 10% of men at the compare RRs for “total prostate cancer” across
highest risk based on a polygenic score using time period or between populations with dif-
the 100 identified risk loci for men of European ferent usage of PSA testing, as RRs will reflect
ancestry have an RR of 2.9 (95% CI 2.8–​3.1) a weighted average of indolent and aggressive
compared to the population average (Al Olama disease. Stratification by the pre-​PSA era and
et al, 2014). PSA era is helpful, as is assessing associations
To date, common genetic variants have gen- separately for fatal or advanced-​ stage disease
erally not been associated with risk of fatal and localized disease. Some studies also look at
disease. An analysis of 10,000 men with pros- associations for high-​grade and low-​grade dis-
tate cancer found that 8 of 47 previously estab- ease separately; however, changes in grading
lished risk alleles were associated with time to over time and between-​pathologist differences in
487

Prostate Cancer 487

grading introduce misclassification into grade-​ Joshu et al, 2011a; Pantarotto et al, 2007; Oefelein
based categorization in large cohort studies. & Resnick, 2004; Kenfield et al, 2011a). A pooled
analysis of five prospective cohort studies found
Tobacco that current smoking was associated with a 40%
Although strongly linked to a number of cancers, higher risk of prostate cancer mortality (95% CI
cigarette smoking does not appear to be asso- 20%–​70%) (Carter et al, 2015).
ciated with overall prostate cancer incidence.
A  meta-​analysis (Islami et  al, 2014)  of 15 stud- Diet
ies prior to 1995 (pre-​PSA screening era), found Studies attempting to elucidate the nutritional
a pooled RR for current smoking and prostate etiology of prostate cancer are many. A Western
cancer risk of 1.06 (95% CI 0.98–​1.15). For the diet has long been regarded as a potentially
18 studies completed after 1995 (the PSA screen- important risk factor for prostate cancer. This
ing era), there was a significant inverse associ- suspicion is based on international compari-
ation with current smoking (RR 0.84, 95% CI sons of prostate cancer mortality rates—​largely
0.79–​0.89), possibly reflecting the inverse asso- regarded as the result of disparate rates of pro-
ciation between smoking and PSA screening. gression of initiated prostate cancer to advanced
This heterogeneity by time period highlights the disease—​and the observation that migrants from
importance of accounting for PSA screening in low-​to high-​risk geographic areas, as well as their
studies of prostate cancer incidence. A previous offspring, assume the higher risk profile of their
meta-​analysis (Huncharek et  al, 2010)  similarly adopted countries. Both observations suggest
found no association between current smoking that strong environmental factors are involved.
and prostate cancer incidence; however, it did Ecologic studies provided the basis for the initial
show a positive association, with risk among the hypotheses regarding diet and prostate cancer,
heaviest smokers measured by cigarettes per day demonstrating the striking disparity in animal
or pack-​years. product and fat consumption between high-​risk
A positive association between smoking and (United States, Sweden) and low-​ risk (Japan,
prostate cancer mortality has been documented China) countries (Rose et  al, 1986). Ecologic
consistently, as noted by the Surgeon General’s study designs, however, cannot adequately con-
2014 report. A  meta-​analysis of 21 prospective trol for the effect of confounders and are limited
cohort studies of smoking and prostate cancer in their ability to isolate important single foods
mortality found that current cigarette smok- or nutrients.
ing was associated with a 24% increased risk Evidence from prospective cohort stud-
of fatal disease (95% CI 18%–​31%), with lit- ies and population-​based case-​control studies
tle evidence of heterogeneity between studies has been frustratingly mixed and inclusive on
(Islami et al, 2014). There was a significant dose-​ the nutritional epidemiology of prostate can-
response relationship between number of ciga- cer. A  report from the World Cancer Research
rettes smoked per day. There was a suggestion Fund/​American Institute for Cancer Research
of increased risk for former smoking (at base- (WCRF/​AICR) Continuous Update Project on
line in the cohort) and subsequent fatal pros- prostate cancer concluded that there were no
tate cancer, with a 6% increase in risk (95% CI dietary factors at the level of probable or con-
0%–​13%). In the Health Professionals Follow-​up vincing evidence for prostate cancer (WCRF/​
Study (HPFS) cohort, Giovannucci et al. (1999) AICR, 2014)  This represented a downgrading
found that former smokers who had quit within of the strength of the evidence for lycopene,
the past 10 years were at increased risk of fatal selenium, and calcium, which were considered
prostate cancer (HR 1.73, 95% CI 1.00–​3.01), “probably” related to prostate cancer in their
but that longer-​term former smokers were not 2007 Expert Report, and for legumes, processed
at significantly increased risk (HR 1.04, 95% CI meat, and foods with vitamin E, which were in
0.66–​1.64). the “Limited—​ suggestive” evidence category
In line with findings on incidence of lethal (WCRF/​ AICR, 2007). The report concluded
disease, studies of smoking and survival among that there is limited but suggestive evidence that
prostate cancer patients suggests that smoking is dairy products and diets high in calcium increase
associated with increased prostate cancer–​specific risk, and that low plasma alpha-​tocopherol and
mortality as well as total mortality (Rieken et al, low plasma selenium increase risk (WCRF/​
2015; Pickles et  al, 2004; Moreira et  al, 2010; AICR, 2014).
48

488 Textbook of Cancer Epidemiology

Fat Intake trend for high-​grade disease. Thus, the associ-

Based on ecologic studies, dietary animal fat ation between ALA and risk of prostate cancer,
(from meat and dairy products) has been of great particularly of advanced or fatal disease, remains
interest in studies of diet and prostate cancer, as unclear.
it stands out as a major attribute of the Western The role of long-​chain n-​3 fatty acids in pros-
diet. However, a meta-​analysis of 14 prospective tate cancer has also been debated; see the discus-
cohort studies found no association between total sion below on fish intake.
fat intake and risk of prostate cancer (Xu et  al,
2015). Intakes of saturated, polyunsaturated, and Survival
monounsaturated fats were also not associated Fewer studies have examined the association
with risk, and fat intakes were not associated with between fat intake and prostate cancer-​specific
risk of advanced-​stage disease. mortality among men with prostate cancer, but
Several prospective studies have evaluated these studies are consistent in finding improved
intake of specific fatty acids, rather than total fat, survival with increased vegetable fats and poorer
in relation to prostate cancer. In Western diets, survival with saturated and animal fats. In the
alpha-​linolenic acid (ALA) is the principal die- PHS, men who consumed more saturated fat in
tary n-​3 fatty acid. Several meta-​analyses of ALA place of carbohydrate in the postdiagnosis diet
intake and prostate cancer have found no asso- were at increased risk of cancer-​ specific and
ciation with overall disease risk, though there all-​cause mortality (Van Blarigan et  al, 2015).
appears to be significant heterogeneity between Increased intake of vegetable fats after diagnosis
studies (Simon et  al, 2009; Carayol et  al, 2010; was associated with lower risk of all-​cause, but
Carleton et  al, 2013). HPFS found that ALA not cancer-​specific, mortality. In HPFS, postdi-
intake was not associated with overall prostate agnosis vegetable fat intake was associated with
cancer risk but was associated with significantly significantly lower risk of both cancer-​specific
increased risk of fatal disease (Giovannucci et al, and all-​cause mortality, whereas intakes of satu-
2007). Two other prospective studies of ALA rated and trans fats were associated only with all-​
intake did not find increased risks for more cause mortality (Richman et al, 2013). A study of
advanced disease; however, both were lim- men diagnosed with prostate cancer in Sweden
ited by low case numbers (Koralek et  al, 2006; found that those reporting a higher intake of sat-
Schuurman et  al, 1999b). A  fourth prospective urated fat at the time of diagnosis were at signif-
study did not examine associations specifically icantly greater risk of prostate cancer mortality
with advanced disease (Park et al, 2007). (Epstein et al, 2012).
A meta-​analysis of six studies of blood or adi-
pose levels of ALA found increased levels were Meat Intake
associated with increased disease risk (Simon Red meat and processed meat have both been
et al, 2009). However, only four of the six stud- studied as possible risk factors for prostate cancer;
ies were nested case-​control studies based on however, there is little evidence of an association.
prospectively collected blood samples, and of A meta-​analysis of 11 prospective studies found
these, two were from the same underlying cohort a combined RR for extreme categories of intake
(the Physicians’ Health Study, PHS). The earlier of 0.98 (95% CI 0.93–​1.04) for total prostate can-
report from the PHS (Gann et  al, 1994)  found cer and 1.01 (95% CI 0.94–​1.09) for advanced
a positive association between blood ALA lev- prostate cancer (eight studies) (Alexander et  al,
els and prostate cancer risk, but the later report 2010). A recent pooled analysis of individual data
(Chavarro et al, 2007), with increased case num- from 15 cohort studies also found no association
bers and longer follow-​up, found a positive asso- between red meat intake and risk of total or fatal
ciation of ALA only for localized disease. Of the prostate cancer (Wu et al, 2016).
two other prospective studies, one found a pos- For processed meat, the meta-​analysis of 11
itive association with overall prostate cancer, prospective studies found an RR for extreme cat-
while the other found no association overall or egories of intake of 1.05 (95% CI 0.99–​1.12) for
for advanced disease. An additional nested case-​ total prostate cancer and 1.10 (95% CI 0.95–​1.27)
control study (Crowe et al, 2008) published after for advanced cancer (eight studies) (Alexander
the meta-​ analysis reported no association for et  al, 2010). There was evidence of publication
plasma ALA and overall or advanced-​stage pros- bias for processed meat studies, and risk esti-
tate cancer risk, but found a significant positive mates were weaker in more recent studies that
489

Prostate Cancer 489

adjusted for more potential confounders. The Fish Intake and Marine Fatty Acids
pooled analysis of 15 cohorts found a suggestion Populations with a high consumption of fish, for
of a slight increase in risk of total prostate can- example in Japan and among Eskimos in Alaska,
cer for the highest category of processed meat have lower rates of prostate cancer than popula-
intake, though there was no significant trend tions with Western food habits, where the fish
across categories (HR 1.04, 95% CI 1.01–​1.08, p-​ intake in general is lower (Nutting et  al, 1993;
trend = 0.29); there was no association for fatal Zhang et  al, 1999; Dewailly et  al, 2003). Fish
disease (Wu et  al, 2016). Similarly, two reports contain long-​chain marine omega-​3 fatty acids,
published after the meta-​analysis also found no which can modify inflammatory pathways and
associations between red or processed meat and thereby affect prostate cancer risk and progres-
risk of total or advanced prostate cancer among sion (Chan et  al, 2005). Indeed, a study among
African Americans in the NIH-​AARP Diet and men without cancer in PCPT found that higher
Health Study (Major et  al, 2011)  and of lethal serum levels of omega-​3 fatty acids were associ-
prostate cancer in the PSA screening era in HPFS ated with lower levels of prostatic inflammation
(Richman et al, 2011a). In line with this, a study (Nash et al, 2015). However, the role of fish and
in the Netherlands found no association between long-​chain fatty acids in prostate cancer have
low-​or no meat consumption and risk of prostate been debated, due to reports from PCPT and the
cancer (Gilsing et al, 2016). Selenium and Vitamin E Cancer Prevention Trial
One possible mechanism by which red meat (SELECT) of significant positive associations
could raise the risk of cancer is through hetero- between high concentrations of serum phospho-
cyclic amines (HCAs) formed during cooking lipid long-​chain omega-​3 fatty acids and risk of
(Sinha et al, 2009; Koutros et al, 2008; Sharma et al, high-​grade disease (Brasky et  al, 2011; Brasky
2010; Sander et  al, 2011). HCAs are mutagenic et al, 2013). The PCPT and SELECT trials had too
compounds formed during cooking of muscle of few advanced cases to study this as an outcome,
meat and fish at high temperatures. Preference so subgroups were based on grade only. Two
for doneness of red meat and calculated intakes of large cohort studies, the European Prospective
common HCAs have been studied with respect to Investigation into Cancer and Nutrition (EPIC)
prostate cancer in several prospective studies with and PHS, have examined serum fatty acid con-
mixed results. Three found no clear associations centrations and risk of prostate cancer by grade
between doneness or HCA intake and risk of pros- and stage. The EPIC cohort found a positive asso-
tate cancer (Rohrmann et al, 2015; Sharma et al, ciation between serum EPA, a long-​chain omega-​
2010; Sander et  al, 2011). Three found positive 3 fatty acid, and risk of high-​grade disease, but
associations between well-​done red meat (Sinha no association with advanced or fatal disease
et al, 2009; Koutros et al, 2008; Cross et al, 2005; (Crowe et al, 2008). The PHS cohort found non-
Major et al, 2011) and for HCA intake (Cross et al, significant inverse associations between total
2005), including for advanced disease. long-​ chain fatty acids and risk of high-​ grade
Overall, intakes of red and processed meat do and low-​grade cancers, significant inverse asso-
not appear related to prostate cancer risk; how- ciations with localized disease, and no significant
ever, well-​done red meat and intake of related association with advanced stage disease. A meta-​
cooking carcinogens may play some role. analysis of nine biomarker studies of long-​chain
fatty acids found no significant association with
Survival total prostate cancer risk (Alexander et al, 2015).
In a study of postdiagnosis meat intake and sur- A 2010 meta-​ analysis of fish intake and
vival among men diagnosed with apparently prostate cancer found no association between
localized prostate cancer, Richman et  al. found fish consumption and incidence of total pros-
suggestive but not statistically significant asso- tate cancer; for the highest versus lowest cate-
ciations between intake of red meat and poultry gory of intake across 12 cohort studies the RR
and risk of lethal prostate cancer (Richman et al, was 1.01 (95% CI 0.90–​1.14) (Szymanski et  al,
2011a). Another study found that a “Western 2010). However, in four cohort studies of prostate
dietary pattern,” characterized by higher intakes cancer-​specific mortality, fish intake was associ-
of red and processed meats, high-​fat dairy, and ated with a significantly lower risk (RR 0.37, 95%
refined grains, was associated with increased risk CI 0.18–​0.74). A study in Iceland, where there is
of prostate cancer-​specific and all-​cause mortal- a tradition of fish oil consumption, found that
ity (Yang et al, 2015a). fish oil consumption in later life was associated
490

490 Textbook of Cancer Epidemiology

with a lower risk of advanced prostate cancer but a significant positive association in three
(Torfadottir et  al, 2013). Additional studies of studies with 10 or more years of follow-​up (RR
serum fatty acid levels and fish intake with risk of1.25, 95% CI 1.09–​1.43). This heterogeneity by
disease by both grade and stage in cohorts with follow-​up time is supported by a report from the
long-​term follow-​up is needed in order to draw HPFS (Wilson et al, 2015), which found a signif-
conclusions about the role of fish, fish oil supple-icant association between total calcium intake
ments, and specific long-​ chain omega-​ 3 fattyand risk of advanced prostate cancer for intake
acids in prostate cancer. 12–​16 years prior to cancer diagnosis, but not for
shorter time periods between intake and diag-
Dairy Products and Calcium nosis. This suggests that calcium may play a role
Dairy products, in addition to containing a sub- early in the carcinogenesis process.
stantial amount of animal fat, are the most com- Possible mechanisms linking dairy or calcium
mon dietary sources of calcium and vitamin and prostate cancer risk include the downregu-
D.  The strong correlation between dairy foods lating effect of high calcium intake on vitamin D
and these nutrients create challenges in trying levels (Giovannucci, 1998) and the positive asso-
to disentangle the independent effects. A  meta-​ ciation between dairy and IGF-​1 levels (Ma et al,
analysis conducted as a part of the AICR/​WCRF 2001). The positive association observed for low-​
Continuous Update Project found a statistically fat or skim milk argues against dairy fat playing
significant increased risk of total prostate cancer an important role in the association.
associated with intake of dairy products and die- Interestingly, the HPFS also found a posi-
tary calcium (Aune et  al, 2015). The combined tive association between phosphorus intake and
estimate across 15 cohort studies found a 7% risk of total, lethal, and high-​grade prostate can-
increased risk per 400 g of dairy products per day cer independent of the association with calcium
(95% CI 2%–​12%) and 5% increased risk per 400 (Wilson et  al, 2015). In contrast to the pattern
mg of dietary calcium (95% CI 2%–​9%). Intakes observed for calcium, the phosphorus associa-
of milk, cheese, and total calcium (from foods tion was strongest for intakes shortly before the
plus supplements) were also positively associated time of diagnosis (0–​4  years). Phosphorus, like
with risk. There was evidence of nonlinear asso- calcium, is concentrated in dairy, but is more
ciations for calcium (dietary and total), with pos- widespread in other foods than is calcium. Fewer
itive associations more pronounced at very high studies have examined phosphorus than calcium,
intakes (1,500 mg/​day and more). particularly with respect to advanced or fatal dis-
Associations according to stage of disease ease, but this should be explored in other studies.
were less clear, with significant positive asso- High phosphorus intake increases parathyroid
ciations for nonadvanced disease but not for hormone, which promotes bone remodeling
advanced disease for both dairy and dietary cal- (Berruti et  al, 1999). Prostate cancer preferen-
cium. However, across five studies of fatal pros- tially metastasizes to bone and is more likely to
tate cancer, the association with dairy yielded an spread to bone with higher remodeling activity
RR very similar to that for total disease, but with (Sturge et al, 2011; Schneider et al, 2005).
a much wider confidence interval (RR 1.11, 95% Overall, there is substantial evidence that
CI 0.92–​1.33) due to limited number of cases for dairy intake is associated with increased prostate
fatal prostate cancer. cancer risk; however, the role of calcium, phos-
Estimates for total calcium were some- phorus, or other components is less clear.
what weaker than for dietary calcium, suggest-
ing that some other component of dairy, rather Survival
than calcium itself, is driving the dietary cal- There have been three studies of postdiagnosis
cium estimates. However, interpretation of this, dairy intake and prostate cancer survival among
too, is complicated, as the only two studies that men diagnosed with apparently localized dis-
examined total calcium and fatal prostate can- ease. HPFS and a Swedish study both found that
cer found a significantly increased risk (RR 1.11, higher postdiagnosis intake of whole milk was
95% CI 1.02–​1.21). In addition, there was signif- associated with worse survival (Pettersson et al,
icant heterogeneity for the total calcium estimate 2012; Downer et  al, 2017), while higher intake
based on study follow-​up time, with a nonsig- of low-​fat dairy was associated with improved
nificant association with total prostate cancer in survival that was statistically significant in HPFS
six studies with less than 10  years of follow-​up, and suggestive in the Swedish population. On the
491

Prostate Cancer 491

other hand, PHS reported that total dairy intake et al, 2011; Tretli et al, 2009), though others have
after diagnosis was associated with increased risk not (Albanes et al, 2011; Holt et al, 2013). In addi-
of all-​cause and prostate cancer-​specific mortal- tion, genetic variants in the vitamin D pathway
ity (Yang et al, 2015b). are associated with risk of recurrence or progres-
sion and prostate cancer–​specific mortality (Holt
et  al, 2010; Shui et  al, 2012; Shui et  al, 2015).
Vitamin D Genetic variants in the vitamin D receptor were
Vitamin D, which is closely related to calcium associated with Gleason score in some studies
homeostasis, has also been considered as a pros- (Chen et al, 2009; Gandini et al, 2014), and high
tate cancer risk factor. Dihydroxyvitamin D expression of the vitamin D receptor protein in
(1,25(OH)2D) is a steroid hormone involved in prostate cancer tissue was associated with lower
regulating differentiation and proliferation of risk of prostate cancer mortality among men
many cell types, including prostate epithelia, with prostate cancer in the HPFS and PHS, with
which express functional vitamin D receptors. adjustment for PSA at diagnosis, Gleason grade,
The main source of vitamin D is endogenous and stage (Hendrickson et al, 2011).
production in the skin resulting from sun expo- Thus, while vitamin D exposure does not
sure, and diet is a secondary source. 1,25(OH)2D seem to be associated with lower risk of incident
is the most biologically active form, whereas prostate cancer, several lines of evidence suggest
hydroxyvitamin D (25(OH)D) is found in much that the vitamin D pathway may play a role in
higher concentrations and better reflects sun and prostate cancer progression.
dietary exposure.
A meta-​analysis of circulating 25(OH) D lev- Lycopene and Tomatoes
els in 14 prospective nested case-​control studies The relationship between tomatoes; lycopene, a
found no association with total prostate cancer carotenoid consumed principally from tomato
risk (OR 1.04, 95% CI 0.99–​1.10) (Gilbert et al, products; and prostate cancer has received a
2011). In six studies of aggressive prostate cancer, great deal of attention (Giovannucci et al, 1995),
defined as high grade or advanced stage, there as preliminary evidence suggesting a significant
was no association (OR 0.98, 95% CI 0.84–​1.15); benefit associated with high intake of tomatoes or
however, there was evidence of heterogeneity lycopene represents a possible avenue for pros-
between studies. Similar null associations were tate cancer prevention. A  meta-​analysis of four
found in another meta-​analysis (Gandini et  al, cohort studies of dietary lycopene intake and
2011). A consortium of cohort studies with 518 prostate cancer risk found a significant inverse
fatal cases and 2986 controls found no associa- association, with a RR of 0.87 (95% CI 0.77–​0.99)
tion between 25(OH)D and risk of fatal pros- (Wang et  al, 2015). The HPFS found a stronger
tate cancer. However, there was evidence that association with lethal cancer (death or distant
the 25(OH)D association may be modified by metastatic disease) than for total prostate can-
genetic variation in several vitamin D-​ related cer, with an HR for lethal cancer of 0.72 (95% CI
genes (Shui et al, 2015). 0.56–​0.94, p-​trend = 0.04) for the highest versus
1,25(OH)2D has been less studied. However, lowest quintile. The association for lethal cancer
a meta-​analysis of seven prospective studies of was stronger among a subcohort of men who
1,25(OH)2D found no association with total pros- received PSA tests (HR 0.47, 95% CI 0.29–​0.75,
tate cancer (OR 1.00, 95% CI 0.87–​1.14). Only p-​trend  =  0.009), suggesting the association is
two studies have looked at 1,25(OH)2D and risk not related to differences in screening or detec-
of aggressive disease (both based on high-​grade, tion (Zu et  al, 2014). Among men diagnosed
or advanced stage, or prostate cancer death) with with aggressive prostate cancers in the CPS II
a suggestive combined odds ratio of 0.86 (95% CI Nutrition Cohort, consistently high lycopene
0.72–​1.02) (Gilbert et al, 2011). intake before and after diagnosis was associated
In spite of these generally null findings for with improved survival; however, there was no
incidence of total or aggressive disease, there association between lycopene intake and survival
is some evidence that vitamin D plays a role in when all prostate cancer cases were included
prostate cancer progression. Several studies have (Wang et al, 2016).
found inverse associations between 25(OH) Cooked or processed tomato products, such
D and survival among prostate cancer patients as tomato sauce, tomato soup, and ketchup, offer
(Brändstedt et  al, 2016; Shui et  al, 2012; Fang more readily bioavailable sources of lycopene
492

492 Textbook of Cancer Epidemiology

than fresh tomatoes (Tonucci et al, 1995; Gartner A meta-​analysis of five cohort studies and
et  al, 1997). Accordingly, some epidemiologic nine case-​control studies found a combined RR
studies have found significant inverse effects for of 0.74 (95% CI 0.63–​0.89) for the highest ver-
tomato sauce while reporting weaker results for sus lowest category of soy intake. The RR among
raw tomato intake and no significant influence studies estimating isoflavone intake was 0.88
for tomato juice (Giovannucci et al, 2002). (95% CI 0.76–​1.02) (Yan & Spitznagel, 2009).
The correlation between dietary estimates of The Multiethnic Cohort Study in Hawaii and
lycopene based on food frequency questionnaires California (Park et al, 2008), which was the larg-
and circulating levels measured in blood are rel- est study in the meta-​analysis, found a hazard
atively low, ranging from 0 to 0.47 (Giovannucci ratio of 0.90 (95% CI 0.80–​1.01, p-​trend = 0.20)
2002). A pooled analysis of 13 prospective stud- for the highest versus lowest tertile of soy food
ies of plasma lycopene found no association consumption for total prostate cancer and 0.78
with total prostate cancer risk, but a significant (95% CI 0.62–​0.98, p-​trend  =  0.05) for high-​
inverse association for advanced or fatal disease, grade or nonlocalized prostate cancer. However,
with an odds ratio of 0.65 (95% CI 0.46–​0.91, p-​ there was no inverse association between soy
trend = 0.03) for the highest versus lowest quin- food and prostate cancer risk among the Japanese
tile (Key et al, 2015). American men in the study, who had higher
Experimental studies suggest that lycopene can soy intakes than the white, Latino, and African
inhibit angiogenesis, perhaps through regulation American men. Two cohort studies in Japan with
of vascular endothelial growth factor (VEGF) and much higher soy intake found suggestive, but not
the PI3K-​Akt and ERK/​p38 signaling pathways statistically significant, inverse associations with
(Yang et al, 2011; Elgass et al, 2012; Huang et al, prostate cancer (Allen et al, 2004; Kurahashi et al,
2013; Chen et al, 2012). Interestingly, three meas- 2007). Additional prospective cohort studies are
ures of tumor angiogenesis, microvessel diameter needed in populations with high soy intake to
and area and irregularity of the vessel lumen, were determine whether there is, in fact, an inverse
all associated with lycopene intake such that those association with disease risk.
with higher intakes had more favorable angiogen-
esis markers (Zu et al, 2014). These angiogenesis Vitamin E and Alpha-​Tocopherol
markers are associated with risk of lethal disease Vitamin E refers to a group of fat-​soluble com-
independent of grade (Mucci et al, 2009). pounds that include tocopherols and tocotri-
In summary, there is consistent epidemiolog- enols. Alpha-​tocopherol is the biologically most
ical evidence that higher lycopene and tomato active form, and current dietary recommenda-
intake is associated with reduced risk of advanced tions for vitamin E in the United States are based
and fatal prostate cancer. on alpha-​tocopherol alone. Gamma-​tocopherol
is the most common tocopherol in the US diet,
Soy/​Phytoestrogens but plasma levels of alpha-​tocopherol are higher
Given the substantially lower incidence of pros- than those of gamma-​tocopherol (Jiang 2014).
tate cancer in Asia, epidemiological attention has Vitamin E is an antioxidant and supports immune
turned to dietary practices that are characteristic response. Possible anticarcinogenic actions of
of these low-​risk populations. Traditional Asian vitamin E include its ability to reduce DNA dam-
diets are notably high in phytoestrogens, chiefly age and inhibit malignant cellular transforma-
from soy-​based products. Animal studies have tion (Meydani 1995; Meydani & Hayek, 1995).
suggested that phytoestrogens may play a role In experimental models, derivatives of vitamin
through estrogenic effects, inhibition of angio- E inhibit growth, induce apoptosis, and enhance
genesis, antioxidant activity, stimulation of apop- therapeutic effects in human prostate cancer cells
tosis, and inhibition of cell growth (Messina et al, (Gunawardena et al, 2000; Ripoll et al, 1986).
1994; Tham et  al, 1998; Adlercreutz & Mazur, Secondary results of the Alpha-​Tocopherol
1997; Bylund et al, 2000; Kyle et al, 1997). Beta-​Carotene Cancer Prevention (ATBC)
Dietary phytoestrogens, naturally occurring Study (ATBC Study Group, 1994), a random-
constituents of plants, are divided into two main ized trial of lung cancer prevention among male
categories:  lignans and isoflavonoids. Lignans smokers in Finland, showed a 32% reduction
occur in whole-​grain bread, seeds, berries, veg- in prostate cancer risk among men assigned to
etables, and tea, while the main source of isofla- alpha-​tocopherol supplementation compared to
vonoids is soybeans and soy products. placebo (Heinonen et  al, 1998). Another large
493

Prostate Cancer 493

trial of a variety of nutrients found that vita- grade 8–​10 (Klein et al, 2011). In addition, only
min E (in combination with selenium and beta-​ 8% of men in SELECT and 4% of men in PHS
carotene) reduced overall cancer mortality (Blot II were current smokers, so neither trial could
et al, 1993). These results, along with laboratory address the effect of vitamin E specifically among
evidence and some epidemiologic support, moti- smokers.
vated two trials of vitamin E supplementation on Interestingly, epidemiological studies of
the risk of prostate cancer. vitamin E and prostate cancer risk tend to sup-
The SELECT trial of selenium and vitamin E port the ATBC observations, with generally
supplementation was conducted among 35,533 null overall associations, but inverse associa-
men from the United States, Canada, and Puerto tions for advanced disease and among smokers.
Rico. The study was planned for 7 to 12  years In the VITamins And Lifestyle (VITAL) study,
but was stopped early because of lack of efficacy a cohort study specifically designed to exam-
for risk reduction (Lippman et  al, 2009). The ine supplement use and future cancer risk,
initial report, based on an average of 5.5 years 10-​year average intake of supplemental vitamin
of treatment, found a nonsignificant suggestion E was not associated with overall prostate can-
of increased prostate cancer risk among men cer risk, but was associated with a reduced risk
receiving 400 IU/​day of alpha-​tocopherol. With of advanced prostate cancer (regionally inva-
additional follow-​up, the vitamin E group was sive or distant metastatic, n  =  123; HR 0.43,
found to have a significant increase in pros- 95% CI 0.19–​1.0 for average intake ≥ 400 IU/​
tate cancer risk (RR 1.17, 99% CI 1.004–​1.36, day vs. none) (Peters et  al, 2008). Other epi-
p = .008) (Klein et al, 2011). Interestingly, there demiological studies have similarly found a
was not a statistically significant increased risk protective association limited to ever smokers,
of prostate cancer in the vitamin E and sele- including a prospective study of dietary vita-
nium combination group (HR 1.05, 95% CI min E intake (Kirsh et al, 2006), and a study of
0.89–​1.22), suggesting the two may interact. In vitamin E supplementation and lethal prostate
fact, SELECT was designed as a four-​group trial cancer risk (Chan et al, 1999).
rather than a factorial trial expressly because A pooled analysis of 13 prospective studies of
of concerns that the two agents may interact blood alpha-​tocopherol and prostate cancer risk
(Lippman et al, 2005). found significant inverse associations overall and
PHS II, conducted contemporaneously with for advanced prostate cancer (regionally invasive,
SELECT, was a randomized trial of vitamin E and distant metastatic, or fatal cancer), with an odds
vitamin C supplement use and prostate cancer ratio for advanced disease of 0.74 (95% CI, 0.59–​
risk among 14,642 US physicians. With a median 0.92; p-​trend = 0.001) for the highest versus low-
of 8 years of follow-​up, there was no effect of 400 est quintile, based on 1,226 advanced cases (Key
IU of vitamin E taken every other day and inci- et  al, 2015). There was significant heterogeneity
dence of prostate cancer (HR 0.97, 95% CI 0.85–​ by disease aggressiveness, with no association for
1.09) (Gaziano et al, 2009). nonadvanced disease. In addition, there was no
Together, the SELECT and PHS II results association among never smokers (OR 0.99, 95%
suggest that vitamin E supplement use is at least CI 0.82–​1.18), and significant inverse associa-
ineffective and perhaps harmful with respect to tions for former and current smokers (OR 0.84,
prostate cancer risk. This is in contrast to the 95% CI 0.72–​0.97 for former; OR 0.82, 95% CI
ATBC findings that spurred these trials. Of note, 0.73–​0.93 for current), although the p-​value for
all men in the ATBC trial were smokers, and the heterogeneity by smoking was not statistically
prostate cancers were diagnosed outside the con- significant.
text of PSA screening and thus were generally Overall, use of vitamin E supplements for
aggressive. The SELECT and PHS II trials were prostate cancer prevention is not supported;
done in the PSA era and could not specifically however, diets higher in alpha-​tocopherol appear
study advanced or fatal prostate cancer. Of 2,279 to be associated with lower risk of advanced
prostate cancers diagnosed in SELECT through disease.
July 2011, only nine were diagnosed with stage
T3 disease, three with N1 disease, and 13 with Selenium
metastatic disease. Even the ability to study The trace element selenium is not an antioxi-
high-​grade cancer was limited, with 613 (27%) dant per se, but plays an important role as an
cases grade 7 and above, only 134 of which were essential element for the antioxidant enzyme
49

494 Textbook of Cancer Epidemiology

glutathione peroxidase as well as other seleno- aggressive prostate cancer, the OR for men in the
proteins involved in exerting antitumor effects, highest versus lowest quintile of nail selenium
including apoptosis and inhibition of cellular was 0.18 (95% CI 0.11–​0.31), and for blood sele-
proliferation (Combs & Combs, 1984; Menter nium was 0.43 (95% CI 0.21–​0.87).
et al, 2000; Redman et al, 1998). Dietary intake of A study in HPFS found that use of selenium
selenium depends on the selenium content of soil supplements of 140 mcg/​day was associated with
in which foods are grown, which varies greatly by significantly increased risk of prostate cancer
geographic area. Ecologic studies have suggested mortality among men diagnosed with localized
an inverse association between selenium soil prostate cancer. The association was independ-
content and prostate cancer incidence (Rayman, ent of prediagnosis supplement use, use of other
2000). Because selenium contents in specific supplements, and stage and grade of disease at
foods vary as a function of the selenium content diagnosis (Kenfield et al, 2014). The authors sug-
of the soil, epidemiological studies of selenium gest the possibility of a U-​shaped relationship
require biological sampling, primarily measur- between selenium status and cancer incidence
ing levels in blood or toenails. Since the activity and progression, with adverse effects at very low
of some selenoenzymes plateau with higher sele- and very high levels.
nium level (Neve 1995), the chemopreventive The results of the SELECT trial do not sup-
effect of selenium may be greatest in populations port the use of selenium supplementation for
with low selenium exposure (Duffield-​ Lillico the prevention of prostate cancer in middle-​aged
et al, 2003). and older men; however, the association between
Like vitamin E, selenium was tested in the selenium and prostate cancer risk and survival is
SELECT trial based on secondary results of other still not completely clear.
randomized trials. The Nutritional Prevention of
Cancer Trial found a 63% reduction in prostate Coffee
cancer risk among men taking selenium supple- There are several potential mechanisms by which
ments (Clark et al, 1996); with additional follow-​ coffee could be associated with a lower risk of
up time, the protective effect was limited to those prostate cancer. Coffee is rich in several biolog-
with low baseline levels of PSA or selenium ically active compounds including caffeine, min-
(Duffield-​Lillico et al, 2003). Another trial of sele- erals, and phytochemicals. In observational and
nium (with vitamin E and beta-​carotene) found a animal studies, long-​term coffee drinking has
reduction in total cancer mortality in China (Blot been associated with improved glucose metabo-
et al, 1993). As discussed above, the SELECT trial lism and insulin secretion in observational and
was stopped early due to lack of efficacy of the animal studies (Tunnicliffe & Shearer, 2008).
supplements (Lippman et  al, 2009). With addi- Moreover, coffee is a potent antioxidant (Svilaas
tional follow-​up, there was still no association et al, 2004; Pulido et al, 2003) and intake may be
between selenium and prostate cancer risk (RR associated with levels of different sex steroid hor-
1.09, 99% CI 0.93–​1.27) (Klein et  al, 2011). In mones (Kempf et al, 2010; Svartberg et al, 2003).
addition, baseline selenium status (measured in Most prior epidemiological studies of cof-
toenails) was not associated with prostate cancer fee and prostate cancer have focused on total
risk among men in the trial, and baseline status incidence of disease, with generally null results.
did not modify the association between selenium However, a report from the HPFS found a sig-
supplementation and risk (Kristal et al, 2014). As nificant 60% decrease in risk of lethal (metastatic
for vitamin E, conclusions about selenium drawn disease or death) prostate cancer with higher
from SELECT are limited by the small number of intakes (Wilson et al, 2011), spurring interest in
advanced and high-​grade cases. the association between coffee and advanced dis-
In contrast to the SELECT results, obser- ease. Several meta-​analyses support an inverse
vational studies of selenium and prostate can- association with fatal or advanced disease (Lu
cer risk have been consistent in finding inverse et  al, 2014; Cao et  al, 2014; Zhong et  al, 2014;
associations. A recent pooled analysis of individ- Discacciati et al, 2014). Combining four studies
ual level data from 15 prospective studies found with a fatal prostate cancer outcome, Discacciati
that nail selenium levels were associated with et al (2014) found an RR of 0.89 (95% CI 0.82–​
lower risk of total and aggressive prostate can- 0.97) per 3-​cups/​day increment in coffee intake.
cer, while blood levels were associated with lower There was also an inverse association with high-​
risk of aggressive disease (Allen et al, 2016). For grade (Gleason 8–​10) disease. These intriguing
495

Prostate Cancer 495

data, while biologically plausible, need to be con- suggestion of a positive association among
firmed in additional study populations with large studies conducted in North America, with an
numbers of fatal or advanced disease. RR of 1.20 (95% CI 1.09–​1.31) for heavy versus
light drinkers, whereas there were no signif-
Summary of Evidence for Diet icant associations in European or Asian stud-
The apparently inconsistent findings in many ies; however, this heterogeneity by region was
studies can probably be attributed to the work- not statistically significant. A  consortium of
ings of often unavoidable epidemiologic bias. 11 cohort studies with data on alcohol intake
When dealing with diet–​disease associations, the in Asia found no association between alcohol
measurement error inherent in dietary assess- intake and risk of prostate cancer mortality
ment instruments and laboratory analyses can (Fowke et al, 2015).
obscure findings to the point where their inter- The meta-​ analysis and Asian consortium
pretation becomes dubious. Cohort studies that analysis did not consider risk of advanced disease
attempt to link prediagnostic dietary exposures or prostate cancer mortality separately, and did
to disease often assess exposure at just one arbi- not consider PSA screening. Several cohort stud-
trary point in time, years before diagnosis, and ies have considered these factors. Three cohorts
cannot account for changing exposure over time. found no associations by stage or grade of dis-
Case-​control studies are, to varying extents, sus- ease (Sutcliffe et al, 2007; Rohrmann et al, 2008;
ceptible to recall bias. However, these sources of Velicer et al, 2006). On the other hand, the NIH-​
bias are common across all cancer sites, while AARP Diet and Health Study, a cohort including
the evidence for diet and prostate cancer appears more than 300,000 men, found a positive asso-
unusually murky. This may be due, in part, to the ciation between alcohol intake and nonadvanced
heterogeneity of prostate cancer, particularly in disease, but no association for advanced disease.
populations with PSA screening; indeed, differ- (Comparing those who have six or more drinks
ent risk factors emerge for incidence of advanced per day to nondrinkers: HR 1.25, 95% CI 1.13–​
or fatal prostate cancer compared to all prostate 1.37, p-​trend < 0.001 for nonadvanced; HR 0.97,
cancer (Giovannucci et al, 2007). Future epidemi- 95% CI 0.73–​1.29, p-​trend = 0.38 for advanced.)
ological studies in prostate cancer must be rigor- (Watters et al, 2010). There was actually a sugges-
ous with respect to controlling for PSA screening tion of an inverse association with fatal disease,
behavior and distinguishing between indolent with an HR of 0.45 for those who have six or more
and more clinically relevant disease. Finally, the drinks per day compared to nondrinkers (95%
long natural history of prostate cancer, compared CI 0.25–​0.81, p-​trend = 0.05); however, there was
even to other cancers, suggests that even long-​ little indication of a reduction of risk for more
term cohort studies may miss the critical window moderate intakes, and the issue of competing
for measuring diet with respect to initiation of risks of death in the very high intake group was
disease. Additional work on the role of early life not examined. In contrast to these results, two
diet is needed. cohorts of alcoholics provided limited evidence
that very heavy drinking may increase the risk of
Alcohol prostate cancer compared to general population
A biological basis for an association between rates (Adami et al, 1992; Tonnesen et al, 1994). In
alcohol and cancer, in general, does exist. addition, a recent study of men in Finland found
A metabolite of alcohol, acetaldehyde, is itself a that both heavy drinking and binge drinking was
carcinogen, and alcohol is known to affect cell associated with increased risk of prostate cancer
membrane integrity, enhance production of free (Dickerman et al, 2016).
radicals, impair immune function, and reduce Two studies within randomized trials have
levels of DNA repair enzymes (Jensen et al, 1996). reported a positive association between alcohol
However, epidemiological studies have found lit- intake and risk of high-​grade cancer. The PCPT
tle evidence that alcohol consumption at moder- trial found that heavy alcohol consumption was
ate levels is associated with prostate cancer risk. associated with increased risk of high-​grade can-
A meta-​ analysis of 43 case-​ control and cer diagnosed either during the study period or
cohort studies of alcohol and prostate cancer upon end-​of-​study biopsy. The clinical relevance
found no significant association, with an RR of of these cancers is uncertain, as men were uni-
1.09 (95% CI 0.98−1.21) for heavy versus light formly biopsied at the end of the study regardless
drinking (Bagnardi et al, 2015). There was some of PSA levels or DRE results (Gong et al, 2009).
496

496 Textbook of Cancer Epidemiology

The ProtecT Study, a trial of treatment of PSA-​ the development of prostate cancer, the obser-
detected prostate cancer in the UK, found a sig- vation that bilateral orchiectomy can result in
nificantly decreased risk of low-​grade (Gleason prostate cancer remission, and the finding that
grade 6)  and suggestive increased risk of high-​ androgens can induce prostate cancer in experi-
grade (Gleason grade 7–​10) with increased alco- mental animals all implicate androgens as, at the
hol intake in a population that received uniform very least, important permissive factors for pros-
PSA testing and biopsy for PSA >3. The mag- tate cancer.
nitude of the RRs was quite low (0.96, 95% CI Most epidemiologic evidence does not
0.93–​0.99 for low grade; 1.04, 95% CI 0.99–​1.08 strongly support the hypothesis that high lev-
for high grade per 10-​g/​day increment) (Zuccolo els of circulating androgens correspond to an
et al, 2013). increased risk of prostate cancer. Little attention
In both the NIH-​ AARP Study and a has been given specifically to the risk of advanced,
Washington state cohort (Watters et  al, 2010; high-​grade, or aggressive cancers in nested case-​
Velicer et  al, 2006), PSA screening was slightly control studies of hormone levels to date. Studies
lower among abstainers than those with moder- have measured both free and total testosterone,
ate alcohol intake. In the NIH-​AARP, those with with the expectation that the former, a readily
the highest intakes also had somewhat lower bioavailable form, would be more closely related
screening rates, similar to those among abstain- to risk. Other testosterone metabolites have also
ers. These studies adjusted for PSA screening been measured in many studies.
behavior, but it did not appear to have a major A pooled analysis of 18 prospective studies
impact on RR estimates. of androgens and prostate cancer risk found no
Overall, moderate alcohol consumption does associations for serum concentrations of testos-
not appear to be associated with prostate cancer terone, calculated free testosterone, dihydro-
risk; the association for very high intakes is less testosterone, dehydroepiandrosterone sulfate,
certain. It is possible that high intakes increase androstenedione, or androstanediol glucuronide
the risk of high-​grade disease. (EHPCCG et al, 2008). Cohort studies published
since then have supported this lack of associa-
Reproductive Factors tion (Gill et  al, 2010; Weiss et  al, 2008; Schenk
See the section “Infections” for a discussion of et  al, 2015; Muller et  al, 2012; Daniels et  al,
sexually transmitted infections in relation to 2010). The Reduction by Dutasteride of Prostate
prostate cancer risk. See “Medical Conditions Cancer Events (REDUCE) trial did find that men
and Treatment” for a discussion of vasectomy. in the placebo group with baseline testosterone
below normal levels were at lower risk of pros-
Hormones tate cancer, but there was no association for tes-
Androgens tosterone levels within the normal range (Muller
There is a long-​standing suspicion—​based on a et al, 2012).
convincing array of circumstantial and experi- In spite of these consistently null results,
mental, but not epidemiologic, evidence—​that interest in androgens has sparked again
endogenous hormones play a role in the etiology with the identification of a common gene
of prostate cancer. Androgens, particularly tes- fusion between the androgen-​regulated gene
tosterone and dihydrotestosterone, are essential TMPRSS2 and members of the ETS family of
for the normal growth and functioning of the transcription factors. (Tomlins et al, 2005) The
prostate. Testosterone diffuses into prostate cells, TMPRSS2:ERG fusion is present in 30%–​50% of
where 90% is irreversibly converted to dihydro- cases of prostate cancer. The role of hormones
testosterone, the more active intracellular andro- on prostate cancer risk and progression may be
gen, by an enzyme, 5-​ alpha-​reductase. Both modified according to presence or absence of
testosterone and dihydrotestosterone bind to a the translocation.
cytoplasmic androgen receptor. The complex Many studies have examined levels of sex
of androgen to androgen receptor then binds to hormone-​binding globulin (SHBG). SHBG is a
specific DNA sites, resulting in increased tran- carrier of androgens and estrogens and is a major
scriptional activity and cell division (Coffey, determinant of the serum concentrations of free
1979). The reliance on androgens for normal testosterone and estradiol; it may also play a role
prostatic growth, the fact that the presence of directly in steroid signaling. In the pooled analy-
functioning testes also appears to be essential for sis of 18 studies it was associated with lower risk
497

Prostate Cancer 497

of prostate cancer, with an odds ratio of 0.86 for 2007). The American Urological Society and
the highest versus lowest quintile (95% CI 0.75–​ American Society of Clinical Oncology have
0.98) (EHPCCG et al, 2008). This association was recommended that clinicians discuss the use of
attenuated by adjustment for IGF-​1 (discussed 5-​alpha reductase inhibitors as chemopreventive
later). Two cohorts found nonsignificant asso- agents for prostate cancer (Kramer et  al, 2009);
ciations of similar magnitude for SHBG (Gill however, the cost-​effectiveness of this has been
et al, 2010; Weiss et al, 2008), whereas two others questioned. An approach targeting higher-​risk
found no suggestion of an association (Schenk men may improve cost-​ effectiveness (Svatek
et al, 2015; Daniels et al, 2010). et al, 2008).

5-​Alpha Reductase Estrogens

5-​alpha reductase converts testosterone to the There is experimental and clinical evidence to
more potent dihydrotestosterone, which, by bind- support both harmful and protective roles of
ing to the androgen receptor, initiates a chain of estrogens in prostate cancer (Nelles et  al, 2011;
molecular events that induces prostate cell divi- Carruba G, 2007). Testosterone may be metabo-
sion. Ross et al (1992) found that young Japanese lized to either dihydrotestosterone or to estradiol
men display lower 5-​ alpha reductase activity in men, and multiple organs express estrogen
than young Caucasian or African American receptors. In addition, estrogen can be produced
men. Consequently, there has been considerable locally in the prostate. Both estrogens and the
speculation that 5-​alpha reductase activity could balance between androgens and estrogens have
be related to prostate cancer risk and could also been assessed in epidemiologic studies.
partially account for the low rate of prostate can- A pooled analysis of 18 cohort studies found
cer observed in the Japanese population. 5-​alpha no association between serum estradiol or cal-
reductase activity is reflected by serum biomark- culated free estradiol and prostate cancer risk
ers, including 3-​alpha-​androstenediol glucuro- (EHPCCG et  al, 2008). Another cohort study
nide and androsterone glucuronide, which are published after the pooled analysis found no
downstream products of dihydrotestosterone association between estrone-​or estradiol-​ to-​
metabolism. Epidemiologic studies have offered testosterone ratio (Schenk et al, 2015).
no compelling evidence that these markers of Two studies have measured estrogens using
5-​alpha reductase activity are related to prostate more sensitive mass spectrometry-​based assays
cancer risk (Gann et al, 1996; Nomura et al, 1996; rather than radioimmunoassay methods, which
Guess et  al, 1997; Vatten et  al, 1997; EHPCCG improves measurement in men, in particular,
et al, 2008). as circulating levels of estrogens are very low.
However, two inhibitors of 5-​alpha reduc- The MrOS cohort reported a strong positive
tase, finasteride and dutasteride, have both been association between estrone and prostate can-
shown in randomized trials to significantly cer risk, with a nearly fourfold increased risk in
reduce the risk of prostate cancer. Over 7  years in the top three quartiles compared to the bot-
of follow-​up in PCPT, men taking finasteride tom quartile. There was no association for the
experienced a 25% reduction in prostate cancer testosterone:estradiol ratio (Daniels et al, 2010).
risk (Thompson et al, 2003). Over 4 years in the The PLCO trial found no association for estrone,
REDUCE trial, men taking dutasteride experi- but found a strong inverse association for the
enced a 23% reduction in risk (Andriole et  al, estradiol:testosterone ratio and a strong positive
2010). While finasteride was associated with a association for the ratio of two estrogen metabo-
reduced risk of total prostate cancer, there was lites, 2-​hydroxyestrone:16-​alpha-​hydroxyestrone
a 27% increase in high-​grade (Gleason 7–​10) with respect to aggressive (high-​grade or -​stage)
disease. This was not seen for dutasteride. The disease (Black et al, 2014). In breast cancer mod-
increase in high-​grade disease with finasteride els, 2-​ hydroxyestrone has weaker estrogenic
has been much debated. It has been suggested activity compared to 16-​alpha-​hydroxyestrone,
that it was the result of the reduction in pros- but the role of estrogen metabolites in prostate
tate volume by finasteride, making the detection cancer models has not been elucidated.
of high-​grade areas of cancer more likely upon Additional studies of estrogens using sensi-
biopsy, and a reduction specifically in the risk tive mass spectrometry assays may shed more
of low-​grade disease by finasteride (Lucia et  al, light on the role of estrogens in prostate cancer.
498

498 Textbook of Cancer Epidemiology

Insulin-​Like Growth Factor 1 and Its Differences in laboratory methods used to meas-
Binding Proteins ure IGFBP-​3 may be one important source of this
The insulin/​IGF system is central to the regu- heterogeneity.
lation of energy metabolism and growth. The IGFBP-​1 shuttles IGF-​1 across blood vessel
peptide hormone IGF-​1 and its binding proteins membranes (Lewitt et  al, 1993), and if serum
(IGFBPs) have been studied in many epidemio- IGFBP-​1 levels are predictive of the amount of
logic studies of prostate cancer, with fewer stud- circulating IGF-​ 1 available to prostate tissue,
ies of insulin, and of the receptors for both IGF they may consequently be predictive of pros-
and insulin. tate cancer risk. The pooled analysis (Roddam
IGF-​1 is a major growth-​regulating molecule et  al, 2008)  and meta-​analysis (Rowlands et  al,
that is known to be a potent mitogen that can 2009)  found no associations between IGFBP-​1
also inhibit apoptosis. It is secreted mainly by the or IGFBP-​2 and prostate cancer risk. However,
liver but is also produced in several other tissues, a large prospective analysis of 957 cases and
including the prostate, in response to growth hor- 1,021 controls, published after the pooled anal-
mone. IGF-​1 is present in large concentrations ysis, found a significant inverse association for
in circulation, which can have systemic or local IGFBP-​1, with an odds ratio for the highest ver-
effects on cell behavior. IGF-​1 exerts its effect by sus lowest quintile of 0.67 (95% CI 0.52–​0.86,
binding with IGF-​1 receptors expressed on pros- p-​trend  =  0.003) (Cao et  al, 2015). This associ-
tate and other types of cells, leading to induc- ation was stronger for low-​grade and lower stage
tion of cellular proliferation. In circulation, 80% cancers.
of IGF-​1 is bound in a ternary complex with its At this point, IGF-​1 is clearly associated with
binding protein, IGFBP-​3 and the glycoprotein increased risk of disease, whereas the role of the
acid-​labile subunit (Jones & Clemmons, 1995). IGFBPs is unclear.
A pooled analysis of data on 3,700 cases and
5,200 controls from 12 prospective cohort stud- Methodologic Issues
ies (Roddam et al, 2008) found an odds ratio for Studying endogenous hormones in relation to
prostate cancer comparing the top to bottom prostate cancer risk is challenging for a number
quintile of serum IGF-​1 of 1.38 (95% CI 1.19–​ of reasons. First, the concentration of many cir-
1.60, p-​trend < 0.001). The mean time between culating hormones has a circadian rhythm, and
blood collection and prostate cancer diagnosis not all epidemiologic studies have accounted for
was 5  years. The association was stronger for this by either matching on blood sampling timing
low-​grade than for high-​grade disease, but did or restricting blood draws to a narrow time-​of-​
not vary by stage of disease. Adjustment for vari- day window. Second, within-​person variation in
ous sex hormone levels also measured in 8 of the circulating hormone levels, coupled with labora-
12 studies did not affect the IGF-​1 results. Results tory measurement error, can introduce exposure
from a 2009 meta-​analysis of 42 retrospective misclassification. It is generally assumed, how-
and prospective studies found a similar signifi- ever, that these errors are nondifferential with
cant and positive association for IGF-​1 and pros- respect to cases and noncases, therefore biasing
tate cancer risk (Rowlands et al, 2009). results toward the null.
There is some debate concerning which, if A single blood measurement is probably
any, of the IGFBPs may be relevant. IGFBP-​3 has not sufficiently reliable to characterize an indi-
received the most attention primarily because vidual’s typical hormonal profile (Nomura et al,
over 95% of circulating IGF-​1 is bound to this 1996) and may explain why the multitude of stud-
protein (Jones & Clemmons, 1995). It has been ies have not been able to detect consistent differ-
hypothesized that higher levels of this binding ences between the average sex hormone levels of
protein would result in less available IGF-​1 to prostate cancer patients and healthy comparison
be taken into cells, resulting in a lower risk of men. Moreover, serum hormone levels may not
prostate cancer. However, the pooled analysis accurately reflect tissue levels. Normal and can-
of 12 studies found no association of IGFBP-​3 cerous prostate cells express a range of enzymes
with risk independent of IGF-​1 (Roddam et  al, involved in steroid metabolism, allowing them
2008). The 2009 meta-​analysis similarly found to convert and produce steroids, independent of
no association for IGFBP-​ 3 among prospec- circulating levels. Alternatively, it may be that sex
tive studies, and also found significant hetero- hormones are merely permissive factors. In their
geneity across studies (Rowlands et  al, 2009). presence, prostate cancer can arise in response
49

Prostate Cancer 499

to cellular events not yet explainable, but in their Zhang et al, 2015). Similar results were also seen
absence (i.e., through castration), the mecha- in a meta-​analysis of four studies of waist circum-
nisms by which the disease develops are blocked. ference and advanced prostate cancer (WCRFI/​
Finally, as with diet, the relevant time period AICR, 2014).
for endogenous hormone levels to affect prostate
cancer development may be earlier in life, with Survival
the result that cohort studies miss the critical Higher BMI is consistently associated with poor
period for measuring hormonal exposures. The outcomes among men diagnosed with prostate
association between height and increased pros- cancer. In a meta-​analysis of six studies of survival
tate cancer risk (discussed in the next section) after prostate cancer diagnosis, the RR of prostate
does support the idea that childhood and ado- cancer mortality was 1.20 (95% CI 0.99–​1.46) for
lescent IGF-​1 exposure—​which is closely associ- a 5-​unit increase in BMI around the time of diag-
ated with attained height—​may be important for nosis or treatment (Discacciati et al, 2012). There
prostate development (Gunnell et al, 2004). was significant heterogeneity in this estimate due
to the inclusion of the largest study, which found
Anthropometric Measures a nonsignificant inverse association with mor-
Obesity and Body Mass Index tality based on 4 years of follow-​up. Two studies
Because obesity can influence endogenous levels among men with prostate cancer in Sweden pub-
of sex hormones (Pasquali et al, 1991; Mantzoros lished after the meta-​analysis also found signifi-
& Georgiadis, 1995), as well as the insulin/​IGF cantly increased risks of prostate cancer–​specific
axis (as discussed earlier), it has been studied in mortality with higher BMI (Bonn et  al, 2014;
many epidemiologic studies. Body mass index Cantarutti et al, 2015). In 16 studies of biochem-
(BMI), measured as height (m)/​weight (kg)2 is ical recurrence after treatment, a 5-​unit increase
the most commonly used measure of obesity in in BMI was associated with an RR of 1.21 (95%
these studies. CI 1.11–​1.31) (Discacciati et al, 2012).
The association between BMI and total pros-
tate cancer incidence is somewhat inconsistent. Body Size in Early Life
(Renehan et al, 2008). However, results are quite Childhood obesity is inconsistently associated
consistent when localized and advanced disease with adult prostate cancer risk. Four studies have
are considered separately, as BMI is associated examined prepubertal body size (8–​ 10  years)
with a significantly lower risk of localized disease (Robinson et  al, 2005; Barba et  al, 2008; Hsing
but a significantly increased risk of advanced et al, 2000a), with two reporting inverse associa-
disease. Because of this heterogeneity, the asso- tions, including for advanced disease (Robinson
ciation between BMI and total prostate cancer et  al, 2005; Barba et  al, 2008), while two others
will vary across populations depending on PSA found no associations (Möller et al, 2015; Hsing
screening and the case mix found in that time et al, 2000a). A study of body size at the time of
and place. A  meta-​ analysis of 13 prospective puberty also found no association (Andersson
studies found a RR per 5-​unit increase in BMI et al, 1995).
of 0.94 (95% CI 0.91–​0.97) for localized prostate The results from the HPFS (Möller et  al,
cancer, and 1.09 (95% CI 1.02–​1.16) for advanced 2015)  found significant inverse associations
prostate cancer (Discacciati et al, 2012). The defi- between obesity at age 10 and risk of advanced
nitions of localized and advanced were a mix of and metastatic disease; however, this associa-
stage and grade, depending on the studies. The tion was not observed with 16 additional years
WCRF/​AICR Continuous Update Project report of follow-​up. It is possible that childhood body
on prostate cancer concluded that greater body size influences risk of prostate cancer less among
fatness is a “probable” cause of advanced pros- older men.
tate cancer (WCRFI/​AICR, 2014). Their meta-​ Adiposity is known to increase estrogen and
analysis of 23 studies of advanced cancer found decrease androgen serum concentrations in men
an RR per 5-​unit increase in BMI of 1.08 (95% (Pasquali et  al, 1991). Hence, a childhood hor-
CI 1.04–​1.12). For 12 studies of prostate cancer monal milieu characterized by low exposure to
mortality, the combined RR per 5-​unit increase the stimulating effect of androgens on the pros-
in BMI was 1.11 (95% CI 1.06–​1.17). This was tate might protect against the disease.
consistent with two meta-​analyses of BMI and The HPFS found an inverse association
prostate cancer mortality (Discacciati et al, 2012; between BMI at age 21 and risk of advanced
50

500 Textbook of Cancer Epidemiology

and lethal (death or distant metastasis) prostate surgery (Whitley et  al, 2011)  or from 1  year
cancer, independent of later life and earlier life before to 5 years after surgery (Joshu et al, 2011b)
body size (Möller et al, 2015). Two other studies were associated with increased risk of biochemi-
found similar inverse associations with advanced cal recurrence.
(Robinson et al, 2005) or fatal (Discacciati et al,
2011) disease; however, other studies have found Height
no associations (Littman et al, 2007; Schuurman There is a consistent positive association between
et al, 2000; Giles et al, 2003; Wright et al, 2007). adult attained height and prostate cancer inci-
A review of studies on total prostate cancer inci- dence and mortality. A systematic review in the
dence suggested no relationship or a weak posi- WCRF/​AICR Continuous Update Project iden-
tive relationship (Robinson et al, 2008). tified 25 studies of height and prostate cancer
Overall, there is no consistent association incidence, 22 of which reported a positive associ-
between childhood and young adult body size ation. Of five studies on prostate cancer mortality,
and prostate cancer risk. four reported positive associations. The dose-​
response meta-​analysis found a significant 4%
Weight Change increase in risk per 5-​cm increase in height (95%
A meta-​ analysis of adult weight gain, from CI 1.03–​1.04) (WCRF/​AICR, 2014). Results were
around age 18 to 25 until study entry in mid or consistent for nonadvanced, advanced, and fatal
late life and risk of prostate cancer found no clear disease. Two large pooled analyses also found
association with overall risk (Keum et al, 2015). very similar results for both incidence and mor-
Among eight prospective studies, the combined tality (Batty et  al, 2010; Emerging Risk Factors
RR for the highest versus lowest weight gain Collaboration, 2012).
category was 0.98 (95% CI 0.91–​1.06). A  dose-​ Male adult height can reflect nutritional sta-
response meta-​analysis of four studies also found tus in early life and is determined in part by
no association, but there was a suggestion of an circulating growth factors and other hormones
inverse association for localized disease (RR for during puberty. It is possible that men who attain
5 kg weight gain of 0.96, 95% CI 0.92–​1.00) and greater height might have had higher exposure to
a positive association for advanced disease at pubertal levels of androgens at a time when the
diagnosis (RR for 5-​kg weight gain of 1.04, 95% prostate gland is commencing its normal devel-
CI 0.99–​1.09). There was a similar suggestion of opment and function (Giovannucci et al, 1997a).
an inverse association in populations with high In preadolescent boys, the production of testos-
rates of PSA screening and a positive association terone markedly increases the level of plasma
in populations with lower rates; however, none IGF-​1 (Keenan et al, 1993), and adult height has
of these differences were statistically significant. been positively correlated with serum IGF-​1 lev-
In line with these suggestive findings, several els (Signorello et al, 2000). Thus, it may be expo-
cohort studies have found significant positive sure to high circulating levels of IGF-​1, more
associations between adult weight gain and pros- than to androgens, that is driving the observed
tate cancer mortality (Bassett et al, 2012; Wright positive associations, given that IGF-​1 has also
et  al, 2007; Rodriguez et  al, 2007). Given that been found to be a risk factor for prostate can-
obesity itself is associated with increased risk of cer. Tall height may also simply be a marker for
advanced and fatal prostate cancer, it is difficult large organ size, where the larger number of cells
to separate the effect of weight gain during adult- at risk would contribute to a positive association
hood from the effect of obesity. (Trichopoulos & Lipworth, 1995).
Three studies have examined weight gain
around or shortly after the time of prostate cancer Infections
diagnosis. Only one study used mortality as the A large body of epidemiologic, genetic, and
outcome, finding a significantly increased risk of molecular pathological data points to the role
total mortality for weight loss of >5% and signif- of chronic inflammation in the pathogenesis
icantly increased risk of prostate cancer–​specific and progression of prostate cancer (Nelson et al,
mortality for weight gain of >5% compared to 2004). The pathways involved in chronic inflam-
stable weight in the 5–​10 years after diagnosis of mation induce cellular damage and compen-
localized prostate cancer (Bonn et al, 2014). Two satory cellular proliferation (Kuper et  al, 2000).
other studies among men treated with prostatec- Clinical prostatitis has been associated with
tomy reported that weight gain in the year before prostate cancer risk (Palapattu et  al, 2004), and
501

Prostate Cancer 501

poor outcomes among men with disease (Irani an excess prostate cancer risk among men with
et  al, 1999). Infectious agents are likely targets serologic evidence of syphilis (Hayes et al, 2000).
involved in the initiation and exacerbation of In prostate tissue, HSV-​2 is no more prevalent
chronic inflammation. Infectious agents may among cases of prostate cancer than among con-
also have direct effects on carcinogenesis through trols (Boldogh et al, 1983; Haid & Sharon, 1984),
the transformation of cells via incorporation of but there is limited evidence that serum antibod-
active oncogenes into the host genome, inhibi- ies to HSV-​2 may predict prostate cancer risk
tion of tumor suppressors, stimulation of prolif- weakly (Baker et al, 1981; Lüleci et al, 1981).
eration signals, or through immune suppression.
Human Papilloma Virus
Sexually Transmitted Infections A major shift in focus occurred when the human
Epidemiological studies have shown associations papilloma virus (HPV) emerged as a strong
between prostate cancer and several indicators of causal factor for other genitourinary cancers.
sexual activity, such as number of partners, rela- Whether this oncogenic virus contributes to the
tions with prostitutes, and age at first marriage development of prostate cancer remains debated,
(Andersson et  al, 1996). Contradictory reports because epidemiologic results have been radi-
have also been published, however, including one cally divergent. Researchers have attempted to
that noted no deficit of prostate cancer mortality compare HPV detection rates in the prostate tis-
among Catholic priests (Ross et al, 1983; Ewings sue of cancer cases to those in the prostate tissue
& Bowie, 1996; Hsieh et al, 1999). of men with benign prostatic hyperplasia; pros-
The suspicion that correlates of sexual tate tissue from nondiseased subjects is typically
behavior may influence prostate cancer risk has unavailable. The results of various studies argue
prompted numerous investigations of individual against HPV as an important risk factor for pros-
infectious agents. In addition, there is evidence tate cancer (Tu et al, 1994; Wideroff et al, 1996;
that some sexually transmitted infections (STIs) Noda et al, 1998; Strickler et al, 1998) and even
contribute to prostatic inflammation and cell suggest that HPV infection in the prostate may
damage, based on measurements of PSA rises be very rare (Tu et al, 1994; Strickler et al, 1998).
during infection (Sutcliffe et  al, 2006a; Sutcliffe Others have implicated HPV type 16 (Moyret-​
et  al, 2011). A  history of STIs has been associ- Lalle et  al, 1995; Suzuki et  al, 1996; Serth et  al,
ated with prostate cancer, with men reporting a 1999), the HPV type known to be most strongly
history of any STI at 50% increased risk (Caini linked to anogenital cancers. Interestingly, HPV
et  al, 2014); however, no specific pathogen has DNA was detected more often in nonmalignant
yet emerged as a causal factor (Sfanos et  al, tissue than in prostate cancer tissue. A combined
2013). Originally, this research effort focused analysis of several studies estimated that 32%,
on gonorrhea, syphilis, chlamydia, and the her- 49%, and 9% of prostate cancer, benign prostatic
pes simplex virus (HSV). Gonorrhea infection hyperplasia, and normal prostate tissue, respec-
has been linked to prostate cancer, with evidence tively, were HPV DNA positive (Cuzick, 1995).
from ecologic data (Heshmat et  al, 1975)  and This observation raises the question of whether
a population-​ based case-​ control study that benign prostatic hyperplasia tissue is an appro-
reported an RR of 1.6 among men who reported priate control group if, in fact, HPV infection is
a history of either gonorrhea or syphilis (Hayes implicated in its own etiology.
et al, 2000). Another study showed a significant Seroepidemiologic studies have not appre-
threefold increase in risk among men with a his- ciably clarified the situation. While some data
tory of syphilis but no association with arseni- reveal serologic evidence of HPV infection to
cal drugs (an early treatment for syphilis) (Lees be the same among cases and controls (Strickler
et  al, 1985). However, many of the studies have et  al, 1998), one study found that those with
been limited by a small sample size, retrospective antibodies against HPV types 16 or 18 had 2.5-​
exposure assessment, reliance on self-​reported fold greater risk of prostate cancer (Dillner et al,
data on infections, and consideration of exposure 1998). This same study reported no association
to only a limited number of STIs. between HPV types 11 and 13 and prostate can-
Seroepidemiologic study has generally cer. An important contributing factor to the con-
detected no differences in the presence of chla- flicting results may lie in the laboratory analyses.
mydia antibodies between case and control sub- HPV results can be quite dependent on the assay,
jects (Dillner et  al, 1998)  but has demonstrated primer sets, or even annealing temperature
502

502 Textbook of Cancer Epidemiology

used (Sinclair et al, 1993; Terris & Peehl, 1997). samples and all adjacent samples, but not in
Moreover, viral detection can be variable in dif- prostate samples from healthy individuals (Chen
ferent specimens from the same patient’s tissue et  al, 2015). Another study found that prostatic
(Terris & Peehl, 1997). P.  acnes was associated with increased prostatic
inflammation (Cohen et  al, 2005). The role of
Trichomonas Vaginalis P. acnes in prostate cancer development and pro-
The parasitic protozoan T.  vaginalis is the most gression warrants further study.
common nonviral STI (Petrin et  al, 1998), with
a prevalence of 5%–​20% among US young adults Physical Activity
(Plitt et al, 2005) and globally affecting 170 mil- Physical activity is suspected to reduce the risk of
lion each year. Typically the infection is asymp- several types of cancer via biological mechanisms
tomatic (Petrin et  al, 1998)  and does not clear, that, although poorly understood, may involve,
such that infection can be detected several years for example, enhanced immune system function
after initial exposure. T.  vaginalis has received (Lee, 1995), changes in the endogenous hormonal
considerably less attention in prostate cancer milieu, or reduced obesity (Simopoulos, 1990).
research than infections such as Chlamydia or Physical activity has not been associated
gonorrhea. In men, T.  vaginalis has been asso- with overall prostate cancer risk. Two prospec-
ciated with an increased risk of clinical prosta- tive cohort studies, HPFS (Giovannucci et  al,
titis (Sutcliffe et  al, 2006b). Trichomonas can 2005)  and the CPS II (Patel et  al, 2005), found
infect and elicit a strong inflammatory response inverse associations between higher levels of rec-
within the prostate (Gardner et  al, 1986; Van reational physical activity and the risk of advanced
Laarhoven, 1967). T.  vaginalis adhesion to epi- or fatal disease, independent of BMI. However,
thelial cells upregulates expression of antiapop- three other cohorts, EPIC (Johnsen et al, 2009),
topic genes including defender against cell death the NIH-​AARP Diet and Health Study (Moore
and cycloxygenase-​2, as well as proinflamma- et  al, 2008), and the Swedish National March
tory genes such as interleukin-​8 and monocyte Cohort (Grotta et  al, 2015), found no associa-
chemoattractant protein (Kucknoor et al, 2005). tions between greater activity and risk of disease
T. vaginalis is adept at evading the host immune by stage or grade. In addition, the PHS found no
system through a number of mechanisms, and association between physical activity and risk of
reinfection does not confer immunity in humans. lethal prostate cancer (Kenfield et al, 2015).
Nevertheless, antibodies to T.  vaginalis can be The assessment of long-​term physical activ-
detected in the serum of infected patients. ity levels is challenging. Study participants are
Two large nested case-​ control studies of often asked to report on the type, intensity, and
T.  vaginalis seroprevalence and prostate cancer duration of their average physical activity, both
found significantly increased risks. HPFS found a currently and in the past. The resulting mis-
43% increase in the odds of total prostate cancer classification may be responsible for the weak
(95% CI 1.00–​2.03) and 76% increase in the odds and often nonsignificant findings. Subgroups
of high-​grade prostate cancer (95% CI 0.97–​3.18) less prone to measurement error, such as, men
among men who were seropositive (Sutcliffe et al, who engage in a serious and regular regimen of
2006b). PHS found significantly increased risk vigorous activity, may offer the best chance of
for both extraprostatic disease (OR 2.17, 95% CI detecting a relationship between exercise and
1.08–​4.37) and lethal disease (OR 2.69, 95% CI prostate cancer if one exists. HPFS was based
1.37–​5.28) (Stark et  al, 2009). One other study on repeated prospective assessments of physical
found no association with self-​reported history activity every 2 years over 14 years of follow-​up,
of T. vaginalis infection (Checkoway et al, 1987). and found significantly lower risks of advanced
However, the infection is most frequently asymp- and fatal disease only for high levels of vigor-
tomatic in men, making self-​report an unreliable ous activity, not for more moderate activity
indicator of infection. (Giovannucci et  al, 2005). It is possible that
repeated assessments of physical activity over
Propionibacterium Acnes time and study populations with a wide range
In a study using RNA-​Seq to look for genomic of activity, including very active participants,
signatures of nine pathogens in prostate tumor are required for an inverse association between
and adjacent normal tissue from healthy men, physical activity and advanced prostate cancer
P. acnes genes were detected in all prostate tumor risk to emerge. Thus, there may be an inverse
503

Prostate Cancer 503

association between very high levels of phys- military exposures (e.g., atomic bomb survivors).
ical activity through adulthood and risk of There is no evidence of excess prostate cancer
advanced prostate cancer. incidence or death among even highly exposed
populations (Boice et al, 1996).
Survival
Physical activity may improve prostate cancer Occupation
survival, and may ameliorate some of the adverse Prostate epithelial cells contain the highest con-
effects of therapy (Bourke et  al, 2015). Physical centration of zinc in the human body (Kolenko
activity has a wide range of whole-​body effects, et al, 2013). This essential trace element plays a
including reduction of cardiovascular mark- key role in production of prostatic fluid by nor-
ers such as insulin and cholesterol (Durstine mal prostate cells, and has also been implicated
& Haskell, 1994; Kokkinos & Fernhall, 1999; in the normal regulation of cell growth and
Kokkinos & Myers, 2010), decrease in BMI, may play a critical role in DNA and RNA repair
reduced risk of type 2 diabetes, and reduction (Kolenko et al, 2013; Chan et al, 1998). Cadmium
of inflammatory markers (Helmrich et al, 1991; is a divalent cation, like zinc, and acts as a zinc
Manson et  al, 1992; Abramson & Vaccarino, antagonist (Kolonel & Winkelstein, 1977; Feustel
2002; Ford ES, 2002). Both obesity and metabolic et  al, 1984; Brys et  al, 1997). For this reason,
syndrome have been associated with increased occupational cadmium exposure has been exten-
risk of advanced prostate cancer and worse sively studied with regard to prostate cancer risk.
prostate cancer–​specific survival and response Early occupational studies that addressed this
to hormonal therapy (Flanagan et  al, 2011). In issue were hindered by small numbers of both
addition, surgery and radiation for treatment of expected and observed prostate cancer cases,
prostate cancer can cause erectile dysfunction, and they reported modest excesses for cadmium-​
incontinence, and radiation proctitis, and ADT exposed workers (Lemen et al, 1976; Kolonel &
can cause loss of muscle mass, fatigue, increase Winkelstein, 1977). These men were engaged
risk of cardiovascular disease, and lower bone in a variety of occupations including smelting,
density (Bourke et al, 2015). alkaline battery manufacturing, welding, and
The few observational studies of activity electroplating. A  later study that was popula-
after diagnosis and prostate cancer progres- tion based and that identified prostate cancer
sion have reported beneficial associations. In cases using a cancer registry did not find any
HPFS, brisk walking among prostate cancer positive association with occupations involving
patients was associated with a lower risk of potential cadmium exposure (Ross et  al, 1979).
total mortality (as expected) and a 40% lower Further studies have also not corroborated the
risk disease-​specific mortality (Kenfield et  al, early reports from the small occupational cohorts
2011b). A  study of PSA recurrence in pros- (Chan et  al, 1998). However, two case-​control
tate cancer patients found similar decreases in studies have used toenail levels of cadmium as
risk with higher activity levels (Richman et al, a more accurate individual-​ level measure of
2011b); this is notable, as PSA recurrence as an exposure with suggestive results. A nested case-​
outcome is less susceptible to reverse causation, control study within the CLUE II cohort found
that is, decreasing activity levels due to progres- no association between cadmium exposure and
sion of disease, than is prostate cancer mortal- total prostate cancer risk, but there was a sugges-
ity. A  meta-​analysis of exercise interventions tion of increased risk for extraprostatic disease,
in men with prostate cancer found a moderate with an odds ratio of 2.06 (p-​trend = 0.12) for the
improvement in quality of life measures over 12 highest versus lowest tertile; however, case num-
weeks in the studies of highest methodological bers were low (Platz et al, 2002). A hospital-​based
quality (Bourke et al, 2015). case-​control study in Italy, in a population with
higher cadmium exposure and lower rates of PSA
Ionizing Radiation testing, found a positive association between toe-
Prostate cancer is exceptional in that it is among nail cadmium and prostate cancer risk, with an
the few malignancies plainly unrelated to radia- OR for the top versus bottom quartile of 4.7 (95%
tion exposure (Boice et al, 1996). Routes of ioniz- CI 1.3–​17.5) (Vinceti et al, 2007). Thus the asso-
ing radiation exposure include medical treatment ciation between cadmium and prostate cancer
(e.g., radiation therapy for cancer), occupational risk is inconclusive; there may be an association
contact (e.g., workers at nuclear facilities), and for more advanced disease.
504

504 Textbook of Cancer Epidemiology

Medical Conditions vasectomy, with a hazard ratio of 1.56 (95% CI

and Treatment 1.03–​2.36); this further supports the conclu-
Vasectomy sion that the association is not due simply to
Results of several epidemiologic studies dur- detection bias (Siddiqui et  al, 2014). Primary
ing the past decades have generated the suspi- treatment choices were also similar between
cion that vasectomy may increase the risk of the two groups, arguing against a difference in
prostate cancer (Bernal-​ Delgado et  al, 1998). treatment preferences leading to a difference in
Two meta-​analyses reported very similar, mar- prostate cancer outcomes.
ginally increased risks for total prostate cancer These results suggest that the vasectomy-​
associated with vasectomy. An analysis of nine prostate cancer association should be revisited
cohort studies found a RR of 1.08 (95% CI 0.87–​ specifically with regard to advanced or fatal
1.34) (Liu et al, 2015), and a study of 10 cohorts disease and with careful consideration of PSA
reported a RR of 1.11 (95% CI 0.98–​1.27) (Shang screening behaviors in cohorts with the appro-
et al, 2015). priate data. It appears that there is a small asso-
The most commonly offered biological expla- ciation of vasectomy with overall prostate cancer
nation for the involvement of vasectomy in the risk, and possibly larger associations with risk of
etiology of prostate cancer is the potential for advanced disease.
this surgery to alter endocrine function (Howard,
1993). Longitudinal data from small numbers of Perinatal Factors
men followed after vasectomy overall reveal no Ecologic data, including observations that cir-
consistent effect on circulating sex hormone culating steroid levels are markedly different in
concentrations (Richards et  al, 1981), although pregnant African American and Chinese versus
one multiethnic study among 850 healthy men white women (Henderson et  al, 1988; Lipworth
with normal PSA levels found that vasectomized et al, 1999), that populations with higher average
men had a lower serum concentration of sex birthweights have a higher incidence of prostate
hormone-​binding globulin and a higher ratio of cancer (Wang et al, 1994; Lawson, 1998), and that
dihydrotestosterone to testosterone (John et  al, the secular trend of increasing worldwide prostate
1995). Immunologic effects have also been sug- cancer incidence was preceded by a correspond-
gested (Flickinger et al, 1999; Verdier et al, 2002), ing increase in average birthweight (Ekbom et al,
as have changes in seminal fluid protein compo- 1996), provide some support for the hypothesis
sition (Batruch et al, 2011). that in utero exposures may influence the future
Alternatively, detection bias has been sug- risk of prostate cancer. Direct testing through ana-
gested as a potential explanation, since vasecto- lytic epidemiologic investigation, however, has not
mized men may be more likely to visit a urologist demonstrated a relationship between factors such
subsequently and would therefore be at higher as birthweight, birth length, placental weight, or
risk of cancer detection by routine prostate maternal age and prostate cancer risk, although
screening than are men who do not have the an inverse association with increasing duration of
motive to visit a urologist. gestation has been reported (Ekbom et  al, 2000;
An update of a previous analysis of vasec- Cook et  al, 2013; Cnattingius et  al, 2009; Nilsen
tomy within the HPFS cohort again found an et al, 2005; Platz et al, 1998; Ekbom et al, 1996).
increased risk of prostate cancer among men However, only one of these studies assessed risk of
reporting a history of vasectomy. The associ- advanced or fatal disease, and this study was una-
ation was stronger for lethal (death or distant ble to rule out associations with high-​stage/​high-​
metastasis) cancer than for overall or localized grade disease (Platz et  al, 1998). A  case-​control
prostate cancer, which argues against detec- study in Sweden found a significantly lower risk of
tion bias as an explanation, as more intensive total prostate cancer and aggressive disease among
PSA testing among vasectomized men would men with low birthweights, <3,000 g, compared
lead to increased detection of localized and to average birthweight and with adjustment for
low-​grade cancers but possibly lower risk of adult factors (Lahmann et al, 2012). Among 1,436
fatal prostate cancer (given enough follow-​ Swedish men followed until age 85, high birth-
up time for screening to impact mortality). weight, ≥4,250 g, was associated with increased
An analysis among a subgroup of men in the risk of both prostate cancer incidence and mortal-
cohort reporting frequent PSA screening found ity compared to average birthweights, apparently
further increased risk of lethal cancer with independent of adult anthropometric factors;
50

TABLE 20-​1   EPIDEMIOLOGY OF PROSTATE CANCER:

RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Age ↑↑
Family history in first-​degree relative ↑↑
African American race ↑↑
Germ-​line mutations in DNA repair genes (BRCA1/​BRCA2) ↑↑
Height ↑

Probable Relationship Exists, Based on Substantial Data

Smoking ↑ (advanced disease)
Dairy intake ↑
Fish intake ↓ (advanced disease)
Lycopene/​tomato intake ↓
Insulin-​like growth factor 1 ↑
Obesity/​body mass index ↑ (advanced disease)
Obesity/​body mass index ↓ (localized disease)
Physical activity (vigorous) ↓ (advanced disease)

Weak, if Any, Relationship Exists, Based on Substantial Data

Total fat intake —​
Alcohol —​
Alpha-​tocopherol supplements —​
Selenium supplements —​
Androgen levels —​
Estrogen levels —​
Sex-​hormone binding globulin levels —​
IGF-​1 binding proteins —​
Childhood/​young adult body size —​
Ionizing radiation —​

Inconsistent Findings or Limited Study to Date

Long-​chain fatty acid intake —​
Alpha-​linolenic acid intake ↑
Red & processed meat intake —​
Calcium intake ↑
Vitamin D ↓ (advanced disease)
Coffee intake ↓ (advanced disease)
Soy/​phytoestrogen intake ↓
Dietary vitamin E intake ↓ (advanced disease)
Dietary selenium intake ↓
History of sexually transmitted infections ↑
Vasectomy ↑ (advanced disease)

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate
to large increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association
506

506 Textbook of Cancer Epidemiology

however, multivariable adjusted results were not an exciting area of study, as men diagnosed with
presented in the report (Eriksson et  al, 2007). prostate cancer seek information on what might
Further analysis of the association between birth- improve their prognosis and quality of life.
weight and perhaps other perinatal factors spe-
cifically with respect to advanced disease may be REFERENCES
warranted. Abramson JL, Vaccarino V. Relationship between
physical activity and inflammation among appar-
CONCLUSION ently healthy middle-​aged and older US adults.
For total prostate cancer risk, there is good evi- Arch Intern Med 2002;162(11):1286–​92.
dence that older age, African American race, Adami HO, McLaughlin JK, Hsing AW, Wolk A,
taller height, and a positive family history are Ekbom A, Holmberg L, Persson I. Alcoholism
established risk factors (Table 20-​1). Moreover, and cancer risk: a population-​based cohort study.
Cancer Causes Control 1992;3:419–​25.
there are now >100 confirmed inherited genetic
Adlercreutz H, Mazur W. Phyto-​ oestrogens and
risks for disease. The results of more recent epi-
Western diseases. Ann Intern Med 1997;29:95–​120.
demiological studies have shed light on a greater
Al Olama AA, Kote-​ Jarai Z, Berndt SI, Conti DV,
number of factors that may influence risk of
Schumacher F, Han Y, Benlloch S, et  al.; Breast
advanced or fatal prostate cancer, or cancer mor- and Prostate Cancer Cohort Consortium (BPC3);
tality after diagnosis. PRACTICAL (Prostate Cancer Association Group
Prostate cancer has an unusual heterogeneity to Investigate Cancer-​ Associated Alterations in
and natural history that complicates interpreta- the Genome) Consortium; COGS (Collaborative
tion of both descriptive and analytic studies. This Oncological Gene-​environment Study) Consortium;
complication arises due to the high prevalence of GAME-​ON/​ELLIPSE Consortium, Cook MB,
subclinical cancers among elderly men, which are Nakagawa H, Wiklund F, Kraft P, Chanock SJ,
now detectable by PSA testing, combined with a Henderson BE, Easton DF, Eeles RA, Haiman CA.
very long natural history even for fatal disease. A meta-​analysis of 87,040 individuals identifies
Hence, diagnostic intensity may substantially 23 new susceptibility loci for prostate cancer. Nat
influence the observed incidence of prostate can- Genet 2014;46(10):1103–​9.
cer and thereby confound comparisons among Albanes D, Mondul AM, Yu K, Parisi D, Horst RL,
geographic areas and among different time peri- Virtamo J, Weinstein SJ. Serum 25-​hydroxy vita-
ods. And extended follow-​up time is required to min D and prostate cancer risk in a large nested
study risk factors with respect to fatal disease. case-​control study. Cancer Epidemiol Biomarkers
In analytic epidemiologic studies, it may be Prev 2011;20(9):1850–​60.
hard or impossible to tell whether identified risk Albertsen PC, Hanley JA, Fine J. 20-​ year out-
factors play a role in malignant transformation or comes following conservative management
of clinically localized prostate cancer. JAMA.
merely in progression from indolent to clinically
2005;293(17):2095–​101.
significant cancer. To deal with this problem, epi-
Alexander DD, Mink PJ, Cushing CA, Sceurman B. A
demiologists more and more often analyze local-
review and meta-​analysis of prospective studies of
ized and advanced cancers separately. A stronger
red and processed meat intake and prostate can-
association with advanced cancers would then cer. Nutr J 2010;9:50.
suggest a promoting rather than an initiating Alexander DD, Bassett JK, Weed DL, Barrett EC,
effect of the particular risk factor. In popula- Watson H, Harris W. Meta-​Analysis of Long-​
tions with high levels of PSA screening, associa- Chain Omega-​ 3 Polyunsaturated Fatty Acids
tions of risk factors mainly with localized and/​ (LCω-​3PUFA) and Prostate Cancer. Nutr Cancer
or low-​grade disease suggests the possibility of 2015;67(4):543–​54.
confounding by PSA screening rather than iden- Allemani C, Weir HK, Carreira H, Harewood R,
tification of true risk factors for disease. Spika D, Wang XS, et  al.; CONCORD Working
However, since prostate cancer began to Group. Global surveillance of cancer sur-
attract more epidemiologic interest in the 1980s, vival 1995–​2009:  analysis of individual data for
numerous etiologic clues have been identified, 25,676,887 patients from 279 population-​based
and aspects of diet, hormones, and sexually registries in 67 countries (CONCORD-​2). Lancet
transmitted infections, among other factors, have 2015;385(9972):977–​1010.
developed as promising topics in the search for Allen NE, Travis RC, Appleby PN, Albanes D, Barnett
causes of prostate cancer. In addition, the role of MJ, Black A, et  al.; Endogenous Hormones,
diet and lifestyle after diagnosis is emerging as Nutritional Biomarkers and Prostate Cancer
507

Prostate Cancer 507

Collaborative Group. Selenium and Prostate Baker LH, Mebust WK, Chin TD, Chapman AL,
Cancer:  Analysis of Individual Participant Data Hinthorn D, Towle D. The relationship of her-
From Fifteen Prospective Studies. J Natl Cancer pesvirus to carcinoma of the prostate. J Urol
Inst. 2016 Jul 6;108(11). 1981;125(3):370–​4.
Allen NE, Sauvaget C, Roddam AW, Appleby P, Bancroft EK, Page EC, Castro E, Lilja H, Vickers A,
Nagano J, Suzuki G, Key TJ, Koyama K. A prospec- Sjoberg D, et  al.; IMPACT Collaborators, Moss
tive study of diet and prostate cancer in Japanese S, Eeles RA. Targeted prostate cancer screening
men. Cancer Causes Control 2004;15(9):911–​20. in BRCA1 and BRCA2 mutation carriers: results
The Alpha-​ Tocopherol, Beta Carotene Cancer from the initial screening round of the IMPACT
Prevention Study Group. The effect of vitamin E study. Eur Urol 2014;66(3):489–​99.
and beta carotene on the incidence of lung cancer Barba M, Terrenato I, Schünemann HJ, Fuhrman B,
and other cancers in male smokers. N Engl J Med Sperati F, Teter B, Gallucci M, D’Amato A, Muti P.
1994;330(15):1029–​35. Indicators of sexual and somatic development and
Amundadottir LT, Sulem P, Gudmundsson J, adolescent body size in relation to prostate can-
Helgason A, Baker A, Agnarsson BA, et  al. A cer risk: results from a case-​control study. Urology
common variant associated with prostate cancer 2008;72(1):183–​7.
in European and African populations. Nat Genet Bassett JK, Severi G, Baglietto L, MacInnis RJ, Hoang
2006;38(6):652–​8. HN, Hopper JL, English DR, Giles GG. Weight
Andersson SO, Baron J, Wolk A, Lindgren C, change and prostate cancer incidence and mortal-
Bergström R, Adami HO. Early life risk factors for ity. Int J Cancer 2012;131(7):1711–​9.
prostate cancer:  a population-​based case-​control Batruch I, Lecker I, Kagedan D, Smith CR, Mullen
study in Sweden. Cancer Epidemiol Biomarkers BJ, Grober E, Lo KC, Diamandis EP, Jarvi KA.
Prev 1995;4(3):187–​92. Proteomic analysis of seminal plasma from nor-
Andersson SO, Baron J, Bergstrom R, Lindgren C, mal volunteers and post-​ vasectomy patients
Wolk A, Adami HO. Lifestyle factors and prostate identifies over 2000 proteins and candidate bio-
cancer risk: a case-​control study in Sweden. Cancer markers of the urogenital system. J Proteome Res
Epidemiol Biomarkers Prev 1996;5(7):509–​13. 2011;10(3):941–​53.
Andriole GL, Crawford ED, Grubb RL 3rd, Buys Batty GD, Barzi F, Woodward M, Jamrozik K, Woo
SS, Chia D, Church TR, et  al.; PLCO Project J, Kim HC, Ueshima H, Huxley RR; Asia Pacific
Team. Mortality results from a randomized Cohort Studies Collaboration. Adult height
prostate-​cancer screening trial. N Engl J Med and cancer mortality in Asia:  the Asia Pacific
2009;360(13):1310–​9. Cohort Studies Collaboration. Ann Oncol
Andriole GL, Crawford ED, Grubb RL 3rd, Buys SS, 2010;21(3):646–​54.
Chia D, Church TR, et  al.; PLCO Project Team. Beilin J, Ball EM, Favaloro JM, Zajac JD. Effect of the
Prostate cancer screening in the randomized androgen receptor CAG repeat polymorphism
Prostate, Lung, Colorectal, and Ovarian Cancer on transcriptional activity:  specificity in prostate
Screening Trial: mortality results after 13 years of and non-​prostate cell lines. J Mol Endrocrinol
follow-​up. J Natl Cancer Inst 2012;104(2):125–​32. 2000;25:85–​96.
Andriole GL, Bostwick DG, Brawley OW, Gomella LG, Beilin J, Harewood L, Frydenberg M, Mameghan H,
Marberger M, Montorsi F, et al.; REDUCE Study Martyres RF, Farish SJ, et al. A case-​control study
Group. Effect of dutasteride on the risk of prostate of the androgen receptor gene CAG repeat poly-
cancer. N Engl J Med 2010;362(13):1192–​202. morphism in Australian prostate carcinoma sub-
Augustsson K, Michaud DS, Rimm EB, Leitzmann jects. Cancer 2001;92:941–​9.
MF, Stampfer MJ, Willett WC, Giovannucci E. Bernal-​ Delgado E, Latour-​ Perez J, Pradas-​ Arnal F,
A prospective study of intake of fish and marine Gomez-​Lopez LI. The association between vasec-
fatty acids and prostate cancer. Cancer Epidemiol tomy and prostate cancer: a systematic review of
Biomarkers Prev 2003;12(1):64–​7. the literature. Fertil Steril 1998;70(2):191–​200.
Aune D, Navarro Rosenblatt DA, Chan DS, Vieira AR, Berry R, Schroeder JJ, French AJ, McDonnell SK,
Vieira R, Greenwood DC, Vatten LJ, Norat T. Dairy Peterson BJ, Cunningham JM, et al. Evidence for
products, calcium, and prostate cancer risk: a sys- a prostate cancer–​susceptibility locus on chromo-
tematic review and meta-​analysis of cohort stud- some 20. Am J Hum Genet 2000;67:82–​91.
ies. Am J Clin Nutr 2015;101(1):87–​117. Berruti A, Dogliotti L, Gorzegno G, Torta M,
Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, Tampellini M, Tucci M, et al. Differential patterns
Fedirko V, et  al. Alcohol consumption and site-​ of bone turnover in relation to bone pain and dis-
specific cancer risk: a comprehensive dose-​response ease extent in bone in cancer patients with skeletal
meta-​analysis. Br J Cancer 2015;112(3):580–​93. metastases. Clin Chem 1999;45(8 Pt 1):1240–​7.
508

508 Textbook of Cancer Epidemiology

Bill-​Axelson A, Holmberg L, Garmo H, Rider JR, Bratt O, Borg A, Kristoffersson U, Lundgren R, Zhang
Taari K, Busch C, et al. Radical prostatectomy or QX, Olsson H. CAG repeat length in the androgen
watchful waiting in early prostate cancer. N Engl J receptor gene is related to age at diagnosis of prostate
Med 2014;370(10):932–​42. cancer and response to endocrine therapy, but not
Black A, Pinsky PF, Grubb RL 3rd, Falk RT, Hsing to prostate cancer risk. Br J Cancer 1999;81: 672–​6.
AW, Chu L, et  al. Sex steroid hormone metab- Brawley OW, Knopf K, Merrill R. The epidemiology of
olism in relation to risk of aggressive prostate prostate cancer part I:  descriptive epidemiology.
cancer. Cancer Epidemiol Biomarkers Prev Semin Urol Oncol 1998;16:187–​92.
2014;23(11):2374–​82. Breast Cancer Linkage Consortium. Cancer risks
Blot WJ, Li JY, Taylor PR, Guo W, Dawsey S, Wang in BRCA2 mutation carriers. J Natl Cancer Inst
GQ, et al. Nutrition intervention trials in Linxian, 1999;91(15):1310–​6.
China:  supplementation with specific vitamin/​ Brys M, Nawrocka AD, Miekos E, Zydek C, Foksinski
mineral combinations, cancer incidence, and M, Barecki A, et al. Zinc and cadmium analysis in
disease-​specific mortality in the general popula- human prostate neoplasms. Biol Trace Elem Res
tion. J Natl Cancer Inst 1993;85(18):1483–​92. 1997;59(1–​3):145–​52.
Bodiwala D, Luscombe CJ, French ME, Liu S, Bylund A, Zhang JX, Bergh A, Damber JE, Widmark
Saxby MF, Jones PW, Fryer AA, Strange RC. A, Johansson A, et  al. Rye bran and soy protein
Polymorphisms in the vitamin D receptor gene, delay growth and increase apoptosis of human
ultraviolet radiation, and susceptibility to prostate LNCaP prostate adenocarcinoma in nude mice.
cancer. Environ Mol Mutagen 2004;43:121–​7. Prostate 2000;42:304–​14.
Boice JD Jr, Land CE, Preston DL. Ionizing radiation. Caini S, Gandini S, Dudas M, Bremer V, Severi E,
In:  Schottenfeld D, Fraumeni JRJr (Eds). Cancer Gherasim A. Sexually transmitted infections and
Epidemiology and Prevention, 2nd ed. New York, prostate cancer risk: a systematic review and meta-​
Oxford University Press, 1996, pp. 319–​54. analysis. Cancer Epidemiol 2014;38(4):329–​38.
Boldogh I, Baskar JF, Mar EC, Huang ES. Human Cancer Genome Atlas Research Network. The
cytomegalovirus and herpes simplex type 2 virus Molecular Taxonomy of Primary Prostate Cancer.
in normal and adenocarcinomatous prostate Cell 2015;163(4):1011–​25.
glands. J Natl Cancer Inst 1983;70(5):819–​26. Cantarutti A, Bonn SE, Adami HO, Grönberg H,
Bonn SE, Wiklund F, Sjölander A, Szulkin R, Stattin Bellocco R, Bälter K. Body mass index and mor-
P, Holmberg E, Grönberg H, Bälter K. Body mass tality in men with prostate cancer. Prostate
index and weight change in men with prostate 2015;75(11):1129–​36.
cancer: progression and mortality. Cancer Causes Cao S, Liu L, Yin X, Wang Y, Liu J, Lu Z. Coffee
Control 2014;25(8):933–​43. consumption and risk of prostate cancer:  a
Bourke L, Smith D, Steed L, Hooper R, Carter A, Catto meta-​ analysis of prospective cohort studies.
J, Albertsen PC, Tombal B, Payne HA, Rosario Carcinogenesis 2014;35(2):256–​61.
DJ. Exercise for Men with Prostate Cancer:  A Cao Y, Nimptsch K, Shui IM, Platz EA, Wu
Systematic Review and Meta-​analysis. Eur Urol K, Pollak MN, Kenfield SA, Stampfer MJ,
2015 [Epub ahead of print]. Giovannucci EL. Prediagnostic plasma IGFBP-​
Brändstedt J, Almquist M, Manjer J, Malm J. Vitamin 1, IGF-​1 and risk of prostate cancer. Int J Cancer
D, PTH, and calcium in relation to survival fol- 2015;136(10):2418–​26.
lowing prostate cancer. Cancer Causes Control. Carayol M, Grosclaude P, Delpierre C. Prospective
2016 May;27(5):669–​77. studies of dietary alpha-​ linolenic acid intake
Brasky TM, Till C, White E, Neuhouser ML, Song and prostate cancer risk: a meta-​analysis. Cancer
X, Goodman P, Thompson IM, King IB, Albanes Causes Control 2010;21(3):347–​55.
D, Kristal AR. Serum phospholipid fatty acids Carleton AJ, Sievenpiper JL, de Souza R, McKeown-​
and prostate cancer risk:  results from the pros- Eyssen G, Jenkins DJ. Case-​control and prospec-
tate cancer prevention trial. Am J Epidemiol tive studies of dietary α-​linolenic acid intake and
2011;173(12):1429–​39. prostate cancer risk:  a meta-​analysis. BMJ Open
Brasky TM, Darke AK, Song X, Tangen CM, Goodman 2013;3(5)pii:e002280.
PJ, Thompson IM, et al. Plasma phospholipid fatty Carpten JD, Nupponen N, Isaacs SD, et al. Germline
acids and prostate cancer risk in the SELECT trial. mutations in the ribonuclease L gene in families
J Natl Cancer Inst 2013;105(15):1132–​41. showing linkage with HPC1. Nature Genetics
Bratt O, Kristoffersson U, Lundgren R, Olsson H. 2002;30:181–​184.
Familial and hereditary prostate cancer in south- Carruba G. Estrogen and prostate cancer: an eclipsed
ern Sweden. A population-​ based case-​ control truth in an androgen-​dominated scenario. J Cell
study. Eur J Cancer 1999;35:272–​7. Biochem 2007;102(4):899–​911.
509

Prostate Cancer 509

Carter BD, Abnet CC, Feskanich D, Freedman ND, prostate cancer at diagnosis:  findings from the
Hartge P, Lewis CE, Ockene JK, Prentice RL, prostate testing for cancer and treatment study
Speizer FE, Thun MJ, Jacobs EJ. Smoking and and a systematic review. Cancer Epidemiol
mortality-​-b
​ eyond established causes. N Engl J Biomarkers Prev 2009;18(11):2874–​81.
Med 2015;372(7):631–​40. Chen ML, Lin YH, Yang CM, Hu ML. Lycopene
Carter BS, Beaty TH, Steinberg GD, Childs B, Walsh inhibits angiogenesis both in vitro and in vivo by
PC. Mendelian inheritance of familial prostate inhibiting MMP-​2/​uPA system through VEGFR2-​
cancer. Proc Natl Acad Sci USA 1992;89:3367–​71. mediated PI3K-​Akt and ERK/​p38 signaling path-
Casey G, Neville PJ, Plummer SJ, et  al. RNASEL ways. Mol Nutr Food Res 2012;56(6):889–​99.
Arg462Gln variant is implicated in up to 13% of Chen Y, Wei J. Identification of Pathogen Signatures
prostate cancer cases. Nat Genet 2002;32(4):581–​3. in Prostate Cancer Using RNA-​seq. PLoS One
Castro E, Goh C, Olmos D, Saunders E, Leongamornlert 2015;10(6):e0128955.
D, Tymrakiewicz M, et al. Germline BRCA muta- Chen YC, Giovannucci E, Lazarus R, Kraft P, Ketkar S,
tions are associated with higher risk of nodal Hunter DJ. Sequence variants of Toll-​like receptor
involvement, distant metastasis, and poor sur- 4 and susceptibility to prostate cancer. Cancer Res
vival outcomes in prostate cancer. J Clin Oncol 2005 Dec 15;65(24):11771–​8.
2013;31(14):1748–​57. Cheng I, Stram DO, Penney KL, Pike M, Le Marchand
Castro E, Goh C, Leongamornlert D, Saunders L, Kolonel LN, Hirschhorn J, Altshuler D,
E, Tymrakiewicz M, Dadaev T, et  al. Effect of Henderson BE, Freedman ML. Common genetic
BRCA Mutations on Metastatic Relapse and variation in IGF1 and prostate cancer risk in the
Cause-​specific Survival After Radical Treatment Multiethnic Cohort. J Natl Cancer Inst 2006;
for Localised Prostate Cancer. Eur Urol 98(2):123–​34.
2015;68(2):186–​93. Ciatto S, Gervasi G, Bonardi R, Frullini P, Zendron P,
Center MM, Jemal A, Lortet-​ Tieulent J, Ward E, Lombardi C, Crocetti E, Zappa M. Determining
Ferlay J, Brawley O, Bray F. International variation overdiagnosis by screening with DRE/​TRUS or
in prostate cancer incidence and mortality rates. PSA (Florence pilot studies, 1991–​1994). Eur J
Eur Urol 2012;61(6):1079–​92. Cancer 2005;41(3):411–​5.
Cerhan JR, Parker AS, Putnam SD, Chiu B C-​ H, Clark LC, Combs GF Jr, Turnbull BW, Slate EH,
Lynch CF, Cohen MB, et  al. Family history and Chalker DK, Chow J, et  al. Effects of selenium
prostate cancer risk in a population-​based cohort supplementation for cancer prevention in patients
of Iowa men. Cancer Epidemiol Biomarkers Prev with carcinoma of the skin. A  randomized con-
1999;8:53–​60. trolled trial. Nutritional Prevention of Cancer
Chan JM, Stampfer MJ, Giovannucci EL. What causes Study Group. JAMA 1996;276(24):1957–​63.
prostate cancer? A brief summary of the epidemi- Cnattingius S, Lundberg F, Sandin S, Grönberg H,
ology. Semin Cancer Biol 1998b;8:263–​73. Iliadou A. Birth characteristics and risk of prostate
Chan JM, Gann PH, Giovannucci EL. Role of diet in cancer: the contribution of genetic factors. Cancer
prostate cancer development and progression. J Epidemiol Biomarkers Prev 2009;18(9):2422–​6.
Clin Oncol 2005;23(32):8152–​60. Review. Coffey DS. Physiological control of prostate growth.
Chan JM, Holick CN, Leitzmann MF, Rimm EB, An overview. In: Prostate Cancer. UICC Technical
Willett WC, Stampfer MJ, Giovannucci EL. Diet Report Series. Vol. 48. Geneva, International
after diagnosis and the risk of prostate cancer pro- Union Against Cancer, 1979.
gression, recurrence, and death (United States). Cohen RJ, Shannon BA, McNeal JE, Shannon T,
Cancer Causes Control 2006;17(2):199–​208. Garrett KL. Propionibacterium acnes associated
Chan JM, Stampfer MJ, Ma J, Rimm EB, Willett WC, with inflammation in radical prostatectomy speci-
Giovannucci EL. Supplemental vitamin E intake mens: a possible link to cancer evolution? J Urol
and prostate cancer risk in a large cohort of men in 2005;173(6):1969–​74.
the United States. Cancer Epidemiol Biomarkers Combs GF Jr, Combs SB. The nutritional biochemis-
Prev 1999;8(10):893–​9. try of selenium. Annu Rev Nutr 1984;4:257–​80.
Chavarro JE, Stampfer MJ, Li H, Campos H, Kurth Cook MB, Gamborg M, Aarestrup J, Sørensen TI,
T, Ma J. A prospective study of polyunsatu- Baker JL. Childhood height and birth weight in
rated fatty acid levels in blood and prostate can- relation to future prostate cancer risk:  a cohort
cer risk. Cancer Epidemiol Biomarkers Prev study based on the copenhagen school health
2007;16(7):1364–​70. records register. Cancer Epidemiol Biomarkers
Chen L, Davey Smith G, Evans DM, Cox A, Lawlor Prev 2013;22(12):2232–​40.
DA, Donovan J, et  al. Genetic variants in the Cooney KA, McCarthy JD, Lange E, Huang L,
vitamin d receptor are associated with advanced Miesfeldt S, Montie JE, et  al. Prostate cancer
510

510 Textbook of Cancer Epidemiology

susceptibility locus on chromosome lq: a confirm- in relation to prostate cancer survival. Int J Cancer.
atory study. J Natl Cancer Inst 1997;89:955–​9. 2017 Feb 10.
Cross AJ, Peters U, Kirsh VA, Andriole GL, Reding D, Duffield-​Lillico AJ, Dalkin BL, Reid ME, Turnbull
Hayes RB, Sinha R. A prospective study of meat BW, Slate EH, Jacobs ET, Marshall JR, Clark LC;
and meat mutagens and prostate cancer risk. Nutritional Prevention of Cancer Study Group.
Cancer Res 2005;65(24):11779–​84. Selenium supplementation, baseline plasma sele-
Crowe FL, Allen NE, Appleby PN, Overvad K, nium status and incidence of prostate cancer: an
Aardestrup IV, Johnsen NF, et al. Fatty acid com- analysis of the complete treatment period of the
position of plasma phospholipids and risk of Nutritional Prevention of Cancer Trial. BJU Int
prostate cancer in a case-​control analysis nested 2003;91(7):608–​12.
within the European Prospective Investigation Durstine JL, Haskell WL. Effects of exercise training
into Cancer and Nutrition. Am J Clin Nutr on plasma lipids and lipoproteins. Exerc Sport Sci
2008;88(5):1353–​63. Rev 1994;22:477–​521.
Cuzick J. Human papillomavirus infection of the pros- Edwards SM, Evans DG, Hope Q, Norman AR,
tate. Cancer Surv 1995;23:91–​5. Barbachano Y, Bullock S, et  al.; UK Genetic
Cuzick J, Fisher G, Kattan MW, Berney D, Oliver T, Prostate Cancer Study Collaborators and
Foster CS, Moller H, Reuter V, Fearn P, Eastham BAUS Section of Oncology. Prostate cancer
J, Scardino P; Transatlantic Prostate Group. Long-​ in BRCA2 germline mutation carriers is asso-
term outcome among men with conservatively ciated with poorer prognosis. Br J Cancer
treated localised prostate cancer. Br J Cancer 2010;103(6):918–​24.
2006;95(9):1186–​94. Eeles RA, Durocher F, Edwards S, Teare D, Badzioch
Daniels NA, Nielson CM, Hoffman AR, Bauer DC; M, Hamoudi R, et al. Linkage analysis of chromo-
Osteoporotic Fractures In Men (MrOS) Study some lq markers in 136 prostate cancer families.
Group. Sex hormones and the risk of incident The Cancer Research Campaign/​British Prostate
prostate cancer. Urology 2010;76(5):1034–​40. group U.K. Familial Prostate Cancer Study
Dewailly E, Mulvad G, Sloth Pedersen H, Hansen Collaborators. Am J Hum Genet 1998;62:653–​8.
JC, Behrendt N, Hart Hansen JP. Inuit are pro- Eeles R, Goh C, Castro E, Bancroft E, Guy M, Al
tected against prostate cancer. Cancer Epidemiol Olama AA, Easton D, Kote-​Jarai Z. The genetic
Biomarkers Prev 2003 Sep;12(9):926–​7. epidemiology of prostate cancer and its clinical
Dickerman BA, Markt SC, Koskenvuo M, Pukkala implications. Nat Rev Urol 2014;11(1):18–​31.
E, Mucci LA, Kaprio J. Alcohol intake, drink- Ekbom A, Hsieh CC, Lipworth L, Wolk A, Pontén
ing patterns, and prostate cancer risk and mor- J, Adami HO, Trichopoulos D. Perinatal char-
tality:  a 30-​ year prospective cohort study of acteristics in relation to incidence of and mor-
Finnish twins. Cancer Causes Control. 2016 tality from prostate cancer. Br Med J 1996;
Sep;27(9):1049–​58. 313(7053):337–​41.
Dillner J, Knekt P, Boman J, Lehtinen M, Af Geijersstam Ekbom A, Wuu J, Adami HO, Lu CM, Lagiou P,
V, Sapp M, et al. Sero-​epidemiological association Trichopoulos D, Hsieh C. Duration of gestation
between human-​papillomavirus infection and risk and prostate cancer risk in offspring. Cancer
of prostate cancer. Int J Cancer 1998;75(4):564–​7. Epidemiol Biomarkers Prev 2000;9(2):221–​3.
Discacciati A, Orsini N, Wolk A. Coffee consumption Elgass S, Cooper A, Chopra M. Lycopene inhibits angi-
and risk of nonaggressive, aggressive and fatal ogenesis inhuman umbilical vein endothelial cells
prostate cancer—​a dose-​response meta-​analysis. and rat aortic rings. Br J Nutr 2012;108(3):431–​9.
Ann Oncol 2014;25(3):584–​91. Emerging Risk Factors Collaboration. Adult height
Discacciati A, Orsini N, Wolk A. Body mass index and the risk of cause-​ specific death and vas-
and incidence of localized and advanced prostate cular morbidity in 1  million people:  individ-
cancer—​ a dose-​ response meta-​analysis of pro- ual participant meta-​ analysis. Int J Epidemiol
spective studies. Ann Oncol 2012;23(7):1665–​71. 2012;41(5):1419–​33.
Discacciati A, Orsini N, Andersson SO, Andrén Endogenous Hormones and Prostate Cancer
O, Johansson JE, Wolk A. Body mass index in Collaborative Group, Roddam AW, Allen NE,
early and middle-​ late adulthood and risk of Appleby P, Key TJ. Endogenous sex hormones
localised, advanced and fatal prostate cancer:  a and prostate cancer:  a collaborative analy-
population-​based prospective study. Br J Cancer sis of 18 prospective studies. J Natl Cancer Inst
2011;105(7):1061–​8. 2008;100(3):170–​83.
Downer MK, Batista JL, Mucci LA, Stampfer MJ, Epstein MM, Kasperzyk JL, Mucci LA, Giovannucci E,
Epstein MM, Håkansson N, Wolk A, Johansson Price A, Wolk A, Håkansson N, Fall K, Andersson
JE, Andrén O, Fall K, Andersson SO. Dairy intake SO, Andrén O. Dietary fatty acid intake and
51

Prostate Cancer 511

prostate cancer survival in Örebro County, Fowke JH, McLerran DF, Gupta PC, He J, Shu XO,
Sweden. Am J Epidemiol 2012;176(3):240–​52. Ramadas K, et al. Associations of body mass index,
Eriksson M, Wedel H, Wallander MA, Krakau I, smoking, and alcohol consumption with prostate
Hugosson J, Carlsson S, Svärdsudd K. The impact cancer mortality in the Asia Cohort Consortium.
of birth weight on prostate cancer incidence and Am J Epidemiol 2015;182(5):381–​9.
mortality in a population-​ based study of men Freedman ML, Haiman CA, Patterson N, McDonald
born in 1913 and followed up from 50 to 85 years GJ, Tandon A, Waliszewska A, et  al. Admixture
of age. Prostate 2007;67(11):1247–​54. mapping identifies 8q24 as a prostate cancer risk
Etzioni R, Gulati R, Cooperberg MR, Penson locus in African-​American men. Proc Natl Acad
DM, Weiss NS, Thompson IM. Limitations Sci U S A 2006;103(38):14068–​73.
of basing screening policies on screening tri- Freedman ML, Pearce CL, Penney KL, Hirschhorn
als:  The US Preventive Services Task Force JN, Kolonel LN, Henderson BE, Altshuler D.
and Prostate Cancer Screening. Med Care Systematic evaluation of genetic variation at
2013;51(4):295–​300. the androgen receptor locus and risk of prostate
Etzioni R, Penson DF, Legler JM, di Tommaso D, cancer in a multiethnic cohort study. Am J Hum
Boer R, Gann PH, Feuer EJ. Overdiagnosis due Genet 2005;76(1):82–​90.
to prostate-​ specific antigen screening:  lessons Friedrichsen DM, Hawley S, Shu J, Humphrey M,
from U.S. prostate cancer incidence trends. J Natl Sabacan L, Iwasaki L, Etzioni R, Ostrander
Cancer Inst 2002;94(13):981–​90. EA, Stanford JL. IGF-​I and IGFBP-​3 polymor-
Ewings P, Bowie C. A case-​control study of cancer phisms and risk of prostate cancer. Prostate
of the prostate in Somerset and east Devon. Br J 2005;65(1):44–​51.
Cancer 1996;74(4):661–​6. Gandini S, Boniol M, Haukka J, Byrnes G, Cox B,
Fang F, Kasperzyk JL, Shui I, Hendrickson W, Hollis Sneyd MJ, Mullie P, Autier P. Meta-​analysis of
BW, Fall K, Ma J, Gaziano JM, Stampfer MJ, Mucci observational studies of serum 25-​hydroxyvitamin
LA, Giovannucci E. Prediagnostic plasma vitamin D levels and colorectal, breast and prostate
D metabolites and mortality among patients with cancer and colorectal adenoma. Int J Cancer
prostate cancer. PLoS One 2011;6(4):e18625. 2011;128(6):1414–​24.
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser Gandini S, Gnagnarella P, Serrano D, Pasquali E,
S, Mathers C, Rebelo M, Parkin DM, Forman D, Raimondi S. Vitamin D receptor polymorphisms
Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence and cancer. Adv Exp Med Biol 2014;810:69–​105.
and Mortality Worldwide: IARC CancerBase No. Gann PH, Hennekens CH, Sacks FM, Grodstein F,
11 [Internet]. Lyon, France: International Agency Giovannucci EL, Stampfer MJ. Prospective study
for Research on Cancer; 2013. Available from: of plasma fatty acids and risk of prostate cancer. J
http://globocan.iarc.fr, accessed 2017. Natl Cancer Inst 1994;86(4):281–​6.
Ferlay J, Bray F, Steliarova-​Foucher E and Forman D. Gann PH, Hennekens CH, Ma J, Longcope C,
Cancer Incidence in Five Continents, CI5plus. Stampfer MJ. Prospective study of sex hormone
IARC CancerBase No. 9.  Lyon:  International levels and risk of prostate cancer. J Natl Cancer
Agency for Research on Cancer; 2014. Available Inst 1996;88:1118–​26.
from: http://​ci5.iarc.fr, accessed on 1/​12/​2015. Gardner WA, Culberson DE, Bennett BD. Trichomonas
Feustel A, Wennrich R. Zinc and cadmium in cell vaginalis in the prostate gland. Archives of Pathology
fractions of prostate cancer tissues of different his- and Laboratory Medicine 1986;110(5):430–​2.
tological grading in comparison to BPH and nor- Gartner C, Stahl W, Sies H. Lycopene is more bioavail-
mal prostate. Urol Res 1984;12(2):147–​50. able from tomato paste than from fresh tomatoes.
Flanagan J, Gray PK, Hahn N, Hayes J, Myers LJ, Am J Clin Nutr 1997;66:116–​22.
Carney-​Doebbeling C, Sweeney CJ. Presence of Gaziano JM, Glynn RJ, Christen WG, Kurth T,
the metabolic syndrome is associated with shorter Belanger C, MacFadyen J, Bubes V, Manson JE,
time to castration-​resistant prostate cancer. Ann Sesso HD, Buring JE. Vitamins E and C in the pre-
Oncol 2011;22(4):801–​7. vention of prostate and total cancer in men:  the
Flickinger CJ, Bush LA, Williams MV, Naaby-​Hansen Physicians' Health Study II randomized controlled
S, Howards SS, Herr JC. Post-​ obstruction rat trial. JAMA 2009;301(1):52–​62.
sperm autoantigens identified by two-​dimensional Ghadirian P, Howe GR, Hislop TG, Maisonneuve
gel electrophoresis and western blotting. J Reprod P. Family history of prostate cancer:  a multi-​
Immunol 1999;43(1):35–​53. center case-​control study in Canada. Int J Cancer
Ford ES. Does exercise reduce inflammation? Physical 1997;70:679–​81.
activity and C-​reactive protein among U.S. adults. Gibbs M, Stanford JL, McIndoe RA, Jarvik GP, Kolb
Epidemiology 2002;13(5):561–​8. S, Goode EL, et  al. Evidence for a rare prostate
512

512 Textbook of Cancer Epidemiology

cancer–​susceptibility locus at chromosome lp36. prostate cancer. Cancer Epidemiol Biomarkers

Am J Hum Genet 1999;64:776–​87. Prev 1997a;6(8):557–​63.
Gilbert R, Martin RM, Beynon R, Harris R, Savovic J, Gong Z, Kristal AR, Schenk JM, Tangen CM,
Zuccolo L, Bekkering GE, Fraser WD, Sterne JA, Goodman PJ, Thompson IM. Alcohol consump-
Metcalfe C. Associations of circulating and dietary tion, finasteride, and prostate cancer risk: results
vitamin D with prostate cancer risk: a systematic from the Prostate Cancer Prevention Trial. Cancer
review and dose-​response meta-​analysis. Cancer 2009;115(16):3661–​9.
Causes Control 2011;22(3):319–​40. Grindedal EM, Møller P, Eeles R, Stormorken AT,
Giles GG, Severi G, English DR, McCredie MR, Bowitz-​ Lothe IM, Landrø SM, Clark N, Kvåle
MacInnis R, Boyle P, Hopper JL. Early growth, R, Shanley S, Maehle L. Germ-​ line mutations
adult body size and prostate cancer risk. Int J in mismatch repair genes associated with pros-
Cancer 2003;103(2):241–​5. tate cancer. Cancer Epidemiol Biomarkers Prev
Gill JK, Wilkens LR, Pollak MN, Stanczyk FZ, Kolonel 2009;18(9):2460–​7.
LN. Androgens, growth factors, and risk of pros- Grönberg H, Damber L, Damber JE. Familial prostate
tate cancer:  the Multiethnic Cohort. Prostate cancer in Sweden. A nationwide register cohort
2010;70(8):906–​15. study. Cancer 1996;77:138–​43.
Gilsing AM, Weijenberg MP, Goldbohm RA, Grönberg H, Damber L, Damber JE, Iselius L.
Dagnelie PC, van den Brandt PA, Schouten LJ. Segregation analysis of prostate cancer in
Vegetarianism, low meat consumption and the Sweden: support for dominant inheritance. Am J
risk of lung, postmenopausal breast and prostate Epidemiol 1997;146:552–​7.
cancer in a population-​based cohort study. Eur J Grotta A, Bottai M, Adami HO, Adams SA, Akre O,
Clin Nutr. 2016 Jun;70(6):723–​9. Blair SN, Mariosa D, Nyrén O, Ye W, Stattin P,
Giovannucci E. A review of epidemiologic studies of Bellocco R, Trolle Lagerros Y. Physical activity and
tomatoes, lycopene, and prostate cancer. Exp Biol body mass index as predictors of prostate cancer
Med (Maywood) 2002;227(10):852–​9. risk. World J Urol 2015;33(10):1495–​502.
Giovannucci E. Dietary influences of 1,25(OH)2 vita- Guess HA, Friedman GD, Sadler MC, Stanczyk
min D in relation to prostate cancer: a hypothesis. FZ, Vogelman JH, Imperato-​ McGinley J, Lobo
Cancer Causes Control 1998;9(6):567–​82. RA, Orentreich N. 5 Alpha-​ reductase activity
Giovannucci E, Stampfer MJ, Krithivas K, Brown M, and prostate cancer:  a case-​control study using
Brufsky A, Talcott J, et al. The CAG repeat within stored sera. Cancer Epidemiol Biomarkers Prev
the androgen receptor gene and its relation- 1997;6:21–​24.
ship to prostate cancer. Proc Natl Acad Sci USA Gunawardena K, Murray DK, Meikle AW. Vitamin
1997b;94:3320–​3. E and other antioxidants inhibit human pros-
Giovannucci E, Rimm EB, Ascherio A, Colditz G, tate cancer cells through apoptosis. Prostate
Spiegelman D, Stampfer MJ, et  al. Smoking and 2000;44(4):287–​95.
risk of total and fatal prostate cancer in United Gunnell D, Oliver SE, Donovan JL, Peters TJ, Gillatt
States health professionals. Cancer Epidemiol D, Persad R, Hamdy FC, Neal DE, Holly JM. Do
Biomarkers Prev 1999;8:277–​82. height-​related variations in insulin-​like growth
Giovannucci EL, Liu Y, Leitzmann MF, Stampfer MJ, factors underlie the associations of stature with
Willett WC. A prospective study of physical activ- adult chronic disease? J Clin Endocrinol Metab
ity and incident and fatal prostate cancer. Arch 2004;89(1):213–​8.
Intern Med 2005;165(9):1005–​10. Haid M, Sharon N. Immunofluorescent evidence of
Giovannucci E, Liu Y, Platz EA, Stampfer MJ, Willett prior herpes simplex virus type-​ 2 infection in
WC. Risk factors for prostate cancer incidence and prostate carcinoma. Urology 1984;24(6):623–​5.
progression in the health professionals follow-​up Haraldsdottir S, Hampel H, Wei L, Wu C, Frankel W,
study. Int J Cancer 2007;121(7):1571–​8. Bekaii-​Saab T, de la Chapelle A, Goldberg RM.
Giovannucci E, Ascherio A, Rimm EB, Stampfer MJ, Prostate cancer incidence in males with Lynch
Colditz GA, Willett WC. Intake of carotenoids syndrome. Genet Med 2014;16(7):553–​7.
and retinol in relation to risk of prostate cancer. J Hayes RB, Pottern LM, Strickler H, Rabkin C,
Natl Cancer Inst 1995;87(23):1767–​76. Pope V, Swanson GM, et  al. Sexual behaviour,
Giovannucci E, Rimm EB, Liu Y, Stampfer MJ, Willett STDs and risks for prostate cancer. Br J Cancer
WC. A prospective study of tomato products, 2000;82(3):718–​25.
lycopene, and prostate cancer risk. J Natl Cancer Hedelin M, Chang ET, Wiklund F, Bellocco R, Klint
Inst 2002;94(5):391–​8. A, Adolfsson J, Shahedi K, Xu J, Adami HO,
Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA, Gronberg H, Balter KA. Association of frequent
Willett WC. Height, body weight, and risk of consumption of fatty fish with prostate cancer
513

