Peer Review in Medical Journals: The Complete Guide

Screening, reviewer invitations, reports, a decision. Knowing what happens at each stage tells you what to fix before you submit and what to do when the letter arrives.

Guide · 19 min read · Directive Publications Editorial Team

What peer review in medical journals is for

The ICMJE Recommendations define peer review as the critical assessment of manuscripts submitted to journals by experts who are usually not part of the editorial staff. They are candid about its limits. The actual value of peer review is widely debated, they say, but the process gives a manuscript a fair hearing among members of the scientific community, helps editors decide which manuscripts suit their journals, and often helps authors and editors improve the quality of reporting.

That description sets three expectations:

• Reviewers advise; editors decide. The ICMJE says a peer-reviewed journal is under no obligation to send a manuscript for review or to follow reviewer recommendations, favourable or negative. The editor is ultimately responsible for everything the journal publishes, and can reject an article at any time before publication, including after acceptance if concerns arise about the integrity of the work.
• Review checks the report, not the underlying research. Unless a journal asks for raw data or an independent analysis, reviewers judge whether the question, methods, analysis and conclusions hold together as reported.
• Not everything in a journal is externally reviewed. Editorials, commentaries and some commissioned pieces may be assessed only by editors. BMJ, for example, states on every article whether it was commissioned and whether it was internally or externally peer reviewed.

Before you cite a paper as peer reviewed in a grant application or guideline document, check its article type and any note on how it was reviewed.

The peer review process, stage by stage

Submission systems label the stages differently, but a research manuscript that is sent for external review typically moves through this sequence:

BMJ's author guidance on peer review shows how this works in practice. Editorial staff run quality checks, then an editor evaluates the article for scope, quality and fit and rejects papers that fall short. For most articles a minimum of two reviews is required, several prospective reviewers may be invited before enough accept, and if a reviewer withdraws, the search starts again. That is why a status such as "reviewers invited" can stay unchanged for some time without anything being wrong.

Of those rows, the editorial assessment is the one you can most influence in advance, through a cover letter that answers the fit question on the first page.

No single timeline applies across journals. The ICMJE asks editors to process manuscripts as promptly as their resources allow, reject unsuitable ones as soon as possible, and answer status requests in a timely manner. If your manuscript has clearly passed a journal's published turnaround figures, a short enquiry quoting the manuscript number is reasonable.

Editorial screening: who decides what goes out for review

The first decision on your manuscript is made by an editor, not a reviewer: the editor-in-chief, a deputy, or a handling editor with expertise in your field. Declining a paper without review, often called a desk rejection, is normal editorial work rather than a verdict on the science, and at a selective journal it is the most likely outcome.

For a medical manuscript, the screening editor can quickly judge a short list of things:

• Fit. Do the topic, readership and article type match the journal?
• Trial registration. The ICMJE recommends that editors require clinical trials to be registered in a public trials registry at or before the time of first patient consent for enrolment, as a condition of consideration. A late registration can end the process before review begins.
• Ethics. For research involving people, the ICMJE expects the methods to state that an independent review body, such as an ethics committee or institutional review board, approved the research.
• Reporting. Some journals require checklists for particular study types, such as CONSORT for randomised trials, and the ICMJE tells authors to send them when requested.
• Declarations. Competing interests, authorship, and any related work that might be regarded as redundant publication.
• Basic soundness. Whether the design can plausibly answer the question.

After a desk rejection, read the reason before acting. If it cites scope or priority, do not rewrite the science; choose a journal that fits better. If it cites a missing registration, ethics statement or checklist, fix that before you submit anywhere. If it names a design or analysis problem, take it seriously, because the next editor is likely to notice it too.

How journals find and select reviewers

The ICMJE makes reviewer selection the journal's responsibility. Editors must give reviewers everything relevant to the evaluation, including supplementary material, and interpret their comments in the light of their declared relationships and activities.

Editors look for reviewers whose expertise matches the manuscript's subject and methods. COPE's Ethical Guidelines for Peer Reviewers ask potential reviewers to represent their expertise accurately, with verifiable contact details, and call impersonating another person during review serious misconduct. They also say who should decline an invitation:

• Anyone currently employed at the same institution as any of the authors
• Recent mentors, mentees, close collaborators or joint grant holders of the authors, with COPE giving the past three years as an example
• Anyone reviewing only to see the manuscript, with no intention of submitting a review
• Anyone with a very similar manuscript in preparation or under consideration at another journal

The ICMJE adds that reviewers must disclose relationships or activities that could bias their opinion, recuse themselves if the potential for bias exists, and never use knowledge of the work before publication to further their own interests.

Some journals invite authors to suggest or oppose reviewers, but the editor chooses. Run every name you propose through the COPE tests above, and see our guide to suggesting reviewers ethically for how to build the list.

Clinical manuscripts may also get specialist scrutiny. The ICMJE notes that some editors ask an independent biostatistician to analyse trial data before accepting a study, and encourages editors to review research protocols and statistical analysis plans as part of peer review. Have both ready to upload at submission.

