Online Program

Towards causally interpretable meta-analysis: transporting inferences from multiple studies to a target population (307885)

Keywords: meta-analysis, causal inference, transportability

For many comparative effectiveness research questions, evidence is available from multiple randomized trials conducted independently in different underlying populations. Standard meta- analysis methods produce results that do not have a clear causal interpretation when each trial samples eligible individuals from a different underlying population (e.g., from centers with different referral patterns or in different geographic locations) and treatment effects vary across the sampled populations. We take steps towards causally interpretable meta-analysis by describing methods for transporting causal inferences from a collection of randomized trials to a new target population. We discuss identifiability conditions for average treatment effects in the target population and provide identification results and propose average treatment effect estimators. We also describe simple methods for sensitivity analysis to violations of the identifiability conditions and discuss extensions of the methods to address non-adherence in the randomized trials. We illustrate the methods using data from a multicenter trial of treatments for chronic hepatitis C infection.