Introduction to

Drug Utilization Research

World Health Organization

WHO International Working Group for Drug Statistics Methodology

WHO Collaborating Centre for Drug Statistics Methodology

WHO Collaborating Centre for Drug Utilization Research and Clinical

Pharmacological Services
 WHO Library Cataloguing-in-Publication Data

Introduction to drug utilization research / WHO International Working Group for Drug
Statistics Methodology, WHO Collaborating Centre for Drug Statistics Methodology,
WHO Collaborating Centre for Drug Utilization Research and Clinical Pharmacological
Services.

1. Drug utilization 2. Research 3. Manuals I.WHO International Working Group for

Drug Statistics Methodology II.WHO Collaborating Centre for Drug
Statistics Methodology III.WHO Collaborating Centre for Drug Utilization Research
and Clinical Pharmacological Services

ISBN 92 4 156234 X (NLM classification: WB 330)

© World Health Organization 2003

All rights reserved. Publications of the World Health Organization can be obtained
from Marketing and Dissemination, World Health Organization, 20 Avenue Appia,
1211 Geneva 27, Switzerland (tel: +41 22 791 2476; fax: +41 22 791 4857;
email: [email protected]). Requests for permission to reproduce or translate WHO
publications – whether for sale or for noncommercial distribution – should be addressed
to Publications, at the above address (fax: +41 22 791 4806; email:
[email protected]).
The designations employed and the presentation of the material in this publication do not
imply the expression of any opinion whatsoever on the part of the World Health
Organization concerning the legal status of any country, territory, city or area or of its
authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on
maps represent approximate border lines for which there may not yet be full agreement.

The mention of specific companies or of certain manufacturers’ products does not imply
that they are endorsed or recommended by the World Health Organization in preference to
others of a similar nature that are not mentioned. Errors and omissions excepted, the names
of proprietary products are distinguished by initial capital letters.

The World Health Organization does not warrant that the information contained in this
publication is complete and correct and shall not be liable for any damages incurred as a
result of its use.

Printed in Oslo, Norway, 2003

3
Contents

Preface: Drug utilization research - the early work ...........................................................................6

Chapter 1: What is drug utilization research and why is it needed? ...............................................8

1.1 Definition and domains............................................................................................................8
1.2 Why drug utilization research? ................................................................................................9
1.2.1 Description of drug use patterns ....................................................................................9
1.2.2 Early signals of irrational use of drugs .......................................................................10
1.2.3 Interventions to improve drug use - follow-up ............................................................10
1.2.4 Quality control of drug use...........................................................................................10
1.3 Drug utilization studies and drug policy decisions................................................................11
1.4 General reading ......................................................................................................................12

Chapter 2: Types of drug use information ........................................................................................13

2.1 Drug-based information .........................................................................................................13
2.1.1 Level of drug use aggregation .....................................................................................13
2.1.2 Indication .....................................................................................................................13
2.1.3 Prescribed daily doses .................................................................................................14
2.2 Problem or encounter-based information ..............................................................................15
2.3 Patient information.................................................................................................................16
2.4 Prescriber information ...........................................................................................................16
2.5 Types of drug utilization study ..............................................................................................17
2.6 Drug costs ..............................................................................................................................17
2.7 General reading ......................................................................................................................18
2.8 Exercises ................................................................................................................................19

Chapter 3: Sources of data on drug utilization ................................................................................20

3.1 Large databases ......................................................................................................................20
3.2 Data from drug regulatory agencies ......................................................................................20
3.3 Supplier (distribution) data ....................................................................................................20
3.4 Practice setting data ...............................................................................................................21
3.4.1 Prescribing data ............................................................................................................21
3.4.2 Dispensing data ............................................................................................................22
3.4.3 Aggregate data .............................................................................................................22
3.4.4 Over-the-counter and pharmacist-prescribed drugs ....................................................22
3.4.5 Telephone and Internet prescribing .............................................................................22
3.5 Community setting data .........................................................................................................23
3.6 Drug use evaluation ...............................................................................................................23
3.7 General reading ......................................................................................................................24
3.8 Exercises ................................................................................................................................24

Chapter 4: Economic aspects of drug use (pharmacoeconomy) ....................................................26

4.1 Introduction ............................................................................................................................26
4.2 Cost-minimization analysis....................................................................................................26
4.3 Cost-effectiveness analysis ....................................................................................................26
4.4 Cost-utility analysis ...............................................................................................................27
4.5 Cost-benefit analysis ..............................................................................................................27
4.6 General reading ......................................................................................................................28
4.7 Exercises ................................................................................................................................28
Chapter 5: Drug classification systems..............................................................................................33
5.1 Different classification systems .............................................................................................33
5.2 The ATC classification system...............................................................................................33
5.3 Ambivalence towards an international classification system ................................................35
5.4 Implementation of the ATC/DDD methodology ...................................................................36
5.5 General reading ......................................................................................................................36
5.6 Exercises ................................................................................................................................37

Chapter 6: Drug utilization metrics and their applications ............................................................38

6.1 The concept of the defined daily dose (DDD).......................................................................38
6.2 Prescribed daily dose and consumed daily dose....................................................................39
6.3 Other units for presentation of volume..................................................................................39
6.4 Cost ......................................................................................................................................39
6.5 General reading .....................................................................................................................40
6.6 Exercises ...............................................................................................................................41

Chapter 7: Solutions to the exercises .................................................................................................74

Acknowledgements...............................................................................................................................84
 Preface:
Drug utilization research - the early work

The development of drug utilization research became clear that we need to know the answers
was sparked by initiatives taken in Northern to the following questions:
Europe and the United Kingdom in the mid- • why drugs are prescribed;
1960s (1, 2). The pioneering work of Arthur • who the prescribers are;
Engel in Sweden and Pieter Siderius in Holland • for whom the prescribers prescribe;
(3) alerted many investigators to the importance • whether patients take their medicines correctly;
of comparing drug use between different count- • what the benefits and risks of the drugs are.
ries and regions. Their demonstration of the
remarkable differences in the sales of antibiotics The ultimate goal of drug utilization research
in six European countries between 1966 and must be to assess whether drug therapy is rational
1967 inspired WHO to organize its first meeting or not. To reach this goal, methods for auditing
on «Drug consumption» in Oslo in 1969 (4). drug therapy towards rationality are necessary.
This led to the constitution of the WHO The early work did not permit detailed compa-
European Drug Utilization Research Group risons of the drug utilization data obtained from
(DURG). different countries because the source and form
The pioneers of this research understood that a of the information varied between them. To
correct interpretation of data on drug utilization overcome this difficulty, researchers in Northern
requires investigations at the patient level. It Ireland (United Kingdom), Norway and Sweden
6

Figure 1 Utilization of insulin and oral antidiabetic drugs in seven European countries from 1971-1980 expressed
in defined daily doses (DDDs) per 1000 inhabitants per day. For comparison the prescribed daily doses (PDD)
per 1000 inhabitants per day of oral antidiabetic drugs are given for Northern Ireland (UK) and Sweden for 1980
(indicated with an asterisk).
developed a new unit of measurement, initially Scandinavica, 1984, Suppl. 683:7-9.
called the agreed daily dose (5) and later the
2. Dukes MNG. Development from Crooks to the
defined daily dose (DDD) (6). This unit was
nineties. In: Auditing Drug Therapy.
defined as the average maintenance dose of Approaches towards rationality at reasonable
the drug when used on its major indication in costs. Stockholm, Swedish Pharmaceutical
adults. The first study used antidiabetic drugs Press, 1992.
as an example: it was found that the sum of the
3. Engel A, Siderius P. The consumption of drugs.
DDDs of insulin and oral antidiabetic drugs
Report on a study 1966-1967. Copenhagen,
(about 20 DDDs per1000 inhabitants per day) WHO Regional Office for Europe, 1968 (EURO
roughly corresponded to the morbidity due to 3101).
diabetes after correction for the number of pati-
ents treated with dietary regimens alone. Among 4. Consumption of drugs. Report on a symposium
in Oslo 1969. Copenhagen, WHO Regional
the first countries to adopt the DDD methodo-
Office for Europe, 1970 (EURO 3102).
logy was the former Czechoslovakia (7) and the
first comprehensive national list of DDDs was 5. Bergman U, et al. The measurement of drug
published in Norway in 1975 (8). Another consumption. Drugs for diabetes in Northern
important methodological advance was the adop- Ireland, Norway, and Sweden. European
Journal of Clinical Pharmacology, 1975,8:83-
tion of the uniform anatomical therapeutic che-
89.
mical (ATC) classification of drugs (see chapter 7
5.2). The use of standardized methodology allo- 6. Bergman U et al., eds. Studies in drug utilizati-
wed meaningful comparisons of drug use in on. Methods and applications. Copenhagen,
different countries to be made (Fig. 1). WHO Regional Office for Europe, 1979 (WHO
Regional Publications, European Series No. 8).
Drug utilization research developed quickly
during the following 30 years and soon became a 7. Stika L et al. Organization of data collection in
respectable subject for consideration at inter- Czechoslovakia. In: Bergman U et al., eds.
national congresses in pharmacology, pharmacy Studies in drug utilization. Methods and appli-
and epidemiology. Particularly rapid develop- cations. WHO Regional Office for Europe,
Copenhagen, 1979 (WHO Regional Publications
ments were seen in Australia (9) and Latin
European Series No. 8) pp.125-136.
America (10). The number of English-language
papers on the subject listed in the Cumulative 8. Baksaas Aasen I et al. Drug dose statistics, list
index medicus rose from 20 in 1973 (when the of defined daily doses for drugs registered in
term «drug utilization« first appeared) to 87 in Norway. Oslo, Norsk Medicinal Depot, 1975.
1980, 167 in 1990, and 486 in 2000.
9. Hall RC. Drug use in Australia. In: Sjöqvist F,
History has taught us that successful research Agenäs I, eds. Drug utilization studies:
in drug utilization requires multidisciplinary col- Implications for medical care. Acta Medica
laboration between clinicians, clinical pharmaco- Scandinavica, 1983, Suppl. XXX:79-80.
logists, pharmacists and epidemiologists.
10. Drug Utilization Research Group, Latin
Without the support of the prescribers, this rese-
America. Multicenter study on self-medication
arch effort will fail to reach its goal of facilita- and self-prescription in six Latin American
ting the rational use of drugs. countries. Clinical Pharmacology and
Therapeutics, 1997, 61:488-493.
References
11. Bergman U, Sjöqvist F. Measurement of drug
1. Wade O. Drug utilization studies - the first utilization in Sweden: methodological and clini-
attempts. Plenary lecture. In: Sjöqvist F, cal implications. Acta Medica Scandinavica,
Agenäs I. eds. Drug utilization studies: implica- 1984, Suppl 683:15-22.
tions for medical care. Acta Medica
 Chapter 1: What is drug utilization research
and why is it needed?