Types of peer review

Labels such as "single-blind" and "open review" have long been used loosely. The ANSI/NISO Z39.106-2023 Standard Terminology for Peer Review, which grew out of a working group set up by STM, the international association of scientific, technical and medical publishers, describes review models with four elements and replaces "blind" with "anonymized". The first element is identity transparency:

The other three elements describe the rest of the process:

• Who the reviewer interacts with. The editor, which is the traditional model; other reviewers, for example by seeing or commenting on each other's reports before recommending a decision; or the authors directly.
• What review information is published. Options range from none, through review summaries or full review reports (always, or only if the author or reviewer opts in), to the submitted manuscript, the decision letters and author responses, and the identities of reviewers and handling editors.
• Post-publication commenting. Open to anyone, or by invitation only.

Terms such as "open" and "transparent" peer review are applied to several different combinations of these elements, so ask about the elements rather than the label. Our guide to transparent peer review explains what the combinations mean for authors and their response letters. Reviews can also move with a manuscript: COPE notes that some publishers transfer reviews between journals in their portfolio, sometimes called portable or cascading review.

Before you submit, find the journal's description of its process, which the ICMJE says journals should publish for all types of manuscripts. If you cannot find one, treat that as a warning sign and check the journal against the red flags of predatory publishing before trusting it with your paper.

Preparing a manuscript for double-anonymized review

Instructions differ between publishers, so follow the journal's own. Guidance from publishers such as Nature Portfolio and IOP Publishing covers points like these:

• Remove author names and affiliations from the manuscript, figures, supplementary files and file names.
• Refer to your own earlier work neutrally: "as previously shown", not "as we have shown".
• Clear author names from the document properties of every file.
• Put acknowledgements and funding where the journal specifies, such as a title page, the cover letter or the submission form.
• Do not sign your response letter at revision.

Medical manuscripts carry an extra risk. An ethics statement naming your institution's committee, or a trial registration number leading to a registry record that names the principal investigator, can reveal who you are. Do not delete required statements; check where the journal wants them. IOP Publishing, for example, asks for ethics statements in the submission system rather than the manuscript.

What reviewers assess in a medical manuscript

Journals give reviewers their own forms, and COPE asks reviewers to read the manuscript, supplementary data and ancillary material, such as required ethics and policy statements, before starting. Whatever the form, expect a reviewer of clinical research to ask questions like these, and make each one easy to answer:

The statistics rows follow the ICMJE's guidance on reporting methods and results, and the ICMJE points authors to the EQUATOR Network for reporting guidelines.

COPE draws a line that helps when comments arrive: the reviewer's job is to comment on the quality and rigour of the work, not to extend it beyond its scope, and reviewers should say which extra investigations are essential to support the claims and which would only strengthen the work. When a report does not make that distinction, it is reasonable to ask the editor which requests are essential.

Confidentiality, conflicts and AI in peer review

The ICMJE treats submitted manuscripts as privileged communications and the authors' confidential property. Editors must not share information about a manuscript, even whether it has been received, with anyone but its authors and reviewers. Reviewers must keep manuscripts confidential, must not publicly discuss the authors' work or appropriate their ideas before publication, and should destroy their copies after submitting their reviews.

Senior authors review too, and these are the points where good intentions most easily cause problems:

• Asking a trainee to help. COPE says reviewers must not involve anyone else, including early career researchers they are mentoring, without first getting the journal's permission, and that a student who performs the review under supervision should be named as the reviewer of record. The ICMJE asks reviewers to acknowledge anyone who helped in their comments to the editor. Our guide to mentoring students through their first peer review sets out how to do that properly.
• Using AI tools. The ICMJE Recommendations, most recently updated in January 2026, say reviewers must follow the journal's policy on AI or ask permission before using it. Confidentiality may prohibit uploading a manuscript to software or AI tools where confidentiality cannot be assured, unless the journal explicitly permits that use. Reviewers who use AI should disclose it and make sure the output is appropriate and valid.
• Requesting citations. COPE says reviewers should not suggest that authors cite the reviewer's or an associate's work merely to increase its citation count or visibility.
• Contacting authors. COPE says reviewers should not contact authors directly without the journal's permission.
• Discovering a conflict partway through. Notify the journal and, as COPE advises, stop looking at the manuscript until it replies.

Confidentiality works in both directions. Do not try to work out who reviewed your paper; if you have evidence that a reviewer had an undisclosed conflict or misused your work, set it out factually for the editor.

Decisions, revisions and reviewer disagreement

Decision names and meanings vary between journals, so read the letter rather than the label. These are the common outcomes:

The letter may not show everything the editor considered. COPE notes that most journals let reviewers send confidential comments to the editor alongside the report you see, though it expects the two to be consistent.

Reviewers often disagree. A study of five years of reviews at the Journal of General Internal Medicine (Kravitz and colleagues, PLoS ONE 2010;5(4):e10072) found that reviewers agreed on whether to reject a manuscript at levels barely beyond chance, yet editors placed considerable weight on their recommendations. When your reports conflict:

1. Look for the editor's steer. A letter that highlights particular comments is telling you which view carries weight.
2. Follow the evidence, not the majority. Make the change best supported by your data and the literature, and explain the choice to both reviewers.
3. Ask before revising if requests cannot both be met. A short query to the editor is quicker than a revision round spent on the wrong change.

