A New Approach to Futility Stopping in Clinical Trials

• Activity Number: 666
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Pei He*+ and Tze Leung Lai and Olivia Yuehwen
• Companies: Stanford University and Stanford University and Stanford University
• Keywords: sequential design ; futility stopping ; survival analysis ; clinical trials

Abstract

Early stopping due to futility, or go/no-go decisions, during interim analysis has become an important feature of clinical trial designs. Current methods for futility stopping in the literature are mostly based on conditional power or predictive power in conjunction with the theory of stochastic curtailment or group sequential design. They have certain drawbacks, especially when the test statistics are nonparametric or when the trial involves survival data that are monitored at pre-specified calendar times. Herein we introduce a new approach to futility stopping in clinical trial designs that can overcome these difficulties. Simulation studies and theoretical analysis show that the method developed is nearly optimal in both parametric and nonparametric problems. Moreover, due to its flexibility, the approach can circumvent the difficulties of "information time" versus "calendar time" in time-sequential clinical trials to compare the failure times of two treatments.