Predicting Data Cut-Off Date in Event-Driven Clinical Trials When Treatment Assignments Blinded

• Activity Number: 408
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Jianming Wang*+ and Zhinuan Yu
• Companies: Celgene Corporation and Celgene Corporation
• Keywords: Data Cut-off ; Event-driven ; Blinded

Abstract

For event-driven clinical trials, the accrual of the event of interest determines the data cut-off date for the statistical analyses. Because of significant resources and strategic planning involved, it is of interest to predict the data cut-off date early and accurately. When the treatment assignment is blinded, a mixture of exponential distributions with a known number of components can be used. In practice, beside potential treatment effect, heterogeneity among subjects is common and a mixture of an unknown number of components could be used. However, modeling a mixture with an unknown number of components and with censored data poses great challenge. In this presentation, we propose another approach to handle heterogeneous population by assuming that pooled event rate following a Gamma or Weibull distribution, as both Gamma and Weibull distributions with shape parameter less than 1 could be used to model a mixture of exponential distributions. The performance of this approach will be evaluated using simulations. An Application in real data will be demonstrated.