Early selection in a randomized phase II clinical trial.

A randomized phase II clinical trial design can be employed when one wishes to be able to select one of several similar therapies or variants of a therapy for inclusion in a subsequent, definitive, phase III trial. It is not necessary in this type of trial to formally identify a superior arm using the usual parameters and stringent criteria employed for hypothesis testing. Simon, Wittes and Ellenberg have described methods and sample sizes for selecting the superior arm from among k possible arms. In this paper, we describe an approach based on statistical selection theory which allows one to potentially make a decision to end accrual to a randomized phase II trial at an interim point of the trial, and to select one of two arms as being worthy of further evaluation in a subsequent study. This method requires that an adequate gap in the number of responses between the two arms be observed at the interim point in order to limit the probability that the selected arm is actually inferior by more than an indifference amount. Published in 2002 by John Wiley & Sons, Ltd.