Multivariate Meta-Analysis with an Increasing Number of Parameters

• Activity Number: 59
• Type: Topic Contributed
• Date/Time: Sunday, August 9, 2015 : 4:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Simina M. Boca and Ruth Pfeiffer and Joshua Sampson*
• Companies: Georgetown University Medical Center and National Cancer Institute and National Cancer Institute
• Keywords: multivariate meta-analysis ; efficiency ; fixed-effects models ; random-effects models

Abstract

When considering multiple studies which each estimate the same set of parameters, such as the association of several treatments or risk factors with a single outcome, researchers have a choice between individually evaluating each parameter through univariate meta-analyses or performing a single multivariate meta-analysis. Our goal is to compare the performance of these two approaches as the number of parameters increases. We show that: (a) for fixed effects meta-analysis, the benefit of using multivariate meta-analysis can substantially increase as the number of parameters increases; (b) for random effects meta-analysis, the improvement is reduced in the presence of high-between study variability and by the need to estimate an increasingly large between-study covariance matrix; (c) for little to no between-study variability, the choice of random effects over fixed effects results in an increasing loss of efficiency as the number of parameters increases.