A network meta-analysis is a statistical method that pools direct and indirect evidence across a group of studies to compare three or more interventions. For example, a network meta-analysis could pool clinical trials assessing different exercise modalities, compared to another exercise type or to no exercise, on the effects of reactive balance (the last mode of defense against a fall).^[1] Pooling direct evidence means that all the studies comparing functional exercise to no exercise are averaged together, all studies comparing strength training to flexibility training are averaged together, and so on. Pooling indirect evidence means that the average effect of flexibility training can be compared to the average effect of functional exercise even though there is no study that compares these two interventions directly.

Because a network meta-analysis allows the comparison of interventions that have not been directly compared in clinical research, these interventions can be ranked in the order of most to least effective for the studied outcome. This method relies on the assumption that the studies in the network are similar in factors that may affect the measured outcomes, like if all of the included studies have participants with similar health conditions and similar socio-demographics.^[2]