A sequential phase II/III trial for binary outcomes.

Abstract

This paper describes a simple two-stage design for randomized clinical trials with a binary outcome. An early stopping criterion applies to results in the experimental arm only and seeks to evaluate that arm with respect to some external standard of efficacy. Stopping rules are similar to those commonly used in phase II designs. This design has no adverse impact on the overall type I error rate, though one must increase sample size to accommodate a modest loss in power.