gms | German Medical Science

Background: Transitivity assumption is the cornerstone of network meta-analysis (NMA). It implies that important effect modifiers are similar across the observed comparisons in a network of interventions. Violation of this assumption compromises the credibility of the indirect estimates and, by extent, the treatment effects of all possible comparisons in the network. Most of the published empirical studies on the evaluation and reporting quality of the underlying assumptions for NMA have focused on the quality of the indirect comparisons.

Objectives: The present study is the first to offer extensive empirical evidence on the evaluation and reporting quality of transitivity assumption using a collection of systematic reviews with NMA. The ultimate goal is to elucidate the prevalence of systematic reviews with conclusions of questionable credibility.

Study design and setting: We used a previous collection of 357 systematic reviews with NMA published between January 2011 and, March 2017.

Results: Only a handful of reviews reported how they planned to evaluate the transitivity assumption. Two in five reviews explicitly mentioned transitivity assumption and mainly in the discussion section. Most reviews described at least one method to evaluate the transitivity assumption. The most prevalent was the narrative evaluation of trial comparability, followed by sensitivity analysis and statistical evaluation of inconsistency. Almost half of the reviews concluded on the plausibility of transitivity assumption, and hence, the reliability of the treatment effects, followed by the consistency evaluation and intervention hierarchy. Approximately one in five reviews revealed that it was difficult to judge the plausibility of transitivity assumption due to limited available data (e.g., few trials and poor reporting of the included trials). In justifying their conclusions about the credibility of the transitivity assumption, most reviews considered the comparability of the trials, followed by the consistency evaluation. Among the reviews that judged transitivity to be questionable or difficult to judge, only three reviews abstained from performing NMA.

Conclusions: The prevalence and the methods' quality to assess the transitivity assumption were low overall. To this aim, we plan to develop a series of extensive recommendations concerning the comprehensive and transparent evaluation of the transitivity assumption in systematic reviews.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.