Reassess inconsistency at the outcome level, apply the appropriate certainty downgrade with a stated reason, and record the change in the Summary of Findings table. GRADE rates certainty per outcome using explicit criteria, so a rating that skipped the inconsistency domain needs a documented revision rather than a silent edit.
- Confirm the scope of the problem. Check whether the inconsistency domain was assessed for that specific outcome and whether the rating reflects it. GRADE judgements are made at the outcome level, so a rating carried over from another outcome or from the review as a whole may not reflect the studies contributing to this result.
- Examine the heterogeneity evidence. Review the effect estimates across studies, the heterogeneity statistics, and the forest plot, and consider whether a fixed-effect or random-effects model changes the interpretation of the pooled result. Cochrane meta-analysis guidance covers effect measures, heterogeneity, and sensitivity analyses for this step.
- Decide the downgrade. If unexplained inconsistency is substantial, apply the appropriate downgrade and write the reason next to the rating. If the inconsistency is explained by a prespecified subgroup or sensitivity analysis, state that explanation instead of downgrading.
- Revise the record. Update the certainty rating, the reason for the change, and the Summary of Findings table so the outcome-level judgement and its footnote match. Cochrane GRADE resources cover Summary of Findings table guidance and GRADEpro workflows for this revision.
- Check downstream consistency. Confirm that the abstract, results text, and any recommendation language reflect the revised rating, since a changed certainty level can alter how the finding is described.
The expected result is a rating whose inconsistency judgement is visible, justified, and consistent across the review and its Summary of Findings table.