Mixed treatment comparison analysis provides internally coherent treatment effect estimates based on overviews of reviews and can reveal inconsistency

Abstract

Objectives: To propose methods for mixed treatment comparisons (MTC) based on pooled summaries of the type produced in overviews of reviews.

Study design and setting: Overviews of reviews (umbrella reviews) summarize the results of multiple systematic reviews into a single document. They report the summary estimates from the original pairwise meta-analyses and discuss them in narrative form, with the intention of identifying the most effective treatment. We present methods for MTC synthesis, tailored for use with overviews of reviews. These generate a single internally consistent summary of all the relative treatment effects and assessments of whether the summary is consistent with the data. These methods are applied to a published overview of treatments for childhood nocturnal enuresis. We apply the methods to both fixed-effect (FE) and random-effects (RE) meta-analyses of the original trials.

Results: The summary relative risks based on FE meta-analyses, as originally published, were highly inconsistent. Those based on RE meta-analyses were consistent and could, given standard assumptions on comparability of treatment effects in meta-analysis, form a basis for coherent decision making.

Conclusion: Along with the summaries from systematic reviews, MTC methods should be used in overviews to provide a single coherent analysis of all treatment comparisons and to check for evidence consistency.