Give clear, respectful, evidence-tied feedback with actionable fixes; cite checklists; flag bias or conflicts; separate major issues from minor edits.
What good peer review achieves
Done well, a review lifts patient-facing science, helps editors reach sound decisions, and guides authors toward a cleaner, more credible paper. Your notes should be fair, specific, and anchored to the study’s question, methods, and claims. Keep the reader in mind: the clinician who will safely apply the result, the trainee who will learn from it, and the patient whose care hinges on reliable evidence. The aim is a clear path from problem to fix, not a verdict delivered from a distance.
Three roles you balance
- Advisor to the editor: say what works, what does not, and why the manuscript should be revised, declined, or moved along.
- Coach for the authors: point to concrete changes that raise clarity, validity, and transparency.
- Guardian of ethics: watch for conflicts, privacy risks, plagiarism, duplicate posting, and unfair claims.
Study-type checkpoints you can use
Use a quick map that matches the design. Link feedback to reporting checklists when helpful, and ask for additions instead of vague rewrites.
| Study type | What to check | Quick prompts you can paste |
|---|---|---|
| Randomised trial | allocation, blinding, pre-registration, sample size, flow of participants, outcome hierarchy | “Please show the randomisation method, concealment, and who was blinded; align outcomes with the registry; add a flow chart.” |
| Cohort or case-control | exposure/Outcome definition, confounding control, missing data, sensitivity runs | “Define exposures and outcomes with sources; justify confounder set; report missingness and any bias checks.” |
| Systematic review/meta-analysis | protocol, search dates, inclusion criteria, risk-of-bias tool, heterogeneity, small-study signals | “Cite the protocol; show full search strings and dates; add a risk-of-bias table; cover heterogeneity and small-study effects.” |
| Diagnostic study | reference standard, spectrum of disease, thresholds, calibration, clinical use | “Clarify the reference standard and disease spectrum; show cut-points and calibration; state the intended use.” |
| Prediction model | derivation/validation split, handling of missing data, overfitting, transportability | “State sample splits; describe imputation; show shrinkage or penalisation; test on an external set.” |
| Qualitative study | sampling strategy, saturation, coding approach, reflexivity, quotes that back themes | “Describe sampling and saturation; explain coding; reflect on researcher position; link themes to quotes.” |
Giving peer review comments in medicine: step-by-step
1) before you accept the invitation
Confirm fit, time, and independence. Decline fast if you lack topic comfort, cannot return a review on the proposed timeline, or have links that could sway judgment. Examples include grants with a competing team, shared authorship in the recent past, or work that overlaps too closely. If the paper is near your niche yet you can keep a clear boundary, disclose the link and let the editor decide.
2) read the paper twice
First pass: read without notes to see the story, audience, and claim. Second pass: mark issues tied to a section, not just sentences. Keep a list of major items that influence validity and a shorter list of minor edits that smooth reading. Quote line numbers only to anchor a point; avoid “gotcha” tone.
3) check core domains
Scope: Do aims, data, and conclusions align. Methods: Are design choices suited to the question. Statistics: Are estimates, intervals, and model checks presented with enough detail to judge trust. Ethics: Are approvals, consent, and data privacy handled. Transparency: Are data, code, and materials shared or explained when not shared.
4) write two parts
Comments to authors: a clear, numbered note that opens with what works and then lists changes in priority order. Comments to editor: a confidential note that states your confidence in the work, any concerns about authorship or overlap, and a bottom-line recommendation with reasons.
5) mark major before minor
Start with issues that affect trust: design mismatches, outcome switching, selective reporting, absent power or precision, analytic choices that shift effect sizes, and unsupported clinical claims. Move next to clarity points: title accuracy, abstract truth, flow, tables that match the text, and figure legends. Close with style edits and small fixes.
6) tie every point to action
Each comment should point to a fix: add, remove, reframe, or show. Ask for missing details, show where a claim outgrows the data, or suggest a clearer figure. When you ask for new analyses, say why they matter to the claim and keep them proportionate to the study’s aim and sample.
How to give peer-review feedback in medical journals
Build a clean structure
Editors and authors value structure that repeats across your reviews. Use short sections with headers such as “Methods,” “Results,” and “Claims.” Number your points. Keep each item to a single ask. If a point spans two sections, split it. This format speeds triage, lowers misreadings, and shows a path to revision.
Lead with strengths
One or two sentences should name the study’s main contribution: a fresh dataset, a tough question, or a careful analysis. Praise that maps to real strengths does more than soften tone; it tells authors what to keep while they change the rest.
Be specific and traceable
Point to tables, figures, and sections instead of pages of prose. Replace “unclear” with the exact sentence that needs a rewrite and, when possible, offer a model sentence. When a method is under-described, list the missing pieces. Specificity shortens cycles and raises clarity for readers down the line.
Hold claims to evidence
Match each claim to the data that back it. If a confidence interval includes clinically trivial and clinically large effects, the paper should say so plainly. If a subgroup run lacks a plan, call it exploratory and ask that it be framed as such. Keep mechanistic explanations modest unless data show otherwise.
