Build a protocol-deviation table that lists every included study, tests it against the protocol's eligibility criteria, and records each deviation with its justification. PRISMA 2020 frames transparent reporting around why a review was done, what methods were used, and what studies were found and included, which gives you the comparison points.
- Extract the protocol's eligibility criteria into separate testable fields: population, intervention or exposure, comparator, outcomes, study design, and any date or setting limits. Expected result: a checklist where each criterion can be marked met or not met for a single study.
- List every study the review includes, then mark each criterion for each study. Expected result: a matrix that shows which studies fail which criterion, so unexplained inclusions stand out.
- For each failed criterion, search the review text and protocol amendments for a stated reason, such as a newly available trial or a clarified definition. Expected result: each deviation is labeled justified (with the reason and where it appears) or unexplained.
- Compare the documented search and selection process against the protocol, since the Cochrane Handbook describes planning searches, selecting sources, developing strategies, and managing records. Expected result: you can see whether the included set follows from the stated selection process.
- Summarize the table into counts of justified and unexplained deviations and flag unexplained ones for the review authors or editors. Expected result: a short list of items that need clarification before the review's conclusions are relied on.
Keep the deviation table separate from risk-of-bias assessment. RoB 2 judges a specific randomized-trial result across bias domains using signalling questions and written justifications; eligibility is a protocol question and uses different criteria. Mixing the two obscures which problem you found.
For guideline work, report the table alongside the review so readers can see which inclusions departed from the protocol and why.