Research ethics and publication integrity: building a defensible submission

Before submitting a manuscript, ask whether another person could trace its important statements to the records that support them. An ethics sentence, an author list and a disclosure form each serve a different purpose. This guide helps a research team assemble those records, resolve inconsistencies and describe uncertainty honestly. The examples concern biomedical publishing; other disciplines and individual journals may apply different policies.

1. Build a record that someone else can check

Create a small submission evidence file with the manuscript version, responsible person, supporting record and unresolved question for each declaration. Keep sensitive originals in an approved institutional location; the working checklist can point to them without duplicating participant information. A file called “ethics final” is less useful than a record identifying the decision, the study it covers and the person who checked it.

Read the declaration against what actually happened. If recruitment dates, study sites or analyses changed, compare those changes with the relevant protocol and permissions. Ask the responsible institutional body about an unclear approval or exemption. Do not manufacture a reference number, assume that retrospective work is automatically exempt, or replace an unresolved issue with reassuring wording.

2. Separate study oversight from permission to publish identifiable material

ICMJE asks authors to document independent review of human research and protect participant privacy. Consent to take part in research and consent to publish identifiable details answer different questions. Removing a name or covering a face does not by itself settle whether someone can still be recognised. The institution’s requirements and the journal’s policy both need checking.

Use a two-column review: “What will the reader see?” and “What authorises that disclosure?” Include images, quotations, rare combinations of personal details and supplementary files. If the answer is uncertain, pause that material and seek an appropriate decision. Keep the review specific to the material; a generic declaration cannot establish that a particular image is suitable for publication.

Reference: ICMJE: protection of research participants

3. Confirm contribution and accountability with the whole team

For journals following ICMJE, authorship combines a substantial contribution, meaningful work on the manuscript, approval of the final version and accountability. Seniority, funding or language editing alone does not establish authorship. Contributors who meet the initial contribution criterion should have a fair opportunity to participate in writing, review and approval.

Before submission, circulate the actual version to everyone on the proposed author list. Record who checked which parts and how concerns were resolved. Agree the author order together. A contribution table is a useful conversation aid, but ticking boxes on someone’s behalf is not their approval. Refer unresolved authorship disputes through the appropriate institutional process.

Reference: ICMJE: authors and contributors

4. Check funding and competing interests as separate statements

A funding statement explains support for the work. A competing-interest disclosure covers relevant financial and non-financial relationships more broadly. A disclosed relationship does not automatically show that the results are biased; readers need enough information to assess its relevance. Use the journal’s requested form and reporting period, rather than inventing a universal time limit.

Ask each contributor to complete their own disclosure. Compare the manuscript statement with those individual responses and record the sponsor’s actual role. If one author reports consultancy work while the manuscript says “no relationships to disclose”, resolve the inconsistency before submitting. Do not delete the individual response to make the documents agree.

Reference: ICMJE: relationships, activities and competing interests

5. Keep AI assistance within a verifiable writing process

ICMJE requires disclosure of relevant AI assistance and keeps responsibility with human authors. AI tools cannot take authorship responsibility, and generated material is not a substitute for the primary source. Check the target journal’s current policy and verify the output yourself.

Keep a brief working log: tool used, purpose, material supplied, proposed changes and the person who checked them. For a rewritten paragraph, compare every factual claim with the original evidence. For a reference, open the actual publication and confirm that it supports the sentence. An attractive synthetic illustration must remain clearly conceptual; it cannot stand in for a study photograph or measured result. Use only material you are authorised to share with the chosen service.

Reference: ICMJE: AI use by authors

6. Work through an inconsistency before it reaches the journal

The following scenario is entirely fictional. It illustrates a review method, not an approved study, a real patient record or a finding about any institution. Assume a team has a documented decision covering its survey, a checked analysis file and a table copied into a manuscript. The analysis file contains 84 completed responses; the manuscript table says 86. One co-author reports paid consultancy, while the draft declaration says there are no relevant relationships.

On small screens, scroll the table horizontally, or focus it and use the arrow keys.

Fictional submission review: evidence, uncertainty and next action
Item	What is established in this example	Action before submission
Response count	The checked analysis file says 84; the manuscript says 86.	Trace the transcription, reconcile every affected statement and obtain analyst confirmation.
Competing interests	One author has reported consultancy; the combined statement omits it.	Clarify relevance using the journal’s form and correct the combined declaration.
Participant illustration	A proposed image has no documented publication permission in the evidence file.	Hold the image until permission or an appropriate alternative is established.
Final approval	The corrected manuscript has not yet been circulated.	Send the corrected version for the required author review and approval.

7. Keep a route open for correcting the record

If an error is found after publication, provide the editor with the affected passage, supporting evidence and the known extent of the problem. Preserve the original files and distinguish established findings from questions still under investigation. Avoid quietly replacing a published PDF.

ICMJE describes linked correction notices and transparent version histories; errors that invalidate results may require a different editorial response, including retraction. The appropriate response depends on what the evidence establishes. A team’s first report should help the editor assess the issue, without declaring in advance that it is either harmless or misconduct.

Reference: ICMJE: corrections and version control

Questions for a practical final check

For each declaration, what record supports it and who checked that record?
Which permissions cover the exact text, images and supplementary material to be published?
Have contributions, disclosures and final-version approvals been confirmed by the people concerned?
Which issues are still unresolved, who owns the next action, and what would close each issue?
If a problem is found later, can the team reconstruct the submitted evidence and contact the editor?

Examples illustrate a working method; they do not establish the readiness of a particular journal or manuscript. This is general guidance: check the current policies of the journal concerned. Source links reviewed 14 September 2026.