What is GRADE?
GRADE (Grading of Recommendations, Assessment, Development and Evaluations) is a systematic approach to rating the certainty of evidence in systematic reviews and the strength of recommendations in clinical practice guidelines. Adopted by over 100 organizations worldwide including WHO, Cochrane, and BMJ, GRADE has become the standard framework for evidence appraisal.
The Four Levels of Evidence Certainty
GRADE classifies evidence certainty into four levels:
- High: We are very confident that the true effect lies close to the estimate of effect
- Moderate: We are moderately confident; the true effect is likely close to the estimate but there is a possibility it is substantially different
- Low: Our confidence is limited; the true effect may be substantially different from the estimate
- Very Low: We have very little confidence; the true effect is likely substantially different from the estimate
Starting Point
In GRADE, evidence from randomized controlled trials (RCTs) starts at "High" certainty, while evidence from observational studies starts at "Low" certainty. From these starting points, certainty can be rated down (for limitations) or rated up (for strengths).
Five Domains for Rating Down
1. Risk of Bias: Assess methodological quality of included studies using tools like RoB 2.0 or ROBINS-I. Serious limitations in study conduct (poor randomization, lack of blinding, high attrition) warrant downgrading.
2. Inconsistency: Examine whether results across studies point in the same direction. High heterogeneity (I² > 50-75%) with no adequate explanation warrants downgrading.
3. Indirectness: Assess whether the evidence directly addresses your PICO question. Differences in population, intervention, comparator, or outcomes between your question and available evidence warrant downgrading.
4. Imprecision: Evaluate whether the confidence interval around the pooled estimate is narrow enough to support a definitive conclusion. Small sample sizes and wide confidence intervals warrant downgrading.
5. Publication Bias: Assess whether selective publication of studies may have skewed results. Funnel plot asymmetry, small study effects, and commercial funding may warrant downgrading.
Three Domains for Rating Up
Observational studies can be rated up for:
- Large effect: A relative risk >2 or <0.5 with no plausible confounding
- Dose-response gradient: A clear dose-response relationship increases confidence
- Residual confounding: When all plausible confounders would reduce the observed effect
Creating Summary of Findings Tables
GRADE Summary of Findings (SoF) tables present the key information for each outcome: the number of studies and participants, the relative and absolute effects, the certainty rating, and footnotes explaining any downgrading. GRADEpro GDT is a free online tool for creating these tables.
Practical Tips
- Apply GRADE to each outcome separately, not to the review as a whole
- Focus on patient-important outcomes, not surrogate measures
- Be transparent about your judgments — document reasons for each rating decision
- Use the GRADE handbook (available free online) for detailed guidance
