A clinical trial design with a fixed maximum number of failures

Abstract

Fushimi has recently analyzed a play-the-winner technique for choosing the better of two dichotomous treatments, in which the maximum number of failures allowed in the trial is fixed in advance. Hare we examine the analogous design when sequentially arriving patients are assigned in pairs to the two treatments. We show by specific numerical calculations that neither design is uniformly better than the other over the entire range of success probabilities.