1.1. Definition and domains not. Sophisticated utilization-oriented pharma-

coepidemiology may focus on the drug (e.g.
Drug utilization; pharmacoepidemiology; dose-effect and concentration-effect relationships),
[ pharmacosurveillance; pharmacovigilance ] the prescriber (e.g. quality indices of the prescrip-
tion), or the patient (e.g. selection of drug and
• Drug utilization research was defined by dose, and comparisons of kidney function, drug
WHO in 1977 as «the marketing, distribution, metabolic phenotype/genotype, age, etc.).
prescription, and use of drugs in a society, with Drug utilization research is thus an essential
special emphasis on the resulting medical, social part of pharmacoepidemiology as it describes the
and economic consequences». Since then, a extent, nature and determinants of drug exposure.
number of other terms have come into use and it Over time, the distinction between these two
is important to understand the interrelationships terms has become less sharp, and they are some-
of the different domains. times used interchangeably. However, while
• Epidemiology has been defined as «the study drug utilization studies often employ various
of the distribution and determinants of health- sources of information that focus on drugs (e.g.
related states and events in the population, and aggregate data from wholesale and prescription
the application of this study to control of health registers) the term epidemiology implies defined
problems». populations in which drug use can be expressed
8 • Pharmacoepidemiology applies epidemiologi- in terms of incidence and prevalence (see
cal methods to studies of the clinical use of chapter 1.2.1).
drugs in populations. A modern definition of Together, drug utilization research and pharma-
pharmacoepidemiology is: «the study of the use coepidemiology may provide insights into the fol-
and effects/side-effects of drugs in large numbers lowing aspects of drug use and drug prescribing.
of people with the purpose of supporting the
rational and cost-effective use of drugs in the • Pattern of use: This covers the extent and pro-
population thereby improving health outcomes». files of drug use and the trends in drug use and
• Pharmacosurveillance and pharmacovigi- costs over time.
lance are terms used to refer to the monitoring • Quality of use: This is determined using audits
of drug safety, for example, by means of spon- to compare actual use to national prescription
taneous adverse-effect reporting systems, case- guidelines or local drug formularies.1 Indices of
control and cohort studies. quality of drug use may include the choice of
drug (compliance with recommended assort-
Pharmacoepidemiology may be drug-oriented, ment), drug cost (compliance with budgetary
emphasizing the safety and effectiveness of indi- recommendations), drug dosage (awareness of
vidual drugs or groups of drugs, or utilization- inter-individual variations in dose requirements
oriented aiming to improve the quality of drug and age-dependence), awareness of drug inter-
therapy through pedagogic (educational) inter- actions and adverse drug reactions, and the pro-
vention. Drug utilization research may also be portion of patients who are aware of or unaware
divided into descriptive and analytical studies. of the costs and benefits of the treatment.
The emphasis of the former has been to describe • Determinants of use: These include user cha-
patterns of drug utilization and to identify pro- racteristics (e.g. sociodemographic parameters
blems deserving more detailed studies. and attitudes towards drugs), prescriber characte-
Analytical studies try to link data on drug utili- ristics (e.g. speciality, education and factors
zation to figures on morbidity, outcome of treat- influencing therapeutic decisions) and drug cha-
ment and quality of care with the ultimate goal racteristics (e.g. therapeutic properties and affor-
of assessing whether drug therapy is rational or dability).
1 An audit in drug use was defined by Crooks (1979) as an examination of the way in which drugs are used in clinical practice carried out at intervals
frequent enough to maintain a generally accepted standard of prescribing.
• Outcomes of use: These are the health out price. Without a knowledge of how drugs are
comes (i.e. the benefits and adverse effects) and being prescribed and used, it is difficult to initiate
the economic consequences. a discussion on rational drug use or to suggest
measures to improve prescribing habits.
The initial focus of pharmacoepidemiology was Information on the past performance of prescri-
on the safety of individual drug products (phar- bers is the linchpin of any auditing system.
macosurveillance), but it now also includes stu- Drug utilization research in itself does not
dies of their beneficial effects. The driving force necessarily provide answers, but it contributes to
behind this development was a growing awa- rational drug use in important ways as described
reness that the health outcomes of drug use in below.
the rigorous setting of randomized clinical trials
are not necessarily the same as the health outco- 1.2.1 Description of drug
mes of drug use in everyday practice. The clini- use patterns
cal trials needed to obtain marketing authori- Drug utilization research can increase our under-
zation for new drugs involve limited numbers of standing of how drugs are being used as follows.
carefully selected patients, who are treated and
followed-up for a relatively short time in strictly • It can be used to estimate the numbers of pati-
controlled conditions. As a result, such trials do ents exposed to specified drugs within a given
not accurately reflect how drug use will affect time period. Such estimates may either refer to 9
health outcomes in everyday practice under every- all drug users, regardless of when they started to
day circumstances. Pharmacoepidemiological use the drug (prevalence), or focus on patients
studies often make useful contributions to our who started to use the drug within the selected
knowledge about effectiveness and safety, because, period (incidence).
unlike clinical trials, they assess drug effects in • It can describe the extent of use at a certain
large, heterogeneous populations of patients over moment and/or in a certain area (e.g. in a coun-
longer periods. try, region, community or hospital). Such des-
Drug utilization research also provides insight criptions are most meaningful when they form
into the efficiency of drug use, i.e. whether a part of a continuous evaluation system, i.e. when
certain drug therapy provides value for money the patterns are followed over time and trends in
and the results of such research can be used to drug use can be discerned.
help to set priorities for the rational allocation of • Researchers can estimate (e.g. on the basis of
health care budgets. epidemiological data on a disease) to what extent
drugs are properly used, overused or underused.
1.2 Why drug utilization • It can determine the pattern or profile of drug
research? use and the extent to which alternative drugs are
Description of drug use pattern; early signals being used to treat particular conditions.
of irrational use of drugs; interventions to • It can be used to compare the observed patterns
improve drug use; quality control cycle; of drug use for the treatment of a certain disease
continuous quality improvement with current recommendations or guidelines.
• It can be used in the application of quality indi-
The principal aim of drug utilization research is cators to patterns of drug utilization. An exam-
to facilitate the rational use of drugs in popu- ple is the so-called DU90% (drug utilization
lations. For the individual patient, the rational 90%), a further development of the «top-ten»
use of a drug implies the prescription of a well- list.
documented drug at an optimal dose, together The DU90% segment reflects the number of
with the correct information, at an affordable drugs that account for 90% of drug prescriptions
 and the adherence to local or national prescrip- mal practice, whether pedagogic interventions
tion guidelines in this segment. This general (education) are required or whether the guide-
indicator can be applied at different levels (e.g. lines should be reviewed in the light of actual
individual prescriber, group of prescribers, hos- practice. These hypotheses should apply to both
pitals, region or county) to obtain a rough esti- under use and over use of drugs.
mate of the quality of prescribing.
• Drug utilization data can be fed back to pre- 1.2.3 Interventions to improve
scribers. This is particularly useful when the drug use - follow-up
drug prescribing by a particular individual can Drug utilization research undertaken in the follo-
be compared with some form of «gold standard» wing ways may enable us to assess whether
or best practice, and with the average prescrip- interventions intended to improve drug use have
tions in the relevant country, region or area. had the desired impact.
• The number of case reports about a drug pro- • The effects of measures taken to ameliorate
blem or adverse effects can be related to the undesirable patterns of drug use (e.g. provision
number of patients exposed to the drug to assess of regional or local formularies, information
the potential magnitude of the problem. If it is campaigns and regulatory policies) should be
possible to detect that the reaction is more com- monitored and evaluated. The researchers
mon in a certain age group, in certain conditions should bear in mind that prescribers may have
10 or at a given dose level, improving the informa- switched to other drugs that are equally undesi-
tion on indications, contraindications and rable. These potential alternative drugs should
appropriate dosages may be sufficient to ensure be included in the survey to assess the full
safer use and avoid withdrawal of the drug from impact of the measure.
the market. • The impact of regulatory changes or changes
in insurance or reimbursement systems should
1.2.2 Early signals of irrational use be assessed using a broad survey. This is neces-
of drugs sary because the total cost to society may remain
Drug utilization research may generate hypothe- the same or may even increase if more expen-
ses that set the agenda for further investigations sive drugs are used as alternatives.
as outlined below, and thus avoid prolonged irra- • The extent to which the promotional activities
tional use of drugs. of the pharmaceutical industry and the educatio-
• Drug utilization patterns and costs between nal activities of the society affect the patterns of
different regions or at different times may be drug use should be assessed.
compared. Hypotheses can be generated to form
the basis for investigations of the reasons for, 1.2.4 Quality control of drug use
and health implications of, the differences Drug use should be controlled according to a
found. Geographical differences and changes in quality control cycle that offers a systematic
drug use over time may have medical, social and framework for continuous quality improve-
economic implications both for the individual ment. The components of such a cycle are illu-
patient and for society, and should therefore be strated on the next page.
identified, explained and, when necessary cor-
rected. After step 4, the cycle begins again with new
• The observed patterns of drug use can be com- analyses, the setting of new targets, and so on.
pared with the current recommendations and The quality control cycle can be applied at
guidelines for the treatment of a certain disease. many levels, ranging from local or regional dis-
Hypotheses can then be generated to determine cussion groups consisting of physicians, clinical
whether discrepancies represent less than opti- pharmacologists or pharmacists to national and
 Step 1. Plan. Analyse current situation to Step 2. Do. Implement the plan on a small
establish a plan for improvment (e.g. analyse scale (e.g. provide feedback on possible
current prescription patterns of individual overuse, underuse or drug misuse of
prescribers, groups of prescribers, or health individual drugs or therapeutic groups).
facilities).

Step 4. Act Revise plan or implement plan Step 3. Check Check to see if expected
on large scale (e.g. guide national imple- results are obtained (e.g. evaluate whether
mentation of plan). prescription patterns really improve).

international initiatives. An important technique and validated from 1992-1994. Since then,
that can be used in conjunction with this cycle annual reviews of drug utilization have been
is benchmarking. By comparing drug utiliza- used to provide background information for
tion data from different localities, it is often decisions on regulatory and reimbursement poli-
possible to detect substantial differences that cies in Estonia; two examples are described
require further evaluation, which may then lead below.
to the identification and promotion of best prac- If physicians have high rates of inappropriate
tice. Such comparisons will be accurate and prescribing, drug regulatory authorities can 11
truthful provided that the data are collected and require educational intervention or impose
aggregated in a standardized, uniform way (see restrictions on specific drugs or on practitioners.
chapter 5). In Estonia, it was decided to stop the import and
use of some hazardous products, such as phena-
1.3 Drug utilization studies and cetine, older sulphonamides and pyrazolones,
drug policy decisions after clarifying and explaining the reasons for
Many of the questions asked in drug utilization this in the national Drug information bulletin,
research and the answers obtained are important which is distributed free by the drug regulatory
for initiating and modifying a rational drug poli- authority to all prescribers in Estonia.
cy at both national and local levels. Two suc- In planning the reimbursement policies, the
cessful examples of the use of such research are total volume of drug use in DDDs was monito-
given below. red carefully. During the 1990s, the use of pre-
scription-only medicines measured as number of
Drug use in Estonia DDDs per capita was less than one third of that
An important reason for undertaking studies of reported from the Nordic countries. This proved
drug use in Estonia after its independence was to be the result of under-treatment of certain
the need to make decisions on drug policy. At chronic diseases (i.e. hypertension and schizo-
the time, no information was available in the phrenia), and therefore the decision was to incre-
country on which drugs were used (sold), or on ase the availability and use of cardiovascular and
the quantities and there was therefore no rationale neuroleptic drugs. Thus, the national drug use
for regulating the drug market. Moreover, in the surveys in Estonia have been used to monitor the
absence of any feedback system it was impos- impact of drug regulatory activities as well as to
sible to gauge the impact of possible future follow the increase in drug expenditure.
interventions. A national drug classification Because data on drug use are only part of the
system was therefore developed for Estonia, and background material relevant to the discussions
a reporting system from wholesalers, based on and decisions on therapeutic strategies - at both
this classification, was implemented, checked the local and national levels - it is difficult to
 evaluate the specific influence of drug utilization rent social-class districts in the catchment areas
research on developments in drug policies. It is, of 11 health centres.1
however, reasonable to assume that such studies
have contributed to a more rational use of drugs 1.4 General reading
in Estonia.1 Bergman U et al. Drug utilization 90% - a sim-
ple method for assessing the quality of drug pre-
Drug use in Latin America scribing. European Journal of Clinical
The second example is the successful work within Pharmacology, 1998, 54:113-118.
the Latin American DURG, in association with Crooks J. Methods of audit in drug use. In:
the WHO Collaborating Centre of Pharmaco- Duchene-Marulla ZP, ed. Advances in pharma-
epidemiology in Barcelona, Spain. cology and therapeutics. Proceedings of 7th
In September 1991, health professionals from International Congress of Pharmacology, Paris,
Spain and eight Latin American countries met in 1978. Oxford, Pergamon Press, 1979:189-195.
Barcelona for the «First Meeting of Latin Diogène E et al. The Cuban experience in
American Groups for Drug Epidemiology». It focusing pharmaceuticals policy to health popu-
was made clear that in most of the countries lation needs: initial results of the National
taking part, data on drug utilization were scarce Pharmacoepidemiology Network (1996-2001).
and fragmentary. Some national drug regulatory European Journal of Clinical Pharmacology,
12 authorities had no access to either quantitative or 2002, in press.
qualitative data on drug consumption and reali- Drug Utilization Research Group, Latin
zed that information on patterns of drug utili- America. Multicenter study on self-medication
zation would be useful for designing drug policy and self-prescription in six Latin American
and educational programmes about drugs. countries. Clinical Pharmacology and
It was agreed at this meeting to set up a Latin Therapeutics, 1997, 61:488-493.
American network (later called DURG-LA), Dukes MNG, ed. Drug Utilization Studies:
with the following aims: Methods and Uses. Copenhagen, WHO Regional
– to promote drug utilization research in Latin Office for Europe, 1993 (WHO Regional
American countries; Publications European Series No. 45)
– to exchange experiences and information Einarson TR, Bergman U, Wiholm BE.
between the participating groups; Principles and practice of pharmacoepidemiolo-
– to use the knowledge generated to give techni- gy. In: Avery’s Drug Treatment, 4th ed. Adis
cal advice to drug regulatory authorities and to International:371-392.
guide teaching of pharmacology; Figueras A et al. Health need, drug registration
– to write and disseminate information aimed at and control in less developed countries - the
improving drug use, and Peruvian case. Pharmacoepidemiology and Drug
– to participate in the training of health pro- Safety, 2001, 10:1-2.
fessionals in pharmacoepidemiology and thera- Laporte JR, Porta M, Capella D. Drug utiliza-
peutics. tion studies: a tool for determining the effecti-
veness of drug use. British Journal of Clinical
Seven further DURG-LA meetings have been Pharmacology, 1983, 16:301-304.
held over the subsequent ten years to promote McGavock H. Handbook of drug use research
drug utilization research. Part of the initial core methodology 1st ed. Newcastle upon Tyne,
group participated in a first multicentre study in United Kingdom Drug Utilization Research
six Latin American countries to examine self- Group, 2000.
medication and self-prescription. The study was Strom BL. Pharmacoepidemiology, 3rd ed.
carried out in a sample of pharmacies from diffe- New York, J Wiley, 2000.
1 The information about DURG-LA was provided in a personal communication by Dr Albert Figueras and Professor Joan-Ramon Laporte,
Barcelona, Spain.
Chapter 2: Types of drug use information

Different types of drug use information are 2.1.2 Indication

required depending on the problem being exami- For drugs with multiple indications, it will usually
ned. These include information about the overall be important to divide data on use according to
use of drugs, drug groups, individual generic indication to allow a correct interpretation of the
compounds or specific products. Often, infor- overall trends. An example is the relative use of
mation about the condition being treated, the drug groups in treating hypertension. The over-
patient and the prescriber is also required. In all data might suggest that the relative use of
addition, data on drug costs will be important in diuretics is comparable to that of ACE inhibitors
ensuring that drugs are used efficiently and eco- and higher than the use of calcium channel bloc-
nomically. These types of drug information are kers (column A in Table 1). However, analysis
described in detail below, together with exam- of the data according to indication may reveal
ples to illustrate the ways in which the informa- that 75% of ACE inhibitors are used to treat
tion can be used to promote the rational use of hypertension whereas only 43% of diuretics are
drugs. used for this indication (most of the high-ceiling
diuretics used are for treating heart failure). The
2.1 Drug-based information picture that emerges of the use of the two drug
Knowledge of the trends in total drug use may groups in the treatment of heart failure is mar-
be useful, but more detailed information invol- kedly altered when use according to indication is
ving aggregation of data on drug use at various taken into account (column B of Table 1). 13
levels, and information on indications, doses and
dosage regimens is usually necessary to answer
Table 1 Relative use of drug groups in the treatment
clinically important questions. of hypertension in Australia in 1998a

