How to write a peer review

Open with a paragraph summarising the work in your own words and stating what it contributes. This shows you understood it and orients the editor.

Follow with numbered major comments, those affecting the validity of the conclusions. Close with minor ones: wording, figures, references, formatting. Separating major from minor is the part editors and authors appreciate most.

Is the question framed so that these data can answer it? Does the design support the intended inference? Is the sample size justified? Are the controls and comparators appropriate?

In quantitative work, check that the tests match the nature of the data, that multiple comparisons are declared, and that effects come with measures of uncertainty rather than significance values alone.

Check that the numbers in the text match the tables and figures, that conclusions do not exceed what the data show, and that limitations are declared.

Review data and code availability, preregistration where applicable, and the conflict-of-interest and ethics approval statements. The system expects these checks from the reviewer, not the editor.

Each comment should state the problem, where it is, and what you expect from the author. "Section 3.2 does not report the exclusion criterion; please specify it" is actionable. "The method is weak" is not.

Criticise the work, not the people. Avoid demanding citations to your own work without cause, and do not impose the study you would have run: the goal is for this manuscript to be sound, not to become a different one.

Keep the tone of the report consistent with the recommendation: a text full of major objections ending in "accept as is" leaves the editor guessing.

If you recommend rejection, state the reason plainly and, where possible, say what would be needed for the work to be publishable elsewhere. Declare conflicts of interest before accepting the review.