Methodology for sample size calculation for phase III clinical trials is well established and widely used. In contrast, for earlier phase clinical trials or pilot studies, although there is an acceptance that the methods used for phase III trials are not appropriate, there is little consensus over methods that should be used. This paper explores this problem from a Bayesian decision-theoretic perspective. The aim is to obtain sample sizes that would be appropriate for studies funded by a large funder such as a public sector body or major pharmaceutical company. The sample sizes obtained are optimal in that they minimise the average number of patients required per successfully identified effective therapy or equivalently maximise the number of effective therapies successfully identified over a long period. It is indicated that the number of patients included in a phase II clinical trial should be approximately 0.03 times that planned to be included in the phase III study. This is similar to that proposed by other researchers in this area, though rather smaller than actually used for many phase II trials.
Copyright © 2011 John Wiley & Sons, Ltd.