Network meta-analysis

A network meta-analysis compares several treatments at once, including some that were never tested directly against each other in the same study.

This term is specific to research interpretation and is defined on AnxietyResearch. Shrinktionary, the network's language layer, doesn't currently carry a canonical entry for it.

A regular meta-analysis usually compares one treatment against one comparison. A network meta-analysis links many studies into a web. If treatment A was tested against B, and B against C, the method can estimate how A compares to C, even though no trial put A and C side by side. The output is often a ranking of treatments.

There are many anxiety treatments and few trials that test them head-to-head. Network meta-analysis helps fill that gap. A well-known example is the 2019 Lancet network meta-analysis of drug treatments for generalized anxiety disorder, which compared many medications in one analysis.

Trials exist for an SSRI against placebo, and for an SNRI against placebo, but few that test the SSRI against the SNRI directly. A network meta-analysis uses the shared placebo comparison to estimate how the two drugs stack up.

The indirect comparisons rest on an assumption that the linked trials are similar enough to connect. If the trials studied very different patients, the network can mislead. Treatment rankings also tend to look more certain than they are. A small gap between two ranked treatments is often not meaningful.