ClinicalTrials.gov data-sharing plans: what an IPD statement really means

01

Separate a sharing plan from shared data

ClinicalTrials.gov lets a responsible party state Yes, No, or Undecided for a plan to share individual participant data and related data dictionaries. This field concerns participant-level observations, not merely the aggregate tables posted in the Results tab. A paper, press release, summary results record, protocol, analysis code, and IPD data set each answer different reproducibility questions.

The word plan is critical. A Yes recorded before a trial finishes can describe a future process. Verify whether the stated start date has arrived, whether the repository or request URL works, whether the study identifier matches, and whether the promised materials are actually listed. Do not report data available when the only evidence is a future intention, and do not report no sharing when the registry says Undecided.

02

Read every component of the IPD statement

ClinicalTrials.gov provides fields for the description of what will be shared, supporting-information types, time frame, access criteria, and a URL. Supporting information can include the protocol, SAP, informed consent form, clinical study report, and analytic code. A useful statement identifies whether all collected IPD or only the data underlying a publication will be included and connects the request to a stable study or repository record.

The time frame should say when access begins and how long it lasts. Access criteria should identify who may request data, for which analyses, who reviews requests, which agreement applies, and how access occurs. A generic available on reasonable request statement can be a starting point, but it does not let a reader evaluate feasibility until the decision maker, process, timing, and conditions are clear.

03

Understand open, controlled, and restricted access

Participant-level clinical data can contain re-identification risk even after direct identifiers are removed, particularly for rare conditions, small populations, unusual events, images, geography, or linked data. NIH describes methods such as de-identification, data-use agreements, restricted enclaves, and controlled access. These protections can be appropriate and should not be portrayed as evidence that investigators are hiding an unfavorable result.

At the same time, controlled access is not self-explanatory. Review eligibility rules, proposal review, permitted purposes, security requirements, publication conditions, costs, timelines, and appeal or contact information. Requirements that are vague, impossible to satisfy, unrelated to participant protection, or inconsistent with the public statement deserve follow-up. This article cannot decide whether a particular restriction is legally or ethically necessary.

04

Compare the registry plan with the paper and repository

ICMJE says reports of clinical trials submitted to its member journals must include a data-sharing statement and trials beginning enrollment on or after January 1, 2019 must include a sharing plan in registration. Its statement elements closely track the registry fields: what data, which documents, timing, duration, access criteria, recipients, analyses, and mechanism. Journal policy is not the same as a universal legal mandate or a guarantee that a request will succeed.

Search the paper and supplement for Data Sharing, Data Availability, or Sharing Statement. Compare that language with the ClinicalTrials.gov plan and Record History. Then open the repository or sponsor portal and verify the NCT number, study title, data version, documentation, access route, and update date. If the plan changed, describe the change and timing rather than silently citing the most favorable version.

05

Test whether the available package can support reuse

A data file without a dictionary, protocol, SAP, variable definitions, coding rules, and information about analysis populations may be difficult to interpret. Conversely, a detailed protocol without participant data does not enable reanalysis. Record which pieces are available and which are promised. Check whether identifiers such as NCT number, DOI, repository accession, and publication match across the package.

Data access also does not validate the original or secondary analysis. A reanalysis can use different estimands, exclusions, missing-data assumptions, multiplicity rules, or models. Readers should distinguish reproduction of the published analysis from a new analysis that asks another question. Shared data can improve scrutiny and discovery, but disagreement between analyses requires methodological evaluation rather than choosing the preferred conclusion.

• →Plan status and update date
• →Data scope
• →Data dictionary
• →Supporting documents
• →Availability window
• →Access criteria and reviewer
• →Repository or request URL
• →NCT and publication match
• →Actual request outcome if known

06

What data-sharing claims do not establish

Open data does not make a peptide effective, safe, approved, or appropriate. No-sharing does not by itself invalidate a study, and controlled access does not prove suppression. Privacy, consent, contracts, intellectual property, repository capacity, ongoing regulatory work, and participant expectations can affect access. Those factors should be described with evidence, not guessed from a checkbox.

For a consumer-facing claim, state the narrow fact: a plan says data will be shared, a repository lists a defined package, a request process exists, or data were obtained and analyzed by named researchers. Keep study limitations, conflicts, changes, and regulatory status visible. Treatment decisions belong with appropriately licensed clinicians using the full evidence and current product labeling, not with a registry transparency score.