Composite Kaplan-Meier and Semiparametric Commensurate Bayesian Methods for Post-Market Medical Device Surveillance with Historical Survival Information

• Activity Number: 236
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Thomas Murray*+ and Brian Hobbs and Ted Lystig and Bradley P. Carlin
• Companies: University of Minnesota and The University of Texas M.D. Anderson Cancer Center and Medtronic, Inc. and University of Minnesota
• Keywords: Bayesian hierarchical modeling ; Commensurate prior ; Evidence synthesis ; Flexible proportional hazards model ; Hazard smoothing ; Non-exchangeable sources

Abstract

Trial investigators often have a primary interest in the estimation of the survival curve in a population for which there exists acceptable historical information from which to borrow strength. However, borrowing strength from a historical trial that is systematically different from the current trial can result in biased conclusions and possibly longer trials. In this paper we propose both composite Kaplan-Meier and model based Bayesian Methods for the purpose of attenuating bias and increasing efficiency when jointly modeling non-exchangeable time-to-event data from two sources of information. The performance of these models regarding survival curve estimation is compared with other common strategies undertaken in the presence of acceptable historical information. We use simulation to show that these methods facilitate attractive bias-variance tradeoffs in a variety of settings. We then further illustrate the mechanics of our methods by fitting them to a pair of post-market surveillance datasets regarding adverse events in persons on dialysis that underwent cardiac revascularization with a bare metal stent. To finish, we discuss the advantages and limitations of these methods.