How do I reconcile effect measures across studies before pooling?

Reconcile effect measures by checking each study's reported measure against the review's prespecified measure, converting only where justified, and logging every derivation with its assumption. Cochrane Handbook Chapter 10 treats the choice of effect measure as a review decision that shapes how results are combined and interpreted, so the reconciliation step starts from that prespecified measure.

1. Fix the target measure before touching study data. Record the measure named in the review protocol (for example a risk ratio, odds ratio, or mean difference) and the direction of benefit. Result: every later decision is judged against one stated target.
2. Extract each study's reported measure, its scale, and the data behind it. Note whether the study reports the target measure directly or reports a different measure with the underlying counts or estimates. Result: a table showing which studies match the target and which need conversion.
3. Convert only where the reported data support it, and write down the assumption each conversion rests on. Result: a conversion log entry per study linking the original value, the derived value, and the assumption used.
4. Flag studies that cannot be harmonised to the target measure. Record the reason and decide whether to exclude them from the pooled estimate or present them separately. Result: a note listing every unharmonised study and its handling.
5. Assemble the harmonised analysis dataset and rerun the pooling on it. Result: one dataset in which every row traces back to a logged derivation or a direct extraction.

Cochrane Handbook Chapter 10 also covers fixed and random effects, heterogeneity, missing data, and sensitivity analyses; use a sensitivity analysis to show how the pooled result responds to the conversion choices you logged. Chapter 15 addresses confidence intervals, absolute and relative effects, and applicability, which matter when you interpret the pooled estimate. GRADE rates certainty at the outcome level using explicit criteria, so carry the harmonisation notes into that judgement when a conversion assumption affects an outcome.