challenging but vital during the early period after
launch. The assessment of pharmacovigilance
findings, particularly the initial few reports of
AEs with low incidence, has obligatorily to include
an epidemiological component. The epidemiolo-
gical interpretation of a finding of excess preva-
lence requires the sophisticated epidemiological
understanding of baseline, or population-expected
risks and application of this understanding to the
vagaries of uncertain ascertainment which charac-
terize spontaneous reports. The usual disposition
of a ‘signal of a potential problem’ from pharma-
covigilance is introduction into the product’s core
safety information (‘product label’). However,
further, recent product withdrawals provide a
litany of more extreme example that need not be
repeated here, but average one or two major pro-
ducts and several minor ones each year. Product
withdrawals are often misunderstood, particularly
by the lay press hungry for a scandal. It is neither
feasible nor desirable to ‘know everything about a
product’ at the time of approval. But judicious
product withdrawal, based on substantial evidence
properly collected and analyzed in the post-mar-
keting environment, is a classic example of a robust
and balanced system, with each component func-
tioning as it should. Pharmacoepidemiology con-
tributes to the pursuit of the best-informed decision
making, with the shared goal of optimization of the
balance between patient benefit and the inevitable
patient adversity.
Prescription-event monitoring (PEM)
This is essentially an extension of traditional,
hands-on epidemiology, which assembles all
patients that are prescribed a drug into a cohort
which is then followed. In the United Kingdom, for
example, through the Drug Safety Research Trust,
all or a sample of this cohort is assembled from the
records of the prescription pricing authority, gen-
erally within the first year or two of initial market-
ing of the product. Each patient can be followed up
with a confidential enquiry for serious AEs using a
form that, in the United Kingdom, is popularly
called by its appearance: the ‘Green card’ (this
term has an entirely different meaning in the
United States, and, curiously, describes a pink
document!). It is a classic example of an observa-
tional as opposed to an experimental method, in
which all uses and all outcomes (events) are
observed, generally without a simultaneously col-
lected comparison population. Thus, data stem-
ming from these sorts of activities are fraught
with analytical and methodological traps. How-
ever, PEM is a good method for generating hypoth-
eses for further testing, usually after reconciliation
with the known pharmacology of the drug of inter-
est, other drugs in the same class, and the natural
history of the disease and kindred disorders. These,
of course, shed further light on these data, and may
be gathered, for example, from the spontaneous
reports system. Indeed, sometimes this is also the
first evidence of an unsuspected drug intolerability,
perhaps in a previously unsuspected subset of the
treated patient population. The ability in certain
European areas and New Zealand to aggregate
prescriptions from entire countries or regions,
often as part of the reimbursement system, is
obviously strategic to this approach.Pregnancy registries
Pregnancy registries (or, more properly, preg-
nancy follow-up studies) are being recommended
with increasing frequency for products that are
likely to be used in women with child-bearing
potential. Inevitably, all new drugs have not
been studied in women who are, or become, preg-
nant, and equally inevitably, labeled warnings to
that effect do not prevent exposures of embryos
and fetuses to new drugs. The anticipated, spon-
taneous incidence of anomalies detectablepost
partumis in the range 3–7%. This wide range is
cited because of the wide variations in criteria such
as severity (e.g. is a minor birthmark a ‘birth anom-
aly’?), degree of scrutiny (follow-up until the age
of four years or beyond is needed to detect some
anomalies; some types of inguinal hernia present-
ing in adulthood are even thought to be congenital),
geography, concomitant disease or toxin exposures
(including tobacco, illicit drugs and alcohol), and
socioeconomic status. The key to a successful
pregnancy registry is that pregnancies should be24.8 THE EMERGING WORLD OF RISK MANAGEMENT 309