Share of ClinicalTrials.gov first results submissions posted after quality control within 90 days, by submission month, February 2017 to June 2026: a data note
For the 61,327 studies whose sponsors first submitted results to ClinicalTrials.gov from February 2017 to June 2026 (52,751 posted after quality control (QC) review by 2026-11-19, 8,576 still pending), the share posted after QC within 90 days of the first submission (P90) was 36.7% for February to December 2017, 69.4% for 2020 and 56.1% for 2025. Monthly P90 peaked at 74.8% in January 2020 and stayed between 33.0% and 43.4% from July 2022 to July 2024; mean monthly P90 was then 47.3% in August to December 2024, 57.2% in January to March 2025 and 55.9% in April to December 2025, and P90 was 40.4% in June 2026. The Kaplan-Meier median time to posting ranged from 40 to 219 days by cohort quarter. The final step, from the submission that met QC criteria to posting, took over 30 days for 9.1% to 33.1% of studies per QC month from October 2021 to December 2024, for none of those posted by the download in QC months February 2025 to April 2026, and for 1.1%, 5.9%, 14.0% and 10.3% in May to August 2026, lower bounds as recent QC months are incomplete. The planned segmented regression gives a level change at April 2025 of +30.1 percentage points (95% CI 20.8 to 39.5) but does not follow this shape: residuals averaged +13.3 points in the second half of 2024, and a placebo break at April 2024 also gives a level change (+11.6, 3.8 to 19.3). The estimate is not read as a change at April 2025, the plan's null rule does not apply, and no cause is named. The plan was committed privately, so its timing cannot be verified. An AI model produced the plan, analysis and text, and three further AI sessions reviewed them; no human has checked the figures. The record holds the note (PDF and Markdown) and a supplement archive: the frozen plan and the feasibility files (internal references replaced, study identifiers removed), the deviations log, the analysis summary as written at the run (the pre-review record), the request record with the size and sha256 of every downloaded page (paging cursors removed), the stop-rule record, the planned estimates (outcomes.json), five monthly and quarterly aggregate tables derived from ClinicalTrials.gov, the after-outcome checks, the analysis code, the checks added for the note and after its reviews (paper_checks.py, paper_checks.json, review_checks.py, review_checks.json, small_cells.py), a manifest with the sha256 of every file and every change made for the deposit (it also lists where the data note supersedes the analysis summary), and a trace of every number in the note. Only aggregate tables and code are deposited: no raw API page, no unit-level row, no count cell of fewer than three studies and no NCT number. Source: National Library of Medicine, ClinicalTrials.gov, a United States Government database; data retrieved 2026-11-19. Plan, code, analysis and text by Claude (an AI model by Anthropic) for the author; reviewed before deposit by two further AI sessions of the same model family (a referee review and a blind outside-reader review) and a fix check by a third; the review by a second AI model that the plan names was not run; not peer reviewed; no human has checked the figures; the human author is responsible for the deposit.
Authors
- Keisuke Watanabe (ORCID: https://orcid.org/0000-0001-6374-3918)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24406039
- Primary Topic
- Research Data Management Practices
- Type
- article
- Field-Weighted Citation Impact
- 0.00