Sample size tables for exact single‐stage phase II designs

Abstract Tables for single‐phase II trials based on the exact binomial distribution are presented. These are preferable to those generated using Fleming's design, which are based on the normal approximation and can give rise to anomalous results. For example, if the upper success rate is accepted, the lower success rate, which the trial is designed to reject, may be included in the final confidence interval for the proportion being estimated. Copyright © 2001 John Wiley & Sons, Ltd.