A Note on the Promising Zone Approach in Adaptive Trial Design

Abstract

The promising zone approach in adaptive trial design has gained popularity since its inception due to the use of the conventional test statistic at the final analysis stage rather than a weighted version that seemingly penalizes the second stage data when the adaptive trial strategy activates a sample size increase. However, this perceived advantage suffers loss of efficiency. This article is to show through mathematical derivation that the weighted test statistic by Cui, Huang, and Wang (CHW) is uniformly more powerful than the Mehta and Pocock method in the statistical promising zone. Due to relatively small chance to fall into the statistical promising zone, in practice, a wider zone defined by the minimally clinically meaningful treatment effect and the target conditional power could be used. The proposed approach is a hybrid approach by which the CHW test statistic is applied within a chosen promising zone (could be wider than the statistical promising zone), outside which the conventional test statistic is applied without a sample size increase. Simulation studies are recommended to facilitate the application of this hybrid approach.