Our guide to responding to peer review comments covers the response letter itself. If you move to another journal after a rejection, the ICMJE says it is helpful to include the previous editors' and reviewers' comments, with your responses, and that editors encourage it.

Appeals and complaints

The ICMJE says journals should clearly state their appeals process and have a system for responding to appeals and complaints. Find that statement before you write anything, and follow it.

An appeal is worth making when the decision rests on something you can show is wrong: a factual error, a misreading of your methods, or a reviewer with a conflict of interest you can document. It is rarely worth making when the editor has judged the paper a poor fit or a low priority, because that judgement is the editor's to make and a second letter will not change it.

Keep the letter short and specific:

1. The manuscript number, title and decision date.
2. Each point you contest, quoting the comment with page and line references.
3. The evidence that the point is mistaken.
4. What you are asking for, such as reconsideration or an additional reviewer.

Send one factual letter. While an appeal is open, do not submit the manuscript elsewhere; if you decide to move on, withdraw the appeal in writing first. Complaints about how a manuscript was handled, as distinct from the decision, should also follow the route the journal publishes.

What peer review cannot reliably catch

Peer review is not an audit of the underlying research, and the evidence suggests it does not reliably catch serious errors in the report either. In a BMJ randomised trial of reviewer training, 607 reviewers were sent test papers into which nine major methodological errors had been deliberately inserted. An analysis of that trial's data published in 2008 (Schroter and colleagues, Journal of the Royal Society of Medicine 2008;101:507-14) found that reviewers identified an average of 2.58 of the nine at baseline, and that short training packages made only a slight difference. A fabricated dataset that is internally consistent gives a reviewer even less to notice.

When a reviewer suspects a problem, such as possible misconduct or substantial similarity to another paper, COPE asks them to tell the editor directly rather than investigate on their own. The ICMJE says editors should then follow procedures such as those detailed by COPE, consider informing the institutions and funders, and may publish an expression of concern while those procedures run. If an investigation proves misconduct, the article should be retracted.

Scrutiny continues after publication. The ICMJE says medical journals should give readers a way to submit comments and criticisms, that authors must respond and cooperate with requests for data, and that investigators should keep the primary data and analytic procedures behind published results for at least ten years.

Because reviewers cannot do this checking for you, build it into your group's routine before submission:

• Have someone other than the original analyst re-run the primary analysis where possible.
• Check that every number in the abstract matches the tables.
• Keep unprocessed original images and record any adjustments.
• Archive the data, code and analysis plan together, labelled by version.

Why review is slow, and what that means for your schedule

Reviewers are volunteers, as BMJ's author guidance reminds authors, and that is why the invitation round can take as long as the review itself.

Incentives have been tested and they do not change the picture much. A quasi-randomized study at Critical Care Medicine (Cotton and colleagues, Critical Care Medicine 2025;53:e1181-e1189) offered reviewers $250 to accept an invitation. A greater share of invitations produced a submitted report in the incentive arm, 49.8% against 42.2%, reviews came back about a day sooner, and editors rated the quality of the reports no differently between the arms. Payment is still an experiment, not the norm, so assume the goodwill model applies.

Three practical consequences for a senior author:

• Make the invitation easy to accept. A clear title, a structured abstract and a named design tell an invited expert in a minute whether the paper is theirs to judge. Vague framing costs you declines you never see.
• Budget for the queue. When a grant report or thesis deadline depends on publication, allow for the invitation round and for a second review round if the decision is major revision.
• Take your turn. COPE asks authors who have benefited from peer review to consider reviewing, accepting only invitations they have the expertise and time to complete. The ICMJE says journals should acknowledge reviewers' contribution and tell them the final decision, and reviews can be added to the peer review section of an ORCID record by a trusted organisation, such as a publisher or a review recognition service, with your permission.

Before you submit: a peer review readiness checklist

1. You have read the journal's description of its review process and know its identity model, reviewer interactions and what it publishes
2. If review is double anonymized, your files follow the journal's anonymization instructions
3. Any clinical trial was registered in a public registry at or before the time the first participant consented to enrolment, and the registry and number appear in the manuscript
4. Ethics approval and consent statements are complete and placed where the journal asks
5. The reporting guideline checklist is completed if the journal requests one
6. The protocol and statistical analysis plan are ready to upload
7. Prespecified and exploratory analyses are labelled as such
8. The data availability statement matches the journal's policy
9. Competing interests are declared on the form the journal requires
10. Any suggested reviewers pass COPE's conflict tests
11. If the paper was rejected elsewhere, previous reviews and your responses are ready to share
12. Every author has approved the final version

Write for the people who will check it

Peer review in medical journals is a sequence of judgements by busy people: an editor asking whether the paper belongs, reviewers asking whether it holds up, and the editor weighing what they said. A manuscript that answers those questions on the page meets fewer surprises.

Directive Publications sets out how its review process works, from editorial checks to decisions, on its peer review page. Read it alongside the checklist above, and when your manuscript is ready you can submit it directly.