Prostate Cancer 513

risk is modified by COX-​2 polymorphism. Int J risk:  a population-​based case-​ control study in
Cancer 2007 Jan 15;120(2):398–​405. China. Cancer Res 2000b;60:5111–​6.
Hedelin M, Klint A, Chang ET, Bellocco R, Johansson Hsing AW, Deng J, Sesterhenn IA, Mostofi FK,
JE, Andersson SO, Heinonen SM, Adlercreutz Stanczyk FZ, Benichou J, Xie T, Gao YT. Body
H, Adami HO, Gronberg H, Balter KA. Dietary size and prostate cancer:  a population-​ based
phytoestrogen, serum enterolactone and risk case-​control study in China. Cancer Epidemiol
of prostate cancer:  the cancer prostate Sweden Biomarkers Prev 2000a;9(12):1335–​41.
study (Sweden). Cancer Causes Control 2006 Huang CS, Chuang CH, Lo TF, Hu ML. Anti-​
Mar;17(2):169–​80. angiogenic effects of lycopene through immu-
Heinonen OP, Albanes D, Virtamo J, Taylor PR, nomodualtion of cytokine secretion in human
Huttunen JK, Hartman AM, et al. Prostate cancer peripheral blood mononuclear cells. J Nutr
and supplementation with alpha-​tocopherol and Biochem 2013;24(2):428–​34.
beta-​carotene:  incidence and mortality in a con- Huncharek M, Haddock KS, Reid R, Kupelnick B.
trolled trial. J Natl Cancer Inst 1998;90(6):440–​6. Smoking as a risk factor for prostate cancer:  a
Helmrich SP, Ragland DR, Leung RW, Paffenbarger meta-​analysis of 24 prospective cohort studies.
RS Jr. Physical activity and reduced occurrence of Am J Public Health 2010;100(4):693–​701.
non-​insulin-​dependent diabetes mellitus. N Engl J Hunter DJ, Riboli E, Haiman CA, Albanes D, Altshuler
Med 1991;325(3):147–​52. D, Chanock SJ, et  al.; National Cancer Institute
Hemminki K, Rawal R, Bermejo JL. Prostate cancer Breast and Prostate Cancer Cohort Consortium.
screening, changing age-​specific incidence trends A candidate gene approach to searching for low-​
and implications on familial risk. Int J Cancer penetrance breast and prostate cancer genes. Nat
2005;113(2):312–​5. Rev Cancer 2005;5(12):977–​85.
Henderson BE, Bernstein L, Ross RK, Depue RH, Hurst R, Hooper L, Norat T, Lau R, Aune D,
Judd HL. The early in utero oestrogen and testos- Greenwood DC, Vieira R, Collings R, Harvey
terone environment of blacks and whites: poten- LJ, Sterne JA, Beynon R, Savović J, Fairweather-​
tial effects on male offspring. Br J Cancer Tait SJ. Selenium and prostate cancer:  system-
1988;57(2):216–​8. atic review and meta-​analysis. Am J Clin Nutr
Hendrickson WK, Flavin R, Kasperzyk JL, Fiorentino 2012;96(1):111–​22.
M, Fang F, Lis R, et al. Vitamin D receptor protein Ingles SA, Ross RK, Yu MC, Irvine RA, La Pera G,
expression in tumor tissue and prostate cancer Haile RW, et al. Association of prostate cancer risk
progression. J Clin Oncol 2011;29(17):2378–​85. with genetic polymorphisms in vitamin D recep-
Heshmat MY, Kovi J, Herson J, Jones GW, Jackson MA. tor and androgen receptor. J Natl Cancer Inst
Epidemiologic association between gonorrhea and 1997;89:166–​70.
prostatic carcinoma. Urology 1975;6(4):457–​60. Irani J, Goujon J-​M, Ragni E, et al. High-​grade inflam-
Hjelmborg JB, Scheike T, Holst K, Skytthe A, Penney mation in prostate cancer as a prognostic factor
KL, Graff RE, et al. The heritability of prostate can- for biochemical recurrence after radical prostatec-
cer in the Nordic Twin Study of Cancer. Cancer tomy. Urology 1999;54(3):467–​72.
Epidemiol Biomarkers Prev 2014;23(11):2303–​10. Islami F, Moreira DM, Boffetta P, Freedland SJ. A
Holt SK, Kolb S, Fu R, Horst R, Feng Z, Stanford JL. systematic review and meta-​analysis of tobacco
Circulating levels of 25-​hydroxyvitamin D and use and prostate cancer mortality and inci-
prostate cancer prognosis. Cancer Epidemiol dence in prospective cohort studies. Eur Urol
2013;37(5):666–​70. 2014;66(6):1054–​64.
Holt SK, Kwon EM, Koopmeiners JS, Lin DW, Feng Z, Jahn JL, Giovannucci EL, Stampfer MJ. The high
Ostrander EA, Peters U, Stanford JL. Vitamin D prevalence of undiagnosed prostate cancer at
pathway gene variants and prostate cancer prog- autopsy: implications for epidemiology and treat-
nosis. Prostate 2010;70(13):1448–​60. ment of prostate cancer in the Prostate-​specific
Howard SS. Possible biological mechanisms for a rela- Antigen-​era. Int J Cancer 2015;137(12):2795–​802.
tionship between vasectomy and prostatic cancer. Jemal A, Fedewa SA, Ma J, Siegel R, Lin CC, Brawley
Eur J Cancer 1993;29A(7):1060–​2. O, Ward EM. Prostate Cancer Incidence and PSA
Hsieh CC, Thanos A, Mitropoulos D, Deliveliotis C, Testing Patterns in Relation to USPSTF Screening
Mantzoros CS, Trichopoulos D. Risk factors for Recommendations. JAMA 2015;314(19):2054–​61.
prostate cancer:  a case-​control study in Greece. Jensen OM, Paine SL, McMichael AJ, Ewertz M.
Int J Cancer 1999;80(5):699–​703. Alcohol. In:  Schottenfeld D, Fraumeni JF (Eds.).
Hsing AW, Gao YT, Wu G, Wang X, Deng J, Chen YL, Cancer Epidemiology and Prevention, 2nd ed.
et al. Polymorphic CAG and CGN repeat lengths New  York, Oxford University Press, 1996, pp
in the androgen receptor gene and prostate cancer 290–​318.
514

514 Textbook of Cancer Epidemiology

Jiang Q. Natural forms of vitamin E:  metabolism, Kenfield SA, Batista JL, Jahn JL, Downer MK, Van
antioxidant, and anti-​inflammatory activities and Blarigan EL, Sesso HD, Giovannucci EL, Stampfer
their role in disease prevention and therapy. Free MJ, Chan JM. Development and Application of a
Radic Biol Med 2014;72:76–​90. Lifestyle Score for Prevention of Lethal Prostate
Johansson JE, Andren O, Andersson SO, Dickman Cancer. J Natl Cancer Inst 2015;108(3).
PW, Holmberg L, Magnuson A, Adami HO. Kenfield SA, Stampfer MJ, Chan JM, Giovannucci E.
Natural history of early, localized prostate cancer. Smoking and prostate cancer survival and recur-
JAMA 2004;291(22):2713–​9. rence. JAMA 2011a;305(24):2548–​55.
John EM, Whittemore AS, Wu AH, Kolonel LN, Kenfield SA, Stampfer MJ, Giovannucci E, Chan JM.
Hislop TG, Howe GR, et  al. Vasectomy and Physical activity and survival after prostate can-
prostate cancer:  results from a multieth- cer diagnosis in the health professionals follow-​up
nic case-​ control study. J Natl Cancer Inst study. J Clin Oncol 2011b;29(6):726–​32.
1995;87(9):662–​9. Kenfield SA, Van Blarigan EL, DuPre N, Stampfer MJ,
John EM, Schwartz GG, Koo J, Van Den Berg D, and L Giovannucci E, Chan JM. Selenium supplemen-
Ingles SA. Sun exposure, vitamin D receptor gene tation and prostate cancer mortality. J Natl Cancer
polymorphisms, and risk of advanced prostate Inst 2014;107(1):360.
cancer. Cancer Res 2005;65:5470–​9. Keum N, Greenwood DC, Lee DH, Kim R, Aune D, Ju
Johnsen NF, Tjønneland A, Thomsen BL, Christensen W, Hu FB, Giovannucci EL. Adult weight gain and
J, Loft S, Friedenreich C, et al. Physical activity and adiposity-​related cancers:  a dose-​response meta-​
risk of prostate cancer in the European Prospective analysis of prospective observational studies. J
Investigation into Cancer and Nutrition (EPIC) Natl Cancer Inst 2015;107(2).
cohort. Int J Cancer 2009;125(4):902–​8. Key TJA, Silcocks PB, Davey GK, Appleby PN, Bishop
Jones JI, Clemmons DR. Insulin-​like growth factors DT. A case-​control study of diet and prostate can-
and their binding proteins:  Biological actions. cer. Br J Cancer 1997;76:678–​87.
Endocrinol Rev 1995;16:3–​34. Key TJ, Appleby PN, Travis RC, Albanes D, Alberg
Joshu CE, Mondul AM, Meinhold CL, Humphreys AJ, Barricarte A, et  al.; Endogenous Hormones
EB, Han M, Walsh PC, Platz EA. Cigarette smok- Nutritional BiomarkersProstate Cancer Collabo­
ing and prostate cancer recurrence after prostatec- rative Group. Carotenoids, retinol, tocopherols, and
tomy. J Natl Cancer Inst 2011a;103(10):835–​8. prostate cancer risk:  pooled analysis of 15 studies.
Joshu CE, Mondul AM, Menke A, Meinhold C, Han Am J Clin Nutr 2015;102(5):1142–​57.
M, Humphreys EB, Freedland SJ, Walsh PC, Platz Kirsh VA, Hayes RB, Mayne ST, Chatterjee N, Subar
EA. Weight gain is associated with an increased AF, Dixon LB, Albanes D, Andriole GL, Urban
risk of prostate cancer recurrence after prosta- DA, Peters U; PLCO Trial. Supplemental and
tectomy in the PSA era. Cancer Prev Res (Phila) dietary vitamin E, beta-​carotene, and vitamin C
2011b;4(4):544–​51. intakes and prostate cancer risk. J Natl Cancer Inst
Kantoff P, Giovannucci E, Brown M. The androgen 2006;98(4):245–​54.
receptor CAG repeat polymorphism and its rela- Klein EA, Thompson IM Jr, Tangen CM, Crowley JJ,
tionship to prostate cancer. Biochem Biophys Acta Lucia MS, Goodman PJ, et al. Vitamin E and the
1998;1378:C1–​5. risk of prostate cancer: the Selenium and Vitamin
Keenan BS, Richards GE, Ponder SW, Dallas FS, E Cancer Prevention Trial (SELECT). JAMA
Nagamani M, Smith ER. Androgen-​ stimulated 2011;306(14):1549–​56.
pubertal growth:  the effects of testosterone and Kokkinos PF, Fernhall B. Physical activity and high
dihydrotestosterone on growth hormone and density lipoprotein cholesterol levels: what is the
insulin-​like growth factor-​1 in the treatment of relationship? Sports Med 1999;28(5):307–​14.
short stature and delayed puberty. J Clin Endocrinol Kokkinos P, Myers J. Exercise and physical activ-
Metab 1993;76(4):996–​1001. ity: clinical outcomes and applications. Circulation
Keetch DW, Rice JP, Suarez BK, Catalona WJ. Familial 2010;122(16):1637–​48.
aspects of prostate cancer: a case control study. J Kolenko V, Teper E, Kutikov A, Uzzo R. Zinc and zinc
Urol 1995;154:2100–​2. transporters in prostate carcinogenesis. Nat Rev
Kempf K, Herder C, Erlund I, Kolb H, Martin S, Urol 2013;10(4):219–​26.
Carstensen M, Koenig W, Sundvall J, Bidel S, Kuha Kolonel L and Winkelstein W Jr. Cadmium and pros-
S, Tuomilehto J. Effects of coffee consumption on tatic carcinoma. Lancet 1977;2(8037):566–​7.
subclinical inflammation and other risk factors Koralek DO, Peters U, Andriole G, Reding D, Kirsh V,
for type 2 diabetes: a clinical trial. Am J Clin Nutr Subar A, Schatzkin A, Hayes R, Leitzmann MF. A
2010;91(4):950–​957. prospective study of dietary alpha-​linolenic acid
51

Prostate Cancer 515

and the risk of prostate cancer (United States). Lawson JS. Prostate cancer, birthweight, and diet.
Cancer Causes Control 2006;17(6):783–​91. Epidemiology 1998;9(2):217.
Kote-​Jarai Z, Leongamornlert D, Saunders E, Lee IM. Exercise and physical health:  cancer and
Tymrakiewicz M, Castro E, Mahmud N, et  al.; immune function. Res Q Exerc Sport 1995;
UKGPCS Collaborators, Goldgar D, Eeles R. 66(4):286–​91.
BRCA2 is a moderate penetrance gene contribut- Lees RE, Steele R, Wardle D. Arsenic, syphilis, and
ing to young-​onset prostate cancer:  implications cancer of the prostate. J Epidemiol Commun
for genetic testing in prostate cancer patients. Br Health 1985;39(3):227–​30.
J Cancer 2011;105(8):1230–​4. Lemen RA, Lee JS, Wagoner JK, Blejer HP. Cancer
Koutros S, Cross AJ, Sandler DP, Hoppin JA, Ma X, mortality among cadmium production workers.
Zheng T, Alavanja MC, Sinha R. Meat and meat Ann Ny Acad Sci 1976;271:273–​9.
mutagens and risk of prostate cancer in the Leongamornlert D, Mahmud N, Tymrakiewicz M,
Agricultural Health Study. Cancer Epidemiol Saunders E, Dadaev T, Castro E, et al.; UKGPCS
Biomarkers Prev 2008;17(1):80–​7. Collaborators, Goldgar D, Eeles R, Kote-​Jarai Z.
Kramer BS, Hagerty KL, Justman S, Somerfield Germline BRCA1 mutations increase prostate
MR, Albertsen PC, Blot WJ, et  al.; American cancer risk. Br J Cancer 2012;106(10):1697–​701.
Society of Clinical Oncology Health Services Leongamornlert D, Saunders E, Dadaev T,
Committee; American Urological Association Tymrakiewicz M, Goh C, Jugurnauth-​ Little S,
Practice Guidelines Committee. Use of 5-​alpha-​ et  al.; UKGPCS Collaborators, Antoniou AC,
reductase inhibitors for prostate cancer che- Eeles R, Kote-​Jarai Z. Frequent germline delete-
moprevention:  American Society of Clinical rious mutations in DNA repair genes in familial
Oncology/​ American Urological Association prostate cancer cases are associated with advanced
2008 Clinical Practice Guideline. J Clin Oncol disease. Br J Cancer 2014;110(6):1663–​72.
2009;27(9):1502–​16. Lesko SM, Rosenberg L, Shapiro S. Family his-
Krieger JN, Verdon M, Siegel N, Holmes KK. Natural tory and prostate cancer risk. Am J Epidemiol
history of urogenital trichomoniasis in men. 1996;144:1041–​7.
Journal of Urology 1993;149(6):1455–​58. Lewitt MS, Saunders H, Cooney GJ, Baxter RC.
Kristal AR, Darke AK, Morris JS, Tangen CM, Effect of human insulin-​ like growth factor
Goodman PJ, Thompson IM, et al. Baseline sele- binding protein-​1 on the half life and action of
nium status and effects of selenium and vitamin administered insulin-​like growth factor in rats. J
e supplementation on prostate cancer risk. J Natl Endocrinol 1993;136:253–​60.
Cancer Inst 2014;106(3):djt456. Li L, Cicek MS, Casey G, Witte JS. No association
Kucknoor A, Mundodi V, Alderete JF. Trichomonas between genetic polymorphisms in IGF-​ I and
vaginalis adherence mediates differential gene IGFBP-​3 and prostate cancer. Cancer Epidemiol
expression in human vaginal epithelial cells. Biomarkers Prev 2004;13(3):497–​8.
Cellular Microbiology 2005;7(6):887–​97. Lilienfeld AM, Levin ML, Kessler II. Cancer in the
Kuper H, Adami HO, Trichopoulos D. Infections as a United States. APHA Monograph. Cambridge,
major preventable cause of human cancer. Journal Harvard University Press, 1972.
of Internal Medicine 2000;248(3):171–​83. Lippman SM, Klein EA, Goodman PJ, Lucia MS,
Kurahashi N, Iwasaki M, Sasazuki S, Otani T, Inoue Thompson IM, Ford LG, et  al. Effect of sele-
M, Tsugane S; Japan Public Health Center-​Based nium and vitamin E on risk of prostate cancer
Prospective Study Group. Soy product and isofla- and other cancers:  the Selenium and Vitamin
vone consumption in relation to prostate cancer in E Cancer Prevention Trial (SELECT). JAMA
Japanese men. Cancer Epidemiol Biomarkers Prev 2009;301(1):39–​51.
2007;16(3):538–​45. Lippman SM, Goodman PJ, Klein EA, Parnes
Kyle E, Neckers L, Takimoto C, Curt G, Bergan R. HL, Thompson IM Jr, Kristal AR, et  al.
Genistein-​ induced apoptosis of prostate can- Designing the Selenium and Vitamin E Cancer
cer cells is preceded by specific decrease in focal Prevention Trial (SELECT). J Natl Cancer Inst
adhesion kinase activity. Molecular Pharmacology 2005;97(2):94–​102.
1997;51:193–​200. Lipworth L, Hsieh CC, Wide L, Ekbom A, Yu SZ, Yu
Lahmann PH, Wallström P2, Lissner L3, Olsson H4, GP, Xu B, Hellerstein S, Carlstrom K, Trichopoulos
Gullberg B2. Measures of birth size in relation D, Adami HO. Maternal pregnancy hormone lev-
to risk of prostate cancer:  the Malmö Diet and els in areas with high incidence (Boston, USA)
Cancer Study, Sweden. J Dev Orig Health Dis and low incidence (Shanghai, China) of breast
2012;3(6):442–​9. cancer. Br J Cancer 1999;79(1):7–​12.
516

516 Textbook of Cancer Epidemiology

Littman AJ, White E, Kristal AR. Anthropometrics Mishina T, Watanabe H, Araki H, Nakao M.
and prostate cancer risk. Am J Epidemiol Epidemiological study of prostate cancer by
2007;165(11):1271–​9. matched-​pair analysis. Prostate 1985;6:423–​36.
Liu LH, Kang R, He J, Zhao SK, Li FT, Wan SP, Zhao Mishra DK, Bid HK, Srivastava DS, Mandhani A,
ZG. Vasectomy and risk of prostate cancer: a sys- and Mittal RD. Association of vitamin D receptor
tematic review and meta-​analysis of cohort stud- gene polymorphism and risk of prostate cancer in
ies. Andrology 2015;3(4):643–​9. India. Urol Int 2005;74:315–​8.
Lu Y, Zhai L, Zeng J, Peng Q, Wang J, Deng Y, Xie L, Mo Möller E, Wilson KM, Batista JL, Mucci LA, Bälter K,
C, Yang S, Li S, Qin X. Coffee consumption and Giovannucci E. Body size across the life course
prostate cancer risk:  an updated meta-​analysis. and prostate cancer in the Health Professionals
Cancer Causes Control 2014;25(5):591–​604. Follow-​up Study. Int J Cancer 2015 [Epub ahead
Lucia MS, Epstein JI, Goodman PJ, Darke AK, of print].
Reuter VE, Civantos F, et  al. Finasteride and Moore SC, Peters TM, Ahn J, Park Y, Schatzkin
high-​grade prostate cancer in the Prostate A, Albanes D, Ballard-​Barbash R, Hollenbeck
Cancer Prevention Trial. J Natl Cancer Inst A, Leitzmann MF. Physical activity in rela-
2007;99(18):1375–​83. tion to total, advanced, and fatal prostate
Lüleci G, Sakizli M, Günalp A, Erkan I, Remzi cancer. Cancer Epidemiol Biomarkers Prev
D. Herpes simplex type 2 neutralization anti- 2008;17(9):2458–​66.
bodies in patients with cancers of the urinary Moreira DM, Antonelli JA, Presti JC Jr, Aronson
bladder, prostate, and cervix. J Surg Oncol WJ, Terris MK, Kane CJ, Amling CL, Freedland
1981;16(4):327–​31. SJ. Association of cigarette smoking with inter-
Ma J, Giovannucci E, Pollak M, Chan JM, Gaziano val to biochemical recurrence after radical pros-
JM, Willett WC, Stampfer MJ. Milk intake, circu- tatectomy:  results from the SEARCH database.
lating levels of insulin-​like growth factor-​I, and Urology 2010;76(5):1218–​23.
risk of colorectal cancer in men. J Natl Cancer Inst Moyer VA; U.S. Preventive Services Task Force.
2001;93(17):1330–​6. Screening for prostate cancer:  U.S. Preventive
Major JM, Cross AJ, Watters JL, Hollenbeck AR, Services Task Force recommendation statement.
Graubard BI, Sinha R. Patterns of meat intake and Ann Intern Med 2012;157(2):120–​34.
risk of prostate cancer among African-​Americans Moyret-​Lalle C, Marçais C, Jacquemier J, Moles JP,
in a large prospective study. Cancer Causes Daver A, Soret JY, Jeanteur P, Ozturk M, Theillet C.
Control 2011;22(12):1691–​8. ras, p53 and HPV status in benign and malignant
Manson JE, Nathan DM, Krolewski AS, Stampfer MJ, prostate tumors. Int J Cancer 1995;64(2):124–​9.
Willett WC, Hennekens CH. A prospective study Mucci LA, Hjelmborg JB, Harris JR, Czene K, Havelick
of exercise and incidence of diabetes among US DJ, Scheike T, et al. Familiar risk and heritability of
male physicians. JAMA 1992;268(1):63–​7. cancer among twins in Nordic countries. JAMA
Mantzoros CS, Georgiadis EI. Body mass and phys- 2015; 13(1):68–​76.
ical activity are important predictors of serum Mucci LA, Powolny A, Giovannucci E, Liao Z,
androgen concentrations in young healthy men. Kenfield SA, Shen R, Stampfer MJ, Clinton SK.
Epidemiology 1995;6:433–​5. Prospective study of prostate tumor angiogen-
McIndoe RA, Stanford JL, Gibbs M, Jarvik GP, esis and cancer-​specific mortality in the health
Brandzel S, Neal CL, et al. Linkage analysis of 49 professionals follow-​ up study. J Clin Oncol
high-​risk families does not support a common 2009;27(33):5627–​33.
familial prostate cancer–​ susceptibility gene at Muir C, Waterhouse J, Mack T, et  al (eds.) Cancer
lq24-​25. Am J Hum Genet 1997;61:347–​53. Incidence in Five Continents, Vol V. International
McLellan DL and Norman RW. Hereditary Agency for Research on Cancer, Lyon, IARC
aspects of prostate cancer. Can Med Assoc J Press, 1987.
1995;153:895–​900. Muller RL, Gerber L, Moreira DM, Andriole G,
Menter DG, Sabichi AL, Lippman SM. Selenium Castro-​Santamaria R, Freedland SJ. Serum testos-
effects on prostate cell growth. Cancer Epidemiol terone and dihydrotestosterone and prostate can-
Biomarkers Prev 2000;9(11):1171–​82. cer risk in the placebo arm of the Reduction by
Messina MJ, Persky V, Setchell KDR, Barnes S. Soy Dutasteride of Prostate Cancer Events trial. Eur
intake and cancer risk: a review of the in vitro and Urol 2012;62(5):757–​64.
in vivo data. Nutr Cancer 1994;21:113–​31. Nam RK, Elhaji Y, Krahn MD, Hakimi J, Ho M, Chu
Meydani M. Vitamin E. Lnacet 1995;345(8943):170–​5. W, et al. Significance of the CAG repeat polymor-
Meydani SN, Hayek MG. Vitamin E and aging immune phism of the androgen receptor gene in prostate
response. Clin Geriatr Med 1995;11(4):567–​76. cancer progression. J Urol 2000;164:567–​72.
517

Prostate Cancer 517

Nam RK, Zhang WW, Trachtenberg J, Jewett MA, of prostate cancer:  a meta-​ analysis. Cancer
Emami M, Vesprini D, et al. Comprehensive assess- Epidemiol Biomarkers Prev 2003;12:1395–​402.
ment of candidate genes and serological mark- Oefelein MG, Resnick MI. Association of tobacco use
ers for the detection of prostate cancer. Cancer with hormone refractory disease and survival of
Epidemiol Biomarkers Prev 2003;12(12):1429–​37. patients with prostate cancer. J Urol 2004;171(6 Pt
Nash SH, Schenk JM, Kristal AR, Goodman PJ, Lucia 1):2281–​4.
MS, Parnes HL, et al. Association between Serum Packer BR, Yeager M, Burdett L, Welch R, Beerman
Phospholipid Fatty Acids and Intraprostatic M, Qi L, et  al. SNP500Cancer:  a public resource
Inflammation in the Placebo Arm of the Prostate for sequence validation, assay development, and
Cancer Prevention Trial. Cancer Prev Res (Phila) frequency analysis for genetic variation in candi-
2015;8(7):590–​6. date genes. Nucleic Acids Res 2006;34(Database
Nelles JL, Hu WY, Prins GS. Estrogen action and issue):D617–​621.
prostate cancer. Expert Rev Endocrinol Metab Palapattu GS, Sutcliffe S, De Marzo AM, Isaacs WB,
2011;6(3):437–​51. Nelson WG. Prostate carcinogenesis and inflam-
Nelson WG, De Marzo AM, DeWeese TL, Isaacs mation:  emerging insights. Carcinogenesis
WB. The role of inflammation in the pathogen- 2004;26(7):1170–​81.
esis of prostate cancer. The Journal of Urology Pantarotto J, Malone S, Dahrouge S, Gallant V, Eapen
2004;172:S6–​S12. L. Smoking is associated with worse outcomes in
Neuhausen SL, Farnham JM, Kort E, Tavtigian SV, patients with prostate cancer treated by radical
Skolnick MH, Cannon-​Albright LA. Prostate can- radiotherapy. BJU Int 2007;99(3):564–​9.
cer susceptibility locus HPC1 in Utah high-​risk Park SY, Murphy SP, Wilkens LR, Henderson BE,
pedigrees. Hum Mol Genet 1999;8:2437–​42. Kolonel LN. Fat and meat intake and prostate can-
Neuhausen SL, Slattery ML, Garner CP, Ding YC, cer risk: the multiethnic cohort study. Int J Cancer
Hoffman M, Brothman AR. Prostate cancer 2007;121(6):1339–​45.
risk and IRS1, IRS2, IGF1, and INS polymor- Park SY, Murphy SP, Wilkens LR, Henderson BE,
phisms: strong association of IRS1 G972R variant Kolonel LN; Multiethnic Cohort Study. Legume
and cancer risk. Prostate 2005;64(2):168–​74. and isoflavone intake and prostate cancer
Neve J. Human selenium supplementation as assessed risk: The Multiethnic Cohort Study. Int J Cancer
by changes in blood selenium concentration and 2008;123(4):927–​32.
glutathione peroxidase activity. J Trace Elem Med Pasquali R, Casimirri F, Cantobelli S, Melchionda
Biol 1995;9(2):65–​73. N, Morselli Labate AM, Fabbri R, Capelli M,
Nilsen TI, Romundstad PR, Troisi R, Vatten LJ. Birth Bortoluzzi L. Effect of obesity and body fat dis-
size and subsequent risk for prostate cancer: a pro- tribution on sex hormones and insulin in men.
spective population-​based study in Norway. Int J Metabolism 1991;40:101–​4.
Cancer 2005;113(6):1002–​4. Patel AV, Rodriguez C, Jacobs EJ, Solomon L, Thun
Noda T, Sasagawa T, Dong Y, Fuse H, Namiki M, MJ, Calle EE. Recreational physical activity
Inoue M. Detection of human papillomavirus and risk of prostate cancer in a large cohort of
(HPV) DNA in archival specimens of benign U.S. men. Cancer Epidemiol Biomarkers Prev
prostatic hyperplasia and prostatic cancer using 2005;14(1):275–​9.
a highly sensitive nested PCR method. Urol Res Penney KL, Stampfer MJ, Jahn JL, Sinnott JA, Flavin
1998;26(3):165–​9. R, Rider JR, Finn S, Giovannucci E, Sesso HD,
Nomura AMY, Stemmermann GN, Chyou PH, Loda M, Mucci LA, Fiorentino M. Gleason
Henderson BE, Stanczyk FZ. Serum androgens grade progression is uncommon. Cancer Res
and prostate cancer. Cancer Epidemiol Biomarkers 2013;73(16):5163–​8.
Prev 1996;5:621–​5. Peters U, Littman AJ, Kristal AR, Patterson RE, Potter
Norrish AE, Skeaff CM, Arribas GL, Sharpe SJ, Jackson JD, White E. Vitamin E and selenium supplemen-
RT. Prostate cancer risk and consumption of fish tation and risk of prostate cancer in the Vitamins
oils: a dietary biomarker-​based case-​control study. and lifestyle (VITAL) study cohort. Cancer Causes
Br J Cancer 1999;81(7):1238–​42. Control 2008;19(1):75–​87.
Nutting PA, Freeman WL, Risser DR, Helgerson Petrin D, Delgaty K, Bhatt R, Garber G. Clinical and
SD, Paisano R, Hisnanick J, et  al. Cancer inci- microbiological aspects of Trichomonas vaginalis.
dence among American Indians and Alaska Clinical Microbiology Reviews 1998;11(2):300–​17.
Natives, 1980 through 1987. Am J Public Health Pettersson A, Kasperzyk JL, Kenfield SA, Richman
1993;83:1589–​98. EL, Chan JM, Willett WC, Stampfer MJ, Mucci
Ntais C, Polycarpou A, and Ioannidis JP. Vitamin LA, Giovannucci EL. Milk and dairy consump-
D receptor gene polymorphisms and risk tion among men with prostate cancer and risk
518

518 Textbook of Cancer Epidemiology

of metastases and prostate cancer death. Cancer Richman EL, Kenfield SA, Chavarro JE, Stampfer
Epidemiol Biomarkers Prev 2012;21(3):428–​36. MJ, Giovannucci EL, Willett WC, Chan JM. Fat
Pickles T, Liu M, Berthelet E, Kim-​Sing C, Kwan W, intake after diagnosis and risk of lethal prostate
Tyldesley S; PROSTATE COHORT OUTCOMES cancer and all-​cause mortality. JAMA Intern Med
INITIATIVE. The effect of smoking on outcome 2013;173(14):1318–​26.
following external radiation for localized prostate Richman EL, Kenfield SA, Stampfer MJ, Giovannucci
cancer. J Urol 2004;171(4):1543–​6. EL, Chan JM. Egg, red meat, and poultry intake
Platz EA, Helzlsouer KJ, Hoffman SC, Morris JS, and risk of lethal prostate cancer in the prostate-​
Baskett CK, Comstock GW. Prediagnostic toenail specific antigen-​ era:  incidence and survival.
cadmium and zinc and subsequent prostate can- Cancer Prev Res (Phila) 2011a;4(12):2110–​21.
cer risk. Prostate 2002;52(4):288–​96. Richman EL, Kenfield SA, Stampfer MJ, Paciorek A,
Platz EA, Giovannucci E, Rimm EB, Curhan Carroll PR, Chan JM. Physical activity after diagno-
GC, Spiegelman D, Colditz GA, Willett WC. sis and risk of prostate cancer progression: data from
Retrospective analysis of birth weight and pros- the cancer of the prostate strategic urologic research
tate cancer in the Health Professionals Follow-​up endeavor. Cancer Res 2011b;71(11):3889–​95.
Study. Am J Epidemiol 1998;147(12):1140–​4. Rieken M, Shariat SF, Kluth LA, Fajkovic H,
Plitt SS, Garfein RS, Gaydos CA, Strathdee SA, Rink M, Karakiewicz PI, et  al. Association of
Sherman SG, Taha TE. Prevalence and correlates Cigarette Smoking and Smoking Cessation with
of chlamydia trachomatis, neisseria gonorrhoeae, Biochemical Recurrence of Prostate Cancer in
trichomonas vaginalis infections, and bacte- Patients Treated with Radical Prostatectomy. Eur
rial vaginosis among a cohort of young injec- Urol 2015;68(6):949–​56.
tion drug users in Baltimore, Maryland. Sexually Ripoll EA, Rama BN, Webber MM. Vitamin E
Transmitted Diseases 2005;32(7):446–​53. enhances the chemotherapeutic effects of adriam-
Pulido R, Hernandez-​ Garcia M, Saura-​ Calixto F. ycin on human prostatic carcinoma cells in vitro. J
Contribution of beverages to the intake of lipo- Urol 1986;136(2):529–​31.
philic and hydrophilic antioxidants in the Spanish Robinson WR, Stevens J, Gammon MD, John
diet. Eur J Clin Nutr. 2003;57(10):1275–​1282. EM. Obesity before age 30  years and risk of
Quinn M and Babb P. Patterns and trends in pros- advanced prostate cancer. Am J Epidemiol
tate cancer incidence, survival, prevalence and 2005;161(12):1107–​14.
mortality: Part I: international comparisons. BJU Robinson WR, Poole C, Godley PA. Systematic review
International 2002;90:162–​173. of prostate cancer’s association with body size in
Rayman MP. The importance of selenium to human childhood and young adulthood. Cancer Causes
health. Lancet 2000;356(9225):233–​41. Control 2008;19(8):793–​803.
Redman C, Scott JA, Baines AT, Basye JL, Clark Roddam AW, Allen NE, Appleby P, Key TJ, Ferrucci L,
LC, Calley C, Roe D, Payne CM, Nelson MA. Carter HB, et al. Insulin-​like growth factors, their
Inhibitory effect of selenomethionine on the binding proteins, and prostate cancer risk: analysis
growth of three selected human tumor cell lines. of individual patient data from 12 prospective stud-
Cancer Lett 1998;125(1–​2):103–​10. ies. Ann Intern Med 2008;149(7):461–​71, W83–​8.
Renehan AG, Zwahlen M, Minder C, O'Dwyer ST, Rodriguez C, Calle EE, Miracle McMahill HL, Tatham
Shalet SM, Egger M. Insulin-​ like growth fac- LM, Wingo PA, Thun MJ, et  al. Family history
tor (IGF)-​I, IGF binding protein-​3, and cancer and risk of fatal prostate cancer. Epidemiology
risk: systematic review and meta-​regression anal- 1997;8:653–​7.
ysis. Lancet. 2004;363(9418):1346–​53. Review. Rodriguez C, Freedland SJ, Deka A, Jacobs EJ,
Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen M. McCullough ML, Patel AV, Thun MJ, Calle EE.
Body-​mass index and incidence of cancer: a system- Body mass index, weight change, and risk of pros-
atic review and meta-​analysis of prospective obser- tate cancer in the Cancer Prevention Study II
vational studies. Lancet 2008;371(9612):569–​78. Nutrition Cohort. Cancer Epidemiol Biomarkers
Rennert H, Zeigler-​ Johnson CM, Addya K, et  al. Prev 2007;16(1):63–​9.
Association of susceptibility alleles in ELAC2/​ Rohrmann S, Nimptsch K, Sinha R, Willett WC,
HPC2, RNASEL/​ HPC1, and MSR1 with pros- Giovannucci EL, Platz EA, Wu K. Intake of Meat
tate cancer severity in European American Mutagens and Risk of Prostate Cancer in a Cohort
and African American men. Cancer Epidemiol of U.S. Health Professionals. Cancer Epidemiol
Biomarkers Prev 2005;14(4):949–​57. Biomarkers Prev 2015;24(10):1557–​63.
Richards IS, Davis FE, Lubell I. Current status of Rohrmann S, Linseisen J, Key TJ, Jensen MK, Overvad
endocrinologic effects of vasectomy. Urology K, Johnsen NF, et  al. Alcohol consumption and
1981;18(1):1–​6. the risk for prostate cancer in the European
519

Prostate Cancer 519

Prospective Investigation into Cancer and IGF1 (CA)19 repeat and IGFBP3 -​202 A/​C geno-
Nutrition. Cancer Epidemiol Biomarkers Prev types and the risk of prostate cancer in Black and
2008;17(5):1282–​7. White men. Cancer Epidemiol Biomarkers Prev
Rose DP, Boyer AP, Wynder EL. International com- 2005;14(2):403–​8.
parisons of mortality rates for cancer of the breast, Schneider A, Kalikin LM, Mattos AC, Keller ET, Allen
ovary, prostate, and colon, and per capita food MJ, Pienta KJ, McCauley LK. Bone turnover medi-
consumption. Cancer 1986;58:2363–​71. ates preferential localization of prostate cancer in
Ross RK, Bernstein L, Lobo RA, Shimizu H, Stanczyk the skeleton. Endocrinology 2005;146(4):1727–​36.
FZ, Pike MC, Henderson BE. 5-​Alpha-​reductase Schröder FH, Hugosson J, Roobol MJ, Tammela TL,
activity and risk of prostate cancer among Ciatto S, Nelen V, et  al.; ERSPC Investigators.
Japanese and U.S. white and black males. Lancet Screening and prostate-​ cancer mortality in a
1992;339(8798):887–​9. randomized European study. N Engl J Med
Ross RK, Paganini-​Hill A, Henderson BE. The etiol- 2009;360(13):1320–​8.
ogy of prostate cancer: what does the epidemiol- Schröder FH, Hugosson J, Roobol MJ, Tammela TL,
ogy suggest? Prostate 1983;4: 333–​44. Ciatto S, Nelen V, et  al.; ERSPC Investigators.
Ross RK, McCurtis JW, Henderson BE, Menck HR, Prostate-​cancer mortality at 11 years of follow-​up.
Mack TM, Martin SP. Descriptive epidemiology N Engl J Med 2012;366(11):981–​90.
of testicular and prostatic cancer in Los Angeles. Schumacher F, Amin Al Olama A, Berndt SI, Benlloch
Br J Cancer 1979;39(3):284–​92. S, Ahmed M, Sheng X, et al., behalf of ELLIPSE,
Rowlands MA, Gunnell D, Harris R, Vatten LJ, Holly PRACTICAL, CaPS, BPC3, PEGASUS. Prostate
JM, Martin RM. Circulating insulin-​like growth cancer meta-​ analysis from more than 143,000
factor peptides and prostate cancer risk:  a sys- men identifies 57 novel prostate cancer suscepti-
tematic review and meta-​analysis. Int J Cancer bility loci. American Society of Human Genetics
2009;124(10):2416–​29. 2016 Annual Meeting, October 18–​ 22, 2016,
Sander A, Linseisen J, Rohrmann S. Intake of heter- Vancouver.
ocyclic aromatic amines and the risk of prostate Schumacher FR, Feigelson HS, Cox DG, Haiman CA,
cancer in the EPIC-​ Heidelberg cohort. Cancer Albanes D, Buring J, et al. A common 8q24 variant
Causes Control 2011;22(1):109–​14. in prostate and breast cancer from a large nested
Saramaki OR, Porkka KP, Vessella RL, Visakorpi T. case-​control study. Cancer Res 2007;67(7):2951–​6.
Genetic aberrations in prostate cancer by micro- Schuurman AG, Zeegers MPA, Goldbohm RA, van
array analysis. Int J Cancer 2006;119(6):1322–​9. den Brandt PA. A case-​cohort study on prostate
Sartor O, Zheng Q, Eastham JA. Androgen receptor cancer risk in relation to family history of prostate
gene CAG repeat length varies in a race-​specific cancer. Epidemiology 1999a;10:192–​5.
fashion in men without prostate cancer. Urology Schuurman AG, van den Brandt PA, Dorant E, Brants
1999;53:378–​80. HA, Goldbohm RA. Association of energy and
Sato K, Qian J, Slezak JM, Lieber MM, Bostwick DG, fat intake with prostate carcinoma risk:  results
Bergstralh EJ, Jenkins RB. Clinical significance from The Netherlands Cohort Study. Cancer
of alterations of chromosome 8 in high-​grade, 1999b;86(6):1019–​27.
advanced, nonmetastatic prostate carcinoma. J Schuurman AG, Goldbohm RA, Dorant E, van den
Natl Cancer Inst 1999;91(18):1574–​80. Brandt PA. Anthropometry in relation to prostate
Savage CJ, Lilja H, Cronin AM, Ulmert D, Vickers AJ. cancer risk in the Netherlands Cohort Study. Am J
Empirical estimates of the lead time distribution Epidemiol 2000;151(6):541–​9.
for prostate cancer based on two independent rep- Serth J, Panitz F, Paeslack U, Kuczyk MA, Jonas U.
resentative cohorts of men not subject to prostate-​ Increased levels of human papillomavirus type 16
specific antigen screening. Cancer Epidemiol DNA in a subset of prostate cancers. Cancer Res
Biomarkers Prev 2010;19(5):1201–​7. 1999;59(4):823–​5.
Schaid DJ, McDonnell SK, Blute ML, Thibodeau SN. Sfanos KS, Isaacs WB, De Marzo AM. Infections and
Evidence for autosomal dominant inheritance of inflammation in prostate cancer. Am J Clin Exp
prostate cancer. Am J Hum Genet 1998;62:1425–​38. Urol 2013;1(1):3–​11.
Schenk JM, Till C, Hsing AW, Stanczyk FZ, Gong Z, Shang Y, Han G, Li J, Zhao J, Cui D, Liu C, Yi S.
Neuhouser ML, et al. Serum androgens and pros- Vasectomy and prostate cancer risk:  a meta-​
tate cancer risk:  results from the placebo arm of analysis of cohort studies. Sci Rep 2015;5:9920.
the Prostate Cancer Prevention Trial. Cancer Sharma S, Cao X, Wilkens LR, Yamamoto J, Lum-​Jones
Causes Control 2015 [Epub ahead of print]. A, Henderson BE, Kolonel LN, Le Marchand L.
Schildkraut JM, Demark-​ Wahnefried W, Wenham Well-​done meat consumption, NAT1 and NAT2
RM, Grubber J, Jeffreys AS, Grambow SC, et  al. acetylator genotypes and prostate cancer risk: the
520

520 Textbook of Cancer Epidemiology

multiethnic cohort study. Cancer Epidemiol androgen receptor gene:  molecular markers for
Biomarkers Prev 2010;19(7):1866–​70. prostate cancer risk. Cancer Res 1997;57:1194–​8.
Shimizu H, Ross RK, Bernstein L, Yatani R, Henderson Stark JR, Judson G, Alderete JF, Mundodi V, Kucknoor
BE, Mack TM. Cancers of the prostate and breast AS, Giovannucci EL, et  al. Prospective study of
among Japanese and white immigrants in Los Trichomonas vaginalis infection and prostate can-
Angeles County. Br J Cancer 1991;63:963–​6. cer incidence and mortality:  Physicians' Health
Shui IM, Mondul AM, Lindström S, Tsilidis KK, Travis Study. J Natl Cancer Inst 2009;101(20):1406–​11.
RC, Gerke T, et  al.; Breast and Prostate Cancer Strickler HD, Burk R, Shah K, Viscidi R, Jackson A,
Cohort Consortium Group. Circulating vitamin Pizza G, et  al. A multifaceted study of human
D, vitamin D-​related genetic variation, and risk papillomavirus and prostate carcinoma. Cancer
of fatal prostate cancer in the National Cancer 1998;82(6):1118–​25.
Institute Breast and Prostate Cancer Cohort Sturge J, Caley MP, Waxman J. Bone metastasis in
Consortium. Cancer 2015;121(12):1949–​56. prostate cancer:  emerging therapeutic strategies.
Shui IM, Mucci LA, Kraft P, Tamimi RM, Lindstrom S, Nat Rev Clin Oncol 2011;8(6):357–​68.
Penney KL, et al. Vitamin D-​related genetic varia- Suarez BK, Lin J, Burmester JK, Broman KW, Weber
tion, plasma vitamin D, and risk of lethal prostate JL, Banerjee TK, et al. A genome screen of mul-
cancer: a prospective nested case-​control study. J tiplex sibships with prostate cancer. Am J Hum
Natl Cancer Inst 2012;104(9):690–​9. Genet 2000;66:933–​44.
Shui IM, Lindström S, Kibel AS, Berndt SI, Campa Sun J, Wiklund F, Hsu FC, Balter K, Zheng SL, Johansson
D, Gerke T, et al. Prostate cancer (PCa) risk vari- JE, et al. Links Interactions of sequence variants in
ants and risk of fatal PCa in the National Cancer interleukin-​1 receptor-​associated kinase4 and the
Institute Breast and Prostate Cancer Cohort toll-​like receptor 6-​1-​10 gene cluster increase pros-
Consortium. Eur Urol 2014;65(6):1069–​75. tate cancer risk. Cancer Epidemiol Biomarkers Prev
Siddiqui MM, Wilson KM, Epstein MM, Rider JR, 2006;15(3):480–​5.
Martin NE, Stampfer MJ, Giovannucci EL, Mucci Surveillance, Epidemiology, and End Results (SEER)
LA. Vasectomy and risk of aggressive prostate Program (www.seer.cancer.gov) SEER*Stat
cancer:  a 24-​year follow-​up study. J Clin Oncol Database: Mortality—​All COD, Aggregated With
2014;32(27):3033–​8. State, Total U.S. (1969–​2012), National Cancer
Signorello LB, Kuper H, Lagiou P, Wuu J, Mucci LA, Institute, DCCPS, Surveillance Research Program,
Trichopoulos D, Adami HO. Lifestyle factors and Surveillance Systems Branch, released April 2015.
insulin-​like growth factor 1 levels among elderly Underlying mortality data provided by NCHS
men. Eur J Cancer Prev 2000;9(3):173–​8. (www.cdc.gov/​nchs).
Simon JA, Chen YH, Bent S. The relation of alpha-​ Sutcliffe S, Giovannucci E, Leitzmann MF, Rimm EB,
linolenic acid to the risk of prostate cancer: a sys- Stampfer MJ, Willett WC, Platz EA. A prospective
tematic review and meta-​analysis. Am J Clin Nutr cohort study of red wine consumption and risk of
2009;89(5):1558S-​1564S. prostate cancer. Int J Cancer 2007;120(7):1529–​35.
Simopoulos AP. Energy imbalance and cancer of the Sutcliffe S, Zenilman JM, Ghanem KG, Jadack RA,
breast, colon and prostate. Med Oncol Tumor Sokoll LJ, Elliott DJ, Nelson WG, De Marzo AM,
Pharmacother 1990;7(2–​3):109–​20. Cole SR, Isaacs WB, Platz EA. Sexually transmitted
Sinclair AL, Nouri AM, Oliver RT, Sexton C, infections and prostatic inflammation/​cell dam-
Dalgleish AG. Bladder and prostate cancer age as measured by serum prostate specific anti-
screening for human papillomavirus by polymer- gen concentration. J Urol 2006a;175(5):1937–​42.
ase chain reaction:  conflicting results using dif- Sutcliffe S, Giovannucci E, Alderete JF, et  al. Plasma
ferent annealing temperatures. Br J Biomed Sci antibodies against Trichomonas vaginalis and
1993;50(4):350–​4. subsequent risk of prostate cancer. Cancer
Sinha R, Park Y, Graubard BI, Leitzmann MF, Epidemiology Biomarkers and Prevention
Hollenbeck A, Schatzkin A, Cross AJ. Meat and 2006b;15(5):939–​45.
meat-​related compounds and risk of prostate can- Sutcliffe S, Nevin RL, Pakpahan R, Elliott DJ, Cole SR,
cer in a large prospective cohort study in the United De Marzo AM, et  al. Prostate involvement dur-
States. Am J Epidemiol 2009;170(9):1165–​77. ing sexually transmitted infections as measured
Smith JR, Freije D, Carpten JD, Gronberg H, Xu J, by prostate-​specific antigen concentration. Br J
Isaacs SD, et  al. Major susceptibility locus for Cancer 2011;105(5):602–​5.
prostate cancer on chromosome 1 suggested by Suzuki H, Komiya A, Aida S, Ito H, Yatani R,
a genome-​wide search. Science 1996;274:1371–​4. Shimazaki J. Detection of human papillomavirus
Stanford JL, Just JJ, Gibbs M, Wicklund KG, Neal CL, DNA and p53 gene mutations in human prostate
Blumenstein BA, et al. Polymorphic repeats in the cancer. Prostate 1996;28(5):318–​24.
521

Prostate Cancer 521

Svartberg J, Midtby M, Bonaa KH, Sundsfjord J, thermally processed tomato-​based food products.
Joakimsen RM, Jorde R. The associations of age, J Agric Food Chem 1995; 43:579–​86.
lifestyle factors and chronic disease with testoster-​ Torfadottir JE, Valdimarsdottir UA, Mucci LA,
one in men: the Tromsø study. Eur J Endocrinol Kasperzyk JL, Fall K, Tryggvadottir L, et  al.
2003;149(2):145–​152. Consumption of fish products across the
Svatek RS, Lee JJ, Roehrborn CG, Lippman SM, Lotan lifespan and prostate cancer risk. PLoS One
Y. Cost-​effectiveness of prostate cancer chemo- 2013;8(4):e59799.
prevention: a quality of life-​years analysis. Cancer Tretli S, Hernes E, Berg JP, Hestvik UE, Robsahm
2008;112(5):1058–​65. TE. Association between serum 25(OH)D
Svilaas A, Sakhi AK, Andersen LF, Svilaas T, Ström and death from prostate cancer. Br J Cancer
EC, Jacobs DR Jr, Ose L, Blomhoff R. Intakes of 2009;100(3):450–​4.
antioxidants in coffee, wine, and vegetables are Trichopoulos D, Lipworth L. Is cancer causation
correlated with plasma carotenoids in humans. J simpler than we thought, but more intractable?
Nutr 2004;134(3):562–​567. Epidemiology 1995;6(4):347–​9.
Szulkin R, Karlsson R, Whitington T, Aly M, Gronberg Tsuchiya N, Wang L, Horikawa Y, Inoue T, Kakinuma
H, Eeles RA, et al.; BPC3 consortium, Maehle L, H, Matsuura S, Sato K, Ogawa O, Kato T, Habuchi
Schleutker J, Lindström S, Wiklund F. Genome-​ T. CA repeat polymorphism in the insulin-​like
Wide Association Study of Prostate Cancer-​ growth factor-​I gene is associated with increased
Specific Survival. Cancer Epidemiol Biomarkers risk of prostate cancer and benign prostatic hyper-
Prev 2015;24(11):1796–​800. plasia. Int J Oncol 2005; 26(1):225–​31.
Szymanski KM, Wheeler DC, Mucci LA. Fish Tu H, Jacobs SC, Mergner WJ, Kyprianou N. Rare
consumption and prostate cancer risk:  a incidence of human papillomavirus types 16 and
review and meta-​ analysis. Am J Clin Nutr 18 in primary and metastatic human prostate can-
2010;92(5):1223–​33. cer. Urology 1994;44(5):726–​31.
Terris MK, Peehl DM. Human papillomavirus detec- Tunnicliffe JM, Shearer J. Coffee, glucose homeosta-
tion by polymerase chain reaction in benign and sis, and insulin resistance:  physiological mecha-
malignant prostate tissue is dependent on the nisms and mediators. Appl Physiol Nutr Metab
primer set utilized. Urology 1997;50(1):150–​6. 2008;33(6):1290–​1300.
Terry P, Lichtenstein P, Feychting M, Ahlbom A, Wolk van Asperen CJ, Brohet RM, Meijers-​ Heijboer EJ,
A. Fatty fish consumption and risk of prostate can- Hoogerbrugge N, Verhoef S, Vasen HF, et  al.;
cer. Lancet 2001;357:1764–​6. Netherlands Collaborative Group on Hereditary
Tham MD, Gardner DC, Haskell LW. Potential Health Breast Cancer (HEBON). Cancer risks in BRCA2
benefits of dietary phytoestrogens:  A review of families: estimates for sites other than breast and
the clinical epidemiological, and mechanistic evi- ovary. J Med Genet 2005;42(9):711–​9.
dence. J Clin Endoc and Metab 1998;83:2223–​35. Van Blarigan EL, Kenfield SA, Yang M, Sesso HD,
Thompson D, Easton DF; Breast Cancer Linkage Ma J, Stampfer MJ, Chan JM, Chavarro JE. Fat
Consortium. Cancer Incidence in BRCA1 mutation intake after prostate cancer diagnosis and mortal-
carriers. J Natl Cancer Inst 2002;94(18):1358–​65. ity in the Physicians' Health Study. Cancer Causes
Thompson IM, Ankerst DP, Chi C, Goodman PJ, Control 2015;26(8):1117–​26.
Tangen CM, Lucia MS, Feng Z, Parnes HL, van Duin M, van Marion R, Vissers K, Watson JE,
Coltman CA Jr. Assessing prostate cancer van Weerden WM, Schroder FH, et  al. High-​
risk: results from the Prostate Cancer Prevention resolution array comparative genomic hybridiza-
Trial. J Natl Cancer Inst 2006;98(8):529–​34. tion of chromosome arm 8q: evaluation of genetic
Thompson IM, Goodman PJ, Tangen CM, Lucia MS, progression markers for prostate cancer. Genes
Miller GJ, Ford LG, et al. The influence of finas- Chromosomes Cancer 2005;44(4):438–​49.
teride on the development of prostate cancer. N Van Laarhoven PHA. Trichomonas vaginalis, a path-
Engl J Med 2003;349(3):215–​24. ogen of prostatitis. Archivum Chirurgicum
Tomlins SA, Rhodes DR, Perner S, Dhanasekaran Neerlandicum 1967;19(3):263–​73.
SM, Mehra R, Sun XW, et al. Recurrent fusion of Vatten LJ, Ursin G, Ross RK, Stanczyk FZ, Lobo RA,
TMPRSS2 and ETS transcription factor genes in Harvei S, Jellum E. Androgens in serum and the
prostate cancer. Science 2005;310(5748):644–​8. risk of prostate cancer: a nested case-​control study
Tonnesen H, Moller H, Andersen JR, Jensen E, Juel K. from the Janus Serum Bank in Norway. Cancer
Cancer morbidity in alcohol abusers. Br J Cancer Epidemiol Biomarkers Prev 1997;6:967–​9.
1994;69:327–​32. Velicer CM, Kristal A, White E. Alcohol use and the
Tonucci LH, Holden JM, Beecher GR, Khachik F, risk of prostate cancer:  results from the VITAL
Davis CS, Mulokozi G. Carotenoid content of cohort study. Nutr Cancer 2006;56(1):50–​6.
52

522 Textbook of Cancer Epidemiology

Verdier Y, Chaffaux S, Boué F. Identification of post-​ and sporadic prostate cancer. Clin Cancer Res
vasectomy sperm auto-​ antigens in fox (Vulpes 2004;10(21):7150–​6.
vulpes) by two-​ dimensional gel electrophore- Wilson KM, Shui IM, Mucci LA, Giovannucci E.
sis and Western blotting. J Reprod Immunol Calcium and phosphorus intake and prostate can-
2002;54(1–​2):65–​80. cer risk:  a 24-​y follow-​up study. Am J Clin Nutr
Vickers AJ, Sjoberg DD, Ulmert D, Vertosick E, 2015;101(1):173–​83.
Roobol MJ, Thompson I, et  al. Empirical esti- Wilson KM, Kasperzyk JL, Rider JR, Kenfield S, van
mates of prostate cancer overdiagnosis by age and Dam RM, Stampfer MJ, Giovannucci E, Mucci LA.
prostate-​specific antigen. BMC Med 2014;12:26. Coffee consumption and prostate cancer risk and
Vinceti M, Venturelli M, Sighinolfi C, Trerotoli P, progression in the Health Professionals Follow-​up
Bonvicini F, Ferrari A, et al. Case-​control study of Study. J Natl Cancer Inst 2011;103(11):876–​84.
toenail cadmium and prostate cancer risk in Italy. World Cancer Research Fund International/​American
Sci Total Environ 2007;373(1):77–​81. Institute for Cancer Research. Continuous Update
Wang Y, Cui R, Xiao Y, Fang J, Xu Q. Effect of Carotene Project Report: Diet, Nutrition, Physical Activity,
and Lycopene on the Risk of Prostate Cancer:  A and Prostate Cancer. 2014. Available at:  www.
Systematic Review and Dose-​ Response Meta-​ wcrf.org/​sites/​default/​files/​Prostate-​Cancer-​2014-​
Analysis of Observational Studies. PLoS One Report.pdf
2015;10(9):e0137427. World Cancer Research Fund/​ American Institute
Wang Y, Jacobs EJ, Newton CC, McCullough ML. for Cancer Research. Food, Nutrition, Physical
Lycopene, tomato products and prostate cancer-​ Activity, and the Prevention of Cancer:  a Global
specific mortality among men diagnosed with Perspective. Washington, DC: AICR, 2007.
nonmetastatic prostate cancer in the Cancer Wright ME, Chang SC, Schatzkin A, Albanes D,
Prevention Study II Nutrition Cohort. Int J Kipnis V, Mouw T, Hurwitz P, Hollenbeck
Cancer. 2016 Jun 15;138(12):2846–​55. A, Leitzmann MF. Prospective study of adi-
Wang X, Guyer B, Paige DM. Differences in gesta- posity and weight change in relation to pros-
tional age-​ specific birthweight among Chinese, tate cancer incidence and mortality. Cancer
Japanese and white Americans. Int J Epidemiol 2007;109(4):675–​84.
1994;23(1):119–​28. Wu K, Spiegelman D, Hou T, Albanes D, Allen NE,
Watters JL, Park Y, Hollenbeck A, Schatzkin A, Berndt SI, et  al. Associations between unproc-
Albanes D. Alcoholic beverages and prostate essed red and processed meat, poultry, seafood
cancer in a prospective US cohort study. Am J and egg intake and the risk of prostate cancer: A
Epidemiol 2010;172(7):773–​80. pooled analysis of 15 prospective cohort studies.
Weiss JM, Huang WY, Rinaldi S, Fears TR, Chatterjee Int J Cancer. 2016 May 15;138(10):2368–​82.
N, Hsing AW, et al. Endogenous sex hormones and Xiang Y, Wang Z, Murakami J, et al. Effects of RNase
the risk of prostate cancer: a prospective study. Int L mutations associated with prostate cancer on
J Cancer 2008;122(10):2345–​50. apoptosis induced by 2',5'-​oligoadenylates. Cancer
Whitley BM, Moreira DM, Thomas JA, Aronson WJ, Res 2003;63(20):6795–​801.
Terris MK, Presti JC Jr, Kane CJ, Amling CL, Xu C, Han FF, Zeng XT, Liu TZ, Li S, Gao ZY. Fat Intake
Freedland SJ; SEARCH Database Study Group. Is Not Linked to Prostate Cancer:  A Systematic
Preoperative weight change and risk of adverse Review and Dose-​Response Meta-​Analysis. PLoS
outcome following radical prostatectomy:  results One 2015;10(7):e0131747.
from the Shared Equal Access Regional Cancer Xu J. Combined analysis of hereditary prostate can-
Hospital database. Prostate Cancer Prostatic Dis cer linkage to lq24-​25: results from 772 hereditary
2011;14(4):361–​6. prostate cancer families from the International
Whittemore AS, Wu AH, Kolonel LN, John EM, Consortium for Prostate Cancer Genetics. Am J
Gallagher RP, Howe GR, et al. Family history and Hum Genet 2000;66:945–​57.
prostate cancer risk in black, white, and Asian men Xu J, Meyers D, Freije D, Isaacs S, Wiley K, Nusskern
in the United States and Canada. Am J Epidemiol D, et  al. Evidence for a prostate cancer suscepti-
1995;141:732–​40. bility locus on the X chromosome. Nat Genet
Wideroff L, Schottenfeld D, Carey TE, Beals T, Fu 1998;20:175–​9.
G, Sakr W, et al. Human papillomavirus DNA in Xu J, Zheng SL, Komiya A, Mychaleckyj JC, Isaacs SD,
malignant and hyperplastic prostate tissue of black Hu JJ, et  al. Germline mutations and sequence
and white males. Prostate 1996;28(2):117–​23. variants of the macrophage scavenger receptor
Wiklund F, Jonsson BA, Brookes AJ, et al. Genetic anal- 1 gene are associated with prostate cancer risk.
ysis of the RNASEL gene in hereditary, familial, Nature Genetics 2002;32:321–​5.
523

Prostate Cancer 523

Yan L, Spitznagel EL. Soy consumption and prostate heart disease, and stroke: an ecological study. Prev
cancer risk in men:  a revisit of a meta-​analysis. Med 1999;28:520–​9.
Am J Clin Nutr 2009;89(4):1155–​63. Zhang X, Zhou G, Sun B, Zhao G, Liu D, Sun J, Liu C,
Yang M, Kenfield SA, Van Blarigan EL, Batista JL, Guo H. Impact of obesity upon prostate cancer-​
Sesso HD, Ma J, Stampfer MJ, Chavarro JE. associated mortality: A meta-​analysis of 17 cohort
Dietary patterns after prostate cancer diagnosis studies. Oncol Lett 2015;9(3):1307–​12.
in relation to disease-​specific and total mortality. Zheng SL, Augustsson-​Balter K, Chang B, Hedelin M,
Cancer Prev Res (Phila) 2015a;8(6):545–​51. Li L, Adami HO, et al. Sequence variants of toll-​
Yang M, Kenfield SA, Van Blarigan EL, Wilson KM, like receptor 4 are associated with prostate cancer
Batista JL, Sesso HD, Ma J, Stampfer MJ, Chavarro risk: results from the CAncer Prostate in Sweden
JE. Dairy intake after prostate cancer diagnosis in Study. Cancer Res 2004;64(8):2918–​22.
relation to disease-​specific and total mortality. Int Zhong S, Chen W, Yu X, Chen Z, Hu Q, Zhao J.
J Cancer 2015b;137(10):2462–​9. Coffee consumption and risk of prostate can-
Yang CM, Yen YT, Huang CS, Hu ML. Growth inhib- cer: an up-​to-​date meta-​analysis. Eur J Clin Nutr
itory efficacy of lycopene and β-​carotene against 2014;68(3):330–​7.
androgen-​ independent prostate tumor cells Zu K, Mucci L, Rosner BA, Clinton SK, Loda M,
xenografted in nude mice. Mol Nutr Food Res Stampfer MJ, Giovannucci E. Dietary lycopene,
2011;55(4):606–​12. angiogenesis, and prostate cancer:  a prospective
Zeegers MP, Kiemeney LA, Nieder AM, Ostrer H. How study in the prostate-​specific antigen era. J Natl
strong is the association between CAG and GGN Cancer Inst 2014;106(2):djt430.
repeat length polymorphisms in the androgen recep- Zuccolo L, Lewis SJ, Donovan JL, Hamdy FC, Neal
tor gene and prostate cancer risk? Cancer Epidemiol DE, Smith GD. Alcohol consumption and PSA-​
Biomarkers Prev 2004;13(11 Pt 1):1765–​71. detected prostate cancer risk-​ -​
a case-​control
Zhang J, Sasaki S, Amano K, Kesteloot H. Fish con- nested in the ProtecT study. Int J Cancer
sumption and mortality from all causes, ischemic 2013;132(9):2176–​85.
524
52

21
Testicular Cancer
L O R E N Z O R I C H I A R D I , G I O V E N A L E M O I R A N O , A N D PA G O N A   L A G I O U

T esticular cancer is highly curable and rela-

tively rare, but it is the most common can-
cer among young men. It offers real challenges
Treatment Orchidectomy remains the standard
treatment. Complementary radiotherapy is often
used in the primary management of semino-
due to its unique descriptive epidemiology and mas, whereas treatment of nonseminomas often
unknown etiology. Incidence of testicular cancer includes a cisplatinum-​based chemotherapy reg-
has been increasing constantly throughout the imen, which may be used also in seminomas as
world since the beginning of the 20th century, an alternative to radiotherapy. In more advanced
but risk factors responsible for the rise in inci- stages, retroperitoneal lymph node dissection
dence remain enigmatic. may be necessary. In patients with metastatic dis-
Misclassification and underascertainment are ease, chemotherapy is the cornerstone of clinical
probably less common in testicular cancer than management.
for most other malignancies:  the organ is visi- Prognosis Introduction of cisplatinum-​ based
ble and palpable, and the origin of the tumor is therapy during the 1970s is one of the success
obvious. A real difficulty in analytic studies, how- stories of modern oncology. Five-​year relative
ever, is measurement of the etiologically relevant survival is nowadays 95% in areas where state-​of-​
exposures, since they are likely to operate early in the-​art technology is available.
life, including in utero.
Progress Research is now focusing mainly
on reduction of treatment-​ related long-​
term
CLINICAL SYNOPSIS
morbidity associated with radiotherapy and
Subgroups Ninety percent of testicular tumors are
chemotherapy.
of germ-​cell origin, and this chapter focuses only
on germ-​cell cancers. Testicular cancer occurs in
DESCRIPTIVE
two main histopathologic types, seminomas and
EPIDEMIOLOGY
nonseminomas, the latter being predominantly
The incidence of testicular cancer is low before
teratomas. Although these two types have differ-
puberty. After puberty there is a dramatic
ent clinical characteristics and peak in incidence
increase in incidence, which peaks at around
at different ages, they have otherwise a similar
age 30 and declines to almost zero by age 60
descriptive epidemiology.
(Fig.  21-​1). Because the disease occurs prima-
Symptoms A painless palpable mass usually per- rily in young men and is virtually nonexistent
ceived as an enlargement of the testicle is the car- in older men, it is generally believed that early
dinal symptom of testicular cancer. life exposures play a role in the development of
Diagnosis Traditionally, orchidectomy with testicular cancer (Richiardi et  al, 2004a; Bray
histopathologic examination, sometimes fro- et  al, 2006). The considerable increase in inci-
zen section examination from an open biopsy, dence rates after puberty has been interpreted by
serve both diagnostic and therapeutic purposes. some to indicate that adult hormones are neces-
Ultrasounds are performed to confirm the pres- sary for the growth and progression of testicular
ence of a testicular mass and to explore the tumors (Moss et al, 1986; Moller & Skakkebaek,
contralateral testis. Tumor markers and abdomi- 1996; Maule et  al, 2012). The age-​specific inci-
nopelvic computed tomography are used for dence patterns for nonsemimonas and semino-
diagnostic and prognostic purposes. mas are different, with nonseminomas having a
526

526 Textbook of Cancer Epidemiology

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 2 4 6 8 10
ASR (world) per 100,000 person-years

FIGURE 21-​1  Age-standardized (to the 2012 world population) incidence rates of testicular cancer.
Source: Ferlay et al, 2013.

peak at around ages 25 to 30, whereas for semi- et al, 2006; Trabert et al, 2015). Since the 1970s,
nomas, peak incidence occurs at around age 35 annual percentage changes of 1% to 5% have been
(Richiardi et al, 2004a). observed in Europe, North America, Central and
The geographic variation in incidence rates South America, and Oceania (Trabert et al, 2015).
of testicular cancer worldwide is considerable A similar increasing trend has been reported also
(Fig. 21-​2). The incidence is highest in west- for the period before the 1970s in the Nordic
ern European populations, with Norway and countries, where cancer registration was already
Switzerland leading at an age-​ standardized active (Richiardi et al, 2004a).
(world) rate above 12 per 100,000 person-​years Age-​period-​cohort analysis reveals that birth
(Ferlay et al, 2013). The lowest rates are observed cohort is a strong determinant of testicular can-
in Asian and African populations, with age-​ cer (Richiardi et  al, 2004a; Znaor et  al, 2015).
standardized rates estimated to be below 1 per Interestingly, in the Scandinavian countries,
100,000 person-​years. the increases in testicular cancer incidence are
Globally, the incidence rates for testicular slowed in cohorts born between the late 1930s
cancer have been increasing dramatically over and early 1940s (Bergström et al, 1996; Richiardi
the past century. The increase has been largely the et  al, 2004a). Apart from this “World War II”
same for seminomas and nonseminomas (Bray effect, in most populations each successive birth

18
New cases per 100,000

16
14
person-years

12
10
8
6
4
2
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Poland Sweden United Kingdom Japan

Spain The Netherlands United States of America

FIGURE 21-​2  Age-specific incidence rates of testicular cancer.

Source: Ferlay et al, 2013.
527

Testicular Cancer 527

cohort has experienced an increased incidence of With few exceptions, pre genome-​wide asso-
testicular cancer. This trend has continued in the ciation studies (GWAS) did not unfold under-
most recent generations, although it has attenu- lying genetic bases for familial testicular cancer.
ated in some high-​incidence European countries One study reported an increased risk of testicular
(Znaor et al, 2015). cancer among carriers of the uncommon “gr/​gr”
There are a few notable exceptions to this oth- deletion, a 1.6-​Mb deletion of the Y chromosome
erwise universal trend:  the US black population, that removes part of the AZFc region and is asso-
for example, has had stable low incidence rates ciated with spermatogenic failure (Nathanson
until the 1990s, when incidence started to increase et al, 2005). A collaborative linkage study on tes-
slightly. In the United States, testicular cancer ticular cancer did not find evidence for genetic
incidence remains much lower among black than linkage, suggesting that genetic susceptibility
white or Hispanic men (McGlynn et  al, 2005; involves multiple moderate risk genes (Crockford
Ghazarian et  al, 2015; Znaor et  al, 2015). Ethnic et al, 2006).
differences have been reported also in other coun- A number of GWAS have been con-
tries. In New Zealand, testicular cancer has been ducted since the late 2000s (Kanetsky et  al,
increasing in all ethnic groups but Māori experi- 2009; Turnbull et  al, 2010; Chung et  al, 2013;
ence a higher incidence than men of a European Schumacher et  al, 2013; Kristiansen et  al, 2015;
origin and a much higher incidence than the Litchfield et  al, 2015a), and a Testicular Cancer
Asian and Pacific population (Sarfati et al, 2011). Consortium has been founded with the primary
The large geographic variations in testicu- aim of examining genetic susceptibility to testic-
lar cancer incidence make migration studies of ular cancer. As recently reviewed, a number of
interest. Migrants tend to maintain the baseline independent polymorphisms have been associ-
risk of their country of origin (Grulich et al, 1995; ated with an at least 30% increased risk of testic-
Ekbom et al, 2003; Myrup et al, 2008; Levine et al, ular cancer (Table 21-​1) (Litchfield et al, 2015b).
2013), with no evidence of change in risk by age KITLG polymorphisms, in particular, have been
at immigration and duration of stay (Ekbom associated with a per-​allele odds ratio above 2,
et al, 2003; Myrup et al, 2007). The second-​and which has been replicated in candidate-​ gene
third-​generation immigrants, however, experi- studies conducted in populations that were not
ence rates that tend to approach those of the host covered by the GWAS (Ferlin et al, 2012; Lessel
population (Myrup et al, 2007; Hemminki & Li, et al, 2012).
2002; Levine et al, 2013). These patterns suggest Some of the genes reported in Table 21-​1
that environmental factors are fundamental in have specific functions related to germ cells.
the development of testicular cancer and should KITLG encodes the ligand for the type III
play a role early in life. receptor tyrosine kinase KIT. The KIT/​KITLG
system plays an important role in the survival,
GENETIC AND MOLECULAR proliferation, motility, and migration of germ
EPIDEMIOLOGY cells (Galan et  al, 2006; Boldajipour & Raz,
Testicular cancer has a strong genetic compo-
nent as suported by different types of studies and
sources of evidence. TABLE 21-​1   MAIN POLYMORPHISMS
ASSOCIATED WITH A 30% OR MORE
Inherited Susceptibility INCREASE IN TESTICULAR CANCER
The early age of onset among familial cases found RISK IN AT LEAST ONE GENOME-WIDE
in some studies (Dong et al, 2001; Forman et al, ASSOCIATION STUDY
1992)  and the high proportion of bilateral tes-
ticular cancer (Heimdal et  al, 1996)  suggest SNP Gene
that there is an important genetic component to
the disease. A  study based on cancer registries rs995030/​rs1508595 KITLG
from the Nordic countries estimated a fourfold rs210138 BAK1
increased risk of testicular cancer among broth- rs4624820 SPRY4
ers and a twofold increased risk among fathers/​ rs4635969 TERT/​CLPTM1L
sons of testicular cancer patients. For men with rs755383 DMRT1
at least two affected relatives, the increase in risk rs8046148 HEATR3
was almost 20-​fold (Kharazmi et al, 2015).
528

528 Textbook of Cancer Epidemiology

2007). BAK1 and SPRY4 are also linked, directly cancer (Skakkebaek et  al, 2001). According
or indirectly, with the KIT/​KITLG pathway, to the notion of the TDS, the aforementioned
and DMRT1 is involved in sex determination four conditions share an environmental etiol-
(Raymond et al, 1998). Among the other genes ogy, acting during fetal life through a disruption
reported in Table 21-​1, TERT is implicated in of the gonadal development (Skakkebaek et  al,
the telomerase function. 2016). The utility of the concept of a TDS for
Testicular cancer is also associated with etiological studies on testicular cancer has been
a number of genetic disorders, including also challenged on account of limited evidence
Klinefelter’s syndrome (47,XXY), Turner’s syn- for shared causes among the components of
drome, male pseudohermaphroditism, testicular the syndrome (Akre & Richiardi, 2009; Thorup
feminization, gonadal dysgenesis, Down’s syn- et al, 2010).
drome, and recessive X-​linked icthyosis (Lutke Another influential hypothesis that has shaped
Holzik et al, 2003; Cools et al, 2006). research on testicular cancer postulates that steroid
sex hormones, including endogenous hormones
Somatic Events and/​or exposure to endocrine disruptors, play a
An isochromosome of the short arm of chro- fundamental etiological role. When describing the
mosome 12 has been identified in precursors individual risk factors in this chapter, we mention
of germ cell cancer and in 80% of all testicular whether an underlying hormonal mechanism has
tumors, regardless of histology (Skotheim & been suggested to explain the association between
Lothe, 2003). that risk factor and testicular cancer risk.
Several studies have evaluated whether semi-
RISK FACTORS nomas and nonseminomas share the same or
The etiology of testicular cancer is elusive. As have different risk factors, and results indicate a
summarized in Table 21-​2, there are few well-​ similar etiology for these two histological sub-
established risk factors, and the causal mecha- types (Stang & Kuss, 2011). Thus, in this chapter,
nisms of these exposures remain to be elucidated. we discuss them together unless otherwise stated.
Molecular similarities between fetal germ cells
and carcinoma in situ (CIS) of the testis—​ a Prenatal Exposures
precursor of germ cell testicular cancer first Substantial data indicate that twinning is associ-
identified in the 1970s (Skakkebaek, 1972) and ated with an increased risk of testicular cancer,
nowadays also termed “testicular intraepithelial and some evidence indicates that exposure to
neoplasia” or “intratubular germ cell neopla- hormonal agents during pregnancy, especially
sia of unspecified type”—​suggest that prenatal diethylstilbestrol, may increase the risk of testic-
and perinatal exposures have a strong impact on ular cancer. Maternal age, smoking during preg-
the risk of testicular cancer (Sonne et al, 2009; nancy, and nausea during pregnancy seem to be
Kristensen et al, 2013; Rijlaarsdam et al, 2015). It unrelated to the risk of testicular cancer. Evidence
has been hypothesized that testicular cancer may on maternal hypertension is conflicting.
derive from primordial germ cells that escaped
normal differentiation (Skakkebaek et al, 1987). Twinning
Postnatal exposures in childhood, adolescence, Twin pregnancies have a higher frequency of
and possibly early adulthood, however, are also intrauterine growth retardation, and higher
likely to have an impact (Richiardi et al, 2007). maternal and lower cord blood estrogen and
For some risk factors, for example height or indi- progesterone levels compared with singleton
cators of hormonal status, it is not possible to dis- pregnancies (Hickey et  al, 2014; Kuijper et  al,
tinguish whether their association with the risk 2015). In a meta-​analysis of 12 studies, twinning
of testicular cancer involves prenatal or postnatal was associated with a 22% increase in testicular
mechanisms or both. cancer risk (Cook et al, 2010). A Swedish study
In 2001, the existence of a testicular dys- compared twins with their singleton siblings
genesis syndrome (TDS) was proposed, encom- (twinning influence), and singletons from twin
passing impaired male fertility, hypospadias, families with singletons from nontwin fami-
cryptorchidism, and testicular cancer. This lies (twin family influence); results pointed to a
has been influential in the design and inter- role of twinning, rather than twin family, influ-
pretation of etiological studies on testicular ence (Chen et al, 2015). Studies suggest that the
529

Testicular Cancer 529

TABLE 21-​2   EPIDEMIOLOGY OF TESTICULAR CANCER: RISK FACTOR SUMMARY

Risk factors Direction of effect*

Well-​Confirmed Risk Factors

Contralateral testicular cancer ↑↑
Cryptorchidism ↑↑
Ethnic group** NA
Family history ↑↑
Tallness ↑

Probable Relationship Exists, Based on Substantial Data

Birth order/​sibship size ↓
Late puberty ↓
Hypospadias ↑↑
Subfertility ↑
Inguinal hernia ↑
Perinatal characteristics (low birth order, short gestational duration, ↑
twinning)

Weak, if Any, Relationship Exists, Based on Substantial Data

Tobacco -​
Alcohol -​
Vasectomy -​

Inconsistent Findings or Limited Study to Date

Baldness*** ↓
Marijuana*** ↑
Fetal exposure to endocrine disruptors (e.g., diethylstilbestrol) ↑
Maternal hypertension**** ↑↓
Immunosuppression ↑
Postnatal exposure to pesticides—​organochlorine compounds ↑

*Arrows indicate the approximate magnitude of the relationship:  ↑, slight to moderate increase in risk; ↑↑, moderate to large
increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; -​, no association
Note: The magnitude of the risk can vary substantially, depending on the exact comparison being made.
** In the United States, the rate of testicular cancer is about six times greater in whites as compared with blacks.
*** Only three studies have been conducted, but results are consistent.
****Both increased and decreased risks have been reported, possibly depending on whether hypertension is mild (increased risk) or
severe (decreased risk).

twinning influence is little modified by zigosity evidence of an association of severe nausea with
or, among dizygotic twins, by the sex of the other testicular cancer risk (Cook et al, 2009).
twin (Neale et al, 2008).
Maternal Hypertension
Nausea during Pregnancy Because gestational hypertension is a sign
Severe nausea has been studied as a proxy of of placental malfunction and the placenta is
levels of estrogens and chorionic gonadotropin instrumental for the production of pregnancy
in the first trimester of pregnancy (Depue et al, hormones, an association between mater-
1987). A meta-​analysis of 12 studies found little nal hypertension and the risk of testicular
530

530 Textbook of Cancer Epidemiology

cancer would support the role of fetal expo- 1979; Brown et  al, 1986; Moller & Skakkebaek,
sure to altered sex hormonal levels in testicu- 1996; Swerdlow et  al, 1987a; Weir et  al, 2000;
lar cancer etiology. A  meta-​analysis suggested Coupland et al, 2004; McGlynn et al, 2006), and
a weak increase in the risk of testicular cancer the same applies for studies in which the informa-
for maternal hypertension (based on eight stud- tion on maternal smoking was obtained prospec-
ies), which was not supported by an increased tively (Pettersson et al, 2007a; Mongraw-​Chaffin
risk of testicular cancer for preeclampsia (based et al, 2009; Tuomisto et al, 2009).
on eight studies) (Cook et  al, 2009). The larg-
est of the studies included in the meta-​analysis Perinatal Characteristics
used registry-​based information to define mild Low birthweight and short gestational duration
and severe gestational hypertension (Pettersson are likely to be associated with an increased risk
et  al, 2008). In that study, severe hypertension of testicular cancer, while other investigated pre-
was associated with a strongly decreased risk of natal characteristics (birth length, jaundice, hav-
testicular cancer, while mild hypertension was ing been breastfed) are unlikely to be risk factors
associated with an increased risk. for testicular cancer.

Maternal Age Birthweight
Maternal age has been largely investigated, A meta-​ analysis of 15 studies found a 34%
because it can be easily measured, it is a risk increased risk of testicular cancer for being born
factor for congenital malformations, and might with a low compared with normal birthweight
be associated with increased levels of maternal (Cook et al, 2010). Studies that obtained informa-
sex hormones in pregnancy (Troisi et  al, 2008). tion on birthweight from maternal reporting
Overall, there is no evidence of an association found higher risk estimates than registry-​based
between maternal age and the risk of testicular studies, but the difference was small and it cannot
cancer (Cook et al, 2009). be convincingly concluded that the meta-​analytic
estimate was affected by recall bias.
Hormonal Agents during Pregnancy The increased risk for low birthweight was
A meta-​analysis of nine studies estimated a rel- not paralleled by a decreased risk among men
ative risk of testicular cancer of approximately with a high birthweight; consistently, in the
2 for exogenous hormones during pregnancy meta-​analysis, there was no association with tes-
(Martin et al, 2008). Out of these studies, there ticular cancer risk when birthweight was used as
was only one cohort with prospectively col- a continuous variable (Cook et al, 2010).
lected data on the use of diethylstilbestrol (DES)
in pregnancy—​DES is a synthetic estrogen that Gestational Duration
has been used in the past to supposedly prevent Gestational duration has been examined as a
spontaneous abortions and pregnancy compli- risk factor for testicular cancer as an indicator
cations and has been strongly associated with both of fetal growth and of exposure to preg-
the risk of adenocarcinoma of the vagina. The nancy hormones. A 31% increased risk of testic-
study found a threefold increased risk of testic- ular cancer for preterm birth was estimated in a
ular cancer, based on seven events (Strohsnitter meta-​analysis of 12 studies; the inverse associa-
et  al, 2001). The other studies on exogenous tion between duration of gestation and testicular
hormones in pregnancy relied on self-​reported cancer risk remained when gestational dura-
information and are potentially affected by recalltion was treated as a continuous variable (Cook
bias and misclassification of the exposure status.et  al, 2010). Similar as for birthweight, studies
based on maternal reporting found higher esti-
Maternal Smoking in Pregnancy mates than registry-​based studies, suggesting a
Because it has been observed that the increasing possible role of recall bias. A Swedish registry-​
rates of smoking in women parallel the increas- based study, published after the meta-​analysis,
ing rates of testicular cancer, maternal smoking reported an increased risk of testicular cancer
during pregnancy has been hypothesized to be in men who were born extremely prematurely
a cause of testicular cancer (Clemmesen, 1997; (after 22–​29 weeks of gestation), but the evi-
Pettersson et al, 2004). However, questionnaire-​ dence for an association with later preterm birth
based case-​control studies on maternal smoking or post-​term birth was much weaker (Crump
have been consistently null (Henderson et  al, et al, 2012).
531

Testicular Cancer 531

Other Perinatal Characteristics sizes (Schnack et al, 2010; Trabert et al, 2013;
Birth length, neonatal jaundice, and having been Le Cornet et al, 2015). Previous studies included
breastfed have been assessed in a number of stud- only few exposed cases and were, thus, incon-
ies on testicular cancer with, overall, null results clusive (Swerdlow et al, 1987b; United Kingdom
(Cook et al, 2010). Testicular Cancer Study Group, 1994a; Prener
et al, 1996; Moller et al, 1996).
Cryptorchidism, Hypospadias, and
Other Congenital Malformations Inguinal Hernia and Other
Cryptorchidism is a well-​ established risk fac- Malformations
tor for testicular cancer, and there is consistent A meta-​ analysis of 19 studies found a 60%
evidence that inguinal hernia, hypospadias, and increased risk of testicular cancer for men who
other genital malformations are also associated had had inguinal hernia (Cook et al, 2010). It has
with an increased risk. been argued that the association between ingui-
nal hernia and the risk of testicular cancer could
Cryptorchidism be explained by confounding from diagnosed or
Cryptorchidism, that is undescended testes, is undiagnosed cryptorchidism (Moller et al, 1996).
a well-​established risk factor for testicular can- However, in a registry-​based study from Sweden
cer. A meta-​analysis of 34 studies has estimated published after the meta-​analysis, the observed
a relative risk of 4.3 (95% CI 3.6–​5.1) (Cook increased risk of testicular cancer for inguinal
et al, 2010), an estimate that is further supported hernia (based on 118 exposed cases) changed
by nationwide studies in Denmark (Schnack only slightly after adjustment for cryptorchidism
et al, 2010) and Sweden (Trabert et al, 2013). In (113 exposed cases) (Trabert et al, 2013).
the Danish study, a family history of crytporchi- The same study also examined “other gen-
dism was not associated with the risk of testicular ital malformations” after exclusion of cryptor-
cancer (Schnack et al, 2010). chidism, hypospadias, and inguinal hernia. This
Although the association with cryptorchi- heterogeneous group of malformations was asso-
dism is strong and well established, it is unclear ciated with a twofold increased risk of testicular
whether the ectopic position of the testis itself cancer, based on 14 exposed cases; the relative
is a cause of testicular cancer or whether the risk did not appreciably change after adjustment
two conditions share a common etiology. Most for cryptorchidism, hypospadias, and inguinal
likely, both mechanisms play a role. In a number hernia. Of note, nongenital malformations eval-
of studies, the risk of testicular cancer has been uated were not associated with testicular cancer
found to be higher in, but not confined to, men risk (Trabert et al, 2013).
who underwent a delayed orchiopexy (Herrinton
et  al, 2003; Walsh et  al, 2007; Pettersson et  al, Puberty and Postnatal Indicators
2007b; Banks et  al, 2013), and it has been sug- of Hormonal Status
gested that, with regard to the risk of testicular Late puberty and, in a smaller number of studies,
cancer, cryptorchidism should be treated before baldness have been consistently associated with a
puberty (Pettersson et  al, 2007b). In men with decreased risk of testicular cancer, while history
unilateral cryptorchidism, although the risk of of severe acne does not seem to be associated.
cancer is increased in both testes, it has been
reported to be much higher in the maldescended Age at Puberty
than the contralateral testis (Akre et  al, 2009; A number of case-​ control studies on testicu-
Banks et al, 2013). lar cancer have retrospectively assessed age at
puberty using indirect indicators, including age
Hypospadias at voice change or age at start shaving compared
Hypospadias, a condition in which the ure- to peers. A  meta-​analysis of 12 studies found a
thral opening is located along the underside decreased risk of testicular cancer associated
rather than at the tip of the penis, is associ- with late puberty and no association with early
ated with an increased risk of testicular cancer. puberty (Maule et  al, 2012). An association
Epidemiological studies on hypospadias are dif- between age at puberty and testicular cancer
ficult, because it is a rare condition, but a two- risk could be explained by shared environmen-
fold increased risk has been consistently reported tal or genetic factors or by differences in cumu-
in nationwide studies based on large sample lative hormonal exposure between men who
532

532 Textbook of Cancer Epidemiology

experienced a late puberty and those who did Diet

not. In addition, since testicular cancer rarely There is little evidence that diet plays a role in
occurs before puberty, a late puberty may imply a testicular cancer etiology. An ecological study
delayed entry in the risk period. had reported that incidence of testicular cancer
was highly correlated with per capita consump-
Baldness tion of cheese, milk, and animal fat (Ganmaa
Baldness has been studied as a proxy for et al, 2002).
increased androgen activity. Three stud- As recently reviewed (McGlynn & Trabert,
ies conducted in three different populations 2012), results from case-​control studies on diets
have consistently found an inverse association high in dairy products, cheese, milk, or fats
between baldness and the risk of testicular can- are conflicting (Davies et  al, 1996; Sigurdson
cer (Petridou et  al, 1997; Trabert et  al, 2011b; et al, 1999; Garner et al, 2003; Stang et al, 2006;
Moirano et al, 2016). McGlynn et al, 2007); the studies are limited by
the difficulty in assessing diet retrospectively to
Acne cover habits in childhood, adolescence, and early
Severe acne is another proxy for increased adulthood (Stang et  al, 2006). Studies assessing
androgen exposure (Pochi, 1982). Most stud- other dietary factors, such as meat, cholesterol,
ies on severe acne and testicular cancer risk and fiber intake, in relation to testicular can-
have reported a null association (Henderson cer have also been inconclusive (Swerdlow et al,
et  al, 1979; Brown et  al, 1987; United Kingdom 1989; Sigurdson et  al, 1999; Bonner et  al, 2002;
Testicular Cancer Study Group, 1994b), whereas Wirehn et al, 2005; Hu et al, 2012). One of these
some have found an inverse association (Depue studies, relying on a cohort of approximately
et al, 1983; Trabert et al, 2011b). 45,000 Swedish men, reported that high base-
line serum cholesterol levels were associated with
Tobacco an increased risk of testicular cancer, with evi-
Studies conducted on tobacco smoking and the dence of a dose-​response relationship (Wirehn
risk of testicular cancer have been consistently et al, 2005).
null (Henderson et  al, 1979; Brown et  al, 1987;
United Kingdom Testicular Cancer Study Group, Physical Activity
1994b; Thune & Lund, 1994; Gallagher et  al, Studies on physical activity and testicular can-
1995; Moller & Skakkebeak, 1996a; Trabert et al, cer show little consistency in results. Positive
2011a; Lacson et al, 2012), with the exception of (Srivastava & Kreiger, 2000), inverse (United
few studies that found some evidence of a posi- Kingdom Testicular Cancer Study Group, 1994a;
tive association between heavy smokers and tes- Gallagher et  al, 1995), and null (Paffenbarger
ticular cancer risk (Srivastava & Kreiger, 2004; et  al, 1997; Thune & Lund, 1994; Littman et  al,
Dusek et al, 2008). 2009)  associations have been reported. These
studies have limitations both in the assessment
Marijuana of physical activity through questionnaires and
Use of marijuana has been associated with an in the choice of the relevant period in life to be
approximately twofold increased risk of testicu- investigated. Only four studies included informa-
lar cancer in three US studies (Daling et al, 2009; tion about physical activity in childhood or ado-
Trabert et al, 2011a; Lacson et al, 2012). The risk lescence, and their results were conflicting, with
seems to increase with younger age at first use reported associations ranging from positive
(Daling et  al, 2009)  and frequency and dura- to inverse to null, depending on the type and
tion of use (Gurney et al, 2015), and seems to be intensity of the activity or on the reporting by
stronger for nonseminomas than for seminomas the mother or the index subject (Gallagher et al,
(Gurney et al, 2015). 1995; Littman et  al, 2009; Srivastava & Kreiger,
2000; Cook et al, 2008).
Alcohol Some studies assessed specific recreational
Studies of alcohol in relation to testicular can- activities separately—​ for example, riding
cer have been consistently null (Henderson activities that can cause testicular traumas—​
et  al, 1979; Brown et  al, 1987; United Kingdom but, overall, no consistent associations with the
Testicular Cancer Study Group, 1994b; Dusek risk of testicular cancer have emerged (Littman
et al, 2008; Lacson et al, 2012; Trabert et al, 2011a). et al, 2009; McGlynn & Trabert, 2012). Of
53

Testicular Cancer 533

interest is a study that reported a positive asso- In addition, adjustment for cryptorchidism was
ciation between use of muscle-​building supple- not always possible.
ments and risk of testicular cancer (Li et al, A registry-​based Swedish study found that
2015). The main ingredients in muscle-​build- brothers, but not sisters, of testicular cancer
ing supplements include creatine and proteins, patients might have impaired fertility, measured
but undeclared anabolic androgenic steroids as decreased number of children and lower fre-
are detectable in many commercially availa- quency of opposite-​sex twins among the child-
ble nonhormonal supplements (Abbate et al, ren (Richiardi & Akre, 2005). Genetic or shared
2015). environmental factors may explain this finding,
which argues in favor of shared causes between
Fertility and Semen Quality cancer-​associated subfertility and testicular
Impaired fertility has been consistently associ- cancer.
ated with an increased risk of testicular cancer;
however, the mechanism underlying this asso- Anthropometric Measures
ciation is not known. It has been hypothesized Tallness is an established risk factor for testicular
that perhaps a growing tumor may affect fertility cancer, while there is limited evidence of an asso-
or that there may be a common etiological link ciation between body mass index and the risk of
between subfertility and testicular cancer. testicular cancer.
Registry-​based studies in Denmark, Norway,
and Sweden revealed that men with testicular Height
cancer father fewer children than population Tall stature has been consistently positively asso-
controls, and this association is apparent sev- ciated with the risk of testicular cancer. A meta-​
eral years prior to the cancer diagnosis (Jacobsen analysis of 13 studies has estimated a 13% increase
et  al, 2000a; Fossa & Kravdal, 2000; Richiardi in risk per 5-​cm increase in height (Lerro et al,
et  al, 2004b). Several questionnaire-​based case-​ 2010). Additional studies published after the
control studies have analyzed different indica- meta-​ analysis have supported this association
tors of subfertility—​ including prior diagnosis (Trabert et  al, 2011b;Giannandrea et  al, 2012;
of infertility, number of children, and time to Levine et  al, 2013; Richiardi et  al, 2014; Wiren
pregnancy—​in association with the risk of tes- et al, 2014). The association between height and
ticular cancer (Haughey et  al, 1989; Swerdlow testicular cancer risk seems to be independent of
et  al, 1989; United Kingdom Testicular Cancer birthweight (Richiardi et al, 2003), age at puberty
Study Group, 1994a; Moller & Skakkebaek, 1999; (McGlynn et  al, 2007; Trabert et  al, 2011b;
Doria-​Rose et al, 2005; Baker et al, 2005; Grasso Richiardi et al, 2014), parental height (Richiardi
et  al, 2016). Results are consistent for number et  al, 2014), and height-​related polymorphisms
of children and reported infertility and less so (Cook et  al, 2011a). One study reported that
for other indicators, such as time to pregnancy. height already at age 9 could be associated with
These studies may be affected by recall bias, and the risk of testicular cancer, although this result
it is not always clear whether analyses consid- was based on stature compared to peers, self-​
ered a lag time between cancer diagnosis and the reported in adulthood (Richiardi et al, 2014).
assessment of fertility to minimize the potential The association between height and the risk
for reverse causation. of testicular cancer could be explained by sev-
A number of cohort studies have fol- eral mechanisms, including shared intrauterine
lowed up men evaluated for fertility problems determinants or shared genetic and hormonal
(Jacobsen et al, 2000b; Raman et al, 2005; Walsh factors, but epidemiological evidence suggests
et al, 2009; Eisenbergh et al, 2015; Hanson et al, that environmental factors affecting growth
2015). These studies have the strength of hav- in infancy and childhood—​for example nutri-
ing prospectively collected information on fer- tion, infections, and chronic diseases—​are more
tility status. The risk of testicular cancer was important than genetic or prenatal factors.
consistently increased, particularly among men
with abnormal semen analyses. However, in Body Mass Index
most of the studies, the increased risk of testic- Body mass index does not appear to be associ-
ular cancer attenuated substantially with time ated with testicular cancer risk, although a weak
since inclusion in the cohort, suggesting some inverse relationship cannot be excluded (Lerro
role of reverse causation and/​or detection bias. et al, 2010).
534

534 Textbook of Cancer Epidemiology

Infections has emerged as a strong candidate risk factor

Based on epidemiological similarities with (Pukkala et  al, 2009; McGlynn & Trabert, 2012;
Hodgkin lymphoma, it has been suggested that Béranger et  al, 2013). Most of these studies are,
testicular cancer may be associated with infec- however, limited by the difficulty in focusing on
tion at later ages (Newell et  al, 1984). An ele- a relevant window of exposure for testicular can-
vated risk of testicular cancer observed among cer etiology, namely, adolescence and very early
patients with acquired immunodeficiency syn- adulthood.
drome (AIDS) (Goedert et  al, 2007; Grulich Evidence is less weak for firefighting and
et al, 2007) and possibly among renal transplant exposure to pesticides. Firefighting has been
patients (Grulich et al, 2007) provides support to classified by the International Agency of
an etiological involvement of the immune system Research on Cancer (IARC) as possibly carcino-
and infections. Most of the studies conducted on genic (Group 2B), also considering the evidence
the role of infections in testicular cancer etiology emerging from six studies on testicular cancer
are, however, null. that were meta-​analyzed by the working group
Although there were initial suggestions of (IARC, 2010). The mechanism and/​ or agents
an association of orchitis and mumps orchi- underlying this association are, nevertheless,
tis with the risk of testicular cancer, over- unknown, and two large studies on firefighters
all evidence suggests a lack of an association did not report an excess risk of testicular cancer
(Trabert et al, 2012). Studies on common child- (Daniels et al, 2014; Pukkala et al, 2014).
hood infections, including rubella, mumps, Evidence does not suggest that farming is
chicken pox, pertussis, and measles, have in associated with the risk of testicular cancer
general found no association with testicu- (Acquavella et al, 1998; Pukkala et al, 2009;
lar cancer risk (Henderson et  al 1979; Brown Béranger et al, 2013). Farmers, however, include
et al, 1987; Swerdlow et al, 1987b; Prener et al, a heterogeneous group of workers typically
1992; United Kingdom Testicular Cancer Study at low cancer risk (Pukkala et al, 2009). Some
Group, 1994b; Petridou et al, 1997; Trabert et al, cohort studies of pesticide users have reported
2012). Two studies have reported that men with an increased risk of testicular cancer, but results
a positive history of sexually transmitted dis- are not entirely consistent (Dich et al, 1996;
eases were at increased risk of testicular can- Kristensen et al, 1996; Zhong & Rafnsson, 1996;
cer (United Kingdom Testicular Cancer Study Fleming et al, 1999; MacFarlane et al, 2009;
Group, 1994b; Gallagher et al, 1995). Koutros et al, 2010; MacFarlane et al, 2010;
One case-​control study on viral infections Frost et al, 2011; Silver et al, 2015). An analy-
and testicular cancer using serum samples drawn sis carried out on 43 chemical agents assessed
before the onset of disease did not find clear evi- through a job-​exposure matrix in Finland found
dence of an association with seroactivity against some evidence of an association of testicular
Epstein Barr virus, cytomegalovirus (Akre et al, cancer with exposure to pesticides—​as well as
1999), or human parvovirus B19 (Tolfvenstam textile dust and some organic solvents (Guo et
et al, 2002). al, 2005).
Some nonoccupational studies on testicular
Ionizing Radiation cancer have analyzed serum levels of organo-
Little research has looked at the association chlorine pesticides, reporting conflicting results
between ionizing radiation and the risk of tes- (Cook et al, 2011b; Giannandrea et al, 2011). Two
ticular cancer. Occupational studies do not sug- studies with prediagnostic serum samples have
gest an association (Yousif et  al, 2010), while it suggested an increased risk of testicular cancer
is difficult to estimate the risk of second primary in association with higher levels of p,p′-​DDE, as
testicular cancer in nontesticular cancer patients well as chlordane and its derivatives (McGlynn
treated with radiotherapy, as most nontesticular et al, 2008; Purdue et al, 2009).
cancers are treated at older ages. Apart from firefighting and use of pesticides,
military personnel have been repeatedly studied
Occupation in relation to testicular cancer risk. There is some
A number of studies have reported an associa- suggestion of a possible increased risk of testicu-
tion between occupations and the risk of testic- lar cancer associated with working on aircraft or
ular cancer, yet initial findings have seldom been operating aircraft (Garland et al, 1988; Foley et al,
replicated; therefore, no occupational exposure 1995; McGlynn & Trabert, 2012).
53

Testicular Cancer 535

Socioeconomic Status cryptorchidism and hypospadias, contralateral

With few exceptions of an inverse association testicular cancer, family history, and height. While
(Sarfati et  al, 2011), socioeconomic status has analytic epidemiologic research has provided
been positively associated with an increased numerous etiologic clues, many of them remain
risk of testicular cancer (Brown et  al, 1987; tentative. Overwhelming evidence indicates the
Pearce et  al, 1987; Haughey et  al, 1989; Prener fundamental importance of environmental fac-
et  al, 1992). Null findings have been reported tors in the etiology of this enigmatic cancer.
in some studies (Richiardi et al, 2004c; Pukkala Prenatal exposures seem to be instrumen-
and Weiderpass, 2002; Schmeisser et  al, 2013; tal in shaping the risk of testicular cancer, but
Richiardi et al, 2014). postnatal exposures acting in childhood, adoles-
cence, and very early adulthood are also impor-
Other Risk Factors tant. Testicular cancer has also a strong genetic
There is strong evidence that birth order and sib- component that is studied through international
ship size are inversely associated with the risk of collaborations and GWAS.
testicular cancer, while testicular trauma and vas- In order to further explain geographic, eth-
ectomy are unlikely to be risk factors. nic, and temporal patterns, we must consider
exposures that differ markedly even within lim-
Birth Order and Sibship Size ited geographic areas that strike whites much
The association between birth order and testic- more than blacks in the United States, and that
ular cancer risk has been investigated in a great have become increasingly prevalent over the last
number of studies, which consistently reported century. To identify factors that tally with this
an inverse association (Cook et  al, 2009; Bevier pattern is a formidable and exciting scientific
et al, 2011). Similar results of an inverse associ- challenge. None of the risk factors so far seriously
ation have been reported for sibship size (Cook considered appear to be promising candidates
et  al, 2009; Bevier et  al, 2011). Although birth to explain the occurrence pattern of testicular
order and sibship size are strongly correlated, two cancer. Studies will have to carefully define rele-
studies found no effect of birth order on testic- vant windows of susceptibility and consider that
ular cancer risk within each category of sibship prenatal and postnatal exposures may interact.
size (Moller & Skakkebaek, 1996; Richiardi et al, Analytical studies based on international col-
2004c). It is relevant to distinguish between these laborations to compare risk factors in different
two variables, as sibship size might be a proxy countries, reach an adequate sample size, and
for parental fertility while birth order might be solve inconsistencies between previous findings
a proxy for prenatal hormonal status or age at will be highly informative.
exposure to specific micro-​organisms.

Trauma REFERENCES
Case-​control studies have reported a positive Abbate V, Kicman AT, Evans-​ Brown M, McVeigh
association between testicular trauma and the J, Cowan DA, Wilson C et  al. Anabolic steroids
risk of testicular cancer, but most likely this asso- detected in bodybuilding dietary supplements—​a
ciation can be attributed to recall bias (United significant risk to public health. Drug Test Anal
Kingdom Testicular Cancer Study Group, 1994a; 2015;7:609–​18.
Akre O, Lipworth L, Tretli S, Linde A, Engstrand L,
Merzenich et al, 2000).
Adami HO et  al. Epstein-​Barr virus and cyto-
megalovirus in relation to testicular-​ cancer
Vasectomy
risk: a nested case-​ control study. Int J Cancer
Vasectomy has convincingly been ruled out as a
1999;82:1–​5.
potential risk factor for testicular cancer (Moss Akre O, Pettersson A, Richiardi L. Risk of contralat-
et  al, 1986; Moller et  al, 1994; United Kingdom eral testicular cancer among men with unilaterally
Testicular Cancer Study Group, 1994a). undescended testis: a meta analysis. Int J Cancer
2009;124:687–​9.
CONCLUSION Akre O, Richiardi L. Does a testicular dysgenesis syn-
A summary of the current state of knowledge for drome exist? Hum Reprod 2009;24:2053–​60.
testicular cancer risk factors is presented in Table Acquavella J, Olsen G, Cole P, Ireland B, Kaneene J,
21-​2. Only few risk factors for testicular can- Schuman S et al. Cancer among farmers: a meta-​
cer are established, including age, ethnic group, analysis. Ann Epidemiol 1998;8:64–​74.
536

536 Textbook of Cancer Epidemiology

Banks K, Tuazon E, Berhane K, Koh CJ, De Filippo RE, Cook MB, Akre O, Forman D, Madigan MP, Richiardi
Chang A et al. Cryptorchidism and testicular germ L, McGlynn KA. A systematic review and meta-​
cell tumors:  comprehensive meta-​analysis reveals analysis of perinatal variables in relation to the risk
that association between these conditions dimin- of testicular cancer: experiences of the mother. Int
ished over time and is modified by clinical charac- J Epidemiol 2009;38:1532–​42.
teristics. Front Endocrinol (Lausanne) 2013;3:182. Cook MB, Akre O, Forman D, Madigan MP, Richiardi
Baker JA, Buck GM, Vena JE, Moysich KB. Fertility L, McGlynn KA. A systematic review and meta-​
patterns prior to testicular cancer diagnosis. analysis of perinatal variables in relation to the
Cancer Causes Control 2005;16:295–​9. risk of testicular cancer:  experiences of the son.
Béranger R, Le Cornet C, Schüz J, Fervers B. Int J Epidemiol 2010;39:1605–​18.
Occupational and environmental exposures asso- Cook MB, Chia VM, Berndt SI, Graubard BI, Chanock
ciated with testicular germ cell tumours: system- SJ, Rubertone MV et al. Genetic contributions to
atic review of prenatal and life-​long exposures. the association between adult height and testicular
PLoS One 2013;8:e77130. germ cell tumors. Int J Epidemiol 2011a;40:731–​9.
Bergström R, Adami HO, Mohner M, Zatonski W, Cook MB, Trabert B, McGlynn KA. Organochlorine
Storm H, Ekbom A et  al. Increase in testicular compounds and testicular dysgenesis syn-
cancer incidence in six European countries:  a drome: human data. Int J Androl 2011b;34:e68–​84.
birth cohort phenomenon. J Natl Cancer Inst Cools M, Drop SL, Wolffenbuttel KP, Oosterhuis JW,
1996;88:727–​33. Looijenga LH. Germ cell tumors in the intersex
Bevier M, Weires M, Thomsen H, Sundquist J, gonad:  old paths, new directions, moving fron-
Hemminki K. Influence of family size and birth tiers. Endocr Rev 2006;27:468–​84.
order on risk of cancer: a population-​based study. Coupland CAC, Forman D, Chilvers CED, Davey G,
BMC Cancer 2011;11:163. Pike MC, Oliver RTD. Maternal risk factors for
Boldajipour B, Raz E. What is left behind-​ quality testicular cancer:  a population-​based case-​control
control in germ cell migration. Sci STKE study (UK). Cancer Causes Control 2004;15:277–​83.
2007;2007:pe16. Crockford GP, Linger R, Hockley S, Dudakia D,
Bonner MR, McCann SE, Moysich KB. Dietary fac- Johnson L, Huddart R et al. Genome-​wide linkage
tors and the risk of testicular cancer. Nutr Cancer screen for testicular germ cell tumour susceptibil-
2002;44:35–​43. ity loci. Hum Mol Genet 2006;15:443–​51.
Bray F, Richiardi L, Ekbom A, Forman D, Pukkala E, Crump C, Sundquist K, Winkleby MA, Sieh W,
Cuminkova M et al. Do testicular seminoma and Sundquist J. Gestational age at birth and risk of
nonseminoma share the same etiology? Evidence testicular cancer. Int J Cancer 2012;131:446–​51.
from an age-​period-​cohort analysis of incidence Daling JR, Doody DR, Sun X, Trabert BL, Weiss NS,
trends in eight European countries. Cancer Chen C et al. Association of marijuana use and the
Epidemiol Biomarkers Prev 2006;15:652–​8. incidence of testicular germ cell tumors. Cancer
Brown LM, Pottern LM, Hoover RN. Prenatal and 2009;115:1215–​23.
perinatal risk factors for testicular cancer. Cancer Daniels RD, Kubale TL, Yiin JH, Dahm MM, Hales
Res 1986;46:4812–​6. TR, Baris D et  al. Mortality and cancer inci-
Brown LM, Pottern LM, Hoover RN. Testicular can- dence in a pooled cohort of US firefighters from
cer in young men: the search for causes of the epi- San Francisco, Chicago and Philadelphia (1950–​
demic increase in the United States. J Epidemiol 2009). Occup Environ Med 2014;71:388–​97.
Commun Health 1987;41:349–​54. Davies TW, Palmer CR, Ruja E, Lipscombe JM.
Chen L, Cnattingius S, Nyman Iliadou A, Oberg AS. Adolescent milk, dairy product and fruit con-
Cancer risks in twins and singletons from twin and sumption and testicular cancer. Br J Cancer
non-​twin families. Int J Cancer 2016;138:1102–​10. 1996;74:657–​60.
Chung CC, Kanetsky PA, Wang Z, Hildebrandt MA, Depue RH, Bernstein L, Ross RK, Judd HL, Henderson
Koster R, Skotheim RI et al. Meta-​analysis identi- BE. Hyperemesis gravidarum in relation to estra-
fies four new loci associated with testicular germ diol levels, pregnancy outcome, and other mater-
cell tumor. Nat Genet 2013;45:680–​5. nal factors:  a seroepidemiologic study. Am J
Clemmesen J. Is pregnancy smoking causal to tes- Obstet Gynecol 1987;156:1137–​41.
tis cancer in sons? A  hypothesis. Acta Oncol Depue RH, Pike MC, Henderson BE. Estrogen expo-
1997;36:59–​63. sure during gestation and risk of testicular cancer.
Cook MB, Zhang Y, Graubard BI, Rubertone MV, J Natl Cancer Inst 1983;71:1151–​5.
Erickson RL, McGlynn KA. Risk of testicular Dich J, Wiklund K, Holm LE. Testicular cancer in pes-
germ-​cell tumours in relation to childhood physi- ticide applicators in Swedish agriculture. Scand J
cal activity. Br J Cancer 2008;98:174–​8. Work Environ Health 1996;22:66.
537

Testicular Cancer 537

Dong C, Lönnstedt I, Hemminki K. Familial testicular (Alberta and British Columbia, Canada). Cancer
cancer and second primary cancers in testicular Causes Control 1995;6:398–​406.
cancer patients by histological type. Eur J Cancer Ganmaa D, Li XM, Wang J, Qin LQ, Wang PY, Sato A.
2001;15:1878–​85. Incidence and mortality of testicular and prostatic
Doria-​Rose VP, Biggs ML, Weiss NS. Subfertility and cancers in relation to world dietary practices. Int J
the risk of testicular germ cell tumors (United Cancer 2002;98:262–​7.
States). Cancer Causes Control 2005;16:651–​6. Garland FC, Gorham ED, Garland CF, Ducatman
Dusek L, Abrahamova J, Lakomy R, Vyzula R, AM. Testicular cancer in U.S. Navy personnel. Am
Koptikova J, Pavlik T et  al. Multivariate analysis J Epidemiol 1988;127:411–​4.
of risk factors for testicular cancer:  a hospital-​ Garner MJ, Birkett NJ, Johnson KC, Shatenstein
based case-​control study in the Czech Republic. B, Ghadirian P, Krerwski D, Canadian Cancer
Neoplasma 2008;55:356–​68. Registries Epidemiology Research Group. Dietary
Eisenberg ML, Li S, Brooks JD, Cullen MR, Baker LC. risk factors for testicular carcinoma. Int J Cancer
Increased risk of cancer in infertile men: analysis 2003;106:934–​41.
of U.S. claims data. J Urol 2015;193:1596–​601. Ghazarian AA, Trabert B, Devesa SS, McGlynn KA.
Ekbom A, Richardi L, Akre O, Montgomery S, Sparén Recent trends in the incidence of testicular germ
P. Testicular cancer risk among Finnish immi- cell tumors in the United States. Andrology
grants:  Perinatal influences. J Natl Cancer Inst 2015;3:13–​8.
2003;95:1238–​40. Giannandrea F, Gandini L, Paoli D, Turci R, Figà-​
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser Talamanca I. Pesticide exposure and serum orga-
S, Mathers C, Rebelo M, Parkin DM, Forman nochlorine residuals among testicular cancer
D, Bray, F. GLOBOCAN 2012 v1.0, Cancer patients and healthy controls. J Environ Sci Health
Incidence and Mortality Worldwide: IARC B 2011;46:780–​7.
CancerBase No. 11 [Internet]. Lyon, France: Giannandrea F, Paoli D, Lombardo F, Lenzi A, Gandini
International Agency for Research on Cancer; L. Case-​control study of anthropometric meas-
2013. Available from: http://globocan.iarc.fr, ures and testicular cancer risk. Front Endocrinol
accessed 2017. (Lausanne) 2012;3:144.
Ferlin A, Pengo M, Pizzol D, Carraro U, Frigo AC, Goedert JJ, Purdue MP, McNeel TS, McGlynn KA,
Foresta C. Variants in KITLG predispose to tes- Engels EA. Risk of germ cell tumors among
ticular germ cell cancer independently from men with HIV/​ acquired immunodeficiency
spermatogenic function. Endocr Relat Cancer syndrome. Cancer Epidemiol Biomarkers Prev
2012;19:101–​8. 2007;16:1266–​9.
Fleming LE, Bean JA, Rudolph M, Hamilton K. Grasso C, Zugna D, Fiano V, Robles Rodriguez N,
Cancer incidence in a cohort of licensed pesti- Maule M, Gillio-​Tos A, et al. Subfertility and Risk
cide applicators in Florida. J Occup Environ Med of Testicular Cancer in the EPSAM Case-​Control
1999;41:279–​88. Study. PLoS One 2016;11:e0169174.
Foley S, Middleton S, Stitson D, Mahoney M. The inci- Grulich AE, McCredie M, Coates M. Cancer inci-
dence of testicular cancer in Royal Air Force per- dence in Asian migrants to New South Wales,
sonnel. Br J Urol 1995;76:495–​6. Australia. Br J Cancer 1995;71:400–​8.
Forman D, Oliver RT, Brett AR, Marsh SG, Moses Grulich AE, van Leeuwen MT, Falster MO, Vajdic
JH, Bodmer JG et al. Familial testicular cancer: a CM. Incidence of cancers in people with HIV/​
report of the UK family register, estimation of risk AIDS compared with immunosuppressed
and an HLA class 1 sib-​pair analysis. Br J Cancer transplant recipients:  a meta-​ analysis. Lancet
1992;65:255–​62. 2007;370:59–​67.
Fossa SD, Kravdal O. Fertility in Norwegian testicular Guo J, Pukkala E, Kyyrönen P, Lindbohm ML, Heikkilä
cancer patients. Br J Cancer 2000;82:737–​41. P, Kauppinen T. Testicular cancer, occupation and
Frost G, Brown T, Harding AH. Mortality and can- exposure to chemical agents among Finnish men.
cer incidence among British agricultural pesticide Cancer Causes Control 2005;16:97–​103.
users. Occup Med (Lond) 2011;61:303–​10. Gurney J, Shaw C, Stanley J, Signal V, Sarfati D.
Galan JJ, De Felici M, Buch B, Rivero MC, Segura Cannabis exposure and risk of testicular can-
A, Royo JL et  al. Association of genetic markers cer: a systematic review and meta-​analysis. BMC
within the KIT and KITLG genes with human Cancer 2015;15:897.
male infertility. Hum Reprod 2006;21:3185–​92. Hanson HA, Anderson RE, Aston KI, Carrell DT,
Gallagher RP, Huchcroft S, Phillips N, Hill GB, Smith KR, Hotaling JM. Subfertility increases risk
Coldman AJ, Coppin C et  al. Physical activ- of testicular cancer:  evidence from population-​
ity, medical history, and risk of testicular cancer based semen samples. Fertil Steril 2015;
S0015-​0282:02038-​5.
538

538 Textbook of Cancer Epidemiology

Haughey BP, Graham S, Brasure J, Zielezny M, Sufrin and women in Norwegian agriculture. Scand J
G, Burnett WS. The epidemiology of testicu- Work Environ Health 1996;22:14–​26.
lar cancer in upstate New York. Am J Epidemiol Kristiansen W, Karlsson R, Rounge TB, Whitington
1989;130:25–​36. T, Andreassen BK, Magnusson PK et al. Two new
Heimdal K, Olsson H, Tretli S, Flodgren P, Borresen loci and gene sets related to sex determination
AL, Fossa SD. Familial testicular cancer in Norway and cancer progression are associated with sus-
and southern Sweden. Br J Cancer 1996;73:964–​9. ceptibility to testicular germ cell tumor. Hum Mol
Hemminki K, Li X. Cancer risks in Nordic immi- Genet 2015;24:4138–​46.
grants and their offspring in Sweden. Eur J Cancer Kuijper EA, Twisk JW, Korsen T, Caanen MR, Kushnir
2002;38:2428–​34. MM, Rockwood AL et al. Mid-​pregnancy, perinatal,
Henderson BE, Benton B, Jing J, Yu MC, Pike MC. and neonatal reproductive endocrinology:  a pro-
Risk factors for cancer of the testis in young men. spective cohort study in twins and singleton control
Int J Cancer 1979;23:598–​602. subjects. Fertil Steril 2015;104:1527–​1534.e9.
Herrinton LJ, Zhao W, Husson G. Management of Lacson JC, Carroll JD, Tuazon E, Castelao EJ,
cryptorchidism and risk of testicular cancer. Am Bernstein L, Cortessis VK. Population-​ based
J Epidemiol 2003;157:602–​5. case-​control study of recreational drug use and
Hickey M, Hart R, Keelan JA. The relationship testis cancer risk confirms an association between
between umbilical cord estrogens and perina- marijuana use and nonseminoma risk. Cancer
tal characteristics. Cancer Epidemiol Biomarkers 2012;118:5374–​83.
Prev 2014;23:946–​52. Le Cornet C, Fervers B, Oksbjerg Dalton S, Feychting
Hu J, La Vecchia C, de Groh M, Negri E, Morrison H, M, Pukkala E, Tynes T et al. Testicular germ cell
Mery L et al. Dietary cholesterol intake and can- tumours and parental occupational exposure to
cer. Ann Oncol 2012;23:491–​500. pesticides:  a register-​ based case-​control study
International Agency for Research on Cancer. in the Nordic countries (NORD-​ TEST study).
Painting, Firefighting, and Shiftwork. IARC Occup Environ Med 2015;72:805–​11.
Monograph on the Evaluation of the Carcinogenic Lerro CC, McGlynn KA, Cook MB. A systematic
Risks to Humans, Vol. 98. Lyon: IARC; 2010. review and meta-​ analysis of the relationship
Jacobsen R, Bostofte E, Engholm G, Hansen J, between body size and testicular cancer. Br J
Skakkebaek NE, Moller H. Fertility and off- Cancer 2010;103:1467–​74.
spring sex ratio of men who develop testicu- Lessel D, Gamulin M, Kulis T, Toliat MR, Grgic M,
lar cancer:  a record linkage study. Hum Reprod Friedrich K, et  al. Replication of genetic sus-
2000a;15:1958–​61. ceptibility loci for testicular germ cell cancer
Jacobsen R, Bostofte E, Engholm G, Hansen J, Olsen in the Croatian population. Carcinogenesis
JH, Skakkebaek NE et  al. Risk of testicular can- 2012;33:1548–​52.
cer in men with abnormal semen characteris- Levine H, Afek A, Shamiss A, Derazne E, Tzur D, Zavdy
tics: cohort study. Br Med J 2000b;321:789–​92. O et al. Risk of germ cell testicular cancer accord-
Kanetsky PA, Mitra N, Vardhanabhuti S, Li M, Vaughn ing to origin: a migrant cohort study in 1,100,000
DJ, Letrero R et al. Common variation in KITLG Israeli men. Int J Cancer 2013;132:1878–​85.
and at 5q31.3 predisposes to testicular germ cell Li N, Hauser R, Holford T, Zhu Y, Zhang Y, Bassig
cancer. Nat Genet 2009;41:811–​5. BA et  al. Muscle-​building supplement use and
Kharazmi E, Hemminki K, Pukkala E, Sundquist K, increased risk of testicular germ cell cancer in
Tryggvadottir L, Tretli S et al. Cancer risk in rela- men from Connecticut and Massachusetts. Br J
tives of testicular cancer patients by histology type Cancer 2015;112:1247–​50.
and age at diagnosis: a joint study from five Nordic Litchfield K, Holroyd A, Lloyd A, Broderick P,
countries. Eur Urol 2015;68:283–​9. Nsengimana J, Eeles R et al. Identification of four
Koutros S, Alavanja MC, Lubin JH, Sandler DP, new susceptibility loci for testicular germ cell
Hoppin JA, Lynch CF et  al. An update of can- tumour. Nat Commun 2015a;6:8690.
cer incidence in the Agricultural Health Study. J Litchfield K, Shipley J, Turnbull C. Common variants
Occup Environ Med 2010;52:1098–​105. identified in genome-​wide association studies of
Kristensen DG, Skakkebæk NE, Rajpert-​De Meyts testicular germ cell tumour:  an update, biologi-
E, Almstrup K. Epigenetic features of testicu- cal insights and clinical application. Andrology
lar germ cell tumours in relation to epigenetic 2015b;3:34–​46.
characteristics of foetal germ cells. Int J Dev Biol Littman AJ, Doody DR, Biggs ML, Weiss NS, Starr JR,
2013;57:309–​17. Schwartz SM. Physical activity in adolescence and
Kristensen P, Andersen A, Irgens LM, Laake P, Bye AS. testicular germ cell cancer risk. Cancer Causes
Incidence and risk factors of cancer among men Control 2009;20:1281–​90.
539

Testicular Cancer 539

Lutke Holzik MF, Sijmons RH, Sleijfer DT, Sonneveld status and related factors: case-​control studies in
DJ, Hoekstra-​Weebers JE, van Echten-​Arends J, Denmark. Int J Cancer 1996;66:287–​93.
Hoekstra HJ. Syndromic aspects of testicular car- Moller H, Skakkebaek NE. Risk of testicular cancer
cinoma. Cancer 2003;97:984–​92. in subfertile men:  case-​control study. Br Med J
MacFarlane E, Benke G, Del Monaco A, Sim MR. 1999;318:559–​62.
Cancer incidence and mortality in a historical Mongraw-​Chaffin ML, Cohn BA, Anglemyer AT,
cohort of Australian pest control workers. Occup Cohen RD, Christianson RE. Maternal smok-
Environ Med 2009;66:818–​23. ing, alcohol, and coffee use during pregnancy
MacFarlane E, Benke G, Del Monaco A, Sim MR. and son’s risk of testicular cancer. Alcohol
Causes of death and incidence of cancer in a 2009;43:241–​5.
cohort of Australian pesticide-​exposed workers. Moss AR, Osmond D, Bacchetti P, Torti FM, Gurgin V.
Ann Epidemiol 2010;20:273–​80. Hormonal risk factors in testicular cancer: a case-​
Martin OV, Shialis T, Lester JN, Scrimshaw MD, control study. Am J Epidemiol 1986;124:39–​52.
Boobis AR, Voulvoulis N. Testicular dysgenesis Myrup C, Westergaard T, Schnack T, Oudin A, Ritz
syndrome and the estrogen hypothesis:  a quan- C, Wohlfahrt J et al. Testicular cancer risk in first-​
titative meta-​analysis. Environ Health Perspect and second-​generation immigrants to Denmark. J
2008;116:149–​57. Natl Cancer Inst 2008;100:41–​7.
Maule M, Malavassi JL, Richiardi L. Age at puberty Nathanson KL, Kanetsky PA, Hawes R, Vaughn DJ,
and risk of testicular cancer: a meta-​analysis. Int J Letrero R, Tucker K et al. The Y deletion gr/​gr and
Androl 2012;35:828–​34. susceptibility to testicular germ cell tumor. Am J
McGlynn KA, Devesa SS, Graubard BI, Castle PE. Hum Genet 2005;77:1034–​43.
Increasing incidence of testicular germ cell Neale RE, Carrière P, Murphy MF, Baade PD.
tumors among black men in the United States. J Testicular cancer in twins:  a meta-​analysis. Br J
Clin Oncol 2005;23:5757–​61. Cancer 2008;98:171–​3.
McGlynn KA, Quraishi SM, Graubard BI, Weber JP, Newell GR, Mills PK, Johnson DE. Epidemiologic com-
Rubertone MV, Erickson RL. Persistent organo- parison of cancer of the testis and Hodgkin’s disease
chlorine pesticides and risk of testicular germ cell among young males. Cancer 1984;54:1117–​23.
tumors. J Natl Cancer Inst 2008;100:663–​71. Paffenbarger RS Jr, Kampert JB, Lee IM. Physical activ-
McGlynn KA, Sakoda LC, Rubertone MV, Sesterhenn ity and health of college men: longitudinal obser-
IA, Lyu C, Graubard BI, Erickson RL. Body size, vations. Int J Sports Med. 1997;18 Suppl 3:S200–​3.
dairy consumption, puberty, and risk of testicular Pearce N, Sheppard RA, Howard JK, Fraser J, Lilley
germ cell tumors. Am J Epidemiol 2007;165:355–​63. BM. Time trends and occupational differences
McGlynn KA, Trabert B. Adolescent and adult in cancer of the testis in New Zealand. Cancer
risk factors for testicular cancer. Nat Rev Urol 1987;59:1677–​82.
2012;9:339–​49. Petridou E, Roukas KI, Dessypris N, Aravantinos G,
McGlynn KA, Zhang Y, Sakoda LC, Rubertone MV, Bafaloukos D, Efraimidis A et  al. Baldness and
Erickson RL, Graubard BI. Maternal smoking and other correlates of sex hormones in relation to tes-
testicular germ cell tumors. Cancer Epidemiol ticular cancer. Int J Cancer 1997;71:982–​5.
Biomarkers Prev 2006;15:1820–​24. Pettersson A, Akre O, Richiardi L, Ekbom A, Kaijser
Merzenich H, Ahrens W, Stang A, Baumgardt-​Elms C, M. Maternal smoking and the epidemic of tes-
Jahn I, Stegmaier C et  al. Sorting the hype from ticular cancer:  a nested case-​control study. Int J
the facts in testicular cancer:  is testicular cancer Cancer 2007a;120:2044–​6.
related to trauma? J Urol 2000;164:2143–​4. Pettersson A, Kaijser M, Richiardi L, Ekbom A,
Moirano G, Zugna D, Grasso C, Lista P, Ciuffreda Akre O. Women smoking and testicular can-
L, Segnan N et  al. Baldness and testicular can- cer:  one epidemic causing another? Int J Cancer
cer:  the EPSAM case-​control study. Andrology 2004;109:941–​4.
2016;4:251–​6. Pettersson A, Richiardi L, Cnattingius S, Kaijser M,
Moller H, Knudsen LB, Lynge E. Risk of testicular can- Akre O. Gestational hypertension, preeclamp-
cer after vasectomy:  cohort study of over 73,000 sia, and risk of testicular cancer. Cancer Res
men. Br Med J 1994;309:295–​9. 2008;68:8832–​6.
Moller H, Prener A, Skakkebaek NE. Testicular cancer, Pettersson A, Richiardi L, Nordenskjold A, Kaijser
cryptorchidism, inguinal hernia, testicular atrophy, M, Akre O. Age at surgery for undescended tes-
and genital malformations: case-​control studies in tis and risk of testicular cancer. N Engl J Med
Denmark. Cancer Causes Control 1996;7:264–​74. 2007b;356:1835–​41.
Moller H, Skakkebaek NE. Risks of testicular cancer Pochi PE. Acne:  endocrinologic aspects. Cutis
and cryptorchidism in relation to socioeconomic 1982;30:212–​4.
540

