ABSTRACT
We consider statistical issues that arise when synthesizing information from multiple studies in the contexts of both traditional meta-analysis, in which the same pair of the treatment groups is assumed to have been studied in all trials, and network meta-analysis, in which the objective may involve drawing inference about treatment groups that may not have been studied in a head-to-head fashion. Formal expressions of the underlying models are provided, with a thorough discussion of the relevant assumptions and measures that need to be taken in order to mitigate the impacts of deviations from the assumptions. In addition, a brief review of the recent literature on best practices for the conduct and reporting of such studies, as well as illustrations of the techniques with simulated data, are provided.
