How to Use GRADE to Assess Certainty of Evidence

GRADE is a transparent framework for rating how certain we can be that an estimated effect is close to the true effect, and for moving from evidence to recommendations. Certainty is rated for each outcome as high, moderate, low or very low. Read the official GRADE guidance for formal use; this is an overview.

How ratings start

• Evidence from randomised trials generally starts as high certainty.
• Evidence from non-randomised studies generally starts as low, though newer approaches (for example with ROBINS-I) may start higher.

Factors that lower certainty

1. Risk of bias in the contributing studies.
2. Inconsistency (unexplained heterogeneity).
3. Indirectness (differences in population, intervention, comparator or outcome from the question).
4. Imprecision (wide confidence intervals, few events).
5. Publication bias.

Factors that can raise certainty (mainly observational evidence)

A large effect, a dose-response gradient, and plausible residual confounding that would reduce the observed effect.

Presenting

A Summary of Findings table lists each important outcome, the number of studies and participants, the effect estimate with a confidence interval, and the certainty rating, with explanations in footnotes.

Illustrative example

Invented for teaching. Three randomised trials show reduced hospital admissions with a nurse-led programme. Two have unclear allocation concealment, the interval is wide and includes small benefit and no effect, so certainty is downgraded twice (risk of bias and imprecision) from high to low.

Reporting GRADE judgements

Give a one-line explanation for every rating decision in the table footnotes, for example "Downgraded one level for imprecision: confidence interval includes both appreciable benefit and no effect". State the wording you use for each certainty level and apply it consistently, so that readers can see how conclusions relate to the strength of evidence. Involve at least two reviewers in the ratings and resolve differences by discussion.

Common pitfalls

• Applying GRADE to a whole review rather than to each outcome.
• Using ratings mechanically without explaining decisions.

Key takeaways

• GRADE rates certainty per outcome using explicit criteria.
• Explain every downgrade or upgrade.

This page is general educational information about research methods. It is not medical, legal or regulatory advice; follow your institution's ethics board and your target journal's instructions.