Improvements in the GRADE approach to network meta-analysis

Article type
Year
Authors
Guyatt G1, Bonner A1, Alexander P1, Brignardello-Petersen R1
1McMaster University, Canada
Abstract
Background: Rating the certainty (synonyms: quality, confidence) in evidence associated with the network estimate of each paired comparison within a network meta-analyses (NMA) presents challenges. The GRADE Working Group has addressed the issue, but when there are a large number of candidate therapies the approach can be onerous. A GRADE working group is now proposing refinements to the GRADE method that increase efficiency of its application.

Objectives: To illustrate proposed refinements in GRADE’s approach to rating certainty of evidence in an NMA.

Methods: Refinement through discussion and iterative testing with application to a NMA of antidepressants.

Results: As initially proposed, the approach involves four steps:
1. Present direct and indirect treatment estimates for each paired comparison.
2. Rate the certainty of each direct and indirect estimate.
3. Present the NMA estimate for each comparison.
4. Rate the certainty of each NMA estimate.
A new insight is that one need not rate direct or indirect estimates using conventional GRADE guidance, but should rather assess aspects of direct comparisons to inform the certainty of the network estimates. What follows is that the judgment regarding precision is based only on the network estimate, and review of the head-to-head trials that inform direct and indirect comparisons need consider only the other four domains (risk of bias, inconsistency, indirectness, publication bias). Thus, the repeated assessment of precision previously suggested is no longer necessary, streamlining the rating process. Another insight enhancing efficiency follows from the guidance that certainty of the network estimate is based on the direct or indirect estimate in which one is more certain. Therefore, if one has a direct comparison in which one has not rated down for any of the four relevant domains, one need not consider the indirect estimate.

Conclusions: Informed decision-making requires rating certainty in individual network estimates. When there are a large number of candidate therapies, the process can be onerous. New insights and associated guidance can streamline GRADE certainty in evidence ratings.