2.1.1 Level of drug use aggregation

Drug group Ab Bc Cd
The level at which data on drug use are aggrega-
ted will depend on the question being asked. ACE inhibitors (C09A) 31.80 36.6 34.8
For example, the question might concern the Calcium channel blockers
relative use of drug groups in the treatment of (C08C) 24.50 28 26.7
hypertension. It would then be appropriate to Diuretics (C03) 29.60 19.4 15.9
aggregate data on diuretics, beta-blockers and Beta-blockers
angiotensin-converting enzyme (ACE) inhibitors,
(C07AA, C07AB) 11.20 11.5 15.7
etc. If, however, the question concerns the rela-
tive use of beta-blockers in hypertension, data at ATII antagonists
the substance (generic drug) level would be nee- (C09CA) 3.00 4.6 6.9
ded. Information on the relative scale of use of
individual products will sometimes be required,
for example to find the market leader or to Source: Australian Drug Utilization
assess the relative use of generic versus branded Subcommittee and BEACH Survey April-
or innovator products. Information down to the December 1998, Sydney University, GPSCU
level of dose strength will be necessary, for 1999.
example, to determine whether there is a trend
a Values are the use of the drug group expressed as a percenta
towards use of higher strengths of antibiotics, or
ge of the total use for these drug groups.
to determine the relative use of strengths of anti- b Based on total use.
depressants to assess whether they are being c Adjusted for the percentage of total use of each group for the
used at effective doses. treatment of hypertension.
d Relative prescribing of these drug groups in hypertension
community practice patient encounters.
 Another example of a situation in which the when making such comparisons. The PDDs dif-
indication is important is antibiotic utilization. fer between countries and ethnic groups, and
In determining whether the use of a particular even between areas or health facilities within the
antibiotic, for example, amoxicillin, is rational, it same country. The PDD will also often differ
will usually be necessary to know what infecti- for different indications of the same drug, so it
ons or problems it is being used to treat. It will sometimes be necessary to reach this level
would therefore be necessary to break down data of detail to interpret overall use data.
on amoxicillin use into indications and compare Data on the use of tricyclic antidepressants
these uses with the appropriate guidelines. If it and selective serotonin reuptake inhibitors
were found that there was substantial use of (SSRIs) in Australia are shown in Table 2 as
amoxicillin to treat acute sore throat, for exam- both DDDs and prescription volumes.
ple, this finding would indicate a problem that The two metrics give different results for the
needed to be addressed. This is because a nar- relative use of the two groups of antidepressant
row-spectrum agent (or no drug) would be a drugs because of the different relationship bet-
more appropriate treatment for a sore throat, and ween the PDDs and the DDDs for the two drug
if amoxicillin is used to treat mononucleosis, groups. On average, the PDD is lower than the
which can present as a sore throat, there is a DDD for the tricyclics and higher for the SSRIs.
high incidence of rash. In this case, knowledge of the PDDs is necessary
14 for clinical interpretation of the data.
2.1.3 Prescribed daily doses The DDD per 1000 inhabitants per day is
The prescribed daily dose (PDD) is the average often used to derive a rough estimate of the pre-
daily dose prescribed, as obtained from a repre- valence of use in the population being studied,
sentative sample of prescriptions. The use of and for chronic diseases it may even be used to
DDD per 1000 inhabitants per day allows aggre- assess the prevalence of a disease when the drug
gation of data across drug groups and compari- is prescribed for a single indication. Such esti-
sons between countries, regions and health faci- mates are valid only if the DDDs and the PDDs
lities. However, the DDD metric may not reflect are similar.
the actual PDDs, and this needs to be considered

Table 2 Relative use of antidepressants in Australia in 1998

Prescription % of total DDD/1000 % of total DDD/1000

volume prescription population/day population/day
(millions)a volume

Tricyclics
(N06AA) 3.53 48.82 8.40 28.09
SSRI
(N06AB) 3.09 42.74 17.20 57.53
Moclobemide
(N06AG02) 0.61 8.44 4.30 14.38

Total 7.23 100.00 29.90 100.00

Source: Australian Drug Utilization Subcommittee, Department of Health & Aged Care, Commonwealth of
Australia, http://www.health.gov.au:80/haf/docs/asm.htm
2.2 Problem or encounter-based sistent. This consistency between data using two
information different approaches (i.e. drug and problem-
based) gives confidence in the result.
Reason for the encounter (the problem); Other questions that might be addressed using
drug treatment versus non-drug treatment; a problem-based approach include the following:
other problems managed; severity of the pro-
blem managed; new or continuing presenta- • Does the severity of hypertension influence the
tion; duration of consultation; medications choice of single or combination therapy?
prescribed for the problem; how the medica- • Is the management of newly-presenting pati-
tions were supplied; other medications pre- ents different to that of patients already receiving
scribed treatment?
• Are there likely to be any drug interactions
with co-prescribed treatments?
Rather than asking how a particular group of
• Is the choice of drug influenced by evidence-
drugs is used, it may be useful to address the
based outcome data?
question of how a particular problem (e.g. sore
throat, hypertension or gastric ulcer) is managed.
For some diseases it may be important to study
The different types of information that may be
the relative use of drug treatment and other the-
required are listed in the box above.
As an example, consider how problem-based
rapeutic approaches to map out and understand 15
pharmacotherapeutic traditions and other thera-
information about the management of hyperten-
peutic approaches. As an example, drug utiliza-
sion might be used. Initially, concordance with
tion research in Estonia has shown that there was
guidelines for drug treatment or non-drug mana-
a reciprocal relationship between the use of hor-
gement of blood pressure and other risk factors
monal contraceptives and the abortion rate
might be assessed. Where drug treatment is
from1989-1997 (Fig. 2).
used, the proportion of patients treated with each
of the drug groups gives an overall picture of
Another example was the excessive use of ulcer
management (column C of Table 1). This is
surgery in Estonia compared to Sweden during
more direct information on how hypertension is
the Soviet era. This was because of the difficul-
managed than that provided by assessing the
ties of obtaining modern anti-ulcer drugs in
overall use of the different drug groups as dis-
Estonia at that time (Fig. 3).
cussed above. In the example shown in Table 1,
the data in columns B and C are reasonably con-
(per 10 000 population

Use of hormonal
Nr of abortions

contraceptives
(DDD/1000/day

Figure 2 Abortion rate and use of hormonal contraceptives in Estonia in 1989-1997.

Source: Kiivet R. Drug utilization studies as support to decisions in drug policy in Estonia. [MD Thesis]
Stockholm, Karolinska Institutet, 1999.
 80 20

60 16
12
40
8
20 4
0 0
1993 1995

anti-ulcer drugs, Estonia surgery, Estonia

anti-ulcer drugs, Stockholm surgery, Stockholm

Figure 3 Treatment of ulcer disease in Estonia and in Stockholm County 1993-1995.

Source: Kiivet R. Drug utilization studies as support to decisions in drug policy in Estonia.
[MD Thesis] Stockholm, Karolinska Institutet, 1999.

2.3 Patient information 2.4 Prescriber information

Age; gender; ethnicity; co-morbidities; Demographic information - age, gender,
16 [ knowledge; beliefs and perceptions ] medical school, years in practice; type of
practice (e.g. specialist or family, rural or
Information on demographic factors and other urban); practice size; patient mix; knowledge
details about the patient will often be useful. about drugs; factors driving prescribing
For example, the age distribution of patients may behaviour
be of critical importance, to assess the likelihood
of severe adverse effects with nonsteroidal anti- The prescriber plays a critical role in determi-
inflammatory drugs (NSAIDs), or whether the ning drug use. Claims have even been made that
drug is being used to treat patients in an age the differences between doctors are greater than
group different from that in which the clinical those between patients and that variations in
trials were performed. The co-morbidities of the drug prescribing behaviour often lack rational
patient group may be important in determining explanations. Dissecting the factors that deter-
the choice of treatment and predicting possible mine prescribing behaviour is therefore often
adverse effects. For instance, in the manage- central to understanding how and why drugs are
ment of hypertension, beta-blockers should not prescribed. Some questions that might be
be used to treat patients with asthma, and ACE addressed using prescriber information include
inhibitors are the preferred treatment in patients the following:
with heart failure. • Are prescribing profiles influenced by the
Qualitative information relating to the know- prescriber’s medical education?
ledge, beliefs and perceptions of patients and • Do the prescribing profiles of specialists differ
their attitudes to drugs will be important in some from those of family practitioners?
cases, for example in assessing the pressures put • Does the age or gender of the prescriber influ-
by patients on their doctors to prescribe antibio- ence the prescribing profile?
tics, or in designing consumer information and • Are there differences in prescribing behaviour
education programmes. between urban and rural practices or between
small and large practices? Do these variations
indicate a need to target education to particular
sectors?
• Who are those prescribers who rapidly adopt often obtained from repeated cross-sectional sur-
recently released drugs? veys (e.g. IMS (Intercontinental Medical
• In assessing rational use of medicines by a Statistics) practice-based data are of this type).
practitioner, has the practice mix been taken into Data collection is continuous, but the practitio-
account? ners surveyed, and therefore the patients, are
• Can the factors that determine and change pre- continually changing. Such data give informa-
scribing behaviour be identified? tion about overall trends, but not about prescri-
bing trends for individual practitioners or practices.
2.5 Types of drug utilization
study Continuous longitudinal studies
Drug utilization studies can be targeted towards In some cases continuous longitudinal data at the
any of the following links in the drug-use chain: individual practitioner and patient level can be
obtained. Claims databases are often able to
– the systems and structures surrounding drug follow individual patients using a unique (but
use (e.g. how drugs are ordered, delivered and anonymous) identifier. These data can provide
administered in a hospital or health care facility); information about concordance with treatment
– the processes of drug use (e.g. what based on the period between prescriptions, co-
drugs are used and how they are used and prescribing, duration of treatment, PDDs and so
does their use comply with the relevant on. As electronic prescribing becomes more 17
criteria, guidelines or restrictions); and common, databases are being developed to pro-
– the outcomes of drug use (e.g. efficacy, vide continuous longitudinal data comprising
adverse drug reactions and the use of resour- full medical and prescribing information at the
ces such as drugs, laboratory tests, hospital individual patient level. Such databases are very
beds or procedures). powerful, and can address a range of issues
including reasons for changes in therapy, adverse
Cross-sectional studies effects and health outcomes.
Cross-sectional data provide a «snapshot» of
drug use at a particular time (e.g. over a year, a
month or a day). Such studies might be used for
2.6 Drug costs
making comparisons with similar data collected
over the same period in a different country, Total drug costs; cost per prescription; cost
health facility or ward, and could be drug-, pro- per treatment day, month or year; cost per
blem-, indication, prescriber- or patient-based. defined daily dose (DDD); cost per prescri-
Alternatively, a cross-sectional study can be car- bed daily dose (PDD); cost as a proportion
ried out before and after an educational or other of gross national product; cost as a propor-
intervention. Studies can simply measure drug tion of total health costs; cost as a propor-
use, or can be criterion-based to assess drug use tion of average income; net cost per health
in relation to guidelines or restrictions. outcome (cost-effectiveness ratio); net cost
per quality adjusted life-year (cost-utility-
Longitudinal studies ratio)
Public health authorities are often interested in
trends in drug use, and longitudinal data are Data on drug costs will always be important in
required for this purpose. Drug-based longi- managing policy related to drug supply, pricing
tudinal data can be on total drug use as obtained and use. Numerous cost metrics can be used and
through a claims database, or the data may be some of these are shown in the box above. For
based on a statistically valid sample of pharma- example, the cost per DDD can usually be used
cies or medical practices. Longitudinal data are to compare the costs of two formulations of the
 same drug. However, it is usually inappropriate ced a marked increase in the cost of anti-psycho-
to use this metric to compare the costs of diffe- tic drugs over the last 5-10 years; the data on use
rent drugs or drug groups as the relationship bet- and cost for these drugs in Australia are illustra-
ween DDD and PDD may vary. ted in Fig. 4.
Estimates of the costs at various levels and
using data aggregated in various ways will be In Australia, there has been little increase in the
required, depending on the circumstances and overall volume of use of antipsychotic drugs,
the perspective taken. A government perspective and the cost increase has been driven by the
might require information on drug costs and cost transfer from the cheap ‘classical’ agents to the
offsets to government to be collected, whereas a much more expensive ‘atypical’ drugs such as
societal perspective would require both govern- clozapine, olanzapine and risperidone resulting
ment and non-government (private sector) costs in an increase in the average cost per prescrip-
and cost offsets to be determined. A patient per- tion. In contrast, both the prescription volume of
spective will be appropriate if questions about antidepressant drugs and the average cost per
affordability and accessibility are being asked. prescription have increased over the same period,
Costs may be determined at government, health due to an ‘add-on’ prescribing effect of the more
facility, hospital, health maintenance organiza- expensive SSRIs.
tion or other levels within the health sector.
18 Costs will often need to be broken down 2.7 General reading
according to drug group or therapeutic area to Einarsson TR, Bergman U, Wiholm BE.
determine, for example, the reason for an increase Principles and practice of pharmacoepidemiology.
in drug costs. For instance, the introduction of In: Speight TM, Holford NH, eds. Avery’s Drug
new, expensive anti-cancer agents may be found Treatment. Place, Adis International, 1999:371-
to be driving the increases in drug costs in a hos- 392.
pital. Changes in drug costs can result from
changes in prescription volumes, quantity per Lee D, Bergman U. Studies of drug utilization.
prescription or in the average cost per prescripti- In: Strom B. ed. Pharmacoepidemiology, 3rd ed.
on. For example, most countries have experien- Chichester, J Wiley, 2000:463-481.

