Combination of Confidence Distributions (CDs) and an Efficient Approach for Meta-Analysis of Heterogeneous Studies

• Activity Number: 301
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: IMS
• Author(s): Dungang Liu*+ and Regina Liu and Minge Xie
• Companies: Rutgers University and Rutgers University and Rutgers University
• Keywords: Confidence distribution ; Heterogeneous studies ; IPD estimator ; Meta-analysis ; Relative efficiency

Abstract

We propose a new methodology for combining summary information from independent studies based on generalized linear models. This work is motivated by the need in meta-analysis to investigate the effect of covariates in clinical trials when covariate designs are heterogeneous among the studies. We develop a meta-analysis approach using multiparameter CDs. We show that the CD-based estimator asymptotically achieves the Cramer-Rao lower bound, and is as efficient as the Individual Patient Data (IPD) estimator derived from the complete individual data. But unlike the IPD estimator, our approach only uses summary information and does not need the complete data. Additionally, it can adapt to heterogeneous studies, such as the case of missing design variables or non-estimable parameters, for which conventional meta-analysis is inadequate. We illustrate our approach with numerical studies.