540 Textbook of Cancer Epidemiology

Prener A, Engholm G, Jensen OM. Genital anoma- Rijlaarsdam MA, van Agthoven T, Gillis AJ, Patel
lies and risk for testicular cancer in Danish men. S, Hayashibara K, Lee KY et  al. Identification
Epidemiology 1996;7:14–​9. of known and novel germ cell cancer-​ specific
Prener A, Hsieh CC, Engholm G, Trichopoulos D, (embryonic) miRs in serum by high-​throughput
Jensen OM. Birth order and risk of testicular can- profiling. Andrology 2015;3:85–​91.
cer. Cancer Causes Control 1992;3:265–​72. Sarfati D, Shaw C, Blakely T, Atkinson J, Stanley J.
Pukkala E, Martinsen JI, Lynge E, Gunnarsdottir HK, Ethnic and socioeconomic trends in testicular
Sparén P, Tryggvadottir L et  al. Occupation and cancer incidence in New Zealand. Int J Cancer
cancer:  follow-​up of 15  million people in five 2011;128:1683–​91.
Nordic countries. Acta Oncol 2009;48:646–​790. Schmeisser N, Conway DI, Stang A, Jahn I, Stegmaier
Pukkala E, Martinsen JI, Weiderpass E, Kjaerheim C, Baumgardt-​ Elms C et  al. A population-​
K, Lynge E, Tryggvadottir L et  al. Cancer inci- based case-​control study on social factors and
dence among firefighters:  45  years of follow-​ up risk of testicular germ cell tumours. BMJ Open
in five Nordic countries. Occup Environ Med 2013;3:e003833.
2014;71:398–​404. Schnack TH, Poulsen G, Myrup C, Wohlfahrt J,
Pukkala E, Weiderpass E. Socio-​ economic differ- Melbye M. Familial coaggregation of cryptorchi-
ences in incidence rates of cancers of the male dism, hypospadias, and testicular germ cell can-
genital organs in Finland, 1971–​95. Int J Cancer cer: a nationwide cohort study. J Natl Cancer Inst
2002;102:643–​8. 2010;102:187–​92.
Purdue MP, Engel LS, Langseth H, Needham LL, Schumacher FR, Wang Z, Skotheim RI, Koster R,
Andersen A, Barr DB et al. Prediagnostic serum Chung CC, Hildebrandt MA et  al. Testicular
concentrations of organochlorine compounds and germ cell tumor susceptibility associated with
risk of testicular germ cell tumors. Environ Health the UCK2 locus on chromosome 1q23. Hum Mol
Perspect 2009;117:1514–​9. Genet 2013;22:2748–​53.
Raman JD, Nobert CF, Goldstein M. Increased inci- Sigurdson AJ, Chang S, Annegers JF, Duphoren CM,
dence of testicular cancer in men presenting with Pillow PC, Amato RJ et  al. A case-​control study
infertility and abnormal semen analysis. J Urol of diet and testicular carcinoma. Nutr Cancer
2005;174:1819–​22. 1999;34:20–​6.
Raymond CS, Shamu CE, Shen MM, Seifert KJ, Silver SR, Bertke SJ, Hines CJ, Alavanja MC, Hoppin
Hirsch B, Hodgkin J et  al. Evidence for evolu- JA, Lubin JH et  al. Cancer incidence and meto-
tionary conservation of sex-​determining genes. lachlor use in the Agricultural Health Study:  an
Nature 1998;391:691–​5. update. Int J Cancer 2015;137:2630–​43.
Richiardi L, Akre O. Fertility among brothers of Skakkebaek NE. Possible carcinoma-​in-​situ of the tes-
patients with testicular cancer. Cancer Epidemiol tis. Lancet 1972;2:516–​7.
Biomarkers Prev 2005;14:2557–​62. Skakkebaek NE, Berthelsen JG, Giwercman A, Müller
Richiardi L, Akre O, Lambe M, Granath F, J. Carcinoma-​in-​situ of the testis: possible origin
Montgomery SM, Ekbom A. Birth order, sib- from gonocytes and precursor of all types of germ
ship size, and risk for germ-​cell testicular cancer. cell tumours except spermatocytoma. Int J Androl
Epidemiology 2004c;15:323–​9. 1987;10:19–​28.
Richiardi L, Akre O, Montgomery SM, Lambe M, Kvist Skakkebaek NE, Rajpert-​ De Meyts E, Buck Louis
U, Ekbom A. Fecundity and twinning rates as meas- GM, Toppari J, Andersson AM, Eisenberg ML
ures of fertility before diagnosis of germ-​cell testic- et  al. Male reproductive disorders and fertility
ular cancer. J Natl Cancer Inst 2004b;96:145–​7. trends: influences of environment and genetic sus-
Richiardi L, Askling J, Granath F, Akre O. Body size at ceptibility. Physiol Rev 2016;96:55–​97.
birth and adulthood and the risk for germ-​cell tes- Skakkebaek NE, Rajpert-​ De Meyts E, Main KM.
ticular cancer. Cancer Epidemiol Biomarkers Prev Testicular dysgenesis syndrome:  an increasingly
2003;12:669–​73. common developmental disorder with environ-
Richiardi L, Bellocco R, Adami H-​ O, Torrång A, mental aspects. Hum Reprod 2001;16:972–​8
Barlow L, Hakulinen T et  al. Testicular cancer Skotheim RI, Lothe RA. The testicular germ cell
incidence in eight Northern European coun- tumour genome. APMIS 2003;111:136–​50.
tries: secular and recent trend. Cancer Epidemiol Sonne SB, Almstrup K, Dalgaard M, Juncker AS,
Biomarkers Prev 2004a;13:2157–​66. Edsgard D, Ruban L et al. Analysis of gene expres-
Richiardi L, Pettersson A, Akre O. Genetic and envi- sion profiles of microdissected cell populations
ronmental risk factors for testicular cancer. Int J indicates that testicular carcinoma in situ is an
Androl 2007;30:230–​40. arrested gonocyte. Cancer Res 2009;69:5241–​50.
Richiardi, L, Vizzini L, Pastore G, Segnan N, Gillio-​ Srivastava A, Kreiger N. Relation of physical activ-
Tos A, Fiano V et al. Lifetime growth and risk of ity to risk of testicular cancer. Am J Epidemiol
testicular cancer. Int J Cancer 2014;135:695–​701. 2000;151:78–​87.
541

Testicular Cancer 541

Srivastava A, Kreiger N. Cigarette smoking and tes- Trabert B, Zugna D, Richiardi L, McGlynn KA, Akre
ticular cancer. Cancer Epidemiol Biomarkers Prev O. Congenital malformations and testicular germ
2004;13:49–​54. cell tumors. Int J Cancer 2013;133:1900–​4.
Stang A, Ahrens W, Baumgardt-​Elms C, Stegmaier Troisi R, Hoover RN, Thadhani R, Hsieh CC, Sluss P,
C, Merzenich H, de Vrese M, Schrezenmeir J, Ballard-​Barbash R et  al. Maternal, prenatal and
Jockel KH. Adolescent milk fat and galactose con- perinatal characteristics and first trimester mater-
sumption and testicular germ cell cancer. Cancer nal serum hormone concentrations. Br J Cancer
Epidemiol Biomarkers Prev 2006;15:2189–​95. 2008;99:1161–​4.
Stang A, Kuss O. Etiologic differences between semi- Tuomisto J, Holl K, Rantakokko P, Koskela P, Hallmans
noma and nonseminoma of the testis: a systematic G, Wadell G et al. Maternal smoking during preg-
review of epidemiologic studies. Hematol Oncol nancy and testicular cancer in the sons: a nested
Clin North Am 2011;25:473–​86. case-​control study and a meta-​ analysis. Eur J
Strohsnitter WC, Noller KL, Hoover RN, Robboy SJ, Cancer 2009;45:1640–​8.
Palmer JR, Titus-​Ernstoff L et  al. Cancer risk in Turnbull C, Rapley EA, Seal S, Pernet D, Renwick A,
men exposed in utero to diethylstilbestrol. J Natl Hughes D et al. Variants near DMRT1, TERT and
Cancer Inst 2001;93:545–​51 ATF7IP are associated with testicular germ cell
Swerdlow AJ, Huttly SR, Smith PG. Prenatal and cancer. Nat Genet 2010;42:604–​7.
familial associations of testicular cancer. Br J United Kingdom Testicular Cancer Study Group.
Cancer 1987a;55:571–​7. Aetiology of testicular cancer:  association with
Swerdlow AJ, Huttly SR, Smith PG. Testicular congenital abnormalities, age at puberty, infertil-
cancer and antecedent diseases. Br J Cancer ity, and exercise. Br Med J 1994a;308:1393–​9.
1987b;55:97–​103. United Kingdom Testicular Cancer Study Group.
Swerdlow AJ, Huttly SRA, Smith PG. Testis can- Social, behavioural and medical factors in the
cer: post-​natal hormonal factors, sexual behaviour aetiology of testicular cancer:  results from the
and fertility. Int J Cancer 1989;43:549–​53. U.K. study. Br J Cancer 1994b;70:513–​20.
Thorup J, McLachlan R, Cortes D, Nation TR, Balic Walsh TJ, Croughan MS, Schembri M, Chan JM,
A, Southwell BR et al. What is new in cryptorchi- Turek PJ. Increased risk of testicular germ cell
dism and hypospadias:  a critical review on the cancer among infertile men. Arch Intern Med
testicular dysgenesis hypothesis. J Pediatr Surg 2009;169:351–​6.
2010;45:2074–​86. Walsh TJ, Dall’Era MA, Croughan MS, Carroll PR,
Thune I, Lund E. Physical activity and the risk of Turek PJ. Prepubertal orchiopexy for cryptorchi-
prostate and testicular cancer:  a cohort study of dism may be associated with lower risk of testicu-
53,000 Norwegian men. Cancer Causes Control lar cancer. J Urol 2007;178:1440–​6.
1994;5:549–​56. Weir HK, Marrett LD, Kreiger N, Darlington GA,
Tolfvenstam T, Papadogiannakis N, Andersen A, Akre Sugar L. Pre-​natal and peri-​natal exposures and
O. No association between human parvovirus risk of testicular germ-​cell cancer. Int J Cancer
B19 and testicular germ cell cancer. J Gen Virol 2000;87:438–​43.
2002;83:2321–​4. Wirehn AB, Tornberg S, Carsten J. Serum cholesterol
Trabert B, Chen J, Devesa SS, Bray F, McGlynn KA. and testicular cancer incidence in 45.000 men fol-
International patterns and trends in testicular lowed for 25 years. Br J Cancer 2005;92:1785–​6.
cancer incidence, overall and by histologic sub- Wirén S, Häggström C, Ulmer H, Manjer J, Bjørge T,
type, 1973–​2007. Andrology 2015;3:4–​12. Nagel G et al. Pooled cohort study on height and
Trabert B, Graubard BI, Erickson RL, McGlynn KA. risk of cancer and cancer death. Cancer Causes
Childhood infections, orchitis and testicular germ Control 2014;25:151–​9.
cell tumours:  a report from the STEED study Yousif L, Blettner M, Hammer GP, Zeeb H. Testicular
and a meta-​analysis of existing data. Br J Cancer cancer risk associated with occupational radiation
2012;106:1331–​4. exposure:  a systematic literature review. J Radiol
Trabert B, Sigurdson AJ, Sweeney AM, Amato Prot 2010;30:389–​406.
RJ, Strom SS, McGlynn KA. Baldness, acne Znaor A, Lortet-​Tieulent J, Laversanne M, Jemal A,
and testicular germ cell tumours. Int J Androl Bray F. International testicular cancer incidence
2011b;34:e59–​67. trends:  generational transitions in 38 countries
Trabert B, Sigurdson AJ, Sweeney AM, Strom SS, 1900–​1990. Cancer Causes Control 2015;26:151–​8.
McGlynn KA. Marijuana use and testicular germ Zhong Y, Rafnsson V. Cancer incidence among Icelandic
cell tumors. Cancer 2011a;117:848–​53. pesticide users. Int J Epidemiol 1996;25:1117–​24.
542
543

22
Urinary Bladder Cancer
MANOLIS KOGEVINAS, JONINE FIGUEROA,
M O N T S E R R AT G A R C I A - C
​ LOSAS, AND LORELEI MUCCI

B ladder cancer is the ninth most common

cancer worldwide (Ferlay et  al, 2013), is
more frequent in men, and is more frequent in
tumors (Ta, carcinoma in situ, or T1) and
muscle-​invasive (MI) tumors (T2–​T4) based on
how far they have invaded into the wall of the
higher-​than lower-​ income countries. Bladder bladder. Among urothelial cell carcinomas of the
cancer occupies an important place in the his- urinary bladder, around 75% present as super-
tory of epidemiology, particularly occupational ficial disease and the remaining as MI, however
epidemiology, in which associations with occu- data suggest that these trends might be changing
pations exposed to aromatic amines were first with increases of low-​grade NMI. Among super-
identified in the 1950s. Bladder cancer is also ficial tumors, most appear as papillary low-​grade
among the first cancers for which an infectious superficial lesions (Ta), and less than 10% as flat
etiology was identified, through a parasitic infec- carcinoma in situ, which is generally high grade.
tion (Schistosoma haematobium). Finally, bladder Symptoms A visible but painless bleeding (hema-
cancer is one of the few examples with consist- turia) is the cardinal symptom of bladder cancer,
ent evidence of interactions between environ- sometimes accompanied by urgency, other void-
mental exposures and genetic polymorphisms in ing problems, or urinary obstruction.
cancer epidemiology. Urothelial cell carcinoma,
Diagnosis Diagnosis is confirmed through visual
also known as transitional cell carcinoma, is the
inspection by a cystoscope, combined with his-
predominant histopathologic type and arises in
topathologic examination of a biopsy specimen
the renal pelvis and ureters as well as in the blad-
or resected tumor tissue. Cytologic examination
der. Therefore, most of the risk factors described
of urine may provide additional information. In
in this chapter for transitional cell cancers of
advanced stages, palpation under general anes-
the bladder are relevant to these sites as well,
thesia is required for local staging, and x-​ray
although the associations may differ in strength.
examination of the upper urinary tract reveals
obstruction of the ureters.
CLINICAL SYNOPSIS Treatment Despite advances in various areas of
Subgroups In high-​ income countries, urothe- clinical medicine, outcomes for patients with
lial cell carcinomas constitute more than 90% bladder cancer have largely remained unchanged.
of malignant tumors. Urothelial cell carcinomas Surgery is the predominant curative treatment.
arise from urothelial cells that line the inside of Depending on the stage of disease, such treat-
the bladder and the urinary tract, such as the ment ranges from transurethral resection to rad-
part of the kidney that connects to the ureter ical cystectomy, followed by various approaches
(called the renal pelvis), the ureters, and the ure- to restore the urinary conduit. Radiotherapy and
thra. Besides urothelial cell carcinomas, 5% to various chemotherapy regimen, including intra-
7% include squamous cell carcinomas, adenocar- vesical chemotherapy and intravesical immuno-
cinomas, undifferentiated carcinomas, and other therapy with Bacillus Calmette-​Guérin (BCG),
rare histologic types such as small-​cell carcino- are used both as an adjuvant to surgical treat-
mas and lymphomas. ment and for palliation in patients with advanced
Bladder cancers are clinically described disease. Occasionally, radiotherapy or chemo-
as superficial or non-​ muscle-​invasive (NMI) therapy alone may be curative.
54

544 Textbook of Cancer Epidemiology

Prognosis In the United States, overall 5-​year rel- treatment of advanced disease. Identification of
ative survival rose from 72% in the mid-​1970s to predictive and prognostic biomarkers that could
83% in cases diagnosed after 1995, however sur- distinguish between superficial tumors likely to
vival has declined slightly to 78% from 2006 to progress into MI tumors, as well as MI tumors of
2012 (SEER). Survival varies widely with stage at better prognosis eligible to be treated only by sur-
diagnosis, with a 70% 5-​year survival for cancers gery, is an important area of research that could
localized to the bladder, 35% for cancers with result in further improvements in the clinical
lymph node invasion, and 4% for metastatic dis- management of bladder cancer. Molecular pro-
ease. Chemotherapy has turned out to be curative filing studies are providing important findings
in a small number of patients even if they have and are likely to lead to the identification of etio-
metastases. Moreover, urinary diversion with logically distinct tumor subtypes as well as novel
continent reservoirs has led to a better quality of clinical biomarkers and targets for treatment.
life for patients receiving radical cystectomy.
DESCRIPTIVE
Molecular Subtyping Clinical and molecular
EPIDEMIOLOGY
characteristics of bladder tumors have identified
The descriptive epidemiology of urinary bladder
at least two distinct pathogenic pathways:  the
cancer reveals interesting patterns of incidence
low-​ grade noninvasive papillary (Ta) pathway
and mortality across populations, which point
normally preceded by hyperplasia and the high-​
to differences in risk factor distributions in these
grade invasive pathway preceded by flat carci-
populations.
noma in situ or dysplasia (Wu, 2005). Somatic
alterations in the FGFR3 and HRAS genes are Incidence
noted to be especially common in the Ta path- There were an estimated 430,000 cases of blad-
way, while structural and functional alterations der cancer occurring worldwide in 2012 (Global
in TP53 and RB are more frequent in the invasive Cancer Observatory [GCO], 2012) with around
pathway. Molecular somatic studies, including 330,000 occurring in men. The incidence of uri-
large-​scale projects such as The Cancer Genome nary bladder cancer is highest in western Europe,
Atlas (TCGA), are identifying somatic alterations North America, and Egypt, and is considerably
that could be actionable targets for therapy, such lower in most of Asia, Africa, and Central and
as in MLL, STAG2, TERT, and APOBEC (Balbas-​ South America (Figs. 22-​1 and 22-​2). Part of the
Martinez 2013; Roberts et  al, 2013; Guo et  al, international variation in incidence is due to geo-
2013; Nickerson 2014; Knowles & Hurst 2015). graphic differences in the prevalence of major
These molecular somatic studies are uncov- risk factors, primarily tobacco smoking but
ering a novel molecular taxonomy of bladder also occupational exposures and infection with
tumors with multiple molecular subtypes likely S.  haematobium in Egypt (Antoni et  al, 2017).
to have different etiology and prognosis (Lerner Some differences may also be due to practices in
et al, 2016). Whole genome and exome sequenc- the registration of low-​grade superficial tumors,
ing studies have identified multiple muta- a diagnostic bias that is also likely to affect time
tional signatures in bladder and other cancer trends. The incidence of urinary bladder cancer
sites (Alexandrov et  al, 2013). Of interest is the had been increasing in past decades, but in recent
APOBEC signature frequently found in bladder years has been generally moderately decreasing
tumors that has been linked to germline vari- or is stable (Chavan et al, 2014). However, inci-
ants in APOBEC3 and environmental exposures dence has been increasing in northern Africa and
(Middlebrooks et  al, 2016). In addition, molec- in some central, southern and eastern European
ular signatures in tumors have been linked to countries. In most countries, mortality has
specific exposures such as smoking (Alexandrov decreased or remained stable over time, partic-
et  al, 2016)  and aristolochic acid (Poon et  al, ularly among men. In the United States, with
2015; Castells et al, 2015). increased longevity, the burden of bladder can-
Progress The US Food and Drug Administration cer has risen from 50,000 new cases diagnosed
(FDA) recently approved the use of novel immu- per year between 1990 and 2000 to an estimated
notherapy agents targeting the checkpoint 77,000 cases in 2016 (SEER).
inhibitors PD-​1 and PDL-​1 for the treatment of The incidence of bladder cancer rises steeply
patients with locally advanced or metastatic uro- with age in both sexes after age 50 (Figs. 22-​3,
thelial carcinoma, thus opening new options for 22-​4). In most high-​income countries, bladder
54

China
Japan

Poland

Spain

Sweden
The Netherlands

United Kingdom

United States of America

0 1 2 3 4 5 6
ASR (world) per 100,000 person-years

FIGURE 22-​1  Age-​standardized (to the 2012 world population) incidence rates of bladder cancer among women.
Source: Ferlay et al, 2013.

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 5 10 15 20 25 30
ASR (world) per 100,000 person-years

FIGURE 22-​2  Age-​standardized (to the 2012 world population) incidence rates of bladder cancer among men.
Source: Ferlay et al, 2013.

70
New cases per 100,000

60
person-years

50
40
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 22-​3  Age-​specific incidence rates of bladder cancer among women.

Source: Ferlay et al, 2013.
546

546 Textbook of Cancer Epidemiology

350

New cases per 100,000

300
250

person-years
200
150
100
50
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 22-​4  Age-​specific incidence rates of bladder cancer among men.

Source: Ferlay et al, 2013.

cancer is three to four times more frequent in 165,000 deaths from this cancer worldwide in
men than in women, attributed in part to differ- 2012 (GCO, 2012) with most deaths occurring in
ences in smoking habits. In southern European Asia (N = 69,294) and Europe (N = 52,411).
countries, the male to female ratio in incidence is
higher (Samanic et al, 2006). Although hormonal GENETIC AND MOLECULAR
factors have also been implicated in the sex dif- EPIDEMIOLOGY
ference, the sex ratio for bladder cancer closely
follows that for lung cancer in all countries, indi- Family History of Bladder Cancer
cating smoking as the more relevant factor. Having a first-​degree relative with bladder can-
Incidence rates are generally higher in urban cer is a known risk factor for developing this dis-
than in rural areas. White men in the United ease, with relative risks ranging from 1.2 to 4.0
States have approximately twice the rate of black (Hemminki 2011; Lin 2006; Murta-​Nascimento
men, while the ethnic difference is less pro- 2007; Mucci et  al, 2016). Data showing cluster-
nounced for women. Studies in migrants from ing of familial bladder cancer with other cancer
Japan and China to the United States, and from sites (e.g., lung and kidney), provide further sup-
Asia and New Zealand to the United Kingdom, port for a role of shared genetic and/​or environ-
invariably show that the risk of bladder cancer in mental factors in the etiology of bladder cancer
migrants approximates that in the host country (Augustsson et  al, 1999; Baastrup et  al, 2008;
with increasing duration of stay. The higher prev- Balbás-​Martinez et  al, 2013). Data suggest that
alence of smoking among lower socioeconomic 7% (95% confidence interval [CI] 4%–​ 11%)
levels and the preponderance of jobs entailing a of the heritability bladder cancer is attributed
high risk of bladder cancer among manual work- to genetic determinants of smoking (Balbás-​
ers would have led to the assumption that the Martinez et al, 2013).
incidence is higher in lower than in upper social
classes, but neither male nor female mortality Genetic Susceptibility
data suggest a consistent social class gradient of Bladder Cancer
(Rosso et al, 1997). The currently known bladder cancer susceptibly
loci identified by candidate gene and genome-​
Mortality wide association studies (GWAS) are summa-
Bladder cancer is the ninth leading cause of death rized in Table 22-​1. Urinary bladder cancer is
from cancer in the United States with 16,390 one of the few sites for which the candidate gene
deaths in 2016 (SEER). The US mortality rate was approach to identify common susceptibility
4.4 per 100,000 men and women per year based genetic loci proved successful. This reflects the
on 2009–​2013 deaths, and death rates have been knowledge of bladder cancer carcinogenic path-
stable over 2004–​2013. There were an estimated ways and polymorphisms in drug and carcinogen
 547
TABLE 22-​1   ODDS RATIO AND P-VALUE ESTIMATES FOR OVERALL RISK AND NOTATION OF WHETHER THERE IS EVIDENCE
OF GENE-ENVIRONMENT INTERACTION (GXE) OR DIFFERENCES BY TUMOR CHARACTERISTICS FOR BLADDER CANCER
SUSCEPTIBILIT Y MARKERS

Locus Marker, Fine-​mapping Nearest genes RAF, ORa P-​valuea GXE? Differences by Refs
risk allele markers controls stage, grade or risk
of invasion?

1p13.3 GSTM1-​deletion GSTM1 0.51 1.45 5.16E-​36 Additive No Garcia-​Closas, 2013;

interaction Rothman, 2010;
with smoking Figueroa, 2015
and high-​risk
occupation
2q37.1 rs11892031, A rs17863783, T UGT1A cluster 0.92 1.17 1.02E-​07 Additive No Garcia-​Closas, 2013;
interaction with Figueroa, 2014; Tang,
smoking and high-​ 2012
risk occupation

3q26.2 rs10936599, C TERC, 0.76 1.18 4.53E-​09 Additive Yes, association Figueroa, 2014
ACTRT3,MYNN, interaction with stronger in low
LRRC34 smoking grade tumors

3q28 rs710521, A TP63 0.73 1.14 1.92E-​11 –​ No Garcia-​Closas, 2013;

Rothman, 2010;
Figueroa, 2014;
Kiemeney et al, 2008

4p16.3 rs798766, T TMEM129, TACC3, 0.19 1.22 7.06E-​25 Additive Yes, association Garcia-​Closas, 2013;
FGFR3 interaction with with low grade, Figueroa, 2014;
smoking and high-​ stage, and low Rothman, 2010;
risk occupation risk of invasion Figueroa, 2015;
tumors. Also data Grotenhuis, 2014;
suggest higher risk Kiemeney et al, 2010
of recurrence and
progression among
never smokers.
(continued)
 548
TABLE 22-​1  CONTINUED

Locus Marker, Fine-​mapping Nearest genes RAF, ORa P-​valuea GXE? Differences by Refs
risk allele markers controls stage, grade or risk
of invasion?

5p11.3 rs401681, C TERT, CLPTM1L 0.54 1.12 4.27E-​11 Additive Yes, association Garcia-​Closas, 2013;
interaction with with low grade and Figueroa, 2014;
smoking low risk of invasion Rothman, 2010;
tumors Rafnar, 2009

8q24 rs9642880, T MYC 0.45 1.24 3.55E-​38 Additive Yes, association Garcia-​Closas, 2013;
interaction with with low grade and Figueroa, 2014;
smoking low risk of invasion Rothman, 2010;
tumors and higher Grotenhuis, 2014;
risk of recurrence Kiemeney, 2008
and progression
among NIMBC

8p22 rs1495741, A NAT2 0.8 1.14 1.77E-​10 Multiplicative No Garcia-​Closas, 2013;

interaction with Figueroa, 2014;
smoking Garcia-​Closas, 2011;
Rothman, 2010

8q24.23 rs2294008, T rs2294008, T PSCA 0.46 1.13 1.27E-​15 Additive No Figueroa, 2014;
interaction with Rothman, 2010; Fu
smoking et al, 2012; Kohar,
2013; Wu, 2009

11p15.5 rs907611, A LSP1 0.31 1.15 4.11E-​08 –​ Yes, low-​grade Figueroa, 2014
tumors

13q34 rs4907479, A MCF2L 0.28 1.12 3.33E-​10 –​ No Figueroa, 2016

15q24 rs11543198, G CYP1A2 0.78 1.41 4.03E-​09 ?? ?? Matsuda, 2015

 549
18q12.3 rs10775480,T/​ rs10775480, T SLC14A1 0.43 1.13 6.15E-​08 Additive Yes, suggestive Figueroa, 2014;
rs10853535, C interaction with association with Garcia Closas, 2011;
smoking high grade and Koutros et al, 2013
high risk of
invasion tumors

18q12.3 rs17674580, T SLC14A1 0.34 1.17 7.60E-​11 ?? No Rafnar et al, 2011

18q12.3 rs1058396, G SLC14A1 0.5 1.14 2.90E-​09 ?? No Rafnar, 2011

19q12 rs8102137, C rs7257330, A CCNE1 0.33 1.13 1.22E-​11 Additive Yes, stronger Figueroa, 2014;
interaction with association with Rothman, 2010; Fu,
smoking high grade and 2014
high risk of
invasion tumors

20p12.2 rs6104690, A gene desert 0.56 1.09 2.19E-​11 –​ No Figueroa, 2014;

Figueroa, 2016

20p12.2 rs62185668, A gene desert 0.24 1.19 1.50E-​11 –​ Yes, stronger Figueroa, 2016;
association with Rafnar et al, 2014
high stage tumors

20p12.2 rs4813953, T gene desert 0.38 1.16 2.10E-​10 –​ Yes, stronger Figueroa, 2016
association with
high-​stage tumors

22q13.1 rs1014971, T CBX6, APOBEC 0.62 1.13 9.76E-​12 Additive Yes, stronger Garcia-​Closas, 2013;
interaction with association with Rothman, 2010;
smoking high risk of Figueroa, 2015
invasion tumors
a
Odds Ratios and P-​values presented are based on pooled or meta-​analysis data for all available bladder cancer GWAS studies when available.
50

550 Textbook of Cancer Epidemiology

metabolism genes that informed the choice Rothman et al, 2010). Gene–​environment inter-
of candidate genes and variants. Interestingly, action analyses have shown that the NAT2 slow
GWAS using an agnostic approach have also acetylation genotype is associated with risk only
identified variants in or nearby genes involved among cigarette smokers (Garcia-​Closas et  al,
in the pathways previously known to be involved 2013). Conversely, this implies that smokers with
in bladder carcinogenesis, including carcinogen the slow acetylator genotype are at higher risk
metabolism, urinary excretion, and oncogenes from smoking than smokers with the rapid/​inter-
and tumor suppressors. mediate acetylation genotypes. Besides exposure
to benzidine exposure from occupation (Carreón
Polymorphisms in Carcinogen et  al, 2006), no significant interactions have
Metabolism Genes been shown between NAT2 genotype and occu-
In the late 19th century, exposure to occupational pational exposures (Figueroa et  al, 2013). This
and environmental chemicals was the leading evidence supports the role of NAT2 acetylation
hypothesis to explain reports of high incidence and detoxification of aromatic amines in bladder
of bladder cancer among dye industry work- carcinogenesis (Table 22-​1). There is limited evi-
ers (Dietrich & Dietrich 2001). This increased dence for differences of the NAT2 risk associa-
risk has been attributed to occupational expo- tion by tumor characteristics or for associations
sure to aromatic amines (Hecht, 2003)  that are with clinical prognosis (Rothman et  al, 2010;
also found in tobacco smoke (Hecht, 2003). Grotenhuis et al, 2014).
Aromatic amines are converted into biologically GSTM1 is a member of the glutathione
active carcinogens during a two-​stage cellular enzymes important for phase II detoxification
detoxification/​bioactivation process. The first of polycyclic hydrocarbons and other chemi-
stage is a hepatic N-​hydroxylation of aromatic cals hypothesized to be involved in bladder car-
amines by the CYP1A2 enzyme, which belongs cinogenesis. In addition, GSTM1 is involved in
to the cytochrome P450 phase I  detoxification oxidative damage, which is another possible
system. The second stage is an enzymatic con- mechanism that may influence its role in blad-
jugation of the N-​hydroxylated aromatic amines der cancer etiology. An inherited deletion in
by phase II detoxification enzymes, such as N-​ GSTM1 has been consistently associated with
acetyltransferases (NATs), glutathione transfer- bladder cancer risk in meta-​and pooled analyses
ases (GSTs), and UDP-​glucuronosyltransferases (Garcia-​Closas et  al, 2005; Figueroa et  al, 2014;
(UGTs). The conjugation facilitates the excretion Rothman et al, 2010). There is an apparent addi-
of the N-​ hydroxylated intermediates via stool tive interaction between GSTM1 and smoking
and urine. It is hypothesized that direct expo- and high-​ risk occupations for bladder cancer
sure to urine enriched with these highly unsta- (Table 22-​1), supporting the presence of biologic
ble conjugates, which is influenced by specific interactions between these two major environ-
genetic polymorphisms, may lead to cancer risk mental risk factors (Garcia-​Closas et  al, 2013;
(Kadlubar et al, 1992). Figueroa et al, 2015).
This knowledge informed candidate gene GWAS have identified new susceptibility
studies focusing on variants in genes encoding markers in genes known to be important for car-
enzymes that metabolize known carcinogens cinogen detoxification, specifically in UGT1A6
found in tobacco products. These studies consist- and CYP1A2 (Rothman et al, 2010; Matsuda et al,
ently showed a modest increased risk of bladder 2015; Table 22-​1. Moreover, there is evidence of
cancer associated with the NAT2 slow acetylation interaction between SNPs in UGT1A6 and smok-
genotype among smokers, which is consistent ing and high-​risk occupations and bladder cancer
with the role of the NAT2 enzyme in aromatic (Garcia-​Closas et al, 2013; Figueroa et al, 2015).
amine metabolism (Garcia-​Closas 2005). GWAS Fine-​mapping and functional genomic studies of
have identified over a dozen low-​penetrant risk the UGT1A6 region have identified a rare SNP
variants for bladder cancer at a level of genome-​ rs17863783 (allelic frequency 1.4%) that showed
wide significance (Table 22-​ 1). Interestingly, a stronger association with risk than the variant
GWAS identified a single nucleotide polymor- identified by GWAS. The protective T allele was
phism (SNP) that captured the majority of NAT2 also associated with increased mRNA expression
acetylation status haplotypes in Europeans, con- of UGT1A6.1 in in vitro exon trap assays and in
firming findings from candidate gene studies human liver tissue samples (Tang et  al, 2012).
(Figueroa et  al, 2014; Garcia-​Closas et  al, 2011; These data suggest that rs17863783 may protect
51

Urinary Bladder Cancer 551

from bladder cancer by increasing the removal findings could have implications for targeted
of carcinogens from bladder epithelium through prevention strategies, although these estimates
higher expression levels of the UGT1A6.1 pro- are too preliminary to translate into specific pub-
tein. rs11543198 in CYP1A2 was identified in a lic health recommendations.
GWAS of Asian populations. The variant allele
of rs11543198 or rs8041357, a highly correlated RISK FACTORS
SNP (r2 = 1.0), is more common in Asians com-
pared to Europeans. This SNP was shown to have Tobacco
a stronger effect in male Asian smokers, support- Cigarettes
ing this marker to potentially influence its effects Cigarette smoking is the most important cause of
on risk through detoxification of bladder carcin- bladder cancer, particularly among men because
ogens (Matsuda et al, 2015). of the higher prevalence of smoking in men com-
pared to women. The association between ciga-
Other Bladder Cancer rette smoking and bladder cancer risk has been
Susceptibility Loci examined in more than 40 case-​control and more
GWAS have identified additional loci whose than 20 cohort studies (Doll et al, 1994; Brennan
function is not yet fully understood. Four sus- et  al, 2000; IARC, 2004a). Nearly all studies
ceptibility loci reside near or in important showed excess risks among smokers, on average
tumor oncogenes/​ tumor suppressors or DNA a two-​to threefold risk of bladder cancer in ever
maintenance genes. Specifically, rs10936599 is smokers compared to nonsmokers. Among heavy
located at 3q26.2, where the telomerase RNA smokers who smoke more than 20 cigarettes per
component of TERC resides; rs710521 at 3q28 day, the relative risk is around 5.  Large cohort
is in the TP63 gene; rs798766 at 4p16.3 is near studies and pooled analyses have shown that
FGFR3; rs401681 at 5p11.3 is in the telomerase the strength of the smoking association is simi-
gene TERT; rs9642880 is at the 8q24 multifo- lar among men and women (Puente et al, 2006;
cal cancer susceptibility region near MYC; and Freedman et  al, 2016), although early studies
rs8102137 at 19q12 is near the cell cycle gene had suggested that women may be more suscep-
CCNE1 (Table 22-​1). Fine-​mapping work on the tible to tobacco carcinogens than men. There is
19q12 region upstream of CCNE1, identified a considerable variation among studies in the risk
functional promoter variant rs7257330, which gradient with number of cigarettes smoked daily,
showed strong allele-​specific binding of nuclear which may be due to different study designs,
proteins in several cell lines. In two GWAS, different ways of smoking, or different types of
rs7257330 was associated only with aggressive tobacco smoked (IARC, 2004a).
bladder cancer (Figueroa et  al, 2014; Rothman The population attributable risk for ever
et  al, 2010; Fu et  al, 2014). Interestingly, all of smoking was calculated by the International
these SNPs apart from rs710521 show evidence Agency for Research on Cancer (IARC, 2004a)
of an interaction with smoking. as 66% (95% CI 61%–​70%) for men and 30%
(95% CI 25%–​35%) for women. The lower attrib-
Polygenic Risk Score and Risk utable proportion among women might reflect
of Bladder Cancer differences in prevalence of smoking or timing
Individually, the common susceptibly loci of smoking patterns between sexes. Subsequent
described previously have weak associations studies suggest the population attributable risk
with bladder cancer risk. However, in combi- for smoking for women was comparable with
nation they can result in substantial risk strati- that for men (Freedman et al, 2011).
fication in the population (Garcia-​Closas et  al, The risk of bladder cancer increases with both
2013; Figueroa et al, 2014). Moreover, polygenic the number of years of smoking and the average
risk captured by currently identified loci could number of cigarettes a person has smoked (Doll
identify individuals that might benefit most from et  al, 1994; Brennan et  al, 2000). The relative
smoking cessation. For instance, Garcia-​Closas risk may, however, level off at high tobacco con-
et  al estimated that the 30-​year absolute risk of sumption levels. In a pooled analysis, the risk of
bladder cancer in US males could range from bladder cancer for those who smoked 15 to 19
2.9% for current smokers in the lowest quar- cigarettes per day was similar to the risk of those
tile of the polygenic risk score to 9.9% for cur- who smoked more than 20 per day (Brennan
rent smokers in the upper score quartile. Such et al, 2000; IARC, 2004a). The leveling off in the
52

552 Textbook of Cancer Epidemiology

exposure–​response pattern may be interpreted The vast majority of studies in industrial-

as an effect of measurement error if heavy smok- ized countries have not evaluated the association
ers tend to underestimate their consumption, between tobacco smoking and nontransitional
or could reflect the interaction between genes cell carcinoma (NTCC), primarily due to the
and the environment or saturation of enzymes low frequency of these cancers in most stud-
(Vineis et al, 2000). ies. One small study in the United States and
The age at which a person starts smoking may a larger pooled analysis in Europe (Fortuny
affect the risk of bladder cancer. A  higher risk et  al, 1999)  identified smoking as an important
among smokers starting before age 20 compared cause of squamous cell bladder carcinomas. In
to those starting after age 20 has been found in the European study, current smoking was sig-
some studies. However, duration of smoking, age nificantly associated with NTCC, and the risk
at start, and age at stopping smoking are strongly increasing with increasing number of pack-​years
correlated and few studies have tried to disentan- of smoking. In a large study in Egypt (Zheng
gle the individual effects. Former smokers are at et  al, 2012), cigarette smoking among men was
lower risk than current smokers. Following ces- associated with an increased risk of urothelial
sation of smoking, the relative risk of bladder carcinoma but not squamous cell carcinoma,
cancer has, in most studies, been found to fall although smoking of both water pipes and cig-
by 20% to 30% within the first 5  years of ces- arettes was associated with an increased risk of
sation. In long-​term ex-​smokers, however, the squamous cell carcinoma (odds ratio [OR] 1.8,
risk seems to remain higher even 20 years after 95% CI 1.2–​2.6).
quitting than in never smokers (IARC, 2004a;
Brennan et al, 2000). Pipes, Cigars, and Other Forms
Blond tobacco (flue-​cured) is the standard of Smoking
tobacco type in northern Europe and North Evidence from epidemiologic studies evaluating
America, while in southern Europe and South pipe, cigar, and other forms of smoking in rela-
America black tobacco (air-​cured) is still widely tion to bladder cancer is more equivocal than
consumed. Studies comparing the two types for cigarette smoking (IARC, 2004a; Shapiro
of tobacco have found an approximately two- et al, 2000). In general, the data support a small
fold higher risk associated with black tobacco excess risk of bladder cancer among heavy pipe
compared to blond tobacco (Clavel et  al, 1989; or cigar smokers. The few reports on other forms
Samanic et al, 2006). The smoke condensate from of tobacco use such as snuff, chewing tobacco,
black tobacco contains a higher concentration and water pipes are mostly negative, but the rela-
of aromatic amines such as 4-​ aminobiphenyl, tive risk estimates were based on small numbers
b-​naphthylamine, and o-​ toluidine than the of exposed cases and controls.
smoke condensate from blond tobacco. Levels
of 4-​aminobiphenyl hemoglobin adducts have Diet
been found to be approximately twice as high Except for a possible protective effect of fruits
in smokers of black compared to smokers of and vegetables, there is little evidence for an
blond tobacco with the same amount of smok- association of other foods with bladder cancer
ing. Moreover, black tobacco smokers have risk (World Cancer Research Fund/​ American
been found to excrete a 1.8-​fold higher level of Institute for Cancer Research [WCRF/​ AICR],
urinary mutagens than blond tobacco smokers 2015). Evidence on coffee, tea, fluid intake,
(Malaveille et al, 1989). and water contaminants is presented in other
Smoking unfiltered cigarettes may be associ- sections.
ated with a 30%–​50% increase in the risk of blad- Fruits, fruit juices, and vegetables such as car-
der cancer as compared to the risk of subjects who rots, cruciferous vegetables, and dark-​green veg-
smoke only filtered cigarettes; a similar pattern is etables have, in numerous studies, appeared to
observed for subjects who inhale into the throat be associated with a lower risk of bladder cancer
as compared to those who do not inhale deeply (Steinmaus et al, 2000; Key et al, 2009). However,
(IARC, 2004a; Samanic et  al, 2006). Most stud- large cohort studies including the European
ies examining exposure to second-​hand smoke Prospective Investigation into Cancer (EPIC)
in relation to bladder cancer have not found an study (Büchner et al, 2009), a cohort of Swedish
increasing risk with number of years of exposure women (Larsson et  al, 2008a), and the Nurses’
(IARC, 2004a). Health study (Holick et  al, 2005)  did not find
53

Urinary Bladder Cancer 553

overall associations with consumption of fruits The association between specific dietary pat-
and vegetables and risk. Results are inconsist- terns and bladder cancer risk has been under-
ent with respect to specific micronutrients. Some taken in the EPIC cohort. Higher adherence to
studies suggest that supplemental or dietary the Mediterranean diet score was associated with
intake of vitamin A and vitamin C, and dietary a lower risk of bladder cancer overall (hazard
intake of folate and carotenoids are associated ratio [HR] 0.84, 95% CI 0.69–​1.03). Moreover,
with a lower risk of bladder cancer (Zeegers the association was stronger among smokers
et al, 2001a; Steinmaus et al, 2000; WCRF/​AICR, (Buckland et  al, 2014). However, adherence to
2015). An inverse association with long-​ term the WCRF/​AICR recommendations for cancer
intake of vitamin E has been observed in some prevention was not associated with bladder can-
studies. Low baseline levels of 25-​hydroxyvitamin cer risk in contrast to most other major cancers
D (25(OH)D) were associated with an increased (Romaguera et al, 2012).
bladder cancer risk in a US cohort of smokers
(Mondul et  al, 2010). A  moderate inverse asso- Total Fluid Intake
ciation with selenium intake has been observed Prolonged exposure of the bladder urothelium
in several studies, including large cohort studies, to carcinogens in the urine has been suggested
but results are not entirely consistent (Brinkman to affect the development of bladder cancer and
et al, 2006). perhaps also to influence the site distribution of
An increased risk of bladder cancer with tumors within the bladder. To test this hypothe-
increased consumption of meat, including pro- sis, epidemiologic studies have evaluated the risk
cessed meat, and exposure to nitrosamines, poul- of bladder cancer in relation to quantity of fluid
try, and fish has been reported, but the overall intake.
evidence is inconclusive (Steinmaus et  al, 2000; Contrary to expectations, many epidemio-
WCRF/​ AICR, 2015). Diets low in fat may be logic studies have found no association or even a
associated with a decreased risk (Steinmaus et al, slight excess risk in subjects with high total fluid
2000). Several studies have investigated the asso- intake, although the results have not been consist-
ciation of milk and dairy products and bladder ent. The strongest evidence of a protective effect
cancer risk, and some large cohort studies found comes from the US Health Professionals Follow-​
a protective effect of consumption of cultured Up Study and the Nurses’ Health Study, which
milk such as sour milk and yogurt (Larsson et al, found a 240-​mL increase in total fluid intake was
2008b). There is no converging evidence to sup- associated with a significant 7% decrease in risk
port or refute a role of other major food groups, among men after adjusting for types of bever-
specific macronutrients (lipids, proteins, carbohy- ages (Michaud et al, 1999). Among women in the
drates), or total energy intake with bladder cancer. Nurses’ Health Study, a modest inverse associated
Cooking process may affect the formation was observed for the highest quartile of average
of carcinogens such as heterocyclic amines, fluid intake compared to the lowest (relative risk
which are formed mainly through grilling or fry- 0.83, 95% CI 0.61–​1.12). A later study found this
ing of meat. While two studies found a twofold protective effect was strongest for female smok-
increased risk of bladder cancer among subjects ers (Zhou et  al, 2014). By contrast, total fluid
with high consumption of fried foods or fried/​ intake was not associated with risk of bladder
grilled meat or fish, another found no significant cancer in the EPIC cohort (HR 1.12, 95% CI
excess risk in subjects with the highest intake of 0.86–​1.45) comparing highest to the lowest tertile
heterocyclic amines compared to those with the of intake (Ros et  al, 2012). A  pooled analysis of
lowest (Augustsson et al, 1999). six case-​control studies of bladder cancer found
Saccharin, a nonnutritive sweetener, is an increased risk with higher intake of tap water
absorbed slowly, remains unmetabolized, and is while no increased risk was observed for the same
excreted through the kidneys. In experimental intake of nontap water (Villanueva et al, 2006).
studies, bladder cancer incidence is higher in rats High fluid intake could reduce the contact of
exposed to high doses of saccharin in utero and carcinogens within the urothelium by diluting
during weaning (Schoenig et al, 1985). While one the urine or increasing the frequency of urina-
epidemiologic study found an excess risk in men tion, whereas there are no plausible biological
consuming artificial sweeteners (saccharin and hypotheses explaining how high total fluid intake
cyclamate), the results of most subsequent stud- could increase the risk. The positive associations
ies did not support this association. in some studies may be attributed to the types
54

554 Textbook of Cancer Epidemiology

of fluid consumed, such as coffee or tap water, of case-​control and cohort studies, the pooled
which may contain carcinogenic constituents. relative risks for bladder cancer were 1.00 (95%
Findings may vary between studies also due to CI 0.92–​1.09) for moderate and 1.02 (95% CI
differences in the definition of total fluid intake. 0.78–​1.33) for heavy alcohol drinkers compared
with nondrinkers (Pelucchi et  al, 2012). Results
Coffee and Tea for specific alcoholic beverages and bladder can-
Since the first reports in 1971 of coffee intake and cer risk are inconsistent.
an increased risk of bladder cancer, numerous
studies have examined this association. Although Reproductive Factors and
some studies suggest a weak positive association Hormones
between heavy coffee consumption and blad- The difference in bladder cancer incidence
der cancer, the overall evidence is inconclusive between men and women seen in most countries,
(Loomis et al, 2016). No association and no dose-​ although not in some African countries, led to
response was observed between intake of coffee the hypothesis that hormonal factors play a role
and risk of urothelial cancer in the EPIC cohort in the occurrence of bladder cancer (Hartge et al,
that examined 233,236 subjects (Ross et al, 2011). 1990). Experimental data indicate that sex steroid
A meta-​analysis including 23 case-​control stud- hormones may modify the effect of chemical-​
ies with 7,690 cases and 13,507 controls, and 5 induced bladder cancer. Several studies have
cohort studies with 700 cases and 229,099 par- found an inverse association between parity and
ticipants identified a positive association with bladder cancer risk among women (Daugherty
heavy coffee consumption in the case-​control but et  al, 2013; Dietrich et  al, 2011), while findings
not the cohort studies (Zhou et al, 2012). are less consistent for age at menarche and use of
The confounding effect of cigarette smok- contraceptives and hormone replacement ther-
ing, a habit closely related to coffee consumption apy. Overall, there is only limited epidemiologic
in some settings, introduces one of the major evidence on this topic.
issues in the evaluation of coffee consumption
as an independent risk factor for bladder cancer. Anthropometric Measures
Residual confounding may bias the results even The EPIC cohort examined 390,000 participants
when adjustment for smoking is done. To avoid using different measures of adiposity and found
such confounding, coffee consumption has been an association between overweight and blad-
examined among nonsmokers in case-​ control der cancer in men but not in women (Roswall
studies and an elevated risk of bladder cancer et al, 2014). The association was not observed in
was found among heavy coffee drinkers (Sala nonsmokers and thus could be due to residual
et al, 2000). confounding by smoking. A 2015 meta-​analysis
Most cohort and case-​control studies have not including EPIC identified a 10% increased
found either no or an inverse association between bladder cancer risk for obese subjects (Sun et al,
tea consumption and bladder cancer. Large 2015). Obesity has been evaluated in relation
cohort studies (Michaud et  al, 1999 Ros et  al, to prognosis, with overweight bladder cancer
2012) and a meta-​analysis (Zhang et al, 2015) do patients having worse prognosis and more peri-
not support an association between tea con- operative complications compared to healthy
sumption and bladder cancer or suggest a protec- weight individuals.
tive effect in subgroups. The 2015 WCRF/​AICR
review shows an approximate 5% lower risk Physical Activity
per cup of tea consumed (WCRF/​AICR, 2015). The most consistent evidence on leisure-​time
Animal studies have shown that compounds in physical activity and bladder cancer risk comes
tea may have inhibitory effects on bladder tumor from a pooled analysis of 12 prospective US
growth possibly due to antioxidant and antipro- and European cohorts with self-​reported phys-
liferative effects of polyphenol compounds pres- ical activity data. The analysis included 9,073
ent in green tea but also found in black teas. bladder cancer cases and suggested a 13% lower
risk (HR 0.87, 95% CI 0.82–​0.92) for the high-
Alcohol est (90th percentile) versus lowest (10th per-
Epidemiologic evidence has, overall, not sup- centile) level of leisure-​time physical activity,
ported a role for alcohol in the etiology of bladder even after adjusting for body mass index (Moore
cancer (WCRF/​AICR, 2007). In a meta-​analysis et al, 2016).
5

Urinary Bladder Cancer 555

Infections of bladder cancer caused by S.  haematobium

The evidence on infectious agents and bladder (Yang et al, 2005). A large case-​control study in
cancer is strongest for squamous cell carcinoma Spain found an approximate 50% increased risk
in relation to infection by S.  haematobium. In of bladder cancer associated with high seroreac-
the early 1990s, the World Health Organization tivity to two polyomaviruses. There was no differ-
estimated that 200 million people were infected ence in seroreactivity with disease stage or grade,
by S. haematobium, an infection that is endemic however (Robles et al, 2013).
in Africa and the eastern Mediterranean. The
first case reports, many from the Nile Delta, sug- Ionizing Radiation
gesting a link between S. haematobium infection Convincing evidence for a link between expo-
and bladder cancer date back to the early 1900s sure to ionizing radiation and risk of bladder
(IARC, 1994). In areas with a high prevalence cancer originates from studies of medically
of S.  haematobium in southern Iraq, more than exposed subjects, of atomic bomb survivors,
50% of all bladder cancers are of the squamous and of occupational groups. X-​ray treatment for
cell type; in northern Iraq, where the infection patients with ankylosing spondylitis substantially
is less common, this proportion is lower. In east increased mortality from various cancers, includ-
African and Middle Eastern countries, squamous ing bladder cancer (Weiss et al, 1994). The risk of
cell carcinoma is much more common than in bladder cancer was increased fourfold in women
Europe and North America, a pattern associ- receiving radiotherapy for benign gynecologic
ated with a high prevalence of S.  haematobium conditions (Boice et al, 1988). An excess risk of
infection. An ecologic correlation between prev- bladder cancer has also been observed among
alence of S.  haematobium infection and death atomic bomb survivors, workers of nuclear
from bladder cancer has been noted in several installations, and participants in the UK nuclear
African countries (IARC, 1994). Based on studies weapon testing programs conducted in Australia
in Africa, the relative risk of NTCC of the blad- and the Pacific Ocean (McGeoghegan & Binks,
der among those infected with S. haematobium is 2000; Richardson et al, 2013).
about fivefold (Vizcaino et al, 1994).
According to an Egyptian study, a clinical Occupation
history of urinary schistosomiasis could explain After smoking, occupational exposures have
16% of bladder cancer cases in this population been identified as the second most important risk
(Bedwani et al, 1998). In a large study in Egypt, factor category for bladder cancer in both men
history of schistosomiasis was associated with an and women. Studies in the 1980s and 1990s sug-
increased risk of both urothelial carcinoma (OR gest that high-​risk occupations may be responsi-
1.9, 95% CI 1.2–​2.9) and squamous cell carci- ble for 4% to 10% of bladder cancer cases in men
noma (OR 1.9, 95% CI 1.2–​3.0) in women and to (Kogevinas et  al, 2003)  and a lower percentage
a lesser extent in men (Zheng et al, 2012). in women (Mannetje et  al, 1999). Occupational
A positive association between cystitis exposures to aromatic amines, polyaromatic
or nonspecific infections of the urinary tract hydrocarbons (PAHs), diesel engine exhaust,
and bladder cancer risk has been consistently and metal working fluids (MWFs) are most con-
reported (Zhou et  al, 2015), including reports sistently associated with an increased risk. The
from patients with spinal cord injury. Possible occupational exposures classified by the IARC
mechanisms include increased absorption of car- as being associated with a higher risk of bladder
cinogenic substances, chronic inflammation, and cancer are listed in Box 22–​1.
the role of bacteria in the endocystic formation Evidence from large cohort studies, pooled
of N-​ nitroso compounds. The varying results analyses of case-​ control studies, and record
point to the difficulties in studying this associa- linkage studies have found that the occupations
tion because early symptoms of bladder cancer linked most consistently with bladder cancer
are similar to those of cystitis. Also, patients may risk are painter, machinist, mechanic, dye work-
recall urinary conditions better than controls. ers, work in the metal industry such as sheet
The prevalence of human papilloma virus metal work and blacksmithing, work in the tex-
(HPV) in bladder cancer biopsy specimens is tile industry, leather workers, hairdressing, dry
around 5% or less, indicating that HPV is not a cleaning, transport work, waiters, and tobacco
major factor in the etiology of the disease. There is workers (Kogevinas et al, 2003; Mannetje et al,
some evidence linking HPV with the occurrence 1999; Colt et al, 2011; Pukkala et al, 2009; Reulen
56

556 Textbook of Cancer Epidemiology

conducted among dyestuff manufacturing work-

BOX 22-​1 ers in the United Kingdom in 1954 (Case et  al,
 OCCUPATIONAL CARCINOGENS 1993), in which exposure to b-​naphthylamine
THAT HAVE BEEN ASSOCIATED entailed a 90-​fold and to benzidine a 14-​fold
WITH BLADDER CANCER AS excess risk of bladder cancer. Subsequent studies
of dyestuff manufacture workers confirmed these
CLASSIFIED BY IARC
findings (Vineis & Pirastu, 1997)  and provided
analyses of the exposure–​response patterns for
SUFFICIENT EVIDENCE
these chemicals. Among workers manufacturing
IN HUMANS aromatic amines, including b-​ naphthylamine,
• Aluminum production benzidine, 4-​aminobiphenyl, and 4-​o-​toluidine,
• 4-​aminobiphenyl relative risks have ranged from 6 to 70.
• auramine production Contemporary studies with extensive follow-​
• benzidine up of aromatic amine exposed workers identify
• magenta production an increased bladder cancer risk after several
• 2-​naphthylamine decades (Pira et al, 2010; Carreon et al, 2010).
• Painting An excess risk of bladder cancer identified in
• Rubber production industry the rubber industry in the early 1950s was asso-
• ortho-​toluidine ciated with the use of an antioxidant containing
b-​naphthylamine (Veys, 2004). The withdrawal
LIMITED EVIDENCE IN HUMANS
of this compound in the rubber industry led to
a reduction of bladder cancer incidence among
• 4-​chloro-​ortho-​toluidine
rubber workers. However, a review of studies
• coal tars and pitches
in the rubber industry in the late 1990s found
• diesel engine exhaust an excess 50% risk in persisted (Kogevinas et al,
• dry-​cleaning 1998). Subsequent studies using biomarkers in
• hairdressers and barbers workers in western Europe still identify increased
• printing processes levels markers of DNA damage (e.g., urothelial
• soot cell DNA adducts) and urine mutagenicity asso-
• tetrachloroethylene ciated with exposure to aromatic amines and
• textile manufacturing PAHs (Peters et  al, 2008). However, a pooled
analysis of rubber workers first employed since
1975 did not find an increased risk of bladder or
any other cancer (Boniol et al, 2016).
et al, 2008; Vlaanderen et al, 2014). These asso- Aromatic amines are present in lower quan-
ciations reflect, in part, past exposure to chem- tities in many other occupations including alu-
icals not currently used, such as benzidine or minum production, hairdressing, painting,
b-​naphthylamine, but also current exposures to printing, and shoemaking. The extent to which
aromatic amines, PAHs, cutting oils, and sol- these compounds contribute to the excess risk of
vents. Various studies have also observed excess bladder cancer observed in these occupations has
risks in white-​collar occupations such as sales not been adequately examined. A  cohort study
workers (Mannetje & Pearce, 2006), even after in farmers identified increased bladder can-
adjusting for potential confounding variables, an cer risk among those exposed to imidazolinone
observation compatible with the socioeconomic herbicides, which are aromatic amines (Koutros
pattern of this disease rather than to exposures et al, 2016).
in the workplace. Finally, studies in China and
Taiwan have reported an increased risk among Polyaromatic Hydrocarbons and Diesel
Chinese herbalists possibly associated with expo- Engine Exhaust
sure to aristolochic acid (see section “Aristolochic Heavy exposure to PAHs, particularly from coal
Acid”). tars and pitches, and to diesel exhaust entails a
moderately increased risk of bladder cancer,
Aromatic Amines although the evidence is not entirely consist-
The first large occupational epidemiology ent. Exposure to PAHs is high for workers in
study directly examining aromatic amines was industries such as the potrooms in aluminum
57

Urinary Bladder Cancer 557

production, coal gasification, roofing, road pav- carcinogenicity. Data from cohort studies mostly
ing, and the transport industry. The results from from Scandinavian countries found hairdressers,
case-​control studies provided evidence of excess barbers, beauticians, and other related workers
risk among transport workers exposed to engine have increased risks for bladder cancer among
exhausts including diesel. Cohort studies among men while an increase was not consistently
other workers exposed to diesel engine exhaust, found in women. Personal use of permanent
such as taxi and truck drivers and operators of hair dyes has been associated with an excess risk
heavy equipment, however, have not found a pos- in some case-​control studies (Gago-​Dominguez
itive association between exposure to PAHs or et  al, 2001)  but the overall results are not con-
diesel exhausts and bladder cancer risk. verging (Henley & Thun, 2001; Kogevinas et al,
2006). A meta-​analysis including 8,500 cases and
Metal Workers and Metal 14,000 controls did not find an increased risk of
Working Fluids bladder cancer for personal use of any type of
An excess risk associated with metalworking, hair dyes compared with no use (OR 0.93, 95%
although only moderately high, is among the CI 0.82–​1.05) in either men or women (Turati
most consistent epidemiologic findings. An ele- et al, 2014).
vated risk of bladder cancer has frequently been
found in aluminum smelter workers, black- Medical Conditions
smiths, foundry workers, furnace operators, and Treatment
machinists, welders, and others. Metal workers Medical Conditions
are exposed to a heterogeneous group of poten- Urinary tract stones and cancer of the renal pel-
tial carcinogens including cuttings oils—​a cate- vis are medical conditions associated with an
gory including numerous agents—​PAHs, metal increased risk of bladder cancer.
fumes, and dusts, as well as combustion gases Urinary tract stones may cause chronic irri-
and vapors. In European Union countries, metal tation and infection or they may be markers of
workers appear to be the largest occupational an underlying chronic infection. The strongest
group associated with increased bladder can- evidence for an association with bladder cancer
cer risk (Kogevinas et al, 2003). A study in New comes from a Swedish population-​based cohort
England, in the United States (Colt et al, 2011), of 61,144 patients hospitalized for kidney or ure-
identified increased bladder cancer risk in men ter stones followed for up to 25 years (Chow et al,
who reported using straight MWFs (OR 1.7, 95% 1997). A  significantly increased risk of bladder
CI 1.1–​2.8) and a slightly lower risk for use of cancer was observed, and the risk was higher
soluble MWFs. among women than among men and higher
Less consistent are the findings for persons in among patients with compared to without a diag-
numerous other occupations such as construc- nosis of urinary tract infection. Several studies
tion worker, electrical fitter, engine driver, food have identified an increased risk of bladder can-
processor and preserver, garage worker, plumber cer among patients with spinal cord injuries, pos-
and welder, railway worker, slaughterer and meat sibly due to higher prevalence of bladder calculi
processor, workers in nuclear plants, and cooks. and infections (Welk et al, 2013).
Studies of patients with cancer of the renal
Hair Dyes pelvis have found increased relative risks, rang-
Between 50% and 80% of women in the United ing from 3 to 40, of a second primary bladder
States, Japan, and the European Union have cancer. In one study (McCredie et  al, 1996),
used hair dyes (IARC, 2010). Hair dyes com- about half of the second primary tumors were
prise numerous compounds including aromatic detected within 2  months, and the excess risk
amines, nitro-​substituted aromatic amines, high decreased with time since the initial tumor diag-
molecular weight complexes, metal salts, and nosis. This pattern could suggest a shared etiol-
other chemicals, although their composition has ogy of the tumors (e.g., tobacco smoking or use
changed considerably throughout the years. of phenacetin). Moreover, the initial increase and
In 2010, the IARC evaluated hairdressing and subsequent decrease in bladder cancer risk over
barbering as occupations entailing exposures time could reflect increased medical surveillance
that are probably carcinogenic to humans, based shortly after diagnosis of a first tumor or a con-
on exposure to hair dyes; personal use of hair comitant decrease in exposure to specific risk
colorants could not be evaluated in terms of its factors such as smoking.
58

558 Textbook of Cancer Epidemiology

An excess risk of bladder cancer has been was associated with tumors of the renal pelvis,
reported among patients with cancers of the the ureter, and bladder, with a two-​to sixfold
prostate, renal parenchyma, testis, and skin increased risk in abusers of drugs containing
(nonmelanoma). There are no identified com- phenacetin (Piper et al, 1985). Since the with-
mon risk factors between these neoplasms and drawal of phenacetin, the consumption of par-
bladder cancer, except perhaps of smoking for acetamol (acetaminophen), an aromatic amine
renal cancer, and the excess occurrence of sec- metabolite of phenacetin, has increased in
ond cancers may be related to treatment, diag- many countries.
nostic surveillance, or other yet unidentified Epidemiological studies have found that use
factors. of paracetamol is associated with, at most, a small,
The association between diabetes and blad- nonsignificant excess risk of bladder cancer in
der cancer has been suggested for more than heavy users (Derby & Jick, 1996; Fortuny et  al,
50 years. A consensus report (Giovannucci et al, 2006). However, use of paracetamol has become
2010)  and subsequent large cohort studies (Lai widespread only recently, so possible long-​term
et  al, 2013)  indicate a 30% higher risk of blad- effects on cancer risk may not yet be possible to
der cancer among patients with type 2 diabetes. document. Only a few studies have examined the
This association may be explained by shared effect of NSAIDs on bladder cancer. Some studies
risk factors such as obesity, diet, and physical found an inverse association (Fortuny et al, 2006),
inactivity, although none of the other risk fac- including a large case-​control study in which reg-
tors are strongly associated with bladder cancer. ular use of nonaspirin, nonselective NSAIDs and
Metabolic disorders typical of diabetes such as ibuprofen was associated with reduced bladder
hyperglycemia, insulin resistance, and hyperin- cancer risk (Baris et al, 2013).
sulinemia may also be intermediate pathways.
A  meta-​analysis, however, examining glycemic Cyclophosphamide
index and glycemic load found no association Cyclophosphamide is mainly used as a single
with bladder cancer (Choi et  al, 2012). Finally, drug or in combination for treatment of non-​
the association between diabetes and bladder Hodgkin lymphoma, other neoplasias, and some
cancer risk could be mediated through diabe- nonneoplastic diseases (Knight et al, 2004), and
tes medications. Most, but not all, studies have is prescribed to half a million patients annu-
found an association between pioglitazone (a ally worldwide. Clinical studies have found a
thiazolidinedione) and bladder cancer risk (see substantially elevated risk of bladder cancer in
section “Medications”). patients treated with cyclophosphamide. Among
Patients with Wagener’s granulomatosis may patients with non-​Hodgkin lymphoma treated
have an increased bladder cancer risk irrespec- with cyclophosphamide, the cumulative risk of
tive of their treatment with cyclophosphamide bladder cancer was 10.7% at 12 years of follow-​up
(Knight et al, 2004). (Pedersen-​Bjergaard et al, 1988). It is not known
which of the three major DNA-​binding metabo-
Medications lites of cyclophosphamide produces the carcino-
Analgesics and anti-​inflammatory drugs, cyclo- genic effect.
phosphamide, pioglitazone, chlornaphazine (not
produced currently), Chinese herbs containing Pioglitazone
aristolochic acid, and barbiturates are the main Pioglitazone is an antidiabetic drug (thiazoli-
medications associated with an increased or, in dinedione) that reduces insulin resistance and
the case of barbiturates and nonsteroidal anti-​ decreases hyperglycemia in patients with type
inflammatory drugs (NSAIDS), decreased risk of 2 diabetes. The association between use of pio-
bladder cancer. glitazone and risk of bladder cancer has been
examined in clinical trials, cohorts including
Analgesics and NSAIDs register–​based studies, and nested case-​control
Heavy use of phenacetin provokes chronic studies in Europe, North America, and Asia.
interstitial nephropathy. Therefore, this widely A  cohort of 145,806 patients newly treated
used analgesic has been gradually withdrawn with antidiabetic drugs between 2000 and 2013
from the market. Since the 1960s, a number found that pioglitazone was associated with an
of reports have reported use of phenacetin increased risk of bladder cancer (HR 1.63, 95% CI
59

Urinary Bladder Cancer 559

1.22–​2.19). There was no association with other tuberculosis drug isoniazid and some laxatives is
antidiabetic drugs in that study (Tuccori et  al, inconclusive.
2016). Most, but not all studies (see, for exam-
ple, Lewis et al, 2015), have found an increased Environmental Exposures
risk of bladder cancer associated with use of Arsenic in Drinking Water
pioglitazone (IARC, 2013a). Exposure-​response The evidence linking arsenic to bladder cancer
relationships with duration or dose were found risk comes mainly from studies in populations
in many high-​quality studies that adjusted for exposed at high levels through drinking water
potential confounders. (IARC, 2004b). Additional evidence is found in
small follow-​up studies of patients treated with
Aristolochic Acid Fowler’s solution (potassium arsenite), in sub-
Aristolochic acid is a natural compound found jects with arsenic poisoning in Japan, and in
in plants that has been used in Chinese tradi- studies of occupationally exposed subjects.
tional medicine for many conditions. Several Exposure to arsenic from drinking water at
case reports from China, Taiwan, and Belgium low levels is widespread, but contamination at
have found consumption of Chinese herbal higher levels is a localized problem. Studies have
products containing Aristolochia species is been conducted in areas with naturally occur-
associated with human nephropathy and uro- ring very high levels of arsenic in drinking water
thelial cancer mostly in the upper urinary tract in Taiwan, Bangladesh, West Bengal in India,
(Nortier et al, 2000). Studies in Chinese herbal- Cordoba province in Argentina, and in north-
ists found an increased risk of bladder cancer ern Chile. Average exposure to arsenic varied in
(Yang et  al, 2013). Animal studies have con- these studies, and was typically above 100 mg/​L
firmed the carcinogenic potential of aristolo- and in some of those areas it was higher than 500
chic acid. DNA adducts specific for aristolochic mg/​L. Many of the studies at high-​level arsenic
acid that were identified in experimental ani- areas are ecologic, but relative risks are frequently
mals were also found in patients with aristolo- so high that potential biases (confounding) are
chic acid nephropathy and urothelial cancer highly unlikely to explain the excess risks.
(Arlt et al, 2007; Poon et al, 2015). A large case-​ Excess bladder cancer mortality was iden-
control study in Taiwan (Lai et al, 2010) found tified in both Chile and Argentina in the
increased risk of urothelial carcinoma and a high-​ arsenic exposure areas compared to the
linear dose–​response relationship between the low-​exposure areas (Hopenhayn-​Rich et al, 1996;
prescribed dose of Mu Tong (a Chinese herb Marshall et al, 2007). In Chile, the elevated risk of
containing aristolochic acid) and risk of uri- bladder cancer has persisted many years after the
nary tract cancer. reduction of arsenic levels in the water (Marshall
et al, 2007), and it was postulated that exposure
Barbiturates in early life affected risk in adults (Steinmaus
In two cohort studies of patients prescribed et al, 2014).
anticonvulsants (phenobarbital, phenytoin, and Blackfoot disease is a severe form of periph-
others) for the treatment of epilepsy, the risk of eral vascular disease common in Taiwan and
bladder cancer was lower than expected (Olsen China and linked to arsenic exposure. A series of
et al, 1989). The hypothesis that treatment inter- ecologic, case-​control, and cohort studies (Huang
acts negatively with smoking has received some et al, 2008) has been conducted in the blackfoot
epidemiologic support. Moreover, experimen- disease–​endemic areas in Taiwan and excess risks
tal studies indicate that phenobarbital induces of bladder cancer—​and other cancers—​have been
drug-​ metabolizing enzymes in the liver and consistently found, with an exposure–​response
may deactivate bladder carcinogens such as 4-​ relationship by years of consumption and by
aminobiphenyl, found in tobacco smoke. level of arsenic in well water (Chiou et al, 2001).
A progressive decrease in bladder cancer mortal-
Other Drugs ity was observed in a high-​arsenic-​exposure area
Clornaphazine, an antineoplastic drug no longer in southwestern Taiwan after the installation of a
used, has been associated with an excess risk tap-​water supply system (Su et al, 2011).
of bladder cancer and is classified as a Group 1 Evidence for exposure to low to moderate
carcinogen by the IARC. Evidence on the levels of arsenic comes mostly from studies in
560

560 Textbook of Cancer Epidemiology

Europe and the United States, but findings are et  al, 2004). A  pooled analysis of case-​control
less consistent (Karagas et  al, 2004; Baastrup studies identified a 50% increased risk among
et al, 2008; Meliker et al, 2010; Baris et al, 2016). subjects with long-​term exposure to disinfection
The excess incidence of bladder cancer in the byproducts at levels around 50 mg/​L, which are
New England region of the United States has levels currently observed in many industrialized
been attributed, in part, to high arsenic con- countries (Villanueva et al, 2004).
tents of well water (Karagas et al, 2004). A large Exposure to chlorinated byproducts in water
case-​control study in New England (Baris et  al, through inhalation and dermal absorption con-
2016) found a positive risk gradient for cumula- tribute to the total exposure to trihalomethanes
tive arsenic exposure from all water sources, with more than exposure through ingestion. Only
an odds ratio of 2.2 (95% CI of 1.3–​3.9) with a one epidemiological study has examined these
lag 40 years for the highest exposed participants routes of exposure and identified increased risks
(97.5th percentile) compared to those in the low- for exposure in showers and baths and for swim-
est quartile of exposure. ming in pools (Villanueva et al, 2007).
Arsenic does not directly damage DNA.
Several mechanisms for its carcinogenicity at Other Water Contaminants
low-​level concentrations have been proposed Exposure to nitrates in drinking water has been
(IARC, 2004b), including the inhibition of DNA examined in case-​ control and cohort studies,
repair mechanisms, induction of oxidative DNA which find no consistent evidence of an increased
damage, and changes in DNA-​methylation pat- bladder cancer risk (Ward et  al, 2003; Zeegers
terns, aneuploidy, and gene amplification. et al, 2006; Espejo-​Herrera et al, 2015). Few eco-
logic or case-​control studies have examined other
Water Disinfection Byproducts water contaminants such as metals (cadmium,
Chlorination byproducts in drinking water nickel), radionucleotides, and tetrachloroeth-
have been associated with bladder cancer risk ylene in relation to bladder cancer risk, and the
in some well-​conducted epidemiologic studies. epidemiologic evidence for excess risk is limited.
Chlorination of drinking water is used in most
industrialized countries for disinfection. During Outdoor Air Pollution
chlorination, chlorine reacts with organic mat- Several cohort and case-​control studies have eval-
ter in water to produce a mixture of byproducts, uated the association between exposure to out-
including trihalomethanes and numerous other door air pollution and bladder cancer, but results
compounds. The levels of these compounds are not consistent (Castano-​Vinyals et  al, 2008;
depend mainly on the amount of chlorine Liu et  al, 2009; Raaschou-​Nielsen et  al, 2011;
added and the water source. Trihalomethanes Visser et  al, 2004; Wilson et  al, 2008). In stud-
are mutagenic to bacteria, and some are car- ies of occupational groups with high exposure to
cinogenic to animals. Epidemiologic studies environmental air-​pollution such as policemen
have used trihalomethanes as indicators of the and taxi and bus drivers, an increased risk was
level of chlorination byproducts. However, found for occupations exposed to PAHs (Bossetti
other compounds such as the MX (3-​chloro-​4-​ et al, 2007) and diesel engine exhaust (Boffetta &
(dichloromethyl)-​5-​hydroxy-​2(5H)-​furanone), Silverman 2001), but not among miners highly
iodomethanes (produced during chloramina- exposed to diesel (Attfield et  al, 2012). Overall,
tion), or halonitromethanes may contribute to the evidence is not conclusive and a 2013 IARC
the toxicity of drinking water to a larger extent working group classified the evidence on air
than trihalomethanes. pollution and bladder cancer as limited (IARC,
Several studies have examined long-​ term 2013b).
exposure to chlorination byproducts by combin-
ing residents’ information from questionnaires Voiding Frequency and Urine pH
with information from water utilities. Positive Voiding frequency affects the retention time
associations between chlorinated drinking water of urine in the bladder. Experimental evidence
and bladder cancer have been reported (Cantor has found low voiding frequency is associated
et  al, 1987). A  study examining ozonization, an with higher formation of DNA adducts to 4-​
alternative to chlorination, found a decreased risk aminobiphenyl (Kadlubar et  al, 1991), an aro-
in subjects administered ozonated water com- matic amine present in tobacco smoke. The
pared to those using chlorinated water (Chevrier strongest epidemiologic evidence comes for a
561

Urinary Bladder Cancer 561

case-​control study in Spain that found a con- the malignancies with evidence of well-​defined
sistent inverse trend in bladder cancer risk with occupational exposures, particularly to aromatic
increasing nighttime voiding frequency in both amines, PAHs, diesel engine exhaust, and MWFs,
sexes (Silverman et  al, 2008). Voiding at least and occupational exposures may play a substantial
two times per night was associated with an role in perhaps 10% of bladder cancers.
approximate 50% reduction in risk. The same Consumption of fresh fruits and vegetables
study showed that the relative risk of smok- and increased total fluid intake had been associ-
ing and bladder cancer was lower among those ated with a lower risk of this cancer, but recent
who voided at least twice per night, indicating evidence is inconsistent. Evidence from large
that increased urination frequency may dimin- cohort studies does not support an association
ish the effect of urinary carcinogens on bladder of bladder cancer risk with heavy coffee con-
cancer risk. sumption. There is clear evidence that artificial
Urine acidity has been shown in experimen- sweeteners are not associated with an excess risk.
tal studies to influence the metabolism of blad- Voiding frequency affects the retention time of
der carcinogens such as aromatic amines and the urine in the bladder, and there is a suggestion
the formation of DNA adducts. Among workers that increased nighttime voiding may be asso-
exposed to benzidine, low urine pH was associ- ciated with decreased risk. There is consistent
ated with higher levels of free urinary benzidine evidence from large cohort studies on a reduced
and 10-​fold higher levels of DNA adducts in bladder cancer risk associated with leisure-​time
urothelial cells. Diet is one of the most impor- physical activity.
tant determinants of urine acidity, with fruits The association of diabetes with cancer has
and vegetables contributing to alkalinizing urine been suggested for more than 50  years, and
pH, while consumption of meat, fish, and dairy evidence from large cohort studies indicates
products contributes to lower urine pH. In a an increased risk for bladder cancer among
large case-​control study, subjects with a consist- patients with type 2 diabetes. Regarding specific
ently acidic urine pH (equal or less than 6.0) had medications, both the suggested decreased risk
an approximately 50% increased bladder cancer with barbiturates, certain analgesics, and anti-​
risk compared to those with higher pH urine. inflammatory drugs, and the detrimental role of
In addition, the effect of smoking on bladder other analgesics, in particular acetaminophen,
cancer was stronger among subjects with acidic await confirmation. Cyclophosphamide, most
urine, indicating that urine acidity may modify commonly used as an anticancer agent, has been
smoking-​related bladder cancer risk (Alguacil associated with an increased risk. Pioglitazone,
et al, 2011). an antidiabetic drug, and aristolochic acid used
in Chinese medicinal herbs are associated with
CONCLUSION urothelial cancer. An infectious etiology of blad-
Bladder cancer has notable descriptive epide- der cancer is suggested by the excess risk associ-
miological patterns with a higher frequency in ated with cystitis, with nonspecific urinary tract
men, more frequently diagnosed in higher-​than infections, and with S.  haematobium. However
lower-​income countries, and with increasing convincing evidence only exists for infection
trends in incidence in past decades although risk with S. haematobium.
has been generally moderately decreasing or is Environmental exposures for bladder cancer
stable in recent years. Urothelial cell (transitional include arsenic in drinking water, and there is
cell) carcinoma is the dominant histologic type emerging evidence that disinfection byproducts
in industrialized countries, while squamous cell in drinking water and air-​pollution may be asso-
carcinoma is a common histologic form in devel- ciated with bladder cancer.
oping countries with a high prevalence of S. hae- The urinary bladder is one of the few
matobium. A summary of the evidence is shown organ sites for which the candidate gene
in Table 22-​2. approach to identify common susceptibil-
Tobacco is by far the main cause of bladder ity loci has proved to be successful due to the
cancer, responsible for about one-​to two-​thirds of knowledge of bladder cancer carcinogenic
bladder cancer cases in different parts of the world. pathways and polymorphisms in drug and car-
Black, in comparison to blond tobacco smoking cinogen metabolism genes, including NAT2
has been associated with a higher risk. It is one of and GSTM1. Interestingly, GWAS using an
562

TABLE 22-​2   EPIDEMIOLOGY OF URINARY BLADDER CANCER:

RISK FACTOR SUMMARY*

Risk factor Direction of effect

Well-​Confirmed risk factors

Cigarette smoking ↑↑
Genetic susceptibility ↑
Infestation with Schistosoma haematobium ↑↑
Arsenic exposure, drinking water ↑
Occupational exposures: aromatic amines and dyes (4-​aminobiphenyl, ↑↑
benzidine, 2-​napthylamine, ortho-​toluidine)
Occupations and industries: Painting, rubber, aluminum production, ↑↑
auramine production, magenta production
Medications: Cyclophosphamide, chlornaphazine, phenacetin, aristolochic ↑
acid (also through diet), pioglitazone
Family history ↑↑
Ionizing Radiation (x-​radiation, gamma-​radiation) ↑

Probable Relationship Exists, Based on Substantial Data

Disinfection byproducts, drinking water ↑
Occupational exposures: diesel engine exhaust, 4-​chloro-​ortho-​toluidine, coal ↑
tar pitch, soot, tetrachloroethylene, metal working fluids
Occupations and industries: barbers and hairdressers, dry cleaning, printing, ↑
textile
Physical activity ↓
Pipes, cigars ↑
Diabetes ↑

Weak, if Any Relationship Exists, Based on Substantial Data

Parity ↓
Milk, diet ↓
Total fluid intake ↓

Inconsistent Findings or Limited Study to Date

Heavy coffee consumption -​
Air-​pollution -​
Hair dyes -​
Other diet (cereals, fruits, vegetables, meat, poultry, fish eggs, tea, alcohol, -​
sweeteners and other)
Obesity -​
Infections (other than S. Haematobium) ↑
Medications: barbiturates, NSAIDS ↓

*Arrows indicate the approximate magnitude of the relationship:  ↑, slight to moderate increase in risk; ↑↑, moderate to large
increase in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk
563

Urinary Bladder Cancer 563

agnostic approach have also identified vari- drugs and other analgesic use and bladder can-
ants in or nearby genes involved in pathways cer in northern New England. Int J Cancer
previously known to be involved in bladder 2013;132:162–​73.
carcinogenesis, including carcinogen metab- Baris D, Waddell R, Beane Freeman LE, Schwenn M,
olism, urinary excretion, and oncogenes and Colt JS et  al. Elevated bladder cancer in north-
tumor suppressors, for example, specifically ern New England: the role of drinking water and
genetic variants at UGT1A6 and CYP1A2. arsenic. J Natl Cancer Inst 2016;108.
Large-​scale tumor genomic projects such as Bedwani R, Renganathan E, El Kwhsky F, Braga C,
Abu Seif HH, Abul Azm T et al. Schistosomiasis
TCGA and ICGC have greatly facilitated prog-
and the risk of bladder cancer in Alexandria,
ress in defining a novel molecular taxonomy
Egypt. Br J Cancer 1998;77: 1186–​89.
of bladder cancer, providing new insights into
Boice JD Jr, Engholm G, Kleinerman RA, Blettner M,
the molecular mechanism of bladder cancer. Stovall M, Lisco H et al. Radiation dose and sec-
The development of promising therapies such ond cancer risk in patients treated for cancer of
as immunotherapy with checkpoint inhibitors the cervix. Radiat Res 1988;116:3–​55.
promises to translate into important improve- Boniol M, Koechlin A, Świątkowska B, Sorahan T,
ments in survival for advanced disease. Wellmann J, Taeger D et  al. Cancer mortality in
cohorts of workers in the European rubber manu-
REFERENCES facturing industry first employed since 1975. Ann
Alexandrov LB, Nik-​Zainal S, Wedge DC, Aparicio Oncol 2016;27:933–​41.
SA, Behjati S, Biankin AV et  al. Signatures of Bosetti C, Boffetta P, La Vecchia C. Occupational expo-
mutational processes in human cancer. Nature sures to polycyclic aromatic hydrocarbons, and res-
2013;500:415–​21. piratory and urinary tract cancers:  a quantitative
Alexandrov LB, Ju YS, Haase K, Van Loo P, review to 2005. Ann Oncol 2007 Mar;18(3):431û46.
Martincorena I, Nik-​Zainal S et  al. Mutational Brennan P, Bogillot O, Cordier S, Greiser E, Schill W,
signatures associated with tobacco smoking in Vineis P et al. Cigarette smoking and bladder can-
human cancer. Science 2016;354:618–​22. cer in men:  a pooled analysis of 11 case-​control
Alguacil J, Kogevinas M, Silverman DT, Malats N, Real studies. Int J Cancer 2000;86:289–​94.
FX, García-​Closas M et al. Urinary pH, cigarette Brinkman M, Buntinx F, Muls E, Zeegers MP. Use of
smoking and bladder cancer risk. Carcinogenesis selenium in chemoprevention of bladder cancer.
2011;32:843–​7. Lancet Oncol 2006;7:766–​74.
Antoni S, Ferlay J, Soerjomataram I, Znaor A, Jemal Büchner FL, Bueno-​ de-​
Mesquita HB, Ros MM,
A, Bray F. Bladder cancer incidence and mortal- Kampman E, Egevad L, Overvad K et  al.
ity: a global overview and recent trends. Eur Urol Consumption of vegetables and fruit and the risk
2017;71:96–​108. of bladder cancer in the European Prospective
Arlt VM, Stiborová M, vom Brocke J, Simões ML, Investigation into Cancer and Nutrition. Int J
Lord GM, Nortier JL et al. Aristolochic acid muta- Cancer 2009;125):2643–​51.
genesis: molecular clues to the aetiology of Balkan Buckland G, Ros MM, Roswall N, Bueno-​de-​Mesquita
endemic nephropathy-​associated urothelial can- HB, Travier N, Tjonneland A et  al. Adherence
cer. Carcinogenesis 2007;28:2253–​61. to the Mediterranean diet and risk of bladder
Augustsson K, Skog K, Jagerstad M, Dickman cancer in the EPIC cohort study. Int J Cancer
PW, Steineck G. Dietary heterocyclic amines 2014;134:2504–​11.
and cancer of the colon, rectum, bladder, and Cantor KP, Hoover R, Hartge P, Mason TJ, Silverman
kidney:  a population-​ based study. Lancet DT, Altman R et al. Bladder cancer, drinking water
1999;353:703–​7. source, and tap water consumption: a case-​control
Baastrup R, Sørensen M, Balstrøm T, Frederiksen study. J Natl Cancer Inst 1987;79:1269–​79.
K, Larsen CL, Tjønneland A et  al. Arsenic in Carreón T, Hein MJ, Viet SM, Hanley KW, Ruder AM,
drinking-​water and risk for cancer in Denmark. Ward EM. Increased bladder cancer risk workers
Environ Health Perspect 2008;116:231–​7. exposed to o-​toluidine and aniline:  a reanalysis.
Balbás-​Martínez C, Sagrera A, Carrillo-​ de-​
Santa-​ Occup Environ Med 2010;67:348–​50.
Pau E, Earl J, Márquez M, Vazquez M, et  al. Carreon T, Ruder AM, Schulte PA, Hayes RB,
Recurrent inactivation of STAG2 in bladder can- Rothman N, Waters M et  al. NAT2 slow acety-
cer is not associated with aneuploidy. Nat Genet lation and bladder cancer in workers exposed to
2013;45:1464–​9. benzidine. Int J Cancer 2006;118:161–​8.
Baris D, Karagas MR, Koutros S, Colt JS, Johnson A, Case RA, Hosker ME, McDonald DB, Pearson JT.
Schwenn M et al. Nonsteroidal anti-​inflammatory Tumours of the urinary bladder in workmen
564

564 Textbook of Cancer Epidemiology

engaged in the manufacture and use of certain Dietrich K, Demidenko E, Schned A, Zens MS,
dyestuff intermediates in the British chemical Heaney J, Karagas MR. Parity, early menopause
industry. Part I.  The role of aniline, benzidine, and the incidence of bladder cancer in women: a
alpha-​naphthylamine, and beta-​naphthylamine. case-​control study and meta-​ analysis. Eur J
1954. Br J Ind Med 1993;50: 389–​411. Cancer 2011;47:592–​9.
Castaño-​Vinyals G, Cantor KP, Malats N, Tardon A, Doll R, Peto R, Wheatley K, Gray R, Sutherland
Garcia-​Closas R, Serra C, Carrato A, Rothman I. Mortality in relation to smoking:  40  years’
N, Vermeulen R, Silverman D, Dosemeci M, observations on male British doctors. Br Med J
Kogevinas M. Air pollution and risk of urinary 1994;309:901–​11.
bladder cancer in a case-​control study in Spain. Espejo-​Herrera N, Cantor KP, Malats N, Silverman
Occup Environ Med 2008 Jan;65(1):56–​60. Epub DT, Tardón A, García-​Closas R et  al. Nitrate in
2007 Jul 18. drinking water and bladder cancer risk in Spain.
Castells X, Karanović S, Ardin M, Tomić K, Xylinas Environ Res 2015;37:299–​307.
E, Durand G et al. Low-​coverage exome sequenc- Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser
ing screen in formalin-​fixed paraffin-​embedded S, Mathers C, Rebelo M, Parkin DM, Forman D,
tumors reveals evidence of exposure to car- Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence
cinogenic aristolochic acid. Cancer Epidemiol and Mortality Worldwide: IARC CancerBase No.
Biomarkers Prev 2015;24:1873–​81. 11 [Internet]. Lyon, France: International Agency
Chavan S, Bray F, Lortet-​Tieulent J, Goodman M, for Research on Cancer; 2013. Available from:
Jemal A. International variations in bladder cancer http://globocan.iarc.fr, accessed 2017.
incidence and mortality. Eur Urol 2014;66:59–​73. Figueroa JD, Koutros S, Colt JS, Kogevinas M, Garcia-​
Chevrier C, Junod B, Cordier S. Does ozonation of Closas M, Real FX et  al. Modification of occupa-
drinking water reduce the risk of bladder cancer? tional exposures on bladder cancer risk by common
Epidemiology 2004;15:605–​14. genetic polymorphisms. J Natl Cancer Inst 2015;107.
Chiou HY, Chiou ST, Hsu YH, Chou YL, Tseng CH, Figueroa JD, Middlebrooks CD, Banday AR, Ye Y,
Wei ML, Chen CJ. Incidence of transitional Garcia-​Closas M, Chatterjee N et al. Identification
cell carcinoma and arsenic in drinking water:  a of a novel susceptibility locus at 13q34 and refine-
follow-​up study of 8,102 residents in an arseniasis-​ ment of the 20p12.2 region as a multi-​ signal
endemic area in northeastern Taiwan. Am J locus associated with bladder cancer risk in indi-
Epidemiol 2001;153:411–​18. viduals of European ancestry. Hum Mol Genet
Choi Y, Giovannucci E, Lee JE. Glycaemic index and gly- 2016;25:1203–​14.
caemic load in relation to risk of diabetes-​related can- Figueroa JD, Ye Y, Siddiq A, Garcia-​ Closas M,
cers: a meta-​analysis. Br J Nutr 2012;108):1934–​47. Chatterjee N, Prokunina-​Olsson L et al. Genome-​
Chow WH, Lindblad P, Gridley G, Nyrén O, wide association study identifies multiple loci
McLaughlin JK. Linet MS et  al. Risk of urinary associated with bladder cancer risk. Hum Mol
tract cancers following kidney or ureter stones. J Genet 2014;23;1387–​98.
Natl Cancer Inst 1997;89:1453–​57. Fortuny J, Kogevinas M, Chang-​Claude J, Gonzalez
Clavel J, Cordier S, Boccon-​ Gibod L, Hemon D. CA, Hours M, Jockel KH et al. Tobacco, occupa-
Tobacco and bladder cancer in males:  increased tion and non-​transitional-​cell carcinoma of the
risk for inhalers and smokers of black tobacco. Int bladder: an international case-​control study. Int J
J Cancer 1989;44:605–​10. Cancer 1999;801:44–​6.
Colt JS, Karagas MR, Schwenn M, Baris D, Johnson Fortuny J, Kogevinas M, Garcia-​Closas M, Real FX,
A, Stewart P et  al. Occupation and bladder can- Tardón A, Garcia-​Closas R et  al. Use of anal-
cer in a population-​based case-​control study in gesics and NSAIDs, genetic predisposition and
Northern New England. Occup Environ Med bladder cancer risk in Spain. Cancer Epidemiol
2011;68:239–​49. Biomarkers Prev 2006;15:1696–​1702.
Daugherty SE, Lacey JV Jr, Pfeiffer RM, Park Y, Freedman ND, Abnet CC, Caporaso NE, Fraumeni JF
Hoover RN, Silverman DT. Reproductive factors Jr, Murphy G, Hartge P et al. Impact of changing
and menopausal hormone therapy and bladder US cigarette smoking patterns on incident can-
cancer risk in the NIH-​AARP Diet and Health cer:  risks of 20 smoking-​related cancers among
Study. Int J Cancer 2013;133:462–​72. the women and men of the NIH-​AARP cohort.
Derby LE, Jick H. Acetaminophen and renal and blad- Int J Epidemiol 2016;45:846–​56.
der cancer. Epidemiology 1996;7:358–​62. Freedman ND, Silverman DT, Hollenbeck AR,
Dietrich H, Dietrich, B. Ludwig Rehn (1849–​ Schatzkin A, Abnet CC. Association between
1930):  pioneering findings on the aetiology of smoking and risk of bladder cancer among men
bladder tumours. World J Urol 2001;19:151–​3. and women. JAMA 2011;306:737–​45.
56

Urinary Bladder Cancer 565

Fu YP, Kohaar I, Moore LE, Lenz P, Figueroa JD, Tang Hemminki K, Bermejo JL, Ji J, Kumar R. Familial
W et  al. The 19q12 bladder cancer GWAS sig- bladder cancer and the related genes. Curr Opin
nal: association with cyclin E function and aggres- Urol 2011;21:386–​392.
sive disease. Cancer Res 2014;74:5808–​18. Henley SJ, Thun MJ. Use of permanent hair dyes and
Fu YP, Kohaar I, Rothman N, Earl J, Figueroa JD, Ye Y bladder-​cancer risk. Int J Cancer 2001;94:903–​6.
et al. Common genetic variants in the PSCA gene Holick CN, De Vivo I, Feskanich D, Giovannucci
influence gene expression and bladder cancer risk. E, Stampfer M, Michaud DS. Intake of fruits
Proc Natl Acad Sci U S A 2012;109:4974–​9. and vegetables, carotenoids, folate, and vita-
Gago-​Dominguez M, Castelao JE, Yuan JM, Yu MC, mins A, C, E and risk of bladder cancer among
Ross RK. Use of permanent hair dyes and bladder women (United States). Cancer Causes Control
cancer risk. Int J Cancer 2001;91:575–​9. 2005;16:1135–​45.
Garcia-​Closas M, Hein DW, Silverman D, Malats N, Hopenhayn-​Rich C, Biggs ML, Fuchs A, Bergoglio R,
Yeager M, Jacobs K et al. A single nucleotide pol- Tello EE, Nicolli H et al. Bladder cancer mortal-
ymorphism tags variation in the arylamine N-​ ity associated with arsenic in drinking water in
acetyltransferase 2 phenotype in populations of Argentina. Epidemiology 1996;7:117–​24.
European background. Pharmacogenet Genomics Huang YK, Huang YL, Hsueh YM, Yang MH, Wu
2011;21:231–​6. MM, Chen SY et  al. Arsenic exposure, urinary
Garcia-​Closas M, Malats N, Silverman D, Dosemeci arsenic speciation, and the incidence of urothelial
M, Kogevinas M, Hein DW et al. NAT2 slow acet- carcinoma: a twelve-​year follow-​up study. Cancer
ylation, GSTM1 null genotype, and risk of bladder Causes Control 2008;19:829–​39.
cancer:  results from the Spanish Bladder Cancer International Agency for Research on Cancer (IARC).
Study and meta-​analyses. Lancet 2005;366:649–​59. Schistosoma, Liver Flukes and Helicobacter
Garcia-​Closas M, Rothman N, Figueroa JD, pylori. IARC Monographs on the Evaluation
Prokunina-​ Olsson L, Han SS, Baris D et  al. of Carcinogenic Risks to Humans, Vol. 61.
Common genetic polymorphisms modify the Lyon: IARC, 1994.
effect of smoking on absolute risk of bladder can- International Agency for Research on Cancer (IARC).
cer. Cancer Res 2013;73:2211–​20. Tobacco Smoke and Involuntary Smoking. IARC
Garcia-​Closas M, Ye Y, Rothman N, Figueroa JD, Monographs on the Evaluation of Carcinogenic
Malats N, Dinney CP et al. A genome-​wide asso- Risks to Humans, Vol. 83. Lyon: IARC, 2004a.
ciation study of bladder cancer identifies a new International Agency for Research on Cancer
susceptibility locus within SLC14A1, a urea trans- (IARC). Some Drinking Water Disinfectants
porter gene on chromosome 18q12.3. Hum Mol and Contaminants, Including Arsenic. IARC
Genet 2011;20:4282–​9. Monographs on the Evaluation of Carcinogenic
Giovannucci E, Harlan DM, Archer MC, Bergenstal Risks to Humans, Vol. 84. Lyon: IARC 2004b.
RM, Gapstur SM, Habel LA et  al. Diabetes International Agency for Research on Cancer (IARC).
and cancer:  a consensus report. Diabetes Care Some Aromatic Amines, Organic Dyes, and
2010;33:1674–​85. Related Exposures. IARC Monographs on the
Global Cancer Observatory (GCO). Cancer today. Evaluation of Carcinogenic Risks to Humans, Vol.
International Agency for Cancer Research. http://​ 99. Lyon: IARC, 2010.
gco.iarc.fr/​today/​home (accessed March 19, 2017). International Agency for Research on Cancer (IARC).
Grotenhuis AJ, Dudek AM, Verhaegh GW, Witjes JA, Pharmaceuticals. IARC Monographs on the
Aben KK, van der Marel SL et al. Prognostic rele- Evaluation of Carcinogenic Risks to Humans, Vol.
vance of urinary bladder cancer susceptibility loci. 100A. Lyon: IARC, 2012.
PLoS One 2014;9:e89164. International Agency for Research on Cancer
Guo G, Sun X, Chen C, Wu S, Huang P, Li Z et  al. (IARC). Some Drugs and Herbal Products. IARC
Whole-​genome and whole-​exome sequencing of Monographs on the Evaluation of Carcinogenic
bladder cancer identifies frequent alterations in Risks to Humans, Vol. 108. Lyon: IARC, 2013a.
genes involved in sister chromatid cohesion and International Agency for Research on Cancer (IARC).
segregation. Nat Genet 2013;45:1459–​63. Outdoor Air Pollution. IARC Monographs on the
Hartge P, Harvey EB, Linehan WM, Silverman DT, Evaluation of Carcinogenic Risks to Humans, Vol.
Sullivan JW, Hoover RN, et al. Unexplained excess 109. Lyon: IARC, 2013b.
risk of bladder cancer in men. J Natl Cancer Inst Kadlubar FF, Dooley KL, Teitel CH, Roberts DW,
1990;82:1636–​40. Benson RW, Butler MA et  al. Frequency of uri-
Hecht SS. Tobacco carcinogens, their biomarkers and nation and its effects on metabolism, phar-
tobacco-​induced cancer. Nature reviews. Cancer macokinetics, blood haemoglobin adduct
2003;3:733–​44. formation, and liver and urinary bladder DNA
56

566 Textbook of Cancer Epidemiology

adduct levels in beagle dogs given the carcinogen Diet and Health Study. J Clin Endocrinol Metab
4-​aminobiphenyl. Cancer Res 1991;51:4371–​7. 2013;98:E497–​502.
Kadlubar FF, Butler MA, Kaderlik KR, Chou HC, Lang Lai MN, Wang SM, Chen PC, Chen YY, Wang JD.
NP. Polymorphisms for aromatic amine metabo- Population-​based case-​control study of Chinese
lism in humans: relevance for human carcinogen- herbal products containing aristolochic acid
esis. Environ Health Perspect 1992;98:69–​74. and urinary tract cancer risk. J Natl Cancer Inst
Karagas MR, Tosteson TD, Morris JS, Demidenko E, 2010;102:179–​86.
Mott LA, Heaney J, Schned A. Incidence of tran- Larsson SC, Andersson SO, Johansson JE, Wolk A.
sitional cell carcinoma of the bladder and arsenic Fruit and vegetable consumption and risk of blad-
exposure in New Hampshire. Cancer Causes der cancer:  a prospective cohort study. Cancer
Control 2004;15:465–​72. Epidemiol Biomarkers Prev 2008a;17:2519–​22.
Key TJ, Appleby PN, Spencer EA, Travis RC, Allen Larsson SC, Andersson SO, Johansson JE, Wolk A.
NE, Thorogood M et  al. Cancer incidence in Cultured milk, yogurt, and dairy intake in rela-
British vegetarians. Br J Cancer 2009;101:192–​7. tion to bladder cancer risk in a prospective study
Kiemeney LA, Sulem P, Besenbacher S, Vermeulen of Swedish women and men. Am J Clin Nutr
SH, Sigurdsson A et  al. A sequence variant at 2008b;88:1083–​7.
4p16.3 confers susceptibility to urinary bladder Lerner SP, McConkey DJ, Hoadley KA, Chan KS, Kim
cancer. Nat Genet 2010;42:415–​9. WY, Radvanyi F et  al. Bladder cancer molecular
Kiemeney LA, Thorlacius S, Sulem P, Geller F, Aben taxonomy:  summary from a consensus meeting.
KK et  al. Sequence variant on 8q24 confers sus- Bladder Cancer 2016;2:37–​47.
ceptibility to urinary bladder cancer. Nat Genet Lewis JD, Habel LA, Quesenberry CP, Strom BL, Peng
2008;40:1307–​12. T, Hedderson MM et al. Pioglitazone use and risk
Knight A, Askling J, Granath F, Sparen P, Ekbom A. of bladder cancer and other common cancers in
Urinary bladder cancer in Wegener’s granulo- persons with diabetes. JAMA 2015;314:265–​77.
matosis: risks and relation to cyclophosphamide. Loomis D, Guyton KZ, Grosse Y, Lauby-​ Secretan
Ann Rheum Dis 2004;63:1307–​11. B, El Ghissassi F, Bouvard V et  al; International
Knowles MA, Hurst CD. Molecular biology of bladder Agency for Research on Cancer Monograph
cancer new insights into pathogenesis and clinical Working Group. Carcinogenicity of drinking cof-
diversity. Nat Rev Cancer 2015;15:25–​41. fee, mate, and very hot beverages. Lancet Oncol
Kogevinas M, Fernandez F, Garcia-​Closas M, Tardon 2016;17:877–​8.
A, Garcia-​Closas R, Serra C et al. Hair dye use is Malaveille C, Vineis P, Esteve J, Ohshima H, Brun G,
not associated with risk for bladder cancer:  evi- Hautefeuille A et al. Levels of mutagens in the urine
dence from a case-​control study in Spain. Eur J of smokers of black and blond tobacco correlate
Cancer 2006;42:1448–​54. with their risk of bladder cancer. Carcinogenesis
Kogevinas M, Sala M, Boffetta P, Kazerouni N, 1989;10:577–​86.
Kromhout H, Hoar-​ Zahm S. Cancer risk in Mannetje A, Kogevinas M, Chang-​Claude J, Cordier
the rubber industry. A  review of the recent epi- S, Gónzalez CA, Hours M et al. Occupation and
demiological evidence. Occup Environ Med bladder cancer in European women. Cancer
1998;55:1–​12. Causes Control 1999;10:209–​17.
Kogevinas M, ‘t Mannetje A, Cordier S, Ranft U, Marshall G, Ferreccio C, Yuan Y, Bates MN, Steinmaus
González CA, Vineis P et al. Occupation and blad- C, Selvin S, Liaw J, Smith AH. Fifty-​year study of
der cancer among men in western Europe. Cancer lung and bladder cancer mortality in Chile related
Causes Control 2003;14:907–​14. to arsenic in drinking water. J Natl Cancer Inst
Koutros S, Baris D, Fischer A, Tang W, Garcia-​Closas 2007;99:920–​8.
M, Karagas MR et  al. Differential urinary spe- Matsuda K, Takahashi A, Middlebrooks CD, Obara W,
cific gravity as a molecular phenotype of the Nasu Y, Inoue K et al. Genome-​wide association
bladder cancer genetic association in the urea study identified SNP on 15q24 associated with
transporter gene, SLC14A1. Int Natl J Cancer bladder cancer risk in Japanese population. Hum
2013;133:3008–​13. Mol Genet 2015;24:1177–​84.
Koutros S, Silverman DT, Alavanja MC, Andreotti McCredie M, Macfarlane GJ, Stewart J, Coates M.
G, Lerro CC, Heltshe S et al. Occupational expo- Second primary cancers following cancers of
sure to pesticides and bladder cancer risk. Int J the kidney and prostate in New South Wales
Epidemiol 2016;45:792–​805. (Australia), 1972–​ 91. Cancer Causes Control
Lai GY, Park Y, Hartge P, Hollenbeck AR, Freedman 1996;7:337–​44.
ND. The association between self-​reported dia- McGeoghegan D, Binks K. The mortality and can-
betes and cancer incidence in the NIH-​AARP cer morbidity experience of workers at the
567

Urinary Bladder Cancer 567

Capenhurst uranium enrichment facility 1946–​ Piper JM, Tonascia J, Matanoski GM. Heavy phenac-
95. J Radiol Prot 2000; 20:381–​401. etin use and bladder cancer in women aged 20 to
Meliker JR, Slotnick MJ, AvRuskin GA, Schottenfeld D, 49 years. N Engl J Med 1985;313:292–​5.
Jacquez GM, Wilson ML et al. Lifetime exposure Pira E, Piolatto G, Negri E, Romano C, Boffetta P,
to arsenic in drinking water and bladder cancer: a Lipworth L et  al. Bladder cancer mortality of
population-​based case-​control study in Michigan, workers exposed to aromatic amines:  a 58-​year
USA. Cancer Causes Control 2010;21:745–​57. follow-​up. J Natl Cancer Inst 2010;102:1096–​9.
Michaud DS, Spiegelman D, Clinton SK, Rimm EB, Poon SL, Huang MN, Choo Y, McPherson JR, Yu
Curhan GC, Willett WC et  al. Fluid intake and W, Heng HL et al. Mutation signatures implicate
the risk of bladder cancer in men. N Engl J Med aristolochic acid in bladder cancer development.
1999;340:1390–​97. Genome Med 2015;7:38.
Mondul AM, Weinstein SJ, Männistö S, Snyder K, Puente D, Hartge P, Greiser E, Cantor KP, King WD,
Horst RL, Virtamo J, Albanes D. Serum vita- González CA et al. A pooled analysis of bladder can-
min D and risk of bladder cancer. Cancer Res cer case-​control studies evaluating smoking in men
2010;70:9218–​23. and women. Cancer Causes Control 2006;17:71–​9.
Moore SC, Lee IM, Weiderpass E, Campbell PT, Pukkala E, Martinsen JI, Lynge E, Gunnarsdottir HK,
Sampson JN, Kitahara CM et  al. Association of Sparén P, Tryggvadottir L et  al. Occupation and
leisure-​time physical activity with risk of 26 types cancer:  follow-​up of 15  million people in five
of cancer in 1.44 million adults. JAMA Intern Med Nordic countries. Acta Oncol 2009;48:646–​790.
2016;176:816–​25. Rafnar T, Sulem P, Stacey SN, Geller F, Gudmundsson
Mucci LA, Hjelmborg J, Harris J, Czene K, Havelick J, Sigurdsson A et  al. Sequence variants at the
D, Scheike T et  al. Familial risk and heritabil- TERT-​CLPTM1L locus associate with many can-
ity of cancer in the Nordic Twin Cohorts. JAMA cer types. Nat Genet 2009;41:221–​7.
2016;315:68–​76. Rafnar T, Sulem P, Thorleifsson G, Vermeulen SH,
Murta-​Nascimento C, Silverman DT, Kogevinas M, Helgason H, Saemundsdottir J et  al. Genome-​
García-​Closas M, Rothman N, Tardón A et  al. wide association study yields variants at 20p12.2
Risk of bladder cancer associated with family his- that associate with urinary bladder cancer. Hum
tory of cancer: do low-​penetrance polymorphisms Mol Genet 2014;23:5545–​57.
account for the increase in risk? Cancer Epidemiol Rafnar T, Vermeulen SH, Sulem P, Thorleifsson G,
Biomarkers Prev 2007;16:1595–​600. Aben KK, Witjes JA et al. European genome-​wide
Nickerson ML, Dancik GM, Im KM, Edwards MG, association study identifies SLC14A1 as a new uri-
Turan S, Brown J et al. Concurrent alterations in nary bladder cancer susceptibility gene. Hum Mol
TERT, KDM6A, and the BRCA pathway in blad- Genet 2011;20:4268–​81.
der cancer. Clin Cancer Res 2014;20:4935–​48. Reulen RC, Kellen E, Buntinx F, Brinkman M, Zeegers
Nortier JL, Martinez MC, Schmeiser HH, Arlt MP. A meta-​analysis on the association between
VM, Bieler CA, Petein M et  al. Urothelial car- bladder cancer and occupation. Scand J Urol
cinoma associated with the use of a Chinese Nephrol Suppl 2008;218:64–​78.
herb (Aristolochia fangchi) N Engl J Med Richardson DB, Wing S, Keil A, Wolf S. Mortality
2000;342:1686–​92. among workers at Oak Ridge National Laboratory.
Olsen JH, Boice JD Jr, Jensen JP, Fraumeni JF Jr. Cancer Am J Ind Med 2013;56:725–​32.
among epileptic patients exposed to anticonvul- Roberts SA, Lawrence MS, Klimczak LJ, Grimm SA,
sant drugs. J Natl Cancer Inst 1989;81:803–​8. Fargo D, Stojanov P et  al. An APOBEC cytidine
Pedersen-​Bjergaard J, Ersboll J, Hansen VL, Sorensen deaminase mutagenesis pattern is widespread in
BL, Christoffersen K, Hou-​ Jensen K et  al. human cancers. Nat Genet 2013;45:970–​6.
Carcinoma of the urinary bladder after treatment Robles C, Viscidi R, Malats N, Silverman DT, Tardon
with cyclophosphamide for non-​Hodgkin’s lym- A, Garcia-​Closas R et al. Bladder cancer and sero-
phoma. N Engl J Med 1988;318:1028–​32. reactivity to BK, JC and Merkel cell polyoma-
Pelucchi C, Galeone C, Tramacere I, Bagnardi V, viruses:  the Spanish bladder cancer study. Int J
Negri E, Islami F et  al. Alcohol drinking and Cancer 2013;133:597–​603.
bladder cancer risk:  a meta-​analysis. Ann Oncol Romaguera D, Vergnaud AC, Peeters PH, van Gils
2012;23:1586–​93. CH, Chan DS, Ferrari P et  al. Is concordance
Peters S, Talaska G, Jönsson BA, Kromhout H, with World Cancer Research Fund/​ American
Vermeulen R. Polycyclic aromatic hydrocar- Institute for Cancer Research guidelines for can-
bon exposure, urinary mutagenicity, and DNA cer prevention related to subsequent risk of can-
adducts in rubber manufacturing workers. Cancer cer? Results from the EPIC study. Am J Clin Nutr
Epidemiol Biomarkers Prev 2008;17:1452–​9. 2012;96:150–​63.
568

568 Textbook of Cancer Epidemiology

Ros MM, Bas Bueno-​de-​Mesquita HB, Büchner FL, Steinmaus CM, Nunez S, Smith AH. Diet and bladder
Aben KK, Kampman E, Egevad L et  al. Fluid cancer:  a meta-​analysis of six dietary variables.
intake and the risk of urothelial cell carcino- Am J Epidemiol 2000;151:693–​702.
mas in the European Prospective Investigation Su CC, Lu JL, Tsai KY, Lian IeB. Reduction in arsenic
into Cancer and Nutrition (EPIC). Int J Cancer intake from water has different impacts on lung can-
2011;128:2695–​708. cer and bladder cancer in an arseniasis endemic area
Ros MM, Bueno-​ de-​
Mesquita HB, Kampman E, in Taiwan. Cancer Causes Control 2011;22:101–​8.
Aben KK, Büchner FL, Jansen EH et  al. Plasma Tang W, Fu YP, Figueroa JD, Malats N, Garcia-​Closas
carotenoids and vitamin C concentrations and M, Chatterjee N et  al. Mapping of the UGT1A
risk of urothelial cell carcinoma in the European locus identifies an uncommon coding variant that
Prospective Investigation into Cancer and affects mRNA expression and protects from blad-
Nutrition. Am J Clin Nutr 2012;96:902–​10. der cancer. Hum Mol Genet 2012;21:1918–​30.
Rosso S, Faggiano F, Zanetti R, Costa G. Social class ‘t Mannetje A, Pearce N. Bladder cancer risk in sales
and cancer survival in Turin, Italy. J Epidemiol workers: artefact or cause for concern? Am J Ind
Community Health. 1997 Feb;51(1):30–​4. Med 2006;49:175–​86.
Roswall N, Freisling H, Bueno-​ de-​
Mesquita Tuccori M, Filion KB, Yin H, Yu OH, Platt RW,
HB, Ros M, Christensen J, Overvad K et  al. Azoulay L. Pioglitazone use and risk of blad-
Anthropometric measures and bladder cancer der cancer:  population based cohort study. BMJ
risk: a prospective study in the EPIC cohort. Int J 2016;352:i1541.
Cancer 2014;135:2918–​29. Turati F, Pelucchi C, Galeone C, Decarli A, La Vecchia
Rothman N, Garcia-​Closas M, Chatterjee N, Malats C. Personal hair dye use and bladder cancer:  a
N, Wu X, Figueroa JD et  al. A multi-​ stage meta-​analysis. Ann Epidemiol 2014;24:151–​9.
genome-​wide association study of bladder cancer Veys CA. Bladder tumours in rubber workers:  a
identifies multiple susceptibility loci. Nat Genet factory study 1946–​ 1995. Occup Med (Lond)
2010;42:978–​84. 2004;54:322–​29.
Sala M, Cordier S, Chang-​Claude J, Donato F, Escolar-​ Villanueva CM, Cantor KP, Cordier S, Jaakkola JJ, King
Pujolar A, Fernandez F et al. Coffee consumption WD, Lynch CF et al. Disinfection byproducts and
and bladder cancer in non-​ smokers:  a pooled bladder cancer:  a pooled analysis. Epidemiology
analysis of case-​control studies in European coun- 2004;15:357–​67.
tries. Cancer Causes and Control 2000;11:925–​31. Villanueva CM, Cantor KP, Grimalt JO, Malats N,
Samanic C, Kogevinas M, Dosemeci M, Malats N, Silverman D, Tardon A et  al. Bladder cancer and
Real FX, Garcia-​Closas M et  al. Smoking and exposure to water disinfection by-​products through
bladder cancer in Spain:  effects of tobacco type, ingestion, bathing, showering and swimming pool
timing, environmental tobacco smoke, and attendance. Am J Epidemiol 2007;165: 148–​56.
gender. Cancer Epidemiol Biomarkers Prev Vineis P, Kogevinas M, Simonato L, Bogillot O, Brennan
2006;15:1348–​54. P, Boffetta P. Levelling-​off the risk of lung and blad-
Schoenig GP, Goldenthal EI, Geil RG, Frith CH, der cancer in heavy smokers: an analysis based on
Richter WR, Carlborg FW. Evaluation of the dose multicentric case-​control studies and a metabolic
response and in utero exposure to saccharin in the interpretation. Mutat Res 2000;463:103–​10.
rat. Food Chem Toxicol 1985;23:475–​90. Vineis P, Pirastu R. Aromatic amines and cancer.
Shapiro JA, Jacobs EJ, Thun MJ. Cigar smoking in men Cancer Causes Control 1997;8:346–​55.
and risk of death from tobacco-​related cancers. J Vizcaino AP, Parkin DM, Boffetta P, Skinner ME.
Natl Cancer Inst 2000;92:333–​37. Bladder cancer:  epidemiology and risk factors
SEER Cancer Statistics Review. National Cancer in Bulawayo, Zimbabwe. Cancer Causes Control
Institute, Bethesda, MA. https://​seer.cancer.gov/​stat- 1994;5:517–​22.
facts/​html/​urinb.html. Accessed March 19, 2017. Vlaanderen J, Straif K, Ruder A, Blair A, Hansen J, Lynge
Silverman DT, Alguacil J, Rothman N, Real FX, E et al. Tetrachloroethylene exposure and bladder
Garcia-​Closas M, Cantor KP et al. Does increased cancer risk: a meta-​analysis of dry-​cleaning-​worker
urination frequency protect against bladder can- studies. Environ Health Perspect 2014;122:661–​6.
cer? Int J Cancer 2008;123:1644–​8. Ward MH, Cantor KP, Riley D, Merkle S, Lynch CF.
Steinmaus C, Ferreccio C, Acevedo J, Yuan Y, Liaw J, Nitrate in public water supplies and risk of bladder
Durán V et al. Increased lung and bladder cancer cancer. Epidemiology 2003;14:183–​90.
incidence in adults after in utero and early-​life Weiss HA, Darby SC, Doll R. Cancer mortality follow-
arsenic exposure. Cancer Epidemiol Biomarkers ing X-​ray treatment for ankylosing spondylitis. Int
Prev 2014;23:1529–​38. J Cancer 1994;59:327–​38.
569

Urinary Bladder Cancer 569

Welk B, McIntyre A, Teasell R, Potter P, Loh E. Bladder from the Netherlands Cohort Study. Br J Cancer
cancer in individuals with spinal cord injuries. 2001a;85:977–​83.
Spinal Cord 2013;51:516–​21. Zeegers MP, Selen RF, Kleinjans JC, Goldbohm
World Cancer Research Fund International/​ RA, van den Brandt PA. Nitrate intake does not
American Institute for Cancer Research (WCRF/​ influence bladder cancer risk:  the Netherlands
AICR). Continuous Update Project Report: Diet, cohort study. Environ Health Perspect
Nutrition, Physical Activity and Bladder Cancer. 2006;114:1527–​31.
2015. Available at wcrf.org/​bladder-​cancer-​2015. Zhang YF, Xu Q, Lu J, Wang P, Zhang HW, Zhou L
Wu XR. Urothelial tumorigenesis: a tale of divergent et al. Tea consumption and the incidence of can-
pathways. Nat Rev Cancer 2005;5:713–​25. cer: a systematic review and meta-​analysis of pro-
Wu X, Ye Y, Kiemeney LA, Sulem P, Rafnar T, Matullo spective observational studies. Eur J Cancer Prev
G et al. Genetic variation in the prostate stem cell 2015;24:353–​62.
antigen gene PSCA confers susceptibility to uri- Zheng YL, Amr S, Saleh DA, Dash C, Ezzat S, Mikhail
nary bladder cancer. Nat Genet 2009;41:991–​5. NN et  al. Urinary bladder cancer risk factors in
Yang H, Yang K, Khafagi A, Tang Y, Carey TE, Opipari Egypt:  a multicenter case-​control study. Cancer
AW et al. Sensitive detection of human papilloma- Epidemiol Biomarkers Prev 2012;21:537–​46.
virus in cervical, head/​neck, and schistosomiasis-​ Zhou J, Kelsey KT, Giovannucci E, Michaud DS. Fluid
associated bladder malignancies. Proc Natl Acad intake and risk of bladder cancer in the Nurses’
Sci U S A 2005;102:7683–​88. Health Studies. Int J Cancer 2014;135:1229–​37.
Yang HY, Wang JD, Lo TC, Chen PC. Occupational Zhou J, Kelsey KT, Smith S, Giovannucci E, Michaud
exposure to herbs containing aristolochic acids DS. Lower urinary tract symptoms and risk of blad-
increases the risk of urothelial carcinoma in der cancer in men: results from the health profes-
Chinese herbalists. J Urol 2013;189:48–​52. sionals follow-​up study. Urology 2015;85:1312–​8.
Zeegers MP, Goldbohm RA, Brandt PA. Are reti- Zhou Y, Tian C, Jia C. A dose-​response meta-​analysis
nol, vitamin C, vitamin E, folate and carotenoids of coffee consumption and bladder cancer. Prev
intake associated with bladder cancer risk? Results Med 2012;55:14–​22.
570
571

23
Kidney Cancer
XUEHONG ZHANG, EUNYOUNG CHO, AND HANS-O
​ LOV ADAMI

K idney cancer in adults includes tumors aris-

ing from the renal parenchyma (also called
renal cell cancer), and the renal pelvis. Although
function of the contralateral kidney. Additional
tools for diagnosis and staging are contrast-​
enhanced computed tomography (CT), ultra-
almost all renal cell cancers are adenocarcino- sound, and magnetic resonance imaging.
mas, renal pelvis cancers are of the transitional Widespread use of modern high-​ sensitivity
cell types, which resemble that of bladder cancer imaging techniques such as CT might entail
(discussed in the previous chapter) and constitute detection of small tumors with a low malignant
less than 10% of all kidney cancer, and thus are potential.
not reviewed here. The majority of kidney cancer Treatment Radical or partial nephrectomy is the
in children is nephroblastoma (also called Wilms’ only curative treatment. Surgery is also used as
tumor), a malignant kidney tumor of embryonal palliative treatment of metastatic disease, as is
origin, constituting about 1% of all kidney can- radiotherapy. The success of chemotherapy in
cers. This tumor is not discussed here, either. We treating kidney cancer has been poor.
only addressed the etiology of renal cell cancer,
Prognosis The clinical course is variable. There
which represents about 90% of all kidney tumors
is a high mortality to incidence ratio for kidney
in adults. We used the term “kidney cancer” in
cancer compared to other urologic malignancies,
this chapter for convenience.
with survival rates relatively high in developed
regions. However, the survival rate was lower in
CLINICAL SYNOPSIS
less developed regions, where kidney cancers are
Histological Subtypes Kidney cancer comprises a
often detected in more advanced stages.
heterogeneous class of tumors arising from dif-
ferent cell types within the nephron. Advances Progress There has been only a small, if any,
in our understanding of the genetics underlying improvement in cure rates over the past decades.
the pathogenesis of renal cell neoplasms have led Further refinements in the treatment of kidney
to a histopathologic classification into five major cancer with immunotherapy, gene-​therapy tech-
histological subtypes:  conventional (clear cell; niques to modify tumor cells, creating tumor
nonpapillary) renal cell carcinoma (75% to 80%); vaccines, and antiangiogenesis modalities are
papillary (chromophilic) renal cell carcinoma needed.
(10% to 15%); chromophobe renal cell carci-
noma; collecting duct carcinoma; and unclassi- DESCRIPTIVE
fied renal cell carcinoma. EPIDEMIOLOGY
Kidney cancer is the 12th most common can-
Symptoms General symptoms are hematuria (50%
cer worldwide, and it was estimated to afflict
to 60% of the patients), abdominal pain (40%),
about 337,860 individuals worldwide in 2012,
and a palpable abdominal or flank mass (30% to
accounting for 2.4% of all cancers (Ferlay et al,
40%). Common nonspecific symptoms and signs
2015). The age-​standardized rates (ASRs) for kid-
(10% to 40%) are hypertension, weight loss, fever,
ney cancer vary across geographic regions (Figs.
malaise, night sweats, hypercalcemia, erythrocy-
23-​1, 23-​2, Table 23-1). About 59% of kidney
tosis, thrombocytosis, and hepatic dysfunction.
cancer cases occur in high-​ income countries,
Diagnosis Intravenous urography is often used with the highest rate observed in North America,
initially for the evaluation of hematuria and Australia/​New Zealand, and Europe, and lowest
572

572 Textbook of Cancer Epidemiology

China

Japan
Poland

Spain

Sweden
The Netherlands
United Kingdom

United States of America

0 1 2 3 4 5 6 7 8 9
ASR (world) per 100,000 person-years

FIGURE 23-​1  Age-​standardized (to the 2012 world population) incidence rates of kidney cancer among women.
Source: Ferlay et al, 2015.

in Africa and Asia (Figs. 23-​1, 23-​2, Table 23-1) in the early 60s (Figs. 23-​3, 23-​4). Kidney cancer
(Ferlay et al, 2015). The ASR of kidney cancer is incidence rates have been rising in most countries
almost 10 times higher in North America than in during the past three decades. The International
Africa in both men and women (Table 23-​1). The Agency for Research on Cancer (IARC) predicts
geographic variation in ASR may be due to dif- a 22% increase in kidney cancer cases from 2012
ferences in diagnostic intensity and autopsy rates to 2020, amounting to about 412,929 cases (Ferlay
as well as to environmental factors that are likely et al, 2015). These increasing rates may partly be
to play an important role in renal cell carcino- explained by the increased diagnostic intensity
genesis. In the United States, the ASR for kidney following the introduction of new imaging meth-
cancer differs by racial or ethnical groups, with ods such as ultrasound and CT.
higher rates observed in the black than the white
population (Siegel et al, 2015). GENETIC AND MOLECULAR
Kidney cancer affects men about twice more EPIDEMIOLOGY
often than women, with rates over 10 per 100,000
in men and over 5 per 100,000 in women in devel- Inherited Susceptibility
oped geographical regions (Ferlay et al, 2015). In Although most kidney cancers are sporadic, sev-
2012, the estimated numbers of new kidney can- eral rare genetic diseases are associated with kid-
cers were about 213,924 in men and 123,936 in ney cancer, including von Hippel-​Lindau (VHL)
women worldwide. The average age at diagnosis is syndrome, hereditary papillary renal carcinoma,

China

Japan
Poland

Spain

Sweden
The Netherlands
United Kingdom

United States of America

0 2 4 6 8 10 12 14 16 18
ASR (world) per 100,000 person-years

FIGURE 23-​2  Age-​standardized (to the 2012 world population) incidence rates of kidney cancer among men.
Source: Ferlay et al, 2015.
573

Kidney Cancer 573

germline mutations in affected patients have

TABLE 23-​1   AGE-STANDARDIZED RATES
been identified (Table 23-​2).
OF KIDNEY CANCER WITHIN SOME Von Hippel-​Lindau syndrome is a domi-
AREAS IN 2012 nantly inherited multisystemic disorder asso-
Area ASR (World), ASR ciated with significant morbidity (Maher
per 100,000 (World),
et  al, 2011). Patients are at risk of develop-
ing tumors in a number of organs:  kidneys—​
per 100,000
often multiple, bilateral tumors—​ pancreas,
Men Women adrenal glands, epididymis, eyes, spine, and
cerebellum (Choyke et  al, 1995; Maher et  al,
World 6.0 3.0 2011). The cumulative risk of kidney cancer
  More developed 12.6 6.2 is more than 70% by the age of 60, and renal
countries cell cancer is the most common cause of death
  Less developed 3.4 1.8 (Maher, 1996).
countries The VHL tumor suppressor gene, located on
Europe 12.3 5.9 chromosome 3p, is involved in both spontaneous
and hereditary kidney cancer and is inactivated
  Northern Europe 11.1 5.8
via several mechanisms, including mutation and
  Western Europe 13.4 6.5 silencing by DNA methylation (Kim & Kaelin,
  Central and Eastern 12.4 6.1 2004; Maher et  al, 2011). This gene is involved
Europe in the etiology of clear cell carcinoma, and loss
  Southern Europe 11.2 4.6 of VHL gene function entails increased expres-
The Americas 9.8 5.2 sion of angiogenetic factors such as vascular
endothelial growth factor (Li & Kaelin, 2011;
  Northern America 15.5 8.3
Maher et al, 2011). Overexpression of endothe-
  Central America 4.1 2.3 lial growth factor in vascularized tumors such as
  South America 5.1 2.7 kidney cancer may promote its growth and pro-
Australia/​New Zealand 12.6 6.1 gression (Jacobsen et al, 2004). The VHL tumor
suppressor protein (pVHL) plays a central role
Asia 3.8 1.9
also in the mammalian oxygen-​sensing pathway
Africa 1.4 1.1 through hypoxia-​inducible factor (HIF) (Kim &
Source: Ferlay et al, 2015.
Kaelin, 2004; Li & Kaelin, 2011). In the absence
of pVHL, HIF induces the expression of several
and Birt-​Hogg-​Dube syndrome (Kim & Kaelin, genes that are related to regulation of angiogen-
2004; Li & Kaelin, 2011; Lopez-​Beltran et  al, esis, cell growth, or cell survival. Other genetic
2009; Maher et  al, 2011). The genes underlying abnormalities associated with kidney cancer
each of these conditions have been cloned, and progression include loss of heterozygosity of
New cases per 100,000

60
50
person-years

40
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age
China Japan Poland Spain Sweden
The Netherlands United Kingdom United States of America

FIGURE 23-​3  Age-​specific incidence rates of kidney cancer among women.

