GROUP SEQUENTIAL DESIGNS FOR ONE-SIDED AND 2-SIDED HYPOTHESIS-TESTING WITH PROVISION FOR EARLY STOPPING IN FAVOR OF THE NULL HYPOTHESIS

In this paper we consider a class of boundaries for group sequential clinical trials that allow for early stopping also when small treatment differences are observed. They are based on the one-parameter family of boundaries introduced by Wang and Tsiatis (Biometrics 43 (1987) 193-199), can be derived exactly for any choice of type I and type II error probabilities, and can be easily applied to both one- and two-sided hypothesis testing. Plans corresponding to different choices of the design parameters are compared in terms of average sample number. Tables of relevant parameters are provided to help in the design of group sequential tests.