Observational Analytical studies: 

Cohort Design
 Epidemiological study designs

Study type
Experimental Observational
Randomized  Field  Commu Descriptive Analytical
controlled  trials
‐nity
trials trials Case report, case Case control,
series, cross cohort,
sectional ecological

2
 Learning Objectives
■ To describe the design of a cohort study, 
and to distinguish it from a randomized 
trial.
■ To illustrate the cohort study design with 
two important examples.
■ To discuss some potential biases in 
cohort studies.
 Other names for Cohort
Studies:
Follow-up studies / Incidence
studies
“In a longitudinal study subjects
are followed over time with
continuous or repeated monitoring
of risk factors or health outcomes,
or both.”
Diagram of a prospective cohort study

Source: Modified from Cahn MA, Auston I, Selden CR, Pomerantz KL. Introduction to HSR, May 23, 1998. National
Information Center on Health Services Research and Health Care Technology (NICHSR), National Library of
Medicine. 1998. Available at: http://www.nlm.nih.gov/nichsr/pres/mla98/cahn/sld034.htm. Accessed July 30, 2008.

Friis, 2010
Gordis, 2014
 Cohort studies: prospective and retrospective
Prospective: onset of study in 2004 follow-up until 2014

Disease (+)
Exposed (+)
No disease (-)
Eligible
Subjects
Disease (+)
Unexposed (-)
No disease (-)

Direction of inquiry

trace back to 1994 Retrospective: onset of study in 2004

Sibai, 2004
 Cohort studies: prospective and
retrospective
 Eligible subjects: study population free from
outcome of interest (vs x-sectional studies
where study population include diseased
individuals)
 Start with assessment of exposure(s)
follow-up/trace  check for outcome(s)
 Calculate incidence rates: incidence studies
(vs. prevalence x-sectional studies)
 Potential loss to follow-up Sibai, 2004
9
 Prospective cohort studies: examples

Framingham Framingham, 6,500 middle- CVD

Study, 1948 Massachusetts aged men and
women
Nurses Health 11 US States 121,000 women CVD & cancer
Study, 1976 aged 34-59

Women’s 40 primary 93,725 CVD, CA,

Health Initiative, clinical centers postmenopausal osteoporotic
1992 in the USA women fractures
50-79 yrs
 Retrospective cohort studies: examples

Employees of Atomic Energy 14,327 men and Cancer

the Sellafield Plant in UK, women (25 yr.
British Nuclear 1947-75 retrospective)
Fluids plant,
1992
Physicians in England and 20,000 NHS Cause-specific
the UK, 1995 Wales: 1962- hospital mortality by
1979 consultants different
specialties
Chernobyl, Ukraine, 1986 250,000 Acute health
2002 individuals effect (deaths,
living within 30 impairment)
Km radius Late effects
around the (CA, congenital
reactor abnormalities)
Prospective (concurrent) cohort
Cohort identified at time of start of study, exposure and
risk factors assessed at beginning of study, outcome
assessed through follow-up as study progresses.

Retrospective (non-concurrent, historical) cohort

Cohort in the past identified at a point in the present,
exposure and risk factors identified from past records,
trace outcome through different sources (mostly used in
occupational studies) or when exposure is rare in the
community

12 Sibai, 2004
Prospective vs, retrospective cohorts

Gordis, 2014
 Single vs Multiple exposures
 Exposed vs non Exposed
 Population based
 Birth Cohorts
Types of Cohort studies

Gordis, 2014
 British Birth Cohorts
• National Survey of Health & Development (NSHD),
established in 1946
• National Child Development Study (NCDS),
established in 1958
• 1970 British Cohort Study (BCS70)
• Millennium Cohort Study (MCS), established in
2000

They have collected information on education and 
employment, family and parenting, physical and mental 
health, and social attitudes, as well as applying cognitive 
tests at various ages.

(Thalib, 2015)
 Examples – Famous Framingham study

 Begun in 1948.
 Framingham is a town about 20 miles
from Boston.
 It was thought that the characteristics of
its population (just under 30,000) would
be appropriate for such a study and
would facilitate follow-up of participants.

(Thalib, 2015)
 Framingham Study
 Residents who were between 30 and 62 years of
age.
 The rationale for using this age range was that
people younger than 30 years would generally be
unlikely to manifest the cardiovascular endpoints
being studied during the proposed 20-year follow-up
period.
 Many persons older than 62 years would already
have established coronary disease, and it would
therefore not be rewarding to study persons in this
age group for incidence of coronary disease.

(Thalib, 2015)
 Framingham Study
 The investigators sought a sample size of
5,000.
 Final sample consisted of 5,127 men and
women who were between 30 and 62 years
of age at the time of study entry and were
free of cardiovascular disease at that time.
 In this study, many “exposures” were defined,
including smoking, obesity, elevated blood
pressure, elevated cholesterol levels, low
levels of physical activity, and other factors.