Source: Ferlay et al, 2015.
574

574 Textbook of Cancer Epidemiology

120

New cases per 100,000

100

person-years
80

60

40

20

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 23-​4  Age-​specific incidence rates of kidney cancer among men.

Source: Ferlay et al, 2015.

chromosome 14q (Li & Kaelin, 2011; Van Poppel cancer risk. Most studies used a candidate gene
et al, 2000). approach and identified genes or single nucleo-
Hereditary papillary renal cell carcinoma, char- tide polymorphisms (SNPs) in certain pathways
acterized by multifocal, bilateral tumors, has an potentially involved in renal carcinogenesis.
autosomal dominant inheritance pattern, distinct Among these genes, glutathione S-​ transferase
from that of other hereditary kidney cancers. It is (GST) polymorphisms (GSTM1, GSTT1, and
caused by a germline mutation of the c-​MET proto-​ GSTP1) received most attention. The GST
oncogene on chromosome 7q or loss of function enzymes are expressed in the kidney and play an
mutations of the fumarate hydratase gene. Apart important role in cellular protection against oxi-
from that, there is no loss of alleles of the short arm dative stress and detoxify toxic chemicals (Hayes &
of chromosome 3 in papillary renal cell carcinoma; Strange, 2000). The GST genes per se appear not
trisomies mainly of chromosomes 7, 16, and 17 and, to be associated with kidney cancer risk, but the
in men, loss of chromosome Y are found. Fumarate associations with intake of cruciferous vegetables
hydratase mutations are also associated with an or tobacco smoke differed by GST status (Moore
increased risk of cutaneous leiomyomata and uter- et al, 2007).
ine fibroids. Birt-​Hogg-​Dube syndrome, caused by NAT2 (N-​acetyl transferase 2), is involved in
germline loss of function mutations of the BHD metabolism of chemicals generated from tobacco
gene, is characterized by fibrofolliculomas, lung smoke and cooked meat. Association with smok-
cysts, and a spectrum of renal carcinomas of vary- ing was much stronger among slow acetylators
ing histological subtypes (chromophobe, oncocy- of NAT2 (Semenza et  al, 2001). Limited stud-
toma, clear-​cell, or papillary) (Kim & Kaelin, 2004; ies have examined the genes involved in DNA
Lopez-​Beltran et al, 2009; Van Poppel et al, 2000). repairs, folate metabolism, vitamin D metabo-
Studies also evaluated the relationship lism, and hormone pathway, cell-​cycle control
between common genetic variants and kidney and apoptosis, and hypertension (Andreotti

TABLE 23-​2   GENES PREDISPOSING TO INHERITED RENAL CELL CARCINOMA

Inherited disease Histology Gene Locus

Von Hippel-​Lindau disease Clear cell VHL 3p25

Hereditary papillary renal cell Papillary (Type I) c-​MET 7q31
carcinoma
Hereditary leiomyomata and Papillary (Type II) Fumarate hydratase (FH) 1q42
renal cell carcinoma
Birt-​Hogg-​Dube syndrome All (esp chromophobe) BHD 17p11
57

Kidney Cancer 575

et al, 2010; Brennan et al, 2007; Dong et al, 2009; smoking (Muscat et  al, 1995)  as well as heavy
Karami et  al, 2009; Moore et  al, 2007). Because cigar smoking (Yuan et  al, 1998). Depending
of the relative small sample size and inconsist- on the prevalence of smoking, the proportion of
ent subgroup findings, large consortium studies, kidney cancer that could be attributed to ciga-
with high-​throughput technologies will shed new rette smoking is between 21% and 30% among
light on the kidney cancer etiology by discovery men, and between 9% and 24% among women
of novel common genetic variants and detections (McLaughlin et al, 1995; Yuan et al, 1998).
of interactions with environmental risk factors Several biological mechanisms of cigarette
(Chow & Devesa, 2008). smoking in renal carcinogenesis have been sug-
Families with kidney cancer have also been gested. First, most of the constituents in ciga-
characterized by constitutionally balanced trans- rette smoke are metabolized or excreted through
locations between chromosomes 3 and 6 or 8 the urinary tract. Animal studies showed that
and between chromosomes 2 and 3 (Van Poppel smoking-​related nitroso compounds, especially
et al, 2000). N-​nitrosodimethylamine, can cause kidney can-
A meta-​analysis observed a twofold-​increased cer (IARC, 1986). N-​ nitrosodimethylamine-​
risk of kidney cancer if a first-​degree relative was induced clear cell renal tumors were identified
affected with the disease (Clague et al, 2009). The with VHL mutations in animal studies (Gnarra,
interactions between environmental risk factors 1998; Shiao et al, 1998). Second, cigarette smok-
and genetic susceptibility are thought to impact ing may increase kidney cancer risk via chronic
the risk of sporadic kidney cancer. tissue hypoxia due to carbon monoxide exposure
(Sharifi & Farrar, 2006). Third, deletions in chro-
Somatic Events mosome 3p, a frequent site of genetic alterations
Somatic mutation of the VHL gene is present in in kidney cancer, was more commonly detected
about 50% of sporadic clear cell kidney cancer in cultured peripheral blood lymphocytes from
and methylation of the gene in another 10%–​20% kidney cancer patients after being treated with
(Kim & Kaelin, 2004). One report identified a benzo[α] pyrene diol epoxide, a major constitute
“hotspot” for mutations of the VHL gene associ- of smoking (Zhu et al, 2008).
ated with exposure to trichloroethylene (Brauch
et  al, 1999). Limited studies have evaluated the Diet
associations between VHL gene inactivation and Considerable evidence from ecologic, case-​
major established environmental risk factors for control, and cohort studies indicates a role of
kidney cancer (Schouten et al, 2005; Smits et al, diet in renal carcinogenesis, however, to date,
2010; van Dijk et al, 2008). there is no convincing evidence of a causal link
with any specific food items or nutrients (WCRF/​
RISK FACTORS AICR, 2015).
Tobacco Foods
Cigarette smoking is an established cause of kid- Increasing fruit and vegetable consumption was
ney cancer, although the excess risks are gener- associated with a reduced risk of kidney cancer in
ally moderate (Humans, 2012b). Compared with a pooled analysis of 13 prospective cohort stud-
never smokers, the risk for kidney cancer among ies (Lee et al, 2009b), but a large European cohort
ever smokers is increased about 50% for men study observed no association (Weikert et  al,
and 20% for women (Hunt et al, 2005). Despite 2006). The IARC classified consumption of pro-
the modest positive associations, a clear dose–​ cessed meat as Group 1 carcinogen (carcinogenic
response relationship was observed for increas- to humans) and red meat as Group  2A (proba-
ing cigarette smoking in both men and women bly carcinogenic to humans) carcinogens largely
(Hunt et  al, 2005). In addition, smoking cessa- based on data for colorectal cancer (Bouvard et al,
tion is associated with decreased relative risk 2015). Both red and processed meat products
observable after about 10 years (Yuan et al, 1998). could contain high levels of heterocyclic amines
Moreover, positive association with passive and DNA-​damaging N-​nitroso compounds (i.e.,
smoking was also reported (Hu & Ugnat, 2005). nitrites and nitrates) generated during processing
Few studies report on tobacco products other or cooking of meat. Experimental models with
than cigarettes, but positive associations were heterocyclic amines demonstrated DNA adducts
observed for smokeless tobacco, cigar, and pipe in different organs, including the kidney (Munro
576

576 Textbook of Cancer Epidemiology

et al, 1993). N-​nitroso compounds could induce foods. Epidemiologic studies on the relation with
renal cell tumors (Nogueira et al, 1993). In addi- kidney cancer have been limited and largely nega-
tion, a high amount of sodium is used as a preserv- tive (Hirvonen et al, 2010; Hogervorst et al, 2008;
ative for processed meat and high-​sodium diets Mucci et al, 2004; Pelucchi et al, 2015). Given its
are linked to high blood pressure (He et al, 2013), ubiquity in foods, monitoring its potential health
a risk factor for kidney cancer (see following). impact is desirable.
Despite these biological data, studies on red
meat and processed meat in relation to kidney Coffee, Tea, and Other Beverages
cancer have been inconclusive (Lee et al, 2008). Although beverage intake might potentially
One study suggests that red/​ processed meats influence kidney cancer risk via a diluting effect
might influence kidney cancer risk via mecha- of carcinogens, alterations of hormone levels, or
nisms related to certain compounds generated other mechanisms, epidemiological studies have
during cooking and that the associations might shown mixed results. It is believed that beverage
be modified by genetic susceptibility (Melkonian intake is at least not positively associated with
et al, 2016). It is desirable to investigate the dose-​ risk of kidney cancer. In fact, a pooled analysis
response relationship, method of cooking, degree of 13 prospective cohort studies observed signifi-
of doneness, and potential interactions and asso- cant inverse associations with coffee and tea con-
ciated mechanisms with genetic markers. Studies sumption (Lee et al, 2007b). Nonetheless, current
on specific types of fruits or vegetables, poultry, evidence is too limited to draw firm conclusions.
milk, margarine, oils, butter, fish, or dietary pat- No clear associations were observed for sweet-
terns are too limited for any definitive conclu- ened carbonated beverages (e.g., soda, cola),
sions (Daniel et al, 2013; Lee et al, 2006; Moore milk, or juice (Boyle et al, 2014; Lee et al, 2007b).
et al, 2007).
Drinking Water “Contamination”
Nutrients Inorganic arsenic in drinking water is well
The results for fat (e.g., vegetable, animal, dairy, absorbed in the gastrointestinal tract, and its
saturated, monounsaturated, polyunsaturated, metabolites are excreted in urine (Straif et  al,
trans fat, and cholesterol), protein, carbohydrate, 2009). These arsenic metabolites have sev-
and antioxidant nutrients common in fruits and eral genotoxic effects, including the induction
vegetables (e.g., carotene; vitamins A, C, and E; of chromosomal abnormalities, inhibition of
and carotenoids) are mixed (WCRF/​AICR, 2015). DNA repair, and an increase in cell proliferation
Fiber is an important component of fruits and (IARC, 2012a). Inorganic arsenic in drinking
vegetables, with a potential role in body weight water is classified as Group 1 carcinogen by the
and blood glucose control, as well as reducing IARC (2012a). Studies from Taiwan (Chen et al,
chronic inflammation (Slavin, 2008). The inverse 1992)  and Argentina (Hopenhayn-​ Rich et  al,
association reported in some case-​control stud- 1998), in areas with chronic exposure to arsenic
ies was also observed in a US cohort but not in a from drinking water, showed a dose-​related asso-
large European cohort (Allen et al, 2009; Daniel ciation with an increased risk of kidney can-
et al, 2013). A few studies found a positive asso- cer. Three other studies observed no significant
ciation between total energy intake and the risk associations (Baastrup et  al, 2008; Kurttio et  al,
of kidney cancer (Mellemgaard et al, 1996; Wolk 1999; Lewis et  al, 1999). The studies were rela-
et al, 1996a). It is difficult, however, to disentan- tively small, and the overall evidence is limited to
gle the individual effects of energy-​driving nutri- suggest that consuming drinking water that con-
ents such as protein and fat from the total energy tains arsenic increases the risk of kidney cancer.
intake because of the high correlations. If posi- Carcinogenic and mutagenic compounds found
tive associations are observed, it is warranted to in chlorinated drinking water have also raised
determine whether the increased risk is related concerns, but evidence is limited with kidney
to individual energy sources or confined to total cancer (IARC, 1991; Morris, 1995).
energy intake (Wolk et al, 1996b).
Artificial Sweeteners
Acrylamide Current evidence does not link saccharin, aspar-
The IARC classified acrylamide as a “probable” tame, and other artificial sweeteners to kidney can-
human carcinogen (Group 2A). Acrylamide was cer (Boyle et al, 2014), although one study found a
detected in commonly consumed fried and baked positive association in men (Wolk et al, 1996b).
57

Kidney Cancer 577

Alcohol Anthropometric Measures

In contrast to the positive associations between Both case-​control and cohort studies conducted
light to moderate alcohol consumption and over- worldwide have consistently shown that excess
all cancer risk (Cao et  al, 2015), alcohol con- body weight (often measured by BMI) is a causal
sumption was inversely associated with risk of risk factor for kidney cancer (Renehan et  al,
developing kidney cancer in a dose-​ response 2008; Wang & Xu, 2014). Overweight and obesity
manner in a pooled analysis of 12 prospective has been estimated to account for over 40% of
studies (Lee et al, 2007a) and two meta-​analyses kidney cancers in the United States and over 30%
(Bellocco et al, 2012; Song et al, 2012). Compared in Europe (Calle & Kaaks, 2004). Higher BMI
with nondrinking, alcohol consumption of 15 g/​ is strongly associated, in a dose-​response man-
day (about 1 drink) was associated with a 20% ner, with an increased kidney cancer risk in men
to 30% lower risk of kidney cancer. Significant and women. The results were robust regardless of
inverse associations were observed for individual whether weight was measured at different ages or
alcoholic beverages including beer, wine, and liq- adjusted for confounders (Bergstrom et al, 2001;
uor, in both women and men. In addition, these Renehan et  al, 2008). One quantitative meta-​
associations appeared not to vary appreciably by analysis of 27 studies observed a pooled relative
body mass index (BMI), history of hypertension, risk of 1.05 (95% CI, 1.04–​1.06) per unit increase
or smoking status. However, there is limited evi- in BMI for men (Ildaphonse et al, 2009), another
dence for higher levels of drinking (e.g., > 50 g/​ meta-​analysis of 28 studies a relative risk of 1.06
d or > 3 drinks/​d). A potential mechanism might (95% CI, 1.05–​1.07) per unit increase in BMI in
be that alcohol consumption enhances insulin women (Mathew et al, 2009).
sensitivity and is inversely associated with hyper- Limited studies have examined kidney can-
insulinemia and diabetes risk (Davies et al, 2002). cer risk in relation to waist circumference or
In addition, alcohol contains antioxidant pheno- waist-​to-​hip ratio (WHR) (Hughes et  al, 2009;
lic compounds and has a diuretic effect to reduce Luo et  al, 2007; Pischon et  al, 2006), or weight
exposure of kidney epithelial cells to carcino- changes (Lindblad et  al, 1994; van Dijk et  al,
genic solutes (Nicodemus et  al, 2004; Pelucchi 2004). Results were often positive, but not robust
et al, 2008). after adjusting for BMI, emphasizing the chal-
lenge to disentangle the influence of these factors
Reproductive Factors and Hormones from the effects of general obesity (BMI).
The gender difference in kidney cancer incidence The mechanisms via which obesity influences
suggests a role of hormonal factors. A  meta-​ kidney cancer development remain unclear, but
analysis of eight case-​control studies and six cohort several hypotheses have been proposed. These
studies showed that parity was associated with an include altered endocrine milieu (e.g., sex steroid
increased risk. Compared to nulliparous women, hormones, biologically active insulin-​like growth
risk increased 23% for parous women, and further factor), insulin resistance, hyperinsulinemia,
increased 36% comparing the highest with lowest altered inflammatory response, altered levels of
parity number (Guan et al, 2013). Early age at first adipokines, and lipid peroxidation and oxida-
birth was associated with a lower kidney cancer risk tive stress (Calle & Kaaks, 2004). In addition,
(Lee et al, 2009a). Studies on other reproductive or hypertension that may be an intermediate step in
hormonal factors, including oral contraceptive use the causal pathway between obesity and kidney
or duration, postmenopausal hormone use (HT) cancer could induce renal damage or could be
or duration, age at menarche, age at menopause, or associated with metabolic or functional changes
history of hysterectomy or oophorectomy, are lim- within the renal tubules that increase the kidney’s
ited, and associations with these factors are not con- susceptibility to carcinogens.
sistent (Karami et al, 2013). Height has been implicated in renal carcin-
The biological mechanism underlying the ogenesis (Green et  al, 2011), but is unlikely to
association between parity and kidney cancer is independently impact cancer risk. Height might
not understood, but pregnancy-​induced physio- rather be a marker for hormonal, nutritional, and
logical changes, such as increased maternal blood other factors that play a causal role.
volume and excess risk of urinary tract infection
might make renal tubular cells more vulnera- Infections
ble to renal stress and inflammation (Chow & Viruses have been implicated in animal exper-
Devesa, 2008; Ramin et al, 2006). imental models as etiologic factors in kidney
578

578 Textbook of Cancer Epidemiology

carcinoma, but no oncogenic viruses have been pyelography, but most cancers were urothelial
associated with kidney cancer in humans in the renal pelvis (Oyen et al, 1988). One study
(Newsom & Vugrin, 1987). However, in a human (Drubay et al, 2014) observed no significant asso-
normal cell culture, an oncogenic protein-​ ciation between kidney cancer mortality and any
initiated transformation of renal proximal tubule type of occupational radiation exposure.
cells—​including the induction of genetic insta-
bility and a deletion of chromosomal region Occupation
on 3p—​is one of the most frequent deletions In contrast to bladder cancer, kidney cancer is
observed (Nanus et al, 1991). The risk of kidney not a typical occupation-​related cancer, but ele-
cancer was increased in patients with HIV infec- vated risk was observed for certain occupational
tion in one study (Baynham et al, 1997), but not exposures. To date, the most extensively exam-
in another (Selik & Rabkin, 1998). ined exposure is trichloroethylene (TCE), a chlo-
Although hepatitis C virus (HCV) infection rinated solvent widely used as metal degreaser
causes cirrhosis and hepatocellular carcinoma, and as general solvent in other industries. An
limited studies have examined its relation to kid- early study observed that kidney cancer patients
ney cancer risk with equivocal results (Gonzalez exposed to TCE showed frequent somatic VHL
et al, 2015; Gordon et al, 2010). mutations (Brauch et al, 1999). Subsequent evi-
dence supports a positive association between
Urinary Tract Infections kidney cancer and TCE but also degreasing agents
See the section “Medical Conditions and Treatment.” and solvents used in iron, steel, dry-​cleaning, and
laundry industries (Karami et al, 2012; Kim et al,
Physical Activity 2014; Vlaanderen et al, 2013). The IARC reclassi-
Limited studies have examined the impact of fied TCE as a human carcinogen (Group 1), based
occupational and/​or recreational physical activ- on sufficient epidemiologic and animal evidence
ity on the risk of kidney cancer, including a few for kidney cancer (Guha et  al, 2012). However,
cohort studies that reported a dose-​ response additional research is needed to identify expo-
relationship with increasing levels of physical sure levels with carcinogenic effects, mechanisms
activity (Mahabir et al, 2004; Moore et al, 2008). of TCE’s actions, and interaction with metabolic
However, it is not clear whether the association gene variants (Purdue, 2013).
is independent of obesity. Physical activity may Exposure to other occupational agents has
influence kidney cancer risk by reducing body also been associated with kidney cancer. Asbestos
weight and hypertension, reducing inflammation is a possible risk factor for kidney cancer (Mattioli
and its associated oxidative stress, and improving et  al, 2002), although unlikely to account for
insulin sensitivity. The limited epidemiologic evi- any important increase in risk (Sali & Boffetta,
dence precludes, however, causal interpretation. 2000). Associations were also observed with gas-
oline and other petroleum products (Lynge et al,
Ionizing Radiation 1997), and further with hydrocarbons (Boffetta
Although the IARC classifies all types of ionizing et al, 1997), lead (Southard et al, 2012), cadmium
radiation as carcinogens to humans (Group  1), (Pesch et al, 2000), and work or exposure related
current evidence suggesting radiation as a cause to dry-​cleaning and laundry (Mandel et al, 1995).
of kidney cancer remains limited (El Ghissassi
et  al, 2009). However, studies, though limited, Medical Conditions and Treatment
generally indicated positive associations. For A number of medical conditions have been asso-
example, positive associations with kidney can- ciated with kidney cancer risk, but the evidence is
cer were observed in a case-​control study (Asal consistent for only few of them. A major concern
et al, 1988) in women with any lifetime radiation in the case-​control studies is recall bias, espe-
therapy received, as well as in Japanese atomic cially in regard to urinary tract conditions.
bomb survivors (Richardson & Hamra, 2010).
Among patients with ankylosing spondylitis End-​Stage Renal Disease
who had received x-​ray treatment, the mortality Acquired cystic kidney disease, which occurs in
due to kidney cancer was significantly increased end-​stage renal disease with progressive devel-
(Weiss et al, 1994). Thorotrast, an alpha-​emitting opment of cysts in a poorly functioning or non-
contrast medium, has been linked to kidney can- functioning kidney, is strongly associated with
cer in patients who had undergone retrograde the development of kidney cancer. Acquired
579

Kidney Cancer 579

cystic kidney disease has been observed in 7% and diabetes (Joh et  al, 2011). The mechanisms
to 22% of patients with end-​stage renal disease underlying the association between diabetes and
prior to dialysis, but the proportion increases to kidney cancer is unclear, but might include hyper-
90% after 10 years of dialysis. A notable sex dif- insulinemia, elevated levels of insulin, insulin-​
ference has been observed, men having a higher like growth factor, inflammatory cytokines, and
incidence and more severe cystic change than renal hypertrophy (Calle & Kaaks, 2004).
women (Marple et al, 1994).
The incidence of kidney cancer in patients Hypertension, Antihypertensive,
with end-​stage renal disease has been reported to and Diuretics
be up to 40 to 100 times higher than that in the Although kidney cancer could sometimes cause
general population (Denton et  al, 2002; Stewart hypertension, there is sufficient evidence that
et  al, 2009). An increased risk has also been hypertension is an independent risk factor
seen in native kidneys after renal transplanta- for kidney cancer. Long-​term hypertension or
tion (Ianhez et al, 2007; Klatte et al, 2010). While increasing levels of blood pressure are, in a dose-​
proliferation of proximal tubular epithelial cells response manner, associated with a higher risk
is the major pathogenetic mechanism of cyst for- of developing kidney cancer (Chow et  al, 2000;
mation, hormones (e.g., estrogens) and growth Flaherty et  al, 2005; Vatten et  al, 2007; Weikert
factors and their receptors may stimulate cell et  al, 2008). In addition, results were robust if
proliferation and promote carcinogenesis. This restricting analysis of hypertension to 5–​10 or
mechanism may also explain, in part, the onset more years prior to cancer diagnosis, arguing
of multiple renal adenomas and bilateral carcino- against hypertension as a consequence of the
mas that develop in patients with acquired cystic cancer.
kidney disease (Bonsib, 2009). Diuretic use (Heath et  al, 1997)  and non-
diuretic antihypertensive use (McCredie &
Urologic Disorders Stewart, 1992) were associated with an increased
Kidney stones have been found to increase the risk of kidney cancer. However, because diu-
risk of kidney cancer in several studies, although retic use is highly correlated with hypertension,
a meta-​analysis found this association only in it is difficult to tease out its independent effect.
men (Cheungpasitporn et al, 2015). Positive asso- Several studies reported that the association
ciations were also observed for kidney infections disappeared after adjusting for hypertension
(Schlehofer et al, 1996), lower urinary tract infec- (Flaherty et  al, 2015; Shapiro et  al, 1999), sug-
tion (Parker et al, 2004), and unspecified urinary gesting that diuretics may be a marker of hyper-
tract infections in women (Kreiger et  al, 1993). tension. Multiple measured blood pressure was
For the reported possible associations between associated with an increased kidney cancer risk
kidney cancer and thyroid disease (Rosenberg and reduction in blood pressure over time was
et al, 1990), thyroid cancer (Tucker et al, 1985), associated with a decreased risk in a cohort of
and lymphoid malignancies (Barista, 1997), the Swedish men (Chow et  al, 2000). This suggests
evidence is weak. that effective control of blood pressure might
reduce kidney cancer risk.
Diabetes Mellitus Hypertension and obesity are highly cor-
The majority of epidemiological studies found related, but each of these factors appears inde-
that patients with type 2 diabetes are more likely pendently associated with kidney cancer risk.
to develop kidney cancer compared with non- Individuals with both conditions experienced
diabetic patients. A meta-​analysis of nine cohort the highest risk (Chow et al, 2000; Weikert et al,
studies showed significant positive associations 2008). The mechanisms underlying the asso-
between diabetes and kidney cancer in both men ciation between hypertension and kidney can-
and women (Larsson & Wolk, 2011). Results cer remain unclear, although lipid peroxidation
were generally attenuated after adjustment for and chronic renal hypoxia have been proposed
BMI, smoking, or hypertension, suggesting that (Sharifi & Farrar, 2006).
it might be difficult to disentangle independ-
ent causal effect of diabetes from these factors Analgesics
(Larsson & Wolk, 2011). Nonetheless, the high- Phenacetin-​ containing analgesics causes renal
est risk was observed among women with multi- pelvic cancer (IARC, 1987), but their role in the
ple comorbidities, such as obesity, hypertension, etiology of kidney cancer is less clear. A number
580

580 Textbook of Cancer Epidemiology

of studies have associated a moderately elevated

TABLE 23-​3   EPIDEMIOLOGY OF KIDNEY
risk with regular or long-​term use (Kreiger et al,
1993; Maclure & MacMahon, 1985; McLaughlin CANCER: RISK FACTOR SUMMARY
et al, 1985). A large case-​control study showed Risk factor Direction
an elevated risk associated with all major types of effect*
of analgesics (Gago-​ Dominguez et al, 1999).
Further, nonaspirin nonsteroidal anti-​inflamma- Well-​Confirmed Risk Factors
tory drug (NSAID) use was linked to an elevated Cigarette smoking ↑
risk of kidney cancer, especially among those with
Obesity ↑
longer duration (i.e., ≥ 10 years) in the Nurses’
Health Study and the Health Professionals Acquired cystic kidney disease ↑↑
Follow-​up Study (Cho et al, 2011). In contrast, Inherited susceptibility (e.g., VHL) ↑
the international multicenter study (McCredie et Hypertension ↑
al, 1995) and the National Institutes of Health–​
American Association of Retired Persons Diet Probable Relationship Exists
and Health Study found no association (Liu et al, Alcohol ↓
2013). Parity ↓
Occupational exposure to ↑
Diet Pills
trichloroethylene or asbestos
Use of diet pills containing amphetamines for
primarily pharmacologic treatment of obesity has Inconsistent Findings or Limited
been reported in a few studies. An increased risk Study to Date
of kidney cancer has been found (McCredie et al,
Dietary factors
1995), including four separate center-​ specific
publications (Yuan et  al, 1998). Sometimes this   Vegetables and fruit ↓
association appeared to be independent of obe-   Red and processed meat ↑
sity, although it is difficult to separate the effects   Protein, carbohydrates, dietary fat —​
of amphetamine use from those of obesity and   Other macronutrients or —​
its consequences. One study found a significant micronutrients
dose-​response relationship (Yuan et al, 1998).
Physical activity ↓
Other Risk Factors Medical conditions
Educational or socioeconomic level has some-  Analgesics ↑
times been associated with kidney cancer. In  Diabetes
case-​control studies, no major differences have  Infections
been found (De Stefani et al, 1998; Hellenthal &
Bermejo, 2012), although others have found an Reproductive factors and hormones —​
inverse relationship (Lindblad et  al, 1997; Yuan (e.g., oral contraceptives)
et al, 1998). Ionizing radiation ↑
Drinking water containing arsenic ↑
CONCLUSION Occupational exposures (e.g., ­cadmium, ↑
Worldwide, the increasing incidence of kidney hydrocarbons, gasoline)
cancer in most populations may be due in part to
incidental detection with new imaging modali- *Arrows indicate the approximate magnitude of the relationship: ↑,
slight to moderate increase in risk; ↑↑, moderate to large increase
ties. Further, the increase is not only restricted in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large
to small local tumors but also includes more decrease in risk; —​, no association
advanced tumors, which may help explain the Note: The magnitude of the risk can vary substantially, depending
on the exact comparison being made.
still high mortality rates. The variation in inci-
dence among populations may have several other
explanations. Cigarette smoking, obesity, and incidence. In high-​income countries the impact
hypertension are well-​ established risk factors of smoking should continue to decrease follow-
for kidney cancer (Table 23-​3). The increased ing declining consumption; however, in low-​
prevalence of obesity and hypertension in many income countries, the incidence rates probably
populations might have contributed to the rising will grow. Accumulating evidence has suggested
581

Kidney Cancer 581

that alcohol consumption, occupational expo- Boffetta P, Jourenkova N, Gustavsson P. Cancer risk
sure to trichloroethylene, physical activity, and from occupational and environmental exposure
high parity in women may play a role in renal car- to polycyclic aromatic hydrocarbons. Cancer
cinogenesis. Although most kidney cancers are Causes Control 1997;8:444–​72.
sporadic and current studies of common genetic Bonsib SM. Renal cystic diseases and renal neo-
variants reported mixed results, genetic varia- plasms:  a mini-​review. Clin J Am Soc Nephrol
tions may also contribute to the etiology of kid- 2009;4:1998–​2007.
ney cancer. Continued research in kidney cancer Bouvard V, Loomis D, Guyton KZ, Grosse Y, Ghissassi
FE, Benbrahim-​Tallaa L et al.; International Agency
is warranted because the etiology remains largely
for Research on Cancer Monograph Working
unknown. With the aim of prevention, the con-
Group. Carcinogenicity of consumption of red and
tinued search for environmental causes should
processed meat. Lancet Oncol 2015;16:1599–​600.
take into account the fact that kidney cancer con- Boyle P, Koechlin A, Autier P. Sweetened car-
sists of different subtypes with specific genetic or bonated beverage consumption and cancer
tumor molecular characteristics. Furthermore, risk: meta-​analysis and review. Eur J Cancer Prev
genetic polymorphisms may have a modulating 2014;23:481–​90.
effect on metabolic activation and detoxifica- Brauch H, Weirich G, Hornauer MA, Storkel S,
tion enzymes. Thus, better understanding of the Wohl T, Bruning T. Trichloroethylene expo-
genetic and molecular processes involved in kid- sure and specific somatic mutations in patients
ney cancer may help analyzing exposure associa- with renal cell carcinoma. J Natl Cancer Inst
tions that are important in both its initiation and 1999;91:854–​61.
progression. Brennan P, McKay J, Moore L, Zaridze D, Mukeria
A, Szeszenia-​ Dabrowska N et  al. Uncommon
REFERENCES CHEK2 mis-​sense variant and reduced risk of
Allen NE, Roddam AW, Sieri S, Boeing H, Jakobsen tobacco-​related cancers: case control study. Hum
MU, Overvad K et  al. A prospective analysis of Mol Genet 2007;16:1794–​801.
the association between macronutrient intake and Calle EE, Kaaks R. Overweight, obesity and can-
renal cell carcinoma in the European Prospective cer:  epidemiological evidence and proposed
Investigation into Cancer and Nutrition. Int J mechanisms. Nat Rev Cancer 2004;4:579–​91.
Cancer 2009;25:982–​87. Cao Y, Willett WC, Rimm EB, Stampfer MJ,
Andreotti G, Boffetta P, Rosenberg PS, Berndt SI, Giovannucci EL. Light to moderate intake of alco-
Karami S, Menashe I et al. Variants in blood pres- hol, drinking patterns, and risk of cancer: results
sure genes and the risk of renal cell carcinoma. from two prospective US cohort studies. BMJ
Carcinogenesis 2010;31:614–​20. 2015;351:h4238.
Asal NR, Geyer JR, Risser DR, Lee ET, Kadamani S, Chen CJ, Chen CW, Wu MM, Kuo TL. Cancer poten-
Cherng N. Risk factors in renal cell carcinoma. tial in liver, lung, bladder and kidney due to
II. Medical history, occupation, multivariate ingested inorganic arsenic in drinking water. Br J
analysis, and conclusions. Cancer Detect Prev Cancer 1992;66:888–​92.
1988;13:263–​79. Cheungpasitporn W, Thongprayoon C, O’Corragain OA,
Baastrup R, Sorensen M, Balstrom T, Frederiksen Edmonds PJ, Ungprasert P, Kittanamongkolchai W
K, Larsen CL, Tjonneland A et  al. Arsenic in et al. The risk of kidney cancer in patients with kid-
drinking-​water and risk for cancer in Denmark. ney stones:  a systematic review and meta-​analysis.
Environ Health Persp 2008;116:231–​37. QJM 2015;108:205–​12.
Barista I. An association between renal cell carcinoma Cho E, Curhan G, Hankinson SE, Kantoff P, Atkins
and lymphoid malignancies:  a case series of eight MB, Stampfer M et  al. Prospective evaluation of
patients. Cancer 1997;80:1004-​ 5; author reply analgesic use and risk of renal cell cancer. Arch
1006–​7. Intern Med 2011;171:1487–​93.
Baynham SA, Katner HP, Cleveland KB. Increased prev- Chow W-​H, Devesa SS. Contemporary epidemiol-
alence of renal cell carcinoma in patients with HIV ogy of renal cell cancer. Cancer Journal 2008;14:
infection. AIDS Patient Care STDS 1997;11:161–​5. 288–​301.
Bellocco R, Pasquali E, Rota M, Bagnardi V, Tramacere Chow WH, Gridley G, Fraumeni JF, Jarvholm B.
I, Scotti L et al. Alcohol drinking and risk of renal Obesity, hypertension, and the risk of kidney can-
cell carcinoma:  results of a meta-​analysis. Ann cer in men. New Engl J Med 2000;343:1305–​11.
Oncol 2012;23:2235–​44. Choyke PL, Glenn GM, Walther MM, Patronas NJ,
Bergstrom A, Pisani P, Tenet V, Wolk A, Adami HO. Linehan WM, Zbar B. von Hippel-​Lindau disease:
Overweight as an avoidable cause of cancer in genetic, clinical, and imaging features. Radiol
Europe. Int J Cancer 2001;91:421–​30. 1995;194:629–​42.
582

582 Textbook of Cancer Epidemiology

Clague J, Lin J, Cassidy A, Matin S, Tannir NM, in chronic hepatitis C infection. Cancer Epidemiol
Tamboli P et  al. Family history and risk of renal Biomarkers Prev 2010;19:1066–​73.
cell carcinoma:  results from a case-​control study Green J, Cairns BJ, Casabonne D, Wright FL, Reeves
and systematic meta-​analysis. Cancer Epidemiol G, Beral V, Million Women Study. Height and can-
Biomarkers Prev 2009;18:801–​07. cer incidence in the Million Women Study: pro-
Daniel CR, Park Y, Chow WH, Graubard BI, spective cohort, and meta-​analysis of prospective
Hollenbeck AR, Sinha R. Intake of fiber and fiber-​ studies of height and total cancer risk. Lancet
rich plant foods is associated with a lower risk of Oncol 2011;12:785–​94.
renal cell carcinoma in a large US cohort. Am J Guan HB, Wu QJ, Gong TT. Parity and kidney cancer
Clin Nutr 2013;97:1036–​43. risk: evidence from epidemiologic studies. Cancer
Davies MJ, Baer DJ, Judd JT, Brown ED, Campbell WS Epidemiol Biomarkers Prev 2013;22:2345–​53.
et  al. Effects of moderate alcohol intake on fast- Guha N, Loomis D, Grosse Y, Lauby-​Secretan B, El
ing insulin and glucose concentrations and insulin Ghissassi F, Bouvard V et  al. Carcinogenicity
sensitivity in postmenopausal women: a random- of trichloroethylene, tetrachloroethylene, some
ized controlled trial. JAMA 2002;287:2559–​62. other chlorinated solvents, and their metabolites.
Denton MD, Magee CC, Ovuworie C, Mauiyyedi Lancet Oncol 2012;13:1192–​3.
S, Pascual M, Colvin RB et  al. Prevalence of Hayes JD, Strange RC. Glutathione S-​transferase poly-
renal cell carcinoma in patients with ESRD pre-​ morphisms and their biological consequences.
transplantation: a pathologic analysis. Kidney Int Pharmacol 2000;61:154–​66.
2002;61:2201–​9. He FJ, Li J, Macgregor GA. Effect of longer term mod-
De Stefani E, Fierro L, Mendilaharsu M, Ronco A, est salt reduction on blood pressure:  Cochrane
Larrinaga MT, Balbi JC et al. Meat intake, “mate” systematic review and meta-​ analysis of ran-
drinking and renal cell cancer in Uruguay: a case-​ domised trials. BMJ 2013;346:f1325.
control study. Br J Cancer 1998;78:1239–​43. Heath CW Jr, Lally CA, Calle EE, McLaughlin JK, Thun
Dong LM, Brennan P, Karami S, Hung RJ, Menashe I, MJ. Hypertension, diuretics, and antihypertensive
Berndt SI et al. An analysis of growth, differentia- medications as possible risk factors for renal cell
tion and apoptosis genes with risk of renal cancer. cancer. Am J Epidemiol 1997;145:607–​13.
PLoS One 2009;4:e4895 Hellenthal NJ, Bermejo CE. The role of socioeco-
Drubay D, Ancelet S, Acker A, Kreuzer M, Laurier nomic status in renal cell carcinoma. Urol Oncol
D, Rage E. Kidney cancer mortality and ionizing 2012;30:89–​94.
radiation among French and German uranium Hirvonen T, Kontto J, Jestoi M, Valsta L, Peltonen K,
miners. Radiat Environ Biophys 2014;53: 505–​13. Pietinen P et  al. Dietary acrylamide intake and
El Ghissassi F, Baan R, Straif K, Grosse Y, Secretan B, the risk of cancer among Finnish male smokers.
Bouvard V et al. WHO International Agency for Cancer Causes Control 2010;21:2223–​9.
Research on Cancer Monograph Working Group Hogervorst JG, Schouten LJ, Konings EJ, Goldbohm
A review of human carcinogens-​part D: radiation. RA, van den Brandt PA. Dietary acrylamide
Lancet Oncol 2009;10:751–​2. intake and the risk of renal cell, bladder, and pros-
Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers tate cancer. Am J Clin Nutr 2008;87:1428–​38.
C, Rebelo M, et al. Cancer incidence and mortality Hopenhayn-​Rich C, Biggs ML, Smith AH. Lung and
worldwide: sources, methods and major patterns in kidney cancer mortality associated with arsenic
GLOBOCAN 2012. Int J Cancer 2015;136: E359–​86. in drinking water in Cordoba, Argentina. Int J
Flaherty KT, Fuchs CS, Colditz GA, Stampfer MJ, Epidemiol 1998;27:561–​9.
Speizer FE, Willett WC et al. A prospective study Hu J, Ugnat AM. Active and passive smoking and risk
of body mass index, hypertension, and smoking of renal cell carcinoma in Canada. Eur J Cancer
and the risk of renal cell carcinoma (United States). 2005;41:770–​8.
Cancer Causes Control 2015;16:1099–​1106. Hughes LA, Schouten LJ, Goldbohm RA, van den
Gago-​Dominguez M, Yuan JM, Castelao JE, Ross RK, Brandt PA, Weijenberg MP. Self-​reported clothing
Yu MC. Regular use of analgesics is a risk factor for size as a proxy measure for body size. Epidemiol
renal cell carcinoma. Br J Cancer 1999;81:542–​8. 2009;20:673–​6.
Gnarra JR. von Hippel-​Lindau gene mutations in human IARC. Tobacco Smoking. IARC Monographs on the
and rodent renal tumors: association with clear cell Evaluation of Carcinogenic Risks to Humans, Vol.
phenotype. J Natl Cancer Inst 1998;90:1685–​7. 38. Lyon, France: IARC, 1986.
Gonzalez HC, Lamerato L, Rogers CG, Gordon SC. IARC. Phenactin and analgesic mixtures containing
Chronic hepatitis C infection as a risk factor for phenacetin. Monographs on the Evaluation of
renal cell carcinoma. Dig Dis Sci 2015;60:1820–​4. the Carcinogenic Risk of Chemicals to Humans.
Gordon SC, Moonka D, Brown KA, Rogers C, Huang An updating of IARC Monographs. IARC
MA, Bhatt N et  al. Risk for renal cell carcinoma Monographs 1987;1–​42 Suppl 7:310–​2.
583

Kidney Cancer 583

IARC. Chlorinated drinking-​water; chlorination by-​ Kurttio P, Pukkala E, Kahelin H, Auvinen A, Pekkanen
products; some other halogenated compounds; J. Arsenic concentrations in well water and risk of
cobalt and cobalt compounds. International bladder and kidney cancer in Finland. Environ
Agency for Research on Cancer (IARC) Working Health Perspect 1999;107:5–​10.
Group, Lyon, 12–​19 June 1990. IARC Monographs Larsson SC, Wolk A. Diabetes mellitus and incidence
on the Evaluation of Carcinogenic Risks to of kidney cancer: a meta-​analysis of cohort stud-
Humans, Vol. 52. Lyon: IARC, 1991. ies. Diabetologia 2011;54:1013–​8.
IARC. Arsenic, Metals, Fibres, and Dusts. IARC Lee JE, Giovannucci E, Smith-​Warner SA, Spiegelman
Monographs on the Evaluation of Carcinogenic D, Willett WC, Curhan GC. Intakes of fruits, veg-
Risks to Humans, Vol. 100C. Lyon: IARC, 2012a. etables, vitamins A, C, and E, and carotenoids
IARC. Personal Habits and Indoor Combustions. IARC and risk of renal cell cancer. Cancer Epidemiol
Monographs on the Evaluation of Carcinogenic Biomarkers Prev 2006;15:2445–​52.
Risks to Humans, Vol. 100E. Lyon: IARC, 2012b. Lee JE, Hankinson SE, Cho E. Reproductive factors
Hunt JD, van der Hel OL, McMillan GP, Boffetta P, and risk of renal cell cancer. Am J Epidemiol
Brennan P. Renal cell carcinoma in relation to cig- 2009a;169:1243–​50.
arette smoking: meta-​analysis of 24 studies. Int J Lee JE, Hunter DJ, Spiegelman D, Adami H-​O, Albanes
Cancer 2005;114:101–​8. D, Bernstein L et al. Alcohol intake and renal cell
Ianhez LE, Lucon M, Nahas WC, Sabbaga E, Saldanha cancer in a pooled analysis of 12 prospective stud-
LB, Lucon AM et al. Renal cell carcinoma in renal ies. J Natl Cancer Inst 2007a;99:801–​10.
transplant patients. Urol 2007;69:462–​4. Lee JE, Hunter DJ, Spiegelman D, Adami H-​ O,
Ildaphonse G, George PS, Mathew A. Obesity and Bernstein L, van den Brandt PA et  al. Intakes of
kidney cancer risk in men: a meta-​analysis (1992–​ coffee, tea, milk, soda and juice and renal cell can-
2008). Asian Pac J Cancer Prev 2009;10:279–​86. cer in a pooled analysis of 13 prospective studies.
Jacobsen J, Grankvist K, Rasmuson T, Bergh A, Int J Cancer 2007b;121:2246–​53.
Landberg G, Ljungberg B. Expression of vascular Lee JE, Mannisto S, Spiegelman D, Hunter DJ, Bernstein
endothelial growth factor protein in human renal L, van den Brandt PA, et al. Intakes of fruit, vege-
cell carcinoma. BJU Int 2004;93:297–​302. tables, and carotenoids and renal cell cancer risk: a
Joh HK, Willett WC, Cho E. Type 2 diabetes and the pooled analysis of 13 prospective studies. Cancer
risk of renal cell cancer in women. Diabetes Care Epidemiol Biomarkers Prev 2009b;18:1730–​39.
2011;34:1552–​6. Lee JE, Spiegelman D, Hunter DJ, Albanes D, Bernstein
Karami S, Brennan P, Rosenberg PS, Navratilova M, L, van den Brandt PA et al. Fat, protein, and meat
Mates D, Zaridze D et  al. Analysis of SNPs and consumption and renal cell cancer risk: a pooled
haplotypes in vitamin D pathway genes and renal analysis of 13 prospective studies. J Natl Cancer
cancer risk. PLoS One 2009;4:e7013. Inst 2008;100:1695–​706.
Karami S, Daugherty SE, Schonfeld SJ, Park Y, Lewis DR, Southwick JW, Ouellet-​Hellstrom R, Rench
Hollenbeck AR, Grubb RL 3rd et  al. Reproductive J, Calderon RL. Drinking water arsenic in Utah: a
factors and kidney cancer risk in 2 US cohort stud- cohort mortality study. Environ Health Perspect
ies, 1993–​2010. Am J Epidemiol 2013;177:1368–​77. 1999;107:359–​65.
Karami S, Lan Q, Rothman N, Stewart PA, Lee KM, Li L, Kaelin WG Jr. New insights into the biology of
Vermeulen R et  al. Occupational trichloroeth- renal cell carcinoma. Hematol Oncol Clin North
ylene exposure and kidney cancer risk: a meta-​ Am 2011;25:667–​86.
analysis. Occup Environ Med 2012;69:858–​67. Lindblad P, Wolk A, Bergstrom R, Adami HO. Diet
Kim I, Ha J, Lee JH, Yoo KM, Rho J. The relationship and risk of renal cell cancer:  a population-​based
between the occupational exposure of trichloro- case-​control study. Cancer Epidemiol Biomarkers
ethylene and kidney cancer. Ann Occup Environ Prev 1997;6:215–​23.
Med 2014;26:12. Lindblad P, Wolk A, Bergstrom R, Persson I, Adami
Kim WY, Kaelin WG. Role of VHL gene mutation in HO. The role of obesity and weight fluctuations
human cancer. J Clin Oncol 2004;22:4991–​5004. in the etiology of renal cell cancer: a population-​
Klatte T, Seitz C, Waldert M, de Martino M, Kikic Z, based case-​ control study. Cancer Epidemiol
Boehmig GA et al. Features and outcomes of renal Biomarkers Prev 1994;3:631–​9.
cell carcinoma of native kidneys in renal trans- Liu W, Park Y, Purdue MP, Giovannucci E, Cho
plant recipients. BJU Int 2010;105:1260–​65. E. A large cohort study of nonsteroidal anti-​
Kreiger N, Marrett LD, Dodds L, Hilditch S, inflammatory drugs and renal cell carcinoma inci-
Darlington GA. Risk factors for renal cell carci- dence in the National Institutes of Health-​AARP
noma: results of a population-​based case-​control Diet and Health Study. Cancer Causes Control
study. Cancer Causes Control 1993;4:101–​10. 2013;24:1865–​73.
584

584 Textbook of Cancer Epidemiology

Lopez-​Beltran A, Carrasco JC, Cheng L, Scarpelli M, of genome-​ wide association study-​ identified
Kirkali Z, Montironi R. 2009 update on the classi- susceptibility loci and meat-​ cooking mutagens
fication of renal epithelial tumors in adults. Int J in the etiology of renal cell carcinoma. Cancer
Urol 2009;16:432–​43. 2016;122:108–​15.
Luo J, Margolis KL, Adami HO, Lopez AM, Lessin Mellemgaard A, McLaughlin JK, Overvad K, Olsen
L, Ye W, Women’s Health Initiative I. Body size, JH. Dietary risk factors for renal cell carcinoma in
weight cycling, and risk of renal cell carcinoma Denmark. Eur J Cancer 1996;32A:673–​82.
among postmenopausal women:  the Women’s Moore LE, Brennan P, Karami S, Hung RJ, Hsu C,
Health Initiative (United States). Am J Epidemiol Boffetta P et al. Glutathione S-​transferase polymor-
2007;166:752–​9. phisms, cruciferous vegetable intake and cancer
Lynge E, Andersen A, Nilsson R, Barlow L, Pukkala E, risk in the Central and Eastern European Kidney
Nordlinder R et al. Risk of cancer and exposure to Cancer Study. Carcinogenesis 2007;28:1960–​4.
gasoline vapors. Am J Epidemiol 1997;145:449–​58. Moore SC, Chow W-​H, Schatzkin A, Adams KF, Park
Maclure M, MacMahon B. Phenactin and cancer of Y, Ballard-​Barbash R et al. Physical activity during
urinary tract. N Engl J Med 1985;313:1479. adulthood and adolescence in relation to renal cell
Mahabir S, Leitzmann MF, Pietinen P, Albanes D, cancer. Am J Epidemiol 2008;168:149–​57.
Virtamo J, Taylor PR. Physical activity and renal Morris RD. Drinking water and cancer. Environ
cell cancer risk in a cohort of male smokers. Int J Health Perspect 1995;103 Suppl 8:225–​31.
Cancer 2004;108:600–​5. Mucci LA, Lindblad P, Steineck G, Adami HO. Dietary
Maher ER. Inherited renal cell carcinoma. Br J Urol acrylamide and risk of renal cell cancer. Int J
1996;78:542–​5. Cancer 2004;109:774–​6.
Maher ER, Neumann HP, Richard S. von Hippel-​ Munro IC, Kennepohl E, Erickson RE, Portoghese PS,
Lindau disease: a clinical and scientific review. Eur Wagner BM, Easterday OD et al. Safety assessment
J Hum Genet 2011;19:617–​23. of ingested heterocyclic amines:  initial report.
Mandel JS, McLaughlin JK, Schlehofer B, Mellemgaard Regul Toxicol Pharmacol 1993; 17(2 Pt 2):S1–​109.
A, Helmert U, Lindblad P et  al. International Muscat JE, Hoffmann D, Wynder EL. The epidemi-
renal-​ cell cancer study. IV. Occupation. Int J ology of renal cell carcinoma. A second look.
Cancer 1995;61:601–​5. Cancer 1995;75:2552–​7.
Marple JT, MacDougall M, Chonko AM. Renal cancer Nanus DM, Lynch SA, Rao PH, Anderson SM,
complicating acquired cystic kidney disease. J Am Jhanwar SC, Albino AP. Transformation of human
Soc Nephrol 1994;4:1951–​6. kidney proximal tubule cells by a src-​containing
Mathew A, George PS, Ildaphonse G. Obesity and kid- retrovirus. Oncogene 1991;6:2105–​11.
ney cancer risk in women: a meta-​analysis (1992–​ Newsom GD, Vugrin D. Etiologic factors in renal cell
2008). Asian Pac J Cancer Prev 2009;10:471–​8. adenocarcinoma. Semin Nephrol 1987;72:109–​16.
Mattioli S, Truffelli D, Baldasseroni A, Risi A, Marchesini Nicodemus KK, Sweeney C, Folsom AR. Evaluation of
B, Giacomini C et al. Occupational risk factors for dietary, medical and lifestyle risk factors for inci-
renal cell cancer:  a case-​control study in northern dent kidney cancer in postmenopausal women.
Italy. J Occup Environ Med 2002;44:1028–​36. Int J Cancer 2004;108:115–​21.
McCredie M, Pommer W, McLaughlin JK, Stewart Nogueira E, Cardesa A, Mohr U. Experimental mod-
JH, Lindblad P, Mandel JS et  al. International els of kidney tumors. J Cancer Res Clin Oncol
renal-​cell cancer study. II. Analgesics. Int J Cancer 1993;119:190–​8.
1995;60:345–​9. Oyen RH, Gielen JL, Van Poppel HP, Verbeken EK,
McCredie M, Stewart JH. Risk factors for kidney can- Van Damme BJ, Baert LV et  al. Renal thorium
cer in New South Wales, Australia. II. Urologic deposition associated with transitional cell carci-
disease, hypertension, obesity, and hormonal fac- noma:  radiologic demonstration in two patients.
tors. Cancer Causes Control 1992;3:323–​31. Radiol 1988;169:705–​7.
McLaughlin JK, Blot WJ, Mehl ES, Fraumeni JF Parker AS, Cerhan JR, Lynch CF, Leibovich BC, Cantor
Jr. Relation of analgesic use to renal cancer: KP. History of urinary tract infection and risk of
population-​ based findings. Natl Cancer Inst renal cell carcinoma. Am J Epidemiol 2004;159:42–​8.
Monogr 1985;69:217–​22. Pelucchi C, Bosetti C, Galeone C, La Vecchia C.
McLaughlin JK, Lindblad P, Mellemgaard A, McCredie Dietary acrylamide and cancer risk:  an updated
M, Mandel JS, Schlehofer B et  al. International meta-​analysis. Int J Cancer 2015;136:2912–​22.
renal-​ cell cancer study. I.  Tobacco use. Int J Pelucchi C, Galeone C, Montella M, Polesel J, Crispo
Cancer 1995;60:194–​8. A, Talamini R et  al. Alcohol consumption and
Melkonian SC, Daniel CR, Ye Y, Tannir NM, Karam renal cell cancer risk in two Italian case-​control
JA, Matin SF et al. Gene-​environment interaction studies. Ann Oncol 2008;19:1003–​8.
58

Kidney Cancer 585

Pesch B, Haerting J, Ranft U, Klimpel A, Oelschlagel Shiao YH, Rice JM, Anderson LM, Diwan BA, Hard
B, Schill W. Occupational risk factors for renal GC. von Hippel-​ Lindau gene mutations in N-​
cell carcinoma: agent-​specific results from a case-​ nitrosodimethylamine-​induced rat renal epithe-
control study in Germany:  MURC Study Group. lial tumors. J Natl Cancer Inst 1998;90:1720–​3.
Multicenter Urothelial and Renal Cancer Study. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2015.
Int J Epidemiol 2000;29:1014–​24. CA Cancer J Clin 2015;65:5–​29.
Pischon T, Lahmann PH, Boeing H, Tjonneland A, Slavin JL. Position of the American Dietetic
Halkjaer J, Overvad K et al. Body size and risk of Association: health implications of dietary fiber. J
renal cell carcinoma in the European Prospective Am Diet Assoc 2008;108:1716–​31.
Investigation into Cancer and Nutrition (EPIC). Smits KM, Schouten LJ, Hudak E, Verhage B, van Dijk
Int J Cancer 2006;118:728–​38. BA, Hulsbergen-​van de Kaa CA et  al. Body mass
Purdue MP. Trichloroethylene and cancer. J Natl index and von Hippel-​Lindau gene mutations in
Cancer Inst 2013;105:844–​6. clear-​cell renal cancer: results of the Netherlands
Ramin SM, Vidaeff AC, Yeomans ER, Gilstrap LC Cohort Study on diet and cancer. Ann Epidemiol
3rd. Chronic renal disease in pregnancy. Obstet 2010;20:401–​4.
Gynecol 2006;108:1531–​9. Song DY, Song S, Song Y, Lee JE. Alcohol intake and
Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen renal cell cancer risk: a meta-​analysis. Br J Cancer
M. Body-​mass index and incidence of cancer:  a 2012;106:1881–​90.
systematic review and meta-​analysis of prospective Southard EB, Roff A, Fortugno T, Richie JP Jr, Kaag
observational studies. Lancet 2008;371:569–​78. M, Chinchilli VM et  al. Lead, calcium uptake, and
Richardson DB, Hamra G. Ionizing radiation and kid- related genetic variants in association with renal cell
ney cancer among Japanese atomic bomb survi- carcinoma risk in a cohort of male Finnish smokers.
vors. Radiat Res 2010;173:837–​42. Cancer Epidemiol Biomarkers Prev 2012;21:191–​201.
Rosenberg AG, Dexeus F, Swanson DA, von Stewart JH, Vajdic CM, van Leeuwen MT, Amin J,
Eschenbach AC. Relationship of thyroid disease Webster AC, Chapman JR et  al. The pattern of
to renal cell carcinoma:  an epidemiologic study. excess cancer in dialysis and transplantation.
Urology 1990;35:492–​8. Nephrology Dialysis Transplant 2009;24:3225–​31.
Sali D, Boffetta P. Kidney cancer and occupational Straif K, Benbrahim-​ Tallaa L, Baan R, Grosse
exposure to asbestos:  a meta-​analysis of occu- Y, Secretan B, El Ghissassi F et  al. WHO
pational cohort studies. Cancer Causes Control International Agency for Research on Cancer
2000;11:37–​47. Monograph Working Group. A review of human
Schlehofer B, Pommer W, Mellemgaard A, Stewart carcinogens—​Part C:  metals, arsenic, dusts, and
JH, McCredie M, Niwa S et al. International renal-​ fibres. Lancet Oncol 2009;10:453–​4.
cell-​cancer study. VI. The role of medical and fam- Tucker MA, Boice JD Jr, Hoffman DA. Second cancer
ily history. Int J Cancer 1996;66:723–​6. following cutaneous melanoma and cancers of the
Schouten LJ, van Dijk BA, Oosterwijk E, Hulsbergen-​ brain, thyroid, connective tissue, bone, and eye in
van de Kaa CA, Kiemeney LA et al. Hypertension, Connecticut, 1935–​82. Natl Cancer Inst Monogr
antihypertensives and mutations in the Von 1985;68:161–​89.
Hippel-​Lindau gene in renal cell carcinoma: results van Dijk BA, Schouten LJ, Kiemeney LA, Goldbohm
from the Netherlands Cohort Study. J Hypertens RA, van den Brandt PA. Relation of height, body
2005;23:1997–​04. mass, energy intake, and physical activity to risk of
Selik RM, Rabkin CS. Higher rates of death with non-​ renal cell carcinoma: results from the Netherlands
AIDS-​defining cancers among HIV-​infected per- Cohort Study. Am J Epidemiol 2004;160:1159–​67.
sons in the USA. In International Conference on van Dijk BA, Schouten LJ, Oosterwijk E, Hulsbergen-​
AIDS, 1998. van de Kaa CA, Kiemeney LA, Goldbohm RA,
Semenza JC, Ziogas A, Largent J, Peel D, Anton-​ et al. Carotenoid and vitamin intake, von Hippel-​
Culver H. Gene-​environment interactions in renal Lindau gene mutations and sporadic renal cell car-
cell carcinoma. Am J Epidemiol 2001;153:851–​9. cinoma. Cancer Causes Control 2008;19:125–​34.
Shapiro JA, Williams MA, Weiss NS, Stergachis A, Van Poppel H, Nilsson S, Algaba F, Bergerheim U, Dal
LaCroix AZ, Barlow WE. Hypertension, antihy- Cin P, Fleming S et al. Precancerous lesions in the
pertensive medication use, and risk of renal cell kidney. Scand J Urol Nephrol 2000; Suppl:136–​65.
carcinoma. Am J Epidemiol 1999;149:521–​30. Vatten LJ, Trichopoulos D, Holmen J, Nilsen TIL.
Sharifi N, Farrar WL. Perturbations in hypoxia detec- Blood pressure and renal cancer risk: the HUNT
tion:  a shared link between hereditary and spo- Study in Norway. Br J Cancer 2007;97:112–​14.
radic tumor formation? Medical Hypotheses Vlaanderen J, Straif K, Pukkala E, Kauppinen T,
2006;66:732–​35. Kyyronen P, Martinsen JI et  al. Occupational
586

586 Textbook of Cancer Epidemiology

exposure to trichloroethylene and perchloroethy- investigation into cancer and nutrition. Am J

lene and the risk of lymphoma, liver, and kidney Epidemiol 2008;167:438–​46.
cancer in four Nordic countries. Occup Environ Weiss HA, Darby SC, Doll R. Cancer mortality follow-
Med 2013;70:393–​401. ing x-​ray treatment for ankylosing spondylitis. Int
Wang F, Xu Y. Body mass index and risk of renal cell J Cancer 1994;59:327–​38.
cancer: a dose-​response meta-​analysis of published Wolk A, Gridley G, Niwa S, Lindblad P, McCredie
cohort studies. Int J Cancer 2014;135:1673–​86. M, Mellemgaard A et  al. International renal
World Cancer Research Fund/​ American Institute cell cancer study. VII. Role of diet. Int J Cancer
for Cancer Research. Continuous Update Project 1996a;65:67–​73.
Report: Diet, Nutrition, Physical Activity, and Wolk A, Lindblad P, Adami HO. Nutrition and renal
Kidney Cancer 2015. cell cancer. Cancer Causes Control 1996b;7:5–​18.
Weikert S, Boeing H, Pischon T, Olsen A, Tjonneland Yuan JM, Castelao JE, Gago-​Dominguez M, Ross RK,
A, Overvad K et al. Fruits and vegetables and renal Yu MC. Hypertension, obesity and their medi-
cell carcinoma: findings from the European pro- cations in relation to renal cell carcinoma. Br J
spective investigation into cancer and nutrition Cancer 1998;77:1508–​13.
(EPIC). Int J Cancer 2006;118:3133–​9. Zhu Y, Horikawa Y, Yang H, Wood CG, Habuchi T,
Weikert S, Boeing H, Pischon T, Weikert C, Olsen A, Wu X. BPDE induced lymphocytic chromosome
Tjonneland A et  al. Blood pressure and risk of 3p deletions may predict renal cell carcinoma risk.
renal cell carcinoma in the European prospective J Urol 2008;179:2416–​21.
587

24
Brain Cancer
D O M I N I Q U E M I C H A U D , D AV I D   S AV I T Z A N D L O R E L E I M U C C I

D espite their relative rarity, brain cancers have

drawn wide attention because of the fear
inspired by their location, the young age at which
in which the diagnostic methods are unequally
available or used is problematic.

they sometimes occur, the generally poor prog- CLINICAL SYNOPSIS

nosis, and their perceived tendency to appear in Subgroups Although primary brain tumors occur
clusters. Nevertheless, few cancers have proven in a wide variety of histopathologic types (Table
as etiologically puzzling as brain cancer. The rea- 24-​1), the two broad groups are gliomas arising
son for such little progress is unclear, but there from neuroepithelial cells and meningiomas,
are several plausible explanations. Brain cancer which are histopathologically benign (although
occurs in an unusually diverse array of histologic they can be life threatening). The clinical fea-
types, with little certainty regarding the extent tures and management of these two types differ
to which they share an etiology. Except for high substantially.
doses of ionizing radiation, a nearly universal Symptoms Clinical manifestations arise through
carcinogen, we lack clear, identified causal agents two mechanisms. First, there are different focal
that can guide future research. symptoms—​related to the anatomic location and
The distinction between the occurrence of local infiltration of the tumor—​such as seizures;
brain tumors in children and adults in regard to changes in personality, mood, initiative, or mem-
the histology of tumors and presumed time per- ory; hemiplegia; aphasia; and visual aberrations.
iod of development suggests that the epidemio- Second, general symptoms dominated by headache
logic considerations also be separated since there are due chiefly to elevated intracranial pressure.
may well be different etiologies in the two groups.
Diagnosis Neuroimaging through magnetic res-
Given how little is currently known about etiol-
onance imaging (MRI) and computed tomog-
ogy, however, it is not yet certain whether the
raphy (CT) is the hallmark of diagnosis, and in
causes of brain tumors differ between children
certain situations complemented with angiogra-
and adults. In the discussion that follows, the
phy to delineate a vascular malformation or to
focus is largely on adult brain tumors, which are
determine the blood supply to a meningioma.
quantitatively dominant in the population and
the focus of the vast majority of studies. Treatment The goal of surgery is complete resec-
Unlike most other cancers, both benign and tion of the tumor, although this is not possible
malignant brain tumors are of concern because in most gliomas and in certain meningiomas.
their distinct anatomic location makes even Substantial palliation can be achieved through
benign tumors a serious threat to health and sur- reduction of tumor volume, however, and acqui-
vival. An additional challenge to researchers is sition of tissue for histopathologic examination
the difficulty of making comprehensive diagno- may guide subsequent adjuvant or palliative radi-
ses because of the inaccessible location of these ation therapy and/​or chemotherapy. While these
tumors. While advanced disease becomes appar- modalities play an important role in the manage-
ent at some point in its clinical course, identifica- ment of many brain tumors, their optimal use is
tion of subclinical disease or of tumors showing complex and depends on the clinical and histo-
only nonspecific symptoms depends on advanced pathologic features of the tumor.
technology. As a result, comparing incidence Prognosis In European countries, the overall rel-
rates across settings, populations, or time periods ative 5-​year survival rate for malignant brain
58

588 Textbook of Cancer Epidemiology

tumors is approximately 20%, with limited var- colleagues (Inskip et al, 1995) inferred that much,
iation between countries (Visser et al, 2015). The perhaps all, of the increase may have been due to
corresponding rate in the United States is around improved completeness of diagnosis.
35% (Howlader et al, 2015). Although the absolute incidence of brain
Progress No therapeutic breakthroughs or any tumors differs across geographic areas, the shape
substantial improvement in long-​term cure due of the age–​incidence curve is similar (Figs. 24-​1,
to more extensive use of radiotherapy have been 24-​2). Considering malignant tumors only of the
documented. Nevertheless, in the United States, brain, cranial nerves, and cranial meninges, data
5-​year relative survival, a good indicator of long-​ from 18 areas in the Surveillance, Epidemiology,
term cure, almost doubled from 22% in 1975 to and End Results (SEER) in 2008–​12 show an
35% in 2002 (Howlader et al, 2015). Changes in annual incidence ranging from 2.2 per 100,000
diagnostic criteria and the use of more sensi- in young adults (20–​24 years) to 19.4 per 100,000
tive, modern imaging techniques resulting in the in those over age 65 years (Howlader et al, 2015).
detection of smaller or slower growing tumors Analysis of age patterns for malignant brain
may have contributed to this apparent progress. tumors indicates a peak in childhood (prior to
age 10), a decline between ages 10 and 25, and
DESCRIPTIVE then a rise through adulthood (Howlader et  al,
EPIDEMIOLOGY 2015). This is also the age pattern for astrocytic
Benign tumors constitute about 66% of all brain tumors, but rarer subtypes more commonly
and central nervous system tumors (Ostrom found in children (e.g., pilocytic astrocytomas)
et  al., 2013). Epidemiologic studies vary in the follow a variety of different age–​incidence curves
extent to which these tumors are considered in (Ostrom et al, 2013).
conjunction with malignant ones, but most stud- The proportions of the various types of brain
ies primarily include malignant brain tumors. tumors differ markedly by sex and, to some extent,
A  detailed histologic classification of brain by race. Approximately 44% of tumors in men are
tumors and their age-​ standardized incidence malignant, while in women, 28% of tumors are
rates is given in Table 24-​1. malignant; this difference reflects a slightly lower
Assessment of temporal trends in adults has rate of malignant tumors in women, but much
been extremely controversial (Inskip et al, 1995), higher rates of meningioma, which are primarily
largely because a true increase in incidence is so benign (Ostrom et al, 2013). Approximately 75%
difficult to distinguish from an apparent increase of all gliomas are astrocytomas or glioblastomas
due to improved diagnosis. During the period multiforme (Ostrom et al, 2013). Data from the
1975–​87, incidence rates increased by 1.4% annu- US Central Brain Tumor Registry are given in
ally and mortality increased by 0.4% annually in Table 24-​1 (Ostrom et al, 2013).
the United States (Howlader et al, 2015). Against International comparisons show similar rates
these modest increments, incidence rates tripled of brain tumors in North American and western
in persons over age 75 years between 1973 (6.0 Europe, with somewhat lower rates in China
per 100,000) and 1985 (18.8 per 100,000) (SEER, and Japan (Figs. 24-​3, 24-​4). The elevated rates
2015). While some fraction of that marked in Sweden and Poland may reflect differences in
increase in the elderly may have been real, it also classification of tumor subtypes, such as inclu-
seems likely that improvements in diagnostic sion of nonmalignant tumors. Nevertheless, the
technology led to diagnoses of brain tumors that variation in incidence rates is far more modest
would not have been recognized as such in earlier than for more common cancers such as those of
periods. In locations and time periods in which the breast, lung, colon, liver, and cervix.
accurate diagnoses are uncertain, the large num- Within the United States, whites have higher
ber of tumors that originate in other sites, such as rates of glioma than blacks, whereas blacks show
the lung or breast, and metastasize to the brain an excess of meningiomas (Ostrom et al, 2013).
create the potential for misclassification of met- Asians have lower rates of both glioma and
astatic as primary brain cancers. Since increases meningioma (Howlader et  al, 2015). Incidence
were more pronounced in countries with the rates among Hispanics fall between whites and
most advanced medical care systems, and were blacks for glioma and are similar to whites for
concentrated in the elderly, who may well have meningioma (Ostrom et  al, 2013). Higher rates
previously been misdiagnosed as having stroke of meningioma have also been observed among
or degenerative neurologic disease, Inskip and individuals of Jewish ancestry (Inskip et  al,
589

TABLE 24-​1   DISTRIBUTION AND INCIDENCE RATES, PER 100,000 PERSON-YEARS,

OF PRIMARY BRAIN AND CENTRAL NERVOUS SYSTEM TUMORS BY HISTOLOGY
(UNITED STATES, CENTRAL BRAIN TUMOR REGISTRY OF THE UNITED STATES,
2007–2011)

% of all histology Mean reported Age-​adjusted

brain tumors
age at diagnosis incidence rate

Tumors of Neuroepithelial Tissue 30.6 56 6.61

Pilocytic astrocytoma 1.4 13 0.34
Diffuse astrocytoma 2.5 48 0.55
Anaplastic astrocytoma 1.7 53 0.37
Unique astrocytoma variants 0.3 23 0.07
Glioblastoma 15.4 64 3.19
Oligodendroglioma 1.2 43 0.26
Anaplastic oligodendroglioma 0.5 49 0.11
Oligoastrocytic tumors 0.9 42 0.21
Ependymal tumors 1.9 44 0.42
Glioma malignant, NOS 2.0 37 0.46
Choroid plexus tumors 0.2 19 0.05
Other neuroepithelial tumors 0.0 32 0.01
Neuronal and mixed
  Neuronal-​glial tumors 1.2 27 0.28
Tumors of the pineal region 0.2 33 0.04
Embryonal tumors 1.1 9 0.26

Tumors of Cranial and Spinal Nerves 8.1 55 1.70

Tumors of Meninges 37.3 65 7.88

Meningioma 36.1 65 7.61
Mesenchymal tumors 0.4 48 0.08
Other neoplasms related to the meninges 0.8 49 0.18

Lymphomas and Hemopoietic Neoplasms 2.1 65 0.46

Germ Cell Tumors and Cysts 0.4 17 0.10

Tumors of the Sellar Region 15.9 50 3.47

Pituitary 15.1 51 3.29
Craniopharyngioma 0.8 42 0.18

Unclassified Tumors 5.5 63 1.19

Hemangioma 1.4 49 0.30
Neoplasm, unspecified 4.1 70 0.88

All other 0.0 60 0.00

Total 100.0 59 21.42

*Rates are per 100,000 and are age-​adjusted to the 2000 US standard population.
590

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 2 4 6 8 10 12
ASR (world) per 100,000 person-years

FIGURE 24-​1  Age-​standardized (to the 2012 world population) incidence rates of brain cancer among women.
Source: Ferlay et al, 2013.

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 2 4 6 8 10
ASR (world) per 100,000 person-years

FIGURE 24-​2  Age-​standardized (to the 2012 world population) incidence rates of brain cancer among men.
Source: Ferlay et al, 2013.

50
New cases per 100,000

40
person-years

30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

TheNetherlands United Kingdom United States of America

FIGURE 24-​3  Age-​specific incidence rates of brain cancer among women.

Source: Ferlay et al, 2013.
591

Brain Cancer 591

50

New cases per 100,000

40

person-years
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

TheNetherlands United Kingdom United States of America

FIGURE 24-​4  Age-​specific incidence rates of brain cancer among men.

Source: Ferlay et al, 2013.

2003), and a recent study suggests that Iran-​born based on case-​ control study data (Shete et  al,
Jews have higher rates of meningioma than Jews 2009, Wrensch et al, 2009); three of these regions
from other origins (Barchana & Liphshitz, 2013). (RTEL, TERT, and CDKN2B) were replicated in
a GWAS analysis consisting primarily of cohort
GENETIC AND MOLECULAR studies (Rajaraman et  al., 2012). Consistent sig-
EPIDEMIOLOGY nals at 7p11.2 (EGFR) have also been reported
Studies of inherited genetic influences on the in three studies (Rajaraman et al., 2012, Wrensch
occurrence of brain tumors and of the distribu- et  al., 2009, Sanson et  al., 2011). One GWAS on
tion and determinants of somatic genetic altera- meningioma identified a new susceptibility locus
tions in these tumors are both active avenues of for meningioma at 10q12.31 (Dobbins et al, 2011).
investigation. Associations between rare famil- Candidate approaches have also provided val-
ial syndromes and risk of brain tumors are well uable data on genetic susceptibility and risk of
known and include neurofibromatosis (type 1 and glioma; potential associations have been noted
2), von Hippel-​Lindau syndrome, Li-​Fraumeni in genes involved in carcinogen metabolism (Gu
syndrome, Turcot’s syndrome, familial adenoma- et al., 2009), DNA repair (Jiang et al, 2013, Gu et al,
tous polyposis (FAP), hereditary nonpolyposis 2009), inflammation/​allergies (Backes et al, 2013,
colorectal cancer (HNPCC), and basal cell nevus Schwartzbaum et  al, 2005, Amirian et  al, 2010,
syndrome. Wiemels et al, 2007), and other immune response
The increase in risk associated with having a genes (Michaud et al, 2013). Research to elucidate
relative with cancer is modest for gliomas (Burch precise pathways that are critical to brain tumor
et al, 1987; Preston-​Martin et al, 1989a; Wrensch development will be an important next step.
& Barger, 1990; Cicuttini et  al, 1997), with rel- Somatic tumor characteristics have also been
ative risks (RRs) of 1.6 or below and in some studied in an effort to identify etiologically dis-
studies very close to null (Wrensch et al, 1997a). tinctive subtypes. Glioblastomas show losses at
A history of brain tumors in relatives was associ- chromosomes 17p, 10q, 13q, 22q, 19q, and 9p
ated with a twofold-​increased risk in one study and gains at chromosome 7 (Inskip et al, 1995).
(Wrensch et al, 1997a). An evaluation of mutations in the tumor sup-
pressor gene p53 demonstrated a high frequency
Genetic Susceptibility of transversion mutations (Li et al, 1998).
Several large genome-​ wide association stud-
ies (GWAS) have been conducted to measure RISK FACTORS
the association between hundreds of thousands While a large number of epidemiological studies
of polymorphisms and the risk of brain tumors. have examined a wide range of risk factors for
These studies identified glioma risk loci in a brain cancer, few of these risk factors are widely
number of genes (TERT, CCDC26, CDKN2A/​ accepted as causal. We present data for risk of
CDKN2B, RTEL1, and PHLDB1). Susceptibility brain tumors overall and for meningioma and
loci in these genes were found in the initial GWAS glioma when available.
592

592 Textbook of Cancer Epidemiology

Tobacco little evidence for the NOC hypothesis, at least as

Tobacco is not associated with brain tumors based it pertains to adult gliomas.
on the evidence gathered thus far. As tobacco Vitamins C and E inhibit nitrosation reac-
smoke contains high numbers of carcinogens, tions in vivo, and intake of these vitamins can
including nitrosamines, and is known to cause a reduce the endogenous formation of NOCs in
number of different cancers, the association with the stomach. Inverse associations with vitamin C
primary brain tumors has been extensively stud- diet or supplemental intake have been observed
ied. Most case-​control and cohort studies have in a few case-​control studies (Preston-​Martin
reported no associations between tobacco smoke et  al, 1989; Hu et  al, 1999; Blowers et  al, 1997;
and the risk of meningiomas or gliomas. In 2004, Burch et al, 1987), but others observed no over-
the US Surgeon General’s Report on the Health all associations (Chen et al, 2002; Lee et al, 1997).
Consequences of Smoking concluded, “the evidence In two large prospective cohort studies, vitamins
is suggestive of no causal relationship between C and E from diet and supplements were not
smoking cigarettes and brain cancer in men and related to risk of glioma (Dubrow et  al, 2010;
women” (US Department of Health and Human Michaud et  al, 2009). As with vitamin intake,
Services, 2004). Findings from studies of maternal findings on fruit and vegetable intake have been
smoking and risk of primary brain tumors in off- inconsistent. In a large prospective cohort study,
spring have been inconsistent, but overall the data fruit and vegetable intake, overall or within sub-
do not support an increase in brain tumor risk groups, was not associated with risk of glioma
with maternal smoking (Johnson et al, 2014). (Holick et al, 2007).
Nitrates in drinking water have also been
Diet studied in relation to brain tumor occurrence,
Compared to the promising data for other cancer in the context of the same hypothesis of an eti-
sites, the evidence supporting dietary influences ologic pathway involving nitrosamines. A  large
on the etiology of adult brain tumors is weak study based in the United States considered water
overall. nitrate intake over a 20-​year period in conjunc-
The most extensively studied category of tion with dietary nitrate and nitrite intake; essen-
foods is cured meats and fish, motivated by a tially no link with risk of gliomas, nor evidence
hypothesized etiologic pathway in which nitrites for effect modification in relation to vitamin C
are converted to n-​nitrosamines (NOCs), estab- intake, was found (Ward et al, 2005).
lished carcinogens in animal systems. In most
case-​control studies of gliomas—​but not in all Tea and Coffee
(Kaplan et al, 1997a) and not always equally for Tea and coffee contain numerous antioxidants
both sexes (Giles et  al, 1994; Lee et  al, 1997)—​ that may have anticarcinogenic properties, and
consumption of cured meats or fish was associ- these beverages have been examined in rela-
ated with an increased glioma risk (Burch et al, tion to numerous cancer types, including brain
1987; Boeing et al, 1993; Giles et al, 1994; Blowers cancer. In a meta-​analysis, inverse associations
et al, 1997; Lee et al, 1997). Evidence supporting were observed for tea (RR 0.88, 95% confidence
an association of cured meat consumption with interval [CI] 0.69–​1.12) and coffee plus tea (RR
meningioma has also been reported (Preston-​ 0.75, 95% CI 0.54–​1.05) comparing highest with
Martin et  al, 1980; Preston-​ Martin & Mack, lowest categories of consumption (Malerba et al,
1991). Attempts to examine nitrate and nitrite 2013). While the overall summary estimates were
consumption directly in these same studies have not statistically significant in the meta-​analysis,
not strengthened the pattern of association and, several of the largest prospective studies reported
in many cases, these compounds have shown significant inverse trends for one or both of these
weaker associations than those found for cured beverages (Michaud et  al, 2010b; Holick et  al,
meats (Boeing et  al, 1993; Giles et  al, 1994). In 2010; Dubrow et al, 2012).
two prospective cohort studies, with over 300
incident cases of glioma in each study, intakes Alcohol
of meats and foods high in nitrites, nitrates, or Total alcohol intake—​as well as beer, wine, and
nitrosamines were not associated with risk of gli- liquor separately—​have been examined in rela-
oma (Michaud et al, 2009; Dubrow et al, 2012). tion to both adult gliomas and meningiomas. The
Data from these two large studies, which are less historically high levels of nitrosamines found in
prone to bias than case-​control studies, provide beer led to the speculation that consumption of
593

Brain Cancer 593

beer may increase the risk of brain tumors. To individuals and 54% in obese individuals, com-
date, no consistent association has been demon- pared to those with a healthy body mass index
strated between consumption of different types (BMI, <25  kg/​m2) (Niedermaier et  al, 2015).
of alcoholic beverages and the risk of gliomas or The relation between obesity and meningioma
meningiomas. Evidence for absence of an asso- may be a mediated by elevated levels of endog-
ciation between beer and liquor consumption enous hormones produced by fat cells. Findings
and glioma is fairly consistent in case-​control for obesity and the risk of glioma have been less
studies (Burch et al, 1987; Preston-​Martin et al, consistent, and the summary association is null
1989a; Ryan et al, 1992a; Lee et al, 1997) and large (Niedermaier et al, 2015). In one large prospec-
prospective cohort studies (Benson et  al, 2008, tive cohort study, individuals who were obese
Braganza et al, 2014). at age 18 were almost at a fourfold higher risk
of glioma compared those reported a healthy
Reproductive Factors weight at age 18 (Moore et al, 2009), however, no
The striking difference between incidence rates associations were noted for BMI at later ages in
in meningioma and glioma by sex suggests that that study.
hormones play a role in the development of these Height has been positively associated with
tumors. For the most part, reproductive factors, risk of glioma in a number of studies. In a large
including age at first birth, parity, menopausal pooling study of 13 prospective cohort stud-
status, and type of menopause, have not been ies and two case-​control studies, a strong posi-
associated with either meningiomas or glioma. tive association was observed for men (≥ 190 vs.
Two prospective cohort studies reported higher 170–​174  cm, pooled odds ratio [OR] 1.70, 95%
risk of glioma among women who had a later CI 1.11–​2.61; p-​trend  =  0.01), and associations
age at menarche (Kabat et al, 2011; Silvera et al, were stronger for glioblastoma than for all glio-
2006), but no association was observed for age at mas, but the associations were weaker in women
menarche in a large European prospective study (Kitahara et  al., 2012). The association between
(Michaud et al, 2010a). height and glioma risk may be explained by early
life biological exposures, such as levels of insulin-​
Hormones like growth factor (IGF) or growth hormones.
Consistent with a lack of association observed with Among children, head circumference at birth
reproductive factors, no associations have been has been positively associated with brain cancer
reported for exogenous hormone use and risk of risk, suggesting that brain pathology may orig-
glioma (Michaud et al, 2010a; Kabat et al, 2011). inate during fetal life (Samuelsen et  al, 2006).
In contrast, three prospective cohort studies have Birthweight and, in boys, greater height in child-
observed a higher risk of meningioma among hood were also associated with a higher risk of
women reporting current hormone therapy use glioma (Kitahara et al, 2012).
(Jhawar et al, 2003; Benson et al, 2010; Michaud
et  al, 2010a). No relation has been consistently Infections
observed with use of oral contraceptives for either Epidemiological evidence of infectious agents
glioma or meningioma risk (Wigertz et  al, 2006; as risk factors in for brain tumors is not conclu-
Huang et al, 2004; Hatch et al, 2005; Custer et al, sive. There has been considerable interest in the
2006; Hochberg et  al, 1990; Silvera et  al, 2006; potential relationship between simian virus 40
Benson et  al, 2008; Michaud et  al, 2010a). The (SV40), a polyomavirus, and brain tumor devel-
biological plausibility of a relation between exog- opment. However, long-​term follow-​up studies
enous hormone use and meningioma risk is sup- of individuals vaccinated with polio vaccines
ported by numerous case reports showing that contaminated with SV40 (1955–​63) did not find
pregnancy can lead to increased growth of menin- an overall increase in the risk of brain tumors
gioma. However, diagnostic bias cannot be ruled (Engels, 2005). Furthermore, no significant
out as an explanation for the elevated risk. association was observed between antibodies to
SV40 (or two other polyomaviruses, JC virus and
Anthropometric Measures BK virus) and primary malignant brain tumors
Studies have consistently reported positive asso- when measured directly in serum collected prior
ciations between body fat and risk for menin- to cancer diagnosis (Rollison et al, 2003).
giomas; in a recent meta-​analysis, the increase Certain infections may decrease the risk of
in risk of meningioma was 21% in overweight brain tumors. The mechanisms are unclear but
594

594 Textbook of Cancer Epidemiology

may be mediated through immune modula- These studies were less prone to reverse cau-
tion. Case-​control studies have reported inverse sation, as the bloods were collected before the
associations between prior infections, including diagnosis of disease. A U-​shaped associated with
chickenpox (varicella-​zoster virus [VZV]) and IgE plasma levels and glioma risk was observed
colds (Schlehofer et al, 1999; Wrensch et al, 1997; in one study (Calboli et al, 2011), similar to the
Wrensch et  al, 2001; Wrensch et  al, 2005)  and observed association in the earlier case-​control
brain tumors. Associations were not seen with study. In a different study (European Prospective
three other herpesviruses (Epstein Barr virus Investigation into Cancer), the risk of glioma
[EBV], cytomegalovirus [CMV], and herpes sim- was inversely related to allergic sensitization,
plex virus [HSV]) (Wrensch et al, 2001; Wrensch that is, specific IgE level ≥ 0.35 kUA/​L (kilo anti-
et al, 2005). Similar findings were observed in a body units per liter) with an OR of 0.73 (95%
prospective study using prediagnostic bloods; CI 0.51–​1.06) (Schlehofer et al, 2011). The third
elevated antibodies to VZV were marginally cohort study similarly reported a lower risk for
associated with lower risk of glioma (OR 0.68, glioma (OR 0.75, 95% CI 0.56–​0.99, >100 kUA/​L
95% CI 0.41–​1.13, highest vs. lowest quartile; p-​ vs. ≤100 kUA/​L); testing positive for total IgE
trend = 0.06), and no associations were observed at least 20 years before diagnosis was associated
for EBV, CMV, and adenovirus (Sjostrom with an even greater risk reduction compared
et al, 2011). with testing negative (OR 0.54, 95% CI 0.30–​
0.99) (Schwartzbaum et al., 2012). All three stud-
Allergies ies reported inverse associations with IgE levels
Observational studies on allergies, such as and risk of glioma, providing additional support
asthma or eczema, and the risk of glioma have for the role of allergies in the etiology of glioma.
been strikingly consistent, and several plausible The high degree of consistency in all reports,
biological mechanisms exist that might explain despite differences in study designs and geo-
the role of allergies in glioma development. In graphical location of the study, suggests that
a meta-​analysis based on 3,450 cases of glioma, an etiological basis for allergies and the related
a 39% decrease in risk of glioma was observed immune response may exist. Taken together,
among those reporting any form of allergy (RR studies on infections and allergies suggest that
0.61, 95% CI 0.55–​ 0.67); the association was the adaptive immune response plays an impor-
similar for those reporting asthma or eczema tant role in glioma etiology. Recent animal mod-
(Linos et al, 2007). Two studies focused on aller- els have demonstrated enhanced glioma growth
gies to specific exposures (e.g., medicine, foods) in mice deficient for histamine decarboxylase
and reported inverse associations for most types (HDC), compared to wild-​type mice, and these
of allergies (Brenner et  al, 2002; Wiemels et  al, mice experience increased frequency of imma-
2002) and a dose-​response relationship between ture myeloid cells and immune suppression of
increasing numbers of specific allergens and CD8 + T cells in the tumor microenvironment
lower risk for glioma. In a case-​control study, (Ahn et al, 2015).
individuals with elevated immunoglobulin (IgE)
levels (>100 kU/​L) had an OR of 0.37 (95% CI Physical Activity
0.22–​0.64), and individuals with intermediate IgE To date, six studies have examined the role of
levels (25 to 100 kU/​L) had an OR of 0.64 (95% physical activity on risk of brain tumors. The
CI 0.42–​0.96) for glioma risk, compared to indi- results were recently summarized in a meta-​
viduals with normal levels (<25 kU/​L) (Wiemels analysis, which concluded that a modest inverse
et  al, 2004). A  similar finding was observed in association of physical activity with meningioma
a follow-​up study, although the association was risk, and a weak association with glioma risk,
weaker than in the first study (Wiemels et  al, may exist (Niedermaier et al, 2015).
2009). An interaction between history of aller-
gies and two of the genetic risk loci detected in Ionizing Radiation
GWAS was recently observed in relation to risk The only well-​ established exogenous cause of
of glioma (Schoemaker et al, 2010). adult brain tumors is exposure to high doses of
The association between blood IgE levels and ionizing radiation. The direct link between local-
risk of glioma was recently examined in three ized exposure to the head and development of
large prospective studies (Calboli et  al, 2011; brain tumors is strongest for meningiomas but
Schlehofer et al, 2011; Schwartzbaum et al, 2012). is also present for gliomas (Ron et al, 1988). The
59

Brain Cancer 595

most informative and widely accepted exposure-​ Preston-​Martin et al, 1989b) indicate fairly strong
response estimates relating ionizing radiation to associations between dental x-​rays and meningi-
brain cancer come from the follow-​up study of omas, acoustic neuromas, and gliomas, with RR
atomic bomb survivors (Preston et al, 2002). This estimates on the order of 2–​4. In a recent study
is a very large population (>80,000 persons) with with detailed information on dental history,
clearly quantified dose information and detailed an elevated risk of meningioma was observed
cancer incidence data, although the nature of the among individuals who reported six or more x-​
acute, high-​dose exposure and other aspects of rays involving full-​mouth series (OR 2.06, 95%
this cohort may limit generalizability. For nerv- CI 1.03–​4.17), but the associations were strongest
ous system tumors in the aggregate, the excess RR among those who had frequent dental x-​rays in
per sievert (Sv) is 1.2 (95% CI 0.6–​2.1). The high- years when doses were higher (Longstreth et al,
est excess risk observed was for schwannomas 2004). In the same study, associations were not
(ERR[SV] 4.5, 95% CI 1.9–​9.2). Corresponding observed for bitewings, lateral cephalometric,
excess relative risk estimates per Sv are 0.56 (95% and panoramic radiographs. In contrast, Ryan et
CI -​0.2–​2.0) for gliomas and 0.64 (95% CI -​0.01–​ al (1992b) found an inverse association of diag-
1.8) for meningioma, with some indication of nostic x-​rays with adult gliomas, whereas other
greater effect among males for all tumor types studies (Burch et al, 1987; Schlehofer et al, 1992)
(Preston et al, 2002). reported no association.
High doses of radiation were once used to Occupational exposure to radiation has also
treat tinea capitis (ringworm), and several of been evaluated, with mixed results. Increases
these patient cohorts have been followed for their in risk have been found in a number of groups
cancer experience. Cohorts of treated children with the potential for elevated exposure to ioniz-
from Israel (Modan et al, 1974) and New York ing radiation (Wilkinson et al, 1987; Checkoway
(Shore et al, 1976) have an increased incidence of et  al, 1988), but the large pooled studies of
brain tumors, both benign and malignant, dur- nuclear industry workers revealed no association
ing adulthood. The magnitude of the association (Gilbert et al, 1993; Carpenter et al, 1994).
was greatest for nerve sheath tumors (RR 33), The circumstances in which high doses of
intermediate for meningioma (RR around 10), ionizing radiation are encountered are limited,
and lowest for glioma (RR 2.6) (Ron et al, 1988). which makes the overall contribution of high-​
The strength of the association was consistently dose ionizing radiation to brain tumor incidence
greater for benign tumors (Little et al, 1998). modest, and of decreasing importance over time
With an additional 16 years of follow-​up of the with the reduced use of radiation for benign con-
10,834 cohort, the RR for meningioma remained ditions (Inskip et  al, 1995). Whether low-​dose
high (overall, RR 4.63; 95% CI 2.43–​9.12) and exposure causes brain cancer is unclear, but it
was elevated regardless of age at exposure or is unlikely to have a major influence on brain
latency period (Sadetzki et al, 2005). In the same tumor etiology at the population level.
study, age at exposure did matter and the high-
est risk of malignant glioma was observed among Occupation
children who were exposed when they were less There is a large, somewhat diffuse literature on
than 5 years old (RR 3.56; 95% CI 0.96–​9.91). a wide range of occupations and occupational
The same pattern of association has been seen exposures in relation to brain tumors (Wrensch
in adults treated with radiation for pituitary et  al, 1993). An abundant number of studies
adenomas (Brada et al, 1992; Tsang et al, 1993), with limited power have reported isolated asso-
in infants treated with x-​rays for thymic enlarge- ciations with one job or another (Olin et al, 1987;
ment (Hildreth et al, 1985), and in children given Brownson et al, 1990; Kaplan et al, 1997b; Cocco
x-​ray therapy for tonsil hypertrophy (Schneider et  al, 1999), and a critical review of the earlier
et al, 1985). These studies provide a clear indica- studies has been published (Thomas & Waxweiler,
tion that ionizing radiation is capable of causing 1986). The topics discussed below are those for
both benign and malignant tumors of the brain, which there is some convergence of the literature.
but the exact nature of the dose-​response gra-
dient is not clear. A series of investigations by Vinyl Chloride
Preston-​Martin and colleagues (Preston-​Martin Summarizing a series of studies, Thomas and
et al, 1980; Preston-​Martin et al, 1982; Preston-​ Waxweiler (1986) reported positive associa-
Martin et al, 1983; Preston-​Martin et al, 1989a; tions between occupational exposure to vinyl
596

596 Textbook of Cancer Epidemiology

chloride and brain tumor risk, with relative risks about the presence of an increased risk (Wong &
in the range of 1.5–​5.0. Some, but not all, of the Raabe, 1989). Reports vary markedly in whether
later studies tend to support a possible positive they are based on a broad census grouping of
association (Hagmar et  al, 1990; Simonato et  al, occupations associated with petroleum and
1991; Wong et  al, 1991). A  large pooled analy- coal products (McLaughlin et al, 1987; Preston-​
sis of workers from 56 plants in North America Martin et al, 1989c), refinery workers (Thériault
and Europe provides the most comprehensive & Goulet, 1979; Wen et  al, 1983; Demers et  al,
data on vinyl chloride and cancer risk with over 1991), or a mixture of refinery and related chem-
22,000 workers and up to 50 years of follow-​up; ical workers (Thomas et  al, 1980; Marsh et  al,
no associations were observed for brain cancer 1991; Teta et al, 1991). The lack of specificity in
(standardized mortality ratio [SMR] 1.18, 95% CI exposure makes it difficult to compare studies or
0.90–​1.52) (Bosetti et  al, 2003). Consistent with for the research to build momentum toward elu-
this finding, a case-​cohort study of workers in one cidating an etiologic pathway.
polymer production plant in Louisville, Kentucky,
where brain cancer mortality had been very high, Electrical Workers
concluded that the excess risk was not associated Conclusions by reviewers of this literature range
with vinyl chloride exposure levels, in sharp con- from perceived support for a weak association
trast to the associations with angiosarcomas in between occupational electric and magnetic
the same plant (Lewis & Rempala, 2003). fields and brain tumors (Kheifets et  al, 1995;
Wrensch et al, 1993) to a more reserved interpre-
Petrochemical Workers tation of the overall evidence (Inskip et al, 1995).
Since the mid-​1970s, at least 15 studies have eval- The reported RRs rarely exceed 2.0 and are more
uated cancer among workers in the petroleum-​ often in the very ambiguous range of 1.2–​1.5.
refining and petrochemical industries. The In this range, the potential for some form of
primary focus in those occupational studies has selection bias for employment in electrical occu-
been solvent exposure, an association addressed pations is difficult to exclude (Kheifets et  al,
specifically in some community-​ based studies 1995). Early studies were based largely on job
as well (Rodvall et  al, 1996). Among the petro- titles on death certificates, cancer registries, or
chemical industry studies are several that report community-​based case-​control studies (Kheifets
positive but imprecise associations between the et al, 1995), often with no explicit effort to exam-
potential for chemical exposure and brain tumor ine exposure to electromagnetic fields. However,
incidence or mortality (Thériault & Goulet, 1979; the sophistication of exposure assessment has
Thomas et  al, 1982; Austin & Schnatter, 1983; risen steadily in both occupational cohort studies
Waxweiler et  al, 1983; Teta et  al, 1991; Cooper (Sahl et  al, 1993; Thériault et  al, 1994; Savitz &
et al, 1997; Sathiakumar et al, 2001; Divine et al, Loomis, 1995; Harrington et al, 1997) and case-​
1999). A number of studies of roughly compara- control studies, primarily in Sweden (Floderus
ble quality, however, fail to corroborate this pat- et al, 1993; Feychting et al, 1997) and in Canada
tern (Wen et  al, 1983; Wong et  al, 1986; Marsh (Villeneuve et al, 2002).
et  al, 1991; Tsai et  al, 1993). The challenge in Early studies generated some strikingly posi-
interpreting these studies is to identify whether tive findings (Lin et al, 1985; Speers et al, 1988),
there is some aspect of petrochemical industry as well as null results (Pearce et al, 1989). Among
employment or exposure that has an etiologic the largest, most detailed studies, the findings
relationship with brain tumors in the aggregate continue to vary, with suggestions of an associ-
or with subsets of brain cancer. A study in Taiwan ation in some (Savitz & Loomis, 1995; Thériault
reported higher risk of brain cancer in subjects et al, 1994) but not all (Sahl et al, 1993; Harrington
residing in municipalities with high levels of pet- et al, 1997; Sorahan, 2012) of the cohort studies
rochemical air pollution (OR 1.65, 95% CI 1.00–​ of electric utility workers. A  pooled analysis of
2.73) (Liu et al, 2008). several of these studies identified a weak positive
The difficulty in summarizing the evidence gradient in risk, with RR estimates rarely exceed-
is that some association may be present, but lit- ing 2.0 even at the most elevated exposure levels
tle progress has been made toward pinpointing (Kheifets et al, 1999).
the exact exposure that could be responsible or The category of “electrical workers” is quite
the type of brain tumor most likely to be affected. diverse, encompassing welders, construc-
Some reviewers draw a more reserved conclusion tion electricians, electric utility linemen, and
597

Brain Cancer 597

electrical equipment operators, and there is lit- be present. A meta-​analysis, including 33 studies
tle consistency across studies in who is included published in the 1980s and 1990s (Khuder et  al,
under this rubric. Efforts to identify occupational 1998), found a 30% excess risk of brain tumors
risks of brain tumors could be more productive among farmers. There are many aspects of lifestyle
if specific jobs were evaluated. The diversity of and occupational exposure that could distinguish
exposures encountered within any one of these agricultural workers, such as reduced contact with
jobs, let alone among the aggregation of electrical some contagious infectious agents in rural areas,
workers, is extensive and includes a wide range of as well as a direct link to specific pesticides or
solvents and polychlorinated biphenyls (PCBs). some other occupational hazard. Several studies
have focused more directly on specific pesticides
(Musicco et al, 1982; Musicco et al, 1988; Ahlbom
Polychlorinated Biphenyls
et al, 1986). Studies with more detail on individ-
Polychlorinated biphenyls were manufactured
ual exposure to pesticides and herbicides have
for a variety of industrial applications, and
not reported positive associations with glioma
because of their persistence in the environment
risk (Yiin et al, 2012; Samanic et al, 2008), but an
and slow excretion in humans, their long-​term
elevated risk was observed for herbicide use and
health effects, including cancer, have been of
meningiomas in one study (Samanic et al, 2008).
concern. The epidemiologic evidence for occu-
pational exposure to PCBs and cancer, including
Rubber and Tire Industry
brain cancer, was summarized by Longnecker
A series of studies from the 1960s through the
et  al (1997). Data from a series of occupational
1980s considered brain tumor incidence and
studies, all with nine or fewer cases, were pooled,
mortality among rubber industry workers.
yielding an RR of 1.3 (95% CI 0.8–​2.0), thus pro-
Several of them reported marked but extremely
viding little evidence of an increased risk of brain
imprecise elevations in risk (Mancuso, 1963;
cancer. No overall excess mortality from brain
Mancuso et  al, 1968; Monson & Fine, 1978;
cancer was reported for cumulative PCB expo-
Preston-​Martin et al, 1989a), while other, meth-
sure estimated using plant-​specific job exposure
odologically similar studies found no support
in a large study of three electrical capacitor man-
for an association (Fox et  al, 1974; McLaughlin
ufacturing plants (Ruder et al, 2014).
et al, 1987; Sorahan et al, 1989). Although some
recent publications reported similarly impre-
Health Professions
cise elevated risks of brain cancer among rub-
Health professionals, including physicians
ber workers (Straif et al, 2000; Wilczynska et al,
(McLaughlin et al, 1987; Preston-​Martin, 1989),
2001), no excess mortality from brain cancer
dentists (Ahlbom et al, 1986), pathologists (Hall
was detectable in a meta-​analysis summarizing
et al, 1991), veterinarians (Blair & Hayes, 1980),
risk estimates from 20 studies of rubber workers
and mixed groups of healthcare industry workers
(RR 0.90, 95% CI 0.79–​1.02) (Borak et al, 2005).
(Thomas et al, 1986; McLaughlin et al, 1987) have
As the technology and industrial hygiene of the
an increased incidence of brain tumors. Even if
rubber industry have changed, the older studies
this increased incidence can be ascribed in part
are of limited value in present-​day evaluation of
to the presumably better access of healthcare
risks. As with petrochemical workers, the num-
workers to health services, there must be a con-
ber of positive reports lends some credence to
cern that it reflects incomplete diagnosis in the
the hypothesis that agents encountered at some
population at large. On the other hand, health-
time in the past in the rubber industry may be a
care workers are likely to have higher exposure to
cause of brain tumors, but many exposures, such
ionizing radiation and to a wide range of poten-
as nitrosamines, are more regulated (and lower)
tially toxic agents such as anesthetics, solvents,
in those industries today and unlikely to be sig-
and pharmaceuticals.
nificant cause of brain cancer mortality.
Agriculture and Pesticides
Several studies have reported an elevated risk of Medical Conditions and Treatment
brain tumors among agricultural workers (Blair Several studies have indicated that persons with
et al, 1985; Musicco et al, 1988; Reif et al, 1989). epilepsy and those who receive anticonvulsant
The magnitude of the increase is inconsistent drugs to treat their epilepsy are at increased risk
and generally modest, but the replication of this of developing brain tumors (White et  al, 1979;
finding suggests that a clue to the etiology may Shirts et  al, 1986; Olsen et  al, 1989; Schlehofer
598

598 Textbook of Cancer Epidemiology

et al, 1992; Wrensch et al, 1997a). However, there study (OR 2.2; 95% CI 1.12–​ 4.67) (Preston-​
is reason to suspect that epilepsy is a consequence Martin et  al, 1989). A  similarly elevated risk of
rather than a cause of the brain tumor, particu- acoustic neuroma was observed among indi-
larly since the increased risk is most pronounced viduals reporting loud noise exposure from any
when a diagnosis of epilepsy and the develop- source in a larger study (OR 1.55, 95% CI 1.04–​
ment of brain cancer are close together in time. 2.30) (Edwards et al, 2006); the associations were
stronger among individuals with exposure to
Other Risk Factors loud noise from machines, power tools, or con-
Head Trauma struction equipment (OR 1.79, 95% CI 1.11–​2.89)
Head trauma has been evaluated in a number of or to loud music (OR 2.25, 95% CI 1.20–​4.23)
studies as a potential risk factor for the develop- (Edwards et al, 2006). No associations were found
ment of brain tumors, but a causal relationship with occupational noise exposures in a recent
seems unlikely. study, but higher risks were observed from self-​
Because of the impression among the pub- reported exposures to noise (e.g., loud music,
lic that a link between head trauma and brain sporting events) (Fisher et  al, 2014), suggesting
cancer might exist, there is great concern about that recall bias may explain the positive findings.
recall bias; brain tumor patients or their fami-
Nonoccupational Electric
lies might overreport or perhaps remember such
and Magnetic Field Exposure
events more completely than persons not affected
In contrast to the sizable and weakly positive
by brain tumors. The strategy to minimize this
results of occupational studies, there is little basis
potential bias has been to focus on more severe,
for further studies of residential magnetic fields
documented head trauma, such as injury result-
and brain tumors.
ing in loss of consciousness or in medical treat-
Several studies have examined the role of
ment. When those stringent criteria are applied,
magnetic fields encountered in homes and
an association, albeit modest, generally remains
through the use of electric appliances in relation
(Ahlbom et  al, 1986; Burch et  al, 1987). Most
to brain cancer. While much of this research has
studies have focused on malignant gliomas (Hu
focused on childhood brain tumors, a few studies
et al, 1998), but head trauma has also been asso-
have considered adult brain tumors. The review
ciated with the risk of meningiomas (Preston-​
by Li et al (1996) identified only four studies of
Martin et  al, 1980; Preston-​Martin et  al, 1983).
residential magnetic field exposure and adult
In a large prospective study of patients hospital-
brain tumors, two of which were too small to
ized due to brain injury in Sweden, however, the
be informative. Of the other two, one reported
RR of all brain tumors was 1.0. Separate analyses
a positive association between wire code and
by severity of trauma, duration of follow-​up, and
central nervous system tumors (Wertheimer &
main histopathologic type consistently indicated
Leeper, 1982), and the other reported no associa-
a lack of association (Nygren et al, 2001).
tion (Feychting & Ahlbom, 1994).
One proposed biological mechanism for the
speculative link between head traumas and the Drinking Water Disinfection Byproducts
risk of brain tumors is that tissue repair after Although most research on the possible carci-
trauma requires increased cellular division, nogenicity induced by the chlorination of public
which could lead to an increased risk of genetic drinking water supplies has focused on bladder
changes, with possible activation of cellular and colorectal cancers, there are studies that
oncogenes (Preston-​Martin et al, 1990). Another suggest a potential association with brain can-
suggested mechanism is cocarcinogenic action cer as well (Cantor et al, 1978; Young et al, 1981;
through inflammation in conjunction with for- Gottleib & Carr, 1982). A later study by Cantor
eign material (Barnett et  al, 1986). Along simi- et al (1999) provided perhaps the strongest evi-
lar lines, an association between acoustic trauma dence to date for such an association based on
in the form of loud noise and acoustic neuromas the quality of the methods and the observation of
has been reported (Preston-​Martin et al, 1989b). a dose-​response gradient, although the associa-
tion was limited to men.
Noise Trauma
Acoustic neuroma (nerve sheath tumor of cranial Cellular Telephones
nerve VIII) was associated with long-​term expo- Cellular telephones emit low-​ power electro-
sure to acoustic noise (trauma) in a case-​control magnetic radiation in the radiofrequency range,
59

Brain Cancer 599

which has been hypothesized to contribute to the number of epidemiologic studies (Gurney et  al,
etiology of brain cancer. The early epidemiologic 2001; Johnson et al, 2014). A sizable number of
studies were largely negative, including a large, studies have evaluated pesticide exposure, mag-
well-​designed study in the United States (Inskip netic fields, and parental occupational exposures,
et  al, 2001)  and other studies from that period but in no case has there been clear support for
in the United States and Europe (Muscat, 2000; an etiologic effect (Ross & Spector, 2006). Other
Johansen et  al, 2001; Auvinen et  al, 2002). An suggestive evidence exists for parental age, birth
extremely ambitious international multicenter defects, fetal growth, and use of prenatal vitamins,
cellular phone case-​control study (Interphone) while insufficient evidence remains for allergic
reported higher risks of glioma among those conditions, infection, and maternal medications
with the highest exposure levels (Group, 2010), (Johnson et al, 2014). Methodologic limitations,
although a lower risk was noted in the second-​ most notably regarding exposure assessment,
highest exposure category (Group, 2010). In con- hinder conclusions regarding environmental
trast, a large nationwide cohort study conducted agents. Given the limited understanding of the
in Denmark comparing phone subscribers to etiology of brain tumors in both children and
nonsubscribers reported no increase in risk of adults, the question of whether risk factors are
central nervous system cancers with 10 or more shared or distinct remains unresolved.
years of phone use (Frei et al, 2011). There con-
tinue to be occasional reports of an association CONCLUSION
between cellular phone use and brain cancer, The epidemiologic study of brain tumors faces
largely confined to one particular set of investi- distinct challenges due to the diverse histologic
gators (Hardell et al, 2013b; Hardell et al, 2013a) types and the possibility of incomplete diagno-
but with other teams also finding some support sis. Rates tend to be higher in North America and
(Coureau et  al, 2014). Given the rapid rise in western Europe than in Asia and South America.
use over a relatively short time, detailed data on Genetic syndromes and genetic susceptibil-
time trends in brain tumor incidence from high ity account for a very small proportion of cases.
quality registries are pertinent to assessing the Many of the risk factors associated with common
plausibility of an adverse effect of cellular phone cancers, such as tobacco, alcohol, and diet, do not
use. In fact, careful analyses of time trends pro- appear to play a role in the risk of brain cancer,
vide no support for an etiologic role of cellular although height has been associated with glioma
phones (Little et  al, 2012; Deltour et  al, 2012). and obesity increases the risk of meningioma. The
While continued monitoring of health regis- best-​established exogenous cause of brain cancer
tries and cohorts is warranted, the current lack is ionizing radiation, as demonstrated in popula-
of changes in incidence rates over time (with tions exposed to high doses by the atomic bomb
increasing mobile phone usage) and null associa- or medical treatment, but it accounts for only a
tions in large cohort studies provide reassurance small proportion of cases. Recent studies suggest
against causal associations (Ahlbom & Feychting, that individuals with allergies, and higher IgE
2011) though the controversy continues. levels, have lower risk of glioma, and while these
are not modifiable exposures, understanding the
RISK FACTORS underlying biological mechanisms that could lead
FOR CHILDHOOD to novel opportunities for treatment (Table 24-​2).
BRAIN TUMORS A number of leads have been reported regard-
As in adults, brain tumors in children include a ing associations with occupation, notably vinyl
diverse array of histologic types, with astrocyto- chloride exposure, petrochemical industry occu-
mas most common (around half of all tumors), pations, electrical industry occupations, PCB
followed by primitive neuroectodermal tumors, exposure, health professions, agriculture/​ pesti-
other gliomas, and ependymomas (Ries et  al, cides, and the rubber and tire industry. Because of
1999). As in adults, the only established risk fac- contradictory findings, small RRs, and methodo-
tor is high-​dose ionizing radiation to the head logic limitations, none of these associations with
and genetic syndromes (Johnson et  al, 2014). work setting or putative agents can be considered
Also, in parallel with studies of adult tumors, to have been established. Cellular telephones con-
NOCs in the form of cured meats and nitros- tinue to be of concern as a potential cause of brain
able drugs are of interest as potential contribu- cancer, but several large, well-​ designed studies
tors, with inconsistent support found in a small have found little empirical evidence to suggest
60

600 Textbook of Cancer Epidemiology

that the use of cell phones is associated with an substantial diversity of subtypes of this tumor,
increased risk. The literature on risk factors for as well as in the challenges of interviewing cases
childhood brain tumors has focused on NOCs and with a disease that has an impact on memory
selected environmental agents, with limited sup- and communication. Perhaps with the continu-
port for associations other than ionizing radiation. ing advances in noninvasive imaging technology,
The epidemiologic study of brain tumors refinements in registration and resulting descrip-
has not been successful in identifying modifia- tive epidemiology, improved biological markers
ble causes that could prevent its occurrence. The for infectious agents and other potential con-
reasons for this limited success may lie in the tributors, as well as unrelenting public concern,
it will be possible to develop novel hypotheses
accelerating the rate of progress.
TABLE 24-​2   EPIDEMIOLOGY OF BRAIN
TUMORS: RISK FACTOR SUMMARY REFERENCES
Ahlbom A, Feychting M. Mobile telephones and brain
Risk factor Direction of effect* tumours. BMJ 2011;343:d6605.
Ahlbom A, Navier IL, Norell S, Olin R Spannare B.
Well-​Confirmed Risk Factors Nonoccupational risk indicators for astrocytomas
Ionizing radiation exposure ↑↑ in adults. Am J Epidemiol 1986;124:334–​7.
Genetic syndromes and ↑ Ahn B, Kohanbash G, Ohkuri T, Kosaka A, Chen X,
susceptibility Ikeura M et  al. Histamine deficiency promotes
accumulation of immunosuppressive immature
Probable Relationship Exists, myeloid cells and growth of murine gliomas.
Based on Substantial Data Oncoimmunology 2015;4:e1047581.
Height ↑ (Glioma)
Amirian E, Liu Y, Scheurer ME, El-​Zein R, Gilbert
MR, Bondy ML. Genetic variants in inflammation
Obesity ↑ (Meningioma) pathway genes and asthma in glioma susceptibil-
Hormone therapy (HT) use ↑ (Meningioma) ity. Neuro Oncol 2010;12:444–​52.
History of allergies ↓↓ (Glioma) Austin SG, Schnatter AR. A case-​control study of
chemical exposures and brain tumors in petro-
Weak, if Any, Relationship chemical workers. J Occup Med 1983;25:321–​20.
Exists, Based on Substantial Data Auvinen A, Hietanen M, Luukkonen R, Koskela R-​S.
Tobacco —​ Brain tumors and salivary gland cancers among cel-
lular telephone users. Epidemiology 2002;13:356–​9.
Meat intake & nitrites from —​ Backes DM, Siddiq A, Cox DG, Calboli FC, Gaziano
processed foods JM, Ma J et al. Single-​nucleotide polymorphisms
Fruits & vegetable intake —​ of allergy-​related genes and risk of adult glioma. J
Alcohol —​ Neurol Oncol 2013;113: 229–​38.
Barchana M, Liphshitz I. High incidence of benign brain
Dental x-​rays —​
meningiomas among Iranian-​born Jews in Israel
Vinyl chloride —​ may be linked to both hereditary and environmental
Exposure to electromagnetic fields —​ factors. Asian Pac J Cancer Prev 2013;14:6049–​53.
Polychlorinated biphenyls —​ Barnett GH, Chou SM, Bay JW. Posttraumatic intra-
cranial meningioma: a case report and review of
Inconsistent Findings Based on the literature. Neurosurgery 1986;18:75–​8.
Substantial Data Benson VS, Pirie K, Green J, Bull D, Casabonne D,
Reeves GK, Beral V. Hormone replacement ther-
Cell phone use ↑
apy and incidence of central nervous system
Limited Study to Date tumours in the million women study. Int J Cancer
2010;127:1692–​8.
Physical activity ↓
Benson VS, Pirie K, Green J, Casabonne D, Beral V.
Pesticides ↑ Lifestyle factors and primary glioma and meningi-
Coffee & tea ↓ oma tumours in the Million Women Study cohort.
Br J Cancer 2008;99:185–​90.
*Arrows indicate the approximate magnitude of the relationship: ↑, Blair A, Hayes HM Jr. Cancer and other causes of
slight to moderate increase in risk; ↑↑, moderate to large increase
in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large death among U.S. veterinarians. Int J Cancer
decrease in risk; —​, no association 1980;25:181–​5.
601

Brain Cancer 601

Blair A, Malker H, Cantor KP, Burmeister L, Wiklund Cooper SP, Labarthe D, Downs T, Burau K, Whitehead
K. Cancer among farmers: a review. Scand J Work L, Vernon S et  al. Cancer mortality among pet-
Environ Health 1985;11:397–​407. roleum refinery and chemical manufacturing
Blowers L, Preston-​Martin S, Mack WJ. Dietary and workers in Texas. J Environ Pathol Toxicol Oncol
other lifestyle factors of women with brain glio- Cancer 1997;16:1–​14.
mas in Los Angeles County (California, USA). Coureau G, Bouvier G, Lebailly P, Fabbro-​Peray P,
Cancer Causes Control 1997;8:5–​12.l. Gruber A, Leffondre K et  al. Mobile phone use
Boeing H, Schlehofer B, Blettner M, Wahrendorf J. Dietary and brain tumours in the CERENAT case-​control
carcinogens and the risk for glioma and meningioma study. Occup Environ Med 2014;71:514–​22.
in Germany. Int J Cancer 1993;53:561–​5. Custer B, Longstreth WT Jr, Phillips LE, Koepsell TD,
Borak J, Slade MD, Russi M. Risks of brain tumors in Van Belle G. Hormonal exposures and the risk of
rubber workers: a metaanalysis. J Occup Environ intracranial meningioma in women: a population-​
Med 2005;47:294–​8. based case-​control study. BMC Cancer 2006;6:152.
Bosetti C, La Vecchia C, Lipworth L, McLaughlin JK. Deltour I, Auvinen A, Feychting M, Johansen C,
Occupational exposure to vinyl chloride and can- Klaeboe L, Sankila R et  al. Mobile phone use
cer risk: a review of the epidemiologic literature. and incidence of glioma in the Nordic countries
Eur J Cancer 2003;12:427–​30. 1979–​ 2008:  consistency check. Epidemiology
Braganza MZ, Rajaraman P, Park Y, Inskip PD, 2012;23:301–​7.
Freedman ND, Hollenbeck AR et  al. Cigarette Demers PA, Vaughan TL, Schommer RR. Occupation,
smoking, alcohol intake, and risk of glioma in the socioeconomic status, and brain tumor mortal-
NIH-​AARP Diet and Health Study. Br J Cancer ity: a death certificate-​based case-​control study. J
2014;110:242–​8. Occup Med 1991;33:1001–​6.
Brenner AV, Linet MS, Fine HA, Shapiro WR, Selker Divine BJ, Hartman CM, Wendt JK. Update of the
RG, Black PM, Inskip PD. History of allergies and Texaco mortality study 1947–​93. I.  Analysis of
autoimmune diseases and risk of brain tumors in overall patterns of mortality among refining,
adults. Int J Cancer 2002;99:252–​9. research, and petrochemical workers. Occup
Brownson RC, Reif JS, Chang JC, Davis JR. An anal- Environ Med 1999;56:167–​73.
ysis of occupational risks for brain cancer. Am J Dobbins SE, Broderick P, Melin B, Feychting M,
Public Health 1990;80:169–​72. Johansen C, Andersson U et  al. Common varia-
Burch JD, Craib KJ, Choi BC, Miller AB, Risch HA, tion at 10p12.31 near MLLT10 influences menin-
Howe GR. An exploratory case-​ control study gioma risk. Nat Genet 2011;43:825–​7.
of brain tumors in adults. J Natl Cancer Inst Dubrow R, Darefsky AS, Freedman ND, Hollenbeck
1987;78:601–​9. AR, Sinha R. Coffee, tea, soda, and caffeine intake
Calboli FC, Cox DG, Buring JE, Gaziano JM, Ma J, in relation to risk of adult glioma in the NIH-​
Stampfer M et al. Prediagnostic plasma IgE levels AARP Diet and Health Study. Cancer Causes
and risk of adult glioma in four prospective cohort Control 2012;23:757–​68.
studies. J Natl Cancer Inst 2011;103:1588–​95. Dubrow R, Darefsky AS, Park Y, Mayne ST, Moore SC,
Carpenter L, Higgins C, Douglas A, Fraser P, Beral V, Kilfoy B et al. Dietary Components related to N-​
Smith P. Combined analyses of mortality in three nitroso compound formation: a prospective study
United Kingdom nuclear industry workforces, of adult glioma. Cancer Epidemiol Biomarkers
1946–​1988. Radiat Res 1994;138:224–​38. Prev 2010;19:1709–​22.
Checkoway H, Pearce N, Crawford-​Brown DJ, Cragle Edwards CG, Schwartzbaum JA, Lonn S, Ahlbom
DL. Radiation doses and cause-​specific mortality A, Feychting M. Exposure to loud noise and
among workers at a nuclear materials fabrication risk of acoustic neuroma. Am J Epidemiol
plant. Am J Epidemiol 1988;127:255–​66. 2006;163:327–​33.
Chen H, Ward MH, Tucker KL, Graubard BI, Engels EA. Cancer risk associated with receipt
McComb RD, Potischman NA et al. Diet and risk of vaccines contaminated with simian virus
of adult glioma in eastern Nebraska, United States. 40:  epidemiologic research. Expert Rev Vaccines
Cancer Causes Control 2002;13:647–​55. 2005;4:197–​206.
Cicuttini FM, Hurley SF, Forbes A, Donnan GA, Feychting M, Ahlbom A. Magnetic fields, leukemia,
Salzberg M, Giles GG et  al. Association of adult and central nervous system tumors in Swedish
glioma with medical conditions, family and repro- adults residing near high-​ voltage power lines.
ductive history. Int J Cancer 1997;71:203–​7. Epidemiology 1994;5:501–​9.
Cocco P, Heineman EF, Dosemeci M. Occupational Feychting M, Forssén U, Floderus B: Occupational and
risk factors for cancer of the central nervous sys- residential magnetic field exposure and leukemia
tem (CNS) among U.S.  women. Am J Ind Med and central nervous system tumors. Epidemiology
1999;36:70–​4. 1997;8:384–​9.
602