10 100
community use
govt cost with clozapine
8 80
DDs/1000/Day

$ million

6 60

use
4 40

2 20
cost
0 0
1990 1991 1992 1993 1994 1995 1996 1997 1998

Figure 4 Antipsychotic drugs - use and cost trends in Australia

2.8 Exercises 4. Antidepressant use
3. Amoxicillin The use of antidepressant drugs (in DDDs per
You note that amoxicillin use expressed as 1000 population per day) and their costs have
DDDs per1000 population per day has been increasing for at least the last five years.
increased over the last two years. What types What types of data would you need to deter-
of drug utilization data would you need to mine the reasons for the change and whether it
evaluate the possible reasons for this? has resulted in positive or negative health out-
comes?

19
 Chapter 3: Sources of data on drug utilization

Drug-use chain; large databases; other sour- drug distribution chain, pharmaceutical and
[ ces; drug use evaluation; pharmacoeconomics ] medical billing or samples of prescriptions. The
databases may be international, national or local
The drug-use chain includes the processes of in scope. They may be diagnosis-linked or non-
drug acquisition, storage, distribution, prescri- diagnosis-linked. Diagnosis-linked data enable
bing, patient compliance and the review of out- drug use to be analysed according to patient cha-
come of treatment. Each of these events is an racteristics, therapeutic groups, diseases or con-
important aspect of drug utilization, and most ditions and, in the best of cases, clinical out-
countries have regulations to cover these aspects. come. A useful analysis requires an understand-
Data are collected, or are available, at national, ing of the sources and organization of the data.
regional and local health facility or household
level and may be derived from quantitative or 3.2 Data from drug regulatory
qualitative studies. Quantitative data may be agencies
used to describe the present situation and the
Drug registration; drug importation
trends in drug prescribing and drug use at vari- [ ]
ous levels of the health care system.
Drug regulatory agencies have the legal respon-
Quantitative data may be routinely collected data
sibility of ensuring the availability of safe, effi-
or obtained from surveys. Qualitative studies
20 assess the appropriateness of drug utilization and
cacious and good-quality drugs in their country.
They are thus the repositories of data on which
generally link prescribing data to reasons (indi-
drugs have been registered for use, withdrawn or
cations) for prescribing. Such studies have been
banned within a country. Regulatory agencies
referred to as «drug utilization review» or «drug
also have inspection and enforcement functions,
utilization evaluation». The process is one of a
and are responsible for supervising the importa-
«therapeutic audit» based on defined criteria and
tion of drugs and for the issuance of permits for
is intended to improve the quality of therapeutic
drug registration.
care. There is an increasing interest in the evalu-
It is possible, therefore, to obtain data on the
ation of the economic impact of clinical care and
number of drugs registered in a country from
medical technology. This has evolved into a dis-
such agencies. Where the agency issues import
cipline dedicated to the study of how pharma-
permits and supervises drug importation, data on
cotherapeutic methods influence resource utili-
product type (i.e. generic or branded), volume,
zation in health care known as pharmacoeco-
port of origin, country of manufacture, batch
nomics (see chapter 4).
number and expiry date may be collected.
The sources of drug utilization data vary from
Where the data reflect total national imports,
country to country depending on the level of
estimates of quantities of drugs in circulation
sophistication of record keeping, data collection,
can be obtained for defined periods and for
analysis and reporting and the operational consi-
various therapeutic groups
derations of the health care system.
It may be difficult to obtain true estimates if
documentation is incomplete and not all trans-
3.1 Large databases
actions are recorded. Information on smuggled
The increasing interest in efficient use of health
goods or goods entering the country through ille-
care resources has resulted in the establishment
gal routes will not be captured by these data.
of computer databases for studies on drug utili-
zation. Some of the databases can generate sta-
tistics for patterns of drug utilization and adverse
3.3 Supplier (distribution) data
drug reactions. Data may be collected on drug Drug importation; local manufacture; cus-
sales, drug movement at various levels of the [ toms service ]
Data on suppliers may be obtained from drug motivate health care providers to adhere to esta-
importers, wholesalers or local manufacturers. blished health care standards.
In countries where permits or licences are
required from drug regulatory authorities and 3.4.1 Prescribing data
ministries of health before importation of drugs, Prescribing data are usually extracted from out-
data may be available from such sources. patient and inpatient prescription forms. Such
Customs services, in the process of clearing data may be easily retrieved where records are
imports from the ports of entry, may collect data computerized and computerized data also facili-
on drugs. However, the codes used by customs tate trend analysis. In the absence of electronic
services are not detailed enough to capture all databases, prescribing data are usually extracted
relevant information. National agencies respon- from patient records or from patient intercept
sible for the collection of excise duty can also studies or retrieved at dispensing points.
provide information on the volume of production Information that may be obtained from pre-
and on distribution of drugs from local manufac- scriptions includes patient demography, drug
turers. name, dosage form, strength, dose, frequency of
Data from these sources can generally be used administration and duration of treatment. Where
to describe total quantities of specific drugs or diagnoses are noted on prescriptions, and parti-
drug groups, origins of supplies and type (i.e. cularly for inpatient prescription, it is possible to
branded or generic). link drug use to indications. Trends in utilization 21
In the absence of a national mechanism for the for specific drugs and diseases can also be esta-
direct capture of data on drug production or blished. As an example, inpatient data may pro-
importation, wholesalers become an important vide a link to empirical treatment of infections as
source of information on drug acquisition. Such opposed to treatment based on microbiological
data are reliable insofar as wholesalers are the assessment. This may be achieved by extracting
only legal entity able to import drugs. In some relevant data from the patient records, but requi-
countries, medical, dental and veterinary practiti- res that the records be of good quality.
oners, as well as pharmacists, can import phar- Prescriptions are a good source of information
maceutical products. It is usually very difficult for determining some of the indicators of drug
to collect comprehensive data from such sources use recommended by WHO including the:
even if there are regulatory requirements about
submitting reports. Public sector procurement – average number of drugs per prescription
practices, however, have reasonable documen- (encounter);
tation but provide data only on that sector. – percentage of drugs prescribed by generic
name;
Practice setting data – percentage of encounters resulting in prescrip-
tion of an antibiotic;
Prescribing data; dispensing data; drug use – percentage of encounters resulting in prescrip-
[ indicators; facility data (aggregate) ] tion of an injection;
– percentage of drugs prescribed from essential
Data from health facilities may be used to evalu- drugs list or formulary, and
ate specific aspects of health provision and drug – average drug cost per encounter.
use and to generate indicators that provide infor-
Prescribing data allow the determination of the
mation on prescribing habits and aspects of pati-
PDD which may differ from the DDD. While
ent care. These indicators can be used to deter-
the DDD is based on the dosages approved in
mine where drug use problems exist, provide a
standard product characteristics with clinical out-
mechanism for monitoring and supervision and
come data from controlled clinical trials, the
 PDD is variable and dependent on factors such can be used to obtain information on various
as severity of illness, body weight, interethnic aspects of drug use including:
differences in drug metabolism and the prescri- – the cost of individual drugs and classes of
bing culture of the health provider. Using DDDs drug;
enables comparison to be made between drug – the most frequently or infrequently used drugs;
groups as the influences of prescribing culture – the most expensive drugs;
and available dosage strengths are eliminated. – the per capita consumption of specific pro-
In some countries, it is a legal requirement ducts;
that prescriptions dispensed by pharmacies and – comparisons of two or more drugs used for the
drug outlets are kept for a minimum period before same indication;
disposal. Where these regulations are adhered – the prevalence of adverse drug reactions;
to, prescription data may be obtainable from – the prevalence of medication errors; and
pharmacies. However, in many developing – the percentage of the budget spent on specific
countries the rule is not generally followed. In drugs or classes of drug.
countries where computerized records of prescri-
bing data are kept, they may be readily retriev- Aggregate data are often useful for comparing
able depending on the depth of the database. the utilization of a particular drug to that of
other drugs and to utilization in other hospitals,
22 3.4.2 Dispensing data regions or countries.
Drug dispensing is a process that ends with a
client leaving a drug outlet with a defined quan- 3.4.4 Over-the-counter and
tity of medication(s) and instructions on how to pharmacist-prescribed drugs
use it (them). The quantity of drugs dispensed Pharmacists and other drug outlet managers may
depends on their availability. Thus information prescribe over-the-counter preparations or phar-
available from dispensers may include: macist-prepared drugs that do not require pre-
– drug(s) prescribed; scription by a physician. Data on such medica-
– dose(s) prescribed; tions may be difficult to obtain especially in
– average number of items per prescription; environments with weak drug regulation and
– percentage of items prescribed that were actu - poor record keeping, but when such information
ally supplied (an indicator of availability); is available from stock or dispensing records, it
– percentage of drugs adequately labelled; broadens the understanding of drug utilization
– quantity of medications dispensed; and patterns.
– cost of each item or prescription.
3.4.5. Telephone and Internet
These data may be obtained from records kept at prescribing
the drug outlet either in electronic or manual Physicians in certain countries may prescribe
form. over the telephone. Prescribing and dispensing
using the Internet also occurs, especially in deve-
3.4.3 Aggregate data loped countries. Most Internet prescriptions are
A number of data sources within the health faci- for nutritional supplements and herbal preparati-
lity or hospital setting can provide aggregate ons. However, as exemplified by sildenafil
data on drug utilization. These sources include (Viagra®), other medicines are also increasingly
procurement records, warehouse drug records, being sold on the Internet. Innovative ways have
pharmacy stock and dispensing records, medica- to be devised to collect information on this type
tion error records, adverse drug reaction records of transaction.
and patient medical records. These data sources
3.5 Community setting data Drug use evaluation can assess the actual pro-
cess of administration or dispensing of a medica-
Household survey; compliance (adherence tion (including appropriate indications, drug
[ to treatment); drug utilization ] selection, dose, route of administration, duration
of treatment and drug interactions) and also the
The drugs available in households have either outcomes of treatment (e.g. cured disease condi-
been prescribed or dispensed at health facilities, tions or decreased levels of a clinical parameter).
purchased at a pharmacy (with or without a pre- The objectives of drug use evaluation include:
scription) or are over-the-counter medications.
The drugs may be for the treatment of a current – ensuring that drug therapy meets current stan--
illness or are left over from a previous illness. It dards of care
is not uncommon for patients to adhere poorly to – controlling drug cost;
the instructions given for taking their dispensed – preventing problems related to medication;
medicines. Thus dispensing data and utilization – evaluating the effectiveness of drug therapy; and
data may not be equivalent because they have – identification of areas of practice that require
not been corrected for non-compliance. further education of practitioners.
Drug utilization by outpatients is best assessed
by performing household surveys, counting left- The problems to be addressed by drug use evalu-
ation may be identified from any of the data des-
over pills or using special devices that allow
cribed in section 3.4 (including prescription indi-
23
electronic counting of the number of times a par-
ticular drug is administered. Drug utilization by cators, dispensing data and aggregate data). The
inpatients can be determined by reviewing treat- main source of data for drug use evaluation is
ment sheets or orders. the patient records. An identifiable authoritative
For both outpatients and inpatients, the data group, such as the drugs and therapeutic com-
on the utilization of a particular drug can be mittee, usually carries out reviews of drug use in
aggregated for a defined population in DDDs. a hospital or health facility. This group has the
Using DDDs has the advantage of allowing com- responsibility for drawing up the guidelines, cri-
parison for example between inpatients and out- teria, indicators and thresholds for the evaluati-
patients. Data on various dosage forms and on. Drug use evaluation may be based on data
generic equivalents of the same medication can collected prospectively (as the drug is being dis-
also be aggregated. pensed or administered) or retrospectively (based
on chart reviews or other data sources).
3.6 Drug use evaluation
• Typical criteria reviewed in prospective studies
Drugs and therapeutic committee; prospec- include the following
tive evaluation; retrospective evaluation; – indications;
criteria setting – drug selection;
– doses prescribed;
Drug use evaluation, sometimes referred to as – dosage form and route of administration;
drug utilization review, is a system of continu- – duration of therapy;
ous, systematic, criteria-based drug evaluation – costs;
that ensures the appropriate use of drugs. It is a – therapeutic duplication;
method of obtaining information to identify pro- – quantity dispensed;
blems related to drug use and if properly develo- – contraindications;
ped, it also provides a means of correcting the – therapeutic outcome
problem and thereby contributes to rational drug – adverse drug reactions; and
therapy. – drug interactions.
 • In retrospective studies, the criteria source that could help you understand the situa-
reviewed include: tion, and (2) some possible advantages and/or
– evaluation of indications; limitations of each of the sources of data you
– monitoring use of high-cost medicines; have listed.
– comparison of prescribing between physicians;
– cost to patient; When evaluating the advantages and
– adverse drug reactions; and limitations of the data, consider the ans-
– drug interactions. wers to the following questions:
• How relevant are the data for
It is possible to incorporate some of the above learning about antibiotics?
criteria into databases thus allowing drug experts • How easy is it to collect these
to evaluate any items that do not meet establis- types of data in your country?
hed criteria. For meaningful results to be obtai- • How much will it cost to collect
ned from drug use evaluation a reasonable num- and process the data and how
ber of records need to be assessed. A minimum long will it take?
of 50 to 75 records per health care facility is • How reliable are these data?
considered adequate. However, the number of
records sampled would depend on the size of the For example, from data from previous
24 facility and the number of prescribers.
surveys, we might obtain the following
useful information: historical utilization
3.7 General reading
rates by facility or geographical area, and
How to investigate drug use in health facilities:
possibly utilization by type of antibiotic,
Selected drug use indicators. Geneva,World
health problem or age. The advantages
Health Organization, 1993 (unpublished docu-
of using historical survey data are that
ment WHO/DAP/93.1; available on request from
they have already been collected and
Department of Essential Drugs and Medicines
carry no additional cost. However, their
Policy, World Health Organization, 1211
limitations include not being able to con-
Geneva, 27, Switzerland).
trol exactly which data have been collec-
ted or from where, not knowing whether
3.8 Exercises
current practices reflect those of the past,
Examine the sources of data listed in the
and having no patient-specific or provi-
Worksheet. Imagine that you want to learn
der-specific information. It would also
about the utilization of antibiotics in your coun-
usually not be possible to find informa-
try. In the spaces provided in the right-hand
tion on dosing of antibiotics.
columns of the worksheet, write down (1) what
kinds of useful data you might gather from each
 Worksheet for section 3.8
Sources of data on drug utilization

