How do I fix a Summary of Findings table that misstates an effect?

Trace the misstated row back to its source analysis and certainty rating, correct the effect estimate and footnotes, then verify the revised table against the underlying analysis records.

A Summary of Findings (SoF) table reports effect estimates and certainty ratings for each outcome in a review. When a row misstates an effect, the error usually sits in the transfer from the analysis to the table, so work backward from the row to the record that generated it.

1. Identify the exact row and column that is wrong: the effect estimate, the confidence interval, the absolute effect, or the certainty rating. Note the outcome and comparison it belongs to. Expected result: a single, specific discrepancy you can describe in one sentence.
2. Open the source analysis for that outcome and comparison and compare the point estimate, confidence interval, and number of participants or studies. Expected result: you can see whether the table copied the wrong value, used a different analysis, or dropped a subgroup.
3. Check the certainty rating against the outcome-level judgement recorded in your GRADE assessment. GRADE rates certainty per outcome using explicit criteria, so the rating in the table should match the judgement for that outcome. Expected result: you know whether the rating or the effect value is the error.
4. Correct the row and update its footnotes so any explanation of imprecision, indirectness, or the analysis used matches the revised value. Expected result: the footnote text and the row agree.
5. Re-read the revised table against the analysis records and the review text, including any statements in the results or conclusions that cite the effect. Expected result: the table, the analysis, and the narrative report the same estimate and certainty.

Cochrane Handbook Chapter 15 covers how to present confidence intervals, exact P values, and absolute and relative effects, which helps you decide which value belongs in the table. Cochrane GRADE resources and the GRADE Working Group describe outcome-level certainty judgements, so use those criteria when the rating is the part that is wrong.