A Guidance of Using Meta-Analysis Method for Data with Rare Event

• Activity Number: 372
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yao Yu* and Yuanyuan Tang and Qi Tang and Shihua Wen
• Companies: AbbVie and AbbVie and AbbVie and AbbVie
• Keywords: Meta-analysis ; rare event ; Bayesian ; Frequentist

Abstract

Meta-analysis is a statistical technique to make trustworthy conclusions through combining findings from individual studies designed to answer the same or similar questions such as drug efficacy or safety. In the past decades, various methods had been proposed for meta-analysis. Different meta-analysis methods may show good performances under different scenarios, but none of the methods shows a uniform advantage over the rest of the methods when rare events occur and effect sizes are small. When events are rare, different methods even provide contradictory conclusions. It is still unclear how to utilize existing meta- analysis methods for efficacy and safety analyses. In this research, we revisited the meta-analyses presented by Bradburn, Michael J. et al (2007) and Lu, Tian et al. (2009) that produced contradictory conclusions using different methods. We perform simulation studies to evaluate the performance, in terms of coverage rate, mean square error (MSE), and power, of the most popular, latest, frequentist or bayesian approaches under different settings. Finally we provide a guidance for meta-analysis with rare events.