Data source Type of information available Advantages and limitations

Drug import records

Drug supply to health

facility
Orders and/or
delivery receipts

Previous reports of
surveys
25
Pharmacy stock cards/
pharmacy ledger book

Pharmacy sales
receipts

Large hospital or
insurance databases

Private
drug outlet sales
records

Community
or household surveys

Drug manufacturing
records

Other sources
 Chapter 4: Economic aspects of drug
use (pharmacoeconomy)

Pharmacoeconomics; types of cost; cost-mini- 4.2 Cost-minimization analysis

mization analysis; cost-effectivness analysis; Cost-minimization analysis is a method of calcu-
cost-benefit analysis; cost-utility analysis lating drug costs to project the least costly drug
or therapeutic modality. Cost minimization also
4.1 Introduction reflects the cost of preparing and administering a
Drug costs per se are important, as they account for dose. This method of cost evaluation is the one
a substantial part of the total cost of health care - used most often in evaluating the cost of a speci-
typically 10-15% in developed countries and up to fic drug. Cost minimization can only be used to
30-40% in some developing countries. However, compare two products that have been shown to
drug costs usually need to be interpreted in the con- be equivalent in dose and therapeutic effect.
text of the overall (net) costs to the health system. Therefore, this method is most useful for compa-
Drugs cost money to buy, but their use may also ring generic and therapeutic equivalents or «me
save costs in other areas. For example, the purchase too» drugs. In many cases, there is no reliable
of one specific type of drug may lead to reductions equivalence between two products and if thera-
in the following: peutic equivalence cannot be demonstrated, then
– use of other drugs; cost-minimization analysis is inappropriate.
– the number of patients requiring hospitalization If a new therapy were no safer or more effecti-
or in the length of stay in hospital; ve than an existing therapy (i.e. there is no incre-
26 – the number of doctor visits required; mental benefit), it would normally justify the
– administration and laboratory costs compared same price as the existing therapy. An example
with those incurred by using another drug to treat would be the introduction of a new ACE inhibi-
the same condition. tor with essentially the same properties as exis-
Assessing the true cost to a health system of ting members of the class; the price would be
using a specific drug will therefore require the cost equivalent to that of the existing drug(s). This is
of acquisition of the drug to be balanced against often not as simple as it may seem, as it requires
both any cost savings resulting from the use of that sound trial-based information on the doses of the
drug and the extra health benefits it may produce. two drugs required for equivalent efficacy. An
On the other hand, costs may arise from adverse alternative is to use the PDDs for the two drugs
drug reactions both in the short- and particularly the in the marketplace to determine the relative pri-
long-term. ces. This is a pragmatic approach, but assumes
Assessing the value for money of using a drug that the two drugs are actually used at equiva-
requires the extra health benefits achieved to be lently effective doses, and this may not always
weighed against the extra net cost. This compari- be the case.
son is usually expressed as an incremental cost-
effectiveness ratio (ICER) which is the net incre- 4.1 Cost-effectiveness analysis
mental cost (costs minus cost offsets) of gaining an Cost-effectiveness analysis involves a more com-
incremental health benefit over another therapy. prehensive look at drug costs. While cost is
Concerns about the cost of medical care in gene- measured in monetary terms, effectiveness is
ral, and pharmaceuticals in particular, are currently determined independently and may be measured
being expressed by all health systems. There is a in terms of a clinical outcome such as number of
general focus on providing quality care within limi- lives saved, complications prevented or diseases
ted financial resources. Decision-makers are increa- cured.
singly dependent on clinical economic data to guide Cost-effectiveness analysis thus measures the
policy formulation and implementation. Some of incremental cost of achieving an incremental
the concepts used in making such decisions include: health benefit expressed as a particular health
cost-minimization, cost-effectiveness, outcome that varies according to the indication
cost-benefit, and cost-utility. for the drug. Examples of ICERs using this
approach are: ratios enables the cost of achieving a health benefit
– the cost per extra patient achieving a 10 mm by treatment with a drug to be assessed against
Hg fall in blood pressure; similar ratios calculated for other health inter-
– the cost per extra asthmatic patient achieving ventions (e.g. surgery or screening by mammo-
a reduction in oral corticosteroid use graphy). It therefore provides a broader context
– the cost per extra episode of febrile neutro- in which to make judgements about the value for
penia avoided; or money of using a particular drug.
– the cost per extra acute rejection episode avo-
ided in patients with kidney transplants.
4.5 Cost-benefit analysis
Cost-benefit analysis is used to value both incre-
It is often difficult to make judgements about the
mental costs and outcomes in monetary terms
relative value for money across a range of drug
and therefore allows a direct calculation of the
groups and health outcomes such as those in the
net monetary cost of achieving a health outcome.
examples given above.
A gain in life-years (survival) may be regarded
as the cost of the productive value to society of
that life-year using, for example, the average
4.4 Cost-utility analysis wage. The methods for valuing gains in quality
Cost-utility analysis is used to determine cost in of life include techniques such as willingness-to-
terms of utilities, especially quantity and quality pay, where the amount that individuals would be
27
of life. This type of analysis is controversial willing to pay for a quality-of-life benefit is
because it is difficult to put a value on health sta- assessed. However, the techniques used to value
tus or on an improvement in health status as per- health outcomes in monetary terms remain
ceived by different individuals or societies. somewhat controversial, with the result that cost-
Unlike cost-benefit analysis, cost-utility analysis benefit analysis is so far not widely used in phar-
is used to compare two different drugs or proce- macoeconomic analyses.
dures whose benefits may be different. Economic analyses such as those described
Cost-utility analysis expresses the value for above may be trial based or modelled. A trial
money in terms of a single type of health out- based analysis uses the incremental benefits and
come. The ICER in this case is usually use of resources in a clinical trial to calculate an
expressed as the incremental cost to gain an ICER, but this may not be as relevant to the use
extra quality-adjusted life-year (QALY). This of the drug as it would be in the marketplace. A
approach incorporates both increases in survival modelled analysis is used to apply the benefits
time (extra life-years) and changes in quality of and use of resources to a local clinical situation,
life (with or without increased survival) into one and to extend the time frame beyond that of a
measure. An increased quality of life is clinical trial. This is particularly important
expressed as a utility value on a scale of 0 (dead) where the benefits of treatment may not be reali-
to one (perfect quality of life). An increased zed until some time in the future. Two examples
duration of life of one year (without change in are the avoidance of liver cancer or transplantati-
quality of life), or an increase in quality of life on for patients with hepatitis C and the prolong-
from 0.5 to 0.7 utility units for five years, would ation of life for hypertensive patients. Short-
both result in a gain of one QALY. This allows term surrogate outcome measures (clearance of
for easy comparison across different types of virus and lowering of blood pressure, respective-
health outcome, but still requires value judge- ly) are used in clinical trials, and need to be
ments to be made about increases in the quality translated by modelling into the longer-term out-
of life (utility) associated with different health comes, which are more relevant to patients and
outcomes. The use of incremental cost-utility policy-makers.
 4.6 General reading Klotgon, has recently been brought to your
How to investigate drug use in health facilities: attention. The two drugs have been compared in
Selected drug use indicators. Geneva,World a large randomized trial in which the primary
Health Organization, 1993 (unpublished docu- outcome of mortality was measured 30 days
ment WHO/DAP/93.1; available on request from after randomization.
the Department of Essential Drugs and
Medicines Policy, World Health Organization, Outcomes in 100 patients
1211 Geneva, 27, Switzerland). No treatment 15 deaths
Schulman KA et al. Pharmacoeconomics: Thrombase 10 deaths
Economic evaluation of pharmaceuticals. In: Klotgon 7 deaths
Strom B. ed. Pharmacoepidemiology, 3rd ed. Drug cost per patient
Chichester, J Wiley, 2000. Thrombase $ 200
Klotgon $ 1000
4.7 Exercises
1. Comparison of antihypertensives You are also aware that the average survival time
You are considering the use of a new alpha-anta- following non-fatal myocardial infarction is
gonist for the treatment of hypertension. It is eight years.
used once daily and you are told that it has been Please answer the following questions. Be
28 tested in trials against enalapril and losartan and prepared to present your findings to the large
it has been found to lower blood pressure to a group.
similar extent to these agents. You already have a. If the hospital budget were unlimited, and if
prazosin on your subsidy list but the producers 1000 patients were to be treated, how many
inform you that they have not carried out trials lives could be saved if patients were treated
against prazosin. The approximate costs for a with Thrombase, compared with no treatment?
month’s supply of the existing drugs are prazosin How many could be saved with Klotgon, com-
$18, enalapril $28 and losartan $35. Beta-bloc- pared with no treatment?
kers and thiazide diuretics are also on your sub-
sidy list at a cost of about $8 for a month’s sup- b. If the hospital’s budget for purchasing throm-
ply but no trials of the new agent have been car- bolytics were $200 000, how many patients
ried out against them. How would you approach could be treated, and how many lives could be
the pricing of the new alpha-antagonist? saved with each of the drugs, compared with
no treatment at all?
2. Thrombolytics for acute myocardial c. What is the incremental cost per life saved, for
infarction (a purely hypothetical exercise) each of the thrombolytic agents, compared
Congratulations! You have just been appointed with no active treatment?
as a member of the formulary committee of a
large teaching hospital in Sydney, Australia. A d. What are the incremental cost-effectiveness
ratios (ICERs), expressed as the incremental
key item on the agenda for the next meeting is a
cost per life-year gained, for each of the
proposal to implement a management protocol
thrombolytic agents, compared with no active
for the treatment of acute myocardial infarction.
treatment?
You have been asked to evaluate the available
evidence and advise to the committee as to e. What is the ICER for Klotgon compared to
which of two available drugs represents the more Thrombase?
cost-effective choice.
f. What will you recommend to the formulary
The draft clinical management protocol cur-
committee?
rently proposes the use of the (hypothetical)
thrombolytic drug Thrombase. A new drug,
 Outcome Low-molecular- Unfractionated P-value
weight heparin heparin

Combined risk of death 318/1607 364/1564 0.016

AMI1 or unstable angina (19.8%) (23.3%)

Percutaneous 236/1607 293/1564 0.002

revascularization ($ 1390 per (14.7%) (18.7%)

procedure)

Major bleeds 102/1569 107/1528 0.57

(6.5%) (7.0%)

Minor bleeds 188/1580 110/1528 <0.001

(11.9%) (7.2%)

3. Unfractionated heparin versus low- Please answer the following questions. Be pre-
molecular-weight heparin pared to present your findings to a large group.
29
Because of your valuable contribution to the a. Calculate the relative risk of the combined
development of a cost-effective treatment proto- (triple) end-point in patients who received
col for acute myocardial infarction, you have low-molecular-weight heparin compared with
been retained as a member of the formulary those who received unfractionated heparin.
committee of the above-mentioned hospital. An b. Calculate the risk difference and the number
agenda item for consideration at your committee’s of patients who need to be treated to prevent a
next meeting concerns a recommendation, from single event with low-molecular-weight hepa-
rin compared with unfractionated heparin.
a very pleasant pharmaceutical company repre-
c. Calculate the ICER for the main clinical out
sentative, that you replace unfractionated heparin
come with low-molecular-weight heparin,
with a low- molecular-weight heparin in the
compared with unfractionated heparin using
management of patients with unstable coronary
drug costs only.
artery disease. She very kindly gives you a sum-
d. Recalculate the ICER for the main clinical
mary of some data from a clinical trial published outcome with low-molecular-weight heparin,
in the New England Journal of Medicine. The compared with unfractionated heparin includ-
outcomes were reported 30 days after randomi- ing the costs of monitoring treatment with
zation. heparin.

