Case–control study designs
Thesewere developed to provide information more
rapidly than when cohorts are followed for pro-
longed periods of time, using traditional hands-on
methods; case-based research is, however, neces-
sarily retrospective. Analysis begins with the char-
acterization of a group of people that already have
the disease of interest, the ‘cases’. Control subjects
are then drawn from a population with exactly the
same attributes as that from which cases are
selected (often a very difficult task!). The antece-
dent demographic, therapeutic and environmental
factors of both groups are documented, often by a
combination of record abstraction and interview. If
differences are found in the proportion (rate) for
some factor between the two groups, then this
becomes suspected as an etiological agent for the
disease of interest. This suspicion is strengthened
when either the discovered factor corresponds with
a predictive hypothesis at the start of the study, or
when there is consistent evidence that would sup-
port its identification (perhaps a biochemical link
between the factor and the disease).
24.3 Drug risk as an
epidemiological problemDrug-related epidemics have occurred, mercifully
relatively infrequently. However, with each unfor-
tunate episode, there is inevitably a variety of reg-
ulatory and clinical fallout. Indeed, the illnesses
associated with ingestion of glycol-tainted linctus
led to the Food, Drugs and Cosmetics Act in
the United States, and the disastrous association
of phocomelia with thalidomide propelled reforms
of drug regulations worldwide. Other famous
examples include, of course, practolol-induced ocu-
lomucocutaneous syndrome, and, more recently,
fenfluramine-induced myocardial fibrosis, isotreti-
noin-associated birth defects and unexpected
heart complications with long-term use of Cox-2
inhibitors.
The major driver for the field of pharmacoepi-
demiology is the nature of the drug development
process itself. Relatively small and often quite
carefully selected clinical trials populations are
followed for only limited periods, during and
after exposure to the agent under study, in the
populations that comprise typical product license
applications and NDA safety summaries. This
leaves, for the post-approval scientific environ-
ment, the challenge to apply methods that can
detect AEs with relatively low frequency or rela-
tively specific risk situations. Those who call for
transfer of these burdens to the pre-approval envir-
onment would benefit from training in epidemiol-
ogy, with the associated understanding that the
only way to understand the real world is to study
the real world!
Pharmacoepidemiology and pharmacovigilance
do not pretend to be able to eliminate the occur-
rence of drug-associated epidemics. The challenge
is to detect and quantitate problems as rapidly and
accurately as possible, so that changes in the ben-
efit–risk balance, as understood at the time of
approval, can be quickly recognized and possible
public health actions considered. Thus, pharma-
covigilance may be understood as ‘epidemiologic
intelligence’. And thus, in turn, the physician phar-
macoepidemiologist is a strong contributor to drug
surveillance departments in industry and drug
safety groups in regulatory groups. Typically,
these epidemiologists will be supervising and/or
providing expert counsel to groups of less specia-
lized or highly trained health scientists who imple-
ment the day-to-day running of these programs.
Teamwork becomes an indispensable skill.
Consideration about the need for and technical
considerations regarding one or more structured
observational studies is a frequently recognized
contribution of the epidemiologist in this enterprise.
Less frequently extolled are the great contributions
which epidemiologists make to consideration of
approaches which, seductive because of their appar-
ent simplicity, willnotbe likely to contribute, as
options forreducing uncertaintyaroundestimates of
possible risk are considered. Observational science
is very complicated, and the opportunities of failure
of study are considerable! When epidemiological
studiesareundertaken,andresultsareknown,itfalls
to the physician epidemiologist to put on the public
health hat and recommend whether an intervention
in the interests of public health might be needed,24.3 DRUG RISK AS AN EPIDEMIOLOGICAL PROBLEM 305