How do I run a sensitivity analysis for a winter-health meta-analysis before grant resubmission?

Run the sensitivity analysis by recomputing the pooled estimate after excluding studies judged high risk of bias, then compare that estimate with the main result against a material-change threshold you set before running the analysis.

1. Define the exclusion rule and the threshold in advance. State which studies will be removed (for example, those with a high-risk RoB 2 judgement for the result being pooled) and what size of shift in the pooled effect counts as material. Record both in the protocol or analysis plan so the comparison is prespecified.
2. Recompute the pooled estimate on the reduced set. Apply the same effect measure, model, and weighting used in the main analysis so the only difference is the study set. Report the new estimate with its confidence interval alongside the main estimate.
3. Build the sensitivity table. Include the main estimate, the restricted estimate, the number of studies and participants removed, and a change flag column marking whether the shift crossed your threshold.
4. Interpret against the threshold. If the restricted estimate stays within the threshold, describe the main result as stable under this exclusion. If it crosses the threshold, treat the finding as sensitive and revise the abstract sentence to reflect the restricted estimate and the excluded studies.

Use RoB 2 to assign risk-of-bias judgements at the result level, since it rates a specific randomized-trial result across bias domains and produces low risk, some concerns, or high risk. Cochrane meta-analysis guidance covers sensitivity analyses and the careful interpretation of pooled results, and GRADE criteria support outcome-level certainty judgements that you can align with the sensitivity finding. Name these sources in the methods section without linking them here.

A revised abstract sentence might read: "After excluding trials at high risk of bias, the pooled effect was [restricted estimate, CI], compared with [main estimate, CI] in the primary analysis." Fill in the values from your table and keep the wording consistent with the change flag.