Reassess publication bias at the outcome level, then apply a GRADE downgrade with a written reason and update the Summary of Findings table. GRADE rates certainty using explicit criteria with outcome-level judgements, so the correction belongs in the outcome row where the concern applies.
- Check whether publication bias was assessed for that outcome. Look for a documented method such as a funnel plot, small-study effects assessment, or a registry and grey-literature search, and record what was found. If no assessment exists, the rating rests on an unexamined concern; note that gap before deciding.
- Decide whether the concern warrants a downgrade. GRADE treats risk of publication bias as one of the criteria that can lower certainty, so apply the downgrade when the assessment or the missing evidence leaves the pooled result open to selective reporting. Write the reason in the row, for example that small studies show larger effects than larger ones or that unpublished results were not identified.
- Apply the downgrade and revise the rating. Move the certainty level down by the amount your review's protocol or methods allow, and keep the reason attached to the outcome rather than to the review as a whole.
- Document the revision. Record the original rating, the new rating, the evidence behind the change, and who made the judgement, so a reader can trace the decision.
- Update the Summary of Findings table. Cochrane GRADE resources cover Summary of Findings table guidance and GRADEpro workflows; revise the certainty column, the footnote explaining the downgrade, and any related commentary so the table matches the revised rating.
If the assessment is inconclusive, leave the rating as unresolved and state what further check would settle it, such as a sensitivity analysis or a search for registered but unpublished studies. Cochrane Handbook Chapter 10 covers sensitivity analyses and careful interpretation of pooled results, which can inform that check.