You decide to investigate the costs of acquiring 4. Celecoxib versus diclofenac

and monitoring treatment with the two drugs and A representative from a very supportive pharma-
note the following: ceutical company addressed your medical staff
during a Saturday seminar last week. She gave
Item Low-molecular- Unfractionated an interesting presentation on the comparative
weight heparin heparin safety of some well-known anti-inflammatory
Monthly drug costs ($) $72.20 $27.09 preparations. At the formulary meeting this
Monthly cost none 5 tests/patient of week, the head of your rheumatology department
monitoring is planning to propose adding celecoxib, a COX-
anticoagulant effect @ $12.40 per test
1
Acute myocardial infarction
 2 inhibitor, to the hospital formulary in place of the committee as a late submission, so you deci-
NSAIDs. He intends to argue that the hospital de to review the evidence and prepare yourself
will save a lot of money by avoiding the compli- for the discussion. You find the following results
cations associated with NSAIDs such as peptic of a clinical trial reported in the Lancet.
ulcers. This item was placed on the agenda of

Mean (SD) arthritis assessment results at week 24

Primary assessments Celecoxib Diclofenac

Baseline Week 24 Baseline Week 24

Physician’s assessmenta 2.9 (0.7) 2.6 (0.8) 3.0 (0.8) 2.6 (0.8)
Patient’s assessmenta 3.0 (0.8) 2.7 (0.9) 3.1 (0.8) 2.8 (0.9)
No. of tender/painful joints 20.3 (14.4) 14.5 (14.1) 21.7 (14.4) 16.4 (14.7)
No. of swollen joints 14.9 (10.2) 10.7 (10.1) 14.3 (9.9) 10.4 (10.0)

30
The following adverse event data were also reported.

Frequency of peptic ulceration and related complications

Celecoxib Diclofenac P-value

(n = 212) (n = 218)

Patients in whom erosion, ulcer or both were detected

Gastric 38 (18%) 74 (34%) <0.001

Duodenal 11 (5%) 23 (11%) <0.009
Ulcer incidence by Helicobacter pylori status
Positive serological test 7/93 (8%) 19/87 (22%) NS
Negative serological test 1/97 (1%) 10/100 (10%) NS
Ulcer frequency by concomitant corticosteroid use
Corticosteroid use 2/80 (3%) 12/102 (12%) NS
No corticosteroid use 6/132 (5%) 21/116 (18%) NS
NS: not significant.

aIndependent assessments, graded from 1 (very good: symptom-free with no limitation of normal activities) to 5 (very poor: very severe symptoms that
are intolerable, and inability to carry out all normal activities).
From your research you also know that: c. Calculate the ICER for the main clinical out-
• One per cent of patients with endoscopic come with celecoxib, compared with the
damage are hospitalized with gastrointestinal NSAID, using drug costs only.
bleeding. d. Recalculate the ICER for the main clinical
outcome with celecoxib, compared with the
• The cost of hospitalization for gastrointestinal NSAID, including the costs of treatment of
bleeding is $1434/patient. gastrointestinal bleeding.
• Ten per cent of patients admitted with gastroin-
testinal bleeding die. 5. Oral montelukast versus an inhaled
• The cost of celecoxib for 60 x 100 mg tablets steroid
is $50. A community-driven asthma awareness group
has donated 10 cartons of montelukast tablets for
• The usual dose of celecoxib is 200 mg bd.
adults treated in your hospital’s asthma clinic.
• The cost of diclofenac is $11.60 for 50 x 50 They feel strongly that this product should be
mg tablets; $14.35 for 100 x 25 mg tablets. made available since, according to the medical
• The usual dose of diclofenac is 50 mg-75 mg bd. adviser of the sponsoring company, it is much
• Answer the following questions. Be prepared more effective and much easier to use than the
to present your findings to a large group. usual «puffers». As this product is fairly new,
the formulary committee has been asked to 31
a.Calculate the relative risk for peptic (i.e. gas- comment on its effectiveness. Since the asthma
tric or duodenal) ulcers in the patients who unit will prepare a submission for including it in
received celecoxib compared with those who the hospital formulary after the supply of dona-
received the NSAID diclofenac. ted drugs is exhausted, you have been asked to
b.Calculate the risk difference and the number comment on the comparative cost-effectiveness
of patients who have to be treated to prevent a of the drug. You begin your assessment by con-
single event with celecoxib, as compared with sidering the following results at 22 weeks after
the NSAID. the baseline assessment.

End-point Placebo Beclometasone Montelukast

Percentage change *FEV1 0.7 [–2.3, 3.7] 13.1 [10.1, 16.2] 7.4 [4.6, 10.1]
Change in daytime
asthma symptom score –0.17 [–0.3, –0.05] –0.62 [–0.75, 0.49] –0.41 [–5.3, –0.29]
Percentage change in total daily
β-agonist use 0.0 [–8.3, 8.3] –40.0 [–48.5, –31.5] –23.9 [–31.4, –16.5]
Change in morning PEFR [l/min] 0.8 [–7.1, 8.6] 39.1 [31.0, 47.1] 23.8 [16.6, 30.9]
Change in evening PEFR [l/min] 0.3 [–7.3, 8.0] 32.1 [24.2, 39.9] 20.8 [13.8, 27.8]
Change in nocturnal awakenings –0.5 [–0.9, –0.1] –2.4 [–2.8, –2.0] –1.7 [–2.07, 1.3]
[nights per week]
Change in eosinophil count
[cells x 103/µl] –0.02 [–0.07, 0.03] –0.07 [–0.12, –0.02] –0.08 [–0.12, –0.03]
Percentage of patients with
asthma attacks 27.3 10.1 15.6

Values are mean [95% CI]. *FEV1 = forced expiratory volume in one second; PEFR = peak expiratory flow rate
 The costs of the two drugs are: a.You wish to compare montelukast with beclo-
– Beclametasone: Australian $26 for 28 days of metasone. Which outcome(s) will you use for
treatment; the comparison? Why?
– Montelukast: Australian $70 for 28 days of b.Calculate the ICER for the main clinical outcome.
treatment. c.Which is the better drug? Why?

Please answer the following questions. Be pre-

pared to present your findings to a large group.

32
Chapter 5: Drug classification systems

A drug classification system represents a com- were originally based on the same main princi-
mon language for describing the drug assortment ples. In the EPhMRA system, drugs are classi-
in a country or region and is a prerequisite for fied in groups at three or four different levels.
national and international comparisons of drug The ATC classification system modifies and
utilization data, which have to be collected and extends the EPhMRA system to include a thera-
aggregated in a uniform way. Access to stan- peutic/pharmacological/chemical subgroup as
dardized and validated information on drug use the fourth level and the chemical substance as
is essential to allow audits of patterns of drug the fifth level (see, for example, the classifi-
utilization, to identify problems in drug use, to cation of glibenclamide in the box below).
initiate educational or other interventions and to
monitor the outcomes of these interventions. The ATC classification is also the basis for the
The main purpose of having an international classification of adverse drug reactions used by
standard is to be able to compare data between the WHO Collaborating Centre for International
countries. A recent example is the international Drug Monitoring in Uppsala, Sweden
focus on creating comparable systems for moni- (www.who-umc.org).
toring cross-national patterns of antibacterial uti-
lization to aid work against bacterial resistance. The main purpose of the ATC classification is as
a tool for presenting drug utilization statistics
5.1 Different classification and it is recommended by WHO for use in inter- 33
systems national comparisons. The EPhMRA classification
system is used worldwide by IMS for providing
ATC classification; AT classification;
market research statistics to the pharmaceutical
EPhMRA; IMS; WHO Collaborating Centre
industry. It should be emphasized that the many
for Drug Statistics Methodology
technical differences between the EPhMRA
classification and the ATC classification mean that
Drugs can be classified in different ways accor-
data prepared using the two classification systems
ding to:
are not directly comparable.
– their mode of action;
In 1996, WHO recognized the need to develop
– their indications; or
the ATC/DDD system from a European to an
– their chemical structure.
international standard in drug utilization studies.
The European WHO Collaborating Centre for
Each classification system will have its advanta-
Drug Statistics Methodology in Oslo, Norway,
ges and limitations and its usefulness will
which is responsible for coordinating the use of
depend on the purpose, the setting used and the
the methodology, was then linked to WHO
user’s knowledge of the methodology.
Headquarters in Geneva. This was intended to
Comparisons between countries may require a
assist WHO in its efforts to ensure universal
classification system different from that needed
access to essential drugs and to stimulate ratio-
for a local comparison (e.g. between different
nal use of drugs particularly in developing coun-
wards in a hospital). Of the various systems pro-
tries.
posed over the years, only two have survived to
attain a dominant position in drug utilization
5.2The ATC classification system
research worldwide. These are the «Anatomical
Therapeutic» (AT) classification developed by the Structure; coding principles; therapeutic use;
European Pharmaceutical Market Research [ pharmaceutical formulations; strengths ]
Association (EPhMRA) and the «Anatomical
Therapeutic Chemical» (ATC) classification deve- The ATC classification system divides the drugs
loped by Norwegian researchers. These systems into different groups according to the organ or
 system on which they act and according to their A medicinal product can be given more than one
chemical, pharmacological and therapeutic prop- ATC code if it is available in two or more
erties. strengths or formulations with clearly different
Drugs are classified in groups at five different therapeutic uses. Two examples of this are given
levels. The drugs are divided into 14 main below:
groups (first level), with two therapeutic/pharma-
cological subgroups (second and third levels). • Sex hormones in certain dosage forms or
The fourth level is a therapeutic/pharmacologi- strengths are used only in the treatment of cancer
cal/chemical subgroup and the fifth level is the and are thus classified under L02 - Endocrine
chemical substance. The second, third and therapy. The other dosage forms and strengths
fourth levels are often used to identify pharma- are classified under G03 - Sex hormones and
cological subgroups when these are considered modulators of the genital system.
to be more appropriate than therapeutic or chem- • Bromocriptine is available in different
ical subgroups. strengths. The low-dose tablets are used as pro-
The complete classification of glibenclamide lactin inhibitors and are classified in G02 - Other
(see box below) illustrates the structure of the gynaecologicals. Bromocriptine tablets in higher
code. strengths are used to treat Parkinson disease and
are classified in N04 - Anti-Parkinson drugs.
34 A Alimentary tract and metabolism
(first level, main anatomical Different formulations with different indications
group) may also be given separate ATC codes, for
example prednisolone is given several ATC
A10 Drugs used in diabetes codes because of the different uses of the diffe-
(second level, main therapeutic rent formulations (see box below).
group)

A10B Oral blood-glucose-lowering A07E A01 Intestinal anti-inflammatory

drugs (third level, therapeutic agents (enemas and rectal
/pharmacological subgroup) foams)
C05A A04 Antihaemorrhoidals for topical
A10B B Sulfonamides, urea derivatives use rectal suppositories)
(fourth level, chemical/therapeutic D07A A03 Dermatological preparations
/pharmacological subgroup) (creams, ointments, lotions)
H02A B06 Corticosteroids for systemic
A10B B01 Glibenclamide use (tablets, injections)
(fifth level, subgroup for R01A D02 Nasal decongestants (nasal
chemical substance) spray, drops) S01B A04
Ophthalmologicals (eye drops)
S02B A03 Otologicals (ear drops)
Thus, in the ATC system all plain glibenclamide
preparations are given the code A10B B01.
Medicinal products are classified according to The ATC system is not strictly a therapeutic clas-
the main therapeutic use of their main active sification system. At all ATC levels, ATC codes
ingredient, on the basic principle of assigning can be assigned according to the pharmacologi-
only one ATC code for each pharmaceutical for- cal properties of the product. Subdivision on the
mulation (i.e. similar ingredients, strength and basis of mechanism of action will understanda-
pharmaceutical form). bly be rather broad, since a very detailed classifi-
cation of this kind would result in having only Because the ATC system separates drugs into
one substance per subgroup, which is better avo- groups at five levels (described above), statistics
ided (e.g. in the case of antidepressants). Some on drug utilization grouped at the five different
ATC groups are subdivided into both chemical levels can be provided. The information avai-
and pharmacological groups (e.g. ATC group lable ranges from figures showing total use of all
J05A - Agents affecting the virus directly). If a drug products classified e.g. in main group C -
new substance fits in both a chemical and phar- Cardiovascular system (first level), to figures for
macological fourth level, the pharmacological the different subgroups (i.e. second, third and
group is normally chosen. fourth level) to figures for the use of the separate
Substances classified as having the same ATC substances.
fourth level should not be considered as pharma-
cotherapeutically equivalent since the profiles for More detailed information can be obtained at the
their mode of action, therapeutic effects, drug lower (i.e. the fourth and fifth) levels. The hig-
interactions and adverse drug reactions may differ. her levels are used if comparison of drug groups
As the drugs available and their uses are cont- is the aim of a study (see Fig. 5). This gives a
inuously changing and expanding, regular revisi- better overview and trends in drug use related to
ons of the ATC system are necessary. An impor- different therapeutic areas can easily be identified.
tant principle is to keep the number of alterations
to a minimum. Before alterations are made, any 5.3 Ambivalence towards an 35
potential difficulties arising for the users of the international classification
ATC system are considered and related to the system
benefits that would be achieved by the alteration. All international standards demand compromises
Changes to the ATC classification would be and a drug classification system is no exception
made when the main use of a drug had clearly to this rule. Drugs may be used for two or more
changed, and when new groups are required to equally important indications, and the main the-
accommodate new substances or to improve the rapeutic use of a drug may differ from one coun-
specificity of the groupings. try to another. This will often result in several

80

70
ACE-inbitors +
angiotensin II-
60 antagonists (C09)
Serum lipid reducing
agents (C10)
50
DDD/1000 inhab.day

Calsium channel
blockers (C08)
40
Diuretics (C03)

