Hybridization of conditional and predictive power for futility assessment in sequential clinical trials with time-to-event outcomes: a resampling approach.

Having the potential to stop a clinical trial early for futility during an interim analysis is an important feature of sequential clinical trials. Conditional power and predictive power methods are widely used to analyze interim data in determining whether a trial is likely to eventually demonstrate both statistically significant and clinically meaningful results. Herein we introduce a new approach to early stopping in clinical trial designs, which can be viewed as a hybridization of the conditional and predictive power methods and has both methods as its special cases. The proposed hybridization, along with a novel resampling method, can be used together to assess futility in sequential trials with time-to-event outcomes. The proposed approach to futility assessment applies to sequential clinical trials with interim analyses planned at either fixed calendar times or number of events. In addition, it does not resort to asymptotics and applies to general test statistics of either single or multiple dimensions.