Cite this dataset in any publication or derivative work.
Not use the data for commercial purposes — including ML training data for
commercial models, redistribution as part of a commercial product/service,
incorporation into proprietary databases offered for sale, or quantitative
trading / commercial market analysis — without first obtaining a separate
commercial license.

For commercial licensing inquiries, contact bret.gaulin@gmail.com.

ClinicalTrials.gov Version History

A longitudinal, per-version snapshot of every protocol amendment ever submitted to ClinicalTrials.gov, normalized into a relational schema. Unlike most ClinicalTrials.gov mirrors that publish only the latest snapshot of each trial, this dataset preserves every version of every trial — making it suitable for event studies, change-tracking, and any analysis where the dynamics of protocol amendments matter.

What's in the `core` config

The core config is one row per (nct_id, nct_version) containing the scalar fields from each trial's protocol-section snapshot. 4,333,631 rows covering ~583K trials across their full version histories from 1999 to today.

from datasets import load_dataset

ds = load_dataset("brbk/clinical_trials_history", "core")
print(ds["train"][0])
# {'nct_id': 'NCT00000125', 'nct_version': 0, 'overall_status': 'ACTIVE_NOT_RECRUITING', ...}

The dataset is sourced directly from ct_norm.nct_core — 96 protocol-section scalar columns. The full column list is in the parquet schema, viewable via the HuggingFace Datasets Viewer or ds.features.

Key columns

Trial identification and version:

Column	Type	Description
`nct_id`	string	ClinicalTrials.gov identifier
`nct_version`	int32	Per-trial version number, aligned 1:1 with CT.gov's 0-based version index. Sorting by `(nct_id, nct_version)` walks each trial's history chronologically.
`ct_gov_version`	int32	Equals `nct_version`. Retained as a separate column for backward compatibility with consumers that switched on it during earlier schema versions.

Status and design:

Column	Type	Source
`overall_status`	string	`statusModule.overallStatus`
`why_stopped`	string	`statusModule.whyStopped`
`study_type`	string	`designModule.studyType`
`allocation`	string	`designModule.designInfo.allocation`
`intervention_model`	string	`designModule.designInfo.interventionModel`
`primary_purpose`	string	`designModule.designInfo.primaryPurpose`
`masking`	string	`designModule.designInfo.maskingInfo.masking`
`enrollment_count`	int32	`designModule.enrollmentInfo.count`
`enrollment_type`	string	`designModule.enrollmentInfo.type`
`has_results`	bool	TRUE if this version exposes a `resultsSection`

Sponsorship:

Column	Type	Source
`lead_sponsor_name`	string	`sponsorCollaboratorsModule.leadSponsor.name`
`lead_sponsor_class`	string	INDUSTRY / FED / NIH / OTHER_GOV / NETWORK / AMBIG / OTHER / INDIV / UNKNOWN
`responsible_party_type`	string	`sponsorCollaboratorsModule.responsibleParty.type`
`organization_name`	string	`identificationModule.organization.fullName`
`organization_class`	string	`identificationModule.organization.class`

Dates (read the Date Field Semantics section below):

Column	Type	Description
`start_date`	date	Trial start; `start_date_precision` indicates day/month/year
`completion_date`	date	Overall completion
`primary_completion_date`	date	Primary endpoint completion
`study_first_submit_date`	date	First-ever submission to CT.gov (constant across versions)
`study_first_post_date`	date	First-ever public posting (constant across versions)
`last_update_submit_date`	date	Date THIS version was submitted by registrant
`last_update_post_date`	date	Date THIS version was posted publicly. Use this for event-study / public-visibility timing.
`last_update_post_date_type`	string	`ACTUAL` (precise) vs `ESTIMATED` (approximate, common pre-2018)

The remaining ~70 columns cover titles, brief/detailed descriptions, eligibility criteria, age ranges, design-info detail, results-disclosure dates, IPD-sharing fields, and other protocol-section scalars. See ds.features for the full list with parquet-native types.

How versions work

Every time a registrant amends a trial's protocol or status on ClinicalTrials.gov, CT.gov assigns a new version. This dataset captures each chronological version as its own row keyed on (nct_id, nct_version). The nct_version index is 0-based and aligns directly with CT.gov's own version numbering — nct_version=0 is the trial's first submission, and the highest nct_version per trial is the latest amendment we've captured.

For each trial, sorting by nct_version produces the chronological version sequence. No special filtering is needed.

Cumulative-history payload (history.changes[], history.originalData.*, etc.) is not in core — it lives in a separate per-trial config (history_metadata) keyed on nct_id. See Planned future configs for the shape; the data is in our normalized pipeline but not yet published to HuggingFace.

Date Field Semantics

This is the most important thing to know about the dataset. Verified directly against the CT.gov API on 2026-05-07 against NCT00000125:

last_update_post_date is per-version, not stale. Each version's row carries the post date for that version, not a value carried forward from a prior version.
last_update_submit_date ≠ last_update_post_date. Submit is when the registrant uploaded the change; post is when CT.gov published it. The gap ranges from 1 day (auto-estimated) to ~30 days (review queue). For event studies tied to market reactions, last_update_post_date is what the public could see.
last_update_post_date_type matters. Older versions (pre-~2018) carry ESTIMATED post dates that are mechanical (often submit+1 day). Newer versions are ACTUAL. Treat ESTIMATED as approximate when timing precision matters.

