Sample Size Re-Estimation at Interim Analysis for a Time-to-Event Endpoint for Data with Nonproportional Hazards

• Activity Number: 534
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Liang Chen*+
• Companies: Pfizer, Inc.
• Keywords: Sample size re-estimation ; Interim analysis ; Time-to-event endpoint ; Non-proportional hazards ; Conditional power ; Piecewise exponential model

Abstract

Randomized clinical trials commonly include one or more planned interim analyses. Sample size may be re-estimated at interim analyses. One of the approaches in sample size re-estimate is to determine sample size to be increased based on conditional power. With a time-to-event endpoint at interim analyses, one may face the problem that observed data does not have proportional hazards, and the problem of how to calculate conditional power if future data is also assumed to be non-proportional hazards. There are many papers discussing sample size determination for data with non-proportional hazards. However, few of them discussed sample size re-estimation for data with non-proportional hazards. We propose the piecewise exponential model to fit the observed data at interim analyses, and estimate the parameters of the model. Piecewise exponential model is quite flexible, and can fit most of time-to-event data quite well (Chen and Dong, 2012 JSM Proceedings). Then data will be simulated based on the estimated piecewise exponential model. Conditional power will be calculated based on simulated data. One can increase the sample size until the conditional power reaches a satisfactory level.