Identify which arms were combined, check whether that combination answers your review question, then split the data or adjust the comparison and recompute any affected estimate.
A multi-arm trial compares several interventions in one study. When a data table merges two or more of those arms into a single row, the pooled numbers no longer describe a single intervention, so the first repair step is to recover the original arm-level results.
- List every arm in the trial and mark which ones the table combined. Compare the table against the trial report to confirm the arm labels, group sizes, and outcomes behind each merged row. The expected result is a clear map of what the merged row actually contains.
- Check the combination against your review question. If your review compares each intervention separately, the merged row hides a distinct comparison and should be split back into its component arms. If the review question genuinely treats those arms as one intervention, keep the combination and state the rationale.
- Split or adjust the data. Rebuild separate rows for each arm with its own group size and outcome counts, or, where the combination is valid, keep the merged row and label it as a combined intervention. Record the change in your data-extraction notes.
- Recompute affected estimates. Recalculate the effect estimate and any pooled result that used the merged row, then compare the new value with the earlier one to see whether the correction changed the direction or size of the finding.
Cochrane Handbook Chapter 10 covers effect measures, missing data, sensitivity analyses, and interpretation of pooled results, which are the parts of the analysis most likely to shift after a correction. Document the correction so a co-author or reviewer can trace the change from the original table to the revised estimate.