Bayesian Approaches for Network Meta-Analysis of Randomized and Nonrandomized Clinical Trial Data
*Brad Carlin, University of Minnesota
Keywords: Bayes, propensity scores, network meta-analysis
Bayesian statistical approaches to network meta-analysis (NMA) are becoming more popular due to their flexibility and interpretability. In this talk, we summarize existing hierarchical Bayesian methods for MTCs with a single outcome, as well as more recent Bayesian approaches for multiple outcomes simultaneously. Our approach models correlation structure between outcomes through both contrast- and arm-based parameterizations that consider any unobserved treatment arms as missing data to be imputed. We extend this model to apply to all types of generalized linear model outcomes, such as count or continuous responses, and indicate how individual patient data (IPD) can be brought to bear if available. We also develop a new measure of inconsistency under our missing data framework, having more straightforward interpretation and implementation than standard methods. Finally, we extend our methods to the case of combining data from both randomized and nonrandomized trials. Here, we illustrate with data from an HIV/AIDS study wherein patients were offered the choice to be randomized to one of two drugs, or, if they preferred, their doctor could make the drug choice for them.