(Thalib, 2015)
 Framingham Study
 New coronary events were identified by
examining the study population every 2
years and by daily surveillance of
hospitalizations at the only hospital in
Framingham.

(Thalib, 2015)
The study tested the following hypotheses
 The incidence of CHD increases with
age. It occurs earlier and more
frequently in males.
 Persons with hypertension develop
CHD at a greater rate than those who
are normotensive.

(Thalib, 2015)
The study hypotheses
Elevated blood cholesterol level is
associated with an increased risk
of CHD.
Tobacco smoking and habitual use
of alcohol are associated with an
increased incidence of CHD.

(Thalib, 2015)
Study hypotheses
 Increased physical activity is associated
with a decrease in the development of
CHD.
 An increase in body weight predisposes
a person to the development of CHD.
 An increased rate of development of
CHD occurs in patients with diabetes
mellitus.
(Thalib, 2015)
It is primarily because of the
Framingham Study, the
epidemiology of cardiovascular
disease and factors related to CHD
are so well known today.

(Thalib, 2015)
 Cohort studies: when to use
1. When there is good evidence for a relationship
between exposure and outcome (from earlier x-
sectional and case-control studies)
2. When outcome is expected to be high
3. When one is interested in examining the association of
many exposures with various outcomes
4. When there are ample funds to conduct the study

Sibai, 2004
25
Cohort studies:
Steps

1. Assembling the cohort

2. Measurement of exposure
3. Measurement of outcome
4. Analysis
Sibai, 2004
26
 Cohort studies: 1. Assembling
the cohort
Characteristics of the cohort
1. People free from outcome of interest (vs. disease)
2. People who are at risk of outcome, they may or may
not be exposed. Inclusion criteria are not linked to
exposure
3. Representativeness of the population of interest: non-
participation affects generalizability
4. Relatively static to ensure low attrition rates
27 Sibai, 2004
 Cohort studies: 2. measurement
of exposure
1. Sources interviews, lab-tests, physical
examination, records.. (criteria of reliability and
validity)

2. Frequency of assessment
People may change exposure during follow-up
time  include changes (time and reason) in data
collection and analysis

3. Components of exposure (intensity, frequency and

duration of exposure)
Sibai, 2004
28
 Cohort studies:3. measurement
of outcome
1. Sources: interviews, lab-tests, physical examination,
records.. Ensure reliability and validity: detailed
description of diagnostic criteria for identification of a case
– accuracy depends on disease under consideration, on
technique(s) used..
2. Frequency of assessment: Depends on objective of study
and available resources
3. Follow-up: Ensure completeness: Loss to follow-up
(attrition) affects validity

Sibai, 2004
29
 Cohort studies: 4. Analysis

General framework: 2x2 table

outcome

Exposure Present Absent TOTAL

Present a b a+b

Absent c d c+d

TOTAL a+c a+d a+b+c+d

Sibai, 2004
30
 Example 1: count data , A prospective study of the effect of
HBP on CVD
CVD
Sibai, 2004 Yes No Total
Present 200 800 1,000
HBP
Absent 900 8,100 9,000
Total 1,100 8,900 10,000

Overall incidence (risk) of outcome in a+c/ 1,100/10,000=11

sample a+b+c+d %
Incidence of CVD among exposed a/a+b 200/1,000=20%
Incidence of CVD among nonexposed c/c+d 900/9,000=10%

Measure(s) of association: Similar to that of X-sectional studies

Relative risk of disease – based on (a/a+b) / 20%/10%=2
incidence – among exposed vs. those (c/c+d)
not exposed RR (interpret)
(200x8100)/(900x800)
Odds ratios OR (ad) / (bc)
= 2.25 (interpret)
 Disease
Exposure Yes No Total

Present a b

Absent c d

Total Fixed total

(x-sectional +
cohort studies)
Interpretation of RR/OR: similar to all study
designs
= 1  No association
> 1  Risk factor
< 1  Protective factor, example below
 Example 2: count data
A prospective study of the effect of smoking on pre-term delivery
Pre-term delivery
No Yes Total
Present 450 50 500
smoking
Absent 1,450 50 1,500
Total 1,900 100 2,000

Overall incidence (risk) of outcome in a+c/ 100/2,000=5%

sample a+b+c+d
Incidence of PTD among exposed a/a+b 50/500=10%
Incidence of PTD among nonexposed c/c+d 50/1,500=3.3%

Measure(s) of association: Similar to that of X-sectional studies

Relative risk of disease – based on (a/a+b) / 10%/3.3%=3

incidence – among exposed vs. those not (c/c+d) (interpret)
exposed RR

Odds ratios OR (ad) / (bc) (50x1450)/(50x450)

