Using
cross-over designs
The fact
that carry-over is a potential problem for cross-over trials is
not a reason to not use them. All scientific progress requires
experiments in which unverifiable assumptions are made. (Although
these assumptions may, in due course, form the basis of further
experiments.) The cross-over trial should be employed with caution
but is nonetheless extremely useful. In the context of drug development,
it is rarely the design of choice in phase III, because in those
conditions for which a cross-over trial might be possible, it
is usually the case that the sample sizes needed to establish
safety of a treatment are larger than needed for efficacy. The
usual advantage of smaller sample sizes associated with the cross-over
trial would not then be relevant.
Cross-over
trials are extremely useful however in phases I and II. The danger
of declaring an ineffective treatment effective because a cross-over
trial has been used is minimal, because carry-over would, if it
occurred, usually bias the effective of treatment downwards. The
potential danger is that a useful treatment may be missed. The
suitability of the cross-over thus becomes a matter of judgement
for the trialist and is hardly likely to concern the wider scientific
community. In any case, the phase III trials to come will deliver
the final verdict to the interim judgement of the cross-over.
Without
cross-over trials dose-finding would be much more onerous in many
areas. Cross-over trials are also supremely useful for establishing
individual response to treatment. This is considered in the next
section.
|