30
Beta blocking agents
(C07)
20

10

0
1990

1991

1992

1993
1994

1995

1996

1997

1998

1999

2001
2000

Figure 5 Total sales of drugs used in cardiovascular disorders in Norway 1990-2001. ATC/DDD version
2002
 possible alternatives for classification, and a national basis to secure consistent use of the
decision has to be made regarding the main use. methodology within a country. As described in
Countries using a drug in a different way from the general introduction, the same substance may
that indicated by the ATC classification may not have several different ATC codes depending on
wish to adopt the ATC classification but prefer to the application form and, to some extent, even
develop national classification systems. the strength. For combination products, specific
However, national traditions have to be weighed guidelines have been established for allocating
against the opportunity to introduce a methodo- ATC codes. Allocating DDDs to the products
logy that permits valid international comparisons necessitates many of the same considerations as
of drug utilization. Indeed, there are now many the allocation of the ATC code. However, in
examples where an enthusiastic application of order to link the drug list with sales figures or
the ATC/DDD methodology has been instrumen- prescription figures to obtain drug utilization sta-
tal in stimulating national research in drug utili- tistics, it is necessary to make appropriate calcu-
zation and in developing an efficient drug con- lations such as the number of DDDs per drug
trol system. package.
Finally, a given country will nearly always
5.4 Implementation of the have medicines and combination products for
ATC/DDD methodology which no ATC codes or DDDs exist. In these
36 cases, it is important to consult the WHO
National drug register; dynamic system;
[ different versions ] Collaborating Centre for Drug Statistics
Methodology in Oslo and request new ATC
codes and DDDs. Once ATC codes and DDDs
As soon as the decision to introduce the
have been linked to the national drug lists, it is
ATC/DDD methodology is taken, it is essential
necessary to update the drug list regularly in
to realize that its proper use inevitably includes
accordance with the annual updates of the
an important and time-consuming first step.
ATC/DDD system.
Each product has to be connected to the appro-
The publication Guidelines for ATC
priate ATC code and DDD (see chapter 6). The
Classification and DDD Assignment (see
linkage between the national drug register and
General reading) gives the information necessary
ATC/DDDs has to be ascertained by persons
for allocating ATC codes and DDDs at a national
with proper knowledge of the methodology.
or local level. All officially assigned ATC codes
Experience has shown that in many countries,
and DDDs are listed in the ATC Index with
health authorities, health policy-makers and rese-
DDDs (see General reading), a publication that
archers have not always allocated adequate
is also available in electronic format and is upda-
resources to this important initial step. Another
ted every year. Training courses in the
problem is that some users seem to be unaware
ATC/DDD methodology are arranged annually
that the ATC/DDD methodology is a dynamic
in Norway and from time to time in other coun-
system to which changes are made continually.
tries. Further information is available on the
This has resulted in several different versions of
web site of the WHO Collaborating Centre for
the ATC/DDD system being used at the same
Drug Statistics Methodology at
time, sometimes even within the same country.
http://www.whocc.no.
It is important to realize that adopting the
ATC/DDD classification of drugs requires
5.5 General reading
resources and the necessary competence to carry
Guidelines for ATC classification and DDD
out the work of allocating ATC codes to the pro-
Assignment. Oslo, Norway, WHO Collaborating
ducts. If possible, this work should be done on a
Centre for Drug Statistics Methodology, 2003.
 ATC Index with DDDs. Oslo, Norway, WHO treatment of schizophrenia. The oral formula-
Collaborating Centre for Drug Statistics tions of Neurol are, however, used mainly in
Methodology, 2003. the treatment of schizophrenia.
Capellà D. Descriptive tools and analysis. In: Discuss whether it would be appropriate to
Dukes MNG ed. Drug utilization studies, met- assign an additional ATC code in N05 for oral
hods and uses. Copenhagen, WHO Regional formulations of Neurol.
Office for Europe, 1993 (WHO Regional 2. Lisuride has been assigned two codes in the
Publications, European Series, No. 45), 55-78. ATC classification system:
Rønning M et al. Different versions of the G02CB02 (Prolactin inhibitors) and
anatomical therapeutic chemical classification N02CA07 (Antimigraine preparations).
Lisuride preparations in high strengths (e.g.
system and the defined daily dose - are drug
tablets of 0.2 mg) are classified in G02CB.
utilization data comparable? European Journal
The recommended dose range for prolactin
of Clinical Pharmacology, 2000, 56:723-727.
inhibition is 0.1-0.2 µg x 3. Low-strength
Rägo L. Estonian regulatory affairs.
preparations (e.g. tablets of 25 µg) are classi-
Regulatory Affairs Journal. 1996, 7:567-573.
fied in N02CA. The recommended dose range
for treatment of migraine is 25 mg x 3.
5.6 Exercises Lisuride is also used for the treatment of
1. «Neurol» is a major tranquillizer belonging to parkinsonism. The recommended dose range
the butyrophenone group of antipsychotics. for this indication is 0.2-0.6 mg daily. 37
The only ATC code for this substance at pre- Discuss whether it would be appropriate to
sent is in N01A X. The parenteral formulati- assign an additional ATC code for lisuride in
ons of Neurol are used for various indications N04.
e.g. in anaesthesia, as antiemetics and in the
 Chapter 6: Drug utilization metrics and
their applications

6.1. The concept of the defined per 1000 inhabitants per day indicates that 1% of
daily dose (DDD) the population on average might receive a certain
drug or group of drugs daily. This estimate is
Definition; DDDs per1000 inhibitants per most useful for chronically used drugs when
day; DDDs per 100 bed-days; there is good agreement between the average
DDDs per inhibitant per year prescribed daily dose (see below) and the DDD.
It may also be important to consider the size of
The historical development of the concept of the the population used as the denominator. Usually
defined daily dose (DDD) and its early applicati- the general utilization is calculated for the total
ons are described in the Preface. population including all age groups, but some
drug groups have very limited use among people
The DDD is the assumed average maintenance below the age of 45 years. To correct for differ-
dose per day for a drug used for its main indi- ences in utilization due to differing age struc-
cation in adults. tures between countries, simple age adjustments
can be made by using the number of inhabitants
It should be emphasized that the DDD is a unit in the relevant age group as the denominator.
of measurement and does not necessarily corres-
pond to the recommended or prescribed daily DDDs per 100 bed-days
38 dose (PDD). Doses for individual patients and The DDDs per 100 bed-days may be applied
patient groups will often differ from the DDD as when drug use by inpatients is considered. The
they must be based on individual characteristics definition of a bed-day may differ between hos-
(e.g. age and weight) and pharmacokinetic consi- pitals or countries, and bed-days should be
derations. adjusted for occupancy rate. The same definition
The DDD is often a compromise based on a should be used when performing comparative
review of the available information about doses studies. As an example, 70 DDDs per 100 bed-
used in various countries. The DDD may even days of hypnotics provide an estimate of the the-
be a dose that is seldom prescribed, because it is rapeutic intensity and suggests that 70% of the
an average of two or more commonly used dose inpatients might receive a DDD of a hypnotic
sizes. every day. This unit is quite useful for bench-
Drug utilization figures should ideally be pre- marking in hospitals.
sented as numbers of DDDs per 1000 inhabitants
per day or, when drug use by inpatients is consi- DDDs per inhabitant per year
dered, as DDDs per 100 bed-days. For antiinfec- The DDDs per inhabitant per year may give an
tives (or other drugs normally used for short estimate of the average number of days for
periods), it is often considered to be most appro- which each inhabitant is treated annually. For
priate to present the figures as numbers of DDDs example, an estimate of five DDDs per inhabi-
per inhabitant per year. tant per year indicates that the utilization is equi-
valent to the treatment of every inhabitant with a
These terms are explained below. five-day course during a certain year.
Alternatively, if the standard treatment period is
DDDs per 1000 inhabitants per day known, the total number of DDDs can be calcu-
Sales or prescription data presented in DDDs per lated as the number of treatment courses, and the
1000 inhabitants per day may provide a rough number of treatment courses can then be related
estimate of the proportion of the study popula- to the total population.
tion treated daily with a particular drug or group
of drugs. As an example, the figure 10 DDDs
6.2 Prescribed daily dose and con- tifying drug utilization, but have certain disad-
sumed daily dose vantages (see below). These units can be applied
The prescribed daily dose (PDD) is defined as only when the use of a single drug or of well-
the average dose prescribed according to a repre- defined products is evaluated. Problems arise,
sentative sample of prescriptions. The PDD can however, when the utilization of whole drug
be determined from studies of prescriptions or groups is considered.
medical or pharmacy records. It is important to
relate the PDD to the diagnosis on which the Grams of active ingredient
dosage is based. The PDD will give the average If utilization is given in terms of grams of active
daily amount of a drug that is actually pre- ingredients, drugs with low potency will account
scribed. When there is a substantial discrepancy for a larger fraction of the total than drugs with
between the PDD and the defined daily dose high potency. Combined products may also con-
(DDD), it is important to take this into considera- tain different amounts of active ingredients from
tion when evaluating and interpreting drug utiliza- plain products, and this difference will not be
tion figures, particularly in terms of morbidity. reflected in the figures.
For drugs where the recommended dosage dif- Number of tablets
fers from one indication to another (e.g. the Counting numbers of tablets does not reflect the
antipsychotics), it is important to link the diag- variations in strengths of tablets, with the result
nosis to the PDD. Pharmacoepidemiological that low-strength preparations contribute relati-
39
information (e.g. on sex, age and whether thera- vely more than high-strength preparations to the
py is mono- or combined) is also important in total numbers. Also, short-acting preparations
order to interpret a PDD. will often contribute more than long-acting pre-
The PDD can vary according to both the ill- parations.
ness treated and national therapeutic traditions.
For instance, for the anti-infectives, PDDs vary Numbers of prescriptions
according to the severity of the infection treated. Numbers of prescriptions do not accurately
The PDDs also vary substantially between dif- reflect total use, unless total quantities of drugs
ferent countries, for example, PDDs are often per prescription are also considered. However,
lower in Asian than in Caucasian populations. counting of prescriptions is valuable in measu-
The fact that PDDs may differ from one country ring the frequency of prescriptions and in evalua-
to another should always be considered when ting the clinical use of drugs (e.g. diagnosis and
making international comparisons. dosages used).
It should be noted that the PDD does not nec- Although they are useful in making national
essarily reflect actual drug utilization. Some comparisons it should be noted that none of
prescribed medications are not dispensed, and these volume units is usually applicable in cross-
the patient does not always take all the medica- national comparisons, as was pointed out during
tions that are dispensed. Specially designed the 1969 WHO symposium in Oslo.
studies including patient interviews are required
to measure actual drug intake at the patient level 6.4 Cost
(i.e. the consumed daily dose). Drug use can be expressed in terms of costs (e.g.
national currency). Cost figures are suitable for
6.3 Other units for presentation an overall analysis of expenditure on drugs.
of volume International comparisons based on cost parame-
Common physical units (e.g. grams, kilograms ters can be misleading and have limited value in
and litres), numbers of packages or tablets and the evaluation of drug use. Price differences
numbers of prescriptions are also used for quan- between alternative preparations and different
 national cost levels make the evaluation difficult. 6.5 General reading
Long-term studies are also difficult due to fluc- Consumption of drugs. Report of a symposium
tuations in currency and changes in prices. in Oslo, 1969. Copenhagen, WHO Regional
When cost data are used, an increase in the use Office for Europe, 1970 (EURO 3102).
of cheaper drugs may have little influence on the Studies in drug utilization: methods and appli-
total level of expenditure on drugs, while a shift cations. Copenhagen, WHO Regional Office for
to more expensive drugs is more readily noticed. Europe 1979 (Regional Publications European
The trends in drug use measured in cost may Series No.8).
therefore look very different from the same drug Bergman U et al. Auditing hospital drug uti-
use measured in DDDs. As an example, the total lization by means of defined daily doses per bed-
drug use in Norway from 1987-1999 measured day. A methodological study. European Journal
in cost (Euros) and in DDDs is shown in Figs 6 of Clinical Pharmacology, 1980, 17:183-187.
and 7.

1400

1200
mill Euro parmacy retail price

40
1000

800

600

400

200

0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

Figure 6 Total sales of drugs in Norway in millions of Euros 1987-1999

3000

2500

2000

1500

1000

500
mill DDD

0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

Figure 7 Total sales of drugs in Norway in millions of DDDs 1987-1999

Baksaas I. Patterns in drug utilization - national Comment on the comparability of these figu-
and international aspects: antihypertensive drugs. res. Discuss how to best present the sales
Acta Medica Scandinavica, 1984, suppl. figures from these two years in the same
683:59-66. article.
Lee D, Bergman U. Studies of drug utiliza-
tion. In: Strom B ed. Pharmacoepidemiology,
3. Annual sales figures given in millions of
3rd ed. Chichester, J Wiley, 2000:463-481.
DDDs are:

6.6 Exercises
Substance A Substance B
1. Assign DDDs for the two antibacterials
1988 1.7 21.6
below according to the following dose
1996 9.1 9.9
recommendations.
Substance A: 500 mg on first day, then 250
Total number of inhabitants: 4 million
mg daily; duration of treatment 14 days.
Substance B: 500 mg on first day, then 250
Calculate total number of four-day courses of
mg daily; duration of treatment five days.
substance A sold per year and the equivalent
number of courses per inhabitant.
2. The DDD for budesonide inhalation powder Calculate total number of eight-day courses of
was changed from 0.3 mg to 0.8 mg in 1991. substance B sold per year and the equivalent
41
The following sales figures from Norway for number of courses per inhabitant.
budesonide inhalation powder are found in
two different books on drug statistics.

1990 9.6 DDDs /1000 inhabitants/day

(DDD = 0.3 mg)

1994 11.6 DDDs /1000 inhabitants/day

(DDD = 0.8 mg)
 Chapter 7: Solutions to the exercises

Chaper 2 - Types of drug use information

2.1. Amoxicillin
The use of a drug expressed as DDD/1000 population/day is derived by calculating the overall
amount of a drug being used over a specified period of time (e.g. a year) and dividing this by
the DDD multiplied by the population and the number of days in the period.

DDD/1000 population/day = Amount used in 1 year (mg)*1000

DDD(mg) x population x 365(days)

The amount used is a function of the number of prescriptions, the number of tablets or capsules
per prescription and the dose size of the tablets or capsules.
An understanding of the above allows the following hypotheses about the reasons for increa-
sed use to be generated and tested.

