ABSTRACT
Network meta-analysis aims to compare multiple interventions by synthesizing studies where each study compares only some of the interventions. This chapter explains how direct and indirect evidence are synthesized to compare interventions linked in a network of trials under the assumptions of consistency and transitivity. It defines the statistical model used for network meta-analysis under consistency, using various perspectives and implementations to shed more light on the model, and describes various ways in which the model may be extended to explore inconsistency. It then shows how network meta-analysis results may be interpreted using tools to rank the interventions. Topics common to all meta-analysis work, such as meta-regression and publication bias, are discussed in the network meta-analysis case, and available software is described.