602 Textbook of Cancer Epidemiology

Fisher JL, Pettersson D, Palmisano S, Schwartzbaum magnetic fields in relation to mortality from brain
JA, Edwards CG, Mathiesen T et  al. Loud noise cancer among electricity generation and transmis-
exposure and acoustic neuroma. Am J Epidemiol sion workers. Occup Environ Med 1997;54:7–​13.
2014;180:58–​67. Hatch EE, Linet MS, Zhang J, Fine HA, Shapiro WR,
Floderus B, Persson T, Stenlund C, Wennberg A, Ost Selker RG et al. Reproductive and hormonal fac-
A, Knave B. Occupational exposure to electro- tors and risk of brain tumors in adult females. Int
magnetic fields in relation to leukemia and brain J Cancer 2005;114:797–​805.
tumors:  a case-​control study in Sweden. Cancer Hildreth NG, Shore RE, Hempelmann LH, Rosenstein
Causes Control 1993;4:465–​76. M. Risk of extrathyroid tumors following radia-
Fox AJ, Lindars DC, Owen R. A survey of occupa- tion treatment in infancy for thymic enlargement.
tional cancer in the rubber and cablemaking Radiat Res 1985;102:378–​91.
industries:  results of five year analysis, 1967–​71. Hochberg F, Toniolo P, Cole P, Salcman M.
Br J Ind Med 1974;31:140–​51. Nonoccupational risk indicators of glioblastoma
Frei P, Poulsen AH, Johansen C, Olsen JH, Steding-​ in adults. J Neuro-​Oncol 1990;8:55–​60.
Jessen M, Schuz J. Use of mobile phones and risk Holick CN, Giovannucci EL, Rosner B, Stampfer MJ,
of brain tumours: update of Danish cohort study. Michaud DS. Prospective study of intake of fruit,
BMJ 2011;343:d6387. vegetables, and carotenoids and the risk of adult
Gilbert ES, Cragle DL, Wiggs LD. Updated analyses glio. ma. Am J Clin Nutr 2007;85:877–​86.
of combined mortality data for workers at the Holick CN, Smith SG, Giovannucci E, Michaud DS.
Hanford Site, Oak Ridge National Laboratory, Coffee, tea, caffeine intake, and risk of adult gli-
and Rocky Flats Weapons Plant. Radiat Res oma in three prospective cohort studies. Cancer
1993;136:408–​21. Epidemiol Biomarkers Prev 2010;19:39–​47.
Giles GG, McNeil JJ, Donnan G, Webley C, Staples MP, Howlader N, Noone AM, Krapcho M, Garshell J,
Ireland PD et  al. Dietary factors and the risk of Miller D, Altekruse SF et al. SEER Cancer Statistics
glioma in adults: results of a case-​control study in Review, 1975–​ 2012. Bethesda, MD:  National
Melbourne, Australia. Int J Cancer 1994;59:357–​62. Cancer Institute, 2015.
Gottleib MS, Carr JK. Case-​ control cancer mortal- Hu J, Johnson KC, Mao Y, Guo L, Zhao X, Jia X et al.
ity study and chlorination of drinking water in Risk factors for glioma in adults:  a case-​control
Louisiana. Environ Health Perspect 1982;46:169–​77. study in northeast China. Cancer Detect Prev
Group IS. Brain tumour risk in relation to mobile 1998;22:100–​8.
telephone use:  results of the INTERPHONE Hu J, La Vecchia C, Negri E, Chatenoud L, Bosetti C,
international case-​control study. Int J Epidemiol Jia X et al. Diet and brain cancer in adults: a case-​
2010;39:675–​94. control study in northeast China. Int J Cancer
Gu J, Liu Y, Kyritsis AP, Bondy ML. Molecular epidemi- 1999;81:20–​3.
ology of primary brain tumors. Neurotherapeutics Huang K, Whelan EA, Ruder AM, Ward EM, Deddens
2009;6:427–​35. JA, Davis-​ King KE et  al. Reproductive factors
Gurney JG, Smith MA, Olshan AF, et al. Clues to the and risk of glioma in women. Cancer Epidemiol
etiology of childhood brain cancer:  N-​ nitroso Biomarkers Prev 2004; 13:1583–​8.
compounds, polyomaviruses, and other factors of Inskip PD, Linet MS, Heineman EF. Etiology of brain
interest. Cancer Invest 2001;19:630–​40. tumors in adults. Epidemiol Rev 1995;17:382–​414.
Hagmar L, Akesson B, Nielsen J, Andersson C, Linden Inskip PD, Tarone RE, Hatch EE, Wilcosky TC, Shapiro
K, Attewell R et al. Mortality and cancer morbidity WR, Selker RG et  al. Cellular telephone use and
in workers exposed to low levels of vinyl chloride brain tumors. N Engl J Med 2001;344:79–​86.
monomer at a polyvinyl chloride processing plant. Inskip PD, Tarone RE, Hatch EE, et  al.
Am J Ind Med 1990;17:553–​65. Sociodemographic indicators and risk of brain
Hardell L, Carlberg M, Hansson Mild K. Use of mobile tumors. Int J Epidemiol 2003;32:225–​33.
phones and cordless phones is associated with Jhawar BS, Fuchs CS, Colditz GA, Stampfer MJ. Sex
increased risk for glioma and acoustic neuroma. steroid hormone exposures and risk for meningi-
Pathophysiology 2013a20:85–​110. oma. J Neurosurg 2003;99:848–​53.
Hardell L, Carlberg M, Soderqvist F, Mild KH. Case-​ Jiang J, Quan XF, Zhang L, Wang YC. The XRCC3
control study of the association between malig- Thr241Met polymorphism influences glioma
nant brain tumours diagnosed between 2007 and risk: a meta-​analysis. APJCP 2013;14:3169–​73.
2009 and mobile and cordless phone use. Int J Johansen C, Boice JD Jr, McLaughlin JK, Olsen JH.
Oncol 2013b;43:1833–​45. Cellular telephones and cancer—​ a nationwide
Harrington JM, McBride DI, Sorahan T, Paddle cohort study in Denmark. J Natl Cancer Inst
GM, van Tongeren M. Occupational exposure to 2001;93:203–​7.
603

Brain Cancer 603

Johnson KJ, Cullen J, Barnholtz-​Sloan JS, Ostrom QT, Little MP, de Vathaire F, Shamsaldin A, Oberlin
Langer CE, Turner MC et  al. Childhood brain O, Campbell S, Grimaud E et al. Risks of brain
tumor epidemiology: a brain tumor epidemiology tumour following treatment for cancer in
consortium review. Cancer Epidemiol Biomarkers childhood:  modification by genetic factors,
Prev 2014;23:2716–​36. radiotherapy and chemotherapy. Int J Cancer
Kabat GC, Park Y, Hollenbeck AR, Schatzkin A, 1998;78:269–​75.
Rohan TE. Reproductive factors and exogenous Little MP, Rajaraman P, Curtis RE, Devesa SS, Inskip
hormone use and risk of adult glioma in women PD, Check DP et al. Mobile phone use and glioma
in the NIH-​AARP Diet and Health Study. Int J risk: comparison of epidemiological study results
Cancer 2011;128:944–​50. with incidence trends in the United States. BMJ
Kaplan S, Etlin S, Novikov I, Modan B. Nutritional 2012;344:e1147.
factors in the etiology of brain tumors: potential Liu CC, Chen CC, Wu TN, Yang CY. Association of
role of nitrosamines, fat, and cholesterol. Am J brain cancer with residential exposure to pet-
Epidemiol 1997a;146:832–​41. rochemical air pollution in Taiwan. J Toxicol
Kaplan S, Etlin S, Novikov I, Modan B. Occupational Environ Health Part 2008;71: 310–​4.
risks for the development of brain tumors. Am J Longnecker MP, Rogan WJ, Lucier G. The human health
Ind Med 1997b;31:15–​20. effects of DDT (dichlorodiphenyl-​trichloroethane)
Kheifets LI, Afifi AA, Buffler PA, Zhang ZW. and PCBs (polychlorinated biphenyls) and an
Occupational electric and magnetic field expo- overview of organochlorines and public health.
sure and brain cancer:  a meta-​analysis. J Occup Annu Rev Public Health 1997;18:211–​44.
Environ Med 1995;37:1327–​41. Malerba S, Galeone C, Pelucchi C, Turati F, Hashibe
Kheifets LI, Gilbert ES, Sussman SS, Guénel P, Sahl M, La Vecchia C, Tavani A. A meta-​analysis of
JD, Savitz DA, Thériault G. Comparative analy- coffee and tea consumption and the risk of glioma
ses of the studies of magnetic fields and cancer in adults. Cancer Causes Control 2013;24:267–​76.
in electric utility workers:  studies from France, Mancuso TF, Ciocco A, El-​Attar AA. An epidemio-
Canada, and the United States. Occup Environ logical approach to the rubber industry:  a study
Med 1999;56:567–​74. based on departmental experience. J Occup Med
Khuder SA, Mutgi AB, Schaub EA. Meta-​analyses 1968;10:213–​32.
of brain cancer and farming. Am J Ind Med Marsh GM, Enterline PE, McCraw D. Mortality pat-
1998;34:252–​60. terns among petroleum refinery and chemical
Kitahara CM, Wang SS, Melin BS, Wang Z, Braganza plant workers. Am J Ind Med 1991;19:29–​42.
M, Inskip PD et  al. Association between adult McLaughlin JK, Malker HSR, Blot W, Malker BK,
height, genetic susceptibility and risk of glioma. Stone BJ, Weiner JA et al. Occupational risks for
Int J Epidemiol 2012;41:1075–​85. intracranial gliomas in Sweden. J Natl Cancer Inst
Lee M, Wrensch M, Miike R. Dietary and tobacco risk 1987;67:253–​7.
factors for adult onset glioma in the San Francisco Michaud DS, Gallo V, Schlehofer B, Tjonneland A,
Bay Area (California, USA). Cancer Causes Olsen A, Overvad K et  al. Reproductive factors
Control 1997;8:13–​24. and exogenous hormone use in relation to risk
Lewis R, Rempala G. A case-​cohort study of angio- of glioma and meningioma in a large European
sarcoma of the liver and brain cancer at a poly- cohort study. Cancer Epidemiol Biomarkers Prev
mer production plant. J Occup Environ Med 2010a;19:2562–​9.
2003;45:538–​45. Michaud DS, Gallo V, Schlehofer B, Tjonneland A,
Li C-​Y, Thériault G, Lin RS. Epidemiological appraisal Olsen A, Overvad K et al. Coffee and tea intake and
of studies of residential exposure to power fre- risk of brain tumors in the European Prospective
quency magnetic fields and adult cancers. Occup Investigation into Cancer and Nutrition cohort
Environ Med 1996;53:505–​10. study. Am J Clin Nutr 2010b;92:1145–​50.
Li Y, Millikan RC, Carozza S, Newman B, Liu E, Michaud DS, Holick CN, Batchelor TT, Giovannucci
Davis R et al. p53 mutations in malignant glioma. E, Hunter DJ. Prospective study of meat intake and
Cancer Epidemiol Biomarkers Prev 1998;7:303–​8. dietary nitrates, nitrites, and nitrosamines and risk
Lin RS, Dischinger PC, Conde J, Farrell KP. of adult glioma. Am J Clin Nutr 2009;90:570–​7.
Occupational exposure to electromagnetic fields Michaud DS, Siddiq A, Cox DG, Backes DM, Calboli FC,
and the occurrence of brain tumors: an analysis of Sughrue ME et al. Mannose-​binding lectin 2 gene
possible associations. J Occup Med 1985;276:413–​9. and risk of adult glioma. PloS One 2013;8:e61117.
Linos E, Raine T, Alonso A, Michaud D. Atopy and Modan B, Baidatz D, Mart H, Steinitz R, Levin SG.
risk of brain tumors:  a meta-​ analysis. J Natl Radiation induced head and neck tumors. Lancet
Cancer Inst 2007;99:1544–​50. 1974;1:277–​9.
604

604 Textbook of Cancer Epidemiology

Monson RR, Fine LJ. Cancer mortality and morbid- Preston-​ Martin S, Mack W. Gliomas and menin-
ity among rubber workers. J Natl Cancer Inst giomas in men in Los Angeles County:  inves-
1978;61:1047–​53. tigation of exposures to n-​nitroso compounds.
Moore SC, Rajaraman P, Dubrow R, Darefsky AS, In:  O’Neill IK et  al (eds.). Relevance to Human
Koebnick C, Hollenbeck A et  al. Height, body Cancer of N-​ Nitroso Compounds, Tobacco
mass index, and physical activity in relation to gli- Smoke and Mycotoxins. IARC Sci. Pub. No. 105.
oma risk. Cancer Res 2009;69:8349–​55. Lyon:  International Agency for Research on
Muscat JE, Malkin M, Thompson S, Shore RE, Cancer, 1991: 1977–​1203.
Stellman SD, McRee D et  al. Handheld cellular Preston-​Martin S, Mack W, Henderson BE. Risk fac-
telephone use and risk of brain cancer. JAMA tors for gliomas and meningiomas in males in Los
2000;284:3001–​7. Angeles County. Cancer Res 1989a;49:6137–​43.
Musicco M, Filipini G, Bordo BM, Melotto A, Morello Preston-​ Martin S, Paganini-​ Hill A, Henderson
G, Berrino F. Gliomas and occupational expo- BE, Pike MC, Wood C. Case-​ control study of
sure to carcinogens:  case-​ control study. Am J intracranial meningiomas in women in Los
Epidemiol 1982;116:782–​90. Angeles County, California. J Natl Cancer Inst
Musicco M, Sant M, Molinari S, Filippini G, Gatta 1980;65:67–​73.
G, Berrino F. A case-​control study of brain glio- Preston-​ Martin S, Pike MC, Ross RK, Jones PA,
mas and occupational exposure to chemical car- Henderson BE. Increased cell division as a cause
cinogens:  the risk to farmers. Am J Epidemiol of human cancer. Cancer Res 1990;50:7415–​21.
1988;128:778–​85. Preston-​ Martin S, Thomas DC, Wright WE,
Niedermaier T, Behrens G, Schmid D, Schlecht I, Henderson BE. Noise trauma in the aetiology of
Fischer B, Leitzmann MF. Body mass index, phys- acoustic neuromas in men in Los Angeles County,
ical activity, and risk of adult meningioma and gli- 1978–​1985. Br J Cancer 1989b;59:783–​6.
oma: a meta-​analysis. Neurology 2015;85:1342–​50. Preston-​ Martin S, Yu MC, Benton B, Henderson
Nygren C, Adami J, Ye W, Bellocco R, af Geijerstam BE. N-​ nitroso compounds and childhood
JL, Borg J et  al. Primary brain tumors follow- brain tumors:  a case-​control study. Cancer Res
ing traumatic brain injury—​a population-​based 1982;42:5240–​5.
cohort study in Sweden. Cancer Causes Control Preston-​Martin S, Yu MC, Henderson BE, Roberts
2001;12:733–​7. C. Risk factors for meningiomas in men in Los
Olin RG, Ahlbom A, Lindberg-​Navier I, Norell SE, Angeles County. J Natl Cancer Inst 1983:70:863–​6.
Spännare B. Occupational factors associated with Rajaraman P, Melin BS, Wang Z, McKean-​Cowdin R,
astrocytomas: a case-​control study. Am J Ind Med Michaud DS, Wang SS et al. Genome-​wide asso-
1987;11:615–​25. ciation study of glioma and meta-​analysis. Hum
Olsen JH, Boice JD Jr, Jensen JPA, Frumeni JF Jr. Cancer Genetics 2012;131:1877–​88.
among epileptic patients exposed to anticonvul- Reif JS, Pearce N, Fraser J. Occupational risks for brain
sant drugs. J Natl Cancer Inst 1989;81:803–​8. cancer:  a New Zealand Cancer Registry-​ Based
Ostrom QT, Gittleman H, Farah P, Ondracek A, Study. J Occup Med 1989;31:863–​7.
Chen Y, Wolinsky Y et  al. CBTRUS statistical Ries LA, Smith MA, Gurney JG et  al. Cancer
report: Primary brain and central nervous system Incidence and Survival among Children and
tumors diagnosed in the United States in 2006–​ Adolescents: United States SEER Program, 1975–​
2010. Neuro Oncol 2013;15 Suppl 2: ii1–​56. 1995. Bethesda, MD:  National Cancer Institute,
Pearce N, Reig, J, Fraser J. Case-​control studies of SEER Program.
cancer in New Zealand electrical workers. Int J Rodvall Y, Ahlbom A, Spännare B, Nise G. Glioma and
Epidemiol 1989;18:55–​9. occupational exposure in Sweden, a case-​control
Preston-​Martin S, Mack W, Henderson BE. Risk fac- study. Occup Environ Med 1996;53:526–​32.
tors for gliomas and meningiomas in males in Los Rollison DE, Helzlsouer KJ, Alberg AJ, Hoffman S,
Angeles County. Cancer Res 1989;49:6137–​43. Hou J, Daniel R et al. Serum antibodies to JC virus,
Preston DL, Ron E, Yonehara S, Kobuke T, Fujii H, BK virus, simian virus 40, and the risk of incident
Kishikawa M et  al. Tumors of the nervous sys- adult astrocytic brain tumors. Cancer Epidemiol
tem and pituitary gland associated with atomic Biomarkers Prev 2003;12:460–​3.
bomb radiation exposure. J Natl Cancer Inst Ross JA, Spector LG. Cancer in children. In D
2002;94:1555–​63. Schottenfeld, JF FraumeniJr (eds.) Cancer
Preston-​ Martin S. Descriptive epidemiology of Epidemiology and Prevention 3 ed. New  York,
primary tumors of the brain, cranial nerves NY: Oxford University Press, 2006: 1251–​68.
and cranial meninges in Los Angeles County. Ruder AM, Hein MJ, Hopf NB, Waters MA. Mortality
Neuroepidemiology 1989;8:283–​95. among 24,865 workers exposed to polychlorinated
605

Brain Cancer 605

biphenyls (PCBs) in three electrical capacitor Schwartzbaum J, Ahlbom A, Malmer B, Lonn S,

manufacturing plants: a ten-​year update. Int J Hyg Brookes AJ, Doss H et al. Polymorphisms associ-
Environ Health 2014;217:176–​87. ated with asthma are inversely related to glioblas-
Ryan P, Lee MS, North JB, McMichael AJ. Risk fac- toma multiforme. Cancer Res 2005;65:6459–​65.
tors for tumors of the brain and meninges: results Schwartzbaum J, Ding B, Johannesen TB, Osnes LT,
from The Adelaide Adult Brain Tumor Study. Int J Karavodin L, Ahlbom A et al. Association between
Cancer 1992a;51:20–​7. prediagnostic IgE levels and risk of glioma. J Natl
Ryan P, Lee MW, North B, McMichael AJ. Amalgam Cancer Inst 2012;104:1251–​9.
fillings, diagnostic dental x-​rays and tumors of the SEER. SEER*Stat Database: Incidence—​SEER 9 Regs
brain and meninges. Eur J Cancer 1992b;28B:91–​5. Research Data. National Cancer Institute. 2015.
Sahl JD, Kelsh MA, Greenland S. Cohort and nested www.seer.cancer.gov.
case-​control studies of hematopoietic cancers Shete S, Hosking FJ, Robertson LB, Dobbins SE,
and brain cancer among electric utility workers. Sanson M, Malmer B et  al. Genome-​wide asso-
Epidemiology 1993;4:104–​14. ciation study identifies five susceptibility loci for
Samanic CM, De Roos AJ, Stewart PA, Rajaraman P, glioma. Nat Genetics 2009;41:899–​904.
Waters MA, Inskip PD. Occupational exposure Shirts SB, Annegers JF, Houser WA, Kurland LT.
to pesticides and risk of adult brain tumors. Am J Cancer incidence in a cohort of patients with
Epidemiol 2008;67:976–​85. severe seizure disorders. J Natl Cancer Inst
Samuelsen SO, Bakketeig LS, Tretli S, Johannesen TB, 1986;77:83–​7.
Magnus P. Head circumference at birth and risk Shore RE, Albert RE, Pasternack BS. Follow-​up study
of brain cancer in childhood: a population-​based of patients treated by x-​ray epilation for tinea capi-
study. Lancet Oncol 2006;7:39–​42. tis. Arch Environ Health 1976;31:17–​24.
Sanson M, Hosking FJ, Shete S, Zelenika D, Dobbins Silvera SA, Miller AB, Rohan TE. Hormonal and
SE, Ma Y et al. Chromosome 7p11.2 (EGFR) var- reproductive factors and risk of glioma: a prospec-
iation influences glioma risk. Hum Mol Genetics tive cohort study. Int J Cancer 2006;118:1321–​4.
2011;20:2897–​904. Simonato L, L-​Abbe KA, Anderson A, Belli S, Comba
Sathiakumar N, Delzell E, Rodu B, Beall C, Myers S. P, Engholm G. A collaborative study of cancer inci-
Cancer incidence among employees at a petro- dence and mortality among vinyl chloride work-
chemical research facility. J Occup Environ Med ers. Scand J Work Environ Health 1991;17:156–​69.
2001;43:166–​74. Sjostrom S, Hjalmars U, Juto P, Wadell G, Hallmans
Savitz DA, Loomis DP. Magnetic field exposure in G, Tjonneland A et  al. Human immunoglobu-
relation to leukemia and brain cancer mortality lin G levels of viruses and associated glioma risk.
among electric utility workers. Am J Epidemiol Cancer Causes Control 2011;22:1259–​66.
1995;141:123–​34. Sorahan T, Parkes HG, Veys CA. Mortality in the
Schlehofer B, Blettner M, Becker N, Martinsohn C, British rubber industry, 1946–​85. Br J Ind Med
Wahrendorf J. Medical risk factors and the devel- 1989;314:1010–​5.
opment of brain tumors. Cancer 1992;69:2541–​7. Sorahan T. Cancer incidence in UK electricity genera-
Schlehofer B, Blettner M, Preston-​Martin S, Niehoff tion and transmission workers, 1973–​2008. Occup
D, Wahrendorf J, Arslan A et  al. Role of medi- Med 2012;62:496–​505.
cal history in brain tumour development: results Speers MA, Dobbins JG, Miller VS. Occupational
from the international adult brain tumour study. exposure and brain cancer mortality:  a prelimi-
Int J Cancer 1999;82:155–​60. nary study of East Texas residents. Am J Ind Med
Schlehofer B, Siegmund B, Linseisen J, Schuz J, 1988;13:629–​38.
Rohrmann S, Becker S et al. Primary brain tumours Straif K, Weiland SK, Bungers M, Holthenrich D,
and specific serum immunoglobulin E:  a case-​ Taeger D, Yi S, Keil U. Exposure to high concentra-
control study nested in the European Prospective tions of nitrosamines and cancer mortality among
Investigation into Cancer and Nutrition cohort. a cohort of rubber workers. Occup Environ Med
Allergy 2011;66:1434–​41. 2000;57:180–​7.
Schneider AB, Shore-​Freedman E, Ryo UY, Bekerman Teta MJ, Ott MG, Schnatter AR. An update of mor-
C, Favus M, Pinsky S. Radiation-​induced tumors tality due to brain neoplasms and other causes
of the head and neck following childhood irradia- among employees of a petrochemical facility. J
tion. Medicine 1985;64:1–​15. Occup Med 1991;33:45–​51.
Schoemaker MJ, Robertson L, Wigertz A, Jones ME, Thériault G, Goldberg M, Miller AB, Armstrong B,
Hosking FJ, Feychting M et al. Interaction between Guénel P, Deadman J et al. Cancer risks associated
5 genetic variants and allergy in glioma risk. Am J with occupational exposure to magnetic fields
Epidemiol 2010;171:1165–​73. among electric utility workers in Ontario and
60

606 Textbook of Cancer Epidemiology

Québec, Canada, and France:  1970–​1989. Am J polymorphisms influence glioma status and
Epidemiol 1994;139:550–​72. serum IgE levels. Cancer Epidemiol Biomarkers
Thériault G, Goulet L. A mortality study of oil refinery Prev 2007;16:1229–​35.
workers. J Occup Med 1979;21:367–​70. Wiemels JL, Wiencke JK, Patoka J, Moghadassi M,
Thomas TL, DeCoufle P, Moure-​Eraso R. Mortality Chew T, McMillan A et  al. Reduced immu-
among workers employed in petroleum refin- noglobulin E and allergy among adults with
ing and petrochemical plants. J Occup Med glioma compared with controls. Cancer Res
1980;22:97–​103. 2004;64:8468–​73.
Thomas TL, Waxweiler RJ. Brain tumors and occupa- Wiemels JL, Wiencke JK, Sison JD, Miike R, McMillan
tional risk factors: a review. Scand J Work Environ A, Wrensch M. History of allergies among
Health 1986;12:1–​15. adults with glioma and controls. Int J Cancer
Thomas TL, Waxweiler RJ, Moure-​Eraso R, Itaya S, 2002;98:609–​15.
Fraumeni JF Jr. Mortality patterns among work- Wiemels JL, Wilson D, Patil C, Patoka J, McCoy L,
ers in three Texas oil refineries. J Occup Med Rice T et al. IgE, allergy, and risk of glioma: update
1982;24:135–​41. from the San Francisco Bay Area Adult Glioma
Tsai SP, Gilstrap EL, Cowles SR, Snyder PJ, Ross Study in the temozolomide era. Int J Cancer
CE. A cohort mortality study of two California 2009;125:680–​7.
refinery and petrochemical plants. J Occup Med Wigertz A, Lonn S, Mathiesen T, Ahlbom A, Hall P,
1993;35:415–​21. Feychting M. Risk of brain tumors associated with
Tsang RW, Laperriere NJ, Simpson WJ, Brierley J, exposure to exogenous female sex hormones. Am
Panzarella T, Smyth HS. Glioma arising after J Epidemiol 2006;164:629–​36.
radiation therapy for pituitary adenoma: a report Wilczynska U, Szadkowska-​ Stanczyk I, Szeszenia-​
of four patients and estimation of risk. Cancer Dabrowska N, Sobala W, Strzelecka A. Cancer
1993;72:2227–​33. mortality in rubber tire workers in Poland. Int J
U.S. Department of Health and Human Services. The Occup Med Environ Health 2001;14:115–​25.
Health Consequences of Smoking: A Report of the Wilkinson GS, Tietjen GL, Wiggs LD, Galke WA,
Surgeon General. Atlanta, GA:  U.S. Department Acquavella JF, Reyes M et  al. Mortality among
of Health and Human Services, Centers for plutonium and other radiation workers at a
Disease Control and Prevention, National Center plutonium weapons facility. Am J Epidemiol
for Chronic Disease Prevention and Health 1987;125:231–​50.
Promotion, Office on Smoking and Health, 2004. Wong O, Morgan RW, Bailey WJ, Swencicki RE,
Villeneuve PJ, Agnew DA, Johnson KC, Mao Y. Claxton K, Kheifets L. An epidemiologic study
Brain cancer and occupational exposure to mag- of petroleum refinery employees. Br J Ind Med
netic fields among men: results from a Canadian 1986;43:6–​17.
population-​ based case-​ control study. Int J Wong O, Raabe GK. A critical review of cancer epide-
Epidemiol 2002;31:210–​7. miology in studies of petroleum industry employ-
Ward MH, Heineman EF, McComb RD, Weisenburger ees, with a quantitative meta-​analysis by cancer
DD. Drinking water and dietary sources of nitrate site. Am J Ind Med 1989;15:283–​310.
and nitrite and risk of glioma. J Occup Environ Wong O, Whorton MD, Foliart DE, Ragland D. An
Med 2005;47:1260–​7. industry-​wide epidemiologic study of vinyl chlor-
Waxweiler RJ, Alexander V, Leffingwell SS, Haring M, ide workers. Am J Ind Med 1991;20:317–​34.
Lloyd JW. Mortality from brain tumor and other Visser O, Ardanaz E, Botta L, Sant M, Tavilla A,
causes in a cohort of petrochemical workers. J Minicozzi P, Group E-​W. Survival of adults with
Natl Cancer Inst 1983;70:75–​81. primary malignant brain tumours in Europe;
Wen CP, Tsai SP, McClellan WA, Gibson RL. Long-​ Results of the EUROCARE-​5 study. Eur J Cancer
term mortality study of oil refinery workers. 2015; S0959-​8049(15)00712-​1. doi:  10.1016/​
I. Mortality of hourly and salaried workers. Am J j.ejca.2015.07.032. [Epub ahead of print]
Epidemiol 1983;118:526–​41. Wrensch MR, Barger GR. Familial factors associ-
Wertheimer N, Leeper E. Adult cancer related to ated with malignant gliomas. Genet Epidemiol
electrical wires near the home. Int J Epidemiol 1990;7:291–​301.
1982;11:345–​55. Wrensch M, Bondy ML, Wiencke J, Yost M.
White SJ, McClean AEM, Howland C. Anticonvulsant Environmental risk factors for primary malig-
drugs and cancer: a cohort study in patients with nant brain tumors:  a review. J Neuro-​ Oncol
severe epilepsy. Lancet 1979;2:458–​60. 1993;17:47–​64.
Wiemels JL, Wiencke JK, Kelsey KT, Moghadassi Wrensch M, Jenkins RB, Chang JS, Yeh RF, Xiao Y,
M, Rice T, Urayama KY et  al. Allergy-​ related Decker PA et  al. Variants in the CDKN2B and
607

Brain Cancer 607

RTEL1 regions are associated with high-​grade gli- Wrensch M, Weinberg A, Wiencke J, Miike R, Sison J,
oma susceptibility. Nat Genetics 2009;41:905–​8. Wiemels J et al. History of chickenpox and shin-
Wrensch M, Lee M, Miike R, Newman B, Barger G, gles and prevalence of antibodies to varicella-​
Davis R et al. Familial and personal medical his- zoster virus and three other herpesviruses among
tory of cancer and nervous system conditions adults with glioma and controls. Am J Epidemiol
among adults with glioma and controls. Am J 2005;161:929–​38.
Epidemiol 1997a;145:581–​93. Yiin JH, Ruder AM, Stewart PA, Waters MA, Carreon
Wrensch M, Weinberg A, Wiencke J, Masters H, Miike T, Butler MA et al. And Brain Cancer Collaborative
R, Barger G, Lee M. Does prior infection with Study G. The Upper Midwest Health Study: a case-​
varicella-​zoster virus influence risk of adult gli- control study of pesticide applicators and risk of
oma? Am J Epidemiol 1997;145:594–​7. glioma. Environ Health 2012;11:39.
Wrensch M, Weinberg A, Wiencke J, Miike R, Barger Young TB, Kanarek MS, Tsiatis AA. Epidemiologic
G, Kelsey K. Prevalence of antibodies to four her- study of drinking water chlorination and
pesviruses among adults with glioma and con- Wisconsin female cancer mortality. J Natl Cancer
trols. Am J Epidemiol 2001;154:161–​5. Inst 1981;67:1191–​8.
608
609

25
Thyroid Cancer
C A R L O L A V E C C H I A , C R I S T I N A B O S E T T I , A N D H A N S -​O L O V   A D A M I

T hyroid cancer was brought to the attention of

the epidemiologists and of the general pub-
lic as well after the Chernobyl nuclear power
antibodies, and an ultrasound scan. Because it is
difficult to distinguish a malignant thyroid nod-
ule from a benign one, a needle biopsy is nec-
plant accident in 1986, when a strong increase in essary to confirm the diagnosis. Sometimes, a
thyroid cancer incidence was observed in indi- histopathology examination following surgical
viduals exposed in childhood and adolescence removal of thyroid tissue is also necessary.
to radioactive iodine in the highly contaminated Treatment Usually, the first treatment for most
areas (Cardis et al, 2006). Apart from this, thyroid thyroid cancers is surgery, including partial or
cancer has received relatively little attention by total thyroidectomy, with or without the removal
the epidemiologic community, since this is a rare of regional enlarged lymph nodes. Depending on
malignancy including several distinct histopath- the type and stage of thyroid cancer, radioactive
ological entities. The interest in thyroid cancer iodine treatment and/​or external radiotherapy
has, however, grown over the last few decade(s), may also be used after thyroid surgery. Thyroid
because a steady increase in the incidence has hormone therapy is also prescribed to treat the
been observed in young women from many cancer or after surgery. Chemotherapy is some-
countries worldwide (La Vecchia et al, 2015). times used to treat advanced or recurrent thyroid
cancers.
CLINICAL SYNOPSIS
Prognosis Survival from thyroid cancer—​
Subgroups Thyroid cancer arises from the epi-
particularly for papillary carcinoma in the
thelial elements of the gland, generally from fol-
young—is high (over 90% 5-​year survival rate
licular cells. It is traditionally separated into the
in Europe or North America), resulting in rela-
two major groups of differentiated (including
tively low mortality rates. Prognosis is lower for
papillary, follicular, and medullary) and undif-
anaplastic and poorly differentiated carcinomas,
ferentiated (anaplastic) carcinoma, based on
particularly in the elderly.
morphology and clinical features. Papillary carci-
nomas represent approximately two-​thirds of all Progress The treatment for differentiated thyroid
thyroid cancers in both sexes; the corresponding cancer did not change appreciably over the last
figures are 10%–​20% for follicular carcinomas, decades, although there are ongoing randomized
5%–​10% for medullary carcinomas, and less than trials searching for the optimal surgery for this
5% for anaplastic carcinomas, other subtypes neoplasm. Also for anaplastic cancers, which are
accounting for less than 1% of all thyroid cancers. more difficult to treat, no progress was evident
over the last decades.
Symptoms Most differentiated thyroid cancers pres-
ent as asymptomatic thyroid nodules. However, DESCRIPTIVE
thyroid nodules are quite frequent particularly in EPIDEMIOLOGY
women, and only about 1 in 20 thyroid nodules Thyroid cancer includes several histopatholog-
are cancers. Hoarseness or soreness for more than ical subtypes, with different etiologies and geo-
3 weeks may also be the first symptoms of thyroid graphic distribution, which makes difficult the
cancer, although these symptoms are unspecific and descriptive epidemiology of this neoplasm. The
may be due to bacterial or viral infections. distribution of thyroid cancer by histological
Diagnosis The diagnosis of thyroid cancer starts subtypes, as reported by epidemiological surveil-
from the palpation of the neck, a blood test to lance in selected countries worldwide, is shown
check the thyroid hormone levels and thyroid in Table 25-​1 (Curado et al, 2007).
 610
TABLE 25-​1   AGE-STANDARDIZED INCIDENCE RATES (ASR, WORLD STANDARD POPULATION) OF THYROID CANCER PER
100,000 AND PERCENT DISTRIBUTION BY SEX AND HISTOLOGICAL SUBT YPE IN SELECTED AREAS COVERED BY A CANCER
REGISTRY, 1998–2002

Country Women Men

ASR Histological subtype (%) ASR Histological subtype (%)

Papillary Follicular Medullary Anaplastic Other/​ Papillary Follicular Medullary Anaplastic Other/​
unknown unknown

Belarus 12.2 86 7 2 1 4 3.3 85 6 3 1 6

Korea 12.2 89 7 1 1 2 2.3 81 9 3 2 5
Israel 11.7 87 7 2 2 2 3.4 78 9 7 3 3
Philippines, 11.1 76 17 1 3 3 2.9 73 18 2 3 4
Manila
USA, SEER 9.1 85 11 2 1 2 3.0 79 13 4 2 3
Italy, North East 8.9 71 13 2 3 11 2.8 59 15 5 6 15
Costa Rica 8.2 89 6 1 1 3 1.7 82 7 4 5 2
Japan, Nagasaki 7.4 82 14 0 1 3 2.0 74 15 3 4 5
New Zealand 5.1 70 21 4 3 3 1.6 67 20 7 2 5
Sweden 3.3 67 16 5 9 4 1.3 57 18 10 9 6
Brazil, Sao Paulo 2.6 66 15 1 1 16 3.6 62 15 4 2 18
Egypt, Gharbiah 2.6 63 15 1 12 8 1.1 56 17 6 15 6
India, New Delhi 2.6 44 18 7 3 28 1.0 41 15 14 5 25
61

Thyroid Cancer 611

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 5 10 15 20 25
ASR (world) per 100,000 person-years

FIGURE 25-​1  Age-​standardized (to the 2012 world population) incidence rates (ASR) of thyroid cancer among women.
Source: Ferlay et al, 2013.

Thyroid cancer is rare overall, although it is Incidence rates are more than twofold higher in
one of the most common neoplasms in child- high-​income countries as compared to low-​and
ren and adults below age 40 years. An estimated middle-​income ones both in women (5.5/​100,000
230,000 new cases of thyroid cancer occurred and 2.6/​100,000, respectively) and in men (2.2/​
in 2012 among women and 70,000 among men, 100,000 and 1.0/​100,000, respectively). High-​risk
with an age-​standardized (world standard pop- areas (incidence over 10/​100,000 women) include
ulation) of about 6/​ 100,000 women and 1.9/​ selected countries of South and North America,
100,000 men (Ferlay et al, 2013). In most areas of Italy in Europe, Japan, and the Pacific islands.
the world, the incidence is in the range of 2–​5/​ International comparisons in thyroid cancer
100,000 among women and 1–​2/​100,000 in men. incidence, however, are hampered by differences
There is a substantial variation in thyroid can- in diagnosis and ascertainment of the disease.
cer incidence across different parts of the world In most countries worldwide, incidence rates
both in women (Fig. 25-​1) and in men (Fig. 25-​2). have been increasing over the last few decades

China

Japan

Poland

Spain

Sweden

The Netherlands

United Kingdom

United States of America

0 1 2 3 4 5 6 7
ASR (world) per 100,000 person-years

FIGURE 25-​2  Age-​standardized (to the 2012 world population) incidence rates (ASR) of thyroid cancer among men.
Source: Ferlay et al, 2013.
612

612 Textbook of Cancer Epidemiology

45

40
New cases per 100,000 person-years
35

30

25

20

15

10

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age
China Japan Poland Spain Sweden
TheNetherlands United Kingdom United States of America

FIGURE 25-​3  Age-​specific incidence rates of thyroid cancer among women.

Source: Ferlay et al, 2013.

(La Vecchia et al, 2015). Between 1975 and 2009, Global estimated number of deaths from thy-
thyroid cancer incidence has increased three- roid cancer were 27,000 in women and 13,000 in
fold in the United States, leading to an apparent men in 2012, with corresponding mortality rates
major improvement in survival (Ho et  al, 2015). of approximately 0.6/​ 100,000 women and 0.3/​
Such increases, however, are most likely due to 100,000 men (Ferlay et  al, 2013). Mortality rates
improved ascertainment, diagnosis, and certifica- have steadily declined over the last decades in most
tion (Baker & Bhatti, 2006; Davies & Welch, 2006; areas of the world, likely because of improved diag-
Vaccarella et  al, 2015). The introduction of neck nosis, management, and treatment (La Vecchia
ultrasonography in the 1980s and then ultrasound-​ et al, 2015) (Fig. 25-​5). Earlier declines (up to the
guided biopsy enabled indeed the detection of late 1980s/​1990s) and subsequent leveling off in
small (<2  mm) indolent cancers. Moreover, fur- mortality rates were observed in women from the
ther increases in thyroid cancer diagnoses have United Kingdom, United States, and Australia and
resulted from the spread of portable ultrasound in men from the United States.
machines and new imaging technologies. In
the US Veteran Affairs Health Care System, for GENETIC AND MOLECULAR
instance, a fourfold increase in thyroid ultrasound EPIDEMIOLOGY
and fine-​needle aspiration between 2001 and 2012
was followed by a doubling of the recorded thyroid Inherited Susceptibility
cancer incidence (Zevallos et al, 2015). There is a strong genetic component for medul-
Thyroid cancer incidence in women increased lary thyroid carcinoma, since about 20% of them
with age particularly up to 40–​ 44  years, and are associated with an autosomal dominant gene
tended to level off or decline thereafter in the with penetrance close to 100%, whereas familial
United States as in selected countries of Europe and genetic factors account for 5%–​15% of non-
and in China, but not in Japan (Fig. 25-​3). Steady medullary (i.e., papillary or follicular) thyroid
increases were seen throughout the life span in cancers (De Lellis et al, 2004; Nose, 2011). Germ-​
men, with a tendency to level off after age 40–​ line mutations in the RET gene are responsible
44 only in Poland, Spain, and China (Fig. 25-​4). for the hereditary tumor syndrome (i.e., multi-
Such different trends with age in various coun- ple endocrine neoplasia type 2, MEN 2) which
tries worldwide can again be interpreted in terms includes three subgroups, MEN 2A, MEN 2B,
of different screening and diagnostic patterns. and familial medullary thyroid carcinoma. The
613

Thyroid Cancer 613

25

New cases per 100,000 person-years

20

15

10

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 25-​4  Age-​specific incidence rates of thyroid cancer among men.

Source: Ferlay et al, 2013.

common malignancies for all three subtypes is stronger for thyroid cancer than for most other
of MEN 2 are medullary thyroid cancer (98%-​ epithelial neoplasms.
100%) and pheochromocytoma (50%).
Other rare familial syndromes associated Somatic Events
with an excess thyroid cancer risk are familial Genetic mutations have been described as
adenomatous polyposis (FAP), Cowden syn- strongly associated with nonmedullary thyroid
drome, Werner’s syndrome, Carney’s complex, cancer (Kondo et  al, 2006). Mutations in the
thyroid cancer developing in up to 12% of indi- BRAF proto-​ oncogene are the most common
viduals with FAP, in at least 10% of those with genetic alterations in thyroid cancer. They were
the Cowden syndrome, and in even 60% of found in one to two out of three papillary can-
those with the Carney’s complex (Nose, 2011; cers, about 15% in anaplastic, but not in follicu-
Steinhagen et al, 2012; Lauper et al, 2013). lar cancers (Kondo et al, 2006). BRAF mutation
For the commonest types of thyroid can- has not been associated with radiation exposure
cer (i.e., nonmedullary forms) the association and is not found in benign thyroid tumors. This
with genetic susceptibility is weaker. In a study mutation is mutually exclusive with other com-
including all thyroid cancers diagnosed between mon genetic alterations, such as RET mutations.
1958 and 2002 in Sweden, the familial relative RET mutations are present in about one-​third of
risk (RR) for papillary carcinoma were 3 and 6 cases, most commonly in pediatric papillary car-
when thyroid cancers were diagnosed in a parent cinomas linked to radiation exposure. A  causal
or a sibling, respectively (Hemminki et al, 2005). link between certain RET alterations and ioniz-
The latest update of that dataset, which contains ing radiation is supported by the high prevalence
information up to 2010, showed that thyroid can- of RET rearrangements in childhood papillary
cer risk is 6.6 times higher in individuals with a carcinomas following the Chernobyl accident
first-​degree relative (i.e., parent, sibling, child) (Kondo et al, 2006).
with the same disease, compared with the gen- The RAS proto-​ oncogene has also been
eral population (Frank et  al, 2014). Moreover, extensively studied in relation to thyroid car-
familial thyroid cancer risk is higher in individu- cinogenesis. A  mutation in the RAS gene
als with multiple first-​degree relatives affected, probably represents an early event in the carcin-
in those with relative(s) diagnosed at a younger ogenic process, leading to alterations in its nor-
age, or those with an affected twin (Fallah et al, mal gatekeeper function (Kondo et  al, 2006).
2013). Thus, the association with family history Mutations in the RAS gene are found at similar
614

614 Textbook of Cancer Epidemiology

France Germany Italy

Death rate per 100,000

Death rate per 100,000

Death rate per 100,000

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

1970 1990 2010 1970 1990 2010 1970 1990 2010

Calendar year Calendar year Calendar year

Netherlands Spain United Kingdom

Death rate per 100,000

Death rate per 100,000

Death rate per 100,000

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

1970 1990 2010 1970 1990 2010 1970 1990 2010

Calendar year Calendar year Calendar year

European Union United States of America Hong Kong

Death rate per 100,000

Death rate per 100,000

Death rate per 100,000

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

1970 1990 2010 1970 1990 2010 1970 1990 2010

Calendar year Calendar year Calendar year

Israel Japan Australia

Death rate per 100,000

Death rate per 100,000

Death rate per 100,000
1.5

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0
1.0
0.5
0.0

1970 1990 2010 1970 1990 2010 1970 1990 2010

Calendar year Calendar year Calendar year

FIGURE 25-​5 Three-​year moving average trends in age-​standardized (world standard population) death rates from
thyroid cancer in selected countries worldwide, 1970–​2012. Men (full line); women (dotted dashed line).
Source: WHO (World Health Organization Statistical Information System).

frequencies in thyroid carcinomas and adeno- undifferentiated (anaplastic) thyroid carcinomas

mas. Transgenic mice with RAS mutations (De Lellis et al, 2004).
show increased follicular cell proliferation and
reduced expression of differentiation markers RISK FACTORS
such as thyroglobulin.
Inactivation of the p53 tumor suppressor Tobacco
gene is rare in patients with differentiated thy- In a collaborative reanalysis of case-​control stud-
roid carcinomas, but common in those with ies conducted worldwide before 1998 including
615

Thyroid Cancer 615

2,725 thyroid cancer cases and 4,776 controls, hormones and indirectly through a consequent
the overall RR of thyroid cancer was 0.6 (95% rise in TSH secretion (Feldt-​Rasmussen, 2001;
CI 0.6–​0.7) for current and 0.9 (95% CI 0.8–​1.1) Dal Maso et al, 2009). Chronic iodine deficiency
for former smokers, with significant trends of is an established risk factor for goiter and follic-
reduced risk with greater frequency and dura- ular thyroid cancer (Franceschi, 1998; Nagataki
tion of smoking (Mack et al, 2003). A meta-​anal- & Nystrom, 2002). However, some ecological
ysis updated to 2013, including 25 case-​control studies suggested that iodine supplementation
studies (with 6,260 thyroid cancer cases) and and iodine excess could increase the incidence of
six cohort studies (with 2,715 thyroid cancers) papillary thyroid cancer (Weissel, 2003).
reported a RR of 0.79 (95% CI 0.70–​0.88) in With reference to iodine intake from diet,
ever smokers as compared with never smokers a study conducted in Hawaii—​ a high iodine
(Cho & Kim, 2014). Similarly, a pooled analy- intake area—​including 191 thyroid cancer cases,
sis of five prospective studies from the United showed a nonsignificant increased RR (1.6) in
States, including a total of 1,003 incident thyroid women with the highest dietary iodine consump-
cancer cases, found a hazard ratio (HR) of 0.68 tion and only a modest increase (1.3) in men
(95% CI 0.55–​0.85) in current as compared to (Kolonel et  al, 1990). Conversely, a study con-
never smokers (Kitahara et al, 2012a). Number of ducted in California, including 608 women with
cigarettes smoked per day among current smok- papillary thyroid cancer, showed that high iodine
ers was also inversely related to medullary can- dietary intakes were associated with an approx-
cer in the pooled analysis of case-​control studies imately halved risk of papillary thyroid cancer
including 67 medullary cancers and 67 matched (Horn-​Ross et al, 2001). However, when toenail
controls (Negri et al, 2002). clipping was used as a biomarker of iodine expo-
Thus, tobacco smoking does not increase the sure, no association with thyroid cancer risk was
risk of thyroid cancer, but—​if anything—​is asso- observed, though correlations between dietary
ciated to a reduced risk. However, the inverse iodine and nail levels of iodine were not optimal.
association between tobacco smoking and thy- Iodine imbalance (i.e., iodine deficiency but
roid cancer does not have a clear biological also iodine excess) is thus a well-​established risk
explanation. Possible mechanisms are related to factor for thyroid cancer, with RR between 1.5
a potential—​ though not clearly established—​ and 2. However, over the last century, the prev-
effect on thyroid hormone levels. Current smok- alence of iodine deficiency has substantially
ers, as compared to never or former smokers, decreased in most countries, particularly in low-​
have indeed lower levels of thyroid stimulating and middle-​ income ones, and adverse effects
hormone (TSH), which may play a role in thy- were rare. Thus, the benefits of correcting iodine
roid cancer development (Williams, 1990; Soldin deficiency outweigh by far any possible risk of
et al, 2009). Cigarette smoking could also poten- iodine supplementation.
tially influence thyroid cancer risk by altering sex
steroid hormone levels (Baron et al, 1990). Fish
A collaborative reanalysis of all case-​ control
Diet studies of thyroid cancer conducted before 1998
A few epidemiological studies provided informa- assessed the association with salt-​water fish and
tion on the potential role of dietary factors in shellfish—​the major natural sources of iodine
thyroid cancer etiology, with inconsistent results in diet—​and reported that high fish consump-
(World Cancer Research Fund and American tion did not increase thyroid cancer risk (RR 0.9,
Institute for Cancer Research, 2007; Dal Maso et 95% CI 0.7–​1.1) (Bosetti et al, 2001) (Table 25-​2).
al, 2009; Liu & Lin, 2014; Cho & Kim, 2015). A reduced risk following high levels of fish con-
sumption was observed in endemic goiter areas
Iodine (RR 0.7, 95% CI 0.5–​0.9), with no association
Iodine intake has long been associated with thy- in iodine-​rich areas (RR 1.1, 95% CI 0.9–​1.5).
roid cancer risk. Indeed, iodine intake is essen- A  subsequent meta-​analysis also suggested that
tial for thyroid functioning, and thyroid cancer high intake of fish may be inversely associated to
is more frequent in areas with iodine deficiency thyroid cancer risk in iodine-​deficient areas but
(Feldt-​Rasmussen, 2001; Dal Maso et  al, 2009). not in iodine-​rich ones (Liu & Lin, 2014).
Iodine deficiency influences thyroid function Overall, elevated fish consumption does not
directly through a reduction in the level of thyroid seem to appreciably increase thyroid cancer risk,
61

616 Textbook of Cancer Epidemiology

TABLE 25-​2   RELATIVE RISKS (RR) a AND 95% CONFIDENCE

INTERVALS (CI) OF THYROID CANCER ACCORDING TO
SELECTED DIETARY FACTORS

N. of studies Categorization RR 95% CI

Fish
Bosetti et al, 2001 13 High versus Low 0.9 0.7–​1.1

Shellfish
Bosetti et al, 2001 13 High versus Low 1.0 0.9–​1.2
Liu & Lin, 2014 9 High versus Low 0.88 0.72–​1.08

Cruciferous vegetables
Bosetti et al, 2002 11 High versus Low 0.9 0.8–​1.1
Liu & Lin, 2014 8 High versus Low 0.93 0.66–​1.29

Other vegetables
Bosetti et al, 2002 11 High versus Low 0.8 0.7–​1.0
Liu & Lin, 2014 5 High versus Low 0.76 0.58–​1.00
a
Adjusted at least for age and sex.

and may have a favorable influence in areas where vegetables are not positively related to thyroid
iodine deficiency is, or was, common. cancer risk, and their effect does not seem to be
substantially different from that of other vegeta-
Vegetables bles, which appear to be moderately protective
Some other aspects of diet have been related to on this neoplasm.
thyroid cancer risk including, in particular, con-
sumption of cruciferous vegetables which con- Other
tain thioglycosides that may be degraded to form Published results on other foods, nutrients, or
goitrogens and induce thyroid tumors in rodents vitamins are limited and inconsistent, and con-
(Kanno et  al, 1990)  (Table 25-​2). In the collabo- sequently no conclusion can be drawn (Dal Maso
rative reanalysis of case-​control studies on thy- et al, 2009; Liu & Lin, 2014). Moreover, no con-
roid cancer conducted before 1998, high intakes sistent association was observed with coffee or
of cruciferous vegetables were not significantly tea drinkers in the collaborative reanalysis of
associated with thyroid cancer risk (RR 0.9, 95% case-​control studies (Mack et al, 2003).
CI 0.8–​1.1) (Bosetti et al, 2002). The results were
similar in studies conducted in iodine-​rich and Alcohol
in endemic goiter areas. Similar findings were The collaborative reanalysis of case-​control stud-
reported in a subsequent meta-​analysis (Liu & ies published before 1998 found an inverse relation
Lin, 2014). between alcohol drinking and thyroid cancer risk,
In the collaborative reanalysis of case-​control with an RR of 0.8 (95% CI 0.7–​0.9, for drinkers ver-
studies, an inverse association between vegeta- sus nondrinkers) (Mack et al, 2003). The RR, how-
bles other than cruciferous was observed, with an ever, become nonsignificant after adjustment for
overall RR of 0.8 (95% CI 0.7–​1.0) for the highest smoking. In a meta-​analysis of nine studies, includ-
level of consumption as compared to the lowest ing 2,503 drinker cases and 1,582 nondrinker
one (Bosetti et al, 2002) (Table 25-​2). This find- cases, the overall RR was 0.81 (95% CI 0.74–​0.88)
ing was confirmed by a systematic review and for light and 0.81 (95% CI 0.71–​0.94) for moderate
a meta-​analysis, reporting RRs for high intake drinking; similarly, dose-​response analyses showed
of noncruciferous vegetable of the range 0.7 to that alcohol was inversely associated with thyroid
0.9 (Liu & Lin, 2014). Therefore, cruciferous cancer risk (Fig. 25-​6) (Bagnardi et al, 2015). In a
617

Thyroid Cancer 617

4.0 RR

2.0

1.0

0.5
0 25 50
Daily alcohol intake, grams

FIGURE 25-​6  Relative risk (RR) functions and corresponding 95% Confidence Intervals describing the dose-​response
relationship between alcohol consumption and thyroid cancer risk by metaregression models.
Source: Bagnardi et al, 2015.

subsequent analysis of the European Prospective OC and hormone replacement therapy (HRT)
Investigation into Cancer and Nutrition (EPIC) for menopause—​ranged between 0.7 and 1.2.
including 556 thyroid cancer cases, drinkers of 1–​ In the collaborative reanalysis of case-​control
1.5 drinks had an HR of 0.77 (95% CI 0.60–​0.98) studies published before 1998 (Negri et al, 1999),
(Sen et al, 2015). Moreover, in a combined analysis the associations of menstrual and reproductive
of five prospective studies, the HR was 0.72 (95% factors with thyroid cancer risk were generally
CI 0.58–​0.90) for 7 or more drinks/​week versus 0 weak, although they appeared stronger among
(Kitahara et al, 2012a). women diagnosed at younger ages. A  modestly
Overall, alcohol consumption does not have increased thyroid cancer risk was observed fol-
a detrimental effect on thyroid cancer and may lowing late age at first birth (RR 1.2, 95% CI 1.0–​
be moderately associated to a reduced risk. As 1.3, for 5-​year delay), late menarche (RR 1.04, 95%
for smoking, the mechanisms explaining a pos- CI 1.0–​1.1, for 1-​year delay), and artificial meno-
sible association with alcohol intake are not pause (RR 1.8, 95% CI 1.4–​2.4). Age at first birth
clearly understood, although effects of alco- was also inversely related to risk of medullary
hol on TSH have been hypothesized (Valeix cancer (Negri et al, 2002). Parity, spontaneous or
et al, 2008). induced abortions, and history of infertility were
not associated with thyroid cancer risk (Negri
Reproductive Factors et al, 1999). In the same pooled analysis, there was
Because thyroid cancer incidence is two-​to a modest increased RR in relation to current OC
threefold higher in women than in men, hormo- use (RR 1.5, 95% CI 1.0–​2.1), which declined with
nal factors might explain this excess. Menarche, increasing time since stopping use (La Vecchia
pregnancy, oral contraceptive (OC) use, and et  al, 1999). No association was found for HRT
estrogens, in fact, are associated with thyroid for menopause. A case-​control study from France
enlargement and with a rise in serum levels of reported moderate positive associations with
TSH and other thyroid hormones. Furthermore, pregnancy, use of lactation suppressant, and early
estrogen receptors are highly expressed in menarche, and inverse ones with breastfeeding,
human and animal thyroid neoplasms (Santin & OC use, and late age at first pregnancy, but all the
Furlanetto, 2011). However, findings from sev- associations were modest and somewhat incon-
eral studies focused on menstrual and reproduc- sistent (Xhaard et al, 2014).
tive factors, as well as on exogenous hormones, A meta-​ analysis of 24 prospective stud-
have been inconsistent (Negri et  al, 1999; Dal ies, including over 5,400 thyroid cancer cases,
Maso et  al, 2009). Most RRs for reproductive, reported an increased thyroid cancer risk fol-
menstrual, and hormonal factors—​ including lowing late age at first pregnancy/​birth (RR 1.56,
618

618 Textbook of Cancer Epidemiology

95% CI 1.01–​2.42) and history of hysterectomy before 1998 found that self-​reported body mass
(RR 1.43, 95% CI 1.15–​1.78), but no associations index (BMI) at diagnosis was positively related
with other menstrual, reproductive, and hormo- to thyroid cancer risk in women, with RRs of 1.2
nal factors (Caini et al, 2015). for the highest category compared with normal
The EPIC cohort study, including 508 thyroid BMI and a RR of 1.1 for each 5-​unit increase
cancer cases, found significant associations with in BMI (Table 25-​3) (Dal Maso et al, 2000). No
history of infertility (HR 1.70, 95% CI 1.12–​2.60), consistent association was, however, observed in
a recent pregnancy (HR 3.87, 95% CI 1.43–​10.46, men. BMI at age 20 years and weight gain from
for ≤ 5 versus >5 years before recruitment), type the same age were not associated with thyroid
of menopause (HR 2.16, 95% CI 1.41–​3.31, for cancer risk.
surgical versus natural menopause), OC use at With reference to weight at diagnosis, the
recruitment (HR 0.48, 95% CI 0.25–​0.92), and RR was 1.2 (95% CI 1.0–​1.4) in women for the
duration of OC use (HR 0.66, 95% CI 0.50–​0.89, highest category versus the lowest one, with a
for ≥ 9 versus ≤ 1 year) (Zamora-​Ros et al, 2015). significant trend in risk (Dal Maso et al, 2000).
Some of these associations may, however, have No association was, however, observed in men.
been partly accounted for by increased surveil- That meta-​analysis showed a positive associa-
lance. Moreover, no significant associations were tion with height in both women (RR 1.2, 95%
observed with number of pregnancies, breast- CI 1.0–​1.4 for the highest as compared to the
feeding, menopausal status, and age at menarche lowest category) and in men (RR 1.5, 95% CI
and at menopause. 1.1–​2.1). The RR of thyroid cancer increased
In conclusion, although some modest linearly according to height, with an RR of
increased risks have been reported for late age at 1.1 for a 5-​cm increase of height in both sexes
first birth, recent pregnancy, late menarche, and combined. A  similar risk pattern emerged for
artificial menopause, the findings for reproduc- papillary and follicular thyroid cancers (Dal
tive factors and thyroid cancer are inconsistent. Maso et  al, 2000). Medullary subtype showed
a stronger RR increase (2.6, 95% CI 1.2–​5.4)
Hormones in individuals in the highest category of height
Elevated secretion of TSH from the pituitary (Negri et al, 2002).
gland has been discussed as a cause of hyperpla- A meta-​analysis of seven prospective studies,
sia and possibly thyroid cancer because it may including a total of 5,154 thyroid cancer cases,
increase mitotic activity in the follicular cells reported a 13% excess risk of thyroid cancer (95%
(Henderson et  al, 1982; Williams, 1990), thus CI 1.04–​1.22) in overweight individuals and of
increasing the probability of genetic and epi- 29% (95% CI 1.18–​1.41) in obese individuals as
genetic changes, mutations, and chromosomal compared to those of normal weight (Zhao et al,
errors related to neoplastic transformations. 2012). The results were somewhat stronger in
A higher level of TSH could occur as a con- men (RR 1.19 and 1.50, for overweight and obese
sequence of puberty, pregnancy and delivery, OC ones, respectively) than in women (RR 1.09 and
use, partial thyroidectomy for benign thyroid 1.19, respectively).
disorders, intake of goitrogens, and radiotherapy In a pooled analysis of seven cohort studies
toward the head and neck area (Pacchiarotti et al, including 393 thyroid cancer cases, an excess risk
1986; Williams, 1990). Molecular studies indicate in relation to height was found in women (HR
that influences mediated by the TSH receptor 1.13, 95% CI 1.02–​1.2, per 5 cm-​increment) but
may be involved in the carcinogenic process, but not in men (HR 0.98, 95% CI 0.86–​1.13) (Wiren
the possible mechanisms have not been clarified, et al, 2014).
and quantification of the issue remains unclear A case-​control study from France, includ-
(Valeix et al, 2008). ing 761 thyroid cancer cases and 825 controls,
reported a higher risk in relation to high body
Anthropometric Measures surface area, great height, and excess body weight,
Body fatness and adult height are associated to a and for women with a high body fat percentage;
possible increased risk of thyroid cancer (World no significant associations were, however, found
Cancer Research Fund and American Institute for BMI (Xhaard et al, 2015).
for Cancer Research, 2007). The collaborative In a pooled analysis of 22 prospective studies
reanalysis of case-​ control studies published (including 833,176 men and 1,260,871 women,
619

Thyroid Cancer 619

TABLE 25-​3   RELATIVE RISKS (RR) a AND 95% CONFIDENCE INTERVALS (CI)

OF THYROID CANCER ACCORDING TO ANTHROPOMETRIC FACTORS

N. of studies Categorization RR 95% CI

Height
Dal Maso et al, 2000 12 High versus Low (Women) 1.2 1.0–​1.4
12 High versus Low (Men) 1.5 1.1–​2.1
Wiren et al, 2014 7 High versus Low (Women) 1.13 1.02–​1.20
7 High versus Low (Men) 0.98 0.86–​1.13
Kitahara et al, 2016 22 5 cm increment 1.07 1.04–​1.10

Weight
Dal Maso et al, 2000 12 High versus Low (Women) 1.2 1.0–​1.4
12 High versus Low (Men) 1.0 0.8–​1.4
Kitahara et al, 2016 22 5 kg/​m2 increment 1.06 1.02–​1.10

Body mass index

Dal Maso et al, 2000 12 High versus Low (Women) 1.2 1.0–​1.4
High versus Low (Men) 1.0 0.8–​1.4
Zhao et al, 2012 3 Obese versus normal (Women) 1.19 1.07–​1.34
4 Obese versus normal (Men) 1.50 1.29–​1.75

a
Adjusted at least for age and sex.

among whom 2,996 incident thyroid cancers and body fat) are affected by thyroid disorders
104 thyroid cancer deaths were identified), all (Franceschi et  al, 1999). Body size might also
anthropometric factors were positively associ- affect iodine requirement and, indirectly, thy-
ated with thyroid cancer incidence. The HRs were roid cancer risk. Moreover, mechanisms related
1.07 (95% CI 1.04–​1.10) for a 5-​cm increment to inflammation, as well as insulin resistance
in height, 1.06 (95% CI 1.02–​1.10) for a 5-​kg/​m2 resulting from adiposity, may also be involved
increase in BMI, 1.03 (95% CI 1.01–​1.05) for a (Pazaitou-​Panayiotou et al, 2013). The associa-
5-​cm increase in waist circumference, 1.13 (95% tion with height indicates the influence of some
CI 1.02–​1.25) for a 5-​kg/​m2 increase in young-​ growth factors or hormones during childhood/​
adult BMI, and 1.07 (95% CI 1.00–​1.15) for a 5-​ adolescence, but the potential role of growth
kg/​m2 increase in adulthood BMI (Kitahara et al, factors in thyroid carcinogenesis is still poorly
2016). Baseline BMI was more strongly associ- defined (Vella et al, 2001).
ated with risk in men as compared with women.
Positive associations were observed for papillary, Physical Activity
follicular, and anaplastic, but not medullary, thy- Physical activity has been inversely related with
roid carcinomas. Somewhat stronger associa- cancer risk through a variety of mechanisms
tions were observed for thyroid cancer mortality. including hormonal, metabolic, and immu-
Thus, overweigh/​ obesity and adult height nologic effects (IARC Working Group on the
are associated with a modestly increased risk Evaluation of Cancer-​ Preventive Strategies,
of thyroid cancer. Such an association is bio- 2002). However, only a few studies analyzed the
logically plausible, although the mechanisms association with thyroid cancer, providing incon-
are still unclear (IARC Working Group on the sistent results. In a pooled analysis of five pro-
Evaluation of Cancer-​ Preventive Strategies, spective studies, including 818 thyroid cancers,
2002). Thyroid hormones are crucial to the individuals reporting the highest level of (total)
development of several body tissues, and physical activity had an 18% increased risk of
anthropometric measures (i.e., height, weight, thyroid cancer (95% CI 1.0–​1.39) (Kitahara et al,
620

620 Textbook of Cancer Epidemiology

2012b). The association was, however, restricted estimated for a dose of 1 Gy, and a dose-​risk
to individuals who were overweight/​obese (RR relationship was observed at least up to 2 Gy.
1.34) as opposed to those who were of normal No reliable estimate of a dose-​ risk relation-
weight (RR 0.98). A  subsequent meta-​analysis, ship was possible for higher doses, due to the
including seven studies with a total of 2,250 thy- low number of individuals exposed. The risk of
roid cancer cases, reported a summary RR of 1.06 radiation-​related thyroid cancer was threefold
(95% CI 0.79–​1.42) for the highest versus the low- higher in iodine-​deficient areas than elsewhere,
est level of physical activity (Schmid et al, 2013). and the administration of potassium iodide as a
The RR was significantly above unity in cohort dietary supplement reduced the risk.
studies (RR 1.28), but nonsignificantly below Among adults exposed to Chernobyl acci-
unity in case-​control ones (RR 0.70). Moreover, dent, as in previous studies on adult populations,
an excess risk of thyroid cancer was evident in the association between ionizing radiation expo-
one study analyzing specifically occupational sure and thyroid cancer is inconsistent (Ron et al,
physical activity (RR 2.94), but not in three stud- 1995). Hence, the thyroid appears less radiosen-
ies analyzing leisure-​time physical activity (RR sitive in adults than in children, and/​or the prev-
0.82) and in three analyzing total physical activ- alence of iodine deficiency is lower in adults than
ity (RR 1.15). in children (Moysich et al, 2002). Several studies
In conclusion, findings are limited and incon- considered adults exposed to 131I for medical pur-
sistent and no clear role of physical activity on poses (UNSCEAR, 2000). Although those studies
thyroid cancer risk has been established. suggested an increased risk, their interpretation
is complicated by the fact that some of these
Ionizing Radiation patients were treated because of thyroid diseases.
Ionizing radiation is the best-​established cause Studies of occupational exposure to low-​ level
of thyroid cancer, particularly in children ionizing radiation, typically in the nuclear indus-
and young adults (United Nations Scientific try, have shown no increased incidence of thy-
Committee on the Effects of Atomic Radiation roid cancer (Cardis et al, 2006).
[UNSCEAR], 2000). A pooled analysis based on Ionizing radiation is therefore a major well-​
seven studies, including 58,000 exposed indi- confirmed cause of thyroid cancer, particularly in
viduals (i.e., irradiated in childhood for medical children and young adults.
conditions and atomic bomb survivors), 61,000
unexposed, and 700 thyroid cancers, reported a Occupation
summary excess RR of 7.7 per Gray (Gy), and an Except for occupational studies of individuals
excess absolute risk of 4.4 per 10,000 persons-​ exposed to ionizing radiation, such as military
year-​Gy (Ron et al, 1995). The dose-​risk relation- service personnel, nuclear industry workers,
ship was consistent with linearity even at a dose and individuals working in medical settings or
of 0.1 Gy, but leveled off at doses higher than 10 underground miners, no data support a rela-
Gy. A latent period of 5–​9 years emerged, with tionship between occupation and thyroid cancer
the highest RR 15–​29 years after exposure, which (UNSCEAR, 2000).
remained elevated 40 or more years after stop-
ping exposure. Women had an approximately Medical Conditions and Treatments
twofold higher RR per Gy of exposure than men. Associations between history of benign thyroid
The RR decreased significantly with increasing diseases (mainly goiter and benign nodules/​
age at exposure. adenomas) and thyroid cancer have been con-
An increased incidence of thyroid can- sistently reported, with excess risks up to four-​to
cer was reported among children (but not fivefold (Dal Maso et al, 2009). In the collabora-
among adults) living in the areas of Belarus and tive reanalysis of case-​control studies published
Ukraine contaminated from iodine 131 (131I) before 1998, goiter and benign nodules/​adeno-
after the Chernobyl accident in 1986 (Moysich mas were the strongest risk factors for thyroid
et al, 2002). In a population-​based case-​control cancer, apart from radiation in childhood (and
study conducted in Belarus and in the Russian genetic factors for medullary cancer) (Franceschi
Federation including 276 thyroid cancer cases et al, 1999). Elevated risks were observed for both
and 1,300 matched-​controls, below age 15 years men and women and in relation to both papil-
at the time of the accident, an approximately lary and follicular cancers, with overall RRs on
sevenfold excess risk of thyroid cancer was the order of five (Table 25-​4). The excess risk
621

Thyroid Cancer 621

TABLE 25-​4   RELATIVE RISKS (RR) a AND 95% CONFIDENCE INTERVALS

(CI) OF THYROID CANCER ACCORDING TO BENIGN THYROID
CONDITIONS

Affected Unaffected RR 95% CI

Cases/​Controls Cases/​Controls

Nodules/​Adenomas
Franceschi et al, 1999 148/​8 1475/​2465 29.9 14.5–​62.0
Memon et al, 2002 34/​0 279/​313 ∞ 12–​∞

Goiter
Franceschi et al, 1999 178/​56 1768/​2939 5.9 4.2–​8.1
Iribarren et al, 2001 –​ –​ 3.4 1.8–​6.2
Memon et al, 2002 21/​4 292/​309 5.3 1.8–​15.3
Truong et at, 2005 30/​10 263/​344 4.2 2.0–​9.1

Goiter or nodules
Horn-​Ross et al, 2001 134/​36 474/​522 4.7 3.1–​7.2

Hyperthyroidism
Franceschi et al, 1999 58/​60 2000/​3150 1.4 1.0–​2.1
Horn-​Ross et al, 2001 19/​17 589/​541 0.6 0.3–​1.2
Iribarren et al, 2001 –​ –​ 0.9 0.3–​2.2
Memon et al, 2002 5/​3 308/​310 1.7 0.4–​7.0

Hypothyroidism
Franceschi et al, 1999 82/​141 1703/​2663 0.9 0.7–​1.3
Horn-​Ross et al, 2001 34/​35 574/​523 0.7 0.4–​1.1
Iribarren et al, 2001 –​ –​ 0.7 0.4–​1.2
Memon et al, 2002 9/​5 304/​308 1.8 0.6–​5.4
a
Adjusted at least for age and sex.

was greatest within 2 to 4  years after diagnosis represent components of the underlying MEN 2
of benign thyroid disease, but persisted 10 years syndrome (Franceschi, 1998; Feldt-​Rasmussen,
or more of follow-​ up. Prior hyperthyroidism 2001; Nagataki & Nystrom, 2002).
was related to a small, nonsignificant increased History of diabetes has been associated to a
thyroid cancer risk that was reduced after allow- modest increased risk of thyroid cancer, the RR
ance for a history of goiter, while hypothyroid- being 1.34 (95% CI 1.11–​1.63) in a meta-​analysis
ism was not associated with cancer risk. Other of 14 cohort and 3 case-​control studies (Yeo et al,
studies from the United States, South Pacific, 2014). The excess risk was, however, found in
and Kuwait reported findings consistent with the women (RR 1.38) but not in men (RR 1.11), and
results of that reanalysis (Horn-​Ross et al, 2001; the results of single studies were highly hetero-
Iribarren et al, 2001; Memon et al, 2002; Truong geneous. In particular, a large prospective study
et al, 2005). from the United States including 818 thyroid can-
Medullary cancer was also directly related cers found no association between self-​reported
to history of thyroid nodules and other medical history of diabetes and thyroid cancer risk (HR
conditions, such as hypertension, allergy, and 1.08, 95% CI 0.83–​1.40) (Kitahara et al, 2012b).
gallbladder disease (Negri et  al, 2002). Some Similarly, a subsequent study based on National
of these conditions (e.g., hypertension) may Health Insurance data from Taiwan, reported a
62

622 Textbook of Cancer Epidemiology

plausible, such an association should be further

TABLE 25-​5   EPIDEMIOLOGY OF THYROID
investigated (Yeo et al, 2014).
CANCER: RISK FACTOR SUMMARY Therefore, history of benign thyroid dis-
Risk factor Direction eases, namely goiter and benign nodules/​ade-
of effecta nomas, are among the major established risk
factors for thyroid cancer, while hyperthyroid-
Well-​Confirmed Risk Factors ism and adult-​onset diabetes are positively but
Familial and genetic factors inconsistently associated with the risk of this
neoplasm.
Papillary carcinoma ↑
Medullary carcinoma ↑↑ CONCLUSION
Iodine imbalance ↑ Thyroid cancer incidence has been increasing over
Ionizing radiation exposure ↑↑ the last few decades in many countries worldwide,
Benign nodules/​adenomas ↑↑ following improved ascertainment, diagnosis,
and certification of the disease, whereas mortality
Goiter ↑↑
has steadily declined, likely because of improved
Probable Relationship Exists, Based on diagnosis, management, and treatment. Selected
Substantial Data risk factors for thyroid cancers and the strength
and direction of their effects are summarized in
Body mass index ↑
Table 25-​5. Thyroid cancers (particularly medul-
Height ↑ lary ones) have strong familial and hence genetic
Cigarette smoking ↓ associations, more than most other neoplasms.
Alcohol use ↓ Ionizing radiation, particularly in children and
young adults, iodine imbalance (i.e., iodine defi-
Weak, if Any, Relationship Exists, ciency and iodine excess), and history of benign
Based on Substantial Data thyroid diseases, namely goiter and benign nod-
Fish —​ ules/​
adenomas, are other well-​ confirmed risk
Vegetable ↓ factors for thyroid cancer. Consequently, the best
approach toward reducing thyroid cancer risk is
Inconsistent Findings or Limited Study to avoid ionizing radiation and iodine deficiency,
to Date particularly in childhood. Avoiding overweight
Reproductive factors (late age at first ↑ and obesity and, possibly, adopting a diet rich
birth, recent pregnancy, late menarche, in vegetables may play a favorable role in thy-
and artificial menopause) roid cancer. Fish consumption does not seem to
appreciably increase thyroid cancer risk, and may
High TSH levels ↑
have a favorable influence in areas where iodine
Physical activity —​ deficiency is, or was, common. Findings are
Occupation, other than those related to —​ inconsistent with reference to other risk factors,
ionizing radiation reproductive factors (late age at first birth, recent
Hyperthyroidism ↑ pregnancy, late menarche, artificial menopause),
Adult-​onset diabetes ↑ high TSH levels, physical activity, occupation
other than those involving exposure to ioniz-
TSH: thyroid stimulating hormone. ing radiation, hyperthyroidism, and adult-​onset
a
Arrows indicate the approximate magnitude of the relationship: ↑, diabetes. No conclusively causal associations
slight to moderate increase in risk; ↑↑, moderate to large increase
in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large emerged with tobacco and alcohol.
decrease in risk; —​, no association

REFERENCES
RR of thyroid cancer of 0.82 (95% CI 0.65–​1.02) Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F,
in relation to history of diabetes. In that study an Fedirko V et  al. Alcohol consumption and site-​
increased risk of thyroid cancer was, however, specific cancer risk: a comprehensive dose-​response
observed for diabetic patients treated with sul- meta-​analysis. Br J Cancer 2015;112:580–​93.
fonylureas (RR 1.88, 95% CI 1.20–​2.95) (Tseng, Baker SR, Bhatti WA. The thyroid cancer epidemic: is
2012). Although a role of diabetes on thyroid it the dark side of the CT revolution? Eur J Radiol
cancer—​ as on other common neoplasms—​ is 2006;60:67–​9.
623

Thyroid Cancer 623

Baron JA, La Vecchia C, Levi F. The antiestrogenic F. GLOBOCAN 2012 v1.0, Cancer Incidence and
effect of cigarette smoking in women. Am J Obstet Mortality Worldwide: IARC CancerBase No. 11
Gynecol 1990;162:502–​14. [Internet]. Lyon, France: International Agency for
Bosetti C, Kolonel L, Negri E, Ron E, Franceschi S, Research on Cancer; 2013. Available from: http://
Dal Maso L et al. A pooled analysis of case-​control globocan.iarc.fr, accessed 2017.
studies of thyroid cancer. VI. Fish and shellfish con- Franceschi S. Iodine intake and thyroid carcinoma-​
sumption. Cancer Causes Control 2001;12:375–​82. a potential risk factor. Exp Clin Endocrinol
Bosetti C, Negri E, Kolonel L, Ron E, Franceschi S, Diabetes 1998;106 Suppl 3: S38–​44.
Preston-​Martin S et al. A pooled analysis of case-​ Franceschi S, Preston-​Martin S, Dal Maso L, Negri
control studies of thyroid cancer. VII. Cruciferous E, La Vecchia C, Mack WJ et  al. A pooled anal-
and other vegetables (International). Cancer ysis of case-​control studies of thyroid cancer. IV.
Causes Control 2002;13:765–​75. Benign thyroid diseases. Cancer Causes Control
Caini S, Gibelli B, Palli D, Saieva C, Ruscica M, Gandini 1999;10:583–​95.
S. Menstrual and reproductive history and use of Frank C, Fallah M, Ji J, Sundquist J, Hemminki K.
exogenous sex hormones and risk of thyroid can- The population impact of familial cancer, a major
cer among women: a meta-​analysis of prospective cause of cancer. Int J Cancer 2014;134:1899–​906.
studies. Cancer Causes Control 2015;26:511–​8. Hemminki K, Eng C, Chen B. Familial risks for non-
Cardis E, Howe G, Ron E, Bebeshko V, Bogdanova medullary thyroid cancer. J Clin Endocrinol
T, Bouville A et  al. Cancer consequences of the Metab 2005;90:5747–​53.
Chernobyl accident:  20  years on. J Radiol Prot Henderson BE, Ross RK, Pike MC, Casagrande JT.
2006;26:127–​40. Endogenous hormones as a major factor in human
Cho YA, Kim J. Thyroid cancer risk and smoking cancer. Cancer Res 1982;42:3232–​9.
status: a meta-​analysis. Cancer Causes Control Ho AS, Davies L, Nixon IJ, Palmer FL, Wang LY, Patel
2014;25:1187–​95. SG et  al. Increasing diagnosis of subclinical thy-
Cho YA, Kim J. Dietary factors affecting thyroid cancer roid cancers leads to spurious improvements in
risk: a meta-​analysis. Nutr Cancer 2015;67:811–​7. survival rates. Cancer 2015;121:1793–​99.
Curado MP, Edwards B, Shin HR, Storm H, Ferlay Horn-​ Ross PL, Morris JS, Lee M, West DW,
J, Heanue M et  al. Cancer Incidence in Five Whittemore AS, McDougall IR et  al. Iodine and
Continents, Vol 9. IARC Sci Publ No. 160. Lyon, thyroid cancer risk among women in a multi-
France:  International Agency for Research on ethnic population:  the Bay Area Thyroid Cancer
Cancer, 2007. Study. Cancer Epidemiol Biomarkers Prev
Dal Maso L, Bosetti C, La Vecchia C, Franceschi S. 2001;10:979–​85.
Risk factors for thyroid cancer: an epidemiologi- IARC Working Group on the Evaluation of Cancer-​
cal review focused on nutritional factors. Cancer Preventive Strategies. Weight Control and Physical
Causes Control 2009;20:75–​86. Activity. IARC Handbooks of Cancer Prevention.
Dal Maso L, La Vecchia C, Franceschi S, Preston-​ Lyon, France: IARC Press, 2002.
Martin S, Ron E, Levi F et al. A pooled analysis of Iribarren C, Haselkorn T, Tekawa IS, Friedman GD.
thyroid cancer studies. V. Anthropometric factors. Cohort study of thyroid cancer in a San Francisco
Cancer Causes Control 2000;11:137–​44. Bay area population. Int J Cancer 2001;93:745–​50.
Davies L, Welch HG. Increasing incidence of thyroid Kanno J, Matsuoka C, Furuta K, Onodera H, Miyajima
cancer in the United States, 1973–​2002. JAMA H, Maekawa A et  al. Tumor promoting effect
2006;295:2164–​7. of goitrogens on the rat thyroid. Toxicol Pathol
De Lellis RA, Lloyd RV, Heitz PU, Eng C. World Health 1990;18:239–​46.
Organization Classification of tumours:  pathol- Kitahara CM, McCullough ML, Franceschi S, Rinaldi
ogy and genetics of tumours of endocrine organs. S, Wolk A, Neta G, et al. Anthropometric factors
Lyon, France: IARC Press, 2004. and thyroid cancer risk by histological subtype:
Fallah M, Pukkala E, Tryggvadottir L, Olsen JH, pooled analysis of 22 prospective studies. Thyroid
Tretli S, Sundquist K, et  al. Risk of thyroid can- 2016;26:306–​18.
cer in first-​degree relatives of patients with non-​ Kitahara CM, Linet MS, Beane Freeman LE, Check
medullary thyroid cancer by histology type and DP, Church TR, Park Y, et al. Cigarette smoking,
age at diagnosis:  a joint study from five Nordic alcohol intake, and thyroid cancer risk: a pooled
countries. J Med Genet 2013;50:373–​82. analysis of five prospective studies in the United
Feldt-​Rasmussen U. Iodine and cancer. Thyroid States. Cancer Causes Control 2012a;23:1615–​24.
2001;11:483–​6. Kitahara CM, Platz EA, Beane Freeman LE, Black A,
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Hsing AW, Linet MS et al. Physical activity, dia-
Mathers C, Rebelo M, Parkin DM, Forman D, Bray, betes, and thyroid cancer risk: a pooled analysis
624

624 Textbook of Cancer Epidemiology

of five prospective studies. Cancer Causes Control serum free thyroid hormones in pregnancy. J
2012b; 23:463–71. Endocrinol Invest 1986;9:185–​189.
Kolonel LN, Hankin JH, Wilkens LR, Fukunaga Pazaitou-​Panayiotou K, Polyzos SA, Mantzoros CS.
FH, Hinds MW. An epidemiologic study of thy- Obesity and thyroid cancer:  epidemiologic asso-
roid cancer in Hawaii. Cancer Causes Control ciations and underlying mechanisms. Obes Rev
1990;1:223–​34. 2013;14:1006–​22.
Kondo T, Ezzat S, Asa SL. Pathogenetic mechanisms in Ron E, Lubin JH, ShoSchneider AB, Tucker MA, Boice
thyroid follicular-​cell neoplasia. Nat Rev Cancer JD, Jr. Thyroid cancer after exposure to exter-
2006;6:292–​306. nal radiation:  a pooled analysis of seven studies.
La Vecchia C, Malvezzi M, Bosetti C, Garavello W, Radiat Res 1995;141:259–​77.
Bertuccio P, Levi F et al. Thyroid cancer mortal- Santin AP, Furlanetto TW. Role of estrogen in thyroid
ity and incidence: a global overview. Int J Cancer function and growth regulation. J Thyroid Res
2015;136:2187–​95. 2011;2011:875125.
La Vecchia C, Ron E, Franceschi S, Dal Maso L, Mark Schmid D, Behrens G, Jochem C, Keimling M,
SD, Chatenoud L et al. A pooled analysis of case-​ Leitzmann M. Physical activity, diabetes, and risk
control studies of thyroid cancer. III. Oral contra- of thyroid cancer: a systematic review and meta-​
ceptives, menopausal replacement therapy and analysis. Eur J Epidemiol 2013;28:945–​58.
other female hormones. Cancer Causes Control Sen A, Tsilidis KK, Allen NE, Rinaldi S, Appleby
1999;10:157–​66. PN, Almquist M et al. Baseline and lifetime alco-
Lauper JM, Krause A, Vaughan TL, Monnat RJ Jr. hol consumption and risk of differentiated thy-
Spectrum and risk of neoplasia in Werner syndrome: roid carcinoma in the EPIC study. Br J Cancer
a systematic review. PLoS One 2013;8:e59709. 2015;113:840–​7.
Liu ZT, Lin AH. Dietary factors and thyroid cancer Soldin OP, Goughenour BE, Gilbert SZ, Landy HJ,
risk:  a meta-​ analysis of observational studies. Soldin SJ. Thyroid hormone levels associated with
Nutr Cancer 2014;66:1165–​78. active and passive cigarette smoking. Thyroid
Mack WJ, Preston-​Martin S, Dal Maso L, Galanti 2009;19:817–​23.
R, Xiang M, Franceschi S, Hallquist A et  al. A Steinhagen E, Guillem JG, Chang G, Salo-​Mullen EE,
pooled analysis of case-​control studies of thyroid Shia J, Fish S et al. The prevalence of thyroid can-
cancer:  cigarette smoking and consumption of cer and benign thyroid disease in patients with
alcohol, coffee, and tea. Cancer Causes Control familial adenomatous polyposis may be higher
2003;14:773–​85. than previously recognized. Clin Colorectal
Memon A, Varghese A, Suresh A. Benign thy- Cancer 2012;11:304–​8.
roid disease and dietary factors in thyroid can- Truong T, Orsi L, Dubourdieu D, Rougier Y, Hemon
cer:  a case-​control study in Kuwait. Br J Cancer D, Guenel P. Role of goiter and of menstrual
2002;86:1745–​50. and reproductive factors in thyroid cancer:  a
Moysich KB, Menezes RJ, Michalek AM. Chernobyl-​ population-​ based case-​ control study in New
related ionising radiation exposure and cancer Caledonia (South Pacific), a very high incidence
risk:  an epidemiological review. Lancet Oncol area. Am J Epidemiol 2005;161:1056–​65.
2002;3:269–​79. Tseng CH. Thyroid cancer risk is not increased in dia-
Nagataki S, Nystrom E. Epidemiology and pri- betic patients. PLoS One 2012;7:e53096.
mary prevention of thyroid cancer. Thyroid United Nations Scientific Committee on the Effects
2002;12:889–​96. of Atomic Radiation (UNSCEAR). Sources and
Negri E, Ron E, Franceschi S, Dal Maso L, Mark SD, effects of ionizing radiation. UNSCEAR 2000
Preston-​Martin S et al. A pooled analysis of case-​ Report to the General Assembly, with Scientific
control studies of thyroid cancer. I.  Methods. Annexes. New York: United Nations, 2000.
Cancer Causes Control 1999;10:131–​42. Vaccarella S, Dal Maso L, Laversanne M, Bray F,
Negri E, Ron E, Franceschi S, La Vecchia C, Preston-​ Plummer M, Franceschi S. The impact of diagnostic
Martin S, Kolonel L, et al. Risk factors for medul- changes on the rise in thyroid cancer incidence: a
lary thyroid carcinoma: a pooled analysis. Cancer population-​based study in selected high-​resource
Causes Control 2002;13:365–​72. countries. Thyroid 2015;25:1127–​36.
Nose V. Familial thyroid cancer: a review. Mod Pathol Valeix P, Faure P, Bertrais S, Vergnaud AC, Dauchet
2011;24 Suppl 2:S19–​33. L, Hercberg S. Effects of light to moder-
Pacchiarotti A, Martino E, Bartalena L, Buratti L, ate alcohol consumption on thyroid volume
Mammoli C, Strigini F, et  al. Serum thyrotropin and thyroid function. Clin Endocrinol (Oxf)
by ultrasensitive immunoradiometric assay and 2008;68:988–​95.
625

Thyroid Cancer 625

Vella V, Sciacca L, Pandini G, Mineo R, Squatrito S, factors for differentiated thyroid cancer in young
Vigneri R et  al. The IGF system in thyroid can- men and women from eastern France:  a case-​
cer: new concepts. Mol Pathol 2001;54:121–​4. control study. Am J Epidemiol 2015;182:202–​14.
Weissel M. Legal augmentation of iodine content in Xhaard C, Rubino C, Clero E, Maillard S, Ren Y, Borson-​
table salt from 10 to 20 mg KI/​kg:  documented Chazot F et al. Menstrual and reproductive factors
effects a decade later. Exp Clin Endocrinol in the risk of differentiated thyroid carcinoma in
Diabetes 2003;111:187–​90. young women in France: a population-​based case-​
Williams ED. TSH and thyroid cancer. Horm Metab control study. Am J Epidemiol 2014;180:1007–​17.
Res Suppl 1990;23:72–​5. Yeo Y, Ma SH, Hwang Y, Horn-​Ross PL, Hsing A,
Wiren S, Haggstrom C, Ulmer H, Manjer J, Bjorge T, Lee KE et al. Diabetes mellitus and risk of thyroid
Nagel G et al. Pooled cohort study on height and cancer: a meta-​analysis. PLoS One 2014;9:e98135.
risk of cancer and cancer death. Cancer Causes Zamora-​Ros R, Rinaldi S, Biessy C, Tjonneland A,
Control 2014;25:151–​9. Halkjaer J, Fournier A et  al. Reproductive and
World Cancer Research Fund and American Institute menstrual factors and risk of differentiated thy-
for Cancer Research. Food Nutrition, Physical roid carcinoma:  the EPIC study. Int J Cancer
Activity and the Prevention of Cancer: A Global 2015;136:1218–​27.
Perspective. Washington, DC: American Institute Zevallos JP, Hartman CM, Kramer JR, Sturgis EM,
for Cancer Research, 2007. Chiao EY. Increased thyroid cancer incidence cor-
World Health Organization Statistical Information responds to increased use of thyroid ultrasound
System. WHO mortality database. Geneva: World and fine-​needle aspiration: a study of the Veterans
Health Organization. Available at http://​ Affairs health care system. Cancer 2015;121:741–​46.
www.who.int/​h ealthinfo/​s tatistics/​m ortality_ Zhao ZG, Guo XG, Ba CX, Wang W, Yang YY, Wang
​rawdata/​en/​index.html (accessed 25 February J et  al. Overweight, obesity and thyroid cancer
2014). risk: a meta-​analysis of cohort studies. J Int Med
Xhaard C, de Vathaire F, Clero E, Maillard S, Ren Res 2012;40:2041–​50.
Y, Borson-​Chazot F et  al. Anthropometric risk
62
627

26
Hodgkin Lymphoma
H E N R I K H J A L G R I M , M A D S M E L B Y E , A N D PA G O N A   L A G I O U

H odgkin lymphoma is named after Thomas

Hodgkin (1798–​1866), a British pathologist,
philanthropist, and traveler, who eventually died
classical Hodgkin lymphoma (Stein et al, 2008b,
Weiss et al, 2008; Anagnostopoulos et al, 2008;
Benharroch et al, 2008).
from dysentery in the Orient. First described Symptoms Hodgkin lymphoma occurs at all ages,
by Hodgkin already in 1832 (“On some mor- but is among the most common malignancies in
bid appearances of the absorbent glands and adolescence and early adulthood (Ferlay et  al,
spleen”) (Hodgkin, 1832), the disease became 2013). The clinical presentation varies, but often
named after him in 1865 (Wilks, 1865). The vast involves painless adenopathy of the neck, which
majority of Hodgkin lymphomas derive from may be accompanied by unspecific symptoms
preapoptotic germinal center B lymphocytes, such as fever, night sweats, itching of the skin,
but from peripheral T-​cells in a small subset of alcohol intolerance, weight loss, and/​or fatigue
cases (<2%) (Kanzler et al, 1996; Muschen et al, (Stein & Morgan, 2004).
2000). Hodgkin lymphoma has a unique histo-
Diagnosis Biopsy is necessary for the microscopic
logical presentation, in that the malignant cells,
identification of the neoplastic Reed-​Sternberg
the characteristic Reed-​Sternberg cells or vari-
cells or their variants. The extent of disease is
ants thereof, typically constitute less than 1% of
determined by imaging techniques (ultrasound,
the tumor lesion, which is instead dominated
computerized tomography [CT], magnetic reso-
by benign inflammatory cells (Stein, 2008).
nance imaging [MRI], and/​or positron emission
Combined with a characteristic clinical presen-
tomography [PET]) and bone marrow biopsy
tation this originally led to the speculation that
(PDQ® Adult Treatment Editorial Board, 2016).
Hodgkin lymphoma is an unusual syndrome
of an inflammatory nature rather than a true Treatment A combination of radiation and che-
neoplasm. However, with the discovery that motherapy is the most common treatment.
the Reed-​Sternberg cells are lymphoid cells of Other regimens such as autologous stem cell
clonal origin, Hodgkin lymphoma is now con- transplantation and, more recently, biologi-
sidered a malignant lymphoma (Jaffe et al, 1998; cal therapies are relevant with more resistant
Harris, 1999). or recurrent disease (PDQ® Adult Treatment
Editorial Board, 2016).
CLINICAL SYNOPSIS Prognosis Hodgkin lymphoma generally has a
Classification WHO recognizes two immuno- favorable prognosis. The 5-​year survival rate has
logically distinct types of Hodgkin lymphoma, improved from 40% in the beginning of the 1960s
namely, classical Hodgkin lymphoma (Stein et (Howlader et  al, 2014)  to currently well above
al, 2008a) and nodular lymphocyte predominant 90% for those with early stage disease (Armitage,
Hodgkin lymphoma (~5% of all Hodgkin lym- 2010), and from 5% to around 90% in some trials
phomas) (Poppema et al, 2008). Based exclusively for those with advanced disease (Engert, 2015).
on histological appearance, classical Hodgkin The high survival has increased the attention
lymphoma is further divided into four variants: toward the multitude of late treatment-​related
nodular sclerosis (~70% of cases), mixed cellular- complications among the survivors, such as car-
ity (20%–​25% of cases), lymphocyte-rich (~5% diovascular disease and secondary malignancies
of cases), and lymphocyte-​depleted (<5% cases) (Ng, 2014).
628

628 Textbook of Cancer Epidemiology

DESCRIPTIVE Age-​Specific Incidence Patterns

EPIDEMIOLOGY The association with socioeconomic condi-
tions also manifests as variations in age-​specific
Incidence and Mortality Hodgkin lymphoma incidence patterns between
In 2012, an estimated 27,500 women and 38,500 populations (Figs. 26-​ 3, 26-​
4). Early interna-
men were diagnosed with Hodgkin lymphoma tional surveys of Hodgkin lymphoma incidence
worldwide. These figures correspond to annual summarized this variation into four archetypi-
age-​standardized (world) incidence rates of 0.7 and cal age-​specific incidence patterns, elements of
1.1 per 100,000 among women and men, respec- which can still be recognized today.
tively, emphasizing that Hodgkin lymphoma is In affluent populations in the Western hemi-
a rare cancer (Ferlay et al, 2013). The total num- sphere, Hodgkin lymphoma occurrence follows
ber of deaths attributed to Hodgkin lymphoma in a bimodal age distribution, with low incidence
2012 was estimated to 25,500, corresponding to among children followed by separate incidence
age-​standardized mortality rates of 0.3 and 0.4 per peaks among younger adults and older adults,
100,000 in women and men (Ferlay et al, 2013). respectively (MacMahon, 1957). This pattern
Hodgkin lymphoma occurrence displays generally applies to both males and females,
considerable geographic variation, which mir- however, the younger adult incidence peak is
rors levels of socioeconomic development of more pronounced among women than among
the underlying population (Figs. 26-​ 1, 26-​2). men, and the older adult peak conversely more
Accordingly, by WHO definitions of socioeco- pronounced among men. Accordingly, while
nomic development, Hodgkin lymphoma inci- Hodgkin lymphoma incidence is the highest for
dence is higher in high-​and middle-​ income males among children and older adults, it may
countries (2.1 per 100,000 persons per year both be the highest for females in adolescents and
sexes combined) than in low-​income countries younger adults (Forman et al, 2013).
(0.6 cases per 100,000 persons per year both Also in deprived populations, age-​ specific
sexes combined) (Ferlay et al, 2013). Hodgkin lymphoma incidence was previously sug-
In addition to socioeconomic conditions, gested to follow a bimodal distribution. However,
Hodgkin lymphoma occurrence also varies by in these settings the “younger” incidence peak was
race and ethnicity both within and between pop- seen among children, notably among boys, rather
ulations. For instance, age-​standardized (world) than among younger adults (Correa & O’Conor,
Hodgkin lymphoma incidence may differ 1971; Gutensohn & Cole, 1977). Reminiscent of
several-​fold between socioeconomically compa- this pattern, the incidence of Hodgkin lymphoma
rable populations, for example, between 0.5 and among boys remains higher in some economically
2.2 per 100,000 persons per year for both sexes less developed countries, for example, in Latin
combined in Japan and in the European Union, America, than in affluent populations such as the
respectively (Ferlay et al, 2013). European Union (Forman et al, 2013).

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 0.5 1 1.5 2 2.5
ASR (world) per 100,000 person-years

FIGURE 26-​1  Age-​standardized (to the 2012 world population) incidence rates of Hodgkin lymphoma among women.
Source: Ferlay et al, 2013.
629

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 0.5 1 1.5 2 2.5 3
ASR (world) per 100,000 person-years

FIGURE 26-​2  Age-​standardized (to the 2012 world population) incidence rates of Hodgkin lymphoma among men.
Source: Ferlay et al, 2013.

5
New cases per 100,000

4
person-years

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 26-​3  Age-​specific incidence rates of Hodgkin lymphoma among women.

Source: Ferlay et al, 2013.
New cases per 100,000

6
5
person-years

4
3
2
1
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 26-​4  Age-​specific incidence rates of Hodgkin lymphoma among men.

Source: Ferlay et al, 2013.
630

630 Textbook of Cancer Epidemiology

In populations in the middle of the socioec- levels of development is mirrored by the occur-
onomic spectrum, for example, in rural areas of rence of the major subtypes of Hodgkin lym-
Western countries, incidence patterns featuring phoma, namely, nodular sclerosis and mixed
both a relatively high incidence among child- cellularity classical Hodgkin lymphoma.
ren and a peak among younger adults have been The incidence peak among younger adults in
reported (Correa & O’Conor, 1971; Gutensohn & Western societies is primarily made up of nodular
Cole, 1977). Finally, a fourth Hodgkin lymphoma sclerosis classical Hodgkin lymphoma. Indeed,
age-​specific incidence pattern has been observed the age-​specific incidence of nodular sclerosis
in Asian populations with a uniformly low inci- classical Hodgkin lymphoma per se is bimod-
dence in the first four to five decades of life and ally distributed with a pronounced younger adult
neither a childhood or a younger adult incidence incidence peak followed by a smaller incidence
peak, consistent with the lower overall incidence peak among older adults (Fig. 26-​6) (Howlader
in these settings (MacMahon, 1957, 1966). et  al, 2014). Among younger adults, the inci-
As suspected from the correlation with socio- dence of nodular sclerosis classical Hodgkin
economic level, the age-​specific incidence pattern lymphoma may be higher in females than in
for Hodgkin lymphoma is not static, but changes in males (Howlader et  al, 2014). Consistent with
conjunction to secular developments. Such a tran- the international variation in overall Hodgkin
sition is evident, for example, from age-​specific lymphoma occurrence, nodular sclerosis classi-
incidence patterns in Connecticut from different cal Hodgkin lymphoma among younger adults is
time calendar periods, showing an augmentation associated with socioeconomic affluence (Clarke
of a preexisting younger adult incidence peak, in et al, 2005).
particular among women (Fig. 26-​5). Analogously, Nodular sclerosis classical Hodgkin lym-
in Asian populations such as that of Singapore a phoma is the most common disease variant in all
younger adult incidence peak has emerged fol- age groups except among children, among whom
lowing improvements in socioeconomic condi- mixed cellularity classical Hodgkin lymphoma
tions and transition toward a westernized life style may dominate (Clavel et al, 2006; Englund et al,
(Hjalgrim, 2012; Hjalgrim et al, 2008). 2015). The incidence of mixed cellularity classical
Hodgkin lymphoma generally increases with age,
Subtype-​Specific Incidence Patterns with a small peak among younger adults in afflu-
The variation in age-​specific incidence patterns ent populations (Fig. 26-​6), and is higher in males
between populations of different socioeconomic than in females at all ages (Howlader et al, 2014).

Males Females
100 100
Rate per 100,000 person-years

1980–89 1980–89
10 10
1960–69

1940–49 1940–49
1 1

1960–69
0.1 0.1
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Age Age

FIGURE 26-​5  Age-​specific incidence rates of Hodgkin lymphoma in different calendar periods in Connecticut, United
States.
Source: Hartge et al, 1994.
631

Hodgkin Lymphoma 631

3.5

Incidnece per 100,000 per year 3

2.5

1.5

0.5

0
00–09 yr 10–19 yr 20–29 yr 30–39 yr 40–49 yr 50–59 yr 60–69 yr 70–79 yr 80+yr
Age group

Nodular sclerosis Lymphocyte rich Mixed cellularity

Not otherwise specified Lymphocyte depleted

FIGURE 26-​6 Age-​specific incidence rates for classical Hodgkin lymphoma in both sexes combined by histologic sub-
type in the United States 2000–​2011; US SEER 18.
Source: Howlader et al, 2014, and Hjalgrim et al, personal communication.

In some (Cozen et al, 1992), but not all investi- diagnostic precision vis-​à-​vis non-​Hodgkin lym-
gations, (Clarke et  al, 2005)  mixed cellularity phomas, previously often mistaken for Hodgkin
classical Hodgkin lymphoma has been inversely lymphomas among the elderly (Glaser & Swartz,
associated with socioeconomic affluence. 1990; Jarrett et al, 2003). Whether this mechan-
The incidence of both the rarer lymphocyte-​ ism completely explains the decreasing incidence
rich and the lymphocyte-​ depleted classical among older adults or if part of it is true, is not
Hodgkin lymphoma subtypes increases with age known. Meanwhile, reduced diagnostic misclas-
(Fig. 26-​6) (Howlader et al, 2014). sification of Hodgkin and non-​ Hodgkin lym-
phomas cannot explain the observed incidence
Incidence Trends increase in the younger age groups, which is
During the last decades of the 20th century, likely to be real.
decreasing age-​standardized incidence rates of
Hodgkin lymphoma were reported in several set- GENETIC AND MOLECULAR
tings (Hjalgrim, 2012). However, age-​stratified EPIDEMIOLOGY
analyses revealed that in reality a decrease in Familial clustering of Hodgkin lymphoma was
incidence had occurred only among older adults, first demonstrated in the mid-​ 20th century
whereas in contrast the incidence had increased (Razis et al, 1959), and has since been confirmed
among adolescents and younger adults, notably in numerous investigations. The currently larg-
among females. With the caution that the repro- est of these is a register-​based investigation of
ducibility of classical Hodgkin lymphoma sub- more than 55,000 relatives of nearly 14,000
type diagnosis is far from perfect (Glaser et  al, Hodgkin lymphoma patients registered in the
2001; Jarrett et al, 2003), the increase in younger five Nordic countries’ population-​based cancer
adults seemed to have occurred for nodular scle- registries (Kharazmi et al, 2015). In this study,
rosis classical Hodgkin lymphoma, whereas the first-​
degree relatives of Hodgkin lymphoma
incidence of other histological subtypes of classi- patients were at 3.3-​fold (95% confidence inter-
cal Hodgkin lymphoma had either been stable or val [CI] 2.8–​ 3.9) increased risk of Hodgkin
decreased (Hjalgrim, 2012). lymphoma patients compared with the general
The decreasing Hodgkin lymphoma incidence population. Higher relative risks (RRs) were
among older adults has coincided with improved observed among patients’ siblings (sixfold) than
632

632 Textbook of Cancer Epidemiology

among their parents and offspring combined similarly been associated with an increased risk
(2.1-​fold), and risk also increased with number of EBV-​ negative Hodgkin lymphoma (Huang
of affected relatives. et al, 2012b; Johnson et al, 2015).
Speculations regarding the contribution of Other HLA alleles, for example, HLA, and
constitutional risk factors to the observed famil- -​B*15:01, have also been suggested to be asso-
ial accumulation have been put firmly to rest by ciated with EBV-​ negative classical Hodgkin
genetic analyses. Indeed, already by 1979 DNA lymphoma, albeit less compellingly (Moutsianas
analyses of more than 2,500 patients had estab- et al, 2011; Johnson et al, 2015).
lished that HLA serotypes A1, B5, B8, and B18 Little is known about genetic predisposi-
are associated with increased risk of Hodgkin tion to Hodgkin lymphoma in non-​Caucasian
lymphoma (Hors & Dausset, 1983). populations. Of note, an increased risk of EBV-​
Increasingly advanced genotyping techniques positive Hodgkin lymphoma and a decreased
combined with information on tumor character- risk of EBV-​negative Hodgkin lymphoma were
istics, in particular regarding tumor Epstein Barr proposed for HLA-​A*02:07 in a Chinese popula-
virus (EBV) status, have expanded the under- tion (Huang et al, 2012a).
standing of genetic susceptibility to Hodgkin Numerous genes outside the HLA region
lymphoma considerably (Diepstra et  al, 2005; have also been associated with Hodgkin lym-
Kushekhar et  al, 2014; Johnson et  al, 2015; Sud phoma risk. Most of these genes have been
et al, 2017; McAulay & Jarrett, 2015). explored in candidate gene studies, and asso-
Accordingly, while some HLA loci have ciations have only occasionally been independ-
been identified that are apparently associated ently confirmed (Kushekhar et  al, 2014; Sud
with the risk of both EBV-​positive and EBV-​ et al, 2017). Meanwhile, the growing number of
negative Hodgkin lymphomas, for example, genome-​wide association studies have reported
close to MICB and HLA-​DRA (Urayama et  al, associations with non-​HLA loci at, for example,
2012), EBV-​specific HLA associations have also 2p16.1 (REL), 3p24.1 (EOMES), 5q31 (IL13),
been reported. Analyses strongly suggest that 6p23.3 (HBS1L-​MYB), 8q24.21 (PVT1), 10p14
in Caucasian populations risk of EBV-​positive (GATA3), and 19p13.3 (TCF3) (Kushekhar
classical Hodgkin lymphoma is associated with et  al., 2014). It falls outside the scope of this
HLA class  I  alleles, specifically HLA-​A*01 car- review to discuss these associations in greater
rying an increased and HLA-​ A*02 carrying depth, but detailed reviews and meta-​analyses
a decreased risk for this Hodgkin lymphoma providing more information are available
variant (Diepstra et  al, 2005; Niens et  al, 2007; in the literature (Kushekhar et  al, 2014; Sud
Hjalgrim et al, 2010). Evidence for associations of et al, 2017).
EBV-​positive classical Hodgkin lymphoma with The association between EBV-​ positive and
other HLA types, for example, HLA-​B37 and  –​ EBV-​ negative Hodgkin lymphoma with HLA
DR10, is less strong (Huang et al, 2012b; Johnson class I and class II alleles/​loci, respectively, may
et al, 2015). indicate subtype-​ specific pathogenic mecha-
For EBV-​negative Hodgkin lymphoma, stud- nisms. Specifically, cytotoxic T lymphocytes may
ies have conversely pointed to associations with be involved in the development of EBV-​positive
variants within the HLA class II region on chro- classical Hodgkin lymphomas and helper
mosome 6 (Diepstra et al, 2005; Kushekhar et al, T lymphocytes conversely in the development
2014; McAulay & Jarrett, 2015; Sud et al, 2017). of classical EBV-​negative Hodgkin lymphomas
In particular, strong associations with rs6903608 (Kushekhar et al, 2014; McAulay & Jarrett, 2015).
have been reported both for EBV-​negative classi- Even more advanced models for constitutional
cal Hodgkin lymphoma (Enciso-​Mora et al, 2010; Hodgkin lymphoma risk, involving also the
Urayama et  al, 2012)  and for nodular sclerosis associated genes outside the HLA region, have
classical Hodgkin lymphoma in younger adults, been suggested by Kushekhar and colleagues
the vast majority of which would be EBV-​negative (Kushekhar et al, 2014).
(Cozen et al, 2012; Frampton et al, 2013). Hodgkin lymphoma also clusters with other
The rs6903608 is in strong linkage disequi- types of lymphomas within families (Chang et al,
librium with HLA-​DRB1*15:01, -​DQA1*01:02, 2005b; Wang et al, 2007; Goldin et al, 2009). The
and -​DQB1*06:02 (McAulay & Jarrett, 2015), and mechanisms underlying this clustering remain to
this may explain why these specific alleles have be identified.
63

Hodgkin Lymphoma 633

RISK FACTORS important risk modifier (Gutensohn & Cole,

The characterization of the archetypical age-​ 1977). Specifically, high and low infectious dis-
specific incidence patterns has been crucial for ease pressure in early childhood would be asso-
the formulation of epidemiological models for ciated with high Hodgkin lymphoma risk in
Hodgkin lymphoma and therefore for the current childhood and early adulthood, respectively
understanding of Hodgkin lymphoma epidemiol- (Abramson, 1974; Gutensohn & Cole, 1977).
ogy. The crux of these models is the suspicion that Indirect support for the late infection hypoth-
multiple epidemiologically and/​ or etiologically esis came from several epidemiological studies
heterogeneous variants of Hodgkin lymphoma of family structure and other characteristics of
exist. This notion was originally inspired by the childhood socioeconomic environment. Here,
bimodal age-​specific incidence pattern of Hodgkin proxies for high childhood infectious pressure,
lymphoma in affluent populations, also provid- for example large sibships, high birth order,
ing a first tool to distinguish between the differ- multifamily house dwelling, and/​or short mater-
ent variants. Thus, it was speculated that Hodgkin nal education were found to be associated with
lymphoma in children (<15  years), in younger reduced Hodgkin lymphoma risk in early adult-
adults (15–​34 years), and in older adults (50 years hood (Gutensohn & Cole, 1981; Bernard et  al,
and older) constitute three epidemiologically dis- 1987; Bonelli et al, 1990; Westergaard et al, 1997;
parate conditions (MacMahon, 1957, 1966). Chang et al, 2004a; Chatenoud et al, 2005).
The uneven age-​distribution of the histolog- These observations for Hodgkin lymphoma
ical Hodgkin lymphoma variants highlighted in younger adults have, however, been diffi-
by the Rye classification in 1966 (Lukes et  al, cult to reproduce in other and in particular
1966) almost immediately attracted interest as a more recent investigations in affluent popula-
biologically more tangible means to distinguish tions (Glaser et  al, 2002; Chang et  al, 2004c;
between suspected heterogeneous disease vari- Hjalgrim et al, 2007). A fascinating explanation
ants. Accordingly, it has been suggested that his- for this could be that secular phenomena, such
tological subtype defines, or at least approximates as increasing preschool attendance or decreas-
better than age does, different variants amenable ing family size, have altered the nuclear family’s
for research (Newell et al, 1970; MacMahon et al, role in childhood infectious disease exposure
1971; Cozen et al, 1992; Morton et al., 2007). (Glaser et al, 2002). Consistent with this notion,
Subsequently, it has been firmly established nursery school or day-​care attendance for more
that EBV is present in the malignant cells in a than a year was associated with a decreased risk
substantial proportion of classical Hodgkin lym- of classical Hodgkin lymphoma in younger/​
phoma cases, and that the virus likely contrib- middle-​aged adults (RR 0.64, 95% CI 0.45–​0.92)
utes to the development of these cases (Kuppers, in an American case-​control study (Chang et al,
2009; Farrell & Jarrett, 2011; Vockerodt et  al, 2004c). A  similar inverse association between
2015). As evidence of epidemiological differences kindergarten attendance and classical Hodgkin
between EBV-​positive and EBV-​negative classi- lymphoma risk among younger adults was also
cal Hodgkin lymphomas has accumulated, it has suggested by a Scandinavian case-​control inves-
been proposed that the two variants of classical tigation (RR 0.78, 95% CI 0.56–​1.09) (Hjalgrim
Hodgkin lymphoma are etiologically distinct et al, 2007).
(Jarrett et al, 1991; Hjalgrim, 2012), and genetic Another potential explanation for the inabil-
susceptibility (see previous section) and other ity to reproduce the earlier observations is that
risk factors have also been studied separately for the composition of Hodgkin lymphomas with
EBV-​positive and EBV-​negative disease. different etiologies among younger adults may
have changed over time. Thus, in the aforemen-
Sibship Size and Social Class tioned Scandinavian case-​control investigation,
Starting from the suspicion of etiologic heteroge- risk of EBV-​ positive classical Hodgkin lym-
neity, the variation in Hodgkin lymphoma inci- phoma among younger adults tended to decrease
dence by level of socioeconomic development with increasing number of both older (RR 0.77,
fostered the so-​ called late infection hypothe- 95% CI 0.56–​1.05) and younger (RR 0.58, 95% CI
sis. According to this, Hodgkin lymphomas in 0.39–​0.86) siblings, whereas no such associations
children and younger adults are caused by an were observed for EBV-​negative Hodgkin lym-
infectious agent, with age at exposure being an phoma (Hjalgrim et al, 2007).
634

634 Textbook of Cancer Epidemiology

It is inherent under the late infection hypoth- in particular TNFAIP3 (A20), and the expression
esis that risk factors for Hodgkin lymphoma of multiple receptor tyrosine kinase pathways
among children are the opposite of those for (Farrell & Jarrett, 2011).
Hodgkin lymphoma among younger adults. EBV is, as mentioned, not present in the
Consistent with this, Hodgkin lymphoma risk malignant cells in all Hodgkin lymphoma cases.
among children was found to increase with Rather, the prevalence of EBV-​positive Hodgkin
number of siblings and birth order in a Danish lymphomas varies by certain demographic fac-
register-​based study (Westergaard et  al, 1997). tors and tumor characteristics (Glaser et al, 1997;
Likewise, the so far largest case-​control study of Lee et al, 2014). A summary of data reported for
Hodgkin lymphoma in childhood reported a pos- more than 13,000 patients showed that EBV-​
itive association with sibship size and an inverse positive Hodgkin lymphoma prevalence differ by
association with maternal educational attain- sex (men 44%; women 27%); age (younger than
ment and household income at birth of the child 15 years 70%; older 41%); classical Hodgkin lym-
(Linabery et  al, 2014). An inverse relationship phoma histology (nodular sclerosis 29%; mixed
with parental education was also suggested in US cellularity 66%; lymphocyte-​ depleted 52%;
register data (Carozza et al, 2010). Various family lymphocyte-​ rich 47%), and geographic locale
structure characteristics, however, were not asso- (Europe 36%; North America 32%; Africa 74%;
ciated with Hodgkin lymphoma risk before the Central and South America 61%; Asia 56%) (Lee
age of 15 years in either a large Swedish (Chang et al, 2014).
et  al, 2004a) or in an American registry-​based Though these demographic (age, sex, and geo-
investigation (von Behren et  al, 2011). Again, graphic setting) and tumor characteristics (histo-
the absence of associations may reflect the limi- logical variants) are mutually related, the observed
tations of family structure as a measure of infec- distribution of EBV-​positive Hodgkin lymphoma
tious disease pressure. mirrors established determinants of EBV-​posi-
tive Hodgkin lymphoma. Specifically, in mutually
Epstein Barr Virus Infection adjusted analyses, age, sex, ethnicity, histolog-
Among microorganisms studied, the evidence ical subtype, and socioeconomic background
for a connection with Hodgkin lymphoma is the are all associated with EBV-​ positive Hodgkin
strongest against the ubiquitous EBV (Evans & lymphoma prevalence (Tables 26-​ 1, 26-​ 2,)
Niederman, 1989)  to the point that it has been (Glaser et al, 1997; Glaser et al, 2008).
classified as causally associated with Hodgkin A causal association between EBV infec-
lymphoma by the International Agency for tion and Hodgkin lymphoma is also supported
Research on Cancer (IARC, 1997). EBV has been by serological findings. While EBV infection
implicated in Hodgkin lymphoma etiology by prevalence does not differ markedly between
molecular-​ biological, serological, and conven- Hodgkin lymphoma patients and healthy peers
tional epidemiological findings. Most compelling (IARC, 1997), the patients stand out with
of these is the presence of EBV in the malignant respect to the profile of antibodies. Here, focus
cells in a subset of cases (Poppema et  al, 1985; has centered on the balance between antibod-
Weiss et al, 1987; Weiss et al, 1989; Uccini et al, ies against two EBV antigens, EBNA1 and
1989; Wu et  al, 1990; Pallesen et  al, 1991). The EBNA2. During normal resolution of acute
EBV infection is latent, and the virus expresses (primary) EBV infection, the ratio of anti-
three antigens in the malignant cells, namely, bodies against EBNA1 and EBNA2 increases
EBV nuclear antigen (EBNA) 1, latent membrane from less to above unity. However, in a series
protein (LMP) 1, and LMP 2A. Each of these of investigations, Mueller and colleagues dem-
viral antigens has the capacity to interfere with onstrated that EBNA1:EBNA2 ratios ≤ 1 are
normal cell functions in ways that could conceiv- typical for EBV-​positive Hodgkin lymphoma
ably contribute to the malignant transformation at and—​ importantly—​ before diagnosis, con-
(Kuppers, 2009; Vockerodt et al, 2015). sistent with an abnormal immune response
Further arguing that the presence of EBV in toward EBV (Chang et  al, 2004b; Levin et  al,
the malignant cells is not coincidental, molecular-​ 2012). Similarly, both before and at diagnosis,
biological differences between EBV-​positive and cell-​free EBV DNA can more often be dem-
EBV-​ negative classical Hodgkin lymphomas onstrated in serum from patients with EBV-​
have been demonstrated, for instance regarding positive Hodgkin than in controls (Gallagher
mutations of genes encoding inhibitors of NF-​κB, et al, 1999; Yang et al, 2002).
635

Hodgkin Lymphoma 635

TABLE 26-​1   ODDS RATIOS (OR) AND 95% CONFIDENCE INTERVALS (CI) FOR
EBV-POSITIVE HODGKIN LYMPHOMA ASSOCIATED WITH HISTOLOGIC
SUBT YPE, REGIONAL ECONOMIC DEVELOPMENT AND SEX ADJUSTED FOR
COUNTRY, CLINICAL STAGE AND COMPONENT CASE SERIES

Variable Comparison Age stratum (years) Adjusted OR 95% CI

Histologic subtype MC vs. NS1 0–​14 7.3 3.8–​14.2

15–​49 13.4 9.0–​19.9
50+ 4.9 2.8–​8.7
Regional economic Less vs. more1 0–​14 6.0 2.0–​18.0
development
15–​49 0.9 0.4–​2.3
50+ 0.8 0.2–​3.0
Sex Female vs. male1 0–​14 0.6 0.3–​1.1
15–​49 0.4 0.3–​0.6
50+ 1.2 0.7–​2.0

MC mixed cellularity; NS nodular sclerosis. 1 represents reference group.

Finally, numerous investigations have shown 2000; Alexander et al, 2003; Hjalgrim et al, 2003;
an increased risk of Hodgkin lymphoma after Hjalgrim et  al, 2007). If causal in nature, mod-
infectious mononucleosis, a common manifes- eling of data in two Scandinavian investigations
tation of primary EBV infection in adolescence suggests that infectious mononucleosis-​induced
(Hjalgrim, 2012). Although findings are equivo- EBV-​ positive classical Hodgkin lymphoma
cal, some investigations find that the risk increase has an incubation period on the order of 3 to
particularly or even specifically concerns EBV-​ 4 years (Fig. 26-​7) (Hjalgrim et al, 2003; Hjalgrim
positive Hodgkin lymphoma (Alexander et  al, et al, 2007).

TABLE 26-​2   ODDS RATIOS (OR) AND 95% CONFIDENCE INTERVALS (CI) FOR EBV-
POSITIVE HODGKIN LYMPHOMA ASSOCIATED WITH AGE ADJUSTED FOR COUNTRY,
CLINICAL STAGE AND COMPONENT CASE SERIES

Variable Strata Males Females

Comparison Regional Histologic Adjusted OR 95% CI Adjusted OR 95% CI

development subtype

Age 0–​14 vs 15–​491 years Less developed NS 10.0 2.6–​39.7 14.3 3.4–​59.9
MC 5.5 1.3–​23.3 7.9 1.7–​36.0
More developed NS 1.6 0.8–​3.0 2.2 1.0–​4.7
MC 0.9 0.4–​1.7 1.2 0.5–​2.7
50 vs. 15–​491 years Less developed NS 1.2 0.4–​3.9 3.6 1.1–​11.3
MC 0.4 0.1–​1.4 1.3 0.4–​4.3
More developed NS 1.3 0.7–​2.4 3.9 2.2–​7.1
MC 0.5 0.3–​0.8 1.4 0.8–​2.8

MC mixed cellularity; NS nodular sclerosis. 1 represents reference group.

63

636 Textbook of Cancer Epidemiology

20

Relative risk 15

10

0
0 5 10 15 20
Time (years) since mononucleosis

Overall EBV pos EBV neg

FIGURE 26-​7  Temporal variation in relative risk of classical Hodgkin lymphoma overall and by Epstein Barr virus
(EBV) status by time since infectious mononucleosis (IM) predicted by statistical modeling based on the age-​, gender-​,
and country-​specific distribution of IM among population controls in the SCALE study.
Source: Hjalgrim et al, 2007.