Hypothesis 1
The number of prescriptions per year has increased.
The information needed to test this hypothesis would be prescription numbers per year adjusted
42 for population changes over the study period. Remember that the DDD/1000 population/day is
corrected for population changes. Another way of addressing this hypothesis for a drug mainly
used acutely in short courses would be to obtain data expressed as the number of amoxicillin
treatment courses/1000 population/year.
If the prescription rate has increased, questions could be asked about the reasons for this.

Have the indications changed?

This would require data over time on the indications for which amoxicillin is used.
Has there been increased promotion for example, to introduce a new brand? This would require
a survey of promotional materials over time.

Hypothesis 2
The amount of amoxicillin per treatment course has increased.
This might be the result of an increase in the average length of the course and/or an increase in
the average PDD. The first possibility could be addressed by a survey of prescribers to find out
about the length of treatment courses, or a survey of prescriptions to calculate the duration of
treatment by dividing the PDD by the total quantity prescribed. To obtain the PDD, a prescrip-
tion survey would be required, either designed for this purpose or making use of data from ong-
oing surveys such as those conducted by IMS.

2.2. Antidepressant use

Both use and cost have been noted to increase over time. The types of data required to deter-
mine the reasons for increased use are similar to those suggested in answer to question 1 with
some differences. In this case, the data on the use of all the antidepressants have been aggrega-
ted so the data on the use of the individual agents and groups (TCAs, SSRIs and monoamine
oxidase (MAO) inhibitors) will need to be disaggregated. In looking at the relative use of
drugs or drug groups, it may be necessary to use both DDD/1000 population/day and prescrip-
 tion numbers. To interpret the data fully, it may be necessary to determine the PDDs for each
drug or drug group. If the relationship between DDD and PDD differs for the different drugs,
the trends based on DDD/1000 population/day may be misleading. For example, the use of the
SSRIs and moclobemide may have increased, while only a small decrease in TCA use has
occurred. This would have increased total antidepressant use with a multiplier effect on cost as
the SSRIs and moclobemide are patented and much more expensive than the older drugs. If
this is what has occurred a number of questions can be asked.

• Is the incidence or prevalence of depression in the community increasing?

• Is there increased awareness of depression by doctors and patients resulting in a
higher proportion of patients with depression being treated?
• If so, is this the result of government educational initiatives, or pharmaceutical
promotion aimed at case-finding and enhancement of compliance.
• Are there changes in the doses being used or in the duration of treatment?
• Has there been a change in the indications for which antidepressant drugs are
used?
• For example, has there been an increase in their use for the treatment of
obsessive–compulsive disorder, panic attacks or chronic pain?

Different types of data will be required to answer some of these questions and special surveys
43
will have to be designed and carried out. Some information (on indications, dose and duration
of treatment) may be available from ongoing prescriber surveys carried out either by academic
units or by commercial sources such as IMS. Data on the incidence and prevalence of depres-
sion may be available from government disease registries or similar sources. Qualitative studi-
es may need to be designed and carried out to determine for example, the degree of awareness
of depression as a problem and the sources that have been used to obtain information about
depression and its treatment.
Cost is a function of price and volume. The issue of volume has been addressed above. A
full assessment of the reasons for cost increases will require information on the price trends for
the drugs over time.

Questions about changes in utilization of a drug or drug group over time require a number of
different types and sources of data.

Chaper 4 - Economic aspects of drug use (pharmacoeconomy)

4.1. Comparison of antihypertensives

The goal of treating hypertension in terms of health outcomes is to prolong life by preventing
cardiovascular events and target organ damage. This is achieved by lowering blood pressure to
a range where absolute cardiovascular risk is essentially reduced to the population level. The
reduction in blood pressure is a surrogate outcome measure, but is accepted by most regulatory
authorities for registration purposes. All the drug groups lower blood pressure to approximate-
ly the same extent. Outcome studies are available for diuretics, beta-blockers and for the ACE
inhibitors, but not for the alpha-antagonists. In terms of subsidy listing, a principle should be
that, to achieve a price premium, a new drug should have demonstrated an increased benefit in
terms of health outcomes.
The company argues that this is a new innovative treatment that has been shown to be equiva-
 lent to losartan and they therefore demand a price equivalent to the A2 antagonists.
You reply that this is just another alpha-antagonist and it should therefore be compared with
prazosin.
The company states that there are no head-to-head trials of the new agent against prazosin
(they have not done any trials and have no intention of doing so) and therefore no evidence on
which to base a comparison. They argue the new agent should be compared with the ACE
inhibitors and A2 antagonists where there are good comparative data.

You reply that the lack of data comparing the new agent with prazosin is their problem, and
that if they want a higher price they should do the studies to demonstrate a health outcome
benefit over prazosin. Indeed, you wonder why prazosin has a price premium over the diure-
tics and beta-blockers and whether this should be reviewed to determine whether the higher
price is justified.
The company now argues that the new innovative drug has a longer half-life than prazosin so
that it can be administered once a day compared to twice a day for prazosin. It would therefore
improve compliance which is a very important consideration in treating hypertension.
You reply that the company has not demonstrated that the once-daily dose leads to improved
compliance or health outcomes and there is little evidence to support this supposition. A small
premium might be considered for the extra convenience for patients who are taking a life-long
44 treatment when they are essentially without symptoms.
The company decides not to proceed with the marketing of the new drug.
Who is right in this story? What price would you offer for this drug? Are you concerned that it
won’t be available?

4.2. Thrombolytics for acute myocardial infarction

a. Of 1000 patients treated with a placebo,150 will die.

Of 1000 patients treated with Drug A (Thrombase), 100 will die, therefore 50 lives will be saved.
Of 1000 patients treated with Drug B (Klotgon) 70 will die, therefore 80 lives will be saved.

b. Treatment with Thrombase

If the budget is $200 000 and the cost of treatment is $200 per patient ($2000/$200), 1000 pati-
ents could be treated and 50 lives saved (see question 1 above).

Treatment with Klotgon

If the budget is $200 000 and the cost of treatment is $1000 per patient ($200 000/$1000), 200
patients could be treated and 80 x 200/1000 = 16 lives could be saved.

c. If 1000 patients are treated with Thrombase, 50 lives are saved.

ICER (Thrombase versus placebo for 1000 patiens)= (1000 x $200-1000 x $0)
50 lives saved

= $200 000 = $4000 per life saved

50

If 1000 patients are treated with Klotgon, 80 lives are saved.

ICER (Klotgon versus placebo for 1000 patients)

= (1000 x $1000 – 1000 x $0) = $1 000 000 = $12 500 per life saved.
80 lives saved 80

d. If 1000 patients are treated with Thrombase, 50 lives are saved. Assuming an increase in
survival time of eight years per patient, 50 x 8 = 400 life-years are gained.

ICER (Thrombase versus placebo for 1000 patients)

= (1000 x $200 – 1000 x $0) = $200 000 = $500 per life-year gained.
400 life-years 400

If 1000 patients are treated with Klotgon, 80 lives are saved. Assuming an increase in survival
time of eight years per patient, 80 x 8 = 640 life-years are gained.

ICER (Klotgon versus placebo for 1000 patients)

= (1000 x $1000 – 1000 x $0) = $1 000 000 = $1 562.50 per life-year gained.
640 life-years 640
45
e. If 1000 patients are treated with Thrombase, 50 lives are saved; if 1000 patients are treated
with Klotgon, 80 lives are saved; therefore, 30 lives are saved by treatment with Klotgon
rather than Thrombase.
Assuming an increase in survival time of eight years per patient, 30 x 8 = 240 life-years
are gained.

ICER (Klotgon versus Thrombase for 1000 patients)

= (1000 x $1000 – 1000 x $200) = $800 000 = $3 333 per life-year gained.
240 life-years 240

4.3. Unfractionated heparin versus low-molecular-weight heparin

a. Relative risk = 19.8% / 23.3% = 0.85.
b. Risk difference = 19.8% / 23.3% = 3.5%
Number of patients who needed to be
treated = 1/0.035 = 29 patients.

c. ICER (1000 patients) = (1000 x 72.20) – (1000 x 27.09) = $45 110

3.5% x 1000 35
= $1 288.86 per event avoided.

d.. ICER (1000 patients)

= (1000 x $72.20) – (1000 x ($27.09 + 5 x $12.40)) = –$16 890
(1000 x 23.3%) – (1000 x 19.8%) 35

Low-molecular-weight heparin is dominant. It is both cheaper and more effective than unfrac-
 tionated heparin when monitoring costs are included.

When a drug is dominant, it is not appropriate to calculate an ICER, as this can produce spuri-
ous results. Why do you think this is?

4.4. Celecoxib versus diclofenac

a. Relative risk = ((38 + 11) / 212) / ((74 + 23) / 218) = 23% / 44% = 0.52
b. Risk difference = 23% – 44% = –21%
Number of patients who have to be treated to prevent a single event = 1/0.21 = 5 patients.
c. Dose of celecoxib = 400 mg/day. One pack contains sufficient drugs for 15 days of treat-
ment. The duration of treatment is 24 weeks = 168 days. Therefore, 168/15 = 11.2 packs
are required at a cost of 11.2 x$50 = $560 per patient.

Dose of diclofenac = 100-150 mg/day. Assume a conservative dose of 100 mg/day.

One pack contains sufficient drugs for 25 days of treatment. The duration of treatment is 168
days. Therefore, 168 / 25 = 6.72 packs are required at a cost of 6.72 x $11.60 = $77.95 per
patient.

46 ICER (1000 patients) = (1000 x $560) – (1000 x $77.95) = $482 050

440 – 230 210
= $2 295.48 per ulcer avoided.

d. Incremental cost per ulcer or erosion avoided

ICER (1000 patients)

= (1000 x $560 + 1000 x 23% x 1% x $1434) – (1000 x $77.95 + 1000 x 44% x 1% x $1434)
(1000 x 0.44) – (1000 x 0.23)

= $479 038.60 = $2281.14 per ulcer or erosion avoided.

210

Incremental cost per hospitalization avoided

ICER (1000 patients)

= (1000 x $560 + 1000 x 23% x 1% x $1434) – (1000 x $77.95 + 1000 x 44% x 1% x $1434)
(1000 x 0.44 x 0.01) – (1000 x 0.23 x 0.01)

= $479 038.60 = $228 113.20 per hospitalization avoided.

2.1
Incremental cost per death avoided
ICER (1000 patients)

= (1000 x $560 + 1000 x 23% x 1% x $1434) – (1000 x $77.95 + 1000 x 44% x 1% x $1434)
(1000 x 0.44 x 0.01 x 0.1) – (1000 x 0.23 x 0.01 x 0.1)
 = $479 038.60 = $2 281 136.20 per death avoided.
0.21

4.5. Oral montelukast versus an inhaled steroid

a/b. There is no “right” answer to this question. What do you think?
c. Beclometasone is both cheaper and more effective than montelukast. Therefore
beclometasone is dominant.

Chapter 5: Drug classification systems

1. It is appropriate to assign an additional ATC code in N05A (Antipsychotics) for oral for-
mulations of «Neurol», because the main indications for the parenteral and oral formula-
tions differ. A medicinal product can be given more than one ATC code if it is available
in two or more strengths or formulations with clearly different therapeutic uses (see
Guidelines for ATC classification and DDD assignment. Oslo, Norway, WHO
Collaborating Centre for Drug Statistics Methodology, version 2003.)

2. It is not appropriate to assign an additional ATC code in N04 (Anti-Parkinson drugs) for
lisuride because the dosages overlap with those used in prolactin inhibition.
47

Chapter 6: Drug utilization metrics and their applications

1. Substance A: 250 mg
Substance B: 300 mg

2. To make the sales figures comparable, it is important to recalculate the figures to reflect
the same DDD version. The most recent DDD version should always be used (i.e. 0.8
mg for budesonide inhalation powder). Recalculation of the 1990 sales figure with the
updated DDD gives 3.6 DDDs/1000 inhabitants/day.

3. Four-day courses, substance A: 1988: 0.43 million courses; 0.1 courses/inhabitant

1996: 2.3 million courses; 0.57 courses/inhabitant

Eight days courses, substance B: 1988: 2.7 million courses; 0.68 courses/inhabitant
1996: 1.2 million courses; 0.31 courses/inhabitant
 Acknowledgements

This manual has been prepared by the WHO CEA: cost-effectiveness analysis
Collaborating Centre for Drug Statistics
Methodology and the WHO Collaborating CMA: cost-minimization analysis ICER: incre-
mental cost-effectiveness ratio
Centre for Drug Utilization Research and
Clinical Pharmacological Services. Particular CUA: cost-utility analysis DURG: WHO
thanks are extended to all the members and European Drug Utilization Research Group
observers of the International Working Group on
Drug Statistics Methodology and the staff of the DDD: defined daily dose
WHO Collaborating Centre for Drug Statistics DU90%: drug utilization 90%
Methodology and in particular to Professor Don
Birkett, Professor Peter de Smet, Professor EPhMRA: European Pharmaceutical Market
David Ofori-Adjei, Dr Ingrid Trolin, Professor Research Organization
Ulf Bergman, Hanne Strøm, Bente Tange Harbø
IMS: International Medical Statistics
and Marit Rønning. A special thank you to
Professor Folke Sjöqvist for his excellent contri- MAO: monoamine oxidase
bution to the editing of the manuscript.
NSAIDs: nonsteroidal anti-inflammatory drugs
List of abbreviations
QUALY: quality-adjusted life-year
48
ACE inhibitors: angiotensin-converting enzyme SSRIs: selective serotonin reuptake inhibitors
inhibitors
TCAs: Tricyclic antidepressants
AT: anatomical therapeutic (classification)

ATC: anatomical therapeutic chemical (classifi-

cation)