When multiple interventions exist for the same problem, evidence-based decision making requires clear and systematic syntheses of available evidence to determine which interventions are most effective. Network meta-analysis extends standard pairwise meta-analytic models to allow direct and indirect evidence from primary studies comparing alternative interventions to be statistically combined in a single analysis. The use of indirect evidence can increase the precision of estimates and allow the relative effectiveness of interventions to be estimated even if particular interventions have never been directly compared in primary studies. The validity of network meta-analyses requires assumptions that may be difficult to verify in reviews of complex social interventions, and the method has rarely been applied outside of medicine.