the natural progression of disease is unknown, then
it can be ethical to recruit patients into a study with
inclusion criteria that include that they arealready
being treated with the drug of interest. Almost any
of the designs discussed above may then be used,
where patients are randomized either to remain on
the treatment of interest or to be withdrawn from
that treatment. All the usual needs for precisely
defined prospective end points and sound statistical
advice before starting the study apply.
Early-phase clinical trials in patients with
cancer often use a two-stage design that has been
promoted by Gehan and others (Gehan, 1979;
Ellenberg, 1989). With progressive, fatal diseases,
the problem of preventing an untoward number
of patients from being treated with a useless ther-
apy increases. These two-stage designs usually
include a small number of open-label treated
patients (usuallyn14) in the first stage. The
proportion and degree of tumor responsiveness
are then used to fix the number of patients in
the second stage of the design which may use an
active comparator or no therapy as the alternative
treatment, depending upon whether an active
comparator therapy can be identified. Such studies
cannot produce fundamental evidence of efficacy,
but in the hands of experienced statisticians and
development teams can predict whether wider
trials are justified.
9.13 Stopping clinical trials
Safety issues
Stopping a clinical trial because of an emergent
safety problem, either by a medical monitor or by a
safety committee, is always a unique situation.
Little useful, generalizable guidance can be pro-
vided here. These are decisions that are always
taken in consultation, and the safety of potential
future trial recruits must be the paramount concern
(including the abrupt cessation of therapy). Trial
suspension is usually the best immediate option,
allowing time for collective thought, notification of
regulatory authorities and wider consultations as
appropriate.
Efficacy issues
Pocock (1992) has succinctly summarized most of
the situations that obtain when it is considered
whether to stop a clinical trial. Efficacy, like safety,
can cause ethical concerns to the pharmaceutical
physician when he or she suspects that patients
will be exposed to alternative therapies which are
suboptimal.
Interim efficacy analyses usually make a mess!
These analyses require either that the overall size
of the trial has to be greater than if no interim
analysis was performed, or that a smalleramust
be accepted as indicating statistical significance at
the end of the whole study.
Pharmaceutical physicians will hear loud com-
plaints about these drawbacks of interim analyses,
especially from senior management with purely
commercial backgrounds. Everyone will want to
know as soon as possible whether ‘the drug is
working’, but lax scientific thinking is behind
these complaints. Common statements are: ‘We
don’t want to stop the study at the halfway stage,
we just want to see how it is going’. When asked
why, the answer is usually something like: ‘There
would be no point in spending more money on the
study if there is no chance of achieving a statisti-
cally significant result’. This is a popular mis-
rationalization: the decision not to stop a study is
a decision to allow it to continue. Any interim
decision introduces a bias on the dataset that is
eventually analyzed.
Spectacularly effective drugs may achieve a
very smallaat the time of the interim analysis.
Stopping the trial by reason of the unethical basis
for treating the patients with anything else is a rare
and pleasant event for the clinical trialist. However,
in that spectacular success, the pharmaceutical
physician should ask whether a minimization
design would have achieved the same thing
with even fewer patients, and thus actually feel
chastened.
It is not the purpose of this chapter to delve into
the mechanics of statistics. However, a few com-
ments about the relationships between values fora
at the stage of an interim and complete statistical
analysis of a clinical trial may be in order. There are
several statistical pointsofviewonthis subject, and9.13 STOPPING CLINICAL TRIALS 113