Comparison of Bayesian and Frequentist Meta-Analytical Approaches for Analyzing Time-to-Event Data

• Activity Number: 160
• Type: Invited
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Brenda Crowe*
• Companies: Eli Lilly and Company
• Keywords: meta-analysis ; survival analysis ; time to event ; Bayesian ; Firth ; rare events

Abstract

Using meta-analysis in health care research is a common practice. I will present the results of a simulation study that investigated meta-analysis methods used for analysis of time-to-event data. The study focused on performance when the event rate is low. We compared the performance of three methods based on the Cox proportional hazards model (standard, Firth correction and Bayesian). Of the methods we included in the simulation, the Firth correction model most consistently predicted the true value of the log hazard ratio. In terms of bias, it did a better job handling survival data when the event count was low than the standard Cox model and the Bayesian model with both the diffuse priors and the more informative priors.