Principles and Practice of Pharmaceutical Medicine

answers to the clinical trial questions were known,
then there would be little point in doing the trial.
There are highly mathematical approaches to
benefit–risk assessment. When a single (binary)
end point of interest can be balanced against a
single adverse event of concern, then the number
of patients required and the number of required
therapeutic events can be defined, and the confi-
dence intervals can be calculated to examine what
the true benefit–risk ratio might be (e.g. for
GUSTO, Willanet al., 1997). The number needed
to treat, number needed to harm (and correspond-
ing reciprocals) can be used to compare drugs for
this purpose. However, this is a highly unusual and
artificial situation, and the sophisticated statistical
answers that result are unlikely to have more than a
partial impact on the more nonnumerical approach
taken by clinicians.
Usually, however, the clinical trialist has to stick
out his or her neck, based upon a highly personal,
nonnumerical assessment of benefit–risk. The
highly mathematical approaches usually work
best in retrospect, and this is the situation neither
of the clinician who must decide whether to pre-
scribe nor the clinical trialist who must decide
whether to commit patients to a particular study
design, both being prospective decisions. Further-
more, both in clinical trials and general medical
practice, it is a rare situation where the benefit to
the patient arises from a single binary variable, and
there are no drugs which possess a single type of
adverse event, whose probability may be confi-
dently, prospectively estimated for any given
patient.Eventhe simplestcase,a drug withsubstan-
tial history and experience, cannot fit the contrived
mathematical approach described above. Penicillin
has three adverse events of primary interest (ana-
phylaxis, bacterial drug resistance and sodium load
at high doses). The mechanism by which infection
recedes, if it is to recede, is only partly due to the
action of the drug, because the extreme variability
introduced by the concomitant condition of the
patient. Whether to prescribe penicillin is a com-
mon decision for doctors and dentists: the mathe-
matical analysis of the benefit–risk ‘ratio’ is
unlikely to affect most prescribing decisions.
The informed consent document is where we
ask patients to make their own benefit-risk assess-

ments, albeit with some guidance (Marsh, 1990).

Certainly, the mathematical approach cannot

be expected on the part of the patient nor will it

be useful in a balanced and fair communication

with the patient about the nature of the clinical

trial.

Benefit–risk, then, is a central part of the prac-

tice of pharmaceutical medicine and its regulation.

It can almost never be reduced to a numerical

exercise. Benefit–risk assessments of clinical trial

data are an important part of all new drug applica-

tions. Good people will differ in their benefit–risk

assessment even when using the same body of

clinical trials data.

9.18 Summary

This chapter has attempted to provide a philosophy

of clinical trials. The place of clinical trials in the

overall development plan and what the clinical

trialist mustknow aboutrather than be able to

actually implement himself or herself has been

emphasized. Almost all clinical trials are unique

because of the infinite combinations of hypothesis

to be addressed, pharmacological properties of the

drug under investigation, the types of patients who

are likely to be available and likely users of the

resulting data. The major categories of trial designs

have been surveyed in some detail; it is hoped that,

when challenged with testing any clinical hypoth-

esis, a good clinical trialist would consider all these

broad categories, select that most relevant to the

clinical situation and then refine the proposed trial

design from that point. Some of the subtle

interactions between statistical, financial and psy-

chological aspects of trial design have been hinted

at. The clinical trialist will only really grow in

this discipline through experience and good

mentorship.

References

Amin HM, Sopchak AM, Esposito BF, Henson LG,

Batenhorst RL, Fox AW, Camporesi EM. 1995.

Naloxone-induced and spontaneous reversal of

depressed ventilatory responses to hypoxia during

116 CH9 PHASE II AND PHASE III CLINICAL STUDIES