Research Integrity in Medical Publishing: The Complete Guide

Medical journals check a paper's integrity against the records behind it: approvals, registrations, disclosures and original data. This guide shows what each check needs.

Guide · 19 min read · Directive Publications Editorial Team

What research integrity in medical publishing covers

Research integrity is what lets a reader rely on a published paper without redoing the work. The study was done as described, the data are real and completely reported, the people named did the work, and any interests that could have shaped it are on the page. In medicine an unreliable paper keeps causing harm for as long as reviews, guidelines and new trials cite it.

The UK's Concordat to Support Research Integrity, refreshed in 2025, rests on five principles: honesty; rigour; transparency and open communication; care and respect; and accountability. The European Code of Conduct for Research Integrity, revised in 2023, names reliability, honesty, respect and accountability.

Research ethics is chiefly about protecting participants through approval, consent and confidentiality. Integrity asks whether the research and its reporting can be trusted. A trial can hold valid ethics approval and still switch its primary outcome, so journals check both.

Responsibility is shared, and each party holds a different part of it:

The standards medical journals apply

Most integrity requirements you meet at submission come from a short list of documents.

• The ICMJE Recommendations. The Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals cover authorship, disclosure, protection of research participants, trial registration, data sharing, overlapping publication, corrections and misconduct. The current version was updated in January 2026 and includes a section on the use of AI in publishing.
• COPE guidance. The Committee on Publication Ethics publishes the guidelines and flowcharts editors follow when they handle concerns; it revised its retraction guidelines and published new expressions of concern guidelines in 2025. COPE retired its Core Practices in 2024 and published a Code of Conduct for its members in July 2026, so older policies may still cite the Core Practices.
• The Declaration of Helsinki. The ICMJE asks that human research be planned, conducted and reported in accordance with the Declaration as revised in 2024.
• Reporting guidelines. CONSORT 2025 for randomised trials, SPIRIT 2025 for trial protocols, PRISMA for systematic reviews, STROBE for observational studies, STARD for diagnostic accuracy studies, CARE for case reports and ARRIVE for animal research. The EQUATOR Network collects them in one place.
• National misconduct frameworks. In the United States, research supported by the Public Health Service falls under 42 CFR Part 93, revised in 2024, and the new rules apply to allegations institutions receive on or after 1 January 2026. UK organisations covered by the Concordat were expected to follow its refreshed edition by 1 April 2026. Many European institutions build their policies on the European Code of Conduct.

The journal's instructions tell you what the manuscript must contain. Your institution's policy, your funders' terms and national rules decide how an allegation is handled.

Ethics approval and patient consent

The ICMJE Recommendations say authors should seek approval from an independent review body, such as an ethics committee or institutional review board, state it in the Methods, and be ready to provide documentation when editors ask. If there is doubt about whether the research followed the Declaration of Helsinki, the authors must explain their approach and show that the review body explicitly approved the doubtful aspects. Approval does not stop an editor forming their own judgement.

Consent to publish identifiable information is a separate step from consent to take part. Under the ICMJE Recommendations, names, initials and hospital numbers should not appear in text, photographs or pedigrees unless the information is essential for scientific purposes and the patient, or a parent or guardian, gives written informed consent for publication. An identifiable patient should be shown the manuscript and told whether identifiable material might be available on the internet. Nonessential identifying details should be removed.

Have these ready before you submit a study involving people or animals:

1. The committee's name, the approval reference and the date, stated in the Methods
2. The committee's written decision if it judged that the study did not need full review
3. How participants gave consent, or the committee's decision to waive it
4. Written consent for publication from any patient who could be identified, on the journal's own form if it has one
5. For animal research, confirmation that institutional and national standards for care and use were followed

A worked example. A case report shows a clinical photograph with the patient's eyes blacked out and gives the patient's initials in the legend. Both are problems. Initials are identifying information, and the ICMJE says that masking the eye region is inadequate protection of anonymity. Remove the initials, crop features that are not clinically essential, and obtain written consent for publication after the patient has seen the manuscript.

Trial registration and complete reporting

The ICMJE defines a clinical trial broadly: any research project that prospectively assigns people or groups to a health-related intervention to study its effect on a health outcome. Behavioural treatments, educational programmes and changes to the process of care count, as well as drugs and devices.

The ICMJE recommends that editors require registration at or before the time of first patient consent for enrolment, in a primary registry of the WHO International Clinical Trials Registry Platform or in ClinicalTrials.gov. The registration date is the date the materials were first submitted. Ethics approval does not meet the requirement, and the ICMJE says retrospective registration, for example at manuscript submission, meets none of its purposes.

Registration exists to prevent selective publication and selective outcome reporting. Authors are responsible for explaining any discrepancy between the results in a registry and those in the journal.

Trial reports submitted to ICMJE journals since 1 July 2018 must include a data sharing statement, and trials that began enrolling on or after 1 January 2019 must include a data sharing plan in the registration. The statement must say whether deidentified individual participant data will be shared, and "undecided" is not an acceptable answer. Journals outside the ICMJE may set different rules, so check the target journal's instructions.

