as a matter of routine, might occasionally lead to a
superior trial design.
9.15 Series of published cases
Some diseases are so rare that the prospects of
conducting a clinical trial are remote. It is unlikely
that enough patients could ever be collected at any
reasonably small number of study sites for any
useful randomization. These diseases may be
found in the literature as case reports. In these
cases, probably the best that can be accomplished
is to collect and retrospectively analyze as many
such cases as possible. If the drug of interest has
been used in a sufficient number of patients, then
retrospective risk ratios for benefit and harm can be
calculated. This may be the strongest evidence that
can ever be collected about a particular drug under
these rare conditions, albeit never as strong as a
controlled clinical trial. One example is the effec-
tiveness of dantrolene in malignant hyperthermia
(Strazis and Fox, 1993).
9.16 Objectives and prerequisites
of pivotal clinical trialsLicensing requirements typically are greater than
reporting data from multicenter ‘phase III’studies.
Specialpopulationsmay requiresmall-scalestudies
to supplement a traditional two-study, large-scale
registration development scheme. Similarly, if (in
the United States) the proposed indication has an
approved Orphan Drug designation, then small-
scale ‘phase II-type’ studies may be all that is
possible due to disease rarity. Furthermore, even
for conventional indications, the resource implica-
tions of pivotal studies are usually much greater
thananyearlier phaseofdevelopment,and efficient
resource utilization becomes exponentially more
important than before. The incorporation of phar-
macoeconomic and humanistic outcomes along-
side the primary registration end points is
becoming essential, and preparatory work is best
done in conjunction with the smaller, earlier stu-
dies and must also factor treatment compliance.9.17 Benefit–risk analysis
The cumulation of all the data from the clinical
trials of a new drug product, assuming a fairly
orthodox regulatory strategy for a typical dossier
or NDA, will form the largest fraction of the appli-
cation. However, these data are also needed for
derivative documents within the application, one
of which is a benefit–risk analysis, which forms the
last part of an Integrated Safety Summary (Section
9 of the NDA), and is a central objective of the
expert report in European applications. These
benefit–risk assessments must be derived from
the clinical study reports and summaries elsewhere
in the applications.
All clinicians constantly weigh benefit–risk in
their daily practice. Their assessment of this ‘ratio’
in everyday practice, using approved drugs, is
usually not as numerical as it sounds. In practice,
clinicians make prescribing decisions based upon
(a) a subset of the published information that might
be available about the drug (labeling, drug repre-
sentatives, comments from colleagues, etc.), (b)
their current and prior experience with this parti-
cular patient and (c) prior experience with other
patients. This prior experience, even if personal,
may or may not be consciously recalled. Further-
more, we all operate algorithms taught us by others
whom we respect, and thus we use others’ experi-
ence with drugs and patients, quite apart from the
often hard-learned lessons from our own therapeu-
tic adventures (pace‘evidence-based medicine’).
Clinical trialists also weigh benefit–risk every
time a protocol is written. Often, unlike for
approved drugs, there is much less information to
go on. In early clinical development, extrapola-
tions are obligatory. However, unlike in general
medical practice, these extrapolations are often
not from clinical experience, but rather from phar-
macokinetic models or animal data, or at best from
patients who are clearly dissimilar from that pro-
posed in the new trial. This is obligatory: if the9.17 BENEFIT–RISK ANALYSIS 115