The evidence linking EBV infection with EBV-​ one in older adults. The peaks in childhood and
positive Hodgkin lymphoma supports the afore- in early adulthood both represent Hodgkin lym-
mentioned hypothesis that Hodgkin lymphomas phomas arising following primary EBV infection,
have different epidemiologies/​etiologies in child- while the peak in older adults represents Hodgkin
ren, young adults, and older adults (MacMahon, lymphomas arising as a results of loss of control of
1957, 1966). Specifically, a four-​ disease model, EBV infection, for example, because of comorbid-
illustrated in Figure 26-​8, has been suggested ity or immune senescence. The fourth component
(Armstrong et al, 1998; Jarrett, 2002). This model in the four-​disease model comprises EBV-​negative
suggests that the age-​specific incidence of EBV-​ classical Hodgkin lymphoma. This entity has a
positive classical Hodgkin lymphoma has three unimodal age-​ distribution that peaks among
peaks, one in children, one in younger adults, and younger adults.
Incidence

Incidence

0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Age group (years) Age group (years)

FIGURE  26-​8  Modeled age-​specific incidence rates of EBV-​positive classical Hodgkin lymphoma (left) and EBV-​
negative classical Hodgkin lymphoma (right).
Source: Hjalgrim, 2012.
637

Hodgkin Lymphoma 637

Infections Other Than EBV encephalitis (RR 7.88, 95% CI 1.97–​31.5), and

The association between Hodgkin lymphoma all gastrointestinal infections combined (OR
risk and childhood infectious disease exposure 1.16, 95% CI 1.00–​1.35). Further analyses sug-
suggested by the late infection model has been gested that associations with sinusitis, tuber-
explored in several case-​control investigations, culosis, and herpes zoster existed for up to
unfortunately without bringing clarity to the 10 years before Hodgkin lymphoma diagnosis.
field. Concerning risk of Hodgkin lymphoma in Theoretically, these associations could reflect a
children, one case-​control study reported paren- mechanism by which infections contribute to
tal recall of any infection to be associated with Hodgkin lymphoma pathogenesis indirectly,
an elevated risk (RR 1.69, 95% CI 0.98–​2.91) for example, through chronic inflammation.
(Linabery et  al, 2014), while in another case-​ However, as also speculated by the authors,
control investigation maternal report of recurrent the aforementioned immune dysfunction pre-
early common infections conversely was asso- ceding Hodgkin lymphoma diagnosis seems a
ciated with decreased Hodgkin lymphoma risk more likely explanation for the observations
(RR 0.5, 95% CI 0.3–​0.9) (Rudant et al, 2011). (Kristinsson et al, 2015).
Self-​reported histories of one or more com-
mon childhood infections have been associated Tobacco
with reduced risk of Hodgkin lymphoma in early Meta-​ analyses of 14 individual investigations
adulthood in several studies (Alexander et  al, including more than 3,000 Hodgkin lymphoma
2000; Glaser et  al, 2005; Montella et  al, 2006; patients have shown that current cigarette smok-
Monnereau et  al, 2007). Even though associa- ers have a 30%–​40% increased risk of Hodgkin
tions like these are compatible with the late infec- lymphoma compared with never smokers
tion model and theoretically could reflect early (Castillo et  al, 2011; Sergentanis et  al, 2013).
modulation of the immune system, other meth- Stratified analyses indicate that the association
odological explanations, such as selection/​partic- applies to both nodular sclerosis (35% increased
ipation/​recall bias, seem more likely. risk) and mixed cellularity classical Hodgkin lym-
Outside the realm of the late infection model, phoma (153% increased risk) (Sergentanis et al,
there is not any solid evidence linking infec- 2013), and to both EBV-​positive (126% increased
tions (other than EBV) to the risk of Hodgkin risk) and EBV-​negative classical Hodgkin lym-
lymphoma, whether EBV-​negative or Hodgkin phoma (40% increased risk) (Castillo et  al,
lymphoma overall. For studies assessing the risk 2011). Contrary to the meta-​analyses using never
associated with infections occurring at shorter smokers as reference, a pooled analysis provided
intervals before Hodgkin lymphoma diagnosis, no evidence of dose-​ dependence in analyses
the role of lymphoma-​related immune impair- restricted to smokers only (Kamper-​Jorgensen
ment that may become increasingly manifest as et al, 2013).
the lymphoma evolves (Newton et al, 2007) must
be considered. Such a mechanism may explain Diet
elevated Hodgkin lymphoma risk reported for There is no consistent evidence of an association
herpes zoster (Serraino et  al, 1991; Kristinsson between Hodgkin lymphoma and specific dietary
et  al, 2015)  and warts in some investigations exposures, with the possible exception of alcohol
(Serraino et al, 1991; Becker et al, 2005). intake (see next section).
Kristinsson and colleagues compared regis-
trations of infectious and inflammatory diseases Alcohol
in the population-​based Swedish inpatient reg- Both case-​control (Bernard et  al, 1987; Besson
ister for 7,414 HL patients and 29,240 matched et  al, 2006; Nieters et  al, 2006; Gorini et  al,
controls (Kristinsson et  al, 2015). When the 2007; Lim et  al, 2007; Monnereau et  al, 2008;
year preceding diagnosis was disregarded, Kanda et al, 2010) and cohort studies (Lim et al,
history of hospitalization for any infection 2007; Kanda et al, 2010) have relatively unani-
was associated with a statistically significant mously suggested inverse associations between
11% increased risk of Hodgkin lymphoma. In alcohol intake and Hodgkin lymphoma risk,
detailed analyses, significantly increased risks albeit estimates have not always been statisti-
were observed with individual conditions cally significant. With some variation between
such as sinusitis (RR 1.81, 95% CI 1.06–​3.07), the studies, reduced risk estimates have been
tuberculosis (RR 1.76, 95% CI 1.01–​ 3.07), reported for all groups of Hodgkin lymphoma,
638

638 Textbook of Cancer Epidemiology

that is, whether stratified by age, EBV-​status, or Studies of the possible association between
histology, the statistical power of such strati- Hodgkin lymphoma risk and body mass index
fied analyses for obvious reasons being limited (BMI) are equally ambiguous. Here, compared
(Bernard et al, 1987; Besson et al, 2006; Nieters with normal weight individuals, elevated risk
et  al, 2006; Gorini et  al, 2007; Lim et  al, 2007; of Hodgkin lymphoma (RR 1.41, 95% CI 1.14–​
Monnereau et  al; 2008, Klatsky et  al, 2009). 1.75) was observed among obese (≥30  kg/​m2),
One study suggested that the reduced risk was but not among overweight individuals in a meta-​
restricted to nonsmokers (Gorini et  al, 2007). analysis of five prospective studies (Larsson &
Because Hodgkin lymphoma may be accompa- Wolk, 2011). The UK Million Women study
nied by alcohol intolerance (Stein & Morgan, found that Hodgkin lymphoma risk increased
2004)  reverse causality must realistically be by 64% for each BMI increase of 10  kg/​ m2
considered when interpreting the observed (Murphy et al, 2013). In contrast, the results of
association. case-​control investigations are somewhat incon-
sistent, with no association between BMI or
Reproductive Factors and Hormones obesity and Hodgkin lymphoma risk reported
The lower incidence among females after their in some studies (Bosetti et al, 2005; Chang et al,
childbearing years (Fig. 26-​ 3) suggests an 2005a; Keegan et  al, 2006; Kanda et  al, 2010),
influence of hormones or reproductive factors while others have reported an increased risk
on the development of Hodgkin lymphoma among obese men, but not among obese women
(Glaser, 1994). However, to date there is no (Willett & Roman, 2006), and opposite asso-
convincing epidemiologic evidence that repro- ciations between BMI and Hodgkin lymphoma
ductive factors influence the risk of Hodgkin risk among younger adults (positive) and older
lymphoma. adults (inverse), respectively (Li et al, 2013).

Anthropometric Measures Physical Activity

Numerous investigations have examined the asso- Three studies have evaluated the possible influ-
ciation between Hodgkin lymphoma and anthro- ence of physical activity on the risk of Hodgkin
pometric measures at different ages, and results lymphoma (Keegan et  al, 2006; Lim et  al, 2007;
have been conflicting. This is true for Hodgkin Parent et  al, 2011). In a meta-​analysis, no evi-
lymphoma among children, where a positive dence of a significant association between the
association with birthweight was reported in one two was observed (RR 0.82, 95% CI 0.47–​1.42)
registry-​based investigation from Sweden (Crump (Vermaete et al, 2013).
et al, 2012), but not in a meta-​analysis of four other
investigations (Papadopoulou et al, 2012). Occupation
More intriguingly, studies have suggested There is no convincing evidence that occupa-
positive associations of Hodgkin lymphoma tion and/​or occupational exposures play any
risk in younger adults with birthweight (Isager major role in the etiology of Hodgkin lymphoma
& Andersen, 1978; Keegan et  al, 2006; Crump (Charbotel et al, 2014).
et  al, 2012)  and with stature in late childhood
and early adolescence (Isager & Andersen, 1978; Ionizing Radiation
Keegan et al, 2006). Reasonably associated with Exposure to therapeutic and diagnostic radiation
both high birthweight and growth in early life, does not appear to increase the risk of Hodgkin
childhood socioeconomic affluence could be lymphoma. Nor do results of the studies of atomic
an underlying common denominator for their bomb survivors support an association with ion-
association with Hodgkin lymphoma risk. izing radiation (Ron, 1998).
Meanwhile, such a common relationship seem-
ingly does not extend to adult stature, which has Ultraviolet Radiation
been associated with both increased (Hancock A role for ultraviolet (UV) light in Hodgkin lym-
et  al, 1976; Gutensohn & Cole, 1981; Keegan phoma has been suggested by associations with
et al, 2006; Murphy et al, 2013) and inconspicu- history of sunburn (RR 0.77, 95% CI 0.63–​0.95)
ous Hodgkin lymphoma risk (Paffenbarger et al, and use of sun lamps (RR 0.81, 95% CI 0.69–​
1977; La Vecchia et al, 1990; Willett & Roman, 0.96) observed in an analysis of 1,320 patients
2006; Lim et al, 2007). and 6,381 controls (Monnereau et  al, 2013).
639

Hodgkin Lymphoma 639

The association was most pronounced for EBV-​ with a variety of autoimmune diseases. In a
positive Hodgkin lymphoma, the risk of which large Swedish registry-​based case-​control study
decreased dose-​ dependently with cumulative of 9,314 Hodgkin lymphoma patients and
lifetime exposure to UV light. Consistent with 37,069 matched controls, statistically signifi-
the speculation that UV radiation influences cantly increased risk of Hodgkin lymphoma was
Hodgkin lymphoma risk, surveys have shown observed among patients with rheumatoid arthri-
that Hodgkin lymphoma occurrence is higher tis (RR 2.2, 95% CI 1.6–​2.8), systemic lupus ery-
in areas with lower than in areas with higher thematosus (RR 5.3, 95% CI 2.5–​11.2), Sjögren’s
levels of ambient UV radiation (van Leeuwen syndrome (RR 4.0, 95% CI 1.3–​12.3), sarcoidosis
et al, 2013; Bowen et al, 2016). (RR 3.7, 95% CI 1.9–​7.4), and idiopathic throm-
Whether the effect of UV light on Hodgkin bocytopenic purpura (RR 15.9, 95% CI 1.8–​142)
lymphoma risk is mediated through regulation of (Kristinsson et  al, 2009). Subanalyses indicated
vitamin D levels remains to be seen. that the association was the strongest for mixed
cellularity classical Hodgkin lymphoma, sug-
Medical Conditions and Treatment gesting that it could be specific to EBV-​positive
Immune Deficiencies Hodgkin lymphoma (Kristinsson et al, 2009).
Both primary and acquired immune deficien- The association between the autoim-
cies have been associated with increased risk mune diseases and Hodgkin lymphoma is not
of Hodgkin lymphoma (Filipovich et al, 1992; fully understood, but may involve persistent
Filipovich et al, 1994; Mueller, 1999; Grulich immune stimulation, autoimmune disease treat-
et al, 2007; Vajdic et al, 2012; Grulich & Vajdic, ment, shared environmental risks, and shared
2015). Most often, Hodgkin lymphomas occur- genetic predisposition (Baecklund et  al, 2014).
ring in immune-​deficient patients are EBV-​pos- Concerning the latter, studies have occasionally
itive (Raphael et al, 2008; Swerdlow et al, 2008; reported increased risk of Hodgkin lymphoma
Van Krieken et al, 200; Gaulard et al, 2008). In among children and/​or other first-​degree rela-
agreement with this, markedly increased risk of tives of patients with, for example, rheumatoid
the often EBV-​positive variants mixed cellularity arthritis (Ekstrom et  al, 2003), sarcoidosis and
(RR 18.3, 95% CI 15.9–​20.9) and lymphocyte-​ ulcerative colitis (Landgren et al, 2006), and mul-
depleted (RR 35.3, 95% CI 24.7–​48.8) classical tiple sclerosis (Hjalgrim et  al, 2004; Landgren
Hodgkin lymphoma was observed in a large fol- et al, 2005). Consistent with this, a recent meta-​
low-​up study of Americans with AIDS, whereas analysis of genome-​ wide association studies
the risk of the less commonly EBV-​positive var- demonstrated overlapping genetic risks between
iant ​ nodular sclerosis classical Hodgkin lym- Hodgkin lymphoma and multiple sclerosis and
phoma was less markedly elevated (RR 4.7, 95% also showed that Hodgkin lymphoma genet-
CI 3.9–​5.5) (Frisch et al, 2001). ically resembles autoimmune diseases more
The mechanism underlying the association than it does other types of cancer (Khankhanian
between immune impairment and Hodgkin lym- et al, 2016).
phoma presumably is that EBV-​infected cells are
allowed to proliferate and undergo malignant Allergy
transformation because of immunological loss Observations of elevated IgE levels and impaired
of control of latent EBV infection combined with Th-​1 immune responses in Hodgkin lymphoma
chronic immune stimulation (Grulich & Vajdic, patients have raised the suspicion of associations
2015). Interestingly, one study has suggested that, with allergies and atopic conditions (Martinez-​
in HIV patients, risk of Hodgkin lymphoma does Maza et al, 2010). However, there is little evidence
not increase linearly with decreasing level of CD4 to support such a relationship in the literature
cells, but is higher for CD4 cell levels of 150–​199 (Martinez-​Maza et al, 2010).
cell/​microL than for those with fewer than 50 cells/​
microL, perhaps suggesting a role of host immunity Medications
in lymphoma development (Biggar et al., 2006). No medications have been strongly or consist-
ently linked to Hodgkin lymphoma risk. Use
Autoimmune Diseases of selective Cox-​2 inhibitors or acetaminophen
Several investigations have reported an increased have been associated with an increased risk of
risk of Hodgkin lymphoma among patients Hodgkin lymphoma; however because the risk
640

640 Textbook of Cancer Epidemiology

was the highest shortly before Hodgkin lym- lymphomas define two such variants with sepa-
phoma diagnosis, reverse causality or confound- rate etiologies. Despite being the less common
ing by indication cannot be excluded (Chang et al, of the two, the natural history of EBV-​positive
2011). A  decreased risk of Hodgkin lymphoma Hodgkin lymphoma is currently the best under-
among long-​ term aspirin users could reflect stood, both with respect to how EBV may con-
aspirin’s inhibitory effect on the NF-​kB pathway, tribute to malignant cell transformation and in
which is constitutively active in Hodgkin lym- relation to constitutional and environmental risk
phoma (Chang et al, 2011). factors. Meanwhile, the understanding of the nat-
ural history of EBV-​negative Hodgkin lymphoma
CONCLUSIONS is meager, the most important elements being a
The descriptive epidemiology of Hodgkin lym- limited understanding of the molecular-​biological
phoma has demonstrated marked variation by mechanisms that contribute to the malignant cell
age, sex, and social class, and over time, strongly transformation and the identification of a small
suggesting that environmental factors contribute number of constitutional risk factors.
to its etiology and that etiologically diverse sub- In future research into the epidemiology of
types of Hodgkin lymphoma exist. classical Hodgkin lymphoma, the distinction
A summary of the current state of knowledge between EBV-​positive and EBV-​negative variants
for Hodgkin lymphoma risk factors is presented seems to be an absolute requirement. Whereas
in Table 26-​3. There is increasing evidence that for EBV-​positive Hodgkin lymphoma efforts will
EBV-​positive and EBV-​negative classical Hodgkin concentrate on understanding the interaction

TABLE 26-​3   EPIDEMIOLOGY OF HODGKIN LYMPHOMA: RISK FACTOR SUMMARY

Risk factor Direction of effect*

Well-​Confirmed Risk Factors

Family history in first-​degree relative ↑↑
EBV infection ↑↑ (EBV-​positive variant)
Immune dysfunction ↑↑ (EBV-​positive variant)
Infectious mononucleosis ↑↑ (EBV-​positive variant)

Probable Relationship Exists, Based on Substantial Data

Current cigarette smoking ↑
Autoimmune disease history ↑
Socioeconomic deprivation in childhood ↑↓ (children and younger adults)

Weak, if Any, Relationship Exists, Based on Substantial Data

Dietary factors —​
Physical activity —​
Radiation —​
Occupation —​
Stature —​

Inconsistent Findings or Limited Study to Date

Ultraviolet light exposure ↓
Obesity ↑
Day-​care attendance ↓ (younger adults)
Long term aspirin use ↓

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate to large increase
in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association
Note: The magnitude of the risk can vary substantially, depending on the exact comparison being made.
641

Hodgkin Lymphoma 641

between host, environment, and virus, for EBV-​ disease:  case control epidemiological study in
negative Hodgkin lymphoma focus will be on the Yorkshire. Br J Cancer 1987;55:85–​90.
presumed infectious etiology, for which there are Besson H, Brennan P, Becker N, De SS, Nieters A, Font
currently no strong candidates. The continued R, Maynadie E et  al. Tobacco smoking, alcohol
attenuation of previously established risk factors drinking and Hodgkin’s lymphoma:  a European
for classical Hodgkin lymphoma among younger multi-​centre case-​control study (Epilymph). Br J
adults will be challenging for this endeavor. Cancer 2006;95:378–​84.
Accordingly, the rapid development of tech- Biggar RJ, Jaffe ES, Goedert JJ, Chaturvedi A, Pfeiffer
R, Engels EA. Hodgkin lymphoma and immu-
niques for single cell genetic sequencing appli-
nodeficiency in persons with HIV/​AIDS. Blood
cable to the rare Reed-​Sternberg cells may hold
2006;108:3786–​91.
greater promise in this regard.
Bonelli L, Vitale V, Bistolfi F, Landucci M, Bruzzi P.
Hodgkin’s disease in adults:  association with
REFERENCES social factors and age at tonsillectomy:  a case-​
Abramson JH. Childhood experience and Hodgkin’s control study. Int J Cancer 1990;45:423–​7.
disease in adults:  an interpretation of incidence Bosetti C, Dal ML, Negri E, Talamini R, Montella M,
data. Israeli J Med Sci 1974;10:1365–​70. Franceschi S, La VC. Re: body mass index and risk
Alexander FE, Jarrett RF, Lawrence D, Armstrong of malignant lymphoma in Scandinavian men and
AA, Freeland J, Fokhale DA et al. Risk factors for women. J Natl Cancer Inst 2005;97;860–​1.
hodgkin’s disease by Epstein-​Barr virus (EBV) sta- Bowen EM, Pfeiffer RM, Linet MS, Liu WT,
tus: prior infection by EBV and other agents. Br J Weisenburger DD, Freedman DM, Cahoon EK.
Cancer 2000;82;1117–​21. Relationship between ambient ultraviolet radi-
Alexander FE, Lawrence DJ, Freeland J, Krajewski AS, ation and Hodgkin lymphoma subtypes in the
Angus B, Taylor G, et al. An epidemiologic study United States. Br J Cancer 2016;114;826–​31.
of index and family infectious mononucleosis Carozza SE, Puumala SE, Chow EJ, Fox EE, Horel S,
and adult Hodgkin’s disease (HD): evidence for a Johnson KJ, et al. Parental educational attainment
specific association with EBV + VE HD in young as an indicator of socioeconomic status and risk of
adults. Int J Cancer 2003;107:298–​302. childhood cancers. Br J Cancer 2010;103:136–​42.
Anagnostopoulos I, Isaacson PG, Stein H. Castillo JJ, Dalia S, Shum H. Meta-​analysis of the
Lymphocyte-​ rich classical Hodgkin lymphoma. association between cigarette smoking and inci-
In: Swerdlow SH, Campo E, Harris NL, Jaffe ES, dence of Hodgkin’s lymphoma. J Clin Oncol
Pileri SA, Stein H et al (eds.), Tumours of haema- 2011;29:3900–​06.
topoietic and lymphoid tissues. 3 ed., pp. 332–​3. Chang ET, Froslev T, Sorensen HT, Pedersen L. A
Lyon: IARC, 2008. nationwide study of aspirin, other non-​steroidal
Armitage JO. Early-​ stage Hodgkin’s lymphoma. N anti-​inflammatory drugs, and Hodgkin lymphoma
Engl J Med 2010; 363;653–​62. risk in Denmark. Br J Cancer 2011;105:1776–​82.
Armstrong AA, Alexander FE, Cartwright RA, Angus B, Chang ET, Hjalgrim H, Smedby KE, Akerman M, Tani
Krajewski AS, Wright DH. et al. Epstein-​Barr virus E, Johnsen HE et al. Body mass index and risk of
and Hodgkin’s disease:  further evidence for the malignant lymphoma in Scandinavian men and
three disease hypothesis. Leukemia 1998;12:1272–​6. women. J Natl Cancer Inst 2005a;97:210–​8.
Baecklund E, Smedby KE, Sutton IA, Askling J, Chang ET, Montgomery SM, Richiardi L, Ehlin A,
Rosenquist R. Lymphoma development in Ekbom A, Lambe M. Number of siblings and
patients with autoimmune and inflammatory dis- risk of Hodgkin’s lymphoma. Cancer Epidemiol
orders:  what are the driving forces? Sem Cancer Biomarkers Prev 2004a;13:1236–​43.
Biolog 2014;24:61–​70. Chang ET, Smedby KE, Hjalgrim H, Porwit-​
Becker N, Deeg E, Rudiger T, and Nieters A. Medical Macdonald A, Roos G, Glimelius, B et al. Family
history and risk for lymphoma:  results of a history of hematopoietic malignancy and risk of
population-​based case-​control study in Germany. lymphoma. J Natl Cancer Inst 2005b;97:1466–​74.
Eur J Cancer 2005;41:133–​42. Chang ET, Zheng T, Lennette ET, Weir EG, Borowitz
Benharroch D, Stein H, Peh SC. Lymphocyte-​depleted M, Mann RB et  al. 2004b. Heterogeneity of risk
classical Hodgkin lymphoma. In:  Swerdlow SH, factors and antibody profiles in Epstein-​Barr virus
Campo E, Harris NL, Jaffe ES, Pileri SA, Stein genome-​ positive and -​ negative Hodgkin lym-
H et  al (eds.), Tumours of Haematopoietic and phoma. J Infect Dis 2004b;189:2271–​81.
Lymphoid Tissues. 3 ed., p. 334. Lyon: IARC, 2008. Chang ET, Zheng T, Weir EG, Borowitz M, Mann
Bernard SM, Cartwright RA, Darwin CM, Richards RB, Spiegelman D et  al. Childhood social
ID, Roberts B, O’Brien C, Bird CC. Hodgkin’s environment and Hodgkin’s lymphoma:  new
642

642 Textbook of Cancer Epidemiology

findings from a population-​based case-​control Evans AS, Niederman JC. Epstein-​Barr virus. In: Evans
study. Cancer Epidemiol Biomarkers Prev AS (ed.) Viral Infections of Humans: Epidemiology
2004c;13:1361–​70. and Control. New  York:  Plenum Medical Book
Charbotel B, Fervers B, Droz JP. Occupational expo- Company, 1989.
sures in rare cancers:  a critical review of the lit- Farrell K, Jarrett RF. 2011. The molecular patho-
erature. Crit Rev Oncol Hematol 2014;90:99–​134. genesis of Hodgkin lymphoma. Histopathology
Chatenoud L, Gallus S, Altieri A, Negri E, Talamini 2011;58:15–​25.
R, Franceschi S, La Vecchia C. Number of sib- Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser
lings and risk of Hodgkin’s and other lymphoid S, Mathers C, Rebelo M, Parkin DM, Forman D,
neoplasms. Cancer Epidemiol Biomarkers Prev Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence
2005;14:552. and Mortality Worldwide: IARC CancerBase No.
Clarke CA, Glaser SL, Keegan TH, Stroup, A. 11 [Internet]. Lyon, France: International Agency
Neighborhood socioeconomic status and Hodgkin’s for Research on Cancer; 2013. Available from:
lymphoma incidence in California. Cancer http://globocan.iarc.fr, accessed 2017.
Epidemiol Biomarkers Prev 2005;14:1441–​7. Filipovich AH, Mathur A, Kamat D, Kersey JH,
Clavel J, Steliarova-​Foucher E, Berger C, Danon S, Shapiro RS. Lymphoproliferative disorders and
Valerianova Z. Hodgkin’s disease incidence and other tumors complicating immunodeficiencies.
survival in European children and adolescents Immunodeficiency 1994;5:91–​112.
(1978–​ 1997):  report from the automated can- Filipovich AH, Mathur A, Kamat D, Shapiro RS.
cer information system project. Eur J Cancer Primary immunodeficiencies: genetic risk factors
2006;42:2037–​49. for lymphoma. Cancer Res 1992;52:5465s–​67s.
Correa P, O’Conor GT. Epidemiologic patterns of Forman D, Brewster DH, Gombe-​Mbalawa C, Kohler
Hodgkin’s disease. Int J Cancer 1971;8:192–​201. B, Pineros M, Steliarova-​Foucher E et al. Cancer
Cozen W, Katz J, Mack TM. Risk patterns of in five continents [online]. Lyon:  International
Hodgkin’s disease in Los Angeles vary by Agency for Research on Cancer, 2013. http://​ci5.
cell type. Cancer Epidemiol Biomarkers Prev iarc.fr/​Default.aspx (accessed June 1, 2015).
1992;1:261–​8. Frampton M, Da Silva Filho MI, Broderick P, Thomsen
Cozen W, Li D, Best T, Van Den Berg DJ, Gourraud PA, H, Forsti A et  al. 2013. Variation at 3p24.1 and
Cortessis VK et al. A genome-​wide meta-​analysis 6q23.3 influences the risk of Hodgkin’s lym-
of nodular sclerosing Hodgkin lymphoma identi- phoma. Nature Com 2013;4:2549.
fies risk loci at 6p21.32. Blood 2012;119:469–​75. Frisch M, Biggar RJ, Engels EA, Goedert JJ. Association
Crump C, Sundquist K, Sieh W, Winkleby MA, of cancer with AIDS-​related immunosuppression
Sundquist J. Perinatal and family risk factors for in adults. JAMA 2001;285;736–​45.
Hodgkin lymphoma in childhood through young Gallagher A, Armstrong AA, Mackenzie J, Shield L,
adulthood. Am J Epidemiol 2012;176:1147–​58. Khan G, Lake A et al. 1999. Detection of Epstein-​
Diepstra A, Niens M, Vellenga E, Van Imhoff GW, Barr virus (EBV) genomes in the serum of
Nolte IM, Schaapveld M et  al. Association with patients with EBV-​associated Hodgkin’s disease.
HLA class  I  in Epstein-​Barr-​virus-​positive and Int J Cancer 1999;84:442–​8.
with HLA class III in Epstein-​Barr-​virus-​negative Gaulard P, Swerdlow SH, Harris NL, Jaffe ES, Sundström,
Hodgkin’s lymphoma. Lancet 2005;365:2216–​24. C. Other iatrogenic immunodeficiency-​associated
Ekstrom K, Hjalgrim H, Brandt L, Baecklund E, disorders. In: Swerdlow SH, Campo E, Harris, NL,
Klareskog L, Ekbom A, Askling J. Risk of malig- Jaffe ES, Pileri SA, Stein H et al (eds.), Tumours of
nant lymphomas in patients with rheuma- Haematopoietic and Lymphoid Tissues. 3 ed., pp.
toid arthritis and in their first-​degree relatives. 350–​1. Lyon: IARC, 2008.
Arthritis Rheum 2003;48:963–​70. Glaser SL. Reproductive factors in Hodgkin’s dis-
Enciso-​Mora V, Broderick P, Ma Y, Jarrett RF, Hjalgrim ease in women:  a review. Am J Epidemiol
H, Hemminki K et al. A genome-​wide association 1994;139:237–​46.
study of Hodgkin’s lymphoma identifies new sus- Glaser SL, Clarke CA, Nugent RA, Stearns CB,
ceptibility loci at 2p16.1 (rel), 8q24.21 and 10p14 Dorfman RF. Social class and risk of Hodgkin’s
(gata3). Nature Gen 2010;42:1126–​30. disease in young-​adult women in 1988–​94. Int J
Engert A. Xvii. Treatment of advanced-​stage Hodgkin Cancer 2002;98:110–​7.
lymphoma. Hematol Oncol 2015;33 Suppl 1:87–​9. Glaser SL, Dorfman RF, Clarke CA. Expert review
Englund A, Hopstadius C, Enblad G, Gustafsson G, of the diagnosis and histologic classification of
Ljungman G. Hodgkin lymphoma:  a survey of Hodgkin disease in a population-​based cancer reg-
children and adolescents treated in Sweden 1985–​ istry: interobserver reliability and impact on inci-
2009. Acta Oncologica 2015;54:41–​8. dence and survival rates. Cancer 2001;92:218–​24.
643

Hodgkin Lymphoma 643

Glaser SL, Gulley ML, Clarke CA, Keegan TH, Chang Hjalgrim H, Rostgaard K, Johnson PC, Lake A, Shield
ET, Shema SJ et al. Racial/​ethnic variation in EBV-​ L, Little AM et  al. HLA-​A alleles and infectious
positive classical Hodgkin lymphoma in California mononucleosis suggest a critical role for cytotoxic
populations. Int J Cancer 2008;123:1499–​507. t-​cell response in EBV-​related Hodgkin lymphoma.
Glaser SL, Keegan TH, Clarke CA, Trinh M, Dorfman RF, Proc Natl Academy Sci USA 2010;107:6400–​5.
Mann RB et al. Exposure to childhood infections and Hjalgrim H, Seow A, Rostgaard K, Friborg J. Changing
risk of Epstein-​Barr virus–​defined Hodgkin’s lym- patterns of Hodgkin lymphoma incidence in
phoma in women. Int J Cancer 2005;115:599–​605. Singapore. Int J Cancer 2008;123:716–​9.
Glaser SL, Lin RJ, Stewart SL, Ambinder RF, Hjalgrim H, Smedby KE, Rostgaard K, Molin D, Hamilton-​
Jarrett RF, Brousset P et  al. Epstein-​Barr virus-​ Dutoit S, Chang ET et al. Infectious mononucleosis,
associated Hodgkin’s disease:  epidemiologic childhood social environment, and risk of Hodgkin
characteristics in international data. Int J Cancer lymphoma. Cancer Res 2007;67:2382–​8.
1997;70:375–​82. Hodgkin T. On some morbid appearances of the
Glaser SL, Swartz WG. Time trends in Hodgkin’s dis- absorbent glands and spleen. Med-​ Chir Trans
ease incidence:  the role of diagnostic accuracy. 1832;17:68–​114.
Cancer 1990;66:2196–​2204. Hors J, Dausset J. HLA and susceptibility to Hodgkin’s
Goldin LR, Bjorkholm M, Kristinsson SY, Turesson disease. Immunol Rev 1983;70:167–​92.
I, Landgren O. Highly increased familial risks Howlader N, Noone AM, Krapcho M, Miller D,
for specific lymphoma subtypes. Br J Hematol Bishop K, Kosary CL, Yu M, Ruhl J, Tatalovich
2009;146:91–​4. Z, Mariotto A, Lewis DR, Chen HS, Feuer EJ,
Gorini G, Stagnaro E, Fontana V, Miligi L, Ramazzotti Cronin KA (eds). SEER Cancer Statistics Review,
V, Amadori D et al. Alcohol consumption and risk 1975–​2014, National Cancer Institute. Bethesda,
of Hodgkin’s lymphoma and multiple myeloma: a MD, https://​seer.cancer.gov/​csr/​1975_​2014/​
multicentre case-​ control study. Ann Oncol , based on November 2016 SEER data submis-
2007;18:143–​48. sion, posted to the SEER web site, April 2017.
Grulich A, Vajdic C. M. The epidemiology of can- Howlader N, Noone AM, Krapcho M, Garshell J,
cers in human immunodeficiency virus infec- Miller D, Altekruse SF et  al. Surveillance,
tion and after organ transplantation. Sem Oncol Epidemiology, and End Results (SEER) program
2015;42:247–​57. (http://​www.seer.cancer.gov/​) SEER*stat data-
Grulich AE, Van Leeuwen MT, Falster MO, Vajdic CM. base:  incidence—​ SEER 18 regs research data +
Incidence of cancers in people with HIV/​AIDS Hurricane Katrina impacted Louisiana cases,
compared with immunosuppressed transplant Nov 2014 sub (2000–​2012) <katrina/​rita popula-
recipients: a meta-​analysis. Lancet 2007;370:59–​67. tion adjustment> -​linked to county attributes—​
Gutensohn N, Cole P. 1977. Epidemiology of total U.S., 1969–​ 2013 counties, National Cancer
Hodgkin’s disease in the young. Int J Cancer Institute, DCCPS, Surveillance Research Program,
1977;19:595–​604. Surveillance Systems Branch, released April 2015,
Gutensohn N, Cole P. Childhood social environ- based on the November 2014 submission. Bethesda,
ment and Hodgkin’s disease. N Engl J Med MD:  National Cancer Institute. [online], 2014.
1981;304:135–​40. Available at http://​seer.cancer.gov/​csr/​1975_​2012/​.
Hancock BW, Mosely R, Coup AJ. Height and Huang X, Hepkema B, Nolte I, Kushekhar K, Jongsma
Hodgkin’s disease. Lancet 1976;2:1364. T, Veenstra R et  al. Hla-​a*02:07 is a protective
Harris NL. Hodgkin’s disease: classification and differ- allele for EBV negative and a susceptibility allele
ential diagnosis. Modern Pathol 1999;12:159–​75. for EBV positive classical Hodgkin lymphoma in
Hartge P, Devesa SS, Fraumeni JF Jr. Hodgkin’s and China. PloS One 2012a;7:e31865.
non-​ Hodgkin’s lymphomas. Cancer Surveys Huang X, Kushekhar K, Nolte I, Kooistra W, Visser
1994;19–​20:423–​53. L, Bouwman I et  al. HLA associations in classi-
Hjalgrim H. On the aetiology of Hodgkin lymphoma. cal Hodgkin lymphoma: EBV status matters. PloS
Dan Med J 2012:59: b4485. One 2012b;7:e39986.
Hjalgrim H, Askling J, Rostgaard K, Hamilton-​Dutoi, IARC. Epstein-​ Barr Virus and Kaposi’s Sarcoma
S, Frisch M, Zhang JS et  al. Characteristics of Herpesvirus/​ Human Herpesvirus 8. IARC
Hodgkin’s lymphoma after infectious mononucle- Monographs on the Evaluation of Carcinogenic
osis. N Engl J Med 2003;349:1324–​32. Risks to Humans, Vol. 70. Lyon: IARC, 1997.
Hjalgrim H, Rasmussen S, Rostgaard K, Nielsen NM, Isager H, Andersen E. Pre-​ morbid factors in
Koch-​Henriksen N, Munksgaard L et al. Familial Hodgkin’s disease. I. Birth weight and growth pat-
clustering of Hodgkin lymphoma and multiple tern from 8 to 14 years of age. Scand J Haematol
sclerosis. J Natl Cancer Inst 2004;96:780–​4. 1978;21:250–​5.
64

644 Textbook of Cancer Epidemiology

Jaffe ES, Harris NL, Diebold J, Muller-​Hermelink HK. of hematologic malignancies. Ann Epidemiol
World Health Organization classification of lym- 2009;9:746–​53.
phomas: a work in progress. Ann Oncol 1998;9 Kristinsson SY, Gao Y, Bjorkholm M, Lund SH, Sjoberg
Suppl 5:s25–​s30. J, Caporaso N et al. Hodgkin lymphoma risk fol-
Jarrett RF. Viruses and Hodgkin’s lymphoma. Ann lowing infectious and chronic inflammatory dis-
Oncol 2002;13(Suppl 1):23–​9. eases: a large population-​based case-​control study
Jarrett RF, Gallagher A, Jones DB, Alexander FE, from Sweden. Int J Hematol 2015;101:563–​8.
Krajewski AS, Kelsey A et al. Detection of Epstein-​ Kristinsson SY, Landgren O, Sjoberg J, Turesson
Barr virus genomes in Hodgkin’s disease: relation I, Bjorkholm M, Goldin LR. Autoimmunity
to age. J Clin Pathol 1991;44:844–​8. and risk for Hodgkin’s lymphoma by subtype.
Jarrett RF, Krajewski AS, Angus B, Freeland J, Haematologica 2009;94:1468–​9.
Taylor PR, Taylor GM et  al. The Scotland and Kuppers R. The biology of Hodgkin’s lymphoma.
Newcastle Epidemiological Study of Hodgkin’s Nature Rev Cancer 2009;9:15–​27.
Disease:  impact of histopathological review and Kushekhar K, Van Den Berg A, Nolte I, Hepkema B,
EBV status on incidence estimates. J Clin Pathol Visser L, Diepstra A. Genetic associations in clas-
2003;56:811–​16. sical Hodgkin lymphoma: a systematic review and
Johnson PC, Mcaulay KA, Montgomery D, Lake insights into susceptibility mechanisms. Cancer
A, Shield L, Gallagher A et  al. Modeling HLA Epidemiol Biomarkers Prev 2014;23:2737–​47.
associations with EBV-​ positive and -​ negative La Vecchia C, Negri E, Parazzini F, Boyle P, D’avanzo
Hodgkin lymphoma suggests distinct mecha- B, Levi F et al. Height and cancer risk in a network
nisms in disease pathogenesis. Int J Cancer of case-​control studies from northern Italy. Int J
2015;137:1066–​75. Cancer 1990;45:275–​79.
Kamper-​Jorgensen M, Rostgaard K, Glaser SL, Zahm Landgren O, Engels EA, Pfeiffer RM, Gridley G,
SH, Cozen W, Smedby KE et al. Cigarette smoking Mellemkjaer L, Olsen JH et al. Autoimmunity and
and risk of Hodgkin lymphoma and its subtypes: a susceptibility to Hodgkin lymphoma: a population-​
pooled analysis from the International Lymphoma based case-​control study in Scandinavia. J Natl
Epidemiology Consortium (Interlymph). Ann Cancer Inst 2006;98:1321–​30.
Oncol 2013;24:2245–​55. Landgren O, Kerstann KF, Gridley G, Mellemkjaer L,
Kanda J, Matsuo K, Suzuki T, Hosono S, Ito H, Hemminki K, Linet MS, Goldin LR. Re:  familial
Ichinohe, T et  al. Association between obe- clustering of Hodgkin lymphoma and multiple
sity and the risk of malignant lymphoma in sclerosis. J Natl Cancer Inst 2005;97:543–​4.
Japanese:  a case-​ control study. Int J Cancer Larsson S, Wol A. Body mass index and risk of non-​
2010;126:2416–​25. Hodgkin’s and Hodgkin’s lymphoma:  a meta-​
Kanzler H, Kuppers R, Hansmann ML, Rajewsky K. analysis of prospective studies. Eur J Cancer
Hodgkin and Reed-​Sternberg cells in Hodgkin’s 2011;47:2422–​30.
disease represent the outgrowth of a dominant Lee JH, Kim Y, Choi JW, Kim YS. Prevalence and prog-
tumor clone derived from (crippled) germinal nostic significance of Epstein-​Barr virus infection
center b cells. J Exper Med 1996;184:1495–​05. in classical Hodgkin’s lymphoma: a meta-​analysis.
Keegan TH, Glaser SL, Clarke CA, Dorfman RF, Arch Med Res 2014;45:417–​31.
Mann RB, Digiuseppe, JA et al. Body size, phys- Levin LI, Chang ET, Ambinder RF, Lennette ET,
ical activity, and risk of Hodgkin’s lymphoma Rubertone MV, Mann RB et  al. Atypical predi-
in women. Cancer Epidemiol Biomarkers Prev agnosis Epstein-​ Barr virus serology restricted
2006;15:1095–​1101. to EBV-​ positive Hodgkin lymphoma. Blood
Khankhanian P, Cozen W, Himmelstein DS, 2012;120:3750–​5.
Madireddy L, Din L, Van Den Berg A et al. Meta-​ Li Q, Chang ET, Bassig BA, Dai M, Qin Q, Gao Y et al.
analysis of genome-​ wide association studies Body size and risk of Hodgkin’s lymphoma by
reveals genetic overlap between Hodgkin lym- age and gender:  a population-​based case-​control
phoma and multiple sclerosis. Int J Epidemiol study in Connecticut and Massachusetts. Cancer
2016;45:728–​40. Causes Control 2013;24:287–​95.
Kharazmi E, Fallah M, Pukkala E, Olsen, JH, Lim U, Morton LM, Subar AF, Baris D, Stolzenberg-​
Tryggvadottir L, Sundquist K et al. Risk of familial Solomon R, Leitzmann M et al. Alcohol, smoking,
classical Hodgkin lymphoma by relationship, his- and body size in relation to incident Hodgkin’s and
tology, age, and sex: a joint study from five Nordic non-​Hodgkin’s lymphoma risk. Am J Epidemiol
countries. Blood 2015;126:1990–​5. 2007;166:697–​708.
Klatsky AL, Li Y, Baer D, Armstrong MA, Udaltsov Linabery AM, Erhardt EB, Fonstad RK, Ambinder RF,
N, Friedman GD. Alcohol consumption and risk Bunin GR, Ross JA et al. Infectious, autoimmune
645

Hodgkin Lymphoma 645

and allergic diseases and risk of Hodgkin lym- prospective Million Women Study. Br J Cancer
phoma in children and adolescents:  a chil- 2013;108:2390–​8.
dren’s oncology group study. Int J Cancer Muschen M, Rajewsky K, Brauninger A, Baur AS,
2014;135:1454–​69. Oudejans JJ, Roers A et  al. Rare occurrence of
Lukes RJ, Craver L, Hall T, Rappaport H, Ruben P. classical Hodgkin’s disease as a T cell lymphoma. J
Report of the nomenclature committee. Cancer Exp Med 2000;191:387–​94.
Res 1966;26:1311. Newell GR, Cole SR, Miettinen OS, Macmahon B. Age
Macmahon B. Epidemiological evidence of the nature differences in the histology of Hodgkin’s disease. J
of Hodgkin’s disease. Cancer 1957;10:1045–​54. Natl Cancer Inst 1970;45:311–​7.
Macmahon B. 1966. Epidemiology of Hodgkin’s dis- Newton R, Crouch S, Ansell P, Simpson J, Willett EV,
ease. Cancer Research 1966;26:1189–​1201. Smith A et  al. Hodgkin’s lymphoma and infec-
Macmahon B, Cole P, Newell GR. Hodgkin’s dis- tion: findings from a UK case-​control study. Br J
ease: one entity or two? Lancet 1971;1:240–​1. Cancer 2007;97:1310–​4.
Martinez-​Maza O, Moreno AD, Cozen W. Ng AK. Current survivorship recommendations for
Epidemiological evidence:  IgE, allergies, and patients with Hodgkin lymphoma:  focus on late
hematopoietic malignancies. In:  Penichet effects. Blood 2014;124:3373–​9.
ML, Jensen-​ Jarolim E (eds.), Cancer and IgE. Niens M, Jarrett RF, Hepkema B, Nolte IM, Diepstra
New York: Springer, 2010: 79–​136. A, Platteel M et al. HLA-​A*02 is associated with
Mcaulay K, Jarrett RF. Human leukocyte antigens a reduced risk and HLA-​A*01 with an increased
and genetic susceptibility to lymphoma. Tissue risk of developing EBV + Hodgkin lymphoma.
Antigens 2015;86:98–​113. Blood 2007;110:3310–​5.
Monnereau A, Glaser SL, Schupp CW, Ekstrom SK, Nieters A, Deeg E, Becker N. Tobacco and alcohol
De SS, Kane E et  al. Exposure to UV radiation consumption and risk of lymphoma:  results of a
and risk of Hodgkin lymphoma: a pooled analysis. population-​based case-​control study in Germany.
Blood 2013;122:3492–​9. Int J Cancer 2006;118:422–​30.
Monnereau A, Orsi L, Troussard X, Berthou C, Fenaux Paffenbarger RS Jr, Wing AL, Hyde RT. Characteristics
P, Marit G, et al. History of infections and vaccina- in youth indicative of adult-​onset Hodgkin’s dis-
tions and risk of lymphoid neoplasms: does influ- ease. J Natl Cancer Inst 1977;58:1489–​91.
enza immunization reduce the risk? Leukemia Pallesen G, Hamilton-​Dutoit SJ, Rowe M, Young LS.
2007;21:2075–​9. Expression of Epstein-​Barr virus latent gene prod-
Monnereau A, Orsi L, Troussard X, Berthou C, Fenaux ucts in tumour cells of Hodgkin’s disease [see
P, Soubeyran P et  al. Cigarette smoking, alcohol comments]. Lancet 1991;337:320–​2.
drinking, and risk of lymphoid neoplasms: results Papadopoulou C, Antonopoulos CN, Sergentanis TN,
of a French case-​control study. Cancer Causes Panagopoulou P, Belechri M, Petridou ET. Is birth
Control 2008;19:1147–​60. weight associated with childhood lymphoma?
Montella M, Maso LD, Crispo A, Talamini R, Bidoli A meta-​analysis. Int J Cancer 2012;130:179–​89.
E, Grimaldi, M et  al. Do childhood diseases Parent ME, Rousseau MC, El-​Zein M, Latreille B, Desy
affect NHL and HL risk? A  case-​control study M, Siemiatycki J. Occupational and recreational
from northern and southern Italy. Leukemia Res physical activity during adult life and the risk of can-
2006;30:917–​22. cer among men. Cancer Epidemiol 2011;35:151–​9.
Morton LM, Turner JJ, Cerhan JR, Linet MS, Treseler PDQ® Adult Treatment Editorial Board. Adult
PA, Clarke CA et al. Proposed classification of lym- Hodgkin Lymphoma Treatment (pdq(r)):  Health
phoid neoplasms for epidemiologic research from Professional Version [online]. Bethesda, MD,
the pathology working group of the International 2016. Available at http://​www.ncbi.nlm.nih.gov/​
Lymphoma Epidemiology Consortium pubmed/​26389473 (accessed May 22, 2016).
(Interlymph). Blood 2007;110:695–​708. Poppema S, Delsol G, Pileri S, Stein H, Swerdlow SH,
Moutsianas L, Enciso-​ Mora V, Ma YP, Leslie S, Warnke RA, Jaffe E. Nodular lymphocyte predomi-
Dilthey A, Broderick P et  al. Multiple Hodgkin nant Hodgkin lymphoma. In: Swerdlow SH, Campo
lymphoma-​associated loci within the HLA region E, Harris NL, Jaffe ES, Pileri SA, Stein H, et  al
at chromosome 6p21.3. Blood 2011;118:670–​4. (eds.), Tumours of Haematopoietic and Lymphoid
Mueller N. Overview of the epidemiology of malig- Tissues. 3 ed., pp. 323–​5. Lyon: IARC, 2008.
nancy in immune deficiency. J Acq Immun Def Poppema S, Van IG, Torensma R, Smit J.
Syndrome 1999;21 Suppl 1:5–​10. Lymphadenopathy morphologically consist-
Murphy F, Kroll ME, Pirie K, Reeves G, Green J, ent with Hodgkin’s disease associated with
Beral V. Body size in relation to incidence of Epstein-​Barr virus infection. Am J Clin Pathol
subtypes of haematological malignancy in the 1985;84:385–​90.
64

646 Textbook of Cancer Epidemiology

Raphael M, Said J, Borisch B, Cesarman E, Harris In:  Swerdlow SH, Campo E, Harris N L, Jaffe
NL. Lymphomas associated with HIV infection. ES, Pileri SA, Stein H, et  al. (eds.), Tumours of
In:  Swerdlow SH, Campo E, Harris N L, Jaffe Haematopoietic and Lymphoid Tissues. 3 ed., pp.
ES, Pileri SA, Stein H, et  al (eds.), Tumours of 343–​9. Lyon: : IARC, 2008.
Haematopoietic and Lymphoid Tissues. 3 ed., pp. Uccini S, Monardo F, Ruco LP, Baroni CD, Faggioni
340–​2. Lyon: IARC, 2008. A, Agliano AM et al. High frequency of Epstein-​
Razis DV, Diamond HD, Craver LF. Hodgkin’s dis- Barr virus genome in HIV-​positive patients with
ease associated with other malignant tumors and Hodgkin’s disease. Lancet 1989;1:1458.
certain non-​neoplastic diseases. Am J Med Sci Urayama KY, Jarrett RF, Hjalgrim H, Diepstra A,
1959;238:327–​35. Kamatani Y, Chabrier A et al. Genome-​wide asso-
Ron E. Ionizing radiation and cancer risk:  evidence ciation study of classical Hodgkin lymphoma and
from epidemiology. Rad Res 1998;150:s30–​41. Epstein-​ Barr virus status-​defined subgroups. J
Rudant J, Orsi L, Monnereau A, Patte C, Pacquement Natl Cancer Inst 2012;104:240–​53.
H, Landman-​Parker J et al. Childhood Hodgkin’s Vajdic CM, Mccaughan GW, Grulich AE. Cancer
lymphoma, non-​Hodgkin’s lymphoma and factors risk after organ transplantation. JAMA
related to the immune system: the ESCALE study 2012;307:663–​4.
(SFCE). Int J Cancer 2011;129: 2236–​47. Van Krieken JH, Onciu M, Elenitoba-​ Johnson KSJ,
Sergentanis TN, Kanavidis P, Michelakos T, Petridou Jaffe ES. Lymphoproliferative diseases associated
ET. Cigarette smoking and risk of lymphoma with primary immune disorders. In:  Swerdlow
in adults:  a comprehensive meta-​ analysis on SH, Campo E, Harris NL, Jaffe, ES, Pileri SA,
Hodgkin and non-​Hodgkin disease. Eur J Cancer Stein H, et al (eds.), Tumours of Haematopoietic
Prev 2013;22:131–​50. and Lymphoid Tissues. 3 ed., pp. 336–​ 9.
Serraino D, Franceschi S, Talamini R, Barra S, Negri E, Lyon: IARC, 2008.
Carbone A, La Vecchia C. Socio-​economic indica- Van Leeuwen MT, Turner JJ, Falster MO, Meagher NS,
tors, infectious disease and Hodgkin’s disease. Int Joske DJ, Grulich AE et al. Latitude gradients for
J Cancer 1991;47:352–​7. lymphoid neoplasm subtypes in Australia support
Stein H. Hodgkin lymphoma—​ introduction. an association with ultraviolet radiation exposure.
In:  Swerdlow SH, Campo E, Harris NL, Jaffe Int J Cancer 2013;133:944–​51.
ES, Pileri SA, Stein H, et  al (eds.), Tumours of Vermaete, NV, Wolter P, Verhoef GE, Kollen BJ,
Haematopoietic and Lymphoid Tissues. 3 ed., Kwakkel G, Schepers L, Gosselink R. Physical activ-
p. 322. Lyon: IARC, 2008. ity and risk of lymphoma: a meta-​analysis. Cancer
Stein H, Delsol G, Pileri S, Weiss LM, Poppema S, Epidemiol Biomarkers Prev 2013;22:1173–​84.
Jaffe E. Classical Hodgkin lymphoma, introduc- Vockerodt M, Yap LF, Shannon-​Lowe C, Curley H,
tion. In: Swerdlow SH, Campo E, Harris N L, Jaffe Wei W, Vrzalikova K, et  al. The Epstein-​ Barr
ES, Pileri SA, Stein H, et  al (eds.), Tumours of virus and the pathogenesis of lymphoma. J Pathol
Haematopoietic and Lymphoid Tissues. 3 ed., pp. 2015;235:312–​22.
326–​9. Lyon: IARC, 2008a. Von Behren J, Spector LG, Mueller BA, Carozza SE,
Stein H, Von Wasielewski R, Poppema S, Maclennan K, Chow EJ, Fox EE et  al. Birth order and risk of
Guenova M. Nodular sclerosis classical Hodgkin childhood cancer: a pooled analysis from five us
lymphoma. In: Swerdlow SH, Campo E, Harris N states. Int J Cancer 2011;128:2709–​16.
L, Jaffe ES, Pileri SA, Stein H, et al (eds.), Tumours Wang SS, Slager SL, Brennan P, Holly EA, De SS,
of Haematopoietic and Lymphoid Tissues. 3 ed., Bernstein L et al. Family history of hematopoietic
p. 330. Lyon: IARC, 2008b. malignancies and risk of non-​Hodgkin lymphoma
Stein SR, Morgan DS. Hodgkin disease. In: Greer JP, (NHL): a pooled analysis of 10 211 cases and 11
Foester J, Lukens JN, Rodgers GM, Paraskevas 905 controls from the International Lymphoma
F, Glader B. (eds.) Wintrobe’s Clinical Epidemiology Consortium (Interlymph). Blood
Hematology. Philadelphia: Lippincott Williams & 2007;109:3479–​88.
Wilkins, 2004. Weiss LM, Movahed LA, Warnke RA, Sklar J.
Sud A, Hemminki K, Houlston RS. Candidate gene Detection of Epstein-​Barr viral genomes in Reed-​
association studies and risk of Hodgkin lym- Sternberg cells of Hodgkin’s disease. N Engl J Med
phoma:  a systematic review and meta-​analysis. 1989;320:502–​6.
Hematol Oncol 2017;35:34–​50. Weiss LM, Strickler JG, Warnke RA, Purtilo DT, Sklar,
Swerdlow SH, Webber SA, Chadbrun A, Ferry JA. J. Epstein-​Barr viral DNA in tissues of Hodgkin’s
Post-​transplant lymphoproliferative disorders. disease. Am J Pathol 1987;129:86–​91.
647

Hodgkin Lymphoma 647

Weiss LM, Von Wasielewski R, Delsol G, Poppema S, Willett EV, Roman E. Obesity and the risk of Hodgkin
Stein H. Mixed cellularity classical Hodgkin lym- lymphoma (United Kingdom). Cancer Causes
phoma. In:  Swerdlow SH, Campo E, Harris NL, Control 2006;17:1103–​06.
Jaffe ES, Pileri SA, Stein H, et al (eds.), Tumours Wu TC, Mann RB, Charache P, Hayward SD, Staal S,
of Haematopoietic and Lymphoid Tissues. 3 ed., Lambe BC et al. Detection of EBV gene expression
p. 331. Lyon: IARC, 2008. in Reed-​Sternberg cells of Hodgkin’s disease. Int J
Westergaard T, Melbye M, Pedersen JB, Frisch M, Cancer 1990;46:801–​4.
Olsen JH, Andersen PK. Birth order, sibship size Yang J, Lin L, Levin L, Mueller N, Liu Q, Weir E, et al.
and risk of Hodgkin’s disease in children and Cell-​free EBV is detected in patients with EBV(+)
young adults: a population-​based study of 31 mil- Hodgkin’s lymphoma (HL) years in advance of
lion person-​years. Int J Cancer 1997;72:977–​81. diagnosis. Oral presentation at the 10th interna-
Wilks S. Cases of enlargement of the lymphatic glands tional symposium on Epstein-​Barr virus & associ-
and spleen (or, Hodgkin’s disease) with remarks. ated malignant diseases. July 16–​21, 2002, Cairns,
Guy’s Hospital Reports 1865;11:56–​67. Australia.
648
649

27
Non-​Hodgkin Lymphoma
K A R I N E K S T R Ö M S M E D B Y, M A D S M E L B Y E , A N D H A N S - ​O L O V   A D A M I

N on-​Hodgkin lymphoma (NHL) is unique

among human malignancies for several
reasons. First, it is distinguished as an entity
published a new classification (Swerdlow et  al,
2016) based on the previous Revised European-​
American Classification of Lymphoid Neoplasms
by exclusion from a much less common group (REAL). The WHO recognizes B-​cell neoplasms,
of lymphomas called Hodgkin lymphoma. T/​NK-​cell neoplasms, and Hodgkin lymphoma.
Second—​and more important—​the incidence of The T-​and B-​cell neoplasms are divided into pre-
NHL has been rising more rapidly than that of cursor lymphoblastic neoplasms (acute lympho-
virtually any other human cancer in many parts blastic leukemia/​lymphoblastic lymphoma), and
of the world. Indeed, this increase has been con- mature neoplasms (Box 27-​1).
sidered almost epidemic and its causes remain Symptoms The most common early signs are pain-
enigmatic, though it suggests exposure to causal less swelling of lymph nodes, fever, night sweats,
agents that might be avoidable. and involuntary loss of weight. Poor appetite,
Finally, no other cancer entity comprises such fatigue, bleeding itchy skin, reddened patches
a heterogeneous group of malignancies as does on the skin, and bone pain may also be reported.
NHL. Originating from lymphocytes, NHL can Many cases of NHL are detected through routine
develop from within organized lymphoid tis- procedures while still asymptomatic.
sue, such as lymph nodes, or from other sites.
Diagnosis Histopathologic diagnosis is prefera-
Many lymphoid neoplasms pass through both
bly based on biopsy material from lymph nodes,
solid tumor and circulating (leukemic) phases.
bone marrow, or an extranodal organ, depend-
Chronic lymphocytic leukemia (CLL) and plas-
ing on site of involvement. Detailed subgrouping
mocytoma/​multiple myeloma now belong to the
requires immunohistochemistry, flow cytometry,
entity of NHL according to the widely accepted
cytogenetics, and molecular genetic analysis.
World Health Organization (WHO) classifi-
Clinical staging involves information on affected
cation (Swerdlow et al, 2016), although they
lymph nodes and spread to the bone marrow,
are classified separately in the International
spleen, or organs outside the lymphatic system,
Classification of Diseases (ICD). The diversity of
and is classified according to the Ann Arbor
NHL subgroups and the inconsistency in classifi-
or Musshoff staging systems. Imaging studies,
cation systems have complicated research on the
including PET (positron emission tomography)
causes of this cancer, but in the past decade possi-
scan to an increasing extent, are used to deter-
ble etiologic heterogeneity among NHL subtypes
mine the localization and extent of spread. Bone-​
has often been recognized.
marrow aspiration and biopsy are often used in
complement, sometimes with lumbar puncture
CLINICAL SYNOPSIS
as well.
Subgroups NHL, divided into B-​cell and T-​cell
neoplasms based on their immunophenotypic Treatment Disease localized to a group of lymph
characteristics, includes numerous specific enti- nodes or to an extranodal organ can be treated
ties that are often associated with particular kar- with radiotherapy. In more advanced stages,
yotypic abnormalities. Due to the wide variety of chemotherapy or immunotherapy (monoclonal
disease entities, classification of NHL has been antibodies), separately or in combination, with
difficult, as reflected in the many taxonomies. In or without sequential radiation therapy, is often
2001, and updated in 2008 and 2016, the WHO used, depending on the type of NHL, symptoms,
650

BOX 27-​1
WORLD HEALTH ORGANIZATION CLASSIFICATION OF
LYMPHOID NEOPLASMS †

B -​C E L L N E O P L A S M S
Precursor B-​Cell Neoplasm
• B lymphoblastic leukemia/​lymphoma, NOS
• B lymphoblastic leukemia/​lymphoma with recurrent genetic abnormalities

Mature B-​Cell Neoplasms*

• Chronic lymphocytic leukemia/​small lymphocytic lymphoma
• B-​cell prolymphocytic leukemia
• Splenic marginal zone lymphoma
• Hairy cell leukemia
• Lymphoplasmacytic lymphoma
• Plasma cell neoplasms
• Extranodal marginal zone B-​cell lymphoma of mucosa-​associated
lymphoid tissue (MALT) lymphoma
• Nodal marginal zone lymphoma
• Follicular lymphoma
• Mantle cell lymphoma
• Diffuse large B-​cell lymphoma (subtypes not listed)
• Burkitt lymphoma

T AND NK-C
​ ELL NEOPLASMS
Precursor T-​Cell Neoplasm
• T lymphoblastic leukemia/​lymphoma

Mature T-​and NK-​Cell Neoplasms

• T-​cell prolymphocytic leukemia
• T-​cell large granular lymphocytic leukemia
• Aggressive NK-​cell leukemia
• Adult T-​cell leukemia/​lymphoma
• Extranodal NK/​T-​cell lymphoma, nasal type
• Enteropathy-​associated T-​cell lymphoma
• Hepatosplenic T-​cell lymphoma
• Subcutaneous panniculitis-​like T-​cell lymphoma
• Mycosis fungoides
• Sezary syndrome
• Primary cutaneous CD30 positive T-​cell lymphoproliferative disorders
• Peripheral T-​cell lymphoma, not otherwise specified
• Angioimmunoblastic T-​cell lymphoma
• Anaplastic large cell lymphoma, ALK positive/​ALK negative

H O D G K I N LY M P H O M A
• Nodular lymphocyte predominant Hodgkin lymphoma
• Nodular sclerosis classical Hodgkin lymphoma
• Mixed cellularity Hodgkin lymphoma
• Lymphocyte-​rich classical Hodgkin lymphoma
• Lymphocyte-​depleted classical Hodgkin lymphoma
*Rare entities and subgroups of listed entities are not included.
651

Non-Hodgkin Lymphoma 651

laboratory values, and location of the lymphoma. 2 to 3 per 100,000 person-​years in Thailand and
Many indolent lymphoma types are not treated China to about 14 per 100,000 person-​ years
until definite symptoms or other signs develop. among whites in the United States and Canada.
New biological anticancer agents are increasingly In general, the incidence is 10% to 60% higher
used, including B-​cell receptor signaling pathway among males than among females (Figs. 27-​1 to
or proteasome inhibitors, but many other types of 27-​4) and is similarly higher among whites than
biological drugs are currently being tested in clin- among blacks, although these figures vary by
ical trials. High-​dose chemotherapy with stem-​ subtype. For example, among whites the male-​
cell transplantation may be used as consolidative to-​female ratio is greater than 2 for Burkitt leu-
treatment at diagnosis in high-​risk patients, or kemia/​lymphoma, mantle cell lymphoma, and
else at relapse after standard treatment. lymphoplasmacytoma, whereas it is close to 1 for
Prognosis and Progress Because NHL is a hetero- marginal zone and follicular NHL (Morton et al,
geneous malignancy, overall survival figures pro- 2006b). For B-​cell NHL the ratio of white-​to-​
vide only crude information. The prognosis may black age-​adjusted incidence is between 2 and 3
vary from an aggressive course bringing death for hairy cell leukemia, mantle cell and follicular
within a few weeks, to an indolent course even lymphoma, and lymphoplasmacytoma. In con-
with no therapy for years or even decades. The trast, peripheral T-​cell NHL and plasma cell neo-
5-​year relative survival rate in the most common plasms are more common among blacks (Morton
aggressive NHL subtype diffuse large B-​cell lym- et al, 2006b).
phoma improved from about 30% in the early As illustrated in Figures 27-​3 and 27-​4, the
1960s to 60%–​70% in the 2000s. In children, rela- age-​specific incidence pattern is different from
tive 5-​year survival approaches 80%–​90%, a great that of Hodgkin lymphoma. The incidence
improvement from the mid-​ 1970s, when few increases steadily with age for almost all sub-
children were alive 5 years after diagnosis. types, except B-​and T-​cell lymphoblastic leuke-
mia/​lymphoma, which are diagnosed primarily
in children.
DESCRIPTIVE
EPIDEMIOLOGY Time Trends
In the past decades, there has been a dramatic
Incidence
increase in the incidence of NHL in most
Worldwide, NHL constitutes the 11th most com- Western countries, on average 2% to 4% annu-
monly diagnosed malignancy, whereas in the ally (Devesa & Fears, 1992; Hjalgrim et al, 1996,
high-​income countries it ranks 8th (Ferlay et al, Shiels et al, 2013). This epidemic increase is
2013). The incidence of NHL varies from around among the most rapid observed for any cancer.

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 2 4 6 8 10 12
ASR (world) per 100,000 person-years

FIGURE 27-​1 
Age-​standardized (to the 2012 world population) incidence rates of non-​Hodgkin lymphoma among
women. Incidence rates for non-​Hodgkin lymphoma include ICD 10 codes C82-​85 and C96.
Source: Ferlay et al, 2013.
652

China
Japan
Poland
Spain
Sweden
The Netherlands
United Kingdom
United States of America
0 2 4 6 8 10 12 14 16
ASR (world) per 100,000 person-years

FIGURE 27-​2 
Age-​standardized (to the 2012 world population) incidence rates of non-​Hodgkin lymphoma among
men. Incidence rates for non-​Hodgkin lymphoma include ICD 10 codes C82-​85 and C96.
Source: Ferlay et al, 2013.

90
New cases per 100,000

80
70
person-years

60
50
40
30
20
10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 27-​3  Age-​specific incidence rates for non-​Hodgkin lymphoma among women include ICD 10 codes C82-​85
and C96.
Source: Ferlay et al, 2013.

140
New cases per 100,000

120
person-years

100
80
60
40
20
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 27-​4  Age-​specific incidence rates for non-​Hodgkin lymphoma among men include ICD 10 codes C82-​85
and C96.
Source: Ferlay et al, 2013.
653

Non-Hodgkin Lymphoma 653

An increase has also been observed in countries The B-​cell phenotype of lymphomas domi-
such as India, Japan, Brazil, and Singapore, as nates in most parts of the world. In Western
well as in Puerto Rico (Devesa & Fears, 1992). countries only 5% to 15% of all NHLs are of T-​or
Changes in classification—​cases previously diag- NK-​cell phenotype (Perry et al, 2016). However,
nosed as Hodgkin lymphoma now classified as in some parts of the Far East and the Caribbean,
NHL—​cannot explain the increase. Indeed, a the T-​and NK-​cell lymphomas constitute the
number of detailed studies including patho- dominant group. Hence, it makes up about 45%
logic reviews indicate that the increase is real of NHL in different areas of Japan, 29% in Hong
(Hartge et al, 1992). According to figures from Kong, 39% in Taiwan, and at least 18% in main-
the United States and the Nordic countries, the land China (reviewed by Pallesen and Hamilton
long-​lasting increase in incidence began to level Dutoit, 1993). In most regions, T-​cell lympho-
off during the 1990s (Shiels et al, 2013; Sandin mas are primarily of the peripheral T-​cell type.
et al, 2006). The reasons for the past increase, Extranodal NK/​ T-​cell lymphoma, nasal type,
as well as for the leveling off in some countries, is the second most common in Asian countries
remain unclear. including China and Thailand, whereas this sub-
type is uncommon in Europe and the United
Subtype Trends States (Perry et al, 2016).
While the overall increase in NHL is unambig-
uous and indeed dramatic, our knowledge of GENETIC AND MOLECULAR
incidence trends by subtype is limited. In the EPIDEMIOLOGY
group of extranodal disease comprising 25% to
30% of all NHL, brain and skin disease show the Inherited Susceptibility
highest proportional rise in incidence (Devesa & Non-​ Hodgkin lymphoma has repeatedly been
Fears, 1992; Groves et al, 2000). For NHL affect- reported to aggregate in certain families. Having
ing the brain, the increase may partly be a result close relatives with this disease or other hema-
of improved diagnostic methods and, in some tolymphoproliferative malignancies increases
countries, may be caused by NHL secondary to the risk two-​to threefold (reviewed in Cerhan &
human immunodeficiency virus (HIV) infection. Slager, 2015). Although familial risks are ele-
However, the incidence of AIDS-​associated NHL vated for multiple lymphoma subtypes, subtype-​
subtypes has decreased markedly in high-​income specific risks are most pronounced with
countries since the introduction of highly active family history of that particular subtype, sug-
antiretroviral therapy (HAART) (van Leeuwen gesting subtype-​specific genetic etiology. In the
et al, 2009; Shiels et al, 2013). past decade, a number of low-​penetrance genetic
Analyses on incidence trends by histologic risk variants have been revealed in large-​scale
subtype are few. The Surveillance, Epidemiology, genome-​wide association studies, notably from
and End Results (SEER) program in the United the InterLymph consortium.
States found high-​ grade NHL to have tripled The InterLymph consortium is an open sci-
among males and doubled among females entific forum for epidemiologic research in
between the periods 1978 to 1983 and 1990 to lymphoma that was formed in 2001 (http://​epi.
1995 (Groves et al, 2000). In a subsequent anal- grants.cancer.gov/​InterLymph/​). Members are
ysis for the period 1992 to 2009, overall inci- international investigators who pool data across
dence rates for diffuse large B-​cell lymphoma studies, and at present more than 25 study cen-
and chronic lymphocytic leukemia/​small lym- ters are involved. Several large-​ scale studies
phocytic lymphoma declined. Rates for marginal of genetic variation have identified a few loci/​
zone lymphoma increased until 2001 and stabi- regions of NHL subtypes worth highlighting.
lized thereafter, while mantle cell lymphoma and In view of NHL originating from cells of the
peripheral T-​cell lymphomas were still increas- immune system, it is of particular interest that
ing through 2009 (Shiels et  al, 2013). The same common subtypes, mainly follicular lymphoma,
subtypes were increasing in Australia up to 2006 but also marginal zone lymphoma and diffuse
along with DLBCL and follicular lymphoma, large B-​cell lymphoma have been associated with
although these trends were in part counteracted variation in the human leukocyte antigen (HLA)
by a decline in NHL/​lymphoma not otherwise region, which regulates antigen presentation to T
specified (van Leeuwen et al, 2014). helper and T cytotoxic cells (Skibola et al, 2014;
654

654 Textbook of Cancer Epidemiology

Cerhan et al, 2014; Vijai et al 2015). Another introducing adverse genetic events. Indeed, chro-
strong susceptibility locus for diffuse large B-​cell mosomal translocations involving the immuno-
lymphoma is located near the EXOC2 (exocyst globulin heavy chain locus as well as somatic
complex component 2) at 6p25, which functions mutations are common findings in many lym-
at the interface between host defense and cell-​ phoma entities. Frequently, the immunoglobulin
death regulation (Cerhan et al, 2014). promoter on chromosome 14 serves as a trans-
Associations with risk of both follicular and location partner, for instance in follicular lym-
diffuse large B-​cell lymphoma have also been phoma, where t(14,18) leads to overexpression
noted with variation in the 8q24 region impli- of the antiapoptotic protein BCL2, and in mantle
cated in MYC activation, a region linked with cell lymphoma, where t(11,14) leads to expression
risk of a number of different cancer forms. For of cyclin D1 promoting proliferation. In sporadic
CLL, over 30 susceptibility loci have so far been Burkitt lymphoma, t(8,14) involves class-​switch
revealed and pathway analyses show a strong recombination, which is also implicated in the
role for apoptosis (Berndt et al, 2016). In sum- activated diffuse large B-​cell lymphoma, but not
mary, most discovered loci have relatively small in the germinal-​center subtype (Koff et al, 2015).
effect sizes (relative risk 0.5–​2.0) and the under- Next-​ generation sequencing studies have
lying mechanisms of association and implica- revealed additional recurrent mutations, pro-
tions for lymphomagenesis are not yet fully viding new insights into disrupted cell-​signaling
understood. Therefore, together with other pathways in lymphomagenesis. These pathways,
aspects pertaining to NHL, screening among relevant for the development of new treatments,
family members of sporadic NHL including involve B-​cell and Toll-​like receptor signaling
CLL cannot be recommended (reviewed in as well as other pathways involving NOTCH,
Cerhan & Slager, 2015). nuclear factor-​κβ and the mitogen-​activated pro-
Many patients with certain rare inherited tein kinase. MYD88 mutations, occurring in sev-
defects of immune function develop primarily eral B-​cell lymphoma subtypes, most notably in
B-​cell lymphomas. These conditions include Waldenström macroglobulinemia, are involved
ataxia-​telangiectasia, Wiskott-​Aldrich syndrome, in toll-​like receptor signaling (Rossi et al, 2013).
common variable immunodeficiency, severe com- The V600E mutation of BRAF, a kinase com-
bined immunodeficiency, X-​linked lymphoprolif- monly mutated in solid tumors, is the key genetic
erative disorder, Nijmegen breakage syndrome, lesion of hairy cell leukemia, which constitutively
hyper-​IgM syndrome, and autoimmune lympho- activates the mitogen-​ activated protein kinase
proliferative syndrome. The mechanisms of lym- pathway (Tiacci et al, 2011).
phomagenesis, related to the underlying disorder,
involve loss of T-​cell control (Wiskott-​Aldrich RISK FACTORS
syndrome), apoptosis defects (autoimmune lym-
phoproliferative syndrome), abnormal DNA Tobacco
repair function (ataxia-​telangiectasia, Nijmegen Most large case-​control and cohort studies have
breakage syndrome), defective T-​cell/​B-​cell inter-
found no association between cigarette smok-
actions (hyper-​ IgM-​syndrome), and perhaps ing and risk of overall NHL (Besson et al, 2006;
chronic antigen stimulation (common variable Fernberg et  al, 2006). However, accumulating
immunodeficiency). However, these genetic dis- evidence suggests an association with risk of fol-
orders explain very few new cases in the general licular lymphoma, particularly among women
population and cannot explain temporal trends in (Freedman et  al, 1998; Diver et  al, 2012; Linet
incidence. et al, 2014) (Table 27-​1). An analysis of 152,958
cohort members in the United States from 1992
Somatic Mutations to 2007 revealed a significant 2.1-​fold increased
In the context of lymphomagenesis, lympho- risk of follicular lymphoma among female cur-
cytes, primarily of B-​cell type, undergo several rent smokers, but no association among men
potentially dangerous genetic changes during cell (Diver et  al, 2012). Similarly, in a large pooled
maturation and as part of the adaptive immune analysis within the InterLymph consortium
response. These include immunoglobulin recom- including 3,530 follicular lymphoma patients and
bination, somatic hypermutation, and class-​ 22,639 controls from Europe, the United States,
switch recombination, all of which involve DNA and Australia, cigarette smoking was associated
double-​strand breaks that also increase risk of with a 20% increase in risk among women, along
65

TABLE 27-​1   EPIDEMIOLOGY OF NON-HODGKIN LYMPHOMA (NHL) SUBT YPES:

RISK FACTOR SUMMARY

Direction of factor (sex if relevant) Main associated NHL subtype(s)

effect*

Well-​Confirmed Risk Factors

HIV infection ↑↑ Aggressive NHL
Organ transplantation ↑↑ Aggressive NHL
Hereditary immunodeficiency syndromes ↑↑ Aggressive NHL
Family history in first-​degree relative ↑ All
Autoimmune and chronic inflammatory disorders ↑ Several
Epstein Barr virus ↑↑ Aggressive NHL
Human T-​cell leukemia/​lymphoma
Adult T-​cell leukemia/​lymphoma
Hepatitis C virus ↑ Several
Human herpesvirus 8 ↑ Primary effusion
Lymphoma
Helicobacter pylori ↑ Gastric MALT
Breast implants ↑ Anaplastic large-​cell lymphoma

Probable Relationship Exists, Based

on Substantial Data
Cigarette smoking (women) ↑ Follicular lymphoma
High body mass index ↑ Diffuse large B-​cell lymphoma
Tall stature ↑ Burkitt lymphoma, Hairy cell
leukemia
Alcohol intake ↓ Several
Farming ↑ Diffuse large B-​cell lymphoma
Ultraviolet radiation ↓ Diffuse large B-​cell lymphoma
Atopic conditions ↓ Several
Hepatitis B virus ↑ Diffuse large B-​cell lymphoma

Weak, if Any, Relationship Exists, Based on

Substantial Data
Pesticides ↑
Solvents ↑
Blood transfusion ↑
Vaccinations ↓

Inconsistent Findings or Limited Study to Date

Dietary factors ↑↓
Reproductive history ↓
Hormone therapy ↓
Physical activity ↓
Vitamin D ↓
Borrelia burgdorferi ↑
Chlamydia psittaci ↑
Campylobacter jejuni ↑

*Arrows indicate the approximate magnitude of the relationship: ↑, slight to moderate increase in risk; ↑↑, moderate to large increase in risk;
↓, slight to moderate decrease in risk; ↓↓, moderate to large decrease in risk; —​, no association.
65

656 Textbook of Cancer Epidemiology

with a significant positive trend for duration of 1999; Tramacere et al, 2012; Morton et al, 2014),
smoking, whereas no associations were noted but several studies also reported no association
among men (Linet et al, 2014). (Chang et al, 2004; Willet et al, 2004; Heinen et al,
The potential association is biologically inter- 2013). Further, putative associations with alco-
esting, because smoking may induce the chro- hol may be both sex-​and subtype-​specific mak-
mosomal translocation t(14;18) (Bell et al, 1995; ing estimates of overall NHL risk in both sexes
Schuler et  al, 2003), which is found in 70% to difficult to interpret. In the large pooled analy-
95% of follicular lymphomas. Genetic variation sis of self-​ reported environmental exposures
may also explain variations in the association (Morton et al, 2014), high alcohol intake was
between smoking and follicular lymphoma risk, associated with a reduced risk of diffuse large B-​
and differences in association by sex, but so far cell lymphoma among men only, and a reduced
our knowledge is limited. risk of follicular lymphoma among women
only. A meta-​analysis in 2012 of 21 case-​con-
Diet trol and 8 cohort studies reported consistently
Research on the influence of dietary factors is reduced risks of NHL in association with mod-
complex per se and may be further complicated erate and heavy drinking among case-​ control
in NHL, if the different subgroups have different as well as cohort studies, supporting a favorable
etiologies (Purdue et al, 2004). There are no estab- role of alcohol drinking (Tramacere et al, 2012).
lished links between dietary factors and risk of Although alcohol intake has been linked with
NHL, and few studies have had sufficient sample immune modulation, the mechanisms behind a
size to robustly evaluate potential heterogeneity putative association are not clear and call for cau-
in risk of NHL subtypes. Intake of fruits, vita- tious interpretation of the nature of the observed
min C, vegetables, lutein, zeaxanthin, and zinc association.
has been inversely associated with NHL in sev-
eral studies mostly of case-​control design (Chang Reproductive Factors
et al, 2005b; Kelemen et al, 2006; Talamini et al, Because a pregnancy results in altered immu-
2006), although no protection has been noted nologic reactions and NHL is greatly influenced
with dietary patterns rich in fruits/​ vegetables by immune dysregulation, it has been suggested
such as the Mediterranean diet (Ollberding et al, that pregnancy could alter the incidence of
2014). The antioxidant or nitrosation-​inhibiting NHL. There is limited evidence of an associa-
properties of fruits and vegetables might be tion between a woman’s reproductive history and
involved in reducing risk of NHL. risk of NHL overall (Adami et al, 1997; Frisch et
Dairy products, as well as total fat and sat- al, 2006). However, a decreased risk of follicular
urated fat intake, have been associated with an lymphoma was noted with increasing number of
increased risk of NHL in several studies (Chiu pregnancies in a pooled analysis of 18 case-​con-
et al, 1996; Purdue et al, 2004; Zheng et al, 2004; trol studies (Kane et al, 2012). In a prospective
Cross et  al, 2006; Talamini et  al, 2006), although cohort study, pregnancy was associated with a
not all (Rorhmann et al, 2011). Also, meat and in borderline reduced risk of B-​cell NHL, in partic-
particular read meat consumption has been asso- ular diffuse large B-​cell lymphoma. Further stud-
ciated with an increased risk of NHL in some case-​ ies of NHL subtype-​specific risks are warranted
control studies (Zheng et  al, 2004; Chang et  al, in this area.
2005d), whereas other prospective studies have
reported no association overall and inconsistent Hormones
results for major NHL subtypes (Cross et al, 2006; Notwithstanding the substantially higher inci-
Daniel et  al, 2012; Rohrmann et  al, 2011). The dence of NHL among men than among women,
underlying mechanisms for the possible impor- the possible role of hormones—​endogenous or
tance of certain meats and lipids are unknown. exogenous—​ has not been adequately investi-
However, in addition to dietary carcinogens, it has gated. The data on hormone replacement therapy
been hypothesized that such products might lead are conflicting (as reviewed in Costas et al, 2014).
to altered immunocompetence (Zheng et al, 2004). In the large pooled case-​control study described
earlier, hormone replacement therapy was asso-
Alcohol ciated with a reduced risk of NHL, primarily
There is some evidence that alcohol intake is diffuse large B-​cell lymphoma, but only with cur-
associated with a reduced risk of NHL (Chiu et al, rent use and with no clear trend with increasing
657

Non-Hodgkin Lymphoma 657

duration of use (Kane et  al, 2013). In the same lymphoma was caused by a vectored virus
setting, hormonal contraceptive use was asso- (Epstein et  al, 1964). The incidence of Burkitt
ciated with an increased risk of follicular lym- lymphoma is highest among children in areas of
phoma, but not diffuse large B-​cell lymphoma Africa with holoendemic malaria, but sporadic
(Kane et al, 2012). cases occur at all ages in other parts of the world.
It has been speculated that African Burkitt lym-
Anthropometric Measures phoma might arise as a result of enhanced B-​cell
It has been speculated whether the obesity epi- proliferation by early EBV infection interacting
demic occurring in many parts of the world is with the antigenic stimulatory effects of malaria.
linked to the increasing incidence in NHL. In Whereas EBV can be demonstrated in more
fact, there is now ample evidence of an associa- than 95% of tumor biopsies of endemic cases in
tion of high BMI with risk of diffuse large B-​cell equatorial Africa, less than 20% of sporadic cases
lymphoma (Castillo et al, 2014), but not convinc- in high-​income countries are EBV positive (Kelly
ingly with other NHL subtypes (Morton et  al, et  al, 2013). Between 50% and 90% of Burkitt
2014). In a meta-​analysis of 10 case-​control and lymphomas in low-​and high-​income countries
6 cohort studies, overweight individuals were outside Africa are EBV associated, but in regions
at a significantly 1.14-​fold increased risk, and with a temperate climate in South America, the
obese individuals at a 1.29-​fold-​increased risk EBV association is weaker than in tropical regions
of diffuse large B-​cell lymphoma (Castillo et  al, in the north of the continent (International
2014). A meta-​regression analysis showed a 14% Agency for Research on Cancer, 2012).
increase in incidence for each 10-​kg/​m2 increase Only a small fraction (usually less than 5%)
in BMI. Results were strikingly similar by sex and of other B-​ cell NHLs in immunocompetent
among cohort and case-​control studies, but dif- patients contain EBV, and no consistent associ-
fered by geographical region, as increased risks ation between histologic type and EBV DNA has
were observed in America and Asia but not in been evident. In contrast, EBV is often demon-
Europe. Because obesity can be characterized as strated in tumor material from T-​cell lympho-
a chronic inflammatory state, an association with mas. These include in particular the extranodal
diffuse large B-​cell lymphoma is biologically plau- NK/​T-​cell lymphoma of the nasal type, but also
sible, since this subtype has also been associated aggressive NK-​cell leukemia and possibly angio-
with diseases characterized by inflammation. immunoblastic T-​ cell lymphoma. Sinonasal
There is accumulating evidence of a moderate T-​cell lymphomas, which are common in Asia
risk increase of NHL overall associated with adult but rare in Europe and the United States, are
height in both sexes (Green et  al, 2011). Height almost always EBV associated. Between 18% and
could potentially account for some of the differ- 70% of other peripheral T-​cell lymphomas con-
ences in risk of NHL by sex (Walter et al, 2013). In tain EBV (Grywalska & Rolinski, 2015).
the large InterLymph pooled analysis in 2014, over Increased incidence of NHL has been related
twofold-​increased risks were noted specifically for to the X-​ linked lymphoproliferative syndrome,
hairy cell leukemia and Burkitt lymphoma in tall Wiskott-​Aldrich syndrome, ataxia-​telangiectasia,
individuals (highest quartile compared to lowest) and other primary immunodeficiency syndromes.
(Morton et al 2014). Interestingly, these subtypes In those cases, EBV is almost always found. Epstein
have the most pronounced male predominance Barr virus also appears to play a central role in
among NHL subtypes. most NHLs arising after organ transplantation. It
is widely accepted that at least some of the tumors
Infections reflect in vivo outgrowth of EBV-​ immortalized
Epstein Barr Virus lymphoblastoid cell lines as a consequence of failed
Epstein Barr virus (EBV), a ubiquitous human immunosurveillance (Ambinder et al, 1999).
herpesvirus, infects and persists in B lympho- Among HIV-​infected individuals, who suf-
cytes of almost all humans without causing sig- fer from a greatly increased risk of NHL (Engels
nificant disease. However, it encodes a number of et al, 2006), approximately half of the lymphomas
genes that possess oncogenic potential by driving contain monoclonal EBV. This percentage rises
cell proliferation or conferring resistance to cell to nearly 100% if the central nervous system is
death (Grywalska & Rolinski, 2015). EBV was affected. Although the incidence of Burkitt lym-
originally discovered in Burkitt lymphoma by phoma is also significantly increased in HIV-​
investigators who believed that African Burkitt infected individuals in Europe and the United
658

658 Textbook of Cancer Epidemiology

States, the proportion being EBV positive is only Hepatitis C Virus

slightly higher compared to HIV-​negative Burkitt A high prevalence of hepatitis C virus (HCV)
lymphoma patients. Primary effusion lympho- infection has been reported in B-​cell NHL pri-
mas, almost exclusively reported in HIV-​positive marily from Mediterranean countries, Japan, and
individuals, show dual infection with EBV and parts of the United States. For most other coun-
human herpes virus 8. tries the association has been modest or absent
(Pozzato et  al, 2016). The histological types of
Human T-​cell Leukemia/​Lymphoma NHL most often associated with HCV are mar-
Virus 1 ginal zone, lymphoplasmacytic, and diffuse large
Several other infectious agents have been asso- B-​cell lymphomas. Active replication of HCV
ciated with NHL, primarily with rare subtypes. in human B or T cells has never been demon-
Human T-​ cell leukemia/​ lymphoma virus 1 strated. Instead, it has been speculated that the
(HTLV-​1) is considered the cause of adult T-​ neoplastic transformation is determined by
cell leukemia/​lymphoma—​a peripheral T-​cell chronic antigenic stimulation of B cells by viral
type. This malignancy is particularly common surface proteins (De Re et  al, 2000; Rosa et  al,
in Japan and the Caribbean, but also occurs in 2005). As part of the support for a causal role,
other areas endemic for HTLV-​1 such as cen- several studies have provided evidence for remis-
tral Africa, the Middle East, South America, sion of the lymphoma after antiviral therapy.
and Papua New Guinea (IARC, 2012; Ishitsuka The International Agency for Cancer Research
& Tamura, 2014). Main transmission routes for (IARC) in their latest evaluation based on exper-
the virus are mother-​to-​child, sexual intercourse, imental and epidemiological evidence concluded
and blood transfusion. The lifetime risk of devel- that there was sufficient evidence in humans to
oping adult T-​cell leukemia/​lymphoma among consider HCV a cause of NHL (IARC, 2012).
HTLV-​1 infected individuals is 3%–​5% (IARC,
2012). Although uncommon, other NHL types Other Infections
have also been found to be associated with the It is well established that infection with
presence of HTLV-​1, but without sufficient evi- Helicobacter pylori may lead to the development
dence to establish causality (Poiesz et al, 2001). of low-​grade mucosa-​associated lymphoid tissue
(MALT) lymphoma (Kusters et al, 2006). MALT
Human Herpesvirus 8 is an extranodal B-​cell lymphoma. The stomach
Human herpesvirus 8 (HHV8), primarily known is one of the most affected organs by this disease,
as a cause of Kaposi’s sarcoma and therefore also and approximately 90% of gastric MALT lym-
named Kaposi sarcoma–​ associated herpesvi- phomas, are infected with H. pylori. The bacterial
rus (KSHV), is found in most patients with pri- infection may provide a chronic antigenic stimu-
mary effusion lymphoma, a condition rarely seen lation that elicits host immune responses capable
and occurring almost exclusively among HIV-​ of promoting clonal B-​cell expansion. A  causal
positive patients (Moore, 2000). Understanding link is substantiated by studies showing regres-
the pathogenic role of HHV8 in primary effu- sion of the tumor in up to 75% of cases after
sion lymphoma is complicated by the fact that elimination of the bacteria by antibiotic therapy
such lymphoma cells are often also infected with (Foster & Portell, 2015).
EBV. However, the lesions have been shown to An association with hepatitis B virus infection
be monoclonal expansions of a single infected and risk of NHL is has been reported in a several
cell. This finding suggests that HHV8 infection studies, although not consistently, and putative
precedes tumor growth, which again supports an mechanisms remain unclear (Dalia et  al, 2013;
etiologic role of HHV8 in these proliferations. Andersen et  al, 2015). Infection with Borrelia
Furthermore, silencing the latent viral genes burgdorferi may cause a lymphoproliferative con-
encoding vCYC and vFLIP induces apoptosis in dition, a pseudolymphoma, which responds to
primary effusion lymphoma cell lines, adding antibiotic treatment. In rare circumstances this
support for the notion that HHV8 is required for condition might evolve into a primary cutaneous
the survival of these cells (Schulz & Cesarman, B-​cell lymphoma (Willemze et al, 1997). Whereas
2015). There is some evidence that also associates an association between B. burgdorferi and cuta-
HHV8 with the plasma cell variant of multicen- neous B-​cell lymphoma has been noted in sev-
tric Castleman’s disease (MCD), another B-​cell eral European countries, there is little evidence
malignancy (Soulier et al, 1995). of an association in North America (Foster &
659

Non-Hodgkin Lymphoma 659

Portell, 2015). A large Scandinavian case-​control suggests an effect of physical activity on

study found a positive association between anti-​ immune status, but few studies have addressed
Borrelia seropositivity and mantle cell lymphoma this issue in the context of NHL subtypes.
(Schollkopf et al, 2008), but this finding remains So far, no or inverse associations have been
to be confirmed by others. reported with nonoccupational physical activ-
In 2004, a case of immunoproliferative small ity and risk of NHL overall (van Veldhoven,
intestinal disease (a rare MALT lymphoma) was 2011; Teras et al, 2012).
shown to regress on antibiotic treatment, and
subsequent testing of archival tumor material Ionizing Radiation
from this and other patients showed evidence There is no evidence that ionizing radiation
of infection with Campylobacter jejuni (Lecuit et causes NHL (Boice, 1992). Low-​dose radiation
al, 2004). Subsequently a few other case reports from diagnostic x-​ray procedures or from occu-
have occurred in the literature as the only evi- pational exposure has not been associated with
dence for an association (Mesnard et al, 2012; an increased risk. Studies among atomic bomb
Coeuret et al, 2014). Some studies have reported survivors have also been negative (Ron, 1998).
an association between Chlamydia psittaci infec- Associations have been found, albeit inconsist-
tion and the occurrence of ocular adnexal lym- ently, only among subjects who received nearly
phoma based on bacterial presence in both lethal doses of ionizing radiation. Under such
tumor tissue and peripheral blood mononu- circumstances, immunosuppression may play
clear cells and on a positive response to anti- a role, perhaps through elimination of sup-
biotic therapy (Ferreri et al, 2004; Ferreri et al, pressor T cells. This depletion might lead to
2005). A meta-​analysis of 458 cases with ocular unregulated proliferation of B lymphocytes
adnexal lymphomas from 10 different countries and, in rare circumstances, to NHL. In an anal-
suggested wide geographic variations in a possi- ysis of risk of death due to NHL in a cohort
ble association (Husain et al, 2007). In 2016, an of >300,000 radiation-​monitored workers, no
increased risk of NHL, and diffuse large B-​cell significant associations were found (Leurad
lymphoma in particular, was reported among et al, 2015).
individuals with Q fever, infected by Coxiella
burnetii (Melenotte et al, 2016). Occupation
The previous observations indicate that there Reports on the association of NHL with occu-
could be more infectious agents associated with pational exposures are inconsistent, but some
this distinct category of infection-​ associated occupations have repeatedly been associated
lymphoid malignancies, in which the agent does with excess risks (D’Amore et  al, 1992; Scherr
not directly infect and transform lymphoid cells & Mueller, 1996; ‘t Mannetje et al, 2016). These
but rather indirectly increases the probabil- include among farmers, pesticide applica-
ity of lymphoid transformation by chronically tors, grain millers, wood and forestry workers,
stimulating the immune system. Chronic infec- chemists, machinists, printers, and workers in
tions may lead to prolonged antigen stimulation the petroleum, rubber, plastic, and synthetics
and lymphocyte proliferation, likely facilitat- industries.
ing clonal expansion, accumulation of genetic Farmers and general populations of rural
events, and lymphoma development. Certain communities have been most extensively stud-
chronic infectious diseases such as tuberculo- ied (Khuder et al, 1999; Blair & Freeman 2009;
sis, malaria, herpes zoster, chronic ear infection, ‘t Mannetje et al, 2016). In a meta-​analysis of 36
bronchitis, and pyelonephritis have in some, but studies (Khuder et al, 1999), farmers had an over-
not all, studies been associated with an increased all 10% increased risk, primarily noticed among
risk of NHL (Cartwright et al, 1988; Franceschi men, and with studies performed among farm-
et  al, 1989; Bernstein & Ross, 1992; Doody ers residing in the United States showing a more
et  al, 1992; La Vecchia et  al, 1992; Askling & elevated risk. In a pooled analysis of 10 case-​con-
Ekbom, 2001). trol studies including in total 10,046 NHL cases
and 12,025 controls, evaluating occupational
Physical Activity risks with a focus on previously reported asso-
Given the influence of immune dysfunction ciations (‘t Mannetje et al, 2016), increased risks
in the etiology of NHL, physical activity could were confirmed for farming occupations includ-
be an important factor. Laboratory evidence ing field crop-​and vegetable-​but not animal
60

660 Textbook of Cancer Epidemiology

farming, charworkers/​cleaners, spray-​painters, Medical Conditions and Treatment

electric wiremen, carpenters, and women’s hair- Immune Modulation
dressers. Subtype-​specific associations were noted Risk of NHL is strongly related to immune mod-
for diffuse large B-​cell lymphoma in crop farm- ulation. In fact, many researchers believe that
ers, diffuse large B-​cell lymphoma, and CLL in the cause(s) of the previous worldwide increase
women’s hairdressers, and for diffuse large B-​cell in incidence is to be found among factors asso-
lymphoma and peripheral T-​cell lymphoma in ciated with immune modulation. Both primary,
textile workers. The main hypothesis for the occu- or genetic, disorders of immune dysfunction and
pational risk of NHL focuses on exposure to pes- acquired states of severe immunosuppression,
ticides, organic solvents, and allergens, but also including HIV/​AIDS and organ transplantation,
includes, for example, infectious agents, dusts, constitute strong and well-​established risk factors
and metals. for NHL, but explain few new cases in the gen-
A number of studies have addressed more eral population. Clinical features shared by the
directly specific exposures that might be involved majority of immunodeficiency-​related lympho-
in the development of NHL. Exposure to pes- mas are diffuse large B-​cell histology, extranodal
ticides has increased continuously since the location, aggressive clinical course, and an asso-
1940s, and case-​control studies have suggested ciation with EBV.
an increased risk in occupations using phenoxy A greatly increased risk of NHL in conjunc-
acids and chlorophenol herbicides, particu- tion with potent immunosuppressive therapy
larly 2.4-​dichlorophenoxyacetic acid (d’Amore following renal, liver, heart, or bone marrow
et al, 1992; Merhi et al, 2007). However, a meta-​ transplantation has been reported consistently.
analysis of nine studies did not confirm an asso- The risk of NHL increases by about 10-​to 20-​fold
ciation with risk of NHL overall (Goodman et al, in renal allograft recipients and up to about 200-​
2015). Thus, there is still no firm evidence of an fold in heart-​transplant recipients, compared to
association. Also, the mechanism by which pesti- the general population (Opelz & Dohler, 2004).
cides might increase the risk of NHL is not clear, Post-​ transplant lymphoproliferative disorders
but chromosomal rearrangements associated constitute a spectrum ranging from early EBV-​
with pesticide application have been reported driven polyclonal proliferations to EBV-​positive
(Kirsch & Lipkowitz, 1992). (80%–​ 90%) or EBV-​ negative NHL, predomi-
In a literature review, d’Amore et  al (1992) nantly of B-​cell origin. However, some lymphoma
noted that a common feature in several of the subtypes such as mantle cell lymphoma and fol-
industrial exposures identified is their capac- licular lymphoma do not seem to occur with
ity to alter the cell-​mediated immune response. increased frequency among organ transplant
However, experimental evidence of pesticide-​ recipients (Clarke et al, 2013). The pathogenesis
induced immunosuppression has not been sub- of post-​transplant lymphoproliferative disorders
stantiated in human studies. Indeed, when the likely follows at least two distinct pathways, with
exposure to Agent Orange (a mixture of two early disease arising during acute immunosup-
phenoxy herbicides) was investigated among pression driven by primary EBV infection, and
Vietnam veterans, the highest incidence of lym- late disease during sustained immunosuppres-
phoma was found in ground troops stationed in sion and aberrant lymphocyte proliferation (van
areas of lowest exposure and among sailors in Leeuwen et al, 2009). Other modulating factors
navy ships off the coast of Vietnam (Breslin et al, include donor and recipient EBV serologic status,
1988; O’Brien et al, 1991). underlying disease, HLA mismatch, and type,
Exposures to organic solvents and chemi- duration, and intensity of the immunosuppres-
cals including benzene, styrene, 1,3-​butadiene, sive treatment (Gibson et al, 2014; Hussain et al,
trichlorethylene, perchlorethylene, creosote, 2016).
lead arsenate, formaldehyde, and paint thin- In HIV-​infected individuals, prior to anti-
ners have been suggested as possible risk fac- retroviral therapy, the relative risk of NHL was
tors for NHL(Mueller, 1996; Guyton KZ et al, on average increased 100-​fold (Aboulafia et  al,
2014). Benzene achieved early attention, but 2004; Grulich & Vajdic, 2015). In the era of anti-
results have been inconclusive (Savitz and retroviral therapy, with increasing CD4 counts,
Andrews, 1997; Lamm et  al, 2005; Kane & NHL incidence has decreased substantially but
Newton, 2010). still remains about 10-​fold increased. Commonly
61

Non-Hodgkin Lymphoma 661

occuring NHL subtypes include diffuse large B-​ as psoriasis (Boffetta et  al, 2001; Gelfand et  al,
cell lymphoma and Burkitt lymphoma, whereas 2003) and inflammatory bowel disorders, mainly
some other subtypes including follicular lym- Crohn’s disease, (Arseneau et  al, 2001; Askling
phoma and mantle cell lymphoma do not occur et  al, 2005; Baecklund et  al, 2014)  have fre-
with increased frequency (Gibson et  al, 2014), quently, but not invariably, been associated with
similar to the organ transplant setting. The path- increased lymphoma risks. Yet other reported
ogenesis of HIV/​ AIDS-​ related lymphomas is associations with immune-​ mediated disorders
complex and most likely varies according to including myositis, hemolytic anemia, and idi-
NHL subtype. In addition to immunodeficiency, opathic trombocytopenic purpura, where an
risk-​modifying factors include B-​ cell activa- occurrence has been noted closely in time with
tion and uncontrolled HIV viremia (Grulich & the NHL diagnosis, may represent paramalig-
Vajdic 2015). nant autoimmune phenomena rather than causal
associations.
Blood Transfusion It cannot be excluded that immunosuppres-
There is little evidence to support an association sive therapy or biological therapy with tumor
between blood transfusion and NHL. Allogeneic necrosis factor antagonists, often given for auto-
blood transfusions can induce immunosuppres- immune conditions, may cause an increase in
sion and increase susceptibility to infections, the incidence of NHL. There is, however, strong
including infections caused by blood-​ borne evidence to support a causal association with
organisms. Therefore, transfusions have been the underlying disease and its severity (Wolfe
hypothesized to influence the risk of NHL. et al, 1998; Baecklund et al, 2006; Theander et al,
Several cohort studies have found a 1.6-​to three- 2006). Markers of severe Sjögren’s syndrome—​
fold increase in risk, whereas most case-​control such as parotid enlargement, hypocomple-
studies found no association (Chow & Holly, mentemia, palpable purpura, and low T CD4+
2002; Zhang et  al, 2004b). A  large nested case-​ cell counts—​have been associated with higher
control study in Sweden that took advantage of excessed risk of NHL (Nishishinya et  al, 2015).
information on exposure prior to disease out- In a large case-​ control study nested within a
come found no association overall with num- cohort of rheumatoid arthritis patients, medium
ber of transfusions or types of blood transfusion overall disease activity was associated with an 8-​
products (Adami et al, 1997a). Analyses of close fold increase in lymphoma risk, and high activ-
to 1,000,000 transfusion recipients in Denmark ity with a 70-​fold risk increase, compared to low
and Sweden showed a substantially increased disease activity when the confounding effect of
risk immediately following the first transfusion, treatment was taken into account in the analysis
which then decreased sharply. This substantial (Baecklund et al, 2006). Thus, the increased risk
risk increase with short follow-​up most likely of lymphomas appears to be confined to a subset
reflects reverse causality due to anemia in rela- of patients with most severe disease.
tion to incipient NHL. A marginal risk increase In celiac disease, highly increased lym-
was observed between 2 and 10 years of follow-​ phoma risks were reported already decades ago
up, but then disappeared (Hjalgrim et al, 2007). (Selby & Gallagher, 1979), while recent studies
Thus, although an association between NHL and have described lower, albeit still increased, risks
blood transfusion is biologically plausible, the (Green et al, 2003; Ilus et al, 2014). Whether the
epidemiologic evidence is, at most, weak. apparent risk-​level discrepancies between old
and new studies relates to an increasing recogni-
Autoimmune and Chronic tion of mild or silent celiac disease (presumably
Inflammatory Diseases at lower risk of NHL), patients diagnosed at an
The association with several autoimmune disor- earlier age, or better management of diagnosed
ders and NHL is well established. Rheumatoid patients is yet unknown. In inflammatory bowel
arthritis, systemic lupus erythematosus, Sjogren’s disorders, some concern has been raised for an
syndrome, celiac disease, dermatitis herpetifor- increased risk of the rare hepatosplenic T-​cell
mis, and autoimmune thyroiditis have all been lymphoma with use of thiopurines and tumor
consistently associated with an increased risk of necrosis factor therapy, especially in combination
NHL (Smedby et al, 2008; Baecklund et al, 2014). (Subramaniam et al, 2014). In 2014, a large meta-​
Similarly, several inflammatory conditions such analysis concluded that tumor necrosis factor
62

662 Textbook of Cancer Epidemiology

therapy alone did not increase risk of lymphoma antibiotics, nonsteroidal anti-​ inflammatory
(Williams et al, 2014). drugs (NSAIDs) and other analgesics, statins,
With regard to lymphoma subtypes, evi- histamine 2–​ receptor antagonists, psychotro-
dence suggests specific associations of Sjögren’s pic drugs, anticonvulsants, antidepressants or
syndrome, rheumatoid arthritis, and systemic antianxiety drugs, amphetamines, and digitalis
lupus erythematosus, with diffuse large B-​cell or digitoxin (Bernstein & Ross, 1992; Kato et
lymphoma (Baecklund et al, 2014), in addition al, 2002; Iwata et al, 2006). Conversely, perhaps
to the well-​established link between Sjögren’s just as many studies have detected no or even
syndrome and MALT lymphoma in the parotid an inverse association between risk of NHL and
gland (Voulgarelis et al, 1999). Traditionally, these same medications (Beiderbeck et al, 2003;
celiac disease has been linked to an uncommon Zhang et al, 2004c; Chang et al, 2005c; Fortuny et
form of T-​cell lymphoma in the small intestine, al, 2006; Ye et al, 2015a). A reduced risk of NHL
referred to as enteropathy-​type T-​cell lymphoma overall in association with statin use represents
(ETTL). one of the more consistent findings in this area,
but knowledge of subtype-​specific associations is
Subclinical Immune Activation still limited (Ye et al, 2015b). As several diseases,
Biomarker studies conducted within large pro- including acquired immunosuppression, auto-
spective cohorts provide some support for the immune disorders, allergies, and infections also
notion that subclinical immune and inflamma- appear to be associated with NHL risk, it is diffi-
tory activation is a risk factor for NHL. Levels cult to determine whether apparent associations
of circulating cytokines and soluble forms of between medications and lymphoma risk are due
B-​lymphocyte surface receptors have been asso- to the effects of the medications themselves or
ciated with risk of both aggressive and indolent rather the underlying disorders.
NHL (Purdue et  al, 2015; Edlefsen et  al, 2014).
Associations have been strongest among NHL Allergy
cases diagnosed closely in time from blood draw, Atopic conditions including allergy, hay fever,
but have also been observed among cases diag- asthma, and eczema have repeatedly, but not
nosed at least 5 and up to 20  years after blood unanimously, been reported to reduce risk of
sampling, arguing against reverse causation as NHL (Grulich et  al, 2005; Morton et  al, 2014).
the only explanation for these findings. More Most reports of inverse associations come from
research is needed to understand the nature and case-​control studies based on self-​reported med-
timing of these observations. ical history (Morton et  al, 2014). In the large
pooled analysis of 14 case-​control studies within
Vaccinations the InterLymph consortium, hay fever was asso-
Bacille Calmette-​Guérin (BCG) vaccination has ciated with an 18% reduced risk of risk of NHL
been associated with increased, decreased, and no overall (OR 0.82, 95% CI 0.77–​0.88), with inverse
altered risk of NHL (Bernard et al, 1984; Bernstein associations specifically noted for several B-​cell
& Ross, 1992; Becker et al, 2004; Villumsen et al, NHL subtypes including diffuse large B-​ cell
2009). Other vaccinations such as attenuated live lymphoma, CLL/​small lymphocytic lymphoma,
polio, cholera, yellow fever, and influenza vacci- mantle cell lymphoma, and Burkitt lymphoma.
nations have, if anything, been associated with a Specific allergies were also noted to decrease risk
decreased risk of NHL (Bernstein & Ross, 1992; of several B-​ cell NHL subtypes. Associations
Holly et al, 1999; Becker et al, 2004; Lankes et al, remained when excluding occurrence of aller-
2009). However, with respect to both the possible gies within 2  years of the NHL diagnosis; for
influence of childhood vaccinations and vaccina- CLL results were similar when excluding expo-
tions in adults, primarily travel vaccinations, the sures within 10 years of NHL diagnoses (Slager
evidence must be regarded as preliminary and et al, 2014).
inconclusive. Still, there is some concern that these results
could be explained by reverse causality, since a
Prior Use of Medication poor immunologic response to allergens could
The existing literature concerning use of dif- be due to the yet undiagnosed lymphoma. In
ferent drugs and NHL risk is contradictory. line with this, a large Scandinavian case-​control
Several studies have found a significantly ele- study found patients with NHL to be signifi-
vated risks of NHL in association with use of cantly less likely to mount IgE antibodies against
63

Non-Hodgkin Lymphoma 663

specific allergens, and there was further a statis- UV-​ induced systemic immune modulation
tically significant inverse association between the (Norval, 2001)  or photoactivation of vitamin
dissemination of NHL disease at time of blood D production (Holick, 2003). A  large number
drawing and seropositivity to specific IgE. These of studies have followed, investigating ambient
latter results could thus indicate that persons or self-​reported UVR levels as well as different
with widespread lymphoma of B-​cell but not T-​ measures of vitamin D and risk of NHL overall
cell origin might have a reduced ability to mount and subtypes. Although findings are not entirely
a response to allergens (Melbye et  al, 2007). consistent, most observations support a possible
Hence, the hypothesis that atopic conditions and protective effect of UVR on lymphomagenesis
allergies is causally related to lymphomagenesis (Kricker et al, 2008; Cahoon et al, 2015). Inverse
has to be further investigated. associations have been described for several
NHL subtypes, but most often for diffuse large B-​
Other Medical Conditions cell lymphoma (Cahoon et al, 2015).
An association between silicone breast implants UVR-​induced effects on immunity are incom-
and anaplastic large-​cell lymphoma of the breast pletely known, but appear to include a downreg-
was first reported in 1997 (Keech & Creech, ulation of antigen-​presenting activity, increased
1997). Since then, a number of cases have been levels of suppressor T-​cells, and a shift toward
reported and a causal relationship is likely. a Th2-​type response (Norval, 2001). Some evi-
Silicone implants are associated with mild to dence of weak inverse associations between dif-
severe scarring of the surrounding tissues. In ferent measures of vitamin D and risk of NHL
some cases, the inflammation may persist and have been provided from prospective studies
progress to monoclonal infiltrates and eventually (Giovannucci et  al, 2006; Ordonez-​Mena et  al,
lymphoma (Bizjak et al, 2015). In this regard, the 2016), but other studies have shown no associa-
pathogenic model may be similar to the develop- tion (Luczynska et al, 2013).
ment of, for example, marginal zone lymphoma
in Sjögren’s syndrome and gastric MALT lym- Hair Dyes
phoma following Helicobacter jejuni infection. The evidence for an association between use of
hair dyes and the risk of NHL is not convincing.
Other Risk Factors However, compounds in hair dyes may be muta-
Ultraviolet Light genic and carcinogenic in laboratory experi-
Numerous studies have described increased risks ments (Ames et  al, 1975), and some products
of malignant lymphomas, including chronic lym- have indeed been removed from the market over
phocytic leukemia, following a diagnosis of skin the past 20  years. The epidemiologic literature
cancer (malignant melanoma, squamous cell car- on this topic is mixed. Several large studies have
cinoma, and basal cell carcinoma). Conversely, not documented an increased risk, whereas oth-
an increased risk of all three forms of skin cancer ers, mostly small, showed a positive association
has been noted following a history of lymphoma (Holly et al, 1998; Zhang et al, 2004a; Benavente
(Frisch & Melbye, 1995; Lens & Newton-​Bishop, et al, 2005; Mendelsohn et al, 2009; Parodi et al,
2005; Famenini et al, 2015). These observations, 2016). In three other large case-​control stud-
along with parallel time trends in incidence of ies (Zhang et  al, 2004a, de Sanjosé et  al, 2006;
skin cancer and NHL, gave rise to the formerly Sangrajrang et al, 2011), a small increased risk of
popular hypothesis that ultraviolet radiation NHL was noted predominantly in relation to hair
(UVR) exposure would increase the risk not only dye use before 1980.
of skin cancer but also of NHL (Zheng et al, 1992;
Melbye et al, 1996). CONCLUSION
However, when investigators began to ana- For several decades the incidence of NHL
lyze this relationship directly, rather the contrary increased more rapidly than that of almost any
has been observed. In two large case-​ control other cancer. Well-​established risk factors such
studies from Australia and Scandinavia, consist- as immunosuppression, autoimmunity/​ inflam-
ent inverse associations between various meas- mation, and certain infectious agents explain at
ures of frequent UVR exposure and risk of NHL most only a small fraction of the cases, and thus
were observed (Hughes et al, 2004; Smedby et al, cannot explain the past increase. However, more
2005). Possible mechanisms behind an inverse observations indicate that high body mass index
association between UVR and NHL risk include and smoking may have contributed to some
64

664 Textbook of Cancer Epidemiology

extent (Table 27-​1). Other possible risk associa- Andersen ES, Omland LH, Jepsen P, Krarup H,
tions with particular food products, medications, Christensen PB, Obel N et  al. Risk of all-​type
pesticides, and hair dyes do not satisfactorily cancer, hepatocellular carcinoma, non-​Hodgkin
explain the observed trends. Although, the lymphoma and pancreatic cancer in patients
strong connection between immune modula- infected with hepatitis B virus. J Viral Hepat
tion (deficiency as well as activation) and NHL 2015;10:828–​34.
risk suggests that we should keep searching for Arseneau KO, Stukenborg GJ, Connors AF Jr,
factors that influence immune modulation, new Cominelli F. The incidence of lymphoid and
myeloid malignancies among hospitalized
hypotheses are needed in order to reveal the eti-
Crohn’s disease patients. Inflamm Bowel Dis
ology of NHL subtypes.
2001;7:106–​12.
Evaluations of epidemiologic patterns of
Askling J, Ekbom A. Risk of non-​Hodgkin’s lymphoma
NHL according to histologic subtype have doc- following tuberculosis. Br J Cancer 2001;84:113–​5.
umented both commonalities and clear differ- Askling J, Brandt L, Lapidus A, Karlen P, Bjorkholm
ences among the different NHL subtypes. These M, Lofberg R et al. Risk of haematopoietic cancer
findings strongly suggest that they may also have in patients with inflammatory bowel disease. Gut
different, but sometimes overlapping, risk factor 2005;54:617–22.
profiles. New leads with regard to mechanisms Baecklund E, Smedby KE, Sutton LA, Askling J,
and pathways of lymphomagenesis will likely be Rosenquist R. Semin Cancer Biol 2014;61–​70.
generated through continued studies of genetic Becker N, Deeg E, Nieters A. Population-​based study
susceptibility and gene–​environment interaction. on lymphoma in Germany: rationale, study design
Such studies also provide fertile ground for new and first results. Leuk Res 2004;28:713–​24.
hypotheses of previously unrecognized environ- Beiderbeck AB, Holly EA, Sturkenboom MC,
mental agents that could be involved in NHL eti- Coebergh JW, Stricker BH, Leufkens HG.
ology. Ongoing research initiatives have already Prescription medications associated with a
led to an increase in our understanding of some decreased risk of non-​Hodgkin’s lymphoma. Am
of the factors involved in NHL pathogenesis and J Epidemiol 2003;157:510–​6.
shown that associations often differ according Bell DA, Liu Y, Cortopassi GA. Occurrence of BCL-​2
to NHL subtype. There still remain more ques- oncogene translocation with increased frequency
tions than answers in relation to the etiology of in the peripheral blood of heavy smokers. J Natl
NHL, but the number of research groups dedi- Cancer Inst 1995;87:223–​4.
Benavente Y, Garcia N, Domingo-​ Domenech E,
cated to NHL research has never been higher
Alvaro T, Font R, Zhang Y et  al. Regular use of
or more productive, which gives hope that our
hair dyes and risk of lymphoma in Spain. Int J
understanding of NHL may greatly improve in
Epidemiol 2005;34:1118–​22.
the coming years. Bernard SM, Cartwright RA, Bird CC, Richards ID,
Lauder I, Roberts BE. Aetiologic factors in lym-
REFERENCES phoid malignancies:  a case-​control epidemiolog-
Aboulafia DM, Pantanowitz L, Dezube BJ. AIDS-​ ical study. Leuk Res 1984;8:681–​9.
related non-​Hodgkin lymphoma: still a problem Bernstein L, Ross RK. Prior medication use and
in the era of HAART. AIDS Read 2004;14:605–​17. health history as risk factors for non-​Hodgkin’s
Adami J, Nyrén O, Bergström R, Ekbom A, lymphoma:  preliminary results from a case-​
McLaughlin JK, Hogman C et al. Blood transfu- control study in Los Angeles County. Cancer Res
sion and non-​Hodgkin lymphoma: lack of asso- 1992;52:5510S–​5S.
ciation. Ann Intern Med 1997a;127:365–​71. Berndt S, Camp N, Skibola S, Vijai J, Wang Z, Gu J
Adami HO, Tsaih S, Lambe M, Hsieh C, Adami J, et  al. Meta-​analysis of genome-​wide association
Trichopoulos D et al. Pregnancy and risk of non-​ studies discovers multiple loci for chronic lym-
Hodgkin’s lymphoma:  a prospective study. Int J phocytic leukemia. Nat Commun 2016;7:10933.
Cancer 1997b;70:155–​8. Besson H, Brennan P, Becker N, Nieters A, De Sanjose
Ambinder RF, Lemas MV, Moore S, Yang J, Fabian S, Font R et al. Tobacco smoking, alcohol drinking
D, Krone C. Epstein-​Barr virus and lymphoma. and non-​Hodgkin’s lymphoma: a European multi-
Cancer Treat Res 1999;99:27–​45. center case-​control study (Epilymph). Int J Cancer
Ames BN, Kammen HO, Yamasaki E. Hair dyes are 2006;119:901–​8.
mutagenic:  identification of a variety of muta- Bizjak M, Selmi C, Praprotnik S, Bruck O, Perricone
genic ingredients. Proc Natl Acad Sci USA C, Ehrenfeld M, Shoenfeld Y. Silicone implants
1975;72:2423–​7.
65

Non-Hodgkin Lymphoma 665

and lymphoma:  the role of inflammation. J Chiu BCH, Cerhan JR, Gapstur SM, Sellers TA, Zheng
Autoimmun 2015;65:64–​73. W, Lutz CT et al. Alcohol consumption and non-​
Blair A, Freeman LB. Epidemiologic studies in agri- Hodgkin lymphoma in a cohort of older women.
cultural populations:  observations and future Br J Cancer 1999;80:1476–​82.
directions. J Agromedicine 2009;14:125–​31. Chow EJ, Holly EA. Blood transfusions and
Boffetta P, Gridley G, Lindelof B. Cancer risk in a non-​Hodgkin’s lymphoma. Epidemiol Rev
population-​ based cohort of patients hospital- 2002;24:269–​79.
ized for psoriasis in Sweden. J Invest Dermatol Clarke CA, Morton LM, Lynch C, Pfeiffer RM, Hall
2001;117:1531–​7. EC, Gibson TM et al. Risk of lymphoma subtypes
Boice JD Jr. Radiation and non-​Hodgkin’s lymphoma. after solid organ transplantation in the United
Cancer Res 1992;52:5489S–​91S. States. Br J Cancer 2013;109:280–​8.
Breslin P, Kang HK, Lee Y, Burt V, Shepard BM. Coeuret S, de La Blanchardière A, Saguet-​Rysanek
Proportionate mortality study of US Army and V, Chèze S, Tavernier M, Arsène D et  al.
US Marine Corps veterans of the Vietnam War. J Campylobacter coli cultured from the stools of a
Occup Med 1988;30:412–​9. patient with immunoproliferative small intestinal
Cahoon EK, Pfeiffer RM, Wheeler DC, Arhancet J, disease. Clin Microbiol Infect. 2014;20:908–​11.
Lin SW, Alexander BH et al. Relationship between Costas L, de Sanjosé S, Infante-​Rivard C. Reproductive
ambient ultraviolet radiation and non-​Hodgkin factors and non-​Hodgkin lymphoma: a systematic
lymphoma subtypes:  a U.S.  population-​ based review. Crit Rev Oncol Hematol 2014;92:181–​93.
study of racial and ethnic groups. Int J Cancer Cross AJ, Ward MH, Schenk M, Kulldorff M, Cozen
2015;136:E432–​41. W, Davis S et al. Meat and meat-​mutagen intake
Cartwright RA, McKinney PA, O’Brien C, Richards and risk of non-​Hodgkin lymphoma: results from
ID, Roberts B, Lauder I et al. Non-​Hodgkin’s lym- a NCI-​SEER case-​control study. Carcinogenesis
phoma:  case control epidemiological study in 2006;27:293–​7.
Yorkshire. Leuk Res 1988;12:81–​8. D’Amore F, Hasle H, Hansen KS. Occupational expo-
Castillo JJ, Ingham RR, Reagan JL, Furman M, Dalia S, sures and non-​ Hodgkin’s lymphoma. Hematol
Mitri J. Obesity is associated with increased rela- Rev 1992;6:183–​99.
tive risk of diffuse large B-​cell lymphoma: a meta-​ Dalia S, Chavez J, Castillo JJ, Sokol L. Hepatitis B
analysis of observational studies. Clin Lymphoma infection increases the risk of non-​Hodgkin lym-
Myeloma Leuk 2014;14:122–​30. phoma:  a meta-​analysis of observational studies.
Cerhan JR, Berndt SI, Vijai J, Ghesquières H, McKay Leuk Res 2013;37:1107–​15.
J, Wang SS et al. Genome-​wide association study Daniel CR, Sinha R, Park Y, Graubard BI, Hollenbeck
identifies multiple susceptibility loci for dif- AR, Morton LM, Cross AJ. Meat intake is not
fuse large B cell lymphoma. Nat Genet 2014 associated with risk of non-​Hodgkin lymphoma
Nov;46(11):1233–​8. doi:10.1038/​ng.3105. Epub in a large prospective cohort of U.S.  men and
2014 Sep 28. women. J Nutr 2012;142:1074–​80.
Cerhan JR, Slager SL. Familial predisposition De Re V, De Vita S, Marzotto A, Rupolo M, Gloghini
and genetic risk factors for lymphoma. Blood A, Pivetta B et  al. Sequence analysis of the
2015;126:2265–​73. immunoglobulin antigen receptor of hepatitis
Chang ET, Smedby KE, Zhang SM, Hjalgrim H, C virus-​ associated non-​ Hodgkin lymphomas
Melbye M, Ost A et  al. Alcohol intake and risk suggests that the malignant cells are derived
of non-​Hodgkin lymphoma in men and women. from the rheumatoid factor-​producing cells that
Cancer Cause Control 2004;15:1067–​76. occur mainly in type II cryoglobulinemia. Blood
Chang ET, Smedby KE, Zhang SM, Hjalgrim H, 2000;96:3578–​84.
Melbye M, Ost A et  al. Dietary factors and De Sanjosé S, Benavente Y, Nieters A, Foretova L,
risk of non-​ Hodgkin lymphoma in men and Maynadié M, Cocco PL et al. Association between
women. Cancer Epidemiol Biomarkers Prev personal hair dyes and lymphoid neoplasms in
2005b;14:512–​20. Europe. Am J Epidemiol 2006;164:47–​55.
Chang ET, Smedby KE, Hjalgrim H, Schollkopf Devesa SS, Fears T. Non-​Hodgkin’s lymphoma time
C, Porwit-​ MacDonald A, Sundstrom C et  al. trends: United States and international data.
Medication use and risk of non-​Hodgkin’s lym- Cancer Res 1992;52:5432S–​40S.
phoma. Am J Epidemiol 2005c;162:965–​74. Diver WR, Patel AV, Thun MJ, Teras LR, Gapstur
Chiu BCH, Cerhan JR, Folsom AR, Sellers TA, Kushi SM. The association between cigarette smok-
LH, Wallace RB et  al. Diet and risk of non-​ ing and non-​ Hodgkin lymphoid neoplasms in
Hodgkin lymphoma in older women. JAMA a large US cohort study. Cancer Causes Control
1996;275:1315–​21. 2012;23:1231–​40.
6