For systematic reviews, PRISMA 2020 asks you to give the register name and registration number, or state that the review was not registered, and to explain any changes from the registered information or protocol.

Before submitting a trial report, compare the manuscript with the registry record:

1. The registration date is on or before the first participant's consent to enrol
2. The registration number appears at the end of the abstract and at the first mention of the trial acronym
3. Primary and secondary outcomes match the registered ones, or every change is explained
4. The data sharing statement matches the registered plan, and the registry is updated if the plan changed
5. The CONSORT checklist is completed if the journal asks for it

Authorship, disclosures and AI use

Authorship

The ICMJE recommends four criteria for authorship: a substantial contribution to the conception or design of the work, or to acquiring, analysing or interpreting its data; drafting the work or reviewing it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all aspects of the work. Everyone named should meet all four, and everyone who meets all four should be named. Acquiring funding, general supervision and writing assistance do not qualify on their own; acknowledge those contributors instead.

If authors ask to add or remove a name after submission, editors should seek an explanation and signed agreement from all listed authors and from the person concerned. Editors do not arbitrate disputes: the ICMJE says the institution where the work was done should be asked to investigate.

Competing interests and funding

Authors are responsible for disclosing all relationships and activities that might bias, or be seen to bias, their work, and ICMJE member journals require the ICMJE Disclosure Form. The ICMJE treats purposeful failure to report relationships specified on the journal's disclosure form as a form of misconduct. Articles should also describe any role the sponsor had in design, analysis and writing, any restrictions on submitting the report for publication, and whether the authors had access to the study data. Reviewers and editors must disclose their own relationships and step aside when a conflict exists.

When you are unsure whether a relationship is relevant, disclose it. As the ICMJE puts it, readers must be able to make their own judgements.

AI use

The ICMJE Recommendations ask journals to require disclosure of AI-assisted technologies at submission. Describe writing assistance in the acknowledgements. If AI was used to conduct the study, describe it in the Methods in enough detail to replicate, including the tool, version and prompts where applicable. AI cannot be an author, and the ICMJE warns that nondisclosure may require corrective action and may be construed as misconduct in some circumstances. Reviewers and editors should not upload a manuscript to an AI system where confidentiality cannot be assured, unless the authors have given explicit permission.

Data, images and the research record

Most integrity questions are eventually settled by the underlying records, so keep them where they can be produced. The ICMJE says investigators have a duty to maintain the primary data and analytic procedures behind published results for at least 10 years, and encourages deposit in a data repository. Your institution or funder may set a longer period.

The ICMJE lists deceptive manipulation of images as a form of data falsification. Recommendations on image integrity from STM, the international association of scientific, technical and medical publishers, set out practical limits worth adopting as lab rules:

• Adjust brightness or contrast only when the change is applied equally across the whole image and to the controls.
• Never use processing to emphasise one region, or the experimental data relative to the control.
• Mark the boundary where lanes from the same gel are placed side by side, and describe it in the legend. Never splice two gels to look like one.
• Give the source, context and reason whenever an image is reused, and obtain any licence needed.
• Keep the unprocessed files, because editors may request source data for any figure before or after publication.

The same recommendations say editors can refuse a manuscript that lacks compelling source data, even without proof of falsification, and that after publication missing source data may lead to an expression of concern or retraction.

Plagiarism, duplicate publication and salami slicing

Similarity software finds overlapping text, but a person decides whether it matters. For how editors weigh a match, see how to read a similarity report the right way.

Overlapping publication has its own rules under the ICMJE Recommendations:

Whenever a manuscript is closely related to other submitted or published work, say so in the cover letter and send the editor copies of the related material. The European Code of Conduct also names salami publication, chopping up results to increase the number of papers, as an unacceptable practice.

Definitions differ here. The US federal definition excludes self-plagiarism from research misconduct, while the UK Concordat counts undisclosed duplicate publication and duplicate submission as misrepresentation. Journals act on duplicate publication either way, because it can lead to data being counted twice and distort the evidence.

Peer review manipulation and paper mills

Peer review only protects the literature when it is independent. The ICMJE Recommendations treat submitted manuscripts as confidential, and reviewers must not discuss the work publicly, use its ideas before publication, or keep copies for their own use.

COPE describes systematic manipulation of the publication process as the repeated use of dishonest or fraudulent practices to influence independent peer assessment, to attribute authorship inappropriately, or to publish fabricated or plagiarised research. Its examples include peer review manipulated through the reviewer contact details that authors enter in submission systems, authorship for sale through paper mills, and swapping in a different manuscript after acceptance. The European Code of Conduct lists cartels of reviewers and authors who review each other's work, and deliberately using paper mills, as unacceptable practices.

COPE guidance lets editors handle suspected paper mill cases collectively and base decisions on all the evidence, including circumstantial evidence, and its retraction guidelines cover batch retractions. Some publishers also use shared screening services, such as the STM Integrity Hub, to look for patterns that indicate paper mills.

