GRADE Working Group approach for NMA
This method helps assess how confident we can be in the results of a network meta-analysis by looking at the quality of both direct and indirect evidence. It breaks the assessment into four steps and uses the well-known GRADE system to rate evidence for each treatment comparison separately, rather than giving one overall rating for the whole network.
At a glance
Use when
Conducting or interpreting a network meta-analysis where confidence in effect estimates needs to be systematically evaluated.
Avoid when
When a rapid or high-level summary of evidence quality is sufficient without per-comparison assessment.
Inputs
Results from a network meta-analysis, including direct and indirect effect estimates, measures of heterogeneity and inconsistency, and study-level risk of bias assessments.
Outputs
Quality of evidence ratings (high, moderate, low, very low) for each treatment comparison in the network, per GRADE criteria.
How it works
The GRADE Working Group approach for network meta-analysis (NMA) provides a structured four-step framework to rate the quality of evidence for treatment effect estimates derived from NMAs. It separately evaluates the quality of direct, indirect, and combined NMA estimates using GRADE criteria such as risk of bias, inconsistency, indirectness, imprecision, and publication bias. The method emphasizes that quality ratings should be assigned per comparison rather than at the network level to avoid misleading conclusions.
- HTA domains
- Clinical Effectiveness
- Assumptions
- The validity of indirect comparisons relies on the assumption of coherence (or consistency) between direct and indirect evidence; the method assumes that GRADE principles can be systematically applied to both types of evidence.
- Strengths
- Provides transparent, standardized quality ratings for each pairwise comparison in an NMA,Separates assessment of direct, indirect, and combined evidence,Avoids misleading summary ratings for entire networks,Builds on the widely accepted GRADE framework
- Limitations
- Requires detailed understanding of GRADE methodology,Can be time-consuming when applied to large networks with many comparisons,Subjectivity in judgment calls (e.g., extent of inconsistency or indirectness)
- Also known as
- GRADE for NMA, GRADE approach to network meta-analysis, Quality rating of NMA evidence
Questions this answers
- › How confident can we be in the treatment effect estimates from a network meta-analysis?
- › What is the quality of direct and indirect evidence for a specific treatment comparison?
- › Does the indirect evidence align with the direct evidence (coherence)?
- › Which GRADE domains (e.g., risk of bias, inconsistency) affect the confidence in NMA results?
- › Should the quality of evidence be downgraded for a particular comparison?
- › How should quality of evidence ratings be presented across a network of interventions?
References & sources
Related methods
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.