Data Sharing, Clinical Trials and AI Use

Transparent data, prospective trial registration and accountable use of computational tools help readers verify pharmaceutical research and reuse it responsibly.

Every research article requires a data-availability statement

Data availability

Authors should preserve the materials needed to support their findings and make non-sensitive data, metadata, code and protocols available in a trusted repository whenever ethical, legal and contractual conditions permit. The manuscript must include one of the following types of statement:

  • data and code are openly available, with repository name, persistent identifier and access link;
  • data are available from the corresponding author on reasonable request, with a brief reason why public deposit was not used;
  • controlled access is available through a named data custodian or application process;
  • data cannot be shared because of participant consent, privacy, commercial, security or legal restrictions, explaining the restriction; or
  • no new data were generated or analysed, where genuinely applicable.

FAIR and responsible sharing

Where possible, shared material should be findable, accessible, interoperable and reusable. Use non-proprietary formats, a data dictionary, units, variable definitions, licence information and a citation to the associated article. Remove direct and indirect identifiers before sharing participant data; de-identification does not by itself override consent or law.

Data review and retention

Editors may request de-identified source data, laboratory records, code, protocols or statistical output during review or an integrity assessment. Inability to provide support for central findings may affect the decision. Authors should retain research records for the period required by their institution, funder, regulator and applicable law.

Clinical-trial registration and reporting

Prospective registration is required for interventional clinical trials. Give the registry, identifier, date and a link in the abstract and manuscript. Registration entries, protocols, statistical analysis plans and reported outcomes must be consistent; explain amendments and unreported or newly introduced outcomes. Randomised trials should follow CONSORT, protocols SPIRIT, and systematic reviews PRISMA, including registration or protocol details when available.

Artificial intelligence and automated tools

No AI authors

Generative-AI systems cannot approve a manuscript, hold copyright, disclose conflicts or accept responsibility and therefore cannot be listed as authors or cited as personal sources.

Human accountability

Authors remain responsible for accuracy, originality, citations, permissions, confidentiality and bias in every part of the submission, including material produced with an AI tool.

Meaningful disclosure

Disclose the tool name and version, provider, purpose and how output was checked when generative AI materially assisted writing, analysis, coding, translation, images or study methods.

Method validity

AI that forms part of the research method must be reported reproducibly, including model or system, inputs, validation, performance measures, human oversight and relevant limitations.

Permitted assistance and required disclosure

Routine spelling, grammar, reference formatting and non-generative software functions generally need no declaration. Material generative assistance should be described in the Methods when it affected the research, or in an acknowledgement or dedicated AI-use statement when it assisted preparation. Authors must verify every factual statement and reference against a reliable source. AI-generated or altered images must be labelled and must not be used to fabricate research observations.

Confidentiality for editors and reviewers

Do not upload confidential manuscripts to public AI systems

Editors and reviewers must not place submitted text, figures, data, review reports or author information into a public or externally retained AI tool without explicit journal authorisation and adequate confidentiality protection. Peer-review judgments must remain the reviewer's own.

Undisclosed or unsafe use

Fabricated citations, invented data, unacknowledged synthetic images, confidential-data exposure or misrepresentation of AI output may be treated as research-integrity concerns. The journal may request prompts, logs, code or validation records when necessary to assess a material claim.

Manuscript preparation Research ethics