GRADE approach to rate the certainty from a network meta-analysis: addressing incoherence

• Published in: Journal of clinical epidemiology Vol. 108; pp. 77 - 85
• Main Authors: Brignardello-Petersen, Romina, Mustafa, Reem A., Siemieniuk, Reed A.C., Murad, M. Hassan, Agoritsas, Thomas, Izcovich, Ariel, Schünemann, Holger J., Guyatt, Gordon H.
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.04.2019; Elsevier Limited
• Subjects: Bias; Certainty in the evidence; Data Interpretation, Statistical; Domains; Epidemiology; Estimates; Evaluation; Evidence-Based Medicine - methods; Evidence-Based Medicine - standards; GRADE; GRADE Approach - methods; Humans; Incoherence; Inconsistency; Internal Medicine; Meta-analysis; Network meta-analysis; Network Meta-Analysis as Topic; Publication Bias; Quality; Quality of the evidence; Systematic reviews; Systematic Reviews as Topic; Working groups; Network meta-analysis; Certainty in the evidence; Inconsistency; Systematic reviews; GRADE; Incoherence; Quality of the evidence
• ISSN: 0895-4356; 1878-5921; 1878-5921
• DOI: 10.1016/j.jclinepi.2018.11.025

This article presents official guidance from the Grading of Recommendations Assessments, Development, and Evaluation (GRADE) working group on how to address incoherence when assessing the certainty in the evidence from network meta-analysis. Incoherence represents important differences between direct and indirect estimates that contribute to a network estimate. Bias due to limitations in study design or publication bias, indirectness, and intransitivity may be responsible for incoherence. Addressing incoherence requires a judgment regarding the importance of the impact on the network estimate. Reviewers need to be alert to the possibility of misguidedly arriving at excessively low ratings of certainty by rating down for both incoherence and other closely related GRADE domains. This article describes and illustrates each of these issues and provides explicit guidance on how to deal with them.