These are thus measures of incidence. The propor-
tion of a population that will experience at least one
seizure or one migraine attack in their lives is a
measure of incidence and would likely be
expressed as a number per thousand (or per hun-
dred thousand) person-years, whereas the propor-
tion of a population suffering from epilepsy or
migraine during the year 2000 is an expression of
prevalence.
In pharmacovigilance terms, the ‘true fre-
quency’ in a treated population in a specified per-
iod, if it was known, of an AE observed in a
marketed product, would be considered an inci-
dence. All too often, the frequency of reported AEs
(definitely not the complete or even estimated
numerator), perhaps weighed against known sales
(scarcely a true denominator), is mistakenly used
to calculate a rate and called an ‘incidence’. At
best, such spontaneous reports data should be
termed ‘reports rates’.
Other, more complex terms are defined and
described in standard textbooks of epidemiology
and statistics (q.v.) and included in two excel-
lent lexicons, the Dictionary of Epidemiology
(Last, 2001) and, more recently, a very useful
Lexicon from the International Society for
Pharmacoeconomics and Outcomes Research
(ISPOR).
24.6 Epidemiology in drug
developmentThe complexities of drug development include a
decision web that is inevitably informed by incom-
plete information. Although past, focused research
may comprise some of the information for the next
step, epidemiological information can be of valu-
able assistance. The capturing and extension of
population-based studies, often concerning the nat-
ural history of disease rather than the pharmacolo-
gical properties of the test agent itself, can guide
the choice of indication, market strategy and even
the viability of an entire project. Furthermore, the
place of existing therapies, in the context of the
natural history of disease, can also be investigated
epidemiologically. ‘If they don’t need it, we can’t
sell it; then let’s not pursue it’ is an aphorism: but
whether they need it is, of course, an epidemiolo-
gical challenge.
Population-based measures of burden of disease
involve a formal quantitation of the opportunity
for a new drug. These measures vary among
organ systems, but typically involve the interaction
of lifestyle interference, duration of disease, pre-
valence, incidence, effectiveness and adverse
effects associated with existing therapies, and
reduction in lifespan. Such objective measures
can be ascertained from population-based studies
and existing national databases, for example from
major ongoing population health surveys, and can
often allow the pharmacoepidemiologist to contri-
bute a substantial and useful evidence base to
inform the difficult and emotion-laden decisions
which must be made by senior executives in drug
development.
During phases II and III, an additional capability
can be offered to the development team that might
hitherto be comprised purely of clinical depart-
ment staff – Are infrequently observed but highly
dangerous AEs being seen in a clinical trials pro-
gram within the expected range for that study
population? If so, then entire development pro-
grams in jeopardy could be saved; if not, appro-
priate actions may be undertaken more rapidly and
decisively.
Under some circumstances, in the United
States, widespread distribution of an investiga-
tional agent prior to NDA approval, involving
large-scale populations, is permitted. Under
these conditions it is, of course, necessary to
monitor safety in such broader use, and usually
with greater scrutiny that might ordinary apply
after product approval. Thus, the best practice is
to structure such programs as observational stu-
dies. AEs are bound to occur; thus providing an
ideal opportunity for early detection of infrequent
but important adverse reactions; conversely, trou-
ble-shooting these, in the context of a sound epi-
demiological and clinical understanding of AEs
associated with the disease itself, and with alter-
native therapies, is also often needed to protect
against false conclusions. Such interpretations
also eventually are translated into labeling, either
by exclusion or inclusion.24.6 EPIDEMIOLOGY IN DRUG DEVELOPMENT 307