Title: Data linking approaches for meta-analysis of individual participant data
Authors: EY Mun - University of North Texas Health Science Center (United States) [presenting]
Abstract: Clinical trials are heterogeneous in key design features, including participants, treatments, comparisons, outcome measures, and settings, creating challenges for feasibility and interpretation for complex multivariate research synthesis. Such between-study heterogeneity has posed a significant barrier to fully utilizing individual participant data for meta-analysis applications despite well-known advantages of analyzing individual participant data in meta-analysis. We present several methodological approaches we have adopted to address between-study heterogeneity for Project INTEGRATE, a large-scale research synthesis project utilizing individual participant data, as well as aggregate data, from multiple independent trials that were developed to prevent alcohol misuse for adolescents and college students. Of the approaches taken for Project INTEGRATE to link data validly across samples and studies, we will focus on a mapping approach and a Bayesian multilevel modeling approach and present data application examples.