Bayesian Prediction Model of Event Times in Randomized Clinical Trial

• Activity Number: 260
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jiang Li* and Wentao Feng and Satrajit Roychoudhury
• Companies: Novartis and Novartis and Novartis Pharmaceuticals
• Keywords: Bayesian prediction ; mixture approach ; clinical trials

Abstract

In trials with time to event endpoint statistical power is primarily determined by the number of events. Thus, it is of interest to predict early and accurately the time of a landmark interim or terminating event. It can also be valuable from logistical perspective of the trial. Different methods are available in statistical literature to predict "landmark" event time (Bagiella and Heitjan (2001), Donovan et al. (2006) etc.). In this work we propose a Bayesian prediction model assuming non-constant event and drop-out rates over time. For better prediction this proposed model takes account of the additional uncertainties due to the blinded treatment arm in randomized studies using mixture approach. The performance of this method is evaluated on the real randomized clinical trials.