*Geek Box: Network Meta-Analysis

*Geek Box: Network Meta-Analysis

In a traditional meta-analysis, single studies are compiled together to obtain an overall summary of effect for the particular treatment/exposure and outcome being investigated. In a network meta-analysis, the effects of three or more interventions may be compared.

This is achieved by combining what is known as ‘direct’ and ‘indirect’ evidence. For example, let’s say we have Drug A, Drug B, and Drug C. And let’s say that a number of studies have compared Drug A vs. Drug C; this would be direct evidence. Now let’s say other studies have compared Drug B vs. Drug C; a network meta-analysis would allow for an indirect comparison between Drug A and Drug B, as they had both been compared to Drug C in other studies.

This means that a network meta-analysis is particularly useful for analysing the comparative effects of different interventions, and can estimate how these interventions rank in effectiveness. An important assumption for the validity of a network meta-analysis is what is known as “transitivity”.

This means that there are no systematic differences in the comparisons other than the treatments being compared, i.e., it is as if participants could have been randomised to any of the treatments in a study and the remaining factors would be similar. Network meta-analysis is a relatively new statistical approach, and is a promising method for determining effectiveness of comparative treatments on a specific condition or outcome.