666 Textbook of Cancer Epidemiology

Doody MM, Linet MS, Glass AG, Friedman GD, non-​ Hodgkin’s lymphoma in the north-​ east of
Pottern LM, Boice JD Jr et  al. Leukemia, lym- Italy:  a hospital-​based case-​control study. Leuk
phoma, and multiple myeloma following selected Res 1989;13:465–​72.
medical conditions. Cancer Causes Control Freedman DS, Tolbert PE, Coates R, Brann EA,
1992;3:449–​56. Kjeldsberg CR. Relation of cigarette smoking to
Edlefsen KL, Martínez-​ Maza O, Madeleine MM, non-​ Hodgkin’s lymphoma among middle-​ aged
Magpantay L, Mirick DK, Kopecky KJ et  al. men. Am J Epidemiol 1998;148:833–​41.
Cytokines in serum in relation to future non-​ Frisch M, Melbye M. New primaries after squamous
Hodgkin lymphoma risk: evidence for associations cell skin cancer. Am J Epidemiol 1995;141:916–​22.
by histologic subtype. Int J Cancer 2014;135:913–​22. Frisch M, Pedersen BV, Wohlfahrt J, Hjalgrim H,
Engels EA, Pfeiffer RM, Goedert JJ, Virgo P, McNeel Biggar RJ. Reproductive patterns and non-​
TS, Scoppa SM, Biggar RJ Trends in cancer risk Hodgkin lymphoma risk in Danish women and
among people with AIDS in the United States men. Eur J Epidemiol 2006;21:673–​9.
1980-2002. HIV/AIDS Cancer Match Study. AIDS Gelfand JM, Berlin J, Van Voorhees A, Margolis DJ.
2006;20:1645–54 Lymphoma rates are low but increased in patients
Epstein MA, Achong BG, Barr YM. Virus particles in with psoriasis:  results from a population-​based
cultured lymphoblasts from Burkitt lymphoma. cohort study in the United Kingdom. Arch
Lancet 1964;1:702–​3. Dermatol 2003;139:1425–​9.
Famenini S, Martires KJ, Zhou H, Xavier MF, Wu JJ. Gibson TM, Engels EA, Clarke CA, Lynch CF,
Melanoma in patients with chronic lymphocytic Weisenburger DD, Morton LM. Risk of diffuse
leukemia and non-​ Hodgkin lymphoma. J Am large B-​cell lymphoma after solid organ trans-
Acad Dermatol 2015;72:78–​84. plantation in the United States. Am J Hematol
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser 2014;89:714–​20.
S, Mathers C, Rebelo M, Parkin DM, Forman D, Giovannucci E, Liu Y, Rimm EB, Hollis BW, Fuchs CS,
Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence Stampfer MJ et al. Prospective study of predictors
and Mortality Worldwide: IARC CancerBase No. of vitamin D status and cancer incidence and mor-
11 [Internet]. Lyon, France: International Agency tality in men. J Natl Cancer Inst 2006;98:451–​9.
for Research on Cancer; 2013. Available from: Goodman JE, Loftus CT, Zu K. 2,4-​
http://globocan.iarc.fr, accessed 2017. Dichlororphenoxyacetic acid and non-​ Hodgkin’s
Fernberg P, Odenbro A, Bellocco R, Boffetta P, lymphoma, gastric cancer and prostate can-
Pawitan Y, Adami J. Tobacco use, body mass index cer:  meta-​analyses of the published literature.
and the risk of malignant lymphomas—​a nation- 2015;25:626–​636.
wide cohort study in Sweden. Int J Cancer 2006 Green J, Cairns BJ, Casabonne D, Wright FL, Reeves
May 1;118(9):2298–​302. G, Beral V; Million Women Study collabora-
Ferreri J, Guidoboni M, Ponzoni M, De Concilliis tors. Height and cancer incidence in the Million
C, Dell’Oro S, Fleischhauer K et  al. Evidence for Women Study:  prospective cohort, and meta-​
an association between Chlamydia psittaci ad analysis of prospective studies of height and total
ocular adnexal lymphomas. J Natl Cancer Inst cancer risk. Lancet Oncol 2011;12:785–​94.
2004;96:586–​94. Green PH, Fleischauer AT, Bhagat G, Goyal R, Jabri
Ferreri J, Ponzoni M, Guidoboni M, De Conciliis B, Neugut AI. Risk of malignancy in patients with
C, Resti AG, Mazzi B et  al. Regression of ocular celiac disease. Am J Med 2003;115:191–​5.
adnexal lymphoma after Chlamydia psittaci–​ Grooves FD, Linet MS, Travis LB, Devesa SS. Cancer
eradicating antibiotic therapy. J Clin Oncol surveillance series:  non-​ Hodgkin’s lymphoma
2005;23:5067–​73. incidence by histologic subtype in the United
Fortuny J, de Sanjose S, Becker N, Maynadie M, States from 1978 through 1995. J Natl Cancer Inst
Cocco PL, Staines A et  al. Statin use and risk of 2000;92:1240–​51.
lymphoid neoplasms:  results from the European Grulich AE, Vajdic CM, Kaldor JM, Hughes AM,
case-​control study EPILYMPH. Cancer Epidemiol Kricker A, Fritschi L et al. Birth order, atopy, and
Biomarkers Prev 2006;15:921–​5. risk of non-​Hodgkin lymphoma. J Natl Cancer
Foster LH1, Portell CA. The role of infectious agents, Inst 2005;97:587–​94.
antibiotics, and antiviral therapy in the treatment Grulich AE, Vajdic CM. The epidemiology of can-
of extranodal marginal zone lymphoma and other cers in human immunodeficiency virus infection
low-​grade lymphomas. Curr Treat Options Oncol and after organ transplantation. Semin Oncol
2015;16:28. 2015;42:247–​57.
Franceschi S, Serraino D, Bidoli E, Talamini R, Guyton KZ, Hogan KA, Scott CS, Cooper GS, Bale
Tirelli U, Carbone A et  al. The epidemiology of AS, Kopylev L et  al. Human health effects of
67

Non-Hodgkin Lymphoma 667

tetrachloroethylene:  key findings and scientific Evaluation of Carcinogenic Risks to Humans, Vol.
issues. Environ Health Perspect 2014;122:325–​34. 100B. Lyon: IARC Press, 2012.
Grywalska E, Rolinski J. Epstein-​Barr virus-​associated Ilus T, Kaukinen K, Virta LJ, Pukkala E, Collin P.
lymphomas. Semin Oncol 2015;42:291–​303. Incidence of malignancies in diagnosed celiac
Hartge P, Devesa SS. Quantification of the impact patients:  a population-​ based estimate. Am J
of known risk factors on time trends in non-​ Gastroenterol 2014;109:1471–​7.
Hodgkin’s lymphoma incidence. Cancer Res Ishitsuka K, Tamura K. Human T-​cell leukaemia virus
1992;52:5566S–​9S. type I  and adult T-​ cell leukaemia-​ lymphoma.
Heinen MM, Verhage BA, Schouten LJ, Goldbohm Lancet Oncol 2014;15:e517–​26.
RA, Schouten HC, van den Brandt PA. Alcohol Iwata H, Matsuo K, Hara S, Takeuchi K, Aoyama
consumption and risk of lymphoid and myeloid T, Murashige N et  al. Use of hydroxy-​methyl-​
neoplasms:  results of the Netherlands cohort glutaryl coenzyme A reductase inhibitors is asso-
study. Int J Cancer 2013;133:1701–​12. ciated with risk of lymphoid malignancies. Cancer
Hjalgrim H, Frisch M, Begtrup K, Melbye M. Recent Sci 2006;97:133–​8.
increase in the incidence of Non-​ Hodgkin’s Kane EV, Newton R. Benzene and the risk of non-​
lymphoma among young men and women in Hodgkin lymphoma:  a review and meta-​analysis
Denmark. Br J Cancer 1996;73:951–​4. of the literature. Cancer Epidemiol 2010;34:7–​12.
Hjalgrim H, Edgren G, Rostgaard K, Reilly M, Tran Kane EV, Roman E, Becker N, Bernstein L, Boffetta
TN, Titlestad KE et  al. Cancer incidence in P, Bracci PM et  al; InterLymph Consortium.
blood transfusion recipients. J Natl Cancer Inst Menstrual and reproductive factors, and hormonal
2007;99:1864–​74. contraception use: associations with non-​Hodgkin
Hjalgrim H, Friborg J, Melbye M. The epidemi- lymphoma in a pooled analysis of InterLymph
ology of Epstein-​ Barr virus and its associ- case-​control studies. Ann Oncol 2012;23:2362–​74.
ation with malignant disease. In:  Arvin A, Kane EV, Bernstein L, Bracci PM, Cerhan JR, Costas
Campadielli-​ Fiume G, Mocarski E, Moore P, L, Dal Maso L et  al; InterLymph Consortium.
Roizman B, Whitley R, Yamanishi K (eds.) Postmenopausal hormone therapy and non-​
Human Herpesviruses:  Biology, Therapy and Hodgkin lymphoma: a pooled analysis of InterLymph
Immunoprophylaxis. Cambridge:  Cambridge case-​control studies. Ann Oncol 2013;24:433–​41.
University Press, 2006. Kato I, Koenig KL, Shore RE, Baptiste MS, Lillquist
Holick, MF. Vitamin D: A millennium perspective. J PP, Frizzera G et al. Use of anti-​inflammatory and
Cell Biochem 2003;88:296–​307. non-​ narcotic analgesic drugs and risk of non-​
Holly EA, Lele C, Bracci PM. Hair-​color products and Hodgkin’s lymphoma (NHL) (United States).
risk for non-​Hodgkin’s lymphoma: a population-​ Cancer Causes Control 2002;13:965–​74.
based study in the San Francisco Bay Area. Am J Keech JA Jr, Creech BJ. Anaplastic T-​cell lymphoma
Public Health 1998;88:1767–​73. in proximity to a saline-​filled breast implant. Plast
Holly EA, Lele C, Bracci PM, McGrath MS. Case-​ Reconstr Surg 1997;100:554–​5.
control study of non-​ Hodgkin’s lymphoma Kelemen LE, Cerhan JR, Lim U, Davis S, Cozen W,
among women and heterosexual men in the San Schenk M et al. Vegetables, fruit, and antioxidant-​
Francisco Bay Area, California. Am J Epidemiol related nutrients and risk of non-​Hodgkin lym-
1999;150:375–​89. phoma: a National Cancer Institute Surveillance,
Hughes AM, Armstrong BK, Vajdic CM, Turner J, Epidemiology, and End Results population-​
Grulich AE, Fritschi, L et  al. Sun exposure may based case-​ control study. Am J Clin Nutr
protect against non-​Hodgkin lymphoma:  a case-​ 2006;83:1401–​10.
control study. Int J Cancer 2004;112:865–​71. Kelly GL, Stylianou J, Rasaiyaah J, Wei W, Thomas
Husain A, Roberts D, Pro B, McLaughlin P, Esmaeli W, Croom-​Carter D et  al. Different patterns of
B. Meta-​ analyses of the association between Epstein-​ Barr virus latency in endemic Burkitt
Chlamydia psittaci and ocular adnexal lymphoma lymphoma (BL) lead to distinct variants within
and the response of ocular adnexal lymphoma to the BL-​ associated gene expression signature. J
antibiotics. Cancer. 2007;110:809–​15. Virol 2013;87:2882–​94.
Hussain SK, Makgoeng SB, Everly MJ, Goodman Koff JL, Chihara D, Phan A, Nastoupil LJ, Williams
MT, Martínez-​Maza O, Morton LM et  al. HLA JN and Flowers CR. To each its own: linking the
and risk of diffuse large B-​cell lymphoma after biology and epidemiology of NHL subtypes. Curr
solid organ transplantation. Transplantation Hematol Malig Rep 2015;10:244–​255.
2016;100:2453–​60. Khuder SA, Mutgi AB, Schaub EA, Tano BD. Meta-​
International Agency for Research on Cancer (IARC). analysis of Hodgkin’s disease among farmers.
Biological agents. IARC Monographs on the Scand J Work Environ Health 1999;25:436–​41.
68

668 Textbook of Cancer Epidemiology

Kirsch IR, Lipkowitz S. A measure of genomic insta- prospective cohort of Chinese women. Cancer Sci
bility and its relevance to lymphomagenesis. 2009;100:1088–​91.
Cancer Res 1992;52:5545S–​6S. Merhi M, Raynal H, Cahuzac E, Vinson F, Cravedi
Kricker A, Armstrong BK, Hughes A-​M, Goumas C, JP, Gamet-​ Payrastre L. Occupational exposure
Smedby KE, Zheng T et al. Personal sun exposure to pesticides and risk of hematopoietic can-
and risk of non-​Hodgkin lymphoma:  a pooled cers: meta-​analysis of case-​control studies. Cancer
analysis from the InterLymph Consortium. Int J Causes Control 2007;18:1209–​26.
Cancer 2008;122:144–​54. Mesnard B, De Vroey B, Maunoury V, Lecuit M.
Kusters JG, van Vliet AH, Kuipers EJ. Pathogenesis of Immunoproliferative small intestinal disease asso-
Helicobacter pylori infection. Clin Microbiol Rev ciated with Campylobacter jejuni. Dig Liver Dis
2006;19:449–​90. 2012;44:799–​800.
La Vecchia C, Negri E, Franceschi S. Medical his- Moore P. The emergence of Kaposi’s sarcoma-​
tory and the risk of non-​Hodgkin’s lymphomas. associated herpesvirus (human herpesvirus 8). N
Cancer Epidemiol Biomarkers Prev 1992;1:533–​6. Engl J Med 2000;343:1411–​3.
Lamm SH, Engel A, Byrd DM. Non-​ Hodgkin Morton LM, Schenk M, Hein DW, Davis S, Zahm
lymphoma and benzene exposure:  a system- SH, Cozen W et  al. Genetic variation in N-​
atic literature review. Chem Biol Interact acetyltransferase 1 (NAT1) and 2 (NAT2) and
2005;153–​54:231–​7. risk of non-​Hodgkin lymphoma. Pharmacogenet
Lankes HA, Fought AJ, Evens AM, Weisenburger Genomics 2006a;16:537–​45.
DD, Chiu BC. Vaccination history and risk of Morton LM, Wang SS, Devesa SS, Hartge P,
non-​Hodgkin lymphoma:  a population-​based, Weisenburger DD, Linet MS. Lymphoma inci-
case-​control study. Cancer Causes Control dence patterns by WHO subtype in the United
2009;20:517–​23. States, 1992–​2001. Blood 2006b;107:265–​76.
Lecuit M, Abachin E, Martin A, Poyart C, Pochart P, Morton LM, Slager SL, Cerhan JR, Wang SS, Vajdic CM,
Suarez F et al. Immunoproliferative small intesti- Skibola CF et  al. Etiologic heterogeneity among
nal disease associated with Campylobacter jejuni. non-​Hodgkin lymphoma subtypes: the InterLymph
N Eng J Med 2004;350:239–​48. Non-​Hodgkin Lymphoma Subtypes Project. J Natl
Lens, MB, Newton-​ Bishop, JA. An association Cancer Inst Monogr 2014;2014:130–​44.
between cutaneous melanoma and non-​Hodgkin’s Mueller NE. Hodgkin’s disease. In:  Schottenfeld D,
lymphoma: pooled analysis of published data with Fraumeni JJr (eds) Cancer Epidemiology and
a review. Ann Oncol 2005;16:460–​5. Prevention. Oxford:  Oxford University Press,
Łuczyńska A, Kaaks R, Rohrmann S, Becker 1996: 893–​919.
S, Linseisen J, Buijsse B et  al. Plasma 25-​ Nishishinya MB, Pereda CA, Muñoz-​ Fernández S,
hydroxyvitamin D concentration and lym- Pego-​ Reigosa JM, Rúa-​ Figueroa I, Andreu JL,
phoma risk:  results of the European Prospective et  al. Identification of lymphoma predictors in
Investigation into Cancer and Nutrition. Am J patients with primary Sjögren’s syndrome:  a sys-
Clin Nutr 2013;98:827–​38. tematic literature review and meta-​ analysis.
Linet MS, Vajdic CM, Morton LM, de Roos AJ, Skibola Rheumatol Int 2015;35:17–​26.
CF, Boffetta P et al. Medical history, lifestyle, fam- Norval, M. Effects of solar radiation on the human
ily history, and occupational risk factors for fol- immune system. J Photochem Photobiol B
licular lymphoma: the InterLymph Non-​Hodgkin 2001;63:28–​40.
Lymphoma Subtypes Project. J Natl Cancer Inst O’Brien TR, Decoufle P, Boyle CA. Non-​Hodgkin’s
Monogr 2014;48:26–​40. lymphoma in a cohort of Vietnam veterans. Am J
Melenotte C, Million M, Audoly G, Gorse A, Public Health 1991;81:758–​60.
Dutronc H, Roland G et  al. B-​cell non-​Hodgkin Ollberding NJ, Aschebrook-​Kilfoy B, Caces DB, Smith
lymphoma linked to Coxiella burnetii. Blood SM, Weisenburger DD, Chiu BC. Dietary patterns
2016;127:113–​21. and the risk of non-​Hodgkin lymphoma. Public
Melbye M, Adami HO, Hjalgrim H, Glimelius B. Health Nutr 2014;17:1531–​7.
Ultraviolet light and non-​Hodgkin’s lymphoma. Opelz G, Dohler B. Lymphomas after solid organ
Acta Oncol 1996;35:655–​57. transplantation:  a collaborative transplant study
Melbye M, Smedby KE, Lehtinen T, Rostgaard K, report. Am J Transplant 2004;4:222–​30.
Glimelius B, Munksgaard L et al. Atopy and risk Ordóñez-​Mena JM, Schöttker B, Fedirko V, Jenab M,
of non-​Hodgkin lymphoma. J Natl Cancer Inst Olsen A, Halkjær J et  al. Pre-​diagnostic vitamin
2007;99:158–​66. D concentrations and cancer risks in older indi-
Mendelsohn JB, Li QZ, Ji BT, Shu XO, Yang G, Li HL viduals: an analysis of cohorts participating in the
et al. Personal use of hair dye and cancer risk in a CHANCES consortium. Eur J Epidemiol 2016
69

Non-Hodgkin Lymphoma 669

Mar;31(3):311–​23. doi:10.1007/​s10654-​015-​0040-​ from next generation sequencing studies. Semin

7. Epub 2015 May 15. Cancer Biol 2013;23:422–​30.
Pallesen G, Hamilton Dutoit SJ. The association of Sandin S, Hjalgrim H, Glimelius B, Rostgaard K,
Epstein-​ Barr virus (EBV) with T cell lympho- Pukkala E, Askling J. Incidence of non-​Hodgkin
proliferations and Hodgkin’s disease:  two new lymphomas in Sweden, Denmark, and Finland—​
developments in the EBV field. Adv Cancer Res an epidemic that was. Cancer Epidemiol
1993;62:179–​238. Biomarkers Prev 2006;15:1295–​300.
Parodi S, Santi I, Marani E, Casella C, Puppo A, Sangrajrang S, Renard H, Kuhaprema T, Pornsopone
Garrone E et al. Lifestyle factors and risk of leuke- P, Arpornwirat W, Brennan P. Personal use of
mia and non-​Hodgkin’s lymphoma: a case-​control hair dyes increased risk of non-​Hodgkin’s lym-
study. Cancer Causes Control 2016;27:367–​75. phoma in Thailand. Asian Pac J Cancer Prev
Poiesz BJ, Papsidero LD, Ehrlich G, Sherman M, Dube 2011;12:2393–​6.
S, Poiesz M et al. Prevalence of HTLV-​I-​associated Savitz DA, Andrews KW. Review of epidemiologic
T-​cell lymphoma. Am J Hematol 2001;66:32–​8. evidence on benzene and lymphatic and hema-
Purdue MP, Bassani DG, Klar NS, Sloan M, Kreiger topoietic cancers. Am J Ind Med 1997;31:287–​95.
N, the Canadian Cancer Registries Epidemiology Scherr PA, Mueller NE. Non-​Hodgkin’s lymphomas.
Research Group. Dietary factors and risk of non-​ In:  Schottenfeld D, Fraumeni JFJr (Eds) Cancer
Hodgkin lymphoma by histologic subtype: a case-​ Epidemiology and Prevention. Oxford and
control study. Cancer Epidemiol Biomarkers Prev New York: Oxford University Press, 1996: 920–​45.
2004;13:1665–​76. Schöllkopf C, Melbye M, Munksgaard L, Smedby KE.
Purdue MP, Lan Q, Kemp TJ, Hildesheim A, Weinstein Rostgaard K, Glimelius B et  al. Borrelia infec-
SJ, Hofmann JN et al. Elevated serum sCD23 and tion and risk of non-​Hodgkin lymphoma. Blood
sCD30 up to two decades prior to diagnosis asso- 2008;111:5524–​9.
ciated with increased risk of non-​Hodgkin lym- Schuler F, Hirt C, Dolken G. Chromosomal transloca-
phoma. Leukemia 2015;29:1429–​31. tion t(14;18) in healthy individuals. Semin Cancer
Perry AM, Diebold J, Nathwani BN, MacLennan Biol 2003;13:203–​9.
KA, Müller-​ Hermelink HK, Bast M et  al. Schulz TF, Cesarman E. Kaposi sarcoma–​associated
Non-​ Hodgkin lymphoma in the developing herpesvirus:  mechanisms of oncogenesis. Curr
world: review of 4539 cases from the International Opin Virol 2015;14:116–​28.
Non-​Hodgkin Lymphoma Classification Project. Skibola CF, Berndt SI, Vijai J, Conde L, Wang Z,
Haematologica. 2016 Oct;101(10):1244–​50. Epub Yeager M et  al. Genome-​wide association study
2016 Jun 27. identifies five susceptibility loci for follicular
Pozzato G, Mazzaro C, Dal Maso L, Mauro E, Zorat lymphoma outside the HLA region. Am J Hum
F, Moratelli G et  al. Hepatitis C virus and non-​ Genet. 2014 Oct 2;95(4):462–​ 71. doi:10.1016/​
Hodgkin’s lymphomas: meta-​analysis of epidemi- j.ajhg.2014.09.004
ology data and therapy options. World J Hepatol. Slager SL, Benavente Y, Blair A, Vermeulen R, Cerhan
2016;8:107–​16. JR, Costantini AS et al. Medical history, lifestyle,
Rohrmann S, Linseisen J, Jakobsen MU, Overvad family history, and occupational risk factors for
K, Raaschou-​ Nielsen O, Tjonneland A et  al. chronic lymphocytic leukemia/​ small lympho-
Consumption of meat and dairy and lymphoma cytic lymphoma:  the InterLymph Non-​Hodgkin
risk in the European Prospective Investigation Lymphoma Subtypes Project. J Natl Cancer Inst
into Cancer and Nutrition. Int J Cancer Monogr 2014 Aug;2014(48):41–​51. doi:10.1093/​
2011;128:623–​34. jncimonographs/​lgu001
Roman E, Ansell P, Bull D. Leukaemia and non-​Hodgkin’s Soulier J, Grollet L, Oksenhendler E, Cacoub P,
lymphoma in children and young adults: are prena- Cazals-​Hatem D, Babinet P et  al. Kaposi’s
tal and neonatal factors important determinants of sarcoma-​associated herpesvirus-​like DNA
disease? Br J Cancer 1997;76:406–​15. sequences in multicentric Castleman’s disease.
Ron E. Ionizing radiation and cancer risk:  evidence Blood 1995;86:1276–​80.
from epidemiology. Radiat Res 1998;150:S30–​S41. Subramaniam K, Yeung D, Grimpen F, Joseph J, Fay K,
Rosa D, Saletti G, De Gregorio E, Zorat F, Comar C, Buckland M et al. Hepatosplenic T-​cell lymphoma,
D’Oro U et al. Activation of naïve B lymphocytes immunosuppressive agents and biologicals:  what
via CD81, a pathogenetic mechanism for hepatitis are the risks? Intern Med J 2014;44:287–​90.
C virus-​associated B lymphocyte disorders. Proc Swerdlow SH, Campo E, Pileri, SA, Harris NL, Stein
Natl Acad Sci USA 2005;102:18544–​9. H, Siebert Ret al. The 2016 revision of the World
Rossi D, Ciardullo C Gaidano G. Genetic aberrations of Health Organization Classification of lymphoid
signaling pathways in lymphomagenesis: revelations neoplasms. Blood 2016;127:2375–​90.
670

670 Textbook of Cancer Epidemiology

Selby WS, Gallagher ND. Malignancy in a 19-​year activity and lymphoid neoplasms in the European
experience of adult celiac disease. Dig Dis Sci Prospective Investigation into Cancer and nutri-
1979;24:684–​8. tion (EPIC). Eur J Cancer 2011;47:748–​60.
Shiels MS, Engels EA, Linet MS, Clarke CA, Hall Vijai J, Wang Z, Berndt SI, Skibola CF, Slager SL, de
HI, Hartge P, Morton LM. The epidemic of non-​ Sanjose S et al. A genome-​wide association study
Hodgkin lymphoma in the United States:  dis- of marginal zone lymphoma shows association to
entangling the effect of HIV, 1992–​ 2009. the HLA region. Nat Commun 2015 Jan 8;6:5751.
2013;6:1069–​78. doi:10.1038/​ncomms6751
Smedby KE, Hjalgrim H, Melbye M, Torrang A, Villumsen M, Sørup S, Jess T, Ravn H, Relander T,
Rostgaard K, Munksgaard L et al. Ultraviolet radi- Baker JL et al. Risk of lymphoma and leukaemia
ation exposure and risk of malignant lymphomas. after bacille Calmette-​Guérin and smallpox vac-
J Natl Cancer Inst 2005;97:199–​209. cination:  a Danish case-​ cohort study. Vaccine
Smedby KE, Vajdic CM, Falster M, Engels EA, 2009;27:6950–​8.
Martínez-​Maza O, Turner J et  al. Autoimmune Voulgarelis M, Dafni UG, Isenberg DA, Moutsopoulos
disorders and risk of non-​ Hodgkin lymphoma HM. Malignant lymphoma in primary Sjogren’s
subtypes: a pooled analysis within the InterLymph syndrome:  a multicenter, retrospective, clini-
Consortium. Blood 2008;111:4029–​38. cal study by the European Concerted Action
Talamini R, Polesel J, Montella M, Dal Maso L, Crovatto on Sjogren’s Syndrome. Arthritis Rheum
M, Crispo A et al. Food groups and risk of non-​ 1999;42:1765–​72.
Hodgkin lymphoma:  a multicenter, case-​control Walter RB, Brasky TM, Buckley SA, Potter JD, White
study in Italy. Int J Cancer 2006;118:11:2871–​6. E. Height as an explanatory factor for sex dif-
Teras LR, Gapstur SM, Diver WR, Birmann BM, ferences in human cancer. J Natl Cancer Inst
Patel AV. Recreational physical activity, leisure 2013;105:860–​8.
sitting time and risk of non-​Hodgkin lymphoid Willemze R, Kerl H, Sterry W, Berti E, Cerroni L,
neoplasms in the American Cancer Society Chimenti S et  al. EORTC classification for pri-
Cancer Prevention Study II Cohort. Int J Cancer mary cutaneous lymphomas:  a proposal from
2012;131:1912–​20. the Cutaneous Lymphoma Study Group of
Theander E, Henriksson G, Ljungberg O, Mandl T, the European Organization for Research and
Manthorpe R, Jacobsson LT. Lymphoma and other Treatment of Cancer. Blood 1997;90:354–​71.
malignancies in primary Sjogren’s syndrome:  a Willet EV, Smith AG, Dovey GJ, Morgan GJ, Parker
cohort study on cancer incidence and lymphoma J, Roman E. Tobacco and alcohol consumption
predictors. Ann Rheum Dis 2006;65:796–​803. and the risk of non-​Hodgkin lymphoma. Cancer
Tiacci E, Trifonov V, Schiavoni G, Holmes A, Kern W, Cause Control 2004;15:771–​80.
Martelli MP et  al. BRAF mutations in hairy-​cell Williams CJ, Peyrin-​Biroulet L, Ford AC. Systematic
leukemia. N Engl J Med 2011;364:2305–​15. review with meta-​ analysis:  malignancies with
Tramacere I, Pelucchi C, Bonifazi M, Bagnardi V, Rota anti-​tumour necrosis factor-​α therapy in inflam-
M, Bellocco R, et al. Alcohol drinking and non- matory bowel disease. Aliment Pharmacol Ther
Hodgkin lymphoma risk: a systematic review and 2014;39:447–​58.
a meta-analysis. Ann Oncol 2012;23:2791–8. Wolfe F. Inflammatory activity, but not methotrexate or
‘t Mannetje A, De Roos AJ, Boffetta P, Vermeulen R, prednisone use predicts non-​Hodgkin’s lymphoma
Benke G, Fritschi L et  al. Occupation and risk in rheumatoid arthritis: a 25-​year study of 1,767 RA
of non-​Hodgkin lymphoma and its subtypes:  a patients [abstract]. Arthritis Rheum 1998;41:188.
pooled analysis from the InterLymph consortium. Ye X, Casaclang N, Mahmud SM. Use of non-​steroidal
Environ Health Perspect 2016;124:396–​405. anti-​inflammatory drugs and risk of non-​
van Leeuwen MT, Vajdic CM, Middleton MG, Hodgkin lymphoma:  a systematic review and
McDonald AM, Law M, Kaldor JM, Grulich meta-​analysis. Hematol Oncol 2015;33:176–​86.
AE. Continuing declines in some but not all Ye X, Mneina A, Johnston JB, Mahmud SM.
HIV-​associated cancers in Australia after wide- Associations between statin use and non-​
spread use of antiretroviral therapy. AIDS Hodgkin lymphoma (NHL) risk and survival:  a
2009;23:2183–​90. meta-​ analysis. Hematol Oncol 2015 Oct 20.
van Leeuwen MT, Turner JJ, Joske DJ, Falster MO, doi: 10.1002/​hon.2265. [Epub ahead of print].
Srasuebkul P, Meagher NS et  al. Lymphoid neo- Zhang Y, Holford TR, Leaderer B, Boyle P, Zahm
plasm incidence by WHO subtype in Australia SH, Flynn S et al. Hair-​coloring product use and
1982–​2006. Int J Cancer 2014;135:2146–​56. risk of non-​Hodgkin’s lymphoma:  a population-​
van Veldhoven CM, Khan AE, Teucher B, Rohrmann S, based case-​control study in Connecticut. Am J
Raaschou-​Nielsen O, Tjønneland A et al. Physical Epidemiol 2004a;159:148–​54.
671

Non-Hodgkin Lymphoma 671

Zhang Y, Holford TR, Leaderer B, Boyle P, Zahm Zheng T, Mayne ST, Boyle P, Holford TR, Liu WL,
SH, Owens PH et  al. Blood transfusion and risk Flannery J. Epidemiology of non-​Hodgkin lym-
of non-​Hodgkin’s lymphoma in Connecticut phoma in Connecticut. 1935–​ 1988. Cancer
women. Am J Epidemiol 2004b;160:325–​30. 1992;70:840–​49.
Zhang Y, Holford TR, Leaderer B, Zahm SH, Boyle P, Zheng T, Holford TR, Leaderer B, Zhang Y, Zahm
Morton LM et  al. Prior medical conditions and SH, Flynn S et  al. Diet and nutrient intakes and
medication use and risk of non-​Hodgkin lym- risk of non-​Hodgkin’s lymphoma in Connecticut
phoma in Connecticut United States women. women. Am J Epidemiol 2004;159:454–​66.
Cancer Causes Control 2004c;15:419–​28.
672
673

28
Leukemias
EVE ROMAN, ALEXANDRA SMITH, AND LORELEI MUCCI

L eukemias constitute a heterogeneous group

of acute and chronic malignancies that
combined are estimated to account for around
CLINICAL SYNOPSIS
Subgroups Historically, leukemias were largely
diagnosed on the basis of the fact that cancer was
2%–​3% of cancers worldwide. Until recently, epi- detected in the blood, and then further grouped
demiological understanding about these complex according to morphology. In recent decades the
cancers has been impeded by a lack of under- taxonomy of leukemias has changed markedly, as
standing around their underlying biology. This knowledge about the links between the various
situation is changing rapidly, as clear differences hematological cancers (leukemias, lymphomas,
in the descriptive patterns of the various subtypes and myelomas) and their relationship with nor-
have led to new hypotheses about their determi- mal hematopoiesis and the immune system has
nants. Nonetheless, while it is challenging for emerged. In 2001, these advances culminated in
population-​based cancer registries to keep pace the World Health Organization’s (WHO) first
with the rapidly evolving diagnostic and clinical consensus classification of hematopoietic and
landscape of leukemia, greater efforts need to be lymphoid tissue tumors which, organized by cell
made to standardize coding, identify all myeloid of origin rather than pattern of spread, incorpo-
and lymphoid subtypes, and track transforma- rated genetic data alongside information on mor-
tions and progressions. phology, immunology, and clinical parameters
Leukemias have both genetic and environ- (Jaffe et al, 2001).
mental components to their etiology, the relative Critically for epidemiology, the WHO’s 2001
contribution of each varying from one subtype to classification and its successors (Swerdlow et al,
another. With respect to exposures, there has been 2008; Vardiman et al, 2009; Campo et al, 2011)
relatively little progress in recent decades; with incorporated refinements to previously defined
known risk factors (e.g., chemotherapy, ionizing categories and recognized several new leukemic
radiation, and certain viral infections) explaining entities including the myelodysplastic syndromes
only a small proportion of the total disease burden, (MDSs), myeloproliferative neoplasms (MPNs),
and many other suggested factors (e.g., nonioniz- and the “overlap” MDS/​MPN syndromes—​all of
ing radiation and antibiotic usage) largely leading which are still assigned a behavior code of one in
to negative results. By contrast, major strides in the latest update of the site-​based ICD-​10 (WHO,
identifying the underpinning genetic and molecu- 2015). Incorporating all myeloid malignancies, as
lar pathogenic processes have been made; and this well as some precursor and mature B-​cell and T-​
increase in biological knowledge will hopefully cell lymphoid neoplasms, the leukemias span the
lead to increased understanding about the deter- WHO classification (Box 28-​1).
minants of the various leukemia subtypes. Symptoms In the clinical setting, acute leuke-
With respect to treatment, hemato-​oncology mias tend to present with features of bone mar-
is one of the most rapidly moving areas of can- row failure (anemia, thrombocytopenia, and
cer research, and the survival improvements seen neutropenia) that results in fatigue, bone pain,
for subtypes such as chronic myeloid leukemia bruising/​bleeding, and fever; and patients with
(CML), acute promyelocytic leukemia (APL), lymphoid neoplasms may also have lymph node
and pediatric precursor B-​lymphoblastic leuke- enlargements. By contrast, patients with chronic
mia (B-​ALL) represent some of the major success remitting/​ relapsing leukemia are often diag-
stories of cancer treatment this century. nosed through routine blood tests when they are
674

BOX 28-​1
WORLD HEALTH ORGANIZATION CLASSIFICATION OF HEMATOPOIETIC
AND LYMPHOID TISSUES

M Y E L O P R O L I F E R AT I V E N E O P L A S M S   ( M P N S )
Chronic myelogenous leukemia, BCR-​ABL1–​positive
Chronic neutrophilic leukemia
Polycythemia vera
Primary myelofibrosis
Essential thrombocythemia
Chronic eosinophilic leukemia, not otherwise specified
Mastocytosis
Myeloproliferative neoplasms, unclassifiable

M Y E L O I D A N D LY M P H O I D N E O P L A S M S A S S O C I AT E D W I T H   E O S I N O P H I L I A
AND ABNORMALITIES OF PDGFRA, PDGFRB, OR FGFR1
Myeloid and lymphoid neoplasms associated with PDGFRA rearrangement
Myeloid neoplasms associated with PDGFRB rearrangement
Myeloid and lymphoid neoplasms associated with FGFR1 abnormalities

M Y E L O D Y S P L A S T I C /​M Y E L O P R O L I F E R AT I V E N E O P L A S M S ( M D S /​M P N )
Chronic myelomonocytic leukemia
Atypical chronic myeloid leukemia, BCR-​ABL1–​negative
Juvenile myelomonocytic leukemia
Myelodysplastic/​myeloproliferative neoplasm, unclassifiable
Provisional entity: refractory anemia with ring sideroblasts and thrombocytosis

MYELODYSPLASTIC SYNDROMES (MDSS)
Refractory cytopenia with unilineage dysplasia
Refractory anemia
Refractory neutropenia
Refractory thrombocytopenia
Refractory anemia with ring sideroblasts
Refractory cytopenia with multilineage dysplasia
Refractory anemia with excess blasts
Myelodysplastic syndrome with isolated del(5q)
Myelodysplastic syndrome, unclassifiable
Childhood myelodysplastic syndrome
Provisional entity: refractory cytopenia of childhood

A C U T E M Y E L O I D L E U K E M I A A N D R E L AT E D N E O P L A S M S
Acute myeloid leukemia with recurrent genetic abnormalities
AML with t(8;21)(q22;q22); RUNX1-​RUNX1T1
AML with inv(16)(p13.1q22) or t(16;16)(p13.1;q22); CBFB-​MYH11
APL with t(15;17)(q22;q12); PML-​RARA
AML with t(9;11)(p22;q23); MLLT3-​MLL
AML with t(6;9)(p23;q34); DEK-​NUP214
AML with inv(3)(q21q26.2) or t(3;3)(q21;q26.2); RPN1-​EVI1
AML (megakaryoblastic) with t(1;22)(p13;q13); RBM15-​MKL1
675

Provisional entity: AML with mutated NPM1

Provisional entity: AML with mutated CEBPA
Acute myeloid leukemia with myelodysplasia-​related changes
Therapy-​related myeloid neoplasms
Acute myeloid leukemia, not otherwise specified
AML with minimal differentiation
AML without maturation
AML with maturation
Acute myelomonocytic leukemia
Acute monoblastic/​monocytic leukemia
Acute erythroid leukemia
Pure erythroid leukemia
Erythroleukemia, erythroid/​myeloid
Acute megakaryoblastic leukemia
Acute basophilic leukemia
Acute panmyelosis with myelofibrosis
Myeloid sarcoma
Myeloid proliferations related to Down’s syndrome
Transient abnormal myelopoiesis
Myeloid leukemia associated with Down’s syndrome
Blastic plasmacytoid dendritic cell neoplasm

ACUTE LEUKEMIAS OF AMBIGUOUS LINEAGE

Acute undifferentiated leukemia
Mixed phenotype acute leukemia with t(9;22)(q34;q11.2); BCR-​ABL1
Mixed phenotype acute leukemia with t(v;11q23); MLL rearranged
Mixed phenotype acute leukemia, B-​myeloid, NOS
Mixed phenotype acute leukemia, T-​myeloid, NOS
Provisional entity: natural killer (NK) cell lymphoblastic leukemia/​lymphoma

P R E C U R S O R LY M P H O I D N E O P L A S M S
B lymphoblastic leukemia/​lymphoma
B lymphoblastic leukemia/​lymphoma, NOS
B lymphoblastic leukemia/​lymphoma with recurrent genetic abnormalities
B lymphoblastic leukemia/​lymphoma with t(9;22)(q34;q11.2); BCR-​ABL 1
B lymphoblastic leukemia/​lymphoma with t(v;11q23); MLL rearranged
B lymphoblastic leukemia/​lymphoma with t(12;21)(p13;q22); TEL-​AML1 (ETV6-​RUNX1)
B lymphoblastic leukemia/​lymphoma with hyperdiploidy
B lymphoblastic leukemia/​lymphoma with hypodiploidy
B lymphoblastic leukemia/​lymphoma with t(5;14)(q31;q32); IL3-​IGH
B lymphoblastic leukemia/​lymphoma with t(1;19)(q23;p13.3); TCF3-​PBX1

T LY M P H O B L A S T I C L E U K E M I A / L​ Y M P H O M A
Mature B-​cell neoplasms
Chronic lymphocytic leukemia/​small lymphocytic lymphoma
Hairy cell leukemia

Mature T-​cell and NK-​cell neoplasms

T-​cell prolymphocytic leukemia
T-​cell large granular lymphocytic leukemia
67

676 Textbook of Cancer Epidemiology

asymptomatic (Swerdlow et  al, 2008; Abel et el, acid and arsenic trioxide. Many new targeted
2012; Howell et al, 2013). AML agents, including monoclonal antibodies,
Diagnosis Using bone marrow aspiration and are currently under investigation; however, as the
biopsy samples, leukemias are diagnosed using majority are used in conjunction with intensive
a combination of information on morphology, chemotherapy, little progress has been made in
cytochemistry, immunophenotype, cytogenet- improving the cure rates of patients who are una-
ics, molecular analysis, and clinical features. ble to tolerate intensive treatments (Dombret &
Depending on the diagnostic pathway, imaging Gardin 2016).
tests (x-​rays, as well as CT, MRI, and PET scans) Prognosis and progress The survival improvements
and lumbar punctures to check for leukemic seen for some leukemia subtypes represent major
cells in the cerebrospinal fluid may also be car- success stories of cancer treatment. In high-​income
ried out, the latter being commonly included in countries, the significant improvements have
clinical guidelines for pediatric disease. On occa- occurred for pediatric precursor B-​ALL, where
sion, patients with lymphoid leukemias, such as survival is now approaching 90% (Gatta et al, 2014;
chronic lymphocytic leukemia (CLL), may also Howlader et al, 2014; Pui et al, 2015); APL, where
undergo lymph node biopsy (Swerdlow et  al, cure rates now exceed 80% (Lo-​Coco et al, 2016);
2008; Vardiman et al, 2009; Campo et al, 2011). and CML, which TKIs have transformed from a
Treatment Leukemias constitute a diverse group comparatively rare fatal cancer to a long-​term con-
of disorders, and this is reflected in the wide dition with a survival that approaches that of the
array of strategies used for their treatment. general population (Höglund et  al, 2013; Smith
Unlike chronic leukemias, which tend to follow a et al, 2014; Apperley, 2015; Thompson et al, 2015).
remitting/​relapsing pathway, acute myeloid leu- Despite these improvements, the outlook for sev-
kemia (AML) and acute lymphoblastic leukemia eral leukemia subtypes, particularly those occur-
(ALL) are potentially curable with intensive che- ring at older ages, remains poor; most notably
motherapy, which can be administered with or for AML, where little progress had been made in
without allogeneic hematopoietic stem cell trans- improving outcomes for the majority of patients
plantation (HSCT). Patients with acute leuke- diagnosed over the last three decades (Burnett
mia require immediate treatment with intensive 2012). Indeed, while the frequency of curative
chemotherapy, whereas patients with chronic therapy is relatively high in younger patients,
leukemia can often be monitored until treatment the inability of some patients, particularly older
is required (“watchful waiting”). In addition, in patients, to tolerate the intensive chemotherapy
order to alleviate symptoms and improve qual- regimens remains problematic.
ity of life, patients with MDSs and MPNs may be The current heterogeneity of leukemia sub-
managed supportively with, for example, transfu- types is demonstrated in Table 28-​ 1, which
sions and erythroid-​stimulating agents (Döhner includes overall and relative 5-​year estimates for
et al, 2010; Malcovati et al, 2013; Eichhorst et al, the leukemias and related precursor neoplasms
2015; Vannucchi et al, 2015). listed in the latest WHO revision (Swerdlow
The treatment landscape for many of these et  al, 2008; Vardiman et  al, 2009; Campo et  al,
disorders is changing rapidly with the devel- 2011). Data are from the United Kingdom’s
opment of targeted therapies (Bachireddy et al, Haematological Malignancy Research Network
2015; Greaves, 2016). The seminal shift began (HMRN; Smith et al, 2011a, 2015; Roman et al,
with the discovery of the Philadelphia chromo- 2016), a population-​based register of hemato-
some/​translocation (BCR-​ABL1), which led to logical malignancies that is served by a single
the development of targeted therapy in the form diagnostic laboratory and covers a population of
of daily oral tyrosine kinase inhibitors (TKIs) for around 4 million people (www.hmrn.org).
the treatment of CML, and Philadelphia chro-
mosome-​ positive ALL. Chemoimmunotherapy DESCRIPTIVE EPIDEMIOLOGY
is now the standard of care for CLL patients Examining global variations in any cancer is
requiring treatment, with newer developments challenging, given the marked variability in the
including B-​cell receptor kinase inhibitors (Byrd quality and coverage of civil and cancer reg-
et al, 2014; Bachireddy, 2015). Likewise, develop- istration. Indeed, around 65% of the world’s
ments have occurred for some AML subtypes, population reside in countries with incomplete
most notably APL, which is treated with retinoic death registration and 80% in countries with
 67
TABLE 28-​1   NUMBER OF DIAGNOSES, MEDIAN AGES (INTER QUARTILE RANGE-​I QR), INCIDENCE RATES PER 100,000, 5-​Y EAR OVERALL
SURVIVAL (OS) AND 5-​Y EAR RELATIVE SURVIVAL (RS): HAEMATOLOGICAL MALIGNANCY RESEARCH NET WORK DIAGNOSES SEPT 2004
TO AUG 2013, FOLLOWED THROUGH TO FEBRUARY 2016

N Median age at Rate per 100,000 (95% Confidence Intervals) Sex rate ratio 5-​year survival
diagnosis (IQR) world 1996 (95% Confidence Intervals)
years
All Male Female Overall Relative

Acute myeloid leukemia 1411 70.6 (57.3–​79.1) 2.58 (2.52–​2.63) 3.11 (3.02–​3.20) 3.02 (3.20–​2.16) 1.33 (1.23–​1.43) 14.4 (12.6–​16.2) 16.3 (14.3–​18.4)
(AML)
AML, not otherwise specified 860 73.7 (62.4–​81.5) 1.41 (1.37–​1.46) 1.80 (1.73–​1.88) 1.10 (1.04–​1.16) 1.64 (1.53–​1.75) 9.2 (7.4–​11.1) 10.6 (8.6–​12.9)
AML with myelodysplasia-​ 198 69.9 (63.4–​75.5) 0.35 (0.32–​0.38) 0.47 (0.42–​0.51) 0.26 (0.21–​0.3) 1.81 (1.53–​2.15) 3.50 (1.7–​6.3) 3.8 (1.8–​7.0)
related changes
AML with NPM1 mutation 104 72.0 (57.3–​79.0) 0.18 (0.16–​0.20) 0.16 (0.12–​0.20) 0.20 (0.16–​0.24) 0.82 (0.6–​1.12) 26.5 (17.8–​36.0) 29.7 (19.9–​40.1)
Acute promyelocytic leukemia 91 47.2 (33.1–​63.1) 0.24 (0.21–​0.27) 0.25 (0.20–​0.30) 0.23 (0.18–​0.27) 1.11 (0.83–​1.47) 63.7 (52.0–​73.3) 66.9 (54.4–​76.7)
(APL)
AML, core binding factor 64 42.9 (27.6–​56.9) 0.18 (0.16–​0.21) 0.24 (0.18–​0.29) 0.13 (0.09–​0.18) 1.80 (1.22–​2.66) 55.5 (42.3–​66.8) 57.3 (43.6–​68.8)
AML, probable therapy 60 72.1 (59.6–​77.9) 0.10 (0.08–​0.12) 0.10 (0.07–​0.13) 0.11 (0.08–​0.14) 0.90 (0.61–​1.33) 1.8 (0.30–​6.70) 2.0 (0.3–​7.1)
related
AML with MLL (11q23) 25 20.3 (13.9–​43.8) 0.10 (0.07–​0.12) 0.08 (0.03–​0.12) 0.12 (0.07–​0.16) 0.63 (0.33–​1.22) 36.2 (17.2–​55.6) 36.5 (17.3–​56.0)
Myelodysplastic syndromes 1194 75.7 (68.5–​81.7) 1.67 (1.63–​1.71) 2.55 (2.48–​2.62) 1.04 (0.98–​1.10) 2.46 (2.33–​2.59) 21.8 (19.4–​24.4) 28.9 (25.7–​32.2)
(MDSs)
Refractory cytopenia with 497 75.7 (69.3–​81.5) 0.68 (0.65–​0.70) 1.15 (1.11–​1.2) 0.33 (0.30–​0.37) 3.45 (3.17–​3.77) 24.0 (20.1–​28.2) 32.2 (26.9–​37.7)
multilineage dysplasia (RCMD)
Refractory anemia with excess 458 74.5 (66.7–​81.3) 0.68 (0.65–​0.71) 0.97 (0.92–​1.01) 0.48 (0.42–​0.53) 2.03 (1.83–​2.25) 8.3 (5.9–​11.3) 10.4 (7.4–​14.1)
blasts (RAEB)
Refractory anemia with ring 213 77.6 (71.4–​83.5) 0.27 (0.25–​0.28) 0.42 (0.39–​0.45) 0.16 (0.14–​0.18) 2.64 (2.29–​3.04) 41.6 (34.6–​48.4) 57.6 (47.7–​66.4)
sideroblasts (RARS)
Myelodysplastic syndrome (5q-​) 26 72.0 (61.7–​78.0) 0.04 (0.03–​0.05) 0.01 (0–​0.02) 0.07 (0.05–​0.09) 0.17 (0.08–​0.39) 55.9 (34.3–​72.9) 70.9 (38.5–​88.3)
Myeloproliferative 2329 70.3 (58.7–​79.2) 4.02 (3.96–​4.08) 4.37 (4.28–​4.46) 3.75 (3.66–​3.83) 1.17 (1.13–​1.2) 70.3 (68.3–​72.2) 90 (87.7–​91.9)
neoplasms (MPNs)
Chronic MPNsa 1818 71.4 (60.7–​79.9) 3.02 (2.97–​3.07) 3.1 (3.02–​3.19) 2.97 (2.9–​3.05) 1.04 (1.01–​1.08) 72.1 (69.8–​74.2) 93.6 (90.9–​95.6)
(continued)
 678
TABLE 28-​1  CONTINUED

N Median age at Rate per 100,000 (95% Confidence Intervals) Sex rate ratio 5-​year survival
diagnosis (IQR) world 1996 (95% Confidence Intervals)
years
All Male Female Overall Relative

Chronic myeloid leukemia 318 59.1 (46.8–​71.1) 0.69 (0.66–​0.72) 0.87 (0.82–​0.93) 0.52 (0.48–​0.57) 1.67 (1.5–​1.85) 78.3 (73.1–​82.7) 90.8 (85.2–​94.3)
(CML)
Myelofibrosis 165 73.7 (65.7–​79.8) 0.25 (0.23–​0.27) 0.35 (0.32–​0.38) 0.16 (0.14–​0.19) 2.14 (1.8–​2.54) 33.1 (25.4–​41.1) 43.4 (33.1–​53.3)
Systemic mastocytosis 26 59.3 (37.6–​69.2) 0.06 (0.04–​0.08) 0.03 (0.01–​0.05) 0.09 (0.06–​0.12) 0.36 (0.18–​0.72) 80.0 (58.5–​91.2) 88.7 (54.1–​97.7)
MDS/​MPN 296 77.2 (69.4–​82.6) 0.4 (0.38–​0.43) 0.59 (0.54–​0.64) 0.27 (0.23–​0.31) 2.20 (1.9–​2.54) 13.7 (9.8–​18.3) 18.2 (13.0–​24.2)
Chronic myelomonocytic 239 77.4 (71.5–​82.9) 0.3 (0.28–​0.32) 0.47 (0.43–​0.5) 0.18 (0.16–​0.21) 2.59 (2.23-​3.00) 13.8 (9.6–​18.8) 18.5 (12.8–​25.0)
leukemia (CMML)
MDS/​MPN, unclassifiable 30 77.5 (67.6–​82.9) 0.04 (0.03–​0.05) 0.05 (0.03–​0.07) 0.03 (0.02–​0.04) 1.72 (0.95–​3.14) 0 0
Atypical chronic myeloid 23 71.4 (66.7–​81.8) 0.04 (0.03–​0.05) 0.06 (0.04–​0.08) 0.02 (0–​0.04) 3.06 (1.47–​6.38) 2.7 (0–​19.9) 3.2 (0–​22.9)
leukemia
Precursor lymphoid 388 15.2 (4.6–​44.6) 1.60 (1.54–​1.66) 1.83 (1.73–​1.92) 1.38 (1.29–​1.46) 1.57 (1.55–​1.59) 61.8 (56.7–​66.6) 64.3 (59.0–​69.1)
neoplasms (ALL)
B-​lymphoblastic leukemia 300 12.8 (3.7–​46.8) 1.28 (1.22–​1.33) 1.36 (1.28–​1.45) 1.19 (1.11–​1.27) 1.15 (1.05–​1.26) 61.9 (56.1–​67.2) 64.4 (58.3–​69.9)
B-​lymphoblastic leukemia, not 139 19.5 (7.5–​53.1) 0.53 (0.49–​0.56) 0.62 (0.56–​0.68) 0.44 (0.38–​0.5) 1.41 (1.20–​1.66) 54.8 (46.0–​62.7) 57.4 (48.1–​65.5)
otherwise specified
with hyperdiploidy 69 4.3 (2.8–​10.0) 0.36 (0.33–​0.39) 0.38 (0.33–​0.44) 0.34 (0.29–​0.39) 1.14 (0.92–​1.41) 85.5 (74.8–​91.9) 87.1 (75.7–​93.4)
with t(9;22) 38 49.3 (35.5–​59.6) 0.11 (0.08–​0.13) 0.11 (0.07–​0.15) 0.11 (0.06–​0.15) 0.99 (0.56–​1.73) 36.2 (21.4–​51.1) 38.4 (22.6–​54.0)
with t(12;21) 35 4.1 (3.0–​5.0) 0.21 (0.18–​0.23) 0.19 (0.15–​0.23) 0.22 (0.18–​0.27) 0.86 (0.65–​1.13) 91.3 (75.4–​97.1) 91.3 (75.4–​97.1)
with MLL rearrangement 13 32.2 (1.0–​56.2) 0.05 (0.03–​0.07) 0.04 (0.01–​0.07) 0.07 (0.03–​0.11) 0.55 (0.20–​1.50) 38.5 (14.1–​62.7) 39.5 (14.4–​64.1)
T-​lymphoblastic leukemia 88 18.5 (11.3–​38.7) 0.33 (0.29–​0.36) 0.46 (0.40–​0.52) 0.19 (0.14–​0.24) 2.45 (1.89–​3.16) 61.5 (50.1–​71.0) 63.3 (51.6–​73.0)
Chronic lymphocytic 2248 71.7 (62.7–​79.0) 3.63 (3.57–​3.68) 5.22 (5.12–​5.31) 2.29 (2.23–​2.36) 2.27 (2.20–​2.35) 65.8 (63.6–​67.8) 83.9 (81.3–​86.1)
leukemia
Hairy cell leukemia 113 66.8 (56.4–​74.5) 0.22 (0.19–​0.24) 0.37 (0.33–​0.41) 0.07 (0.05–​0.09) 5.10 (3.93–​6.62) 80.4 (71.1–​86.9) 96.1 (73.3–​99.5)
T-​cell leukemias 139 73.6 (62.0–​78.9) 0.22 (0.20–​0.25) 0.23 (0.18–​0.27) 0.22 (0.19–​0.25) 1.03 (0.82–​1.29) 64.5 (55.5–​72.1) 81.9 (69.9–​89.5)
T-​cell large granular 110 71.3 (61.7–​77.2) 0.18 (0.16–​0.21) 0.19 (0.15–​0.23) 0.18 (0.16–​0.21) 1.02 (0.78–​1.33) 78.7 (69.4–​85.5) 95 (75.1–​99.1)
lymphocytosis
T-​cell prolymphocytic leukemia 29 77.8 (73.8–​83.4) 0.04 (0.03–​0.05) 0.04 (0.02–​0.06) 0.04 (0.02–​0.05) 1.08 (0.55–​2.12) 12.3 (3.5–​27.2) 17.1 (4.5–​36.4)
679

Leukemias 679

60

New cases per 100,000

50

person-years
40

30

20

10
0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 28-​1  Age-​specific incidence rates of leukemia among women.

Source: Ferlay et al, 2013.

no cancer registration (Parkin, 2006; Mathers & been applied in order to deal with this phenome-
Boerma, 2010; Ferlay et  al, 2015). For hemato- non (Derolf et al, 2009; HAEMACARE Working
logical malignancies, this is further exacerbated Group, 2010; Gavin et al, 2015; Ruhl et al, 2015).
by the inconsistent implementation of WHO’s In brief, the US Surveillance Epidemiology and
2001 classification, which although internation- End Results (US SEER) registries, as well as sev-
ally adopted into clinical practice was not uni- eral specialist registries, allow primary AML
formly embraced by population-​ based cancer registrations in patients with a prior MDS unless
registries. This is evidenced in Figures 28-​1 and the two diagnoses are ≤ 21 days apart (Ruhl et al,
28-​2, which, for the eight countries included 2015); whereas ENCR (European Network of
in this textbook, presents the 2012 age-​specific Cancer Registries) guidelines state that only the
incidence rates for all leukemias combined for first tumor (in this case MDS) should be counted
males and females respectively; data on finer in incidence statistics, unless progression occurs
groupings are unavailable (Ferlay et  al, 2013). within 90 days, in which case the original MDS
This adherence to past coding practice partly code should be replaced by the appropriate
reflects the fact that the range and depth of data AML code (HAEMACARE Working Group,
used to diagnose leukemia and other hematolog- 2010; Gavin et  al, 2015). Such protocol vari-
ical cancers can be challenging to access system- ations are likely to have contributed to the fact
atically; potentially forming a barrier not only that leukemia rates in the United States for all
to the collection of diagnostic data at the level subtypes combined tend to be marginally higher
of detail required to implement the latest WHO than those in Europe; but whether or not leuke-
classification but also to complete ascertain- mia is four times more common in the United
ment, with many registries struggling to capture States than in China awaits more in-​depth study
information on cancers such as MDS and MPN, (Figs.28.1, 28.2).
which are still assigned a behavior code of one While using national data to draw conclusions
and grouped with the D-​codes in the latest ICD-​ about global variation in overall disease burden is
10 update (WHO, 2015). problematic, there are marked geographic patterns
The capture and coding problems that sur- for some leukemia subtypes, most notably for
round leukemia are further complicated by the adult T-​cell leukemia/​lymphoma (ATLL), which is
fact that many are characterized by their ability endemic in parts of Central Africa, the Caribbean,
to progress and transform; at least a third of com- and Japan, but is rare elsewhere. This geographic
mon MDS subtypes, for example, will eventually patterning, which is predicated on the HTLV-​1, is
progress to AML (Swerdlow et al, 2008; Vardiman further discussed in the risk factors section.
et al, 2009). Unfortunately, there is a lack of con- Overall, leukemias tend to be more common
sistency in the registration policies that have in males than females and, as with most cancers,
680

680 Textbook of Cancer Epidemiology

100

New cases per 100,000 80

person-years
60

40

20

0
0–14 15–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75+
Age

China Japan Poland Spain Sweden

The Netherlands United Kingdom United States of America

FIGURE 28-​2  Age-​specific incidence rates of leukemia among men.

Source: Ferlay et al, 2013.

incidence increases with increasing age (Figs. include “myeloid malignancies with germline
28.1, 28.2). However, the age and sex distribu- predisposition” as a new provisional diagnostic
tions vary markedly from one subtype to another, category (Czuchlewski & Peterson, 2016).
as is illustrated in Table 28-​1. Precursor T-​and A number of predisposition syndromes are
B-​cell lymphoblastic leukemias are primarily strongly associated with increased leukemia
diseases of children and young adults, with spo- risk, particularly in children and young adults
radic cases occurring at older ages. On the other (Seif, 2011; Schmiegelow, 2016). One of the most
hand, malignancies arising from myeloid cells widely known is with trisomy 21 (Down syn-
and more mature immunocompetent lymphoid drome); affected individuals are about 20 times
cells predominate in adults, with sporadic cases more likely to develop acute leukemia (Krivit
of some occasionally occurring at younger ages. & Good, 1957; Zipursky & Doyle, 1993; Hasle
With respect to gender, there is a clear male bias et  al, 2000; Hill et  al, 2003; Zeller et  al, 2005).
among many subtypes; the strongest difference is With respect to AML, children with trisomy 21
for the mature B-​cell neoplasm hairy cell leuke- present with a unique megakaryoblastic subtype
mia, which has an age-​standardized sex-​rate ratio characterized by a somatic mutation in GATA1,
(male/​female) of 5.1 (95% Confidence Interval which currently has its own WHO ICD-​O3 code,
[CI] 3.9–​6.6). In the context of such pronounced and generally responds well to therapy (Hitzler,
effects, which are evident across all ages and most 2007; Swerdlow et al, 2008; Vardiman et al, 2009;
subtypes, conditions that exhibit no strong sex Schmiegelow, 2016). By contrast, ALL in children
differences (e.g., AML with NPM1 mutation and with trisomy 21 shares broadly similar features to
APL), as well as those that are more common in that in those without trisomy 21; but the outcome
females (e.g., systematic mastocytosis and MDS is often poorer, largely due to the toxicity-​related
5q-​) stand out. At present, the underpinning issues that the syndrome confers (Whitlock et al,
reasons for these sex differences remain, for the 2005; Zeller et al, 2005; Schmiegelow, 2016).
most part, unknown. A number of other comparatively rare germ-
line predisposition syndromes also result in an
GENETIC PREDISPOSITION increased risk of leukemia, including tumor sup-
As evidenced by the 2001 WHO classification pressor syndromes (e.g., Li-​ Fraumeni), DNA
and its successors, genetic and molecular data repair syndromes (e.g., Fanconi anemia), bone
play a central, and ever-​expanding, role in leu- marrow failure syndromes (e.g., Diamond-​
kemia diagnostics, prognostics, and treatment Blackfan), and immunodeficiency syndromes
decision-​ making (Jaffe et  al, 2001; Swerdlow (e.g., Wiskott-​Aldrich). Currently, more than 20
et  al, 2008; Vardiman et  al, 2009; Campo et  al, conditions have been identified, and this list is
2011). Furthermore, the forthcoming update of growing as the use of targeted germline sequenc-
the WHO classification will, for the first time, ing increases (Seif, 2011; Enciso-​ Mora et  al,
681

Leukemias 681

2012; Geyer & Jacobson, 2012; Sava et  al, 2014; potential etiological role of exposure to infections
Cerhan & Slager, 2015; Babushok et  al, 2016; in early life. Indeed, the suggestion that the pattern
Czuchlewski & Peterson, 2016; Schmiegelow, and timing of an infectious disease in infancy may
2016). Importantly, although accounting for be a determinant of the type of leukemia com-
probably no more than 2% of the leukemias mon in children was first proposed 100 years ago
diagnosed, such characterization has clear clin- (Ward, 1917; Cooke, 1954). Since then the poten-
ical relevance, as patients with predisposition tial etiological role of infection in the develop-
syndromes tend to be prone to therapy-​induced ment of childhood leukemia has been the focus of
toxicities. many epidemiological studies. Accordingly, with
a view to quantifying children’s likely exposure at
RISK FACTORS various time-​points, a wide range of proxies have
Like all malignancies, leukemias have both been employed including family measures of soci-
genetic and environmental components to their oeconomic status and residential location (Poole
etiology, the relative contribution of each varying et al, 2006; Smith et al, 2006); parental indicators
from one subtype to another. In addition to the of social contact outside the home (Kinlen et  al,
biological, physical, and chemical associations 2002; Fear et al, 2005; Chang et al, 2007; Keegan
briefly touched on in the following sections, it is et  al, 2012); markers of the child’s social activity
likely that more specific etiological insights will such as birth order (Dockerty et  al, 2001; Schüz
arise as new data about the pathological and clin- et  al, 2015); and preschool group attendance
ical diversity of the various leukemia subtypes (Neglia et  al, 2000; Ma et  al, 2002; Gilham et  al,
continues to emerge. 2005; Urayama et al, 2011; Rudant et al, 2015) as
well as infectious illness histories of both the child-
Infections ren and their mothers (Jourdan-​Da Silva et  al,
The RNA retrovirus HTLV-​1 is a necessary, but 2004; Rosenbaum et al, 2005; Roman et al, 2007;
not sufficient, cause of the comparatively rare MacArthur et al, 2008). To date, findings have been
adult T-​cell leukemia/​lymphoma (ATLL), which equivocal, and no specific agents have been iden-
develops in approximately 5% of those infected tified, although the observation that IL10 levels
with the virus (Swerdlow et  al, 2008; Gessain are reduced at birth in children who subsequently
& Cassar, 2012; Ishitsuka & Tamura, 2014; develop ALL suggests that immune dysregulation
Tsukasaki & Tobinai, 2014). HTLV-​1 is thought to may have a role to play (Chang et al, 2011).
have originated from nonhuman primates, and is
endemic in parts of Japan, South America, Papua Ionizing Radiation
New Guinea, Africa, and the Middle East—​but First reported among radiobiologists and clini-
hardly ever occurs elsewhere, with an estimated cians, studies of survivors of the atomic bombs
5 to 20  million individuals infected worldwide. dropped on Nagasaki and Hiroshima in August
ATLL has a male predominance, timing of first 1945 confirmed the association between expo-
exposure is thought to be key to its development; sure to high doses of ionizing radiation and acute
the strongest risk factor is high HTLV-​1 proviral leukemia (Henshaw et al, 1944; Folley et al, 1952;
load in the peripheral blood. Hsu et al, 2013). For AML, the excess risk associ-
In common with many lymphoma subtypes, ated with high and intermediate doses of ionizing
and mediated by immunosuppression, associa- radiation begins around 2 years after exposure and
tions between the HIV retrovirus and some lym- peaks a few years later. Longer-​term follow-​up of
phoid leukemias are well established. A  large Japanese A-​bomb survivors, as well as epidemio-
US follow-​up study of HIV-​infected individuals logical studies of therapeutically and occupation-
reported a doubling of risk for ALL (B-​cell and T-​ ally exposed groups, have also found associations
cell combined), but no increase for CLL (Gibson with CML and ALL, but not the mature B-​cell neo-
et  al, 2014). With respect to ALL, while specific plasm CLL (Koshurnikova et al, 1994; Muirhead
associations with viruses other than HIV have et al, 2009; Hsu et al, 2013; Gluzman et al, 2015;
not been consistently found, there is considerable McKeown et al, 2015; Radivoyevitch et al, 2016).
interest in the relationship between infection and Exposures in utero have long been thought to
pediatric precursor B-​ALL:  the incidence peak be important determinants of certain childhood
around 3  years of age in high-​income countries cancers, and supporting evidence that this may
having led to several hypotheses concerning the be so was provided around 70 years ago, when a
682

682 Textbook of Cancer Epidemiology

case-​control study of childhood leukemia reported Kheifets et  al, 2008; Elliott et  al, 2013; Koeman
that diagnostic radiography of pregnant women et  al, 2014). Most concern has surrounded the
was associated with an increased risk of leukemia potential association with childhood B-​ ALL,
in their offspring (Giles et al, 1956; Stewart, Webb, where results from epidemiological studies have
& Hewitt, 1958; Wakeford, 2008). Initially greeted been more consistent; pooled analyses gener-
with some skepticism, this association was subse- ally estimating relative risks of 1.5 to 2.0 in the
quently confirmed in several other studies (Bithell highest exposure categories (Ahlbom et al, 2000;
& Stiller, 1988; Boice & Miller, 1999). Modern x-​ Pedersen et al, 2015; Schüz et al, 2016). The expla-
ray procedures involve much lower doses than in nation for these elevated risks remains unknown,
the 1950s and 1960s, and radiological examination but it is generally agreed that selection bias may
of pregnant women and young children is gener- account for some of the excess observed.
ally subject to careful regulation and monitoring
(Schulze-​Rath et al, 2008; Buls et al, 2009; Austin Medicines and Drugs
& Frush, 2011). More recently, concern has been The risk of AML after receiving chemotherapy has
expressed about the potential overuse of computed been recognized since the 1960s, initially in sur-
tomography (CT) in children and young adults, vivors of Hodgkin lymphoma and subsequently
and several large cohorts have now been established for other cancers including breast, ovarian, mye-
to investigate this (Pearce et al, 2012; Mathews et al, loma, and non-​malignant disorders (Grünwald &
2013; Huang et al, 2014; Boice, 2015). Rosner, 1982; Curtis et al, 1984; Kaldor et al, 1987,
In the 1980s, reports from the United 1990). It is estimated that 10% to 20% of AML
Kingdom suggesting that leukemia rates were occurs as a consequence of exposure to cytotoxic
increased in children and young adults living chemotherapy and/​or radiation therapy (Leone
in the vicinity of nuclear establishments led not et al, 2007, 2010; Granfeldt Østgård et al, 2015). In
only to concern about general environmental a US cohort of 426,068 patients treated with che-
exposure but also to concern about the possible motherapy between 1975 and 2008, AML risk was
deleterious effects of periconceptional exposure nearly five times higher than that in the general
of fathers to ionizing radiation in the work- population (Morton et al, 2013).
place (Gardner et al, 1990; Urquhart et al, 1991; In 2001, WHO recognized therapy-​ related
Roman et al, 1999). Since then, many ecological, myeloid neoplasms as a distinct entity, includ-
case-​control, and cohort studies have been con- ing the subtypes t-​AML, t-​MDS, and t-​MDS/​
ducted, and the topic remains an active area of MPN. It is generally recognized that there are
research (Roman et  al, 1999; Boice et  al, 2003; two well-​defined groups, depending on whether
White-​Koning et  al, 2004; Laurier et  al, 2014; the patient received alkylating agents (e.g., mel-
Wakeford, 2014a). Furthermore, in addition to phalan, cyclosphosphamide, and chlorambucil)
exposure to ionizing radiation, several compet- or topoisomerase II inhibitors (e.g., etoposide,
ing hypotheses, most notably Kinlen’s hypothesis doxorubicin, and mitoxantrone). Patients treated
relating to population mixing, have been sug- with alkylating agents tend to present with MDS
gested as possible explanations for the excesses within 5–​10  years, whereas those treated with
frequently reported in the vicinity of nuclear topoisomerase inhibitors are more likely to pres-
installations (Kinlen 1997, 2015; van Laar et  al, ent with t-​AML within 2 to 3  years. Different
2014; Wakeford, 2014b). cytogenetic abnormalities are also observed, the
former generally having chromosome(s) 7 and/​
Extremely Low Frequency or 5 monosomies or deletions, and the latter
Electro-​magnetic Fields presenting with balanced translocations involv-
Generated by the production and use of electric- ing the MLL gene on chromosome 11q23 (Smith
ity, extremely low frequency electro-​ magnetic et  al, 1996). However, newer agents used in the
fields (ELF-​EMFs) have been the subject of many treatment of malignant and nonmalignant condi-
epidemiological studies; primarily in relation tions, including “antimetabolites” (e.g., azathio-
to occupational exposure of adults and residen- prine and fludarabine) and immunomodulatory
tial exposure of children. However, in contrast agents (e.g., thalidomide and lenalidomide) may
to ionizing radiation, no biological pathways also increase AML risk (Mailankody et al, 2011;
have been identified and findings for adults Smith et al, 2011b; Palumbo et al, 2014).
have been inconsistent (Feychting & Ahlbom, In addition to chemotherapy, weaker asso-
1994; Verkasalo, 1996; Tynes & Haldorsen, 2003; ciations have been claimed for several drugs,
683

Leukemias 683

including the antibiotic chloramphenicol effects for alcohol (Ma et al, 2009; Leal et al, 2014;
and nonsteroidal anti-​ inflammatory drugs Morton et  al, 2014), and equivocal effects for
(NSAIDS); although medical-​ record linkage most dietary factors (Ma et al, 2009; Chen et al,
studies have generally failed to confirm any asso- 2013; Yamamura et al, 2013; Morton et al, 2014).
ciations (Doody et al, 1996; Traversa et al, 1998; Most studies have, however, reported a weak pos-
Smith et al, 2000; Bonaventure et al, 2015). itive association with smoking (Musselman et al,
2013; Fircanis et al, 2014; Morton et al, 2014); the
Chemical Work Place Exposures 2004 US Surgeon General’s Report concluding
A number of potentially hazardous workplace that, “The evidence is sufficient to infer a causal
chemicals have been associated with leukemia. relationship between smoking and acute myeloid
In 2009, the International Agency for Research leukemia.”
on Cancer (IARC) determined that the accumu-
lated evidence was sufficient to conclude that Anthropometric Measures
exposure to benzene (a key component of sev- Heavy birthweight (most commonly defined as
eral manufacturing processes including leather, ≥ 3,500g) and/​or accelerated fetal growth is one
rubber, paints, detergents, pesticides, and poly- of the few established risk factors for childhood
mers) could cause AML, and possibly ALL and leukaemia, principally B-​ ALL (Hjalgrim et al
CLL (Baan et al, 2009; IARC, 2012). In addition, 2003, 2004; Milne et al, 2013; Roman et al, 2013).
exposures to compounds such as butadiene and With respect to leukemia at older ages (lymphoid
styrene used in rubber-​manufacturing were also and myeloid), the evidence for associations with
highlighted as potential causes of AML, CML, anthropometric measures (in-​utero, childhood,
and CLL; as was formaldehyde, where pregnancy and/​or adulthood) is far less consistent (Wolk et
exposure was linked to leukemia in offspring as al, 2001; Reeves et al, 2007; Fernberg et al, 2007;
well as to AML in the individual. Although the Söderberg et al, 2009; Strom et al, 2009; Castillo
IARC working group concluded that the evidence et al, 2012; Saberi Hosnijeh et al, 2013). In this
for formaldehyde was sufficient for AML, the context, it is also important to note that size at
association has not been consistently observed birth, as well as various measures of growth
(Blair et  al, 1990; Cole et  al, 2010; Checkoway throughout life, have been associated with sev-
et al, 2012, 2015; IARC, 2012; Gentry et al, 2013; eral cancers (Ahlgren et al, 2007; Green et al,
Coggon et al, 2014). 2011). The generality of this relationship lead-
With respect to other potentially adverse ing to suggestions that a common pathway could
occupational exposures, modest increased rates be involved; favored candidates including insu-
of leukemia among farmers and other agricul- lin-​like growth factors, estrogens and stem cells
tural workers have been observed in some, but (Ahlgren et al, 2007; Ross J, 2012; Milne et al
not all, studies (IARC, 2012). However, the expo- 2013; O’Neill 2015).
sures encountered in such occupations, which
include fertilizers, pesticides, fuel products, and CONCLUSION
veterinary medicines as well as a range of physi- Descriptive patterns and trends generated by
cal and biological agents, are complex and defin- population-​based cancer registries provide the
itive associations have yet to emerge. In order to bedrock for the development of testable etio-
investigate such potential relationships in more logical hypotheses. The evolving classification
depth, future epidemiological studies would ben- of hematological malignancies has, however,
efit from the use of exposure biomarkers. proved to be a major challenge, where poor
coding standardization and incomplete capture
Smoking, Diet, and Alcohol have continued to inhibit reliable comparisons
The marked life-​style and within-​country socio- between populations and overtime. While this
economic associations consistently observed for situation is improving in some countries and spe-
commoner cancers such as lung, breast, and col- cialist registries, greater efforts need to be made
orectum, are not major features of leukemia epide- by others to identify all myeloid and lymphoid
miology (Ji & Hemminki, 2005; Bhayat et al, 2009; subtypes characterized in the latest WHO clas-
Smith et  al, 2011a); suggesting that lifestyle fac- sification, as well as to track cancer progressions
tors are not major determinants of leukemogen- and transformations.
esis. In particular, epidemiological studies have In common with other cancers, leuke-
generally reported nonexistent and/​or protective mias have both genetic and environmental
684

684 Textbook of Cancer Epidemiology

processes have been made; and this increase in

TABLE 28-​2  EPIDEMIOLOGY
biological knowledge will hopefully lead to an
OF LEUKEMIA: RISK FACTOR SUMMARY.
increased understanding of the determinants of
Risk factor Direction the various leukemia subtypes. A summary of the
of effect* current state of knowledge for leukemia risk fac-
tors is presented in Table 28-​2.
Well-​confirmed risk factors With respect to treatment, hemato-​oncology
Trisomy 21 (acute leukaemias) ↑↑ is one of the most rapidly moving areas of can-
Other inherited predispostion ↑↑
cer research, and the survival improvements
syndromes
seen for subtypes such as CML, APL, and B-​
ALL represent some of the major success sto-
HTLV-​1 infection (ATLL) ↑↑
ries of cancer treatment this century. Despite
HIV infection ↑↑ these improvements, the outlook for several
Ionizing radiation ↑↑ leukemia subtypes, particularly those occur-
Chemotherapy for previous cancers ↑↑ ring at older ages, remains poor. Many prom-
ising novel therapies and targeted agents are,
Heavy birtheight (B-​ALL in children) ↑
however, being tested. If these curative and life-​
Probable relationship exists, based prolonging treatments prove to be successful,
on substantial data the prevalence of individuals living with leuke-
Benzene ↑
mia, or surviving it, will increase markedly; and,
as populations age, epidemiological attention
Butadiene and styrene ↑ may well to shift to the health problems of these
Smoking (AML) ↑ long-​term survivors.
Weak, if any relationship exists,
based on substantial data REFERENCES
Abel GA, Friese CR, Neville BA, Wilson KM, Hastings
Extremely low frequency magnetic ↑
BT, Earle CC et al. Referrals for suspected hemato-
fields
logic malignancy: a survey of primary care physi-
Nonsteroidal anti-​inflammatory drugs ↑ cians. Am J Hematol 2012;87:634–​6.
Farming occupations ↑ Ahlbom A, Day N, Feychting M, Roman E, Skinner
J, Dockerty J et  al. A pooled analysis of mag-
Inconsistent findings or limited netic fields and childhood leukaemia. Br J Cancer
study to date 2000;83:692–​8.
Obesity ↑ Ahlgren M, Wohlfahrt J, Olsen LW, Sørensen TIA, and
Melbye M. Birth weight and risk of cancer. Cancer
Chloramphenicol ↑
2007, 110, 412–​9.
Formaldehyde ↑ Apperley JF. Chronic myeloid leukaemia. Lancet Lond
Engl 2015;385:1447–​59.
*Arrows indicate the approximate magnitude of the relationship: ↑,
slight to moderate increase in risk; ↑↑, moderate to large increase Austin LM, Frush DP. Compendium of national
in risk; ↓, slight to moderate decrease in risk; ↓↓, moderate to large guidelines for imaging the pregnant patient. Am
decrease in risk J Roentgenol 2011;197:W737–​46.
Baan R, Grosse Y, Straif K, Secretan B, El Ghissassi F,
Bouvard V et al. A review of human carcinogens—​
components to their etiology, the relative con-
Part F: Chemical agents and related occupations.
tribution of each varying from one subtype to
Lancet Oncol 2009;10:1143–​4.
another. With respect to exposures, there has Babushok DV, Bessler M, Olson TS. Genetic predis-
been relatively little progress in recent decades; position to myelodysplastic syndrome and acute
with known risk factors (e.g., chemotherapy, myeloid leukemia in children and young adults.
ionizing radiation, and certain viral infections) Leuk Lymphoma 2016;57:520–​36.
explaining only a small proportion of the total Bachireddy P, Burkhardt UE, Rajasagi M, Wu C.
disease burden, and many other suggested fac- Haematological malignancies:  at the forefront of
tors (e.g., nonionizing radiation and antibi- immunotherapeutic innovation. Nat Rev Cancer
otic usage) largely leading to negative results. 2015;15:201–​15.
By contrast, major strides in understanding the Bhayat F, Das-​ Gupta E, Smith C, McKeever T,
underpinning genetic and molecular pathogenic Hubbard R. The incidence of and mortality from
685

Leukemias 685

leukaemias in the UK: a general population-​based Checkoway H, Boffetta P, Mundt DJ, Mundt KA.
study. BMC Cancer 2009;9:252. Critical review and synthesis of the epidemio-
Bithell JF, Stiller CA. A new calculation of the car- logic evidence on formaldehyde exposure and
cinogenic risk of obstetric x-​ raying. Stat Med risk of leukemia and other lymphohemato-
1988;7:857–​64. poietic malignancies. Cancer Causes Control
Blair A, Saracci R, Stewart PA, Hayes RB, Shy C. 2012;23:1747–​66.
Epidemiologic evidence on the relationship Checkoway H, Dell LD, Boffetta P, Gallagher AE,
between formaldehyde exposure and cancer. Crawford L, Lees PS et al. Formaldehyde exposure
Scand J Work Environ Health 1990;16:381–​93. and mortality risks from acute myeloid leukemia
Boice JD. Radiation epidemiology and recent paedi- and other lymphohematopoietic malignancies in
atric computed tomography studies. Ann ICRP the US National Cancer Institute cohort study
2015;44:236–​48. of workers in formaldehyde industries. J Occup
Boice JD, Bigbee WL, Mumma MT, Blot WJ. Cancer Environ Med 2015;57:785–​94.
mortality in counties near two former nuclear Chen G-​C, Lv D-​B, Pang Z, Liu QF. Fruits and veg-
materials processing facilities in Pennsylvania, etables consumption and risk of non-​Hodgkin’s
1950–​1995. Health Phys 2003;85:691–​700. lymphoma: a meta-​analysis of observational stud-
Boice JD, Miller RW. Childhood and adult cancer ies. Int J Cancer 2013;133:190–​200.
after intrauterine exposure to ionizing radiation. Coggon D, Ntani G, Harris EC, Palmer KT. Upper
Teratology 1999;59:227–​33. airway cancer, myeloid leukemia, and other
Bonaventure A, Simpson J, Ansell P, Roman E, cancers in a cohort of British chemical work-
Lightfoot T. Prescription drug use during preg- ers exposed to formaldehyde. Am J Epidemiol
nancy and risk of childhood cancer—​Is there an 2014;179:1301–​11.
association? Cancer Epidemiol 2015;39:73–​8. Cole P, Adami H-​ O, Trichopoulos D, Mandel J.
Buls N, Covens P, Nieboer K, Van Schuerbeek P, Formaldehyde and lymphohematopoietic can-
Devacht P, Eloot L et al. Dealing with pregnancy cers: a review of two recent studies. Regul Toxicol
in radiology:  a thin line between science, social Pharmacol 2010;58:161–​6.
and regulatory aspects. JBR-​BTR 2009;92:271–​9. Cooke JV. The occurrence of leukemia. Blood
Burnett AK. Treatment of acute myeloid leukemia: are 1954;9:340–​7.
we making progress? ASH Educ Program Book Curtis RE, Hankey BF, Myers MH, Young JL Jr. Risk
2012;2012:1–​6. of leukemia associated with the first course of
Byrd J, Jones J, Woyach J, Johnson A, Flynn J. cancer treatment: an analysis of the Surveillance,
Entering the Era of Targeted Therapy for Chronic Epidemiology, and End Results Program experi-
Lymphocytic Leukemia: Impact on the Practicing ence. J Natl Cancer Inst 1984;72:531–​44.
Clinician. J Clin Oncol 2014;32:3039–47. Czuchlewski DR, Peterson LC. Myeloid neoplasms
Campo E, Swerdlow SH, Harris NL, Pileri S, Stein with germline predisposition:  a new provisional
H, Jaffe ES. The 2008 WHO classification entity within the World Health Organization
of lymphoid neoplasms and beyond:  evolv- Classification. Surg Pathol Clin 2016;9:165–​76.
ing concepts and practical applications. Blood Derolf AR, Kristinsson SY, Andersson TM-​ L,
2011;117:5019–​32. Landgren O, Dickman PW, Björkholm M.
Castillo JJ, Reagan JL, Ingham RR, Furman M, Dalia Improved patient survival for acute myeloid leu-
S, Merhi B et  al. Obesity but not overweight kemia: a population-​based study of 9729 patients
increases the incidence and mortality of leukemia diagnosed in Sweden between 1973 and 2005.
in adults:  a meta-​analysis of prospective cohort Blood 2009;113:3666–​72.
studies. Leuk Res 2012;36:868–​75. Dockerty JD, Draper G, Vincent T, Rowan SD, Bunch
Cerhan JR, Slager SL. Familial predisposition KJ. Case-​control study of parental age, parity and
and genetic risk factors for lymphoma. Blood socioeconomic level in relation to childhood can-
2015;126:2265–​73. cers. Int J Epidemiol 2001;30:1428–​37.
Chang JS, Metayer C, Fear NT, Reinier K, Yin X, Döhner H, Estey EH, Amadori S, Appelbaum FR,
Urayama K et al. Parental social contact in the work Büchner T, Burnett AK et al. Diagnosis and man-
place and the risk of childhood acute lymphoblas- agement of acute myeloid leukemia in adults: rec-
tic leukaemia. Br J Cancer 2007;97:1315–​21. ommendations from an international expert
Chang JS, Zhou M, Buffler PA, Chokkalingam AP, panel, on behalf of the European LeukemiaNet.
Metayer C, Wiemels JL. Profound deficit of Blood 2010;115:453–​74.
IL10 at birth in children who develop childhood Dombret H, Gardin C. An update of current treat-
acute lymphoblastic leukemia. Cancer Epidemiol ments for adult acute myeloid leukemia. Blood
Biomark Prev 2011;20:1736–​40. 2016;127: 53–61.
68

686 Textbook of Cancer Epidemiology

Doody MM, Linet MS, Glass AG, Curtis RE, Pottern Gatta G, Botta L, Rossi S, Aareleid T, Bielska-​Lasota M,
LM, Rush BB et al. Risks of non-​Hodgkin’s lym- Clavel J et al. Childhood cancer survival in Europe
phoma, multiple myeloma, and leukemia associ- 1999–​2007: results of EUROCARE-​5—​a population-​
ated with common medications. Epidemiol Camb based study. Lancet Oncol 2014;15:35–​47.
Mass 1996;7:131–​9. Gavin A, Rous B, Marcos-​Gragera R, Middleton R,
Eichhorst B, Robak T, Montserrat E, Ghia P, Hillmen Steliarova-​Foucher E, Maynadie M et al. Towards
P, Hallek M et  al. Chronic lymphocytic leukae- optimal clinical and epidemiological registra-
mia:  ESMO Clinical Practice Guidelines for tion of haematological malignancies:  guidelines
diagnosis, treatment and follow-​up. Ann Oncol for recording progressions, transformations
2015;26:v78–​84. and multiple diagnoses. Eur J Cancer 1990
Elliott P, Shaddick G, Douglass M, de Hoogh K, Briggs 2015;51:1109–​22.
DJ, Toledano MB. Adult cancers near high-​voltage Gentry PR, Rodricks JV, Turnbull D, Bachand A, Van
overhead power lines. Epidemiol Camb Mass Landingham C, Shipp AM et  al. Formaldehyde
2013;24:184–​90. exposure and leukemia: critical review and reeval-
Enciso-​Mora V, Hosking FJ, Sheridan E, Kinsey SE, uation of the results from a study that is the focus
Lightfoot T, Roman E et  al. Common genetic for evidence of biological plausibility. Crit Rev
variation contributes significantly to the risk of Toxicol 2013;43:661–​70.
childhood B-​cell precursor acute lymphoblastic Gessain A, Cassar O. Epidemiological aspects and
leukemia. Leukemia 2012;26:2212–​5. world distribution of HTLV-​ 1 infection. Front
Fear NT, Simpson J, Roman E. Childhood cancer Microbiol 2012;3:388.
and social contact:  the role of paternal occupa- Geyer MB, Jacobson JS. Epidemiology of hemato-
tion (United Kingdom). Cancer Causes Control logical malignancies of children, adolescents,
2005;16:1091–​7. and young adults. In:  Cairo MS, Perkins SL
Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers (eds.) Hematological Malignancies in Children,
C, Rebelo M et  al. Cancer incidence and mor- Adolescents and Young Adults. World Scientific,
tality worldwide:  sources, methods and major 2012:  1–​ 13. Hackensack, NJ:  World Scientific.
patterns in GLOBOCAN 2012. Int J Cancer Available from:  http://​www.worldscientific.com/​
2015;136:E359–​86. doi/​abs/​10.1142/​9789814299619_​0001.
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser Gibson TM, Morton LM, Shiels MS, Clarke CA, Engels
S, Mathers C, Rebelo M, Parkin DM, Forman D, EA. Risk of non-​Hodgkin lymphoma subtypes in
Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence HIV-​infected people during the HAART era:  a
and Mortality Worldwide: IARC CancerBase No. population-​based study. AIDS 2014;28:2313–​8.
11 [Internet]. Lyon, France: International Agency Giles D, Hewitt D, Stewart A, Webb J. Malignant dis-
for Research on Cancer; 2013. Available from: ease in childhood and diagnostic irradiation in
http://globocan.iarc.fr, accessed 2017. utero. Lancet 1956;271:447.
Fernberg P, Odenbro A, Bellocco R, Boffetta P, Gilham C, Peto J, Simpson J, Roman E, Eden TO,
Pawitan Y, Zendehdel K et al. Tobacco use, body Greaves MF et al. Day care in infancy and risk of
mass index, and the risk of leukemia and multiple childhood acute lymphoblastic leukaemia: findings
myeloma:  a nationwide cohort study in Sweden. from UK case-​control study. BMJ 2005;330:1294.
Cancer Res 2007;67:5983–​6. Gluzman DF, Sklyarenko LM, Zavelevich MP, Koval
Feychting M, Ahlbom A. Magnetic fields, leukemia, SV, Ivanivska TS, Rodionova NK. Overview on
and central nervous system tumors in Swedish association of different types of leukemias with
adults residing near high-​ voltage power lines. radiation exposure. Exp Oncol 2015;37:89–​93.
Epidemiol Camb Mass 1994;5:501–​9. Granfeldt Østgård LS, Medeiros BC, Sengeløv H,
Fircanis S, Merriam P, Khan N, Castillo JJ. The relation Nørgaard M, Andersen MK, Dufva IH et  al.
between cigarette smoking and risk of acute mye- Epidemiology and clinical significance of second-
loid leukemia: an updated meta-​analysis of epide- ary and therapy-​related acute myeloid leukemia: a
miological studies. Am J Hematol 2014;89:E125–​32. national population-​ based cohort study. J Clin
Folley JH, Borges W, Yamawaki T. Incidence of leuke- Oncol 2015;33:3641–​9.
mia in survivors of the atomic bomb in Hiroshima Greaves M. Leukaemia “firsts” in cancer research and
and Nagasaki, Japan. Am J Med 1952;13:311–​21. treatment. Nat Rev Cancer 2016;16:163–​72.
Gardner MJ, Snee MP, Hall AJ, Powell CA, Downes Green J, Cairns B, Casabonne D, Wright F, Reeves
S, Terrell JD. Results of case-​control study of leu- G, Beral V. Height and cancer incidence in the
kaemia and lymphoma among young people near Million Women Study:  prospective cohort, and
Sellafield nuclear plant in West Cumbria. BMJ meta-​analysis of prospective studies of height and
1990;300:423–​9. total cancer risk. Lancet Onclogy 2011; 12:785-​94
687

Leukemias 687

Grünwald HW, Rosner F. Acute myeloid leukemia fol- Carcinogenic Risks to Humans, Vol. 100 F. This
lowing treatment of Hodgkin’s disease:  a review. publication represents the views and expert opin-
Cancer 1982;50:676–​83. ions of an IARC working group on the evaluation
HAEMACARE Working Group. Manual for cod- of carcinogenic risks to humans, which met in
ing and reporting haematological malignancies. Lyon, October 20–​27, 2009. Lyon: IARC, 2012.
Tumori 2010;96:i–​A32. Ishitsuka K, Tamura K. Human T-​cell leukaemia virus
Hasle H, Clemmensen IH, Mikkelsen M. Risks of type I  and adult T-​ cell leukaemia-​
lymphoma.
leukaemia and solid tumours in individuals with Lancet Oncol 2014;15:e517–​26.
Down’s syndrome. Lancet 2000;355:165–​9. Jaffe ES, Harris NL, Stein H, Vardiman JW. Pathology
Henshaw PS, Hawkins JW. Incidence of leukemia in and Genetics of Tumours of Haematopoietic and
physicians. J Natl Cancer Inst 1944;4:339–​46. Lymphoid Tissues. Lyon and Oxford: IARC Press;
Hill DA, Gridley G, Cnattingius S, Mellemkjaer L, Oxford University Press (distributor), 2001.
Linet M, Adami HO et  al. Mortality and cancer Ji J, Hemminki K. Occurrences of leukemia subtypes
incidence among individuals with Down syn- by socioeconomic and occupational groups in
drome. Arch Intern Med 2003;163:705–​11. Sweden. J Occup Environ Med Am Coll Occup
Hitzler JK. Acute megakaryoblastic leukemia in Down Environ Med 2005;47:1131–​40.
syndrome. Pediatr Blood Cancer 2007;49:1066–​9. Jourdan-​Da Silva N, Perel Y, Méchinaud F, Plouvier
Hjalgrim LL, Westergaard T, Rostgaard K, E, Gandemer V, Lutz P et  al. Infectious diseases
Schmiegelow K, Melbye M, Hjalgrim H et al. Birth in the first year of life, perinatal characteris-
weight as a risk factor for childhood leukemia: a tics and childhood acute leukaemia. Br J Cancer
meta-​analysis of 18 epidemiologic studies. Am. 2004;90:139–​45.
J. Epidemiol. 2003; 158: 724–​35. Kaldor JM, Day NE, Band P, Choi NW, Clarke EA,
Hjalgrim LL, Rostgaard K, Hjalgrim H, Westergaard Coleman MP et al. Second malignancies following
T, Thomassen H, Forestier E et  al. Birth weight testicular cancer, ovarian cancer and hodgkin’s dis-
and risk for childhood leukemia in Denmark, ease:  an international collaborative study among
Sweden, Norway, and Iceland. J. Natl. Cancer Inst. cancer registries. Int J Cancer 1987;39:571–​85.
2004; 96: 1549–​56. Kaldor JM, Day NE, Clarke EA, Van Leeuwen FE,
Höglund M, Sandin F, Hellström K, Björeman M, Henry-​Amar M, Fiorentino MV et  al. Leukemia
Björkholm M, Brune M et  al. Tyrosine kinase following Hodgkin’s disease. N Engl J Med
inhibitor usage, treatment outcome, and 1990;322:7–​13.
prognostic scores in CML:  report from the Keegan TJ, Bunch KJ, Vincent TJ, King JC, O’Neill KA,
population-​based Swedish CML registry. Blood Kendall GM et al. Case-​control study of paternal
2013;122:1284–​92. occupation and childhood leukaemia in Great
Howell DA, Smith AG, Jack A, Patmore R, Macleod Britain, 1962-​2006. Br J Cancer 2012;107:1652–​9.
U, Mironska E et al. Time-​to-​diagnosis and symp- Kheifets L, Monroe J, Vergara X, Mezei G, Afifi AA.
toms of myeloma, lymphomas and leukaemias: a Occupational electromagnetic fields and leukemia
report from the Haematological Malignancy and brain cancer: an update to two meta-​analyses.
Research Network. BMC Hematol 2013;13:9. J Occup Environ 2008;50:677–​88.
Howlader N, Noone A, Krapcho M, Garshell J, Miller Kinlen L. Population mixing and childhood leukae-
D, Altekruse S et al. SEER Cancer Statistics Review, mia. Eur J Epidemiol 2015;30:1331.
1975–​ 2012. Bethesda, MD:  National Cancer Kinlen LJ. High-​contact paternal occupations, infec-
Institute. http://​seer.cancer.gov/​csr/​1975_​2012/​. tion and childhood leukaemia:  five studies of
Based on November 2014 SEER Data Submission, unusual population-​mixing of adults. Br J Cancer
posted to the SEER website, April 2015, 2014. 1997;76:1539–​45.
Hsu W-​ L, Preston DL, Soda M, Sugiyama H, Kinlen L, Jiang J, Hemminki K. A case-​control study
Funamoto S, Kodama K et  al. The incidence of of childhood leukaemia and paternal occupa-
leukemia, lymphoma and multiple myeloma tional contact level in rural Sweden. Br J Cancer
among atomic bomb survivors: 1950-​2001. Radiat 2002;86:732–​7.
Res 2013;179:361–​82. Koeman T, van den Brandt PA, Slottje P, Schouten LJ,
Huang WY, Muo CH, Lin CY, Jen YM, Yang MH, Lin Goldbohm RA, Kromhout H et  al. Occupational
JC et al. Paediatric head CT scan and subsequent extremely low-​frequency magnetic field exposure
risk of malignancy and benign brain tumour:  a and selected cancer outcomes in a prospective Dutch
nation-​wide population-​based cohort study. Br J cohort. Cancer Causes Control 2014;25:203–​14.
Cancer 2014;110:2354–​60. Koshurnikova NA, Buldakov LA, Bysogolov GD,
IARC, ed. Chemical Agents and Related Occupations. Bolotnikova MG, Komleva NS, Peternikova VS.
IARC Monographs on the Evaluation of Mortality from malignancies of the hematopoietic
68

688 Textbook of Cancer Epidemiology

and lymphatic tissues among personnel of the myeloid leukemia in the NIH–​AARP cohort. Am
first nuclear plant in the USSR. Sci Total Environ J Epidemiol 2009:kwp371.
1994;142:19–​23. McKeown SR, Hatfield P, Prestwich RJD, Shaffer
Krivit W, Good RA. Simultaneous occurrence of mon- RE, Taylor RE. Radiotherapy for benign disease;
golism and leukemia: report of a nationwide sur- assessing the risk of radiation-​induced cancer fol-
vey. AMA J Dis Child 1957;94:289–​93. lowing exposure to intermediate dose radiation.
Laurier D, Grosche B, Auvinen A, Clavel J, Cobaleda Br J Radiol 2015;88:20150405.
C, Dehos A et al. Childhood leukaemia risks: from Milne E, Greenop KR, Metayer C, Schuz J, Petreidou
unexplained findings near nuclear installations to E, Pombo-​de-​Oliveira M et  al. Fetal growth and
recommendations for future research. J Radiol childhood acute lymphoblastic leukemia: findings
Prot 2014;34:R53–​68. from the childhood leukemia international con-
Leal AD, Thompson CA, Wang AH, Vierkant RA, sortium. Int. J. Cancer 2013; 133, 2968–​79.
Habermann TM, Ross JA et  al. Anthropometric, Morton LM, Dores GM, Tucker MA, Kim CJ, Onel K,
medical history and lifestyle risk factors for myelo- Gilbert ES et  al. Evolving risk of therapy-​related
proliferative neoplasms in the Iowa Women’s Health acute myeloid leukemia following cancer chemo-
Study cohort. Int J Cancer 2014;134:1741–​50. therapy among adults in the United States, 1975–​
Leone G, Fianchi L, Pagano L, Voso MT. Incidence 2008. Blood 2013;121:2996–​3004.
and susceptibility to therapy-​related myeloid neo- Morton LM, Slager SL, Cerhan JR, Wang SS, Vajdic
plasms. Chem Biol Interact 2010;184:39–​45. CM, Skibola CF et  al. Etiologic heterogeneity
Leone G, Pagano L, Ben-​ Yehuda D, Voso MT. among non-​ Hodgkin lymphoma subtypes:  the
Therapy-​ related leukemia and myelodyspla- Interlymph Non-​ Hodgkin Lymphoma Subtypes
sia:  susceptibility and incidence. Haematologica Project. JNCI Monogr 2014;2014:130–​44.
2007;92:1389–​98. Muirhead CR, O’Hagan JA, Haylock RGE, Phillipson
Lo-​Coco F, Cicconi L, Breccia M. Current standard MA, Willcock T, Berridge GL et al. Mortality and
treatment of adult acute promyelocytic leukaemia. cancer incidence following occupational radiation
Br J Haematol 2016;172:841–​54. exposure: third analysis of the National Registry for
MacArthur AC, McBride ML, Spinelli JJ, Tamaro S, Radiation Workers. Br J Cancer 2009;100:206–​12.
Gallagher RP, Theriault GP. Risk of childhood leu- Musselman JR, Blair CK, Cerhan JR, Nguyen P, Hirsch
kemia associated with vaccination, infection, and B, Ross JA. Risk of adult acute and chronic mye-
medication use in childhood:  the Cross-​Canada loid leukemia with cigarette smoking and cessa-
Childhood Leukemia Study. Am J Epidemiol tion. Cancer Epidemiol 2013;37:410–​6.
2008;167:598–​606. Neglia JP, Linet MS, Shu XO, Severson RK, Potter JD,
Mailankody S, Pfeiffer RM, Kristinsson SY, Korde N, Mertens AC et  al. Patterns of infection and day
Bjorkholm M, Goldin LR et al. Risk of acute mye- care utilization and risk of childhood acute lym-
loid leukemia and myelodysplastic syndromes phoblastic leukaemia. Br J Cancer 2000;82:234–​40.
after multiple myeloma and its precursor disease O’Neill KA, Murphy MF, Bunch KJ, Puumal S, Carozza
(MGUS). Blood 2011;118:4086–​92. S, Chow E et  al. Infant birthweight and risk of
Malcovati L, Hellström-​Lindberg E, Bowen D, Adès childhood cancer:  international population-​
L, Cermak J, Del Cañizo C et  al. Diagnosis and based case control studies of 40 000 cases. Int.
treatment of primary myelodysplastic syndromes J. Epidemiol.2015, 44, 153–​168.
in adults:  recommendations from the European Palumbo A, Bringhen S, Kumar SK, Lupparelli G,
LeukemiaNet. Blood 2013;122:2943–​64. Usmani S, Waage A et al. Second primary malig-
Mathers C, Boerma T. Mortality measurement mat- nancies with lenalidomide therapy for newly diag-
ters:  improving data collection and estimation nosed myeloma:  a meta-​ analysis of individual
methods for child and adult mortality. PLoS Med patient data. Lancet Oncol 2014;15:333–​42.
2010;7:e1000265. Parkin D. The evolution of the population-​based can-
Mathews JD, Forsythe AV, Brady Z, Butler MW, cer registry. Nat Rev Cancer 2006;6:603–​12.
Goergen SK, Byrnes GB et al. Cancer risk in 680 000 Pearce MS, Salotti JA, Little MP, McHugh K, Lee C,
people exposed to computed tomography scans in Kim KP et al. Radiation exposure from CT scans
childhood or adolescence:  data linkage study of in childhood and subsequent risk of leukaemia
11 million Australians. BMJ 2013;346:f2360. and brain tumours:  a retrospective cohort study.
Ma X, Buffler PA, Selvin S, Matthay KK, Wiencke JK, Lancet 2012;380:499–​505.
Wiemels JL et al. Daycare attendance and risk of Pedersen C, Johansen C, Schüz J, Olsen JH, Raaschou-​
childhood acute lymphoblastic leukaemia. Br J Nielsen O. Residential exposure to extremely
Cancer 2002;86:1419–​24. low-​frequency magnetic fields and risk of child-
Ma X, Park Y, Mayne ST, Wang R, Sinha R, hood leukaemia, CNS tumour and lymphoma in
Hollenbeck AR et  al. Diet, lifestyle, and acute Denmark. Br J Cancer 2015;113:1370–​4.
689

Leukemias 689

Poole C, Greenland S, Luetters C, Kelsey JL, Mezei G. myeloid leukemia in the European Prospective
Socioeconomic status and childhood leukaemia: a Investigation into Cancer and Nutrition (EPIC).
review. Int J Epidemiol 2006;35:370–​84. Cancer Causes Control 2013;24:427–​38.
Pui C-​H, Yang JJ, Hunger SP, Pieters R, Schrappe M, Sava GP, Speedy HE, Houlston RS. Candidate gene
Biondi A et  al. Childhood acute lymphoblastic association studies and risk of chronic lympho-
leukemia:  progress through collaboration. J Clin cytic leukemia:  a systematic review and meta-​
Oncol 2015;33:2938–​48. analysis. Leuk Lymphoma 2014;55:160–​7.
Radivoyevitch T, Sachs RK, Gale RP, Smith MR, Hill Schmiegelow K. Treatment-​related toxicities in child-
BT. Ionizing radiation exposures in treatments ren with acute lymphoblastic leukaemia pre-
of solid neoplasms are not associated with subse- disposition syndromes. Eur J Med Genet 2016,
quent increased risks of chronic lymphocytic leu- doi: 10.1016/​j.ejmg.2016.02.006.
kemia. Leuk Res 2016;43:9–​12. Schulze-​Rath R, Hammer GP, Blettner M. Are pre-​or
Reeves GK, Pirie K, Beral V, Green J, Spencer E, Bull postnatal diagnostic x-​rays a risk factor for child-
D, and the Million Women Study Collaboration. hood cancer? A systematic review. Radiat Environ
Cancer incidence and mortality in relation to body Biophys 2008;47:301–​12.
mass index in the Million Women Study: cohort Schüz J, Dasenbrock C, Ravazzani P, Röösli M, Schär
study. BMJ 2007;335:1134. P, Bounds PL et al. Extremely low-​frequency mag-
Roman E, Doyle P, Maconochie N, Davies G, Smith netic fields and risk of childhood leukemia:  A
PG, Beral V. Cancer in children of nuclear indus- risk assessment by the ARIMMORA consortium.
try employees:  report on children aged under Bioelectromagnetics 2016. doi: 10.1002/​bem.21963.
25 years from nuclear industry family study. BMJ Schüz J, Luta G, Erdmann F, Ferro G, Bautz A, Simony
1999;318:1443–​50. SB et al. Birth order and risk of childhood cancer
Roman E, Simpson J, Ansell P, Kinsey S, Mitchell CD, in the Danish birth cohort of 1973–​2010. Cancer
McKinney PA et  al. Childhood acute lympho- Causes Control 2015;26:1575–​82.
blastic leukemia and infections in the first year of Seif AE. Pediatric leukemia predisposition syn-
life: a report from the United Kingdom Childhood dromes:  clues to understanding leukemogenesis.
Cancer Study. Am J Epidemiol 2007;165:496–​504. Cancer Genet 2011;204:227–​44.
Roman E, Lightfoot T, Smith AG, Forman M, Linet Smith A, Crouch S, Lax S, Li J, Painter D, Howell D
M, Robison L et  al. Childhood acute lympho- et  al. Lymphoma incidence, survival and preva-
blastic leukaemia and birthweight:  insights from lence 2004–​2014: sub-​type analyses from the UK’s
a pooled analysis of case-​ control data from Haematological Malignancy Research Network.
Germany, the United Kingdom and the United Br J Cancer 2015;112:1575–​84.
States. Eur. J. Cancer 2013; 49:1437–​47. Smith A, Howell D, Patmore R, Jack A, Roman E.
Roman E, Smith A, Appleton S, Crouch S, Kelly Incidence of haematological malignancy by
R, Kinsey C et  al. Myeloid malignancies in the sub-​ type:  a report from the Haematological
real-​world:  occurrence, progression and survival Malignancy Research Network. Br J Cancer
in the UK’s population-​ based Haematological 2011a;105:1684–​92.
Malignancy Research Network 2004-​15. Cancer Smith A, Roman E, Simpson J, Ansell P, Fear NT, Eden
Epidemiology 2016; 42:186-​98 T. Childhood leukaemia and socioeconomic sta-
Rosenbaum PF, Buck GM, Brecher ML. Allergy and tus:  fact or artefact? A  report from the United
infectious disease histories and the risk of child- Kingdom childhood cancer study (UKCCS). Int J
hood acute lymphoblastic leukaemia. Paediatr Epidemiol 2006;35:1504–​13.
Perinat Epidemiol 2005;19:152–​64. Smith AG, Dovey GJ, Cartwright RA. Topical chloram-
Ross JA. Birth weight and childhood leukemia: time to phenicol and the risk of acute leukaemia in adults.
tackle bigger lessons. Pediatr. Blood Cancer 2012; Pharmacoepidemiol Drug Saf 2000;9:215–​9.
58(1), 1–​2. Smith AG, Painter D, Howell DA, Evans P, Smith
Rudant J, Lightfoot T, Urayama KY, Petridou E, Dockerty G, Patmore R et  al. Determinants of survival in
JD, Magnani C et al. Childhood acute lymphoblastic patients with chronic myeloid leukaemia treated
leukemia and indicators of early immune stimula- in the new era of oral therapy:  findings from a
tion: a childhood leukemia international consortium UK population-​based patient cohort. BMJ Open
study. Am J Epidemiol 2015;181:549–​62. 2014;4:e004266.
Ruhl J, Adamo M, Dickie L. Hematopoietic and lym- Smith A, Howell D, Patmore R et  al. Incidence of
phoid neoplasm coding manual. National Cancer haematological malignancy by sub-​type: a report
Institute, Bethesda, MD 20850-​9765, 2015. from the Haematological Malignancy Research
Saberi Hosnijeh F, Romieu I, Gallo V, Riboli E, Network. Br J Cancer 2011a;105:1684–​92.
Tjønneland A, Halkjær J et  al. Anthropometric Smith A, Roman E, Simpson J et  al. Childhood
characteristics and risk of lymphoid and leukaemia and socioeconomic status:  fact or
690

690 Textbook of Cancer Epidemiology

artefact? A  report from the United Kingdom neoplasms: ESMO clinical practice guidelines for
childhood cancer study (UKCCS). Int J Epidemiol diagnosis, treatment and follow-​up. Ann Oncol
2006;35:1504–​13. 2015;26:v85–​99.
Smith MA, McCaffrey RP, Karp JE. The secondary Vardiman JW, Thiele J, Arber DA, Brunning RD,
leukemias:  challenges and research directions. J Borowitz MJ, Porwit A et  al. The 2008 revision
Natl Cancer Inst 1996;88:407–​18. of the World Health Organization (WHO) clas-
Smith MR, Neuberg D, Flinn IW, Grever MR, Lazarus sification of myeloid neoplasms and acute leu-
HM, Rowe JM et  al. Incidence of therapy-​ kemia:  rationale and important changes. Blood
related myeloid neoplasia after initial therapy for 2009;114:937–​51.
chronic lymphocytic leukemia with fludarabine-​ Verkasalo PK. Magnetic fields and leukemia—​risk for
cyclophosphamide versus fludarabine:  long-​term adults living close to power lines. Scand J Work
follow-​up of US Intergroup Study E2997. Blood Environ Health 1996;22 Suppl 2:1–​56.
2011b;118:3525–​7. Wakeford R. Childhood leukaemia following med-
Stewart A, Webb J, Hewitt D. A survey of childhood ical diagnostic exposure to ionizing radiation
malignancies. Br Med J 1958;1:1495–​508. in utero or after birth. Radiat Prot Dosimetry
Swerdlow, SH, Campo, E, Harris NL et  al. WHO 2008;132:166–​74.
Classification of Tumours of Haematopoietic and Wakeford R. The risk of leukaemia in young children
Lymphoid Tissues. 4 ed. Lyon, France: International from exposure to tritium and carbon-​14 in the
Agency for Research on Cancer, 2008. discharges of German nuclear power stations and
Söderberg KC, Kaprio J, Verkasalo PK, Pukkala E, in the fallout from atmospheric nuclear weapons
Koskenvuo M, Lundqvist E et  al. Overweight, testing. Radiat Environ Biophys 2014a;53:365–​79.
obesity and risk of haematological malignancies: a Wakeford R. Childhood leukaemia and nuclear instal-
cohort study of Swedish and Finnish twins. Eur J lations:  the long and winding road. Br J Cancer
Cancer 2009;45:1232–​8. 2014b;111:1681–​3.
Thompson PA, Kantarjian HM, Cortes JE. Diagnosis Ward G. The infective theory of acute leukaemia. Br J
and treatment of chronic myeloid leukemia in Child Dis 1917;14:10–​20.
2015. Mayo Clin Proc 2015;90:1440–​54. White-​Koning ML, Hémon D, Laurier D, Tirmarche
Traversa G, Menniti-​Ippolito F, Da Cas R, Mele A, M, Jougla E, Goubin A et  al. Incidence of child-
Pulsoni A, Mandelli F. Drug use and acute leuke- hood leukaemia in the vicinity of nuclear sites in
mia. Pharmacoepidemiol Drug Saf 1998;7:113–​23. France, 1990–​1998. Br J Cancer 2004;91:916–​22.
Tsukasaki K, Tobinai K. Human T-​cell lymphotropic Whitlock JA, Sather HN, Gaynon P, Robison LL, Wells
virus type I-​ associated adult T-​ cell leukemia-​ RJ, Trigg M et al. Clinical characteristics and out-
lymphoma:  new directions in clinical research. come of children with Down syndrome and acute
Clin Cancer Res Off J Am Assoc Cancer Res lymphoblastic leukemia:  a Children’s Cancer
2014;20:5217–​25. Group study. Blood 2005;106:4043–​9.
Tynes T, Haldorsen T. Residential and occupational WHO. International Statistical Classification of
exposure to 50 Hz magnetic fields and hemato- Diseases and Related Health Problems, 10th
logical cancers in Norway. Cancer Causes Control Revision. ICD-​ 10 Version:2015. World Health
2003;14:715–​20. Organization, 2015. http://​apps.who.int/​classifica-
Urayama KY, Ma X, Selvin S, Metayer C, Chokkalingam tions/​icd10/​browse/​2015/​en.
AP, Wiemels JL et al. Early life exposure to infec- Wolk A, Gridley G, Svensson M, Nyrén O, McLaughlin
tions and risk of childhood acute lymphoblastic JK, Fraumeni JF et al. A prospective study of obe-
leukemia. Int J Cancer 2011;128:1632–​43. sity and cancer risk (Sweden). Cancer Causes
Urquhart JD, Black RJ, Muirhead MJ, Sharp L, Control 2001;12:13–​21.
Maxwell M, Eden OB et  al. Case-​control study Yamamura Y, Oum R, Elhor Gbito KY, Garcia-​Manero
of leukaemia and non-​Hodgkin’s lymphoma in G, Strom SS. Dietary intake of vegetables, fruits,
children in Caithness near the Dounreay nuclear and meats/​beans as potential risk factors of acute
installation. BMJ 1991;302:687–​92. myeloid leukemia:  a Texas case-​ control study.
van Laar M, Stark DP, McKinney P, Parslow RC, Nutr Cancer 2013;65:1132–​40.
Kinsey SE, Picton SV et al. Population mixing for Zeller B, Gustafsson G, Forestier E, Abrahamsson
leukaemia, lymphoma and CNS tumours in teen- J, Clausen N, Heldrup J et  al. Acute leukae-
agers and young adults in England, 1996–​2005. mia in children with Down syndrome:  a
BMC Cancer 2014;14:698. population-​ based Nordic study. Br J Haematol
Vannucchi AM, Barbui T, Cervantes F, Harrison 2005;128:797–​804.
C, Kiladjian JJ, Kröger N et  al. Philadelphia Zipursky A, Doyle J. Leukemia in newborn infants
chromosome-​negative chronic myeloproliferative with Down syndrome. Leuk Res 1993;17:195.
691

EPILOGUE

The epidemiologic search for cancer causes has cancers, consortia pooling data from tens of thou-
been conducted in earnest for almost 70  years. sands of cases have now identified more than 160
This book documents the striking success that common variants association with breast cancer,
the field has had in establishing these causes for more than 180 with prostate cancer, and more
many cancer sites—​lung, breast, stomach, colo- than 75 associated with colorectal cancer. The
rectal, liver, cervical, endometrial, and skin can- smaller numbers of associations with other can-
cers, for instance—​although there are sites for cers largely reflect overwhelming external causes
which we still have few definitively established (e.g., tobacco use and lung cancer) or the smaller
causes, including leukemias, lymphomas, and numbers of cases available even to international
pancreas and brain cancers. It is natural to ask consortia. In the latter case, aggregation of more
whether we have reached an age of “diminishing cases in the next decade will almost certainly
returns” during which ever greater efforts will be lead to substantial increases in the numbers of
made to establish weaker and less prevalent risk genetic variants unambiguously associated with
factors for several cancer sites. these cancers. Finding these genetic variants
If the past is prologue to the future, then the permits us to assess gene–​environment interac-
pace of new discoveries in recent decades sug- tions with real and well-​measured genetic risks,
gests that we are still in a golden age of epidemio- instead of using the weaker surrogate of family
logic inquiry into the causes of cancer. Table E-​1 history. Interestingly, results so far suggest that
lists factors that have been generally accepted most environmental factors interact with the
in the past 30 years as highly probable causes of genetic risk on a simple multiplicative scale, and
certain specific cancers. Table E-​2 is an equally synergistic supra-​or inframultiplicative interac-
important list of exposures that recent research tions are rare. Understanding the physiological
has shown are probably not major causes of the role of these genetic variants should open up new
cancers listed, although previous research had understanding of the underlying mechanisms
suggested associations. Many of these findings of specific cancers. Whether this understanding
have been pinned down since the early 1990s, leads directly to preventive advice, chemopreven-
suggesting that the pace of epidemiologic discov- tion, and/​or therapies, only time will tell.
ery is accelerating. It is interesting then, that epidemiology is far
Since 2003, the completion of the Human from universally accepted as a successful science.
Genome Project, the discovery that much of The pages of our journals are full of thoughtful,
common genetic variation can be captured by self-​critical analyses of our methods and their
genotyping a subset of all variants through the limitations. Occasionally these doubts receive a
information from the International HapMap public airing. Almost everybody, epidemiologists
Project, and the technological developments and critics alike, accepts that the problems stem
that have made it possible to measure more than not only from the application of epidemiologic
500,000 single nucleotide polymorphisms in a methods but also from the fact that much of our
single assay, have all combined to permit us to work takes place under the spotlight of public
search genome-​wide for common variants asso- scrutiny. Thus, it is inevitable that a single prom-
ciated with specific cancers. For the common inent “positive” study of a prevalent exposure
692

692 Epilogue

TABLE E-​1   RISK FACTORS FOR SPECIFIC CANCERS ACCEPTED AS CAUSAL

SINCE THE BEGINNING OF THE 1980S

Risk factor Cancer

Helicobacter pylori infection Stomach

HPV infection Cervix
HBV, HCV infection Liver
Passive smoking Lung
Postmenopausal hormones Breast, Endometrium, Ovary
Tubal ligation (protective) Ovary
Physical activity (protective) Colorectum
Obesity Kidney, Breast (postmenopausal), Esophagus
(adenocarcinoma), other sites
Smoking Pancreas, Liver, Colorectal, other sites
Reflux symptoms Esophagus (adenocarcinoma)
Aflatoxin Liver
Aristolochic acid Upper urothelial
Inherited genetic variants (>160) Breast
Inherited genetic variants (>180) Prostate
Inherited genetic variants (>75) Colorectal
Inherited genetic variants (varies) Several other cancer sites

HPV, human papillomavirus; HBV, hepatitis B virus; HCV, hepatitis C virus.

will receive more publicity than several “nega- or more subsequent negative studies will be pub-
tive” (and perhaps better-​designed) studies of the licized as “overturning” this wisdom. The atten-
same association. tion our studies receive is a mixed blessing. It is
When the first positive study of an associa- certainly nice to be noticed, but if we all agree
tion receives wide press attention, the association that it is rare for a single epidemiologic study to
sometimes becomes accepted wisdom. Then one be definitive, then do most studies deserve any
public notice at all? The answer is no, but in an
TABLE E-​2   SOME SELECTED
age in which the public is interested in the causes
HYPOTHESIZED RISK FACTORS
of disease, especially cancer, and thus reads the
newspapers and watches the TV stations that
ACCEPTED AS PROBABLY NOT MAJOR
publicize the latest findings, would we really
CAUSES OF THE LISTED CANCERS,
prefer to be ignored and have our findings go
BASED ON EPIDEMIOLOGIC RESEARCH
completely unheeded? Probably not. Yet it does
SINCE THE BEGINNING OF THE 1980S
behoove us to be self-​critical about the ways we
Risk factor Cancer communicate our findings.
In the Epilogue to the first edition we asked
HSV II infection Cervix whether readers of a second edition of this text-
PCBs and DDT Breast book would find much new information in 5
Dietary fat Breast or 10  years time. Asking the same question in
the third edition, we continue to feel optimistic
Coffee Pancreas
about continued progress.
Reserpine (an antihypertensive) Breast There are several reasons for our optimism:
Saccharine Bladder
Cell phone use Brain • Our understanding of study design and
methods for limiting bias in epidemiologic
HSV II, herpes simplex virus II; PCBs, polychlorinated biphenyls; studies continues to improve. Overall study
DDT, dichlorodiphenyltrichloroethane.
693

Epilogue 693

quality has improved, and average sample (­chapter 9), and recent changes in breast cancer
sizes have increased. incidence (­chapter 16), should give us confidence
• Prospective studies are able to provide in our approach and methods. We can be proud
larger numbers of cases of common of recent achievements in our field. Any individ-
cancers than was previously possible and ual study result may be less than certain, but of
are beginning to provide data on less one thing we can be sure—​the field of cancer epi-
common cancer sites and cancer subtypes. demiology has much more to contribute.
• The recognition that single studies are Nevertheless, the large fractions of prostate,
often underpowered, particularly for esophagus, brain, lymphoma, leukemia, and
assessment of effect modification, has led other cancers that remain unexplained indicate
the continued formation of consortia to that we still have much work to do. There is rea-
permit analysis of pooled primary data son to be concerned that a combination of fac-
to increase power. This also helps provide tors including reduced funding, large secondary
the most definitive evidence available at dataset availability, and, perhaps, a lack of risk-​
an earlier stage, eliminating some of the taking among younger investigators, is leading
back-​and-​forth in the literature that can to a kind of stasis in our field (Adami & Nyren,
contribute to confusion. 2016). It is increasingly hard to start new and
• Many cancer sites have been understudied. innovative studies, and for many epidemiolo-
If lymphoma or brain cancer had received gists, great efforts are needed merely to continue
the attention that has been given to more and/​or fully analyze existing studies. The art of
common cancers such as breast and generating hypotheses from descriptive studies
colorectal cancer, for instance, we would seems less frequently practiced, despite intrigu-
probably know more about their causes. ing geographic differences and temporal trends
• Newer methods of exposure assessment for many cancers (Adami & Nyren, 2016). New
are becoming available. Better analytic studies are needed in addition to data-​mining of
techniques are reducing the amount of existing studies, and there are many parts of the
specimen required to measure nutrients world in which epidemiology studies are almost
and exogenous chemicals in blood unknown.
and other tissues, and are decreasing In addition, some once promising areas
the random laboratory error in these such as the role of diet seem to be producing
measurements. mainly null results. We know this is not entirely
• The availability of catalogs of common due to exposure misclassification, because the
between-​person genetic variation make same cohort studies in which results for can-
it possible to define the genetic variants cer are largely null have reported robust die-
responsible for much of the inherited tary associations with cardiovascular disease,
component of risk of specific cancers, at diabetes, and diverticular disease. A  plausible
least for those sites for which adequate hypothesis is that it is diet during childhood,
numbers of samples from cases have been adolescence, and young adulthood (or even in
assembled. utero) that is associated with cancers decades
• Newer methods of cancer definition are later, but exposures in early life are much
being validated. Entities that look similar harder to document than exposures in middle
under the microscope may have different age. Greater attention to early life exposures is
etiologic factors, and molecular techniques characteristic not only of diet but also of many
may help separate entities that have other exposures for which robust associations
previously been lumped together. cannot be seen in middle life. The precedent of
radiation exposure around the time of puberty
For all these reasons, we feel that epidemi- being much more strongly associated with
ology still has much to offer the field of cancer breast cancer risk than exposure later in life
research. Our recent track record in explaining is a long-​established indication of the need to
substantial parts of mysteries such as the decline examine early life exposures where possible.
in stomach cancer incidence (­chapter  10), the If and when there is a fourth edition of this
increase in adenocarcinoma of the esophagus textbook, we hope that some of these issues will
694

694 Epilogue

have been addressed and cancer epidemiology Reference

will still be in rude health. We have no doubt Adami H-​O, Nyren O. Enigmas, priorities and oppor-
however, that many future discoveries await in tunities in cancer epidemiology. Eur J Epidemiol
our field and we eagerly look forward to learning 2016;31:1161–​71.
about these in the next decade.
695

INDEX

Note: Page references followed by a t indicate tables; f indicate figures.

