Abstract:
|
Network meta-analysis allows for the aggregation of study results comparing various treatments on a common outcome. For instance, one might be interested in estimating the comparative effectiveness of several competing medications using randomized controlled trial results. Due to diversity in study planning, many potential sources of heterogeneity threaten the legitimacy of such summary analysis. These sources include differences in population composition and in the provided version of a common treatment. We describe the counterfactual perspective that allows for the definition of causal effects in a network meta-analysis. This framework allows for a formal integration of these types of heterogeneity into the estimation procedure. As in other contexts, the causal inference approach allows for a clear description of when estimation is feasible, allowing for a more informed interpretation of analytical results.
|