Abstract

Over the past few decades, clinicians have been forced to deal with a revolution in the proposed procedures for the progression of a drug through phase II trials. Historically, a single-arm phase II design, in which all patients receive the experimental treatment, was employed as the principal way to determine whether to proceed to a randomised controlled phase III trial. However, single-arm trials are commonly associated with high type-I error rates, as well as biased estimated treatment effects. As a result, it has become increasingly clear that this classical phase II design performs poorly in predicting the likelihood of phase III success. Consequently, there has been much interest recently in the use of randomised phase II designs, and a more commonly accepted phase II development plan today has become to utilize a single-arm designed trial, followed by a pilot randomised study if the null hypothesis is rejected at the single-arm stage. In my talk, I will detail how the most common single-arm and randomised two-arm trials are optimally designed and the roles each play today, before discussing a simple way in which the expected sample size of the single-arm followed by randomised two-arm development plan may be reduced over current practice.