Multiple‐arm superiority and non‐inferiority designs with various endpoints

Multiple‐arm dose–response superiority trials are widely studied for continuous and binary endpoints, while non‐inferiority designs have been studied recently in two‐arm trials. In this paper, a unified asymptotic formulation of a sample size calculation for k ‐arm ( k >0) trials with different endpoints (continuous, binary and survival endpoints) is derived for both superiority and non‐inferiority designs. The proposed method covers the sample size calculation for single‐arm and k ‐arm ( k ⩾2) designs with survival endpoints, which has not been covered in the statistic literature. A simple, closed form for power and sample size calculations is derived from a contrast test. Application examples are provided. The effect of the contrasts on the power is discussed, and a SAS program for sample size calculation is provided and ready to use. Copyright © 2006 John Wiley & Sons, Ltd.