Predicting Analysis Time in Event-Driven Clinical Trials When There Are Multiple Correlated Time-to-Event Endpoints

• Activity Number: 244
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jianming Wang*+ and Zhinuan Yu
• Companies: Celgene and Celgene
• Keywords: clinical trials ; analysis time prediction ; event-driven ; Bayesian ; surrogate

Abstract

For clinical trials with time-to-event endpoints, predicting the accrual of those events of interest could help to determine the timing of the interim and final analyses. For example, in many oncology studies, overall survival (OS) is the key efficacy endpoint, with planned interim and final analyses. Other time to event endpoints, such as progression free survival (PFS) and time to progression (TTP), are usually considered as surrogate endpoints. Currently available methods in literature mostly deal with predicting analysis time when there is only one time-to-event endpoint. One might predict the accrual of events separately without considering the correlation. However, as OS, TTP and PFS are usually correlated, a joint model that could consider the dependency among endpoints would be appealing in practice since predicting of OS could borrow strength from the accumulating surrogate information from TTP or PFS. In this presentation, we describe a general parametric model to simultaneously predicting the accrual of both OS and TTP, with a dependency structure of OS on TTP. PFS will be handled separately since it could be derived from OS and TTP.