Methods to Minimize Power Loss at Interim Futility Analyses and Final Analyses Due to Violations of Sample Size Assumptions in Clinical Trials with Survival Endpoints

• Activity Number: 350
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Joseph Adair*+
• Companies: PPD
• Keywords: Survival ; Power ; Futility

Abstract

In clinical trials with a survival endpoint where interim analyses include a futility component with a one-sided boundary, power may be diminished. Unlike symmetric superiority/futility bounds, one-sided futility bounds can be set without increasing the chance of stopping for superiority. This may be done to appease a regulatory agency for which it is desirable to have a greater chance to stop for futility if the treatments do not differ. When futility bounds are one-sided, power loss associated with stopping for futility is not accounted for by many commonly used software packages. The log-rank test relies upon an assumption of proportional hazards and is generally assumed to be robust to small departures from this assumption. Through simulation we show that the log-rank test is not robust to departures of this assumption. Specifically, given a delay in the onset of separation between survival curves, proportional hazards does not hold and standard software packages underestimate the number of events needed, decreasing power. Methods to minimize power loss are examined. A case-study and several simulations illustrate the issue and solutions are discussed.