Testing Violations of Exponential Assumption in Cancer Clinical Trials with Survival Endpoints

• Activity Number: 171
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Gang Han*+ and Michael Schell and Daniel Zelterman and Christos Hatzis and Lajos Pusztai
• Companies: Yale and Moffitt Cancer Center & Research Institute and Yale and Yale Cancer Center and Yale Cancer Center
• Keywords: piecewise exponential estimate ; uniformly most powerful unbiased test ; personalized cancer therapy ; invasive breast cancer

Abstract

The development of personalized cancer therapy leads to clinical trials with smaller sample sizes compared with trials involving complete disease entities. The use of exponential survival modeling has the potential of gaining 35% more efficiency or saving 28% required sample size, making the personalized therapy trials increasingly feasible. However, the use of exponential survival has not been fully accepted in practice due to the lack of knowledge about the exponential assumption. We propose a test for identifying violations of the exponential assumption using a reduced piecewise exponential approach. Compared with an alternative goodness-of-fit test, which suffers from the inflation of type I error under various censoring mechanisms, the proposed test can maintain the correct type I error. We conduct power analysis using simulated data based on different types of the cancer survival distribution in the national registry database. This approach was applied in a genomic predictor study of distant relapse-free survival following Taxane-Anthracycline chemotherapy for invasive breast cancer to compare patient groups corresponding to different disease subtypes.