= 3.2 (interpret)
 Example 3: count data
A cross-sectional/cohort study of the effect of smoking on
CHD

CHD
Smoking Present Absent Total Prevalence Prevalence Prevalence
(risk) (incidence) (incidence)
of disease RR OR
Nonsmoker 200 800 1000 20.0% 1.00 (ref) 1.00 (ref)
Occasional 30 100 130 23.1% 1.16 1.20
Mild 120 300 420 28.6% 1.43 1.60
Moderate 100 220 320 31.3% 1.57 1.82
Heavy 200 300 500 40.0% 2.00 2.67
Total 650 1720 2370 32.0%
 Example 4: cohort study- person time data

A hypothetical example of a cohort study of the association

between HPB and CVD mortality in 4 male subjects, follow-
up period 10 years

Group exposed to HBP Group not exposed to HBP

(D,2yrs) (D, 8yrs)

(A, 10yrs) (A, 10yrs)

In each group
n=2 study f/u period =10yrs CVD Deaths =1 Alive = 1

Examine the distribution and the timing of death

 Group exposed to HBP Group not exposed to HBP
(D,2yrs) (D, 8yrs)

(A, 10yrs) (A, 10yrs)

CVD deaths CVD deaths

Yes No Tot (count) Yes No Tot person years
Present 1 1 2 1 1 12
HBP
Absent 1 1 2 1 1 18
Total 2 2 4 2 2 30

RR or OR 1 RR (1/12)/(1/18)=1.5

Limitation of the risk and Correction for the limitation in

RR/OR Risks and RR is to take into
consideration follow-up time
= Rate and rate ratio
What is a rate?
It is a proportion with the numerator representing
the number of diseased/deaths/events/condition
and the denominator representing the number of
person-years calculated for all study subjects.

The difference between risk and rate is in the

denominator. Risk includes counts only while
rate takes into consideration the years of
observation.
 More examples
Exposed Grp (E) Non exposed Grp (NE)

D, 2yrs D, 5yrs

D, 3yrs D, 6yrs

D, 5yrs D, 8yrs
D, 7yrs
n=10, A, 10yrs n=10, A, 10yrs

In both groups N=20, D=10, A=10, f/u time =10yrs,

risk = 50% in both groups (Risk ratio = 1)

But the rates (instantaneous incidence rate, incidence density,

hazard, force of morbidity) are different
pys for E = 140 pys for NE = 168
rate for E = 10/140 = 7.1/100 py rate for NE = 10/168 = 5.9/100 py
Rate ratio (RR) = (7.1/5.9) = 1.2 (interpretation)
 Last example

Calculate the p-yrs and rate for the following cohort:

 1st subject followed up for 5 yrs and died from
CVD
 2nd subject followed up for 6 yrs and refused to
continue with study (attrition.., lost to follow-up)
 3rd subject followed up for 7 years and died from
car accident
 4th subject remained free from CVD until end of
study (10 yrs)
 5th subject had to leave country after only two yr
Total -30 person years
Interpret the answer
Rate = 1/30
 Objective: To investigate changes in mental health and other
needs, as well as clinical and diagnostic ‘caseness’, in a sample of
adolescents over a 6-month period following entry into a Young
Offenders Institution in the UK.
 Design: Prospective cohort study.
 Setting: One Young Offenders Institution between November 2006
and August 2009.
 Participants: 219 male adolescents aged 15–18 years (M=16.56;
SD=0.6) were assessed at baseline (median=4; range 0–26 days
following reception into custody) on the Salford Needs Assessment
Schedule for Adolescents (SNASA) and Kiddie Schedule for
Affective Disorders and Schizophrenia (K-SADS).
41
Applications of observational studies

Bonita et al, 2006 42

Advantages and disadvantages of
observational studies

Bonita et al, 2006 43

Ability to prove causation

Bonita et al, 2006

44
 References:
• Gordis L, (2009). Epidemiology, 5th ed. Philadelphia: Elsevier 
Saunders.
• Bonita R, Beaglehole R, Kjellstrom T, 2006.  Basic Epidemiology.  
Geneva: WHO.
• Friis R, (2010) . Epidemiology 101.  Sudbury, MA: Jones and Bartlett 
Publishers.
• Coggon, D., Rose, G. A., & Barker, D. J. P. (2003). Epidemiology for 
the uninitiated. London: BMJ Books. 
• Cohort and Case-Control Studies. In Public Health Research
Methods. Guest G, Namey EE, editors. Sage Publications;2014, p.
197-222
• Some slides were extracted with permission from presentation 
prepared by:
– Thalib L. Case Control presentation. Qatar University 2015
– Sibai Abla. Descriptive Analytic studies . A. Cohort Studies (CS). AUB 2004
– Abdul Rahim H. Cohort and Case‐control Presentation. Qatar University 2015