For authors, the rules are simple:

• Never buy, sell or trade an authorship position, and treat an offer of one as a warning sign.
• Suggest reviewers only when the journal asks, and only independent experts with verifiable institutional addresses.
• Keep your own manuscript files, so you can show what you submitted at each stage.

What counts as research misconduct

The definitions share a core but differ at the edges, and the difference decides what an institution can find.

A finding under the US federal definition has three further conditions. There must be a significant departure from accepted practices of the relevant research community, the misconduct must be committed intentionally, knowingly or recklessly, and the allegation must be proven by a preponderance of the evidence.

Below misconduct sits a wider band that the UK Concordat calls questionable research practices: avoidable errors and practices that fall short of intentional misconduct, arising from lack of knowledge, negligence or deliberate action.

Two practical consequences follow. Whether a case is misconduct depends on the policy that applies, so institutions should state their definition in writing. And a journal does not need a misconduct finding to act: COPE's guidelines let an editor retract a paper when they no longer have confidence in its findings, whether the cause was honest error or misconduct.

Who investigates: institutions, journals and regulators

Journals and institutions have different powers: a journal cannot investigate an institution's staff, and an institution cannot correct a journal's record. COPE's guidance on cooperation between research institutions and journals sets out the division:

• Institutions investigate. Allegations of research misconduct should be investigated by institutions, which should publish a contact for research integrity concerns and tell journals about findings that affect the reliability or attribution of published work.
• Journals raise concerns and correct the record. Journals should tell institutions when they suspect misconduct, provide evidence, cooperate with investigations and publish corrections or retractions based on the findings.
• Confidentiality limits what can be shared. An institution may not be able to share details until its investigation ends, and a journal may publish an expression of concern in the meantime.

The ICMJE gives editors the same sequence, adding that they should consider informing funders. National bodies add a layer in some countries: in the United States, institutions with Public Health Service support handle allegations through assessment, inquiry and investigation, and the Office of Research Integrity reviews their findings.

If a journal contacts you about a concern, answer within its deadline, send original data rather than a summary, and tell your research integrity office. If you are raising a concern, put it in writing with specific identifiers, such as the DOI and figure number. The steps an office should follow from intake to closure are in a practical checklist for research integrity offices.

Correcting the record after publication

Journals have three main tools, and the choice follows how far the paper can still be relied on.

COPE's retraction guidelines state that the purpose of retraction is to correct the literature and ensure its integrity, not to punish the authors, and that the editor makes the final decision. The ICMJE adds an option for honest error: when a misclassification or miscalculation changes the results but the underlying science is sound, a journal can retract the paper and, after further review, republish the corrected version with an explanation.

Authors have a duty on the citing side too. The ICMJE Recommendations make authors responsible for checking that no reference cites a retracted article except when discussing the retraction. For MEDLINE journals, search PubMed with "Retracted publication [pt]". The Retraction Watch Database, which Crossref acquired in 2023, is another source.

If you find an error in your own paper, report it to the journal in writing rather than waiting to be asked. Our guide to when and how to correct the published record walks through that letter.

Consequences for the authors involved

Consequences depend on the finding, the institution, the funder and the country, but some are predictable.

The record. Retraction notices are permanent and linked to the article. The ICMJE Recommendations also say that the validity of earlier work by the author of a fraudulent paper cannot be assumed. Editors may ask the author's institution to confirm the validity of other papers in their journal, retract them, or publish a notice that their validity is uncertain.

Employment. Institutions apply their own disciplinary procedures, so outcomes depend on local policy and law.

Funding. For US Public Health Service research, HHS can impose administrative actions including letters of reprimand, correction or retraction of the research record, supervision requirements on grants, special review of the person's funding requests, and exclusion from Public Health Service advisory roles. It can also seek to recover funds and separately pursue suspension or debarment. The Office of Research Integrity publishes summaries of cases in which administrative actions were imposed.

Co-authors. Every author agrees to be accountable for the work under the ICMJE criteria, so co-authors who did nothing wrong still share a retracted paper. Question an unexpected result, and ask to see raw data, before you approve submission.

What institutions should put in place

Standing arrangements prevent many integrity problems and make the rest easier to resolve:

1. A written definition of research misconduct, matched to national rules and funder terms
2. A research integrity contact published prominently on the institution's website, as COPE guidance recommends
3. Template letters for telling editors about findings that affect published work
4. Data retention rules covering raw images and analysis code, for at least the 10 years the ICMJE expects
5. An authorship agreement based on the ICMJE criteria, completed when a project starts
6. A check that clinical trials are registered before the first participant is enrolled
7. Guidance on disclosure and on AI use that follows the current ICMJE Recommendations
8. Image processing rules for every lab that produces figures
9. Protection for people who raise concerns in good faith

In the UK, organisations following the Concordat must also publish an annual statement that includes the number and types of misconduct allegations and investigations.

Directive Publications describes its peer review process, including the checks a submission passes through, and brings its author guidance and publication ethics policies together on its For Authors page, so authors and research offices can compare them with the standards in this guide before submitting.