How do I respond to reviewer comments for a medical journal? (template and examples)

What goes into a response to reviewers?

A revision is a package, not a single letter. Editors and reviewers check your revision against the original comments one by one, so everything you send should make that check fast. Elsevier's guidance on responding to reviewers lists four parts, and most medical journals ask for the same set in their submission system:

Some journals also want updated reporting checklists, revised figures as separate files and new supplementary tables. Read the decision letter line by line before you start.

How do you write a point-by-point response letter?

Most published advice agrees on the same core method. Hywel Williams' paper on replying to referees boils it down to three rules: respond completely, respond politely and respond with evidence. William Noble's "Ten simple rules" adds the mechanics. In practice:

1. Wait a day before you write anything. First reactions to criticism are rarely the ones you want on paper.
2. Copy every comment, verbatim, into a new document. Do not paraphrase or shorten. Number them (Editor 1, R1.1, R1.2, R2.1 and so on) and split any comment that raises more than one issue into separate numbered points.
3. Triage. Label each point as agree and fix, clarify, new analysis, or disagree. This shows the real workload and which points the co-authors need to discuss.
4. Revise the manuscript first, starting with the biggest structural changes. Polishing sentences that a later change will delete wastes time.
5. Answer each point directly in the first sentence ("We agree", "Yes, we have added…", "We respectfully disagree…"), then explain.
6. Say exactly what changed and where. Give the page and line numbers, state whether they refer to the revised or the tracked version, and quote the new text so the reviewer does not have to open the manuscript.
7. Use consistent typography. For example, reviewer comments in bold, responses in plain text and quoted manuscript changes indented or in italics. Explain the convention in one line at the top.
8. Open with a short overview of the major changes (new analyses, restructured sections) before the detailed replies.
9. Run a consistency check last. Every number quoted in the response must match the revised abstract, tables and text. New analyses often change a denominator in one place and not another.

Response to reviewers template (fill-in)

Copy this structure and replace everything in square brackets.

Response to reviewers: [Manuscript ID], "[Title]"

Dear Dr [Editor's surname],

Thank you for the opportunity to revise our manuscript. We thank the reviewers for their careful reading. We have addressed every comment below. Reviewer comments are in bold, our responses are in plain text, and changes to the manuscript are quoted in italics. Page and line numbers refer to the [revised clean / tracked-changes] manuscript.

Summary of main changes

1. [New analysis, e.g. sensitivity analysis excluding X; new Supplementary Table S3]
2. [Restructured section, e.g. Methods: statistical analysis rewritten]
3. [Other major change]

Editor

E1. [Editor's comment, verbatim] Response: [Direct answer in the first sentence. What you did and why.] Change: Page [X], lines [X–Y]: "[new or revised text]"

Reviewer 1

R1.1 [Comment, verbatim] Response: [...] Change: Page [X], lines [X–Y]: "[...]"

R1.2 [Comment, verbatim] Response: [If no change was made, say so and give the reason with evidence.] Change: None. [Or: added to Limitations, page X, lines X–Y.]

Reviewer 2

R2.1 [...]

Sincerely, [Corresponding author, on behalf of all authors]

Example responses (synthetic)

The three exchanges below come from a synthetic retrospective cohort study (gestational diabetes and neonatal hypoglycaemia). The study, numbers and reviewers are invented for illustration.

Example 1: a clarification.

R1.3 How were missing BMI values handled? This is not described. Response: Agreed; this was missing from the Methods. Pre-pregnancy BMI was missing for 6.1% of participants (synthetic). We used multiple imputation by chained equations with 20 imputed datasets and now report a complete-case analysis as a sensitivity analysis. Change: Methods, page 7, lines 142–149: "Pre-pregnancy BMI was missing for 6.1% of participants. Missing values were imputed using multiple imputation by chained equations (20 datasets)…" Results of the complete-case analysis are in new Supplementary Table S4.

Example 2: a request for a new analysis.

R2.2 The association may be driven by women with pregestational diabetes misclassified as GDM. Please exclude women diagnosed before 20 weeks. Response: We agree this is an important check and have run it. After excluding 214 women diagnosed before 20 weeks (synthetic), the adjusted odds ratio was 1.38 (95% CI 1.05–1.81), compared with 1.42 (95% CI 1.10–1.83) in the main analysis. The conclusion is unchanged. Change: Methods, page 8, lines 160–163 (new sensitivity analysis described); Results, page 11, lines 231–234; new Supplementary Table S5.

Example 3: a polite disagreement.

