Compare each reviewer's domain judgements and the evidence behind them to find where the profiles diverge, then reconcile with a written rationale or escalate to a third methodologist.
GRADE rates certainty of evidence through explicit criteria and outcome-level judgements, so a disagreement between two profiles usually sits in a specific domain or outcome rather than across the whole rating. Locating that point is the first task.
- Align the two profiles outcome by outcome and list the certainty rating each reviewer assigned. Mark every outcome where the ratings differ; outcomes that match need no further work.
- For each divergent outcome, compare the domain judgements behind the rating: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Identify which domain each reviewer judged differently and the study data or reasoning cited for it.
- Re-examine the underlying evidence for that domain, such as confidence intervals, exact P values, and absolute and relative effects, and check whether the two reviewers were interpreting the same data or different subsets.
- Reconcile the divergence in writing: record the agreed rating, the domain that changed, and the rationale. If the reviewers still disagree after reviewing the same evidence, escalate the outcome to a third methodologist for an independent judgement.
- Document the final consensus profile with its rationale, so the resolved rating and the reasoning behind it are traceable in the review or guideline record.
Cochrane's GRADE resources and Summary of Findings guidance describe the domain structure and GRADEpro workflows that support this outcome-level comparison. Cochrane Handbook Chapter 15 covers the interpretation of confidence intervals, exact P values, and absolute and relative effects that often underlie an imprecision or inconsistency disagreement.