Grading of recommendations assessment, development, and evaluation concept article 5: addressing intransitivity in a network meta-analysis

J Clin Epidemiol. 2023 Aug:160:151-159. doi: 10.1016/j.jclinepi.2023.06.010. Epub 2023 Jun 20.

Abstract

Objectives: This article describes considerations for addressing intransitivity when assessing the certainty of the evidence from network meta-analysis (NMA) using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. Intransitivity is induced by effect modification, that is, when the magnitude of the effect between an intervention and outcome differs depending on the level of another factor.

Study design and setting: To develop this GRADE concept paper, the lead authors conducted iterative discussions, computer simulations, and presentations to the GRADE project group and at GRADE working group meetings. The GRADE Working Group formally approved the article in July 2022.

Results: NMA authors can have a higher or a lower threshold to rate down the certainty of the evidence due to intransitivity, which depends on the extent of their concerns regarding the trustworthiness of indirect comparisons, and their view of the relative problems with rating down excessively or insufficiently. NMA authors should consider three main factors when addressing intransitivity: the credibility of effect modification, the strength of the effect modification, and the distribution of effect modifiers across the direct comparisons. To avoid double counting limitations of the evidence, authors should consider the relationship between intransitivity and other GRADE domains.

Conclusion: NMA authors face theoretic and pragmatic challenges and in most situations need to assess intransitivity without the availability of empirical data. Thus, explicitness regarding perspective is crucial.

Keywords: Certainty of evidence; GRADE; Indirect evidence; Intransitivity; Network meta-analysis; Quality of evidence.

Publication types

  • Meta-Analysis

MeSH terms

  • GRADE Approach*
  • Humans
  • Network Meta-Analysis