Abortions (induced or spontaneous) of testicles, 530

breast cancer and, 392, 404, 407t of vagina, 17
ovarian cancer and, 465 Adenocarcinoma in situ (AIS), in cervical cancer, 421
thyroid cancer and, 617 Adenomatous polyposis coli (APC) genetic
Acetaminophen, 18, 471, 558, 561, 639 mutation, 247
Achalasia, esophageal cancer and, 200, 201t Adult T-​cell leukemia/​lymphoma, 650
Acne, testicular cancer and, 531, 532 Affymetrix GeneChip array, genotyping method, 91
Acquired cystic kidney disease, 578–​579, 580t Aflatoxin exposure
Acquired immunodeficiency syndrome (AIDS). See also biomarkers of, 103, 104, 105
HIV/​AIDS; Human immunodeficiency virus HBV synergism with, 284
cancer association with, 15 hepatitis B virus interaction, 104
global deaths from, 46t liver cancer and, 280–​281, 282t, 283–​284
HIV and, 15 smoking-​related cancers and, 94
infant survival and, 14 African Americans
Kaposi’s sarcoma and, 15 breast cancer and, 383
nasopharyngeal cancer and, 170, 172t colorectal cancer and, 262
non-​Hodgkin lymphoma and, 653 laryngeal cancer and, 342
organ transplantation and, 15 lung cancer and, 328, 331t, 332, 334
skin cancer and, 366–​367 ovarian cancer and, 460
testicular cancer and, 534 pancreatic cancer and, 310
Acromegaly, colorectal cancer and, 245f, 261, 263 prostate cancer and, 482, 489, 492, 497, 504, 505t
Acute hepatic porphyrias, 280 stomach cancer and, 14
Acute lymphoblastic leukemia, 649, 676 women and breast cancer, 383
Acute lymphocytic leukemia (children), 53t Aggressive NK-​cell leukemia, 650, 657
Acute myeloid leukemia (AML), 8, 11, 108 AIDS. See Acquired immunodeficiency syndrome
Acute promyelocytic leukemia, 673 Air pollution
Adenocarcinoma. See also specific cancers bladder cancer and, 560, 561, 562t
of bladder, 543 brain cancer and, 596
of breasts, 381 breast cancer and, 406
of cervix, 421, 423, 424, 426, 430 DALYs and, 45–​46
of colon and rectum, 243 lung cancer and, 328, 341–​342
of endometrium, 441, 450 Alcohol consumption, 16
of esophagus, 183–​191, 193–​197, 200–​202, 693 basal cell carcinoma and, 364
of gallbladder, 277 bladder cancer and, 554
H. pylori and, 199 brain cancer and, 592–​593
of kidneys, 571 breast cancer and, 16, 51t, 125, 389–​390
of liver, 277 cardia cancer and, 222
of lungs, 327, 328, 331–​334, 338 colorectal cancer and, 16, 51t, 255–​256
of pancreas, 309, 317 endometrial cancer and, 449
of prostate gland, 481 esophageal cancer and, 16, 191–​194, 201t
of stomach, 213, 216–​218 global cancer rates and, 16
69

696 Index

Alcohol consumption (cont.) Anorexia, 213

Hodgkin lymphoma and, 637 Anthropometric measures
kidney cancer and, 577 basal cell carcinoma and, 365
laryngeal cancer and, 16 bladder cancer and, 554
leukemia and, 683 breast cancer and, 400–​401
liver cancer and, 13, 16, 51t, 282t, 285 cervical cancer and, 426
lung cancer and, 337 colorectal cancer and, 256–​257
melanoma and, 364 endometrial cancer and, 445
non-​Hodgkin lymphoma and, 656 Hodgkin lymphoma and, 637–​638, 638–​639
oral health and, 202 kidney cancer and, 577
oral/​pharyngeal cancer and, 16, 113, 140, 145, 146–​147 liver cancer and, 287
ovarian cancer and, 464 lung cancer and, 338
pancreatic cancer and, 314 melanoma and, 365
pharyngeal cancer and, 16 nasopharyngeal cancer and, 169
prostate cancer and, 495–​496 non-​Hodgkin lymphoma and, 657
skin cancer and, 364 oral/​pharyngeal cancer and, 147
testicular cancer and, 532 ovarian cancer and, 469
thyroid cancer and, 616–​617 pancreatic cancer and, 315
Alcohol-​metabolizing polymorphisms, 189 prostate cancer and, 93, 499–​500
Alleles skin cancer and, 365
HLA alleles and NPC risk, 164 testicular cancer and, 533
identification methods, 87–​88 thyroid cancer and, 618–​620, 622t
sharing methods, 87 Antioxidants
Allergies oral/​pharyngeal cancer and, 145
brain cancer and, 594 stomach cancer and, 51t, 223–​227, 233t
Hodgkin lymphoma and, 639 Apoptosis (programmed cell death)
non-​Hodgkin lymphoma and, 662–​663 breast cancer and, 384, 396, 398
pancreatic cancer and, 318 cervical cancer and, 428, 433
American Cancer Society (ACS), 4 colorectal cancer and, 71–​72, 255, 260, 264
American Indians EBV and, 106t
breast cancer and, 383 endometrial cancer and, 445
colorectal cancer and, 262 KSHV and, 106t
Anal cancer, 106t liver cancer and, 280–​281, 288, 290, 293
Analgesic medications melanoma and, 360
protective value of, 18 oral/​pharyngeal cancer and, 141
renal cancer and, 18 ovarian cancer and, 469
urinary tract tumors and, 18 prostate cancer and, 492, 494
Analysis methods role in carcinogenesis, 70–​72, 71f, 81
effect measures, 125 squamous cell carcinoma and, 365
interaction, 125–​127, 126t stomach cancer and, 221, 228
meta-​analysis, 127 Arsenic
Anaplastic astrocytoma, 589t bladder cancer and, 10
Anaplastic large cell lymphoma, ALK positive/ environmental exposure to, 10–​11
​ALK negative, 650 lung cancer and, 7
Anaplastic oligodendroglioma, 589t occupational exposure to, 7
Androgens scrotal cancer and, 6
breast cancer and, 393–​394, 399–​401 Asbestos
endometrial cancer and, 445 emphysema (asbestosis) and, 7
esophageal adenocarcinoma and, 198 environmental exposure to, 11
liver cancer and, 286 esophageal cancer and, 7
ovarian cancer and, 462, 469 kidney cancer and, 7
prostate cancer and, 496–​497, 499–​500 lung cancer and, 7, 11, 339
Anemia mesothelioma and, 7, 11
hypochromic, 200 occupational exposure to, 7
Angioimmunoblastic T-​cell lymphoma, 650 pulmonary fibrosis and, 7
Angiosarcoma of the liver (ASL), 282t stomach cancer and, 7
thorotrast and, 6 Ashkenazi Jews, 86, 90
vinyl chloride and, 9–​10 Asians/​Asian countries
697

Index 697

bladder cancer and, 551 Basal cell carcinoma (BCC)

breast cancer, 383, 389, 393 DNA repair and, 358
breast cancer and, 393 gene mapping of, 359
colorectal cancer and, 245, 256, 262 pathogenesis of, 358, 360
endometrial cancer and, 442, 443 prognosis, 355
esophageal cancer and, 183, 184, 189, 196, 197 progress, 355
Hodgkin lymphoma and, 630 risk factors
liver cancer and, 50, 286–​287 alcohol use, 364
lung cancer and, 330t, 341 anthropometric measures, 365
mortality rate comparison, 48, 51 artificial UV radiation, 369–​370
nasopharyngeal cancer and, 163, 164, 167, 172 diet, 363
non-​Hodgkin lymphoma and, 654 exogenous hormones, 364–​365
oral/​pharyngeal cancer and, 141 green tea, 363
ovarian cancer, 460 height, 365
pancreatic cancer and, 315 infections, 365
prostate cancer and, 51, 482, 484–​485, 492, 495 inherited susceptibility, 358
skin cancer and, 356 ionizing radiation, 366
stomach cancer and, 53, 227 medical conditions, treatment, 366–​367
testicular cancer, 526 menopausal hormone therapy, 364
ASL. See Angiosarcoma of the liver obesity, 365
Aspirin reproductive, 364
colorectal cancer and, 18 solar UV radiation, 367–​369
esophageal cancer and, 18 tobacco use, 361
ovarian cancer and, 471 trauma, 370
pancreatic cancer and, 318 smoothened gene in, 358
skin cancer (BCC/​SCC) and, 367 solar UV radiation and, 355
stomach cancer and, 232 somatic events in, 360
Astrocytoma, 61, 589t treatment, 355
Ataxia-​telangiectasia tumor-​suppressor genes and, 358
ATM gene and, 85t Basal cell nevus syndrome, 591
characteristics of, 66 BCC. See Basal cell carcinoma
congenital immunodeficiency and, 14 B-​cell neoplasms, 650. See also specific
Atomic bomb survivors B-​cell neoplasms
basal cell carcinoma and, 366 Benign prostatic hyperplasia, 481, 501
bladder cancer and, 555 Benzene
brain cancer and, 595, 599 acute myeloid leukemia and, 7–​8, 11
breast cancer and, 402 environmental exposure to, 11–​12
colorectal cancer and, 259 leukemia and, 7–​8, 11, 51t
Hodgkin lymphoma and, 638 lymphoid leukemia and, 8
kidney disease and, 578 occupational exposure to, 7–​8
leukemia and, 4–​5, 681 Beta-​carotene
liver cancer and, 292 leukoplakia and, 145
lung cancer and, 338 oral/​pharyngeal cancer and, 145
non-​Hodgkin lymphoma and, 638 Betel leaf chewing, oral/​pharyngeal cancer and,
ovarian cancer and, 470 142, 144
prostate cancer and, 503 Bias
stomach cancer and, 230 in biomarker-​based studies, 101
thyroid cancer and, 620 in epidemiological studies, 124–​125
Australia antigen, 13 Biliary tract cancer, 277
Autoimmune diseases descriptive epidemiology, 294
Hodgkin lymphoma and, 638–​639 incidence rates, 35t, 37t, 39t, 41t
non-​Hodgkin lymphoma and, 661 risk factors
Autoimmune hepatitis, 367 diet, 294–​295
gallbladder disease, 294
Baldness, testicular cancer and, 529t, 531–​532 hormones, 295
Barrett’s esophagus inherited susceptibility, 295
esophageal cancer and, 187–​189, 191, 194, 198–​199, 202 mechanisms, 295
GERD and, 202 obesity, 294–​295
698

698 Index

Biomarker-​based studies microbiologic exposure, 106t

confounding bias in, 101, 104, 104f obesity, 554, 558, 562t
Framingham Heart Study, 101 occupational exposure, 8, 10–​11, 51t, 555–​557
gene-​environment, 104 outdoor air pollution, 560
gene-​environment studies, 104 parity, 577
gene-​gene cancer pathogenesis, 104 physical activity, 554
information bias in, 101 reproductive factors, 554
interaction/​effect modification, 104 S. haematobium, 106t, 543, 544, 555, 561, 562t
random error in, 104–​105 summary of, 562t
selection bias in, 101–​102 tobacco, 4, 51t, 551–​552
Biomarkers, 101–​109 total fluid intake, 553–​554
applications, 102f voiding frequency, urine pH, 560–​561
classes/​examples of, 101, 102t subgroups, 543
defined, 101 symptoms, 543
epidemiological applications of, 105–​108 treatment, 543
group 1 agents, 107t–​108t tumor-​suppressor genes and, 550, 551, 563
identification of human carcinogens, 104–​107 Blood transfusion, non-​Hodgkin lymphoma and, 655t,
influences on, 102 658, 661
low-​risk disease alleles and, 88 Bloom’s syndrome
Mendelian randomization and, 81 BLM gene and, 85t
misclassifications of, 103 characteristics of, 66
variation in, 103–​104, 103t oral/​pharyngeal cancer and, 140
Birt-​Hogg Dube syndrome, 574f B-​lymphoblastic leukemia, 673
Birth order BMI (body mass index)
Hodgkin lymphoma and, 633 breast cancer and, 400–​401
leukemias and, 681 colorectal cancer and, 256
stomach cancer and, 233 endometrial cancer and, 445
testicular cancer and, 529t, 535 esophageal cancer and, 193, 194, 200–​201
Blackfoot disease (in Taiwan), 10–​11 GERD and, 200, 202
Bladder cancer, 543–​563 Hodgkin lymphoma and, 637–​638
chromosome 8q24 and, 92 kidney cancer and, 575–​576
diagnosis, 543 liver cancer and, 287
epidemiology lung cancer and, 338
descriptive, 544–​546 oral/​pharyngeal cancer and, 147
genetic and molecular, 546, 547t–​549t, 550–​551 pancreatic cancer and, 315
genome-​wide association studies, 547–​549t testicular cancer and, 533
global mortality rates, 51t thyroid cancer and, 619, 622t
global trends and patterns, 51t Body weight. See Obesity; Weight change
incidence of, 544, 545f, 546 Borrelia burgdorferi, 658–​659
incidence rates, 46t, 545f–​546f Brain cancer, 587–​600
molecular subtyping, 544 children and, 587, 588, 593, 595, 598, 599–​600
occupational exposure and, 8, 10–​11 cranial/​spinal nerve tumors, 589t
polygenic risk score, 551 diagnosis, 587
polymorphisms in, 543, 546–​547, 550–​551, 561 distribution, 589t
prognosis, 544 DNA repair and, 591
progress, 544 epidemiology
risk factors descriptive, 588–​591
alcohol, 554 genetic and molecular, 591
anthropometric measures, 554 genome-​wide association studies, 591
diet, 552–​554, 561 germ cell tumors, cysts, 589t
environmental exposure, 543, 544, 559–​561 hemopoietic neoplasms, 589t
green tea, 554 incidence rates, 590f–​591f
hormones, 554 lymphomas, 589t
infections, 555 meningeal tumors, 589t
inherited susceptibility, 546, 550, 562t neuroepithelial tissue tumors, 589t
insulin resistance, 556 person-​time studies, 115–​116, 116f
ionizing radiation, 555 polymorphisms in, 591
kidney cancer, 553 prognosis, 587–​588
medical conditions, 557–​558 progress, 588
medications, 558–​559 risk factors
69

Index 699

alcohol use, 592–​593 anthropometric measures, 400–​401

allergies, 594 benign breast disease, 405
cellular phones, 598–​599 BMI, 400–​401
for childhood brain tumors, 599–​600 body fat distribution, 401
diet, 592 breastfeeding, 391–​392, 406
drinking water contamination, 598 diabetes, 403
electric/​magnetic field exposure, 598 diet, 114, 125, 385–​389
head and noise trauma, 535, 598 environmental exposures, 406
head/​noise trauma, 535 environmental pollution, 405–​406
hormones, 593 estrogen metabolites, 393
infections, 593–​594 estrogen plus progestin use, 398–​399
inherited susceptibility, 591, 599 estrogen receptors, 381, 383, 389, 392–​393, 399
ionizing radiation, 594–​595 fertility drugs, 403
medical conditions, treatments, 597–​598 gene mutations, 87
obesity, 593, 599 genetic predisposition, 85–​86, 92, 112, 311, 320t
occupational exposure, 595–​597 hormonal chemoprevention, 399–​400
parity, 593 hormones, 18–​20, 118, 198, 392–​399
physical activity, 594 hysterectomy, 398
pregnancy, 593 infections, 401
reproductive factors, 593 infertility, 403
summary of, 600t inherited susceptibility, 382–​384
tea and coffee, 592 insulin/​insulin-​like growth factor 1, 395, 407t
tobacco, 592 insulin resistance, 396, 400, 401
sellar region tumors, 589t ionizing radiation, 5, 338, 402
subgroups, 587 lifestyle, 91
symptoms, 587 lung cancer treatment, 341
treatment, 587 mammographic density, 405
tumor-​suppressor genes and, 591 medical conditions, treatments, 403–​404
Breast cancer, 381–​407 menopausal hormone therapy, 381, 383, 401
Ashkenazi Jews and, 90 menopause, 392
breast-​feeding protective benefits, 15 night shift work, 403
consortia of studies for, 92 NSAIDs, 404
diagnostic methods, 381 obesity, 19–​20, 51t, 382, 403, 406
DNA repair and, 388, 391 occupational exposure, 120, 402–​403
early-​onset breast cancer, 85–​86 oral contraceptives, 17, 117, 396–​397
endometrial cancer and, 448 overweight, obesity, 19–​20, 51t
epidemiology ovulatory cycles, 391
descriptive, 382–​383 parity, 391, 403
genetic/​molecular, 383–​384 perinatal exposures, 404
extracellular matrix and, 73–​74 physical activity, 401–​402
genetic variants, 62, 92 physical inactivity, 51t
global trends and patterns, 51t postmenopausal hormones, 397–​398
GWAS studies, 92 pregnancy, 15, 385, 390, 391–​392, 396,
HER2 type, 62, 381 404, 406
high-​penetrance gene mutations in, progesterone, 62, 392–​395, 393, 394–​395, 407t
383–​384 progestins, 17–​18, 396
incidence rates, 5, 46t, 382–​383, 382f–​383f prolactin, 391, 395–​396, 406, 407t
low-​penetrance gene mutations in, 384 reproductive factors, 15–​16, 390–​392
luminal A/​luminal B types, 62, 381 statins, 404
moderate-​penetrance gene mutations in, 384 summary of, 407t
mortality rates, 49f, 51t tobacco use, 112, 384–​385
Nurses’ Health Study, 386 in utero/​perinatal exposures, 404
polygenic risk score, 93 subgroups, 381
polymorphisms in, 383–​384 survival rates, 53t
prognosis, 381–​382 symptoms, 381
progress, 382 treatment, 69, 381
risk factors triple negative/​basal-​like type, 62, 381
abortions, 392 tumor-​suppressor genes and, 384
alcohol use, 16, 51t, 125, 389–​390 Women’s Health Initiative study, 386
androgens, 393–​394, 399–​401 Breast Cancer Association Consortium, 92
70

700 Index

Breastfeeding linkage disequilibrium, tagging SNPs, 88–​89

breast cancer and, 391–​392, 406 low-​risk disease alleles, identification, 88
endometrial cancer and, 448, 451t Cancer Genome Atlas (TCGA) Project, 63, 222, 485
esophageal cancer and, 198 Cancer stem cells, 61–​62
ovarian cancer and, 466, 468t, 472t, 473 Candidate-​gene studies
thyroid cancer and, 617–​618 description, 91
Breast-​ovarian cancer syndromes, 84, 85, 461 of melanoma, 359
Bronchial cancer, 51t of nasopharyngeal cancer, 165
Burkitt lymphoma, 106t, 650, 651, 657 of testicular cancer, 527
Butter consumption Carcinogenesis, 57–​77
colorectal cancer and, 249 apoptosis and, 70–​72, 71f
kidney cancer and, 576 cancer genes and, 62
oral/​pharyngeal cancer and, 146 cancer genomics and, 62–​63
cell-​cell–​extracellular matrix interactions, 59, 65,
Calcium intake 68–​69, 72, 74
colorectal cancer and, 51t, 253–​255, 263, 265 cell cycle, checkpoints, and, 65–​66, 65f
liver cancer and, 284 cellular models, 59–​61, 60f
ovarian cancer, 463 c-​Myc transcription factor, 70
ovarian cancer and, 463 Darwinian view of, 59
pancreatic cancer, 313 epigenetic gene silencing and, 64–​65
prostate cancer and, 480 ErbB-​2 (HER2) receptor, 69
Campylobacter jejuni, 659 exogenous mutagenesis and, 63–​64
Cancer. See also specific cancer sites and names extrinsic factors, 59, 65, 72–​75
of specific cancers genetic carcinogenesis, 63
benign vs. malignant tumors, 58t genetic mechanisms, 58–​59
cellular origin of, 59–​61 genomic instability disorders and, 66
clinical behavior of, 61–​62 growth-​regulatory signaling pathways, 68–​70
definition, 57–​58 IARC classifications, 105
genetic epidemiology of, 81–​96 modified metabolism, 70
metastasis (defined), 59 multi-​step/​two-​hit theory of, 63, 64f
molecular approach to p53 pathways and, 66–​68
understanding, 62–​73 progression, 57–​58, 58–​59, 73–​77
mortality from, 31–​34, 44–​52, 46t, 47f, 48f, 49f proto-​oncogenes and, 62, 66, 69, 77
origins of. See carcinogenesis Ras proto-​oncogene, 69–​70
progression of, 57–​58, 73–​77 retinoblastoma (RB) pathways and, 68, 68f
Cancer burden, 31–​54 single-​cell model, 59
determinants, limitations, 32t–​33t, 50t tumor suppressor genes and, 62–​67, 69, 72, 77
globalization and, 31 Carcinoids, 213
age-​standardized rates, 47f, 48f Carcinoma-​in-​situ (CIS)
GLOBOCAN data, 46 defined, 58–​59
leading causes of death, 46t HPV and, 13
mortality determinants, 50t squamous cell carcinoma and, 421
survival rate differences, 52–​53, 53t Carcinosarcomas, 441
trends and patterns, 46–​51 Carney’s complex, 613
measures, 31–​46 Carotenoids
age standardization, 34, 35t–​38t bladder cancer and, 553
cumulative incidence, 34, 39t–​42t breast cancer and, 387–​388
incidence, 31, 33–​34 lung cancer and, 336
mortality, 31–​34, 44–​52, 46t, 47f, 48f, 49f oral/​pharyngeal cancer and, 145
prevalence, 34, 43, 45, 47–​48, 50, 54 Caucasians
survival, 43–​44, 53t breast cancer and, 393
temporal trends, 44–​46 cancer and, 331
Cancer genes. See also DNA repair; Epidemiologic EBV-​positive Hodgkin lymphoma and, 632
studies; Gene identification strategies; nasopharyngeal cancer and, 165
Proto-​oncogenes; Tumor suppressor genes oral/​pharyngeal cancer and, 141
disruption mechanisms, 81–​83 prostate cancer and, 482, 497, 504
epidemiologic studies, 89–​91 Causal inference
genotyping methods in, 91–​92 general principles, 127–​129
gene identification strategies, 83–​88 Hill Criteria, 128t
701

Index 701

IARC classification, 129, 129t sexual behavior, 51t, 426

process STIs, 421, 423, 426, 429
disease in a specific person (level III), 130 systemic inflammatory diseases, 433
general case (level II), 129–​130 systemic lupus erythematosus, 433
individual study (level I), 129 tobacco use, 51t, 424–​425
Causal pie model, 112t somatic mutations, 424
Causation (of disease) subgroups, 421
associations, 115, 122 survival rates, 53t
contributing factors, 112 symptoms, 421
epidemiology and, 111 treatment, 422
ethical considerations, 123 tumor-​suppressor genes and, 428, 433
genetic factors, 83 vaccine for, 14
observational epidemiology and, 114 Cervical dysplasia, 433
Celiac disease Cervical intraepithelial neoplasia (CIN), 13, 421
esophageal disease and, 200 Chance, in epidemiological studies
non-​Hodgkin lymphoma and, 661–​662 confidence intervals, 20, 122, 199, 616t, 617f, 619t,
Cell-​cell–​extracellular matrix interactions, 59, 65, 621t, 635t, 677t–​678t
68–​69, 72, 74 P-​value, 121–​122, 330
B-​cell prolymphocytic leukemia, 650 Chemical dyes
Cellular phones, brain cancer and, 598–​599 occupational exposure to, 8
Cervical cancer, 421–​434. See also Human papilloma scrotal cancer and, 6
virus (HPV) infection urinary tract cancer and, 8
diagnosis, 422 Childbearing. See Parity
DNA repair and, 424, 428f, 461 Children
dysplastic, noninvasive precancerous lesions, 421 basal cell carcinoma and, 368
epidemiology brain tumors in, 587, 588, 593, 595, 598, 599–​600
descriptive, 422–​423 breast cancer and, 401, 407t
genetic and molecular, 423–​424 colorectal cancer in, 265
global trends and patterns, 51t Epstein Barr virus in, 169
HPV and, 104 H. pylori infection in, 233
IARC sponsored study, 13 HBV infection in, 287–​288, 344
incidence rates, 46t, 422f–​423f kidney cancer in, 571
mortality rates, 50–​51, 51t, 423 leukemias in, 53t
NPC and, 159 pancreatic cancer and, 315, 317
prognosis, progress, 422 prostate cancer and, 499, 505t
proto-​oncogenes and, 424 retinoblastoma in, 63, 68
risk factors testicular cancer and, 533, 535
anthropometric measures, 426 vaccination programs for, 289, 432
C. trachomatis infection, 427 years of life lost (YLL) measure, 45
diet, 425 China. See Asians/​Asian countries
diethylstilbestrol, 426, 434t Chlamydia pneumoniae infection, 338
hormones, 425–​426 Chlamydia psittaci infection, 659
HPV infection, 13, 104, 421, 423, 424–​425, 427–​433 Chlamydia trachomatis infection, 427
hysterectomy, 422 Chlorinated hydrocarbon compounds, pancreatic
infections, 426–​427 cancer, 316, 319t
inflammatory bowel disease, 433 Chlorinated water, pancreatic cancer and, 318
inherited susceptibility, 423–​424 Chlorination byproducts, bladder cancer and, 560
ionizing radiation, 5, 230, 433 Cholangiocarcinoma, 106t, 277, 282t, 292
medical conditions, 433 Cholecystectomy
microbiologic exposure, 12, 106t colorectal cancer and, 259–​260
obesity, 426, 434t pancreatic cancer and, 317
occupational exposure, 113, 433 Choroid plexus tumors, 589t
oral contraceptives, 17, 424, 426 Chromium
parity, 424, 425, 431f, 434t environmental exposure to, 12
physical activity, 433 lung cancer and, 8
pregnancy, 425, 432 occupational exposure to, 8
progesterone, 426 Chromophobe renal cell carcinoma, 571
reproductive factors, 425 Chronic lymphocytic leukemia (CLL), 366, 649, 650, 663
rheumatoid arthritis, 433 Chronic myelogenous leukemia, 60
702

702 Index

Chronic myeloid leukemia (CML), 60, 673 circulating insulin-​like growth factor, 263

Chronic obstructive pulmonary disease (COPD), 46t diet, 19, 51t, 114, 226
Cigar/​cigarette smoking. See Tobacco use diet/​dietary patterns, 251–​255
Circadian rhythms, 103 gender, 262
Clear cell carcinoma, 441, 571 genetics, 84, 120, 249
Clonorchis sinensis, 106t, 292 height, 262–​263
Clonorchis sinensis (liver fluke), 292 hypertension, 251
Cockayne syndrome, 358 infections, 257
Coffee intake inflammatory bowel disease, 248
bladder cancer and, 554 inherited susceptibility, 84, 120, 249, 250
brain cancer and, 592 ionizing radiation, 259
liver cancer and, 282t, 285 Lynch Syndrome, 84, 85t, 216, 219, 249
lung cancer and, 336 medical conditions, 259–​261
oral/​pharyngeal cancer and, 145 metabolic syndrome, 244t, 256–​257, 260
ovarian cancer and, 463–​464 NSAIDs, 18
pancreatic cancer and, 313–​314 obesity, 19, 51t
prostate cancer and, 494–​495 occupational exposure, 259
stomach cancer and, 223 parity, 256
Cohort studies, 116–​118 physical activity, 257–​259
closed/​open cohorts, 118 physical inactivity, 51t
defined, 117f probable, 244t
as exposure-​based, 117–​118 race and ethnicity, 262
as patently/​conceptually longitudinal, 118 sedentary lifestyle, 258–​259
Collecting duct carcinoma, 6 suboptimal calcium intake, 51t
Colorectal cancer (CRC), 243–​265 tobacco use, 250–​251
adenoma-​cancer sequence, 247–​248 weak, 244t
apoptosis and, 71–​72 well-​confirmed, 244t
chemoprevention of, 263–​264 screening methods, 264–​265
chromosomal instability in, 246–​247 serrated pathway, 248
chromosome 8q24 and, 92 somatic mutations of, 246–​247
CpG island methylator phenotype, 247 subgroups, 243
diagnosis, 243 survival rates, 53t, 245
DNA repair and, 247 susceptibility to, 249–​250
epidemiology symptoms, 243
descriptive, 245–​246 treatment, 75, 243
genetic, molecular, 246–​250 tumor-​suppressor genes and, 247
family history, in first-​degree relatives, 249 U.S. survival rates, 52f
FAP and, 84, 85t, 248–​249 Confidence intervals, 20, 122, 199, 616t, 617f, 619t, 621t,
genetic variants, 92 635t, 635t, 677t–​678t
global trends and patterns, 51t Confounding bias, 101, 102, 104, 104f
hereditary nonpolyposis CRC, 84, 85t, 216, 219, 249, 461 Congenital immunodeficiency
incidence rates, 35t, 37t, 46t, 245f, 246f, 261f–​262f ataxia telangiectasia and, 14
Mendelian randomization and, 93 in infants, 14
methylation and, 221 Wiskott-​Aldrich syndrome and, 14
microsatellite instability in, 247 Contralateral testicular cancer, 529t
mortality rates, 7, 46, 46t, 49f, 50, 51t, 52f, 243 Corrected survival, 43
natural history of, 247–​248 Cowden syndrome, 85t, 613
pathogenesis of, 250, 257, 260–​261, 265 Coxiella burnetii, 659
PLCO screening trial, 146 Craniopharyngioma, 589t
polymorphisms in, 250 Cryptorchidism, 529t, 531, 533, 535
prognosis, 243–​244 Cumulative incidence
progression, 245 as measure of disease burden, 34, 39t–​42t
risk factors Cyclophosphamide, bladder cancer and, 558, 561, 562t
acromegaly, 245f, 261, 263 Cystic fibrosis
age, 261 pancreatic cancer and, 318
alcohol use, 16, 51t, 255–​256 Cystic kidney disease, 578–​579, 580t
anthropometric measures, 256–​257
aspirin, 18 Dairy products
BMI, 256 ovarian cancer and, 463, 472t
703

Index 703

pancreatic cancer and, 313 ovarian cancer and, 462–​464, 472t

prostate cancer and, 490–​491 pancreatic cancer and, 313, 320t
testicular cancer and, 532 prostate cancer and, 115, 449, 487–​495
DALYs. See Disability-​adjusted life years (DALYs) skin cancer and, 363–​364
measure testicular cancer and, 532
Descriptive epidemiology thyroid cancer and, 615–​616
of biliary tract cancer, 294 tracheal cancer and, 51t
of bladder cancer, 544–​546 Diethylstilbestrol (DES), 17
of brain cancer, 588–​591 cervical cancer and, 426, 434t
of breast cancer, 382–​383 testicular cancer and, 528, 529t, 530
of cervical cancer, 422–​423 uterine cancer and, 404
of colorectal cancer, 245–​246 vaginal cancer and, 17
of endometrial cancer, 442–​443 Diffuse astrocytoma, 589t
of esophageal cancer, 184–​188 Diffuse large B-​cell lymphoma, 650
of Hodgkin lymphoma, 628–​630 Disability-​adjusted life years (DALYs) measure, 45–​46
of kidney cancer, 571–​572 Disease burden (of cancer), 31–​54
of liver cancer, 278–​280 determinants, limitations, 32t–​33t, 50t
of lung cancer, 328 globalization and, 31
of nasopharyngeal cancer, 160–​163 age-​standardized rates, 47f, 48f
of non-​Hodgkin lymphoma, 651–​653 GLOBOCAN data, 46
of oral cancer, 138–​139 leading causes of death, 46t
of ovarian cancer, 459–​461 mortality determinants, 50t
of pancreatic cancer, 310 survival rate differences, 52–​53, 53t
of skin cancer, 353–​358 trends and patterns, 46–​51
of stomach cancer, 14, 214–​216, 214f–​215f measures, 31–​46
of testicular cancer, 525–​527 age standardization, 34, 35t–​38t
of thyroid cancer, 609–​612 cumulative incidence, 34, 39t–​42t
Diabetes mellitus incidence, 31, 33–​34
breast cancer and, 403 mortality, 31–​34, 44–​52, 46t, 47f, 48f, 49f
colorectal cancer and, 260–​261 prevalence, 34, 43, 45, 47–​48, 50, 54
endometrial cancer and, 449 survival, 43–​44, 53t
kidney cancer and, 579 temporal trends, 44–​46
liver cancer and, 282t, 287 DNA repair, 81
ovarian cancer and, 471 biomarkers and, 102t
pancreatic cancer and, 316, 319t bladder cancer and, 560
squamous cell carcinoma and, 367 brain cancer and, 591
thyroid cancer and, 622 breast cancer and, 388, 391
2013, death rate, 46t cervical cancer and, 424, 428f, 461
in Werner’s syndrome, 66 colorectal cancer and, 247
Diarrheal diseases, 46t defined, 95
Diet/​dietary patterns, 18–​20 kidney cancer and, 574, 576
biliary tract cancer and, 295 leukemia and, 680
bladder cancer and, 552–​554 liver cancer and, 284, 285, 288
brain cancer and, 592 mechanisms of action, 62–​66, 64f
breast cancer and, 19, 385–​389 melanoma and, 359, 365
bronchial cancer and, 51t nasopharyngeal cancer and, 165
cervical cancer and, 425 pancreatic cancer and, 312
colorectal cancer and, 19, 51t, 114, 226, 251–​255 prostate cancer and, 485–​486, 495, 505t
endometrial cancer and, 448–​449 stomach cancer and, 221, 228
esophageal cancer and, 194–​198 Drinking water contamination
Hodgkin lymphoma and, 637 brain cancer and, 598
kidney cancer and, 575–​576 lung cancer and, 342
laryngeal cancer and, 343–​344 Ductal adenocarcinomas, 309
leukemia and, 683 Ductal carcinoma in situ, 381
liver cancer and, 283–​285 Duodenal ulceration, H. pylori and, 14
lung cancer and, 335–​337 Dysplastic, noninvasive precancerous lesions, 421
nasopharyngeal cancer and, 167–​169
non-​Hodgkin lymphoma and, 656 Early-​onset breast cancer, 85–​86
oral/​pharyngeal cancer and, 144–​146 EBV. See Epstein Barr virus
704

704 Index

Egg consumption, ovarian cancer and, 464 risk prediction models, 450–​452

Electrical workers, brain cancer and, 596–​597 somatic events/​mutations, 444–​445
Electro-​magnetic field exposure subgroups, 441
brain cancer and, 598 symptoms, 441
leukemia and, 682 treatment, 441–​442
Embryonal tumors, 589t tumor-​suppressor genes and, 444–​445
ENCODE (Encyclopedia of DNA Elements), 93 Endometrial hyperplasia, 398
Endocrine neoplasia (types 2A and 2B), 84 Endometrioid cancers, 459, 462
Endogenous hormones Endometriosis
breast cancer and, 392 endometrial cancer and, 450
endometrial cancer and, 445–​446 ovarian cancer and, 471
testicular cancer and, 528 End-​stage renal disease, 578–​579
Endometrial adenocarcinomas, 441 Enteropathy-​associated T-​cell lymphoma, 650
Endometrial cancer, 441–​452 Environmental exposures. See also Microbiologic envi-
diagnosis, 441 ronment, risk factors; Occupational exposure;
epidemiology Physical environment, risk factors
descriptive, 442–​443 bladder cancer and, 543, 544, 546, 559–​561
genetic and molecular, 443–​445 breast cancer and, 405–​406
exogenous estrogens, progestins and, 18 familial aggregation and, 83
incidence rates, 442f–​443f genotype interaction with, 89, 90–​91
mortality data, 44 kidney cancer and, 546
polymorphisms in, 443–​444 leukemias and, 682
prognosis, 442 liver cancer and, 281
progress, 442 lung cancer and, 345
risk factors nasopharyngeal cancer and, 162, 165, 166, 168
androgens, 445 non-​Hodgkin lymphoma and, 656
anthropometric measures, 445 occupational disease and, 11–​12
breastfeeding, 448, 451t Environmental Protection Agency (EPA), 11
childbirth, 15–​16 Ependymal tumors, 589t
diabetes, 449 Epidemiologic studies. See also Descriptive epidemiology;
diet, 448–​449 Genetic and molecular epidemiology; Genetic
endogenous estrogens, 445–​446 epidemiology; specific disorders
endometriosis, 450 analysis methods
environmental xenoestrogens, 450 effect measures, 125
estrogen plus progestin use, 398 interaction, 125–​127, 126t
estrogen receptors, 444, 447 meta-​analysis, 127
exogenous hormones, 18, 446–​447 cancer risks determined by
hormonal chemoprevention, 399 alcohol consumption, 16
hypertension, 449–​450, 451t cigarette smoking, 3–​4
hysterectomy, 442 diet, 18–​20
infections, 450 iatrogenic exposure, 17–​18
inherited susceptibility, 85t, 443–​444 immunodeficiency, 14
insulin-​like growth factor 1, 445 ionizing radiation, 5–​7
insulin resistance, 449 microbiologic environment, 12–​15
IUDs, 426 occupational exposure, 6–​10
metabolic syndrome, 450 physical environment, 10–​12
nulliparity, 447, 448, 451t reproductive factors, 15–​16
obesity, 441, 443, 445, 449, 452 chance, role of
oral contraceptives, 17, 447 confidence intervals, 122
PCOS, 446, 450, 451t P-​value, 121–​122
physical activity, 449 cohort studies, 116–​118
polycystic ovary syndrome, 450 genetic, 89–​94
post-​menopause, 441, 447 case-​control studies, 89–​90
pregnancy, 448, 452 case-​only design, 90
progesterone, 441, 444, 445, 446–​448 family-​based association studies, 90
progestins, 446–​447 gene-​environment interaction, polygenic models of
reproductive factors, 16, 447–​448 disease risk, 90–​91
socioeconomic status, 450 genotyping methods used in, 91–​92
tamoxifen, 450 HPV/​cervical cancer, 13–​14
705

Index 705

Korea, hepatitis vaccine, 13 pesticide exposure, 200

study design pickled vegetables, 197, 201t
analytic studies, 115 Plummer-​Vinson syndrome, 200
case control studies, 118–​120 probable relationships, 201t
cohort studies, 116–​118 reproductive factors, 198
descriptive studies, 114 tobacco use, 4, 51t, 191–​194, 192f, 201t
ecologic studies, 114–​115 weak relationships, 201t
person-​time, study base, 115–​116 well-​confirmed, 201t
systematic errors symptoms, 183
bias, 124–​125 treatment, 183
confounding, 123–​124 tumor-​suppressor genes and, 189, 191
experimental study, 122–​123 Esophageal squamous cell carcinoma. See Squamous cell
Epidermal growth factor, 69, 142, 188, 280, 424 carcinoma, of esophagus
Epidermodysplasia verruciformis, 359 Esophageal squamous dysplasia, 187
Epithelial tumors, malignant, 459 Estrogen metabolites
Epstein Barr virus associated gastric cancers (EBVaGC), 229 breast cancer and, 393
Epstein Barr virus (EBV) prostate cancer and, 497
cancer association, 62, 159–​160 Estrogen-​progesterone contraceptives, 426
carcinogen classification, 106t Estrogens/​estrogen receptors (ERs)
Hodgkin lymphoma and, 15, 106t, 632–​637, 636f breast cancer and, 381, 383, 389, 392–​393, 399
immunodeficiency and, 15 cervical cancer and, 425–​426
IMRT treatment, 159–​160 endometrial cancer and, 444, 447
nasopharyngeal cancer and, 106t, 169–​170 genome-​wide association studies, 92
non-​Hodgkin lymphoma and, 657–​658 lung cancer and, 398
stomach cancer and, 229 pancreatic cancer and, 314
ERSPC. See European Randomized Study of Screening prostate cancer and, 497
for Prostate Cancer thyroid cancer and, 617
Esophageal adenocarcinoma, 183, 184, 187–​188, 189, Etiology
190–​191, 193–​194, 198 causality and, 111–​113, 112f, 112t
Esophageal cancer, 183–​202 interventional epidemiology and, 113–​114
diagnosis, 183 observational epidemiology and, 114
epidemiology European Molecular Biology Laboratory, 92
descriptive, 184–​188 European Randomized Study of Screening for Prostate
genetic and molecular, 188–​191 Cancer (ERSPC), 484
esophageal adenocarcinoma subgroup, 183 Extracellular matrix (ECM)
global trends and patterns, 51t breast cancer and, 73–​74
incidence rates, 35t, 37t, 39t, 41t, 47f, 184f–​185f cancer origins and, 59, 65
prognosis, 53, 183 Extranodal marginal zone B-​cell lymphoma of
progression, 183–​184 mucosa-​associated lymphoid tissues (MALT)
risk factors lymphoma, 650, 658
achalasia, 200 Extranodal NK/​T-​cell lymphoma, nasal type, 650
alcohol consumption, 16, 51t, 191–​194, 201t Extravasation, in metastasis, 59
androgens, 198 Eye (conjunctival) cancer, 106t
aspirin, 18
breastfeeding, 198 Familial adenomatous polyposis (FAP), 84, 85t
celiac disease, 200 brain tumors and, 591
diet, 51t colorectal cancer and, 244t, 248
GERD, 200–​201 thyroid cancer and, 613
green tea, 197 Familial aggregation. See also Inherited susceptibility
H. pylori infection, 199 brain cancer and, 595–​597
hot drinks, 197–​198 colorectal cancer and, 250
HPV, 199 nasopharyngeal cancer and, 163.164
inconsistent findings, 201t non-​Hodgkin lymphoma and, 653
maté tea, 49, 149, 150 ovarian cancer and, 461
obesity, 189, 202 stomach cancer and, 218–​219
occupational exposure, 7, 199–​200 Familial breast cancer, 85–​86, 92, 112, 311, 320t
oral health, 202 Familial diffuse gastric cancer (FDGC), 218, 219
parity, 198 Familial medullary thyroid cancer, 85t
Paterson-​Kelly syndrome, 200 Familial papillary renal carcinoma syndrome, 85t
706

706 Index

Familial susceptibility. See Inherited susceptibility Gastric corpus atrophy, 216

Fanconi anemia, 66, 140 Gastric lymphomas, 213
FAP. See Familial adenomatous polyposis Gastroesophageal reflux disease (GERD), 143t, 149
Fat intake. See also Butter consumption; Olive oil Barrett’s esophagus and, 202
lung cancer and, 336 esophageal cancer and, 200–​201, 201t
oral/​pharyngeal cancer and, 146 oropharyngeal cancer and, 143t, 149
ovarian cancer and, 464 Gene disruption mechanisms
prostate cancer and, 488 genetic variation, 81–​82
stomach cancer and, 226 mutations, polymorphisms, 82–​83
Ferguson-​Smith syndrome, 359 Gene-​environment studies, 104
Fertility drugs Gene-​gene cancer pathogenesis studies, 104
breast cancer and, 403 Gene identification strategies
ovarian cancer and, 465, 472t alleles
Fiber intake (dietary fiber) identification methods, 87–​88
breast cancer and, 386–​387 sharing methods, 87
colorectal cancer and, 254–​255 alternative methods, 86–​87
oral/​pharyngeal cancer and, 146 high penetrance genes for cancer syndromes, 84–​86
stomach cancer and, 225 identifying rare alleles, 87–​88
testicular cancer and, 532 limitations of segregation and link analysis, 86
Fish consumption linkage analysis
liver cancer and, 284 early-​onset breast cancer study, 85
prostate cancer and, 488–​490, 505t for identifying chromosomal location, 84
Folate (vitamin B9) limitations of, 86
bladder cancer and, 553 positional cloning, 84, 85
breast cancer and, 388 segregation analysis, 83–​84, 85
colorectal cancer and, 252–​253 limitations of, 86
esophageal cancer and, 195 Gene mapping
Follicular cancer, 621 in basal cell carcinoma, 359
Follicular lymphoma, 650 in squamous cell carcinoma, 359
Formaldehyde Gene penetrance, for cancer syndromes, 84–​86
leukemia and, 8, 51t Genes (human gene). See also specific genes
nasopharyngeal cancer and, 8 components of, 81–​82
occupational exposure to, 8–​9 familial aggregation of, 83
Framingham Heart Study, 101 high penetrance, for cancer syndromes, 84
Fruit consumption inherited cancer-​predisposition genes, 85t
bladder cancer and, 552–​553, 561 mutations and polymorphisms in, 82–​83, 82f
breast cancer and, 389 variations in, 81–​82
bronchial cancer and, 51t Gene silencing, epigenetic mechanisms of, 64–​65
colorectal cancer and, 19 Genetic and molecular epidemiology. See also
esophageal cancer and, 51t, 196 Inherited susceptibility; Somatic events/​
kidney cancer and, 575 mutations
laryngeal cancer and, 343–​344 of bladder cancer, 546, 547t–​549t, 550–​551
lung cancer and, 335–​336 of brain cancer, 591
nasopharyngeal cancer and, 168 of breast cancer, 383–​384
ovarian cancer and, 464 of cervical cancer, 423–​424
stomach cancer and, 222, 225 of colorectal cancer, 246–​250
tracheal cancer and, 51t of endometrial cancer, 443–​445
Frying/​grilling, stomach cancer and, 226–​227 of esophageal cancer, 188–​191
of Hodgkin lymphoma, 630–​633
Gallbladder disease of kidney cancer, 572–​575
adenocarcinomas and, 277 of leukemias, 680–​681
biliary tract cancer and, 294 of liver cancer, 280–​281
incidence rates, 46t of nasopharyngeal cancer, 163–​166
liver cancer and, 282t of non-​Hodgkin lymphoma, 653–​654
obesity and, 19 of oral cancer, 139–​142
pancreatic cancer and, 317 of ovarian cancer, 461–​462
survival rates, 277 of pancreatic cancer, 311–​312
thyroid cancer and, 621 of skin cancer, 358–​361
Garlic, colorectal cancer and, 255 of stomach cancer, 219–​222
70

Index 707

of testicular cancer, 527–​528 in melanoma, 359

of thyroid cancer, 612–​614 in pleural mesothelioma, 345
Genetic carcinogenesis, 63 in stomach cancer, 218–​219
Genetic epidemiologic studies. See Epidemiologic Glioblastoma, 589t
studies Glioma malignant, NOS, 589t
Genetic epidemiology, 81–​97 Glutathione S-​transferase (GST)
biological function assignment, 93–​94 bladder cancer and, 550
gene disruption mechanisms breast cancer and, 388
genetic variation, 81–​82 kidney cancer and, 574
mutations, polymorphisms, 82–​83 liver cancer and, 280
gene identification strategies lung cancer and, 337
allele-​sharing methods, 87 nasopharyngeal cancer and, 165
alternative methods, 86–​87 oral/​oropharyngeal cancer and, 140
high penetrance genes for cancer syndromes, 84–​86 ovarian cancer and, 451
identifying rare alleles, 87–​88 skin cancer and, 358
link analysis for identifying chromosomal stomach cancer and, 223
location, 84 Goiter, 615, 621, 622
positional cloning supplanted by regional Gorlin syndrome, 358
sequencing, 84 Green tea
segregation analysis, 83–​84 basal cell carcinoma and, 363
segregation/​linkage analysis limitations, 86 bladder cancer and, 554
genome typing methods, 91–​92 esophageal cancer and, 197
linkage disequilibrium, tagging SNPs, 88–​89 liver cancer and, 282t, 284
Mendelian randomization, 93 lung cancer and, 336
polygenic risk scores, 93 nasopharyngeal cancer and, 169
somatic mutations, 94 ovarian cancer and, 463–​464
types of studies, 89–​91 stomach cancer and, 223
whole-​genome scans, collaboration, consortia, 92 Growth-​regulatory signaling pathways, 68–​70
Genetic heterogeneity, 86 c-​Myc transcription factor, 70
Genetic polymorphisms. See Polymorphisms (genetic ErbB-​2 (HER2) receptor, 69
polymorphisms) Ras proto-​oncogene, 69–​70
Genetic susceptibility. See Inherited susceptibility retinoblastoma (RB) pathway, 68f
Genetic variation, 81–​82 GSTM1 gene
Genital cancers, 13 bladder cancer and, 547t, 550, 561–​562
Genome-​wide association studies (GWAS) kidney cancer and, 574
biological pathways, 92 liver cancer and, 280
of bladder cancer, 547–​549t lung cancer and, 337f
of brain cancer, 591 nasopharyngeal cancer and, 165
of breast cancer, 92 oral/​oropharyngeal cancer and, 140–​141
cancer genetic variants, 83, 88, 91–​92, 93 skin cancer and, 358–​359
of liver cancer, 280 GSTP1 gene
of lung cancer, 104, 330–​332 kidney cancer and, 574
of melanoma, 360 skin cancer and, 358–​359
of nasopharyngeal cancer, 165–​166 stomach cancer and, 219
of prostate cancer, 92, 444, 486 GSTT1 gene
of skin cancer, 358 kidney cancer and, 574
of squamous cell carcinoma, 359 liver cancer and, 280
of stomach cancer, 219, 220t lung cancer and, 337f
of testicular cancer, 527 nasopharyngeal cancer and, 165
Genomic instability disorders, 66 oral/​oropharyngeal cancer and, 140–​141
Genotype-​Tissue Expression (GTEx) project, 94 skin cancer and, 358–​359
Genotyping methods, used in epidemiologic
studies, 91–​92 Hair dyes
Affymetrix GeneChip array, 91 bladder cancer and, 557, 562t
Illumina BeadArray, 91 non-​Hodgkin lymphoma and, 663
Sequenom MassARRAY, 91 Hairy cell leukemia, 650–651, 654, 657
TaqMan, 91 HapMap Project, 88, 691
Germ cell tumors, cysts, 589t HBV. See Hepatitis B virus (HBV) infection
Germline mutations HCC (hepatocellular carcinoma). See Liver cancer
708

708 Index

Head trauma, brain cancer and, 535, 598 in lung cancer, 330

Height in melanoma, 84, 359
breast cancer and, 400 in oral, oropharyngeal cancer, 140
kidney cancer and, 577 in oral and pharyngeal cancer, 140
lung cancer and, 338 in ovarian cancer, 461
pancreatic cancer and, 315, 319t population-​based kin-​cohort study, 86
prostate cancer and, 93, 500 in prostate cancer, 84–​85
testicular cancer and, 533 Hill Criteria for inferring causation, 128t
thyroid cancer and, 618, 622t Hispanic Americans
Helicobacter pylori, 14, 106t breast cancer and, 383
esophageal cancer and, 199 ovarian cancer and, 460
non-​Hodgkin lymphoma and, 658 testicular cancer and, 527
pancreatic cancer and, 315, 319t HIV/​AIDS. See also Acquired immunodeficiency syn-
stomach cancer and, 14, 219, 225, 227–​228, 233t drome; Human immunodeficiency virus
Hemangioma, 589t Hodgkin lymphoma and, 534
Hematochromatosis, 280 non-​Hodgkin lymphoma and, 653, 660–​661
Hemochromatosis occupational exposure and, 172t
liver cancer oral/​pharyngeal cancer and, 148
risk factors skin cancer and, 366–​367
hemochromatosis, 282t 2013, death rate, 46t
Hemopoietic cancers HNPCC. See Hereditary nonpolyposis colorectal cancer
benzene and, 11–​12 Hodgkin lymphoma, 627–​641
Hemorrhagic stroke classifications, 627, 650
2013, death rate, 46t diagnosis, 627
Hepatic cancer. See Liver cancer epidemiology
Hepatitis B virus (HBV) infection, 13, 106t descriptive, 628–​630
aflatoxin synergism with, 104, 284 genetic and molecular, 630–​633
liver cancer and, 277, 282t, 287–​290 incidence rates, 36t, 38t, 40t, 42t, 46t, 628f–​629f, 630,
nasopharyngeal cancer and, 170 630f–​631f, 633
pancreatic cancer and, 315 mortality rate, 628
Hepatitis C virus (HCV), 106t pathogenesis of, 637
liver cancer and, 277, 290–​292 prognosis, 627
non-​Hodgkin lymphoma and, 658 risk factors
Hepatobiliary cancer, 277–​278 alcohol use, 637
Hepatocellular carcinoma (HCC). See Liver cancer allergies, 639
Hepatosplenic T-​cell lymphoma, 650 anthropometric measures, 637–​638
Hereditary breast-​ovarian syndrome, 85t autoimmune diseases, 638–​639
Hereditary diffuse gastric cancer (HDGC), 218 birth order, 633
Hereditary leiomyomata and renal cell carcinoma, 574t BMI, 637–​638
Hereditary melanoma syndrome, p16/​CDKN2A gene diet, 637
and, 85t EBV, 15, 106t, 632–​633, 633–​636, 636f
Hereditary nonpolyposis colorectal cancer (HNPCC), aka HCV, 106t
Lynch Syndrome, 84, 85t, 216, 219, 249, 461, 591 herpes zoster, 637
Hereditary papillary renal cell carcinoma, 574t HIV-​1, 106t
Hereditary retinoblastoma, 85t hormones, 637
Heritability. See Inherited susceptibility immune deficiencies, 638
Herpes simplex viruses 1 (HSV-​1) infections, 106t, 637
brain cancer and, 594 inherited susceptibility, 630–​632
cervical cancer and, 426 ionizing radiation, 5
oral/​pharyngeal cancer and, 148–​149 medications, 639
prostate cancer and, 501 microbiologic, 106t
HER2 type breast cancer, 62, 381 obesity, 637, 640t
High-​penetrance polymorphisms occupation exposure, 638
in breast cancer, 85, 383–​384 physical activity, 638
in early-​onset breast-​ovarian cancer, 84, 85–​86 reproductive factors, 637
in esophageal cancer, 188 sibship size, 633
in familial adenomatous polyposis, 84 sinusitis, 637
in hereditary nonpolyposis colorectal cancer, 84 socioeconomic status, 633
linkage studies of, 85–​86 summary of, 640t
709

Index 709

systemic lupus erythematosus, 637 Human leukocyte antigen (HLA) genes, NPC and,
tobacco use, 637 164–​165
tuberculosis, 637 Human papilloma virus (HPV) infection, 13
ultraviolet radiation, 638 causative for oral/​pharyngeal cancer, 137
subgroups, 649 cell regulating mechanisms, 428f
survival rate, 627 cervical cancer and, 13, 104, 421, 423, 424–​425,
symptoms, 627 427–​433
treatment, 627 detection, 428–​429
Homonymous virus hepatitis (HVA), 13 epidemiology, 429–​431
Hormone replacement therapy (HRT) esophageal cancer, 199
biliary tract cancer and, 295 esophageal cancer and, 199
bladder cancer, 554 nasopharyngeal cancer and, 170
colorectal cancer and, 263 oral/​pharyngeal cancer and, 147–​148
endometrial cancer, 442–​443 prevention
esophageal cancer and, 198 through screening, 431–​432
lung cancer and, 338, 346t through vaccination, 14, 432–​433
ovarian cancer and, 468t prostate cancer and, 501–​502
pancreatic cancer and, 314 reproductive factors, 425
thyroid cancer and, 617 sexual behaviors, 148, 425, 426
Hormones, hormone receptors. See also specific tobacco use and, 104
hormones variants, 106t
basal cell carcinoma and, 364–​365 virology, 427–​428
biomarkers, molecular epidemiology, and, 103t Human T-​cell leukemia virus type 1, 62, 106t
bladder cancer and, 554 Human T-​cell lymphotrophic viruses
brain cancer and, 593 (HTLVs), 12, 106t
breast cancer and, 18–​20, 118, 198, 392–​396, 392–​399 Hypergonadotropic hypogonadism, 463
cervical cancer and, 425–​426 Hyperinsulinemia
esophageal cancer and, 198 colorectal cancer and, 260
Hodgkin lymphoma and, 637 pancreatic cancer and, 315
liver cancer and, 286 Hyperkeratosis palmaris et plantaris, 188
lung cancer and, 338 Hypertension
nasopharyngeal cancer and, 169 colorectal cancer and, 251
non-​Hodgkin lymphoma and, 656–​657 endometrial cancer, 449–​450
pancreatic cancer and, 314 endometrial cancer and, 449–​450, 451t
squamous cell carcinoma and, 364–​365 kidney cancer and, 571, 574–​575, 577–​579, 580t
thyroid cancer and, 618 maternal hypertension, 529–​530
Hot drinks, esophageal cancer and, 197–​198 testicular cancer and, 528, 529t
HPMS1, hPMS2 genes thyroid cancer and, 621–​622
endometrial cancer and, 85t Hyperthyroidism, 621, 622, 623
gastrointestinal cancer and, 85t Hypochromic anemia, 200
HTLVS. See Human T-​cell lymphotrophic viruses Hypopharynx tumors, 149
Human Genome Project, 81 Hypospadias, 528, 529t, 531, 535
Human herpesvirus 8 (HHV8), 658 Hysterectomy
Human immunodeficiency virus (HIV). See also Acquired breast cancer and, 398
immunodeficiency syndrome; HIV/​AIDS cervical cancer and, 422
AIDS and, 15 endometrial cancer and, 442
anal cancer and, 106t kidney cancer and, 577
cervical cancer and, 106t, 434t ovarian cancer and, 460, 466, 472t, 473
eye (conjunctiva) cancer and, 106t pancreatic cancer and, 314
HHV8 and, 658 thyroid cancer and, 617–​618
Hodgkin lymphoma and, 106t, 638
HPV and, 427, 430 IARC. See International Agency for
HPV-​positive malignancies and, 148 Research on Cancer
Kaposi sarcoma and, 106t Iatrogenic exposure, as risk factor, 17–​18
kidney cancer and, 578 analgesic medications, 18
nasopharyngeal cancer and, 163 diethylstilbestrol, 17
non-​Hodgkin lymphoma and, 106t, 653, 655t, 657–​658 exogenous estrogens/​progestins, 17–​18
oral cavity cancer and, 147 oral contraceptives, 17
pancreatic cancer and, 315 IBD. See Inflammatory bowel disease
710

710 Index

ICARE study, oral and pharyngeal cancer, 139–​140 Hodgkin lymphoma and, 106t, 637
Identical by descent (IBD), allele-​sharing measure, 87 kidney cancer and, 577–​578
Identical by state (IBS), allele-​sharing measure, 87 laryngeal cancer and, 344
Illumina BeadArray, genotyping method, 91 leukemia and, 681
Immunodeficiency. See also Acquired immunodeficiency lung cancer and, 338
syndrome; Human immunodeficiency virus melanoma and, 365
ataxia-​telangiectasia and, 66, 657 Merkel cell carcinoma and, 365–​366
Bloom’s syndrome and, 66 non-​Hodgkin lymphoma and, 15, 106t, 657–​659
congenital immunodeficiency, 14 oral/​pharyngeal cancer and, 147–​149
Hodgkin lymphoma and, 638 ovarian cancer and, 470
liver cancer and, 291 pancreatic cancer and, 315
non-​Hodgkin lymphoma and, 14–​15, 655t, 657, prostate cancer and, 500–​502
660–​661 SCC and, 365
Wiskott-​Aldrich syndrome and, 14, 654 skin cancer and, 366–​367
IMRT. See Intensity-​modulated radiotherapy testicular cancer and, 534
Incidence rates Inflammatory bowel disease (IBD)
of biliary tract cancer, 35t, 37t, 39t, 41t cervical cancer and, 433
of bladder cancer, 47f, 545f–​546f colorectal cancer and, 259
of brain/​nervous system cancers, 47f, 590f–​591f HPV and, 433
of breast cancer, 47f, 382f–​383f Information bias, 102
of cervical cancer, 47f, 422, 423f Inherited cancer-​predisposition genes, 85t
of colorectal cancer, 35t, 37t, 47f, 245f, 246f, 261f–​262f Inherited susceptibility
of EBV-​positive Hodgkin lymphoma, 636, 636f of basal cell carcinoma, 358
of endometrial cancer, 443f of biliary tract cancer, 295
of esophageal cancer, 35t, 37t, 39t, 41t, 47f, 184, of bladder cancer, 546, 550, 562t,
184f–​185f, 185f–​186f of brain cancer, 591, 599
of gallbladder cancer, 47f of breast cancer, 382–​384
of Hodgkin lymphoma, 47f, 628, 629f, 630, of cervical cancer, 423–​424
630f–​631f, 633 of colorectal cancer, 250
of keratinocyte cancers, 356f of EBV-​related NPC risk, 170
of kidney cancer, 47f, 573f–​574f of endometrial cancer, 85t, 443–​444
of laryngeal cancer, 35t, 37t, 39t, 41t, 343f, 344f of Hodgkin lymphoma, 630–​632
of leukemias, 36t, 38t, 40t, 42t, 47f, 48f, 51t, ​53t, of kidney cancer, 546, 572–​575
677t–​678t, 679, 679f–​680f of leukemias, 680–​681
of liver cancer, 47f, 278f–​279f, 279f of liver cancer, 280
of lung cancer, 47f, 328, 328f–​329, 328f–329f, ​330t of lung cancer, 328, 332, 334, 346t
of melanoma, 357f of nasopharyngeal cancer, 163–​164
of multiple myeloma, 47f of non-​Hodgkin lymphoma, 653–​654
of nasopharyngeal cancer, 162f, 169 non-​Hodgkin lymphoma
of non-​Hodgkin lymphoma, 36t, 38t, 40t, 42t, 47f, and, 653–​654, 664
651f–​652f of oral and oropharyngeal cancer, 139–​141
of oral and pharyngeal cancer, 35t, 37t, 39t, 41t, of ovarian cancer, 85, 461
47f–​48f, 138f–​139f, 142 of prostate cancer, 92, 485
of ovarian cancer, 47f, 460f of stomach cancer, 218–​219
of pancreatic cancer, 35, 37t, 39t, 41t, 47f, 51f, 310, of testicular cancer, 527–​528
310f, 311f–​312f of thyroid cancer, 612–​613
of prostate cancer, 484f, 485 thyroid cancer and, 613, 622t
of stomach cancer, 34, 213–​214, 215f–​216f Insulin-​like growth factor 1 (IGF-​1)
of testicular cancer, 525–​526, 526f brain cancer and, 593
of thyroid cancer, 36t, 38t, 40t, 42t, 46, 47f, 47f, 610t, breast cancer and, 395, 407t
611f–​612f, 612f–​613 colorectal cancer and, 244t, 260, 261, 263
of uterine cancer, 47f diabetes mellitus and, 316
Infections. See also specific infections endometrial cancer and, 445
BCC and, 365 kidney cancer and, 577, 579
bladder cancer and, 543, 555 leukemias and, 683
brain cancer and, 593–​594 pancreatic cancer and, 315
breast cancer and, 401 prostate cancer, 486, 498, 505t
cervical cancer and, 426–​427 Insulin resistance
endometrial cancer and, 450 bladder cancer and, 556
71

Index 711

breast cancer and, 396, 400, 401 Framingham Heart Study, 101

colorectal cancer and, 256–​258 2013, death rate, 46t
endometrial cancer and, 449 Isothiocyanates
kidney cancer and, 577 lung cancer and, 336–​337, 346t
liver cancer and, 291, 293 stomach cancer and, 223
pancreatic cancer and, 315, 316 IUDs (intrauterine devices), endometrial cancer and, 426
thyroid cancer and, 619
Intensity-​modulated radiotherapy (IMRT) Kaposi’s sarcoma
for Epstein Barr virus, 159–​160 AIDS and, 15
for nasopharyngeal cancer, 159 HHV8 and, 658
Interassay variability, 103 microbiologic risk factors, 106t
Interindividual variability, 103 non-​Hodgkin lymphoma and, 106t, 658
International Agency for Research on Kaposi’s sarcoma-​associated herpesvirus (KSHV), 658
Cancer (IARC) Keratinocyte cancers (KCs), 355–​356. See also Basal cell
on cancer rates of immigrants, 19 carcinoma; Squamous cell carcinoma
on chewing betel nuts, 144, 167 autoimmune hepatitis and, 367
classification of carcinogenicity of agents, mixtures, descriptive epidemiology of, 355–​356
processes, 129t exogenous hormones and, 364
formaldehyde review, 8, 9 incidence rates, 356f
HPV cross-​sectional study, 13–​14 melanoma and, 367
on identification of human carcinogens, 105 Kidney cancer (renal cell cancer), 571–​581
oral/​pharyngeal cancer data, 138 diagnosis, 571
on salted fish consumption, 167 DNA repair and, 574, 576
smoking–​lung cancer link, 12 epidemiology
International Head and Neck Cancer Epidemiology descriptive, 571–​572
(INHANCE), 144, 145, 146–​147 genetic and molecular, 572–​575
International Lung Cancer Consortium, 331 histological subtypes, 571
Interventional epidemiology, 113–​114 incidence rates, 36t, 38t, 40t, 42t, 46t, 573f–​574f
of lung cancer, 113 pathogenesis of, 571
Intestinal gland cells cancer, 243 polymorphisms in, 574, 581
Intraepithelial carcinomas, 462 predisposing genes, 574
Intrahepatic bile duct cancer, 52f prognosis, 571
Intraindividual variability, 103 progress, 571
Iodine intake proto-​oncogenes and, 574
thyroid cancer and, 615 risk factors
Ionizing radiation acrylamide, 576
bladder cancer and, 555 alcohol, 577
brain cancer and, 594–​595 anthropometric measures, 577
breast cancer and, 5, 338, 402 artificial sweeteners, 576
cervical cancer and, 5, 230, 433 bladder cancer, 553
colorectal cancer and, 259 BMI, 575–​576
Hodgkin lymphoma and, 5, 638 contaminated drinking water, 576
kidney cancer and, 578 diabetes mellitus, 579
leukemia and, 4–​5, 114, 681–​682 diet, 575–​576
liver cancer and, 282t, 292 end-​stage renal disease, 578–​579
lung cancer and, 5, 6, 338–​339 environmental exposures, 546
nasopharyngeal cancer and, 170–​171 height, 577
non-​Hodgkin lymphoma and, 659 hypertension, 571, 574–​575, 577–​579, 580t
oral/​pharyngeal cancer and, 149 hysterectomy, 577
ovarian cancer and, 5, 470 infections, 577–​578
pancreatic cancer and, 316 inherited susceptibility, 546, 572, 575–​576
prostate cancer and, 503 insulin resistance, 577
skin cancer and, 4 ionizing radiation, 578
stomach cancer and, 230 kidney/​ureter stones, 557
testicular cancer and, 534 medications, 579–​580
thyroid cancer and, 5–​6, 620, 622t obesity, 19, 575–​580
Iron overload, esophageal cancer and, 198 occupational exposure, 6, 7, 293, 578
Ischemic heart disease parity, 580t, 581
death rate data, 4 physical activity, 578
712

712 Index

Kidney cancer (cont.) acute promyelocytic, 673

pregnancy, 577 aggressive NK-​cell, 657
reproductive factors, 577 B-​cell neoplasms, 650
socioeconomic status, 580 B-​lymphoblastic (B-​ALL), 673
thyroid cancer, 579 Burkitt lymphoma, 651, 657
tobacco, 4, 575 chronic lymphocytic (CLL), 366, 649, 663
tobacco use, 4 chronic myelogenous, 60
urologic disorders, 579 chronic myeloid, 60, 673
urothelial cell carcinomas, 543 hairy cell, 651, 654, 657
waist-​to-​hip ratio, 577 hematopoietic/​lymphoid tissues, 674–​675
symptoms, 571 human T-​cell, virus type 1, 62, 106t
treatment, 571 mixed lineage (MLL) translocations, 106t, 108
tumor-​suppressor genes and, 573 myeloid, 4
KSHV. See Kaposi’s sarcoma-​associated herpesvirus myeloproliferative neoplasms, 677t
nonsolid, 58
Lactose/​galactose intake, ovarian cancer and, 463 T and NK-​cell neoplasms, 650–651, 657, 658
Large cell carcinoma, 327 survival rates, 53t
Laryngeal cancer, 342–​344 symptoms, 673, 676
genetic polymorphisms in, 140, 141 treatment, 676
incidence rates, 35t, 37t, 39t, 41t, 343f, 344f Leukoplakia
low-​penetrance polymorphisms in, 140, 141 beta-​carotene and, 145
mortality rate, 9–​10 vitamin C and, 145
risk factors Life Span Study, of atomic bomb survivors, 292
alcohol use, 16 Li-​Fraumeni syndrome
diet, 343–​344 brain cancer and, 591
infections, 344 melanoma and, 359
occupational exposure, 344 oral/​pharyngeal cancer and, 140
tobacco use, 4, 343 p53 gene and, 85t, 86–​87
Leiomyosarcomas, 213 stomach cancer and, 219
Leukemias, 673–​684 Linkage analysis, 84
descriptive epidemiology, 676, 679–​680 early-​onset breast cancer study, 85
diagnosis, 676, 677t–​678t for identiying chromosomal location, 84
DNA repair and, 680 limitations of, 86
epidemiology Linkage disequilibrium, 88–​89
descriptive, 676, 679–​680 Lip cancer, 137
genetic and molecular, 680–​681 Liver cancer, 277–​294. See also Angiosarcoma of the
global trends and patterns, 51t liver; Cholangiocarcinoma
incidence rates, 36t, 38t, 40t, 42t, 47f, 51t–​53t, 677t–​678t diagnosis, 277
prognosis, progress, 676 DNA repair and, 284, 285, 288
risk factors epidemiology
alcohol use, 683 descriptive, 278–​280
biologic factors, 106t genetic, molecular, 280–​281
birth order, 681 genome-​wide association studies, 280
diet, 683 global trends and patterns, 51t
electro-​magnetic fields, 682 hepatitis and, 12–​13
formaldehyde, 8 incidence rates, 46t, 278f–​279f
HTLVs, 12 mortality rates, 47, 49f, 51t
infections, 681 pathogenesis of, 280
inherited susceptibility, 680–​681 prognosis, 277
ionizing radiation, 4–​5, 114, 681–​682 progression, 278
medicines, drugs, 682–​683 risk factors
microbiologic environment, 12 aflatoxin contamination, 282t, 283–​284
occupational exposure, 6, 7–​9, 11, 51t, 683 alcohol use, 13, 16, 51t, 282t, 285
overweight, obesity, 51t androgens, 286
summary of, 684t angiosarcoma, 282t
tobacco use, 51t, 683 anthropometric measures, 287
subgroups BMI, 287
acute lymphoblastic, 649 cholangiocarcinoma, 282t
acute myeloid, 108 coffee consumption, 282t, 285
713

Index 713

diabetes mellitus, 282t, 287 observational, 114

diagnostic thorium dioxide, 282t genetic polymorphisms in, 330–​332, 331t
diet, 283–​285 global trends and patterns, 51t
drug use, 51t GWAS of, 330–​332
environmental exposures, 281 high-​penetrance genes and, 330
gallbladder disease, 282t incidence rates, 46t, 328, 328f–​329f, 328f–329f, ​330t
green tea, 282t, 284 interaction/​effect modification and, 104
HBC virus, 290–​292 Mendelian randomization and, 93
HBV virus, 282t, 287–​290 mortality rates, 46t, 47–​49, 49f, 51
hemochromatosis, 282t nested case-​control studies, 119
hormones, 286–​287 never-​smoking variant, 188
immunodeficiency, 291 non-​small-​cell lung cancer, 63, 74, 75, 327
inherited metabolic disorders, 282t occupational exposure and, 6, 7, 8, 10, 12
inherited susceptibility, 280 PLCO cancer screening trial, 146
insulin resistance, 291, 293 polymorphisms in, 330, 331t
ionizing radiation, 282t, 292 prognosis, 327
liver cirrhosis, 13 progress, 327–​328
liver flukes, 292 proto-​oncogenes and, 332
macro-​/​micronutrients, 284 risk factors
medical conditions, 293–​294 air pollution, 341–​342, 342t
microbiologic exposure, 106t alcohol, 337
microorganisms, 12 anthropometric measures, 338
nonalcoholic fatty liver disease, 282t, 293 BMI, 338
nonalcoholic steatohepatitis, 282t diet, 335–​337
obesity, 282t, 287 drinking water contamination, 342
occupational exposures, 282t, 293 environmental exposures, 345
parity, 282t, 286, 295 family history, 330, 346t
porphyria cutanea tarda, 282t green tea, 336
probable relationships, 282t height, 338
reproductive, 285–​286 hormones, 338
thorotrast, 6 infections, 338
tobacco use, 4, 13, 51t, 281, 282t, 283 inherited susceptibility, 328, 332, 334, 346t
somatic events, 280–​281 ionizing radiation
subgroups, 277 medical conditions, treatment, 341
survival rates, 53t nasopharyngeal cancer, 167
symptoms, 277 obesity, 346t
treatment, 277 occupational exposure, 5–​11, 127, 327, 339–​341,
tumor-​suppressor genes and, 280–​281, 288, 290 340t, 346t
U.S. survival rates, 53t passive smoking, 12, 114, 334–​335, 345, 346t
Liver cirrhosis, 13, 280 progesterone, 338, 398
Liver flukes, liver cancer and, 292 tobacco, 3–​4, 12, 111, 121, 123–​124, 126t, 127, 130,
Loss of heterozygosity (LOH), 81 327, 333–​335
Low-​penetrance polymorphisms, 140–​141 somatic cells, genetic events, and, 332–​333, 333t
in breast cancer, 384 subgroups, 327
in colorectal cancer, 249–​250 survival rates, 47–​48, 53t
in laryngeal cancer, 140, 141 symptoms, 327
in oral, oropharyngeal cancer, 140 treatment, 327
in ovarian cancer, 461 tumor-​suppressor genes and, 330, 332, 345
Luminal A/​luminal B breast cancer, 62, 381 2013, death rate, 46t
Lung cancer, 327–​346 Lutein, lung cancer and, 336
cohort studies, 117–​119 Lycopene
confounding and, 104f lung cancer and, 336
deaths, U.S., 3 oral/​pharyngeal cancer and, 145
descriptive studies, 114 prostate cancer and, 491–​492
diagnosis, 327 Lymphatic cancer, benzene and, 11–​12
ecologic studies, 115 B lymphoblastic leukemia/​lymphoma, NOS, 650
epidemiology Lymphocyte-​depleted classical Hodgkin lymphoma,
descriptive, 328 630, 650
interventional, 113 Lymphocyte-​rich classical Hodgkin lymphoma, 650
714

714 Index

Lymphoma. See Hodgkin lymphoma; Non-​Hodgkin susceptibility genes in, 359

lymphoma tumor-​suppressor genes and, 359, 361
Lymphoplasmacytic lymphoma, 650 Melatonin levels, 103
Lynch Syndrome. See Hereditary nonpolyposis Men, cancer incidence in
colorectal cancer biliary tract cancer, 37t, 41t
brain cancer, 590f
Macronutrients/​micronutrients, liver cancer and, cancer burden
284–​285 age standardization, 37t–​38t
Magnesium, esophageal cancer and, 195 cumulative incidence, 41t–​42t
Magnetic field exposure. See Electro-​magnetic field global mortality rates, 49f
exposure colorectal cancer, 18, 37t, 52f, 246f, 262
Males. See Men esophageal cancer, 37t, 41t, 184
Mantle cell lymphoma, 650 Hodgkin lymphoma, 628f–​629f
Marijuana, testicular cancer and, 532 laryngeal cancer, 37t, 41t, 47f, 343f
Maté tea consumption leukemia, 680f
esophageal cancer and, 149, 150 liver cancer, 279f
oral/​pharyngeal cancer and, 149–​150 lung cancer, 328, 329f
Meat consumption melanoma, 357f
bladder cancer and, 553 nasopharyngeal cancer, 161f, 162f
colorectal cancer and, 19, 51t, 114, 226, 251–​252 nonmelanoma skin cancer, 48f
kidney cancer and, 575–​576 oral/​pharyngeal cancer, 138, 139f, 148
lung cancer and, 336 pancreatic cancer, 35t, 41t
oral/​pharyngeal cancer and, 144–​145 stomach cancer, 37t, 41t, 50, 222
ovarian cancer and, 464 thyroid cancer, 611f
prostate cancer and, 449, 487, 488–​489 Mendelian inheritance, 83–​84, 87, 96
stomach cancer and, 226 MEN1 gene, multiple endocrine syndrome and, 85t
testicular cancer and, 532 Meningeal tumors, 589t
Medical conditions, treatment Menopausal hormone therapy (MHT)
bladder cancer and, 557–​558 basal cell/​squamous cell carcinoma and, 364
brain cancer and, 597–​598 breast cancer and, 381, 383, 401
cervical cancer and, 433 liver cancer and, 286
endometrial cancer and, 449–​450 ovarian cancer and, 467, 469
lung cancer and, 341 Menopause
non-​Hodgkin lymphoma and, 662 breast cancer and, 392, 397–​398
thyroid cancer and, 621–​622 endometrial cancer and, 441, 447
Mediterranean diet, breast cancer and, 386 ovarian cancer and, 392
Medullary cancer, thyroid cancer and, 621–​622 Merkel cell carcinoma, 362t, 365–​367
Melanoma, 356f–​357f Mesothelioma
DNA repair and, 359, 365 asbestos and, 7, 11
genome-​wide association studies, 360 pathogenesis of, 345
incidence, 356, 357f, 358 Metabolic syndrome
Li-​Fraumeni syndrome and, 359 colorectal cancer and, 244t, 256–​257, 260
proto-​oncogenes and, 360 endometrial cancer and, 450
risk factors pancreatic cancer and, 315
alcohol use, 364 prostate cancer and, 503
anthropometric measures, 365 Metastasis, mechanisms of, 59
diabetes, 367 Metastatic colonization, 59
diet, 363–​364 MET gene, familial papillary renal carcinoma syndrome
exogenous hormones, 365 and, 85t
infections, 365 Microbiologic environment, risk factors, 12–​15
obesity, 365 bladder cancer and, 106t
Parkinson’s disease, 367 cervical cancer and, 12, 13–​14
pregnancy, 364 Helicobacter pylori and, 14
reproductive, 364 hepatitis viruses, 12–​13
solar UV radiation, 367–​369 HIV/​AIDS and, 15
tobacco use, 362–​363 Hodgkin lymphoma and, 106t
skin color and prevalence, 356 HPV infection and, 13–​14
solar UV radiation and, 355 immunodeficiency and, 14–​15
somatic events, 360–​361 Kaposi’s sarcoma and, 15
715

Index 715

liver cancer and, 12, 106t genetic, 164

stomach cancer and, 12, 14, 106, 199 genome-​wide association studies, 165–​166
Micrometastases, 59 historical background, 159
Micronutrient intake. See also specific micronutrients incidence rates, 47f–​48f, 160, 160f–​162f, 169
breast cancer and, 387–​389 male-​specific data, 161f, 162f
esophageal cancer and, 194–​196 physical activity, 170
oral/​pharyngeal cancer and, 145 prognosis, 160
Microsatellite instability (MIN), 87, 96 proto-​oncogenes and, 166
Microsporum canis, lung cancer and, 338 risk factors
Missense mutations, 82, 96, 188, 359 anthropometric measures, 169
Mixed cellularity Hodgkin lymphoma, 650 betel nut, 167
Mixed lineage (MLL) translocations of leukemia, 106t, 108 cannabis smoking, 167
Moderate-​penetrance polymorphisms, breast cancer cervical cancer, 159
and, 384 diet, 165, 167
Mortality rates dust exposure, 171
Asians/​Asian countries, 48 EBV infection, 106t, 169–​170
bladder cancer, 51t environmental exposures, 162, 165, 166, 168
breast cancer, 49f, 51t formaldehyde, 8, 171
cancer burden and, 9, 31–​34, 44–​52, 46t, 47f, 48f, 49f green tea, 169
cervical cancer, 50–​51, 51t, 423 hormones/​hormone receptors, 169
chromium exposure and, 8 human papilloma virus, 170
colorectal cancer, 7, 46, 46t, 49f, 50, 51t, 52f, 243 inconsistent findings, 172t
endometrial cancer, 44 inherited susceptibility, 162–​164
environmental factors, 18–​19 ionizing radiation, 170–​171
formaldehyde exposure and, 9 lung cancer, 167
Hodgkin lymphoma, 628 passive smoking, 167
liver cancer, 6, 47, 49f, 51t possible factors, 172t
lung cancer, 46t, 47–​49, 49f, 51 salt-​preserved fish, 166, 168
lung cancer and, 21 tea consumption, 169
as measure of disease burden, 31–​34, 44–​52 tobacco smoking, 165, 166, 167, 170
nonmelanoma skin cancer, 44 well-​confirmed factors, 172t
pancreatic cancer, 21, 31, 44, 48, 51, 51t wood fire smoke, 167
PCB exposure and, 9 somatic mutations of, 166
prostate cancer, 49f, 50t, 483–​484 subgroups, 159
skin cancer, 44 survival data, 160
stomach cancer, 47, 49f, 50, 51t symptoms, 159
thyroid cancer, 44 treatment, 159–​160, 172–​173
tracheal cancer, 51t tumor-​suppressor genes and, 165, 166
uterine cancer, 49f women-​specific data, 160f, 161f, 162f
WHO data, 31 National Cancer Institute Breast & Cancer Association
Mouthwash, oral/​pharyngeal cancer, 149 Consortium, 92
Mucinous cancers, 459, 462 NCI Breast and Prostate Cohort Consortium, 91, 92
Multicentric Castleman’s disease (MCD), 658 NCI Cancer Cohort Consortium, 147
Multiple endocrine neoplasia 2, RET gene and, 85t Neoplasm, unspecified, 589t
Multiple endocrine syndrome (MEN1), MEN1 gene Netherlands Cohort Study, 145–​146
and, 85t Neuroepithelial tissue tumors, 589t
Multiple myeloma, 47f Neurofibromatosis (type 1 and 2), 85t, 591
Mutations/​polymorphisms, in human genes. See Neuronal/​mixed neuronal-​glial tumors, 589t
Polymorphisms Never-​smoking lung cancer, 188
Mycosis fungoides, 650 Nevoid basal cell carcinoma syndrome, 85t
Myeloid leukemia, 4 NF2 gene, neurofibromatosis 2 and, 85t
Nickel
Nasopharyngeal carcinoma (NPC), 160–​172 cigarettes and, 4
diagnosis, 159 lung cancer and, 9
DNA repair and, 165 nasal cancer and, 9
epidemiology occupational exposure to, 9
descriptive, 160–​163 scrotal cancer and, 6
genetic/​molecular, 165 Night shift work, breast cancer and, 403
Epstein Barr virus association, 159–​160 NIH Epigenomics Roadmap, 93
716

716 Index

N-​nitroso compounds Nonsteroidal anti-​inflammatory drugs (NSAIDs)

colorectal cancer and, 252 breast cancer and, 404
stomach cancer and, 225–​226 colorectal cancer and, 232
Nodal marginal zone lymphoma, 650 ovarian cancer and, 471
Nodular lymphocyte predominant Hodgkin lymphoma, pancreatic cancer and, 318
627, 650 skin cancer (BCC/​SCC) and, 367
Nodular sclerosis classical Hodgkin stomach cancer and, 232
lymphoma, 630, 650 Nurses’ Health Study, Nurses’ Health Study II, 386
Noise trauma, brain cancer and, 535, 598
Nonalcoholic fatty liver disease (NAFLD), 282t, 293 Obesity
Nonalcoholic steatohepatitis (NASH), 282t basal cell carcinoma and, 365
Nonendometrioid carcinomas (type 2), 441 biliary tract cancer and, 294, 295
Non-​Hodgkin lymphoma, 649–​664 bladder cancer and, 554, 558, 562t
diagnosis, 649 brain cancer and, 593, 599
epidemiology breast cancer and, 19–​20, 51t, 382, 403, 406
descriptive, 651–​653 cervical cancer and, 426, 434t
genetic and molecular, 653–​654 colorectal cancer and, 19, 51t, 255, 256, 257, 260
immunodeficiency and, 14–​15 endometrial cancer and, 441, 443, 445, 449, 452
incidence rates, 36t, 38t, 40t, 42t, 46t, 651f–​652f esophageal cancer and, 189, 202
prognosis, 651 gallbladder cancer and, 19
progress, 651 Hodgkin lymphoma and, 637, 640t
risk factors kidney cancer and, 19, 575–​576, 577, 578, 579, 580
AIDS, 14–​15 leukemia and, 51t
alcohol use, 656 liver cancer and, 50, 281, 282t, 287
allergies, 662–​663 lung cancer and, 346t
anthropometric measures, 657 melanoma and, 365
autoimmune diseases, 661 nonalcoholic fatty liver disease and, 287
blood transfusions, 655t, 658, 661 non-​Hodgkin lymphoma and, 657
chronic inflammatory diseases, 661 oral/​pharyngeal cancer and, 147
diet, 656 ovarian cancer, 469, 472t, 473
environmental exposures, 656 pancreatic cancer and, 19, 51t, 315, 318, 319t
hair dyes, 663 prostate cancer and, 499–​500, 502, 503, 505t
HCV, 106t SCC and, 365
HIV-​1, 106t steatohepatitis and, 287
hormones, 656–​657 thyroid cancer and, 618, 619, 622
immune modulation, 660–​661 Observational epidemiology, 93, 114, 122–​123, 128, 145.
immunodeficiency, 660–​661 See also Epidemiologic studies
infections, 15, 106t, 657–​659 Observational studies (experimental studies), 114
inherited susceptibility, 653–​654, 664 Observed survival, 43
ionizing radiation, 659 Occupational exposure. See also Environmental
medications, 662 exposures
obesity, 657 bladder cancer and, 8, 10–​11, 51t, 555–​557
occupational exposure, 11–​12, 558, 659–​660 brain cancer and, 595–​597
physical activity, 659 breast cancer and, 120, 402–​403
reproductive factors, 656 cervical cancer and, 113, 433
skin cancer, 366 colorectal cancer and, 259
subclinical immune activation, 662 esophageal cancer and, 7, 199–​200
systemic lupus erythematosus, 637 Hodgkin lymphoma and, 638
tobacco use, 654, 656 IARC classifications, 556b
ultraviolet light, 663 kidney cancer and, 6, 7, 293, 578
vaccinations, 662 laryngeal cancer and, 344
somatic mutations, 654 leukemia and, 6, 7–​9, 11, 51t, 683
subgroups, 649 liver cancer and, 282t, 293
symptoms, 649 lung cancer and, 5–​11, 5–​12, 127, 327, 339–​341, 340t
treatment, 649, 651 nasal cancer and, 9
Nonmelanoma skin cancer, 33–​34 non-​Hodgkin lymphoma and, 11–​12, 558, 659–​660
Nonsense mutations, 82, 96 oral/​pharyngeal cancer and, 149
Non-​small-​cell lung cancer, 63, 327 ovarian cancer and, 470–​471
Nonsolid leukemia, 58 pancreatic cancer and, 231, 316
71

Index 717

pleural mesothelioma and, 345 well-​confirmed factors, 143t

prostate cancer and, 503 SEER-​Medicare study findings, 148–​149
skin cancer and, 10 somatic mutations, 141–​142
stomach cancer and, 6, 7, 8, 11, 21, 230–​231, 233t symptoms, 137
testicular cancer and, 534 treatment, 137
thyroid cancer and, 620 tumor-​suppressor genes and, 141
urinary tract cancer and, 11 WHO data, 137
Oculocutaneous albinism, 359 World Cancer Research Fund data, 144, 146, 147
Oligoastrocytic tumors, 589t Oral contraceptives (OCs)
Oligodendroglioma, 589t BCC, SCC, and, 364–​365
Olive oil breast cancer and, 17, 117, 396–​397
breast cancer and, 386 cervical cancer and, 17, 424, 426
oral/​pharyngeal cancer and, 146 endometrial cancer and, 17, 447
Omega-​3 fatty acids ovarian cancer and, 17, 466–​467
basal cell carcinoma and, 363 thyroid cancer and, 617
colorectal cancer and, 263 Oral health
prostate cancer and, 488, 489 esophageal cancer and, 202
skin cancer and, 363 stomach cancer and, 229
Oncogenes, 85t. See also names of specific genes Organ transplantation
Opisthorchis viverrini (liver fluke), 106t, 292 cancer risk factors, 15
Oral and pharyngeal cancer, 137–​150. See also cervical cancer and, 433, 434t
Nasopharyngeal cancer EBV and, 657
associated cancers, 137 nasopharyngeal cancer and, 170
descriptive epidemiology, 138–​139 non-​Hodgkin lymphoma and, 655, 661
diagnosis, prognosis, 137 skin cancer and, 365, 366
epidemiology Osteoporosis
descriptive, 138–​139 in Werner’s syndrome, 66
genetic/​molecular, 140 Osteosarcomas, 87
gene mutations Outdoor air pollution
high-​penetrance, 140 bladder cancer and, 560
low-​penetrance, 140–​141 lung cancer and, 341, 342
ICARE study, 139–​140 Ovarian cancer, 459–​473
incidence rates, 35t, 37t, 39t, 41t, 47f–​48f, 138f–​140f, 142 breast-​ovarian cancer syndromes, 84, 85, 461
INHANCE data, 144, 145, 146–​147 diagnosis, 459
morbidity rates, 21, 51t epidemiology
Netherlands Cohort Study, 145 descriptive, 459–​461
prognosis, 137–​138 genetic and molecular, 461–​462
risk factors global survival rates, 53t
alcohol use, 16, 51t, 113, 140, 145, 146–​147 high-​penetrance germline mutations, 461
anthropometric measures, 147 incidence rates, 46t, 459–​460, 460–​461, 460f
betel nuts, 144, 167 low-​penetrance gene polymorphisms, 461
diet/​dietary patterns, 144–​146 pathogenesis of, 468t, 469, 473
genetic, 139–​140 PLCO cancer screening trial, 146
GERD, 143t, 149 polymorphisms in, 461
HPV type 16, 137 prognosis, 459
inconsistent findings, 143t risk factors
infections, 147–​149 alcohol, 464
inherited susceptibility, 139–​140 analgesic medication, 471
ionizing radiation, 149 androgens, 462, 469
maté consumption, 149–​150 anthropometric measures, 469
medical conditions, treatment, 149 breastfeeding, 466, 468t, 472t, 473
microbiologic exposure, 106t coffee and tea, 463–​464
mouthwash, 149 diabetes mellitus, 471
occupational exposure, 149 diet, 462–​464, 472t
physical activity, 149 endometriosis, 471
probable relationships, 143t fertility drugs, 465
reproductive factors, 147 gene mutations, 87
tobacco use, 4, 51t, 112f, 113, 140, 142–​144, 147, 148 green tea, 463–​464
weak relationships, 143t high ovulatory cycles, 467–​469
718

718 Index

Ovarian cancer (cont.) diabetes mellitus, 316, 319t

hysterectomy, 460, 466, 472t, 473 diet, 313, 320t
infections, 470 estrogen receptors, 314
inherited susceptibility, 85, 461 familial cancer syndromes, 319t
ionizing radiation, 5, 470 gallbladder disease, 317
lactose intolerance, 463 H. pylori infection, 315, 319t
menopausal hormone therapy, 467, 469 HBV, 315
menopause, 392 HIV/​HAART therapy, 315
nulliparity, 464–​465 hyperinsulinemia, 315
obesity, 469, 472t, 473 hysterectomy, 314
occupational exposure, 470–​471 infections, 315
oral contraceptives, 17, 466–​467 insulin resistance, 315
PCOS, 471 ionizing radiation, 316
physical activity, 449, 469–​470 Lynch II syndrome, 249
pregnancy, 463, 465, 468t, 469 obesity, 19, 51t, 315, 319t
progesterone, 461, 465, 468t, 469 occupational exposure, 231, 316
reproductive factors, 15–​16, 447, 464–​466 overweight, obesity, 19, 51t
socioeconomic status, 460 pancreatitis, 317–​318, 319t
talcum powder, 468t, 471–​473 partial gastrectomy, 318
tobacco use, 462 passive smoking, 313
tubal ligation, 459, 466, 468t, 472t, 473 periodontal disease, 315
somatic events/​mutations, 461–​462 pernicious anemia, 317
subgroups, 459 physical activity, 315–​316
symptoms, 459 pregnancy, 314
treatment, 459 primary sclerosing cholangitis, 318
tumor-​suppressor genes and, 462 reproductive, 314
Ovulatory cycles summary of, 319t
breast cancer and, 391 taller height, 315, 319t
ovarian cancer and, 467–​469 tobacco use, 4, 51t, 312–​313, 319t, 320t
subgroups, 309
Pacific Islander Americans survival rates, 52f, 309–​310
lung cancer and, 330t symptoms and signs, 309
nasopharyngeal cancer and, 163 treatment, 309
Pancreatic cancer, 309–​320 Pancreatitis, pancreatic cancer and, 317–​318, 319t
cancer-​disposition genes, 85t Papillary cancer, thyroid cancer and, 621
cigarettes and, 4 Parity
diagnosis, 309 bladder cancer, 577
DNA repair and, 312 brain cancer, 593
epidemiology breast cancer and, 391, 403
descriptive, 310 cervical cancer and, 424, 425, 431f, 434t
genetic, molecular, 311–​312 colorectal cancer and, 256
familial occurrence of, 318 endometrial cancer, 447, 448, 451, 452
genetic variations and, 188 esophageal cancer and, 198
hormones, 314–​315 kidney cancer, 580t, 581
incidence rates, 35, 37t, 39t, 41t, 46t, 51t, 310f–​312f liver cancer and, 282t, 286, 295
mortality rates, 21, 31, 44, 48, 51t ovarian cancer, 459, 464–​465, 467, 468t, 469, 472t
progress, prognosis, 309–​310 pancreatic cancer and, 314
risk factors prostate cancer, 487
alcohol use, 314 as risk factor, 15
allergies, 318 skin cancer and, 364
alpha-​tocopherol, 313 thyroid cancer and, 617
anthropometric measures, 315 Passive smoking, lung cancer and, 12, 114, 334–​335,
aspirin, NSAIDs, 318 345, 346t
BMI, 315 Patched gene, in Gorlin syndrome, 358
chlorinated water, 318 Paterson-​Kelly syndrome, esophageal cancer and, 200
cholecystectomy, 317 Pathogenesis of cancer
coffee consumption, 313–​314 basal cell carcinoma, 358, 360
cystic fibrosis, 318 candidate cancer genes and, 88
dairy foods, 313 chronic myeloid leukemia, 60
719

Index 719

colorectal cancer, 250, 257, 260–​261, 265 Pituitary tumors, 589t

gene-​gene interactions, 104 Plasma cell neoplasms, 650
HIV/​AIDS-​related lymphomas, 661 Pleiotropy, 92
Hodgkin lymphoma, 637 Pleural mesothelioma (cancer of the pleura), 344–​345
kidney cancer, 571 Plummer-​Vinson syndrome, 200
liver cancer, 280 Polychlorinated biphenyls (PCBs)
mesothelioma, 345 brain cancer and, 597, 600t
ovarian cancer, 468t, 469, 473 cohort studies, 9
post-​transplant lymphoproliferative disorders, 660 endometrial cancer and, 450
P16/​CDKN2A gene, hereditary melanoma syndrome liver cancer and, 293
and, 85t occupational exposure to, 9
PCOS. See Polycystic ovary syndrome skin cancer and, 366
Pedigree, defined, 96 Polycyclic aromatic hydrocarbons (PAHs), 104, 106
Penile cancer, 106t breast cancer and, 406
Peptic ulcer, 213, 231 cervical cancer and, 424
Perinatal exposures esophageal cancer and, 193, 198, 201t
breast cancer and, 404 lung cancer and, 332, 340–​341, 346t
prostate cancer and, 504, 506 oral/​pharyngeal cancer and, 141, 147
testicular cancer and, 528, 529t, 530–​531 pancreatic cancer and, 312
Periodontal disease, pancreatic cancer and, 315 stomach cancer and, 226–​227
Peripheral T-​cell lymphoma, NOS, 650 Polycystic ovary syndrome (PCOS)
Pernicious anemia, pancreatic cancer and, 317 endometrial cancer and, 446, 450, 451t
Person-​time, 115–​116 ovarian cancer and, 471
Pesticide exposure Polygenic risk score
brain cancer and, 597 for bladder cancer, 551
esophageal cancer and, 200 for breast cancer, 93
Petrochemical workers, brain cancer and, 596, 597 Polymerase chain reaction (PCR), 91, 96, 290, 428–​430
Peutz-​Jeghers syndrome, 320t Polymorphisms (genetic polymorphisms). See also
Physical activity High-​penetrance polymorphisms; Low-​
bladder cancer and, 554 penetrance polymorphisms; Moderate-​
brain cancer and, 594 penetrance polymorphisms; Single nucleotide
cervical cancer and, 433 polymorphisms; Somatic events/​mutations
colorectal cancer and, 257–​259 alcohol-​metabolizing polymorphisms, 189
endometrial cancer and, 449 biomarkers, 103t
Hodgkin lymphoma, 638 in bladder cancer, 543, 546–​547, 550–​551, 561
kidney cancer and, 578 in brain cancer, 591
nasopharyngeal cancer and, 170 in breast cancer, 383–​384
non-​Hodgkin lymphoma and, 659 candidate-​gene studies and, 165
oral/​pharyngeal cancer and, 149 case-​control studies and, 89–​90, 120
ovarian cancer and, 449, 469–​470 characterizing/​cataloging of, 89
pancreatic cancer and, 315–​316 in colorectal cancer, 250
prostate cancer and, 502–​503 contributions to phenotypic differences, 83
stomach cancer and, 229–​230 defined, 82–​83, 96
testicular cancer and, 532–​533 in endometrial cancer, 443–​444
thyroid cancer and, 620 genetic association studies, 120
Physical environment, risk factors, 10–​12 genotyping methods and, 91–​92
Physical inactivity GWAS and, 165–​166
breast cancer and, 51t in kidney cancer, 574, 581
colorectal cancer and, 51t in laryngeal cancer, 140, 141
Phytoestrogens in lung cancer, 330–​332, 331t
breast cancer and, 389 in oral, oropharyngeal cancer, 140
prostate cancer and, 492 in ovarian cancer, 461
in soy/​prostate cancer and, 492 in prostate cancer, 486
Picked vegetables relationship to diseases, 89
esophageal cancer and, 197, 201t single nucleotide polymorphisms (SNPs), 120–​121,
nasopharyngeal cancer and, 167 156, 188, 250, 574
stomach cancer and, 227 in skin cancer, 358–​359
Pilocytic astrocytoma, 589t in stomach cancer, 219
Pineal region tumors, 589t in testicular cancer, 527, 533
720

720 Index

Pooling Project of Prospective Studies of Diet and polymorphisms in, 486

Cancer, 389 prognosis, 481–​482
Porphyria cutanea tarda, 280, 282t progress, 482
Porphyromonas gingivalis infection, 315 PSA testing, 33
Positional cloning, 84, 85, 86, 96 risk factors
Post-​transplant lymphoproliferative disorders, 660 age, 484–​485
Pregnancy alcohol, 495–​496
brain cancer and, 593 alpha-​tocopherol, 487, 492–​493, 505t
breast cancer and, 15, 385, 390, 391–​392, 396, 404, 406 androgens, 496–​497, 499–​500
cervical cancer and, 425, 432 anthropometric measures, 93, 499–​500
diethylstilbesterol and, 17 BMI, 499
endometrial cancer and, 448, 452 body size in early life, 499–​500
hormonal agents during, 529–​530 coffee intake, 494–​495
kidney cancer and, 577 dairy foods, 490
melanoma and, 364 diet, 115, 449, 487–​495
ovarian cancer and, 463, 465, 468t, 469 estrogens, 198, 497
pancreatic cancer and, 314 fish consumption, 488–​490, 505t
smoking during, 530 5-​alpha reductase, 497
thyroid cancer and, 617–​618, 622t height, 93, 500
Premature aging, in Werner’s syndrome, 66 IGF-​1, 486, 498, 505t
Prenatal exposures infections, 500–​502
breast cancer and, 406 inherited susceptibility, 92, 485
testicular cancer and, 528, 535 ionizing radiation, 503
Preneoplasia, 185 meat consumption, 449, 487, 488–​489
Prevalence, as measure of disease burden, 34, 43, 45, metabolic syndrome, 503
47–​48, 50, 54 obesity, 499–​500, 502, 503, 505t
Primary cutaneous CD30 positive T-​cell lymphoprolif- occupational exposure, 503
erative disorders, 650 perinatal exposures, 504, 506
Primary sclerosing cholangitis, 318 physical activity, 502–​503
Progesterone phytoestrogens, 492
breast cancer and, 62, 392–​395, 407t reproductive factors, 496–​499
cervical cancer and, 426 selenium, 487, 489, 493–​494
endometrial cancer and, 441, 444, 445, 446–​448 sexual behaviors, 501
lung cancer and, 338, 398 STIs, 496, 501, 505t, 506
ovarian cancer and, 461, 465, 468t, 469 tobacco, 487
testicular cancer and, 528 vasectomy, 504, 505t
Progestins vitamin E, 487, 489, 492–​493
breast cancer and, 17–​18, 396 weight change, 500
endometrial cancer and, 446–​447 subgroups, 481
Prolactin survival data, 52f, 488, 489, 490–​491
breast cancer and, 391, 395–​396, 406, 407t symptoms, 481
liver cancer and, 286 treatment, 259, 481
Propionibacterium acnes, 502 tumor-​suppressor genes and, 501
Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer Prostate Cancer Prevention Trial (PCPT), 485
screening trial, 146 Prostate specific antigen (PSA) testing, 33
Prostate cancer, 481–​506 Proto-​oncogenes
benign prostatic hyperplasia and, 481, 501 cervical cancer and, 424
diagnosis, 33, 481 kidney cancer and, 574
DNA repair and, 485–​486, 495, 505t lung cancer and, 332
epidemiology melanoma and, 360
descriptive, 482–​485 nasopharyngeal cancer and, 166
genetic and molecular, 485–​486 origins of cancer and, 62–​63, 66, 68–​70
ERSPC screening trial, 484 stomach cancer and, 221, 228
genetic variants, 92 thyroid cancer and, 613
genome-​wide association studies, 92, 444, 486 PTCH gene, nevoid basal cell carcinoma syndrome
incidence rates, 37t, 41t, 45, 48f, 52f, 482f–​484f and, 85t
migrant studies, 485 PTEN gene, Cowden syndrome and, 85t
mortality rates, 49f, 50t, 483–​484 Pulmonary fibrosis, asbestos and, 7
PLCO screening trial, 146, 484 P-​value, 121–​122, 330
721

Index 721

Randomized control trials SCC. See Squamous cell carcinoma

ethical considerations, 123 Schistosoma haematobium, 106t, 543, 562t
experimental studies, 114 Scrotal cancer, occupational risk factors, 6
interventional epidemiology and, 113 SEER-​Medicare linkage study, oral and pharyngeal
Relative survival, 43–​44 cancer, 148–​149
Renal cell cancer. See Kidney cancer Segregation analysis, 83–​84, 86
Reproductive factors, 15–​16 Selection bias, 89–​90, 102–​103, 124, 132, 279–​280
bladder cancer and, 554 Selenium
brain cancer and, 593 breast cancer and, 388
breast cancer and, 15–​16, 390–​392 esophageal cancer and, 195
cervical cancer and, 425 lung cancer and, 336
endometrial cancer and, 16, 447–​448 oral/​pharyngeal cancer and, 145
Hodgkin lymphoma and, 637 prostate cancer and, 487, 489, 493–​494
kidney cancer and, 577 Sellar region tumors, 589t
liver cancer and, 285–​286 Sequenom MassARRAY, genotyping method, 91
non-​Hodgkin lymphoma and, 656 Serous carcinoma, 441, 459, 462
oral/​pharyngeal cancer and, 147 Sexual behaviors
ovarian cancer and, 15–​16, 447, 464–​466 cervical cancer and, 421, 426
pancreatic cancer and, 314 HPV infection and, 148, 425, 426
skin cancer and, 364 oral contraceptives and, 425–​426
thyroid cancer and, 617–​618 prostate cancer and, 501
uterine cancer and, 15–​16 Sexually transmitted infections (STIs)
Respiratory infections, 46t cervical cancer and, 421, 423, 426, 429
Restriction fragment length polymorphism (RFLP), 96 prostate cancer and, 496, 501, 505t, 506
Retinoblastoma, Rb gene and, 84 testicular cancer and, 534
Retinol, oral/​pharyngeal cancer and, 145 Sezary syndrome, 650
RET proto-​oncogene, 85t Sibship size
Rheumatoid arthritis (RA), 433 Hodgkin lymphoma and, 633
Risk factor summaries. See also under names of specific testicular cancer and, 529t, 535
cancers Silent mutations, 82, 96
bladder cancer, 562t Silica
brain cancer, 600t esophageal cancer and, 198, 200
breast cancer, 407t lung cancer and, 338, 339, 340, 340t, 346t
for cancers accepted as causal since the beginning of stomach cancer and, 230
the 1980s, 692t Simian virus 40, 67, 338, 345, 346t, 593
for cancers accepted as probably not major causes Single nucleotide polymorphisms (SNPs), 120–​121, 156,
since the beginning of the 1980s, 692t 188, 250, 574
cervical cancer, 434t databases of, 89
colorectal cancer, 244t genotyping platforms, 91–​92
endometrial cancer, 451t GWAS studies, 92, 220t
esophageal squamous cell carcinoma, 201t nasopharyngeal cancer and, 165
Hodgkin lymphoma, 640t stomach cancer and, 220t
kidney cancer, 580t tagging of, and linkage disequilibrium, 88–​89
leukemia, 684t Skin cancer, 355–​370. See also Basal cell carcinoma;
liver cancer, 282t Squamous cell carcinoma
lung cancer, 346t detoxifying proteins and, 358–​359
non-​Hodgkin lymphoma, 655t diagnosis, 328, 355
ovarian cancer, 472t DNA repair defects and, 358
pancreatic cancer, 319t epidemiology
prostate cancer, 505t descriptive, 353–​358
skin cancer, 362t genetic and molecular, 358–​361
stomach cancer, 233t genome-​wide association studies, 358
testicular cancer, 529t incidence rates, 47f, 48f
thyroid cancer, 622 keratinocyte cancers and, 355–​356
melanoma form, 355, 356, 358
S. haematobium, bladder cancer and, 106t, 543, 544, 555, Merkel cell carcinoma, 362t, 365–​367
561, 562t mortality rate, 44
Salivary gland cancer, 137, 164 nonmelanoma skin cancer, 33–​34, 44, 47f
Salt intake, stomach cancer and, 225 polymorphisms in, 358–​359
72

722 Index

Skin cancer (cont.) of stomach cancer, 219, 221–​222

prognosis, 355 of testicular cancer, 528
progress, 355 of thyroid cancer, 613–​614
risk factors Spinal nerves tumors, 589t
alpha-​tocopherol, 363 Splenic marginal zone lymphoma, 650
anthropometric measures, 365 Squamous cell carcinoma (SCC), 327
arsenic, 11 of the cervix, 421
artificial UV radiation, 369–​370 dietary risk factors, 363
aspirin, NSAIDs, 367 of esophagus (See also Esophageal cancer)
chemical induction, 188 BMI and, 194
diet, 363–​364 clinical course, 183
exogenous hormones, 364–​365 diet and, 194–​198
infections, 365–​366 geographic distribution, 184–​185
inherited susceptibility, 358 incidence, 183
ionizing radiation, 4, 366 risk factors, 4, 188, 191–​194, 192f, 201t
medical conditions, treatment, 366–​367 secular trends, 185–​186
non-​Hodgkin lymphoma, 366 somatic mutations, 189–​191
occupational exposure, 10, 366 survival rate, 183–​184
physical activity, 366 treatment, 183
reproductive, 364 variants, 187, 188, 189
solar UV radiation, 10, 94, 188, 355, 358, 367–​368 gene mapping of, 359
tobacco use, 361–​363 menopausal hormone therapy and, 364
trauma, 369, 370 obesity and, 365
subgroups, 355 of skin (See also Skin cancer)
symptoms, 355 alcohol risks, 364
treatment, 355 anthropometric risk factors, 365
xeroderma pigmentosum, 85t, 358 artificial UV radiation risks, 369–​370
Small lymphocytic lymphoma, 650 exogenous hormone risks, 364–​365
Smoothened genes, in basal cell carcinoma, 358 genetic risk factors, 359
SNPs. See Single nucleotide polymorphisms genome-​wide association studies, 359
Snuff, oral cancer risk factors, 144 infection risks, 365
Socioeconomic status, 633 ionizing radiation risks, 366
endometrial cancer and, 450 medical conditions, treatment risks, 366–​367
kidney cancer and, 580 occupation exposure risks, 366
ovarian cancer and, 460 physical activity risks, 366
stomach cancer and, 233 prognosis, 355
testicular cancer and, 535 progress, 355
Soft-​tissue sarcomas, 87 solar UV radiation and, 10, 355
Somatic cells solar UV radiation risks, 367–​369
defined, 96 somatic events, 360
germline cells comparison, 81 tobacco use risks, 361–​362
lung cancer and, 332–​333 trauma risks, 370
telomerase expression and, 73 type 1 diabetes risk, 367
Somatic events/​mutations solar UV radiation and, 355
of basal cell carcinoma, 360 somatic events in, 360
of cervical cancer, 424 tumor-​suppressor genes and, 359, 360
of colorectal cancer, 246–​247 ultraviolet radiation and, 10
defined, 96 Statins, breast cancer and, 404
of endometrial cancer, 444–​445 Stomach cancer, 213–​234
of esophageal cancer, 189–​191 cardia cancer and, 217–​218
of kidney cancer, 575 death rate, 46t, 52f
of liver cancer, 280–​281 diagnosis, treatment, 213
of melanoma, 360–​361 dietary intervention trials, 224–​225
of nasopharyngeal cancer, 166 DNA repair and, 221, 228
of non-​Hodgkin lymphoma, 654 epidemiology
of oral and oropharyngeal cancer, 141–​142 descriptive, 214–​216, 214f–​215f
of ovarian cancer, 461–​462 genetic/​molecular, 219, 221–​222
of pleural mesothelioma, 345 familial occurrence of, 218
of squamous cell carcinoma, 360 genetic polymorphisms in, 219
723

Index 723

global survival rates, 53t Systemic lupus erythematosus (SLE)

global trends and patterns, 51t Hodgkin lymphoma and, 637
GWAS of, 219, 220t HPV and, 433
incidence rates, 34, 35t, 46t, 52, 213–​214, 215f–​216f non-​Hodgkin lymphoma and, 637
mortality rates, 47, 49f, 50, 51t
occupational exposure and, 7, 8, 230–​231 T-​and NK-​cell neoplasms, 650. See also specific
polymorphisms in, 219 neoplasms
prognosis, progress, 213 TaqMan, genotyping method, 91
proto-​oncogenes and, 221, 228 Tea intake
risk factors bladder cancer and, 554
acid inhibition, 232 brain cancer and, 592
alcohol use, 16, 222 lung cancer and, 336
alpha-​tocopherol, 224 nasopharyngeal cancer and, 169
analgesic medications, 18 oral/​pharyngeal cancer and, 145
aspirin/​NSAIDs, 232 ovarian cancer and, 463–​464
birth order, 233 stomach cancer and, 223
diet, antioxidants, 51t, 222, 223–​227, 233t Temporal trends, in disease burden, 44–​46
EBV, 229 disability-​adjusted life years (DALYs), 45–​46
green tea, 223 years of life lost (YLL), 45
H. pylori, 14, 219, 225, 227–​228, 233t Testicular cancer, 525–​535
inherited susceptibility, 218–​219 diagnosis, 525
ionizing radiation, 230, 233t epidemiology
microbiologic exposure, 12, 14, 106, 199 descriptive, 525–​527
N-​nitroso compounds, 225 genetic and molecular, 527–​528
non-​H. pylori gastric microbiota, 228–​229 genome wide association studies, 527
occupational exposure, 6, 7, 8, 11, 21, incidence rates, 525–​526, 526f
230–​231, 233t polymorphisms in, 527, 533
oral health, 229 prevalence, 43
partial gastrectomy, 231–​232 prognosis, 48, 525
peptic ulcer, 231 progress, 525
physical activity, 229–​230 risk factors
salt intake, 225, 233t acne, 532
socioeconomic status, 233 age at puberty, 531–​532
tea and coffee, 223 alcohol, 532
tobacco use, 51t, 222, 233t anthropometric measures, 533
secular trends, 216–​218 baldness, 532
somatic mutations, 219, 221–​222 birth order, 529t, 535
survival rates, 52f, 53, 53t, 213 birth order, sibship size, 535
symptoms, 213 body mass index, 533
treatment, 213 congenital malformations, 531
tumor-​suppressor genes and, 219, 221 cryptorchidism, 529t, 531, 533, 535
Stop codon, defined, 96 diet, 532
Study designs,114–​120. See under Epidemiologic studies diethylstilbestrol, 528, 529t, 530
Subcutaneous panniculitis-​like T-​cell lymphoma, 650 endogenous hormone exposure, 528
Susceptibility genes fertility, semen quality, 533
in breast cancer, 85 height, 533
family patterns, 83 hypertension, 528, 529t
implication in rare cancer mutations, 88 hypospadias, 528, 529t, 531, 535
linkage analysis and, 84, 86 infections, 534
in melanoma, 359 inguinal hernia, 531
in nasopharyngeal cancer, 164 inherited susceptibility, 527–​528
role in genotyping methods, 91 ionizing radiation, 534
Swedish Cancer Registry, 218 marijuana, 532
Systematic errors, in epidemiologic studies maternal/​gestational hypertension, 529–​530
bias, 124–​125 occupational exposure, 534
confounding, 123–​124 perinatal exposures, 528, 529t, 530–​531
experimental study, 122–​123 physical activity, 532–​533
Systemic androgen deprivation therapy (ADT), 481 prenatal exposures, 528–​530, 535
Systemic inflammatory diseases (SIDs), 433 progesterone, 528
724

724 Index

Testicular cancer (cont.) tobacco use, 614–​615, 622t

sibship size, 529t, 535 somatic events/​mutations, 613–​614
socioeconomic status, 535 subgroups, 85t, 609, 613
STIs, 534 symptoms, 609
tobacco, 532 treatment, 609
trauma, 535 tumor-​suppressor genes and, 614
vasectomy, 529t, 535 T lymphoblastic leukemia/​lymphoma, 650
somatic events, 528 Tobacco use
subgroups, 525 black (air-​cured) tobacco, 144
survival data, 230 bladder cancer and, 4, 51t, 551–​552
treatment, 525 brain cancer and, 592
Testicular dysgenesis syndrome (TDS), 528 cervical cancer and, 51t, 424–​425
Thorotrast chewing tobacco, 142
kidney cancer and, 578 cigarette epidemic, 3–​4
leukemia and, 6, 51t cohort studies, 4
liver cancer and, 6, 292 colorectal cancer and, 250–​251
lung cancer and, 345 confounding effects of, 335
Thyroid cancer, 609–​623 effects on histology, gender, race, 334
diagnosis, 609 esophageal cancer and, 4, 191–​194, 192f, 201t
EPIC cohort study, 618 Great Britain cohort studies, 4
epidemiology hepatocellular carcinoma and, 13
descriptive, 609–​612 Hodgkin lymphoma and, 637
genetic and molecular, 612–​614 HPV and, 104
incidence rates, 36t, 38t, 40t, 42t, 46, 47f, 610t, kidney cancer and, 4, 575
611f–​612f laryngeal cancer and, 4, 343
mortality rate, 44 leukemia and, 51t, 683
prognosis, 609 liver cancer and, 4, 13, 51t, 281, 282t, 283
progress, 609 lung cancer and, 3–​4, 12, 111, 121, 123–​124, 126t, 127,
proto-​oncogenes and, 613 130, 327, 333–​335
risk factors myeloid leukemia and, 4
alcohol use, 616–​617 non-​Hodgkin lymphoma and, 654, 656
anthropometric measures, 618–​620, 622t oral health and, 202
benign thyroid conditions, 621t oral/​pharyngeal cancer and, 4, 112f, 113, 140,
BMI, 619, 622t 142–​144, 147, 148
breastfeeding, 617–​618 pancreatic cancer and, 4, 51t, 312–​313, 319t, 320t
diabetes, 622 passive smoking, 12, 114, 313, 334–​335, 345, 346t
diet, 615–​616 second-​hand tobacco smoke, 334–​335, 552
fish consumption, 615–​616 skin cancer and, 361–​362, 361–​363
gallbladder disease, 621 smokeless tobacco, 144
goiter, 615, 621, 622 snuff, risk factors, 144
height, 622t testicular cancer and, 532
hormones, 618 thyroid cancer and, 614–​615, 622t
hypertension, 621–​622 tracheal cancer and, 51t
hysterectomy, 617 uterine cancer and, 51t, 104
inherited susceptibility, 613, 622t uterine cervix cancer and, 104
insulin resistance, 619 Tongue cancer, 137
iodine imbalance, 622t Tracheal cancer, 51t
iodine intake, 615 Transmission disequilibrium test (TDT), 90
ionizing radiation, 5–​6, 620, 622t Trauma
kidney cancer, 579 head/​noise, and brain cancer, 535
medical conditions, treatment, 621–​622 skin cancer and, 369
medullary cancer, 621–​622 testicular cancer and, 535
obesity, 618, 619, 622 Trichomonas Vaginalis, 502
occupational exposure, 620 Triple negative/​basal-​like type breast cancer, 62, 381
papillary/​follicular cancers, 621 Tubal ligation, ovarian cancer and, 459, 466, 468t,
physical activity, 620 472t, 473
pregnancy, 617–​618, 622t Tuberculosis
reproductive factors, 617–​618 smoking, lung cancer, and, 338
summary of, 622t 2013, death rate, 46t
725

Index 725

Tumor-​suppressor genes diethylstilbestrol and, 17

basal cell carcinoma and, 358 microbiologic exposure and, 106t
bladder cancer and, 550, 551, 563 Variability, in biomarkers, 103–​104
brain cancer and, 591 Vasectomy
breast cancer and, 384 prostate cancer and, 504, 505t
cancer origins and, 62–​72 testicular cancer and, 529t, 535
cervical cancer and, 428, 433 Vegetable consumption
colorectal cancer and, 247 bladder cancer and, 552–​553, 561
defined, 97 breast cancer and, 389
endometrial cancer and, 444–​445 bronchial cancer and, 51t
esophageal cancer and, 189, 191 colorectal cancer and, 19
Gorlin syndrome and, 358 esophageal cancer and, 51t, 196
kidney cancer and, 573 kidney cancer and, 575
liver cancer and, 280–​281, 288, 290 laryngeal cancer and, 343–​344
loss of heterozygosity and, 95 lung cancer and, 335–​336
lung cancer and, 330, 332, 345 nasopharyngeal cancer and, 168
melanoma and, 359, 361 ovarian cancer and, 462–​463, 464
nasopharyngeal cancer and, 165, 166 stomach cancer and, 222, 225
oral/​pharyngeal cancer and, 141 thyroid cancer and, 616
ovarian cancer and, 462 tracheal cancer and, 51t
prostate cancer and, 501 Vinyl chloride (VC)
roles of, 62–​67, 69, 72 ASL and, 9–​10
squamous cell carcinoma and, 359, 360 brain cancer and, 596–​597
stomach cancer and, 219, 221 occupational exposure to, 9–​10
thyroid cancer and, 614 scrotal cancer and, 6
types of, 85t Vitamin A (retinol)
Turcot’s syndrome, 591 bladder cancer and, 553
Twin studies, monozygotic vs. dizygotic, 84 esophageal cancer and, 194–​195
Tylosis, 188 kidney cancer and, 576
lung cancer and, 336
Ultraviolet radiation (UVR) oral/​pharyngeal cancer and, 145
basal cell carcinoma and, 10 stomach cancer and, 224
Hodgkin lymphoma and, 638, 640t Vitamin B12, esophageal cancer and, 195
non-​Hodgkin lymphoma and, 663 Vitamin C
skin cancer and, 10 bladder cancer and, 553
squamous cell carcinoma and, 10 breast cancer and, 388
Unclassified renal cell carcinoma, 571 esophageal cancer and, 194–​195
Urinary bladder cancer. See Bladder cancer kidney cancer and, 576
Urinary tract cancer, 8, 11 leukoplakia and, 145
Uterine cancer, 164 stomach cancer and, 223
global mortality rates, 51t Vitamin D
global trends and patterns, 51t bladder cancer and, 553
incidence rates, 46t breast cancer and, 388
mortality rates, 49f colorectal cancer and, 254
risk factors ovarian cancer and, 463
diethylstilbestrol, 404 pancreatic cancer, 313
reproductive factors, 15 prostate cancer and, 491
tobacco use and, 104 Vitamin E (alpha-​tocopherol)
bladder cancer and, 553
Vaccination breast cancer and, 388
for cervical cancer, 13–​14, 51 esophageal cancer and, 195
for HCC, 278, 289 kidney cancer and, 576
for hepatitis infections, 12, 13, 289–​290 oral/​pharyngeal cancer and, 145–​146
for HPV, 421, 430, 432–​434 pancreatic cancer and, 313
for non-​Hodgkin lymphoma, 655t prostate cancer and, 487, 489, 492–​493, 505t
non-​Hodgkin lymphoma and, 662 skin cancer and, 363
for polio, 593 stomach cancer and, 223–​224
for tuberculosis, 662 Von Hippel-​Lindau
Vaginal cancer syndrome, 85t, 572–​573, 574t, 591
726

726 Index

Weight change oral/​pharyngeal cancer, 138f, 148

colorectal cancer and, 256 pancreatic cancer, 35t, 37t, 39t
kidney cancer and, 577 stomach cancer, 35t, 39t, 39t, 50, 214f, 215f, 222
prostate cancer and, 500 thyroid cancer, 611f–​612f
Werner’s syndrome, 66, 359, 613 Women’s Health Initiative, 386
Wilms tumor syndrome, 85t World Cancer Research Fund
Wiskott-​Aldrich syndrome, 14, 654 on liver cancer, 283
Women, cancer incidence in on oral/​pharyngeal cancer, 144, 145, 146, 147
biliary tract cancer, 35t, 39t World Cancer Research Fund/​American Institute for
brain cancer, 590f Cancer Research (WCRF/​AICR), 283
breast cancer, 381–​407 World Health Organization (WHO)
cancer burden cervical abnormalities classification, 421
age standardized data, 35t–​36t, 52f global mortality data, 31
cumulative incidence, 39t–​40t nasopharyngeal cancer data, 159
global mortality rates, 49f oral/​pharyngeal cancer data, 137, 147
colorectal cancer, 18, 35t, 52f, 245f, 262
esophageal cancer, 35t, 39t, 47f, 184–​185 Xeroderma pigmentosum
Hodgkin lymphoma, 628f–​629f basal cell carcinoma and, 358
laryngeal cancer, 35t, 39t, 47f, 343f oral/​oropharyngeal cancer and, 140
leukemia, 679f skin cancer and, 367
liver cancer, 278f–​279f squamous cell carcinoma and, 359
lung cancer, 48, 54, 328, 328f–​329f
melanoma, 356f, 357f Years of life lost (YLL) measure, 45
nasopharyngeal cancer, 160f, 161f, 162f
nonmelanoma skin cancer, 47f Zinc, esophageal cancer and, 194–​195
72
728
729
730
731
732