|
Abstract:
|
Simon's 2-stage design is a single-arm design, in which null hypothesis (H0) that response rate, p, is at most p0 is tested against alternative (H1) that p is at least p1. If there are fewer than k1 responders out of n1 enrolled in Stage 1, trial is terminated to accept H0. If at least k1 responders are observed, n2 additional subjects are enrolled in Stage 2. If there are fewer than k responders at the end of Stage 2, H0 is accepted; otherwise, H1 is accepted. For given Type I and II errors, design parameters n1, n2, k1, and k are estimated to meet certain optimality criteria. Without a control arm, results could be unreliable, especially when p0 and p1 values are unrealistic. We propose a 2-stage design with a control arm, and provide a SAS program to calculate optimal solutions for sample size and critical values at Stages 1 and 2 for given Type I and Type II errors, if they exist.
|
