11.2 Clinical-Trial Data-Sharing Statements and FAIR Principles

Clinical-Trial Data-Sharing Statements and FAIR Principles

Clinical-trial data sharing can support verification, new analyses, and participant value, but it must be planned and governed. “Share everything” and “share nothing” are not adequate default answers. The appropriate route depends on the data, consent, privacy risk, agreements, law, platform, and scientific purpose.

ICMJE statement requirement

Since July 1, 2018, manuscripts submitted to ICMJE journals that report clinical-trial results must contain a data-sharing statement. The statement must indicate:

1. whether individual deidentified participant data, including data dictionaries, will be shared;
2. which data will be shared;
3. whether related documents such as the protocol or statistical analysis plan will be available;
4. when data will become available and for how long;
5. access criteria, including with whom, for what types of analyses, and by what mechanism.

A statement may say that data will not be shared. “Undecided” is not acceptable. ICMJE does not require unrestricted public download or promise that every request will be approved.

For trials beginning enrollment on or after January 1, 2019, the registration must include a data-sharing plan under ICMJE policy. If that plan changes, the change should be reflected in the registry and in the statement submitted and published with the manuscript. A change is not necessarily forbidden, but it should not be concealed.

Separate data categories

Define the asset rather than saying “study data.” Possible materials include:

• participant-level analysis data;
• raw or source data, which may not be shareable in the proposed form;
• data dictionaries and metadata;
• statistical code;
• protocol, amendments, and statistical analysis plan;
• clinical study report or selected appendices;
• case-report forms;
• consent language relevant to secondary use.

“Deidentified” does not mean zero reidentification risk, especially for rare diseases, genomics, imaging, dates, free text, or linked datasets. Determine the applicable legal standard and use data minimization, disclosure-risk assessment, secure access, and contractual controls as appropriate.

Access models

Data may be supplied through an open repository, controlled repository, secure analysis environment, sponsor portal, data enclave, or direct agreement. Controlled access can require a research proposal, qualified investigators, ethics review where applicable, a data-use agreement, analysis in a secure workspace, publication commitments, and output disclosure review.

Governance differs among Vivli, the YODA Project, ClinicalStudyDataRequest.com, institutional repositories, and sponsor systems. Some use independent review; others use data-contributor review or different combinations. Do not claim that every named platform has one identical independent panel or approval test. Check the current platform policy.

A review committee may consider scientific purpose, feasibility, team qualifications, requested variables, participant risk, conflicts, consent, and security. It should apply published criteria consistently. A request should not be rejected merely because findings may be unfavorable, but access can be denied when it conflicts with participant protection, law, consent, or a legitimate governance rule.

FAIR principles

FAIR provides stewardship principles:

• Findable: persistent identifiers, rich metadata, indexing, and searchable registration.
• Accessible: a defined retrieval method using an appropriate protocol, with authentication and authorization where necessary; metadata may remain available even if data access ends.
• Interoperable: formal, accessible, shared, broadly applicable languages and vocabularies, with qualified links to other data.
• Reusable: rich descriptions, clear usage conditions, provenance, and relevant community standards.

FAIR is not synonymous with open. A controlled dataset can be FAIR when metadata are findable, the access route is clear, formats are interoperable, and reuse conditions are explicit. Conversely, a downloadable spreadsheet without metadata, provenance, or a license may be open but not very reusable.

Clinical standards such as CDISC SDTM or ADaM can support interoperability in appropriate contexts, but FAIR does not require those two standards for every biomedical dataset. Use domain-appropriate structures and vocabularies.

Align plans, consent, and publications

Plan data sharing before consent and contracting when possible. The protocol, consent materials, registry plan, privacy notice, data-processing arrangement, sponsor–investigator agreement, and publication statement should not contradict one another.

If older consent language is silent or restrictive, do not assume deidentification automatically authorizes every international reuse. Seek privacy, ethics, and legal advice and use an access model proportionate to risk. For new trials, explain foreseeable secondary research and sharing honestly without promising uses the team cannot govern.

Secondary users and credit

ICMJE expects secondary users to comply with the access terms, cite the dataset with a persistent identifier when available, acknowledge the data source, and explain how the new analysis differs from prior analyses. Those who generated and shared the data deserve substantial credit, but data access does not automatically confer authorship. Apply authorship criteria to actual collaboration.

A secondary analysis should be registered or otherwise made transparent when appropriate, use a prespecified plan, preserve provenance, and avoid presenting an exploratory finding as if it were the original trial's primary conclusion.

Operational workflow

For a manuscript:

1. identify the ICMJE and venue requirements;
2. verify the registered plan and any amendments;
3. inventory the exact data and documents;
4. confirm consent, privacy, contractual, and intellectual-property constraints;
5. choose an access mechanism and accountable owner;
6. state availability dates, duration, users, purposes, review criteria, and mechanism clearly;
7. ensure the manuscript, registry, repository, and request page agree;
8. retain decisions and update public records when the plan changes.