Mind tone
Keep it professional
Replace “obviously” with “the data do not show X.” Avoid personal remarks. Write as if your name will be known, even when the process is blind. Calm, specific, and steady beats sharp zingers every time.
Respect privacy and property
Do not share the manuscript or its data outside the journal. Do not use ideas or text you learned from the submission. If you spot a related preprint or prior posting by the same group, flag it for the editor and ask how to proceed.
Ethics, policies, and helpful links
Confidentiality, conflict disclosure, and fair treatment sit at the core of sound reviewing. Many journals point reviewers to shared guidance from editorial bodies and reporting networks. When a point in your review leans on a standard, link it so authors know the yardstick. Two anchors many editors cite are the COPE peer reviewer guidance and the ICMJE recommendations. For reporting depth, point authors to the EQUATOR Network or a study-specific checklist such as PRISMA or CONSORT.
When your review changes course
If you realise mid-review that a tie or bias exists, stop and tell the editor. If new work appears that alters how you see the manuscript, alert the editor and ask whether to proceed. If you need more time, ask before the due date.
Template you can adapt
Borrow this outline, tweak phrasing, and keep it handy. You can save a version for trials, another for observational work, and one for reviews. The headings stay the same; the asks change.
Title: [repeat the manuscript title] Summary (2–3 lines): [what the paper tries to do and what it finds] Overall view: [strengths in one line + one line on the main fix needed] Major points 1) [Design or methods] — [what is missing + action] 2) [Outcomes] — [align registry/text; add hierarchy] 3) [Analysis] — [justify model; add checks] 4) [Claims] — [limit scope to data; adjust language] Minor points 5) [Clarity] — [short phrase; action] 6) [Tables/figures] — [align with text; add legend detail] 7) [Style] — [term consistency; abbreviations] Comments to editor (confidential) • Suitability for the journal’s audience? • Any overlap or ethical flags? • Bottom line: Reject / Major revision / Minor revision / Accept
Language that lowers friction
Short, neutral phrases keep heat down and progress up. Swap vague labels for direct asks, and pick verbs that signal the next step.
| Goal | Helpful wording | Why it works |
|---|---|---|
| Ask for detail | “Please describe the allocation and who remained masked at each step.” | Names the missing piece and the location where it belongs. |
| Dial back a claim | “The data show association, not causation; suggest softening the language in the title and abstract.” | Pairs a boundary with a specific edit. |
| Request a new run | “Given missingness patterns, a complete-case run plus one imputation run would help gauge stability.” | Sets scope and purpose for the ask. |
| Flag bias | “Please disclose funding and roles; note how they might shape interpretation.” | Links transparency to reading the result. |
| Tighten structure | “Merge duplicate methods; move design choices to the Methods; shorten the Discussion.” | Shows placement and scale instead of a blunt “too long.” |
Common pitfalls and how to avoid them
Overshooting the ask
A review is not a fresh study. Avoid demands that need new data or a new site unless the claim rests on such work. When in doubt, ask the editor whether a narrower claim would fit the journal.
Vague labels
Words like “unclear,” “weak,” or “interesting” tell authors nothing. Replace them with a short description of the issue and the edit that would fix it. If many small edits repeat, show one model fix and say “apply across the text.”
Missed harm
Look for privacy gaps, stigmatising language, and group harms. Ask for lay-friendly wording in abstracts and press notes so readers do not overread a finding or misread a risk.
Ignored readers
Match your asks to end users plainly. A busy clinician needs size and direction of effect, thresholds, and practical trade-offs. A policy reader needs population, setting, and generalisability. Ask for tables and figures that serve those needs.
Calibrate asks to journal scope
Not every manuscript belongs in the same venue. A pragmatic regional journal may prize clear methods and bedside value; a methods-heavy outlet may value deeper modelling or measures. Shape your asks to the fit. If the design is sound yet the paper sits outside scope, say so in confidential note and suggest a home that matches the audience. When the work fits but space is tight, guide authors toward leaner tables, a tighter title, and figures that carry the message without extra ink.
Speed, format, and delivery
Reply to the invitation fast, return the review on time, and use the journal’s form or portal headings. Keep a local copy. If the review is open, sign your name only if you can keep the same tone you would use in a blind process and if no safety risk exists. If the journal invites you to review a revision, check whether prior asks were met and state where gaps remain.
A quick checklist before you hit submit
- Have you stated the paper’s contribution in two lines.
- Are major issues listed before minor edits.
- Does each point ask for a clear action.
- Have you linked to a reporting checklist when useful.
- Did you avoid personal remarks and loaded words.
- Did you add a short confidential note to the editor.
- Did you declare ties and confirm confidentiality.
Further reading and model standards
Many journals ask reviewers to follow shared standards. You can learn the basics in an hour and then keep the links handy for later work. Start with the COPE peer reviewer guidance, the ICMJE section on reviewer duties, and a reporting hub such as the EQUATOR Network.