R2.4 Please adjust the main model for gestational age at delivery. Response: We considered this carefully. Because gestational diabetes may lead to earlier delivery, gestational age is plausibly on the causal pathway between exposure and outcome, and adjusting for a mediator would estimate a different quantity from the total effect we set out to measure. We therefore kept the main model unchanged, but we now report the model adjusted for gestational age as a secondary analysis (aOR 1.29, 95% CI 0.98–1.70, synthetic) and discuss the difference. Change: Supplementary Table S6; Discussion, page 14, lines 301–308.

Notice the pattern: a direct first sentence, the evidence, the exact change and its location. In Example 3 the authors did not simply refuse. They ran the analysis, reported it, and explained why it is not the main result. If your study needs a reporting checklist the reviewer asked about, our guide to choosing the right reporting guideline helps you find it.

How do you respond when you disagree with a reviewer?

Disagreeing is acceptable. Disagreeing without an argument is not. Both the Elsevier and Taylor & Francis author guidance say that respectful disagreement is fine when you back it with a sound, evidence-based case.

• Concede what you can. Most disagreements are partial. Accept the valid part, fix it, and disagree only with the rest.
• Assume the misunderstanding is your fault. If a reviewer misread a passage, other readers will too. Noble advises rewriting the passage even when you think it was already clear.
• Cite evidence: a methods paper, a reporting guideline, your own data. Opinion against opinion persuades nobody.
• Offer a compromise: a sensitivity analysis, a supplementary table or a sentence in the Limitations.
• Never skip the comment. An unanswered point reads as an ignored point.

What if two reviewers ask for opposite changes?

Contradicting requests are common, for example one reviewer wants a shorter Discussion and another wants a new paragraph on mechanisms.

1. Choose the option you can defend scientifically.
2. In the reply to each reviewer, say what you did and acknowledge that the other reviewer asked for something different, without naming who asked for what in a way that sets them against each other.
3. Explain the conflict and your choice in the cover letter to the editor. Noble recommends using the letter to the editor for exactly this kind of issue. The editor makes the final call.
4. If the conflict is fundamental, for example two incompatible analysis strategies, it is reasonable to ask the editor which direction to take before you do the work.

What if the reviewer asks for new analyses or new data?

Noble's advice is to do what the reviewer asks whenever it is feasible, even if you doubt it will add much, then report the result and explain whether it belongs in the main paper or the supplement. Updated tables are easier to judge when the letter shows the old and new results side by side, as Taylor & Francis suggests.

When an analysis is impossible, for example the variable was never collected, say so plainly and add it to the Limitations. Never estimate or approximate a result you did not run. If a co-author is still running an analysis while you draft, keep an explicit placeholder such as [XX] highlighted in the draft, and check before sending that none remain. Reviewers who ask for effect sizes and intervals instead of bare P values are common; our guide to reporting P values and confidence intervals covers the format they expect.

What tone should a response letter use?

Write as a colleague answering a colleague: brief, specific and calm.

If a comment is rude or seems biased, answer the scientific content only and raise the concern with the editor in the cover letter.

How should you track changes and cite page and line numbers?

• Keep continuous line numbering switched on in the revised manuscript.
• Tracked and clean versions paginate differently. Decide which version your page and line numbers refer to, say so at the top of the letter, and stick to it.
• Generate the line numbers after the last edit. A late change in the Introduction shifts every reference after it.
• If the journal's system creates its own PDF, check the numbers against that PDF before you submit.

What do you do after a minor revision, a major revision or a reject-and-resubmit?

Decision labels and deadlines vary between journals, so the decision letter is the authority. As a rule of thumb:

Williams advises that, where possible, authors consider revising and resubmitting rather than sending the paper elsewhere. After a straight rejection, the same comments often come back from the next journal's reviewers, so fix them first. When you resubmit, the cover letter for journal submission should state that the paper is a revision and summarise what changed.

Can you use AI to draft a response to reviewers?

AI can help organise comments into a numbered table, draft first replies and check that every comment has an answer. CliniPaper parses the reviews into a point-by-point table and drafts responses for you to edit. The authors remain responsible for every statement and number in the letter, and you decide what to concede. Two cautions: Springer Nature says manuscripts and peer review reports must not be shared with unsecured or public AI systems, and many journals require you to disclose AI use. See our guide to the AI disclosure statement for what to write.

Common mistakes reviewers flag

Vickers and colleagues' guide in European Urology catalogues the mistakes authors make most often. These are the ones that most often send a revision back for another round:

• Comments paraphrased, merged or silently skipped.
• "Done" or "See revised manuscript" with no location and no quoted text.
• Page and line numbers that do not match the version the reviewer opens.
• A new analysis reported in the letter but not added to the manuscript or supplement, or the reverse.
• Numbers that changed in the Results but not in the abstract or tables.
• Defensive or sarcastic replies, or arguing without evidence.
• Changes made that no reviewer asked for, without flagging them.
• Missing the deadline without asking the editor for an extension.