Using Surrogate Endpoints in Adaptive Designs with Delayed Treatment Effect

Abstract

Delayed treatment effects have been commonly observed in clinical trials, which bring more challenges to the interim decision making particularly in adaptive designs setting. An improper interim analysis (IA) may falsely stop a promising study based on the traditional conditional power (CP) approach assuming the observed treatment effect will carry over for the entire study. For such scenario, a short-term surrogate endpoint which is predictive of the primary long-term outcome can be extremely useful for a more accurate CP calculation and adaptative decision. In this article, we propose using a surrogate endpoint in the IA to improve the CP calculation in designing an adaptive sample size reestimation or event size reestimation study. Through theoretical derivation and extensive simulations, we show that our proposed approach demonstrates the practical feasibility and benefits of using a surrogate endpoint for adaptive designs with delayed treatment effects. The average overall power is shown to be significantly higher than conventional event size reestimation and group sequential design when there is a delayed treatment effect in primary survival endpoint. We also demonstrate proposed approach in a case study of Phase III non-small cell lung cancer (NSCLC) trial with delayed treatment effect. Finally, we give recommendations on how this method could be implemented in confirmatory clinical trials.