Question	Field
When did the registrant submit this version?	`last_update_submit_date`
When did the public first see this version?	`last_update_post_date`
How precise is that post date?	`last_update_post_date_type`
When was the trial first registered?	`study_first_post_date`

Why this exists

Existing public ClinicalTrials.gov datasets — including AACT, the OpenFDA mirror, and HuggingFace's flat snapshot mirrors — publish only the current state of each trial. Anything that requires historical version data (event studies on protocol amendments, longitudinal modeling of trial trajectories, change-frequency analyses) requires reconstructing the full version history from CT.gov's undocumented internal API.

This dataset publishes the reconstructed history directly. Companion configs (planned, see below) will publish the JSON Patch operations between versions and the leaf-level scalar changes they describe.

Example queries

Find all trials whose status changed within the last 30 days

import polars as pl
from datasets import load_dataset

ds = load_dataset("brbk/clinical_trials_history", "core", split="train")
df = pl.from_arrow(ds.data.table)

changes = (
    df.sort(["nct_id", "nct_version"])
      .with_columns(prev_status=pl.col("overall_status").shift().over("nct_id"))
      .filter(
          pl.col("overall_status") != pl.col("prev_status"),
          pl.col("prev_status").is_not_null(),
          pl.col("last_update_post_date") >= pl.date(2026, 4, 8),
      )
      .select("nct_id", "nct_version", "prev_status", "overall_status",
              "last_update_post_date")
)

Completion-date pushes ("trial slipped")

slips = (
    df.sort(["nct_id", "nct_version"])
      .with_columns(prev_completion=pl.col("completion_date").shift().over("nct_id"))
      .filter(pl.col("completion_date") > pl.col("prev_completion"))
      .with_columns(slip_days=(pl.col("completion_date") - pl.col("prev_completion")).dt.total_days())
      .select("nct_id", "nct_version", "prev_completion", "completion_date",
              "slip_days", "last_update_post_date")
)

Update cadence

Refreshed weekly. Each refresh is committed to a Git tag (v2026.05.08, v2026.05.15, ...) so consumers can pin to a specific snapshot:

ds = load_dataset("brbk/clinical_trials_history", "core",
                  revision="v2026.05.15")

The most recent commit on main is always the latest snapshot.

Changelog

v2026.05.15 — schema simplification (`core` config)

Removed nct_version = -1 rows. These were a transitional artifact of an earlier "ended-trial snapshot" pattern. The cumulative-history data those rows carried (history.changes[], history.originalData.*, history.lastUpdateVersions.*, history.outcomesUpdateCount) is captured in our normalized pipeline on a per-trial basis and will be published as a separate history_metadata config — see Planned future configs.
Removed is_ended_snapshot column. It existed solely to discriminate the -1 rows from the per-version rows. With -1 rows gone, the column is meaningless.
Row count: 4,699,547 → 4,333,631 (-368,751 reflecting the removal of one -1 row per ended trial). The per-version data for each trial is unchanged; only the synthetic sentinel rows were dropped.
No filter change required. Code that previously used WHERE nct_version >= 0 to exclude the sentinel rows continues to work unchanged (it's now a no-op filter). Code that relied on nct_version = -1 rows directly will need to migrate to the history_metadata config when it ships.

v2026.05.08 — initial release

First public release. 4,699,547 rows across core config; per-version protocol snapshots plus -1 ended-trial sentinel rows. Pin to this revision for reproducibility of analyses written before the v2026.05.15 schema simplification.

Planned future configs

Same dataset URL, joined either on nct_id (per-trial configs) or (nct_id, nct_version) (per-version configs) — these will be added as additional configs without disrupting core:

Config	Grain	Source	Join key
`history_metadata`	One row per trial (cumulative metadata)	`ct_norm.nct_history_metadata` — `originalData.`, `lastUpdateVersions.`, `outcomesUpdateCount`. 583,902 rows. Covers every trial (active and ended).	`nct_id`
`version_history`	One row per CT.gov version-change event	`ct_norm.nct_version_history` — `history.changes[]` (`change_date`, `status`, `study_type`, `module_labels`). 4.3M rows.	`nct_id`, `ct_gov_version`
`history_original_outcomes`	One row per outcome arm, first-submitted state	`ct_norm.nct_history_original_{primary,secondary,other}_outcomes` — `originalData.{primary,secondary,other}Outcomes[]`	`nct_id`, `ord`
`version_patches`	One row per JSON Patch op per version transition	`ct_norm.nct_version_patches` (RFC 6902 ops)	`nct_id`, `from_version`, `to_version`, `op`, `path`
`patch_scalar_values`	One row per leaf scalar change	`ct_norm.nct_patch_scalar_values` (typed values: `value_str`/`value_num`/`value_bool`)	`nct_id`, `from_version`, `path`
`interventions`	One row per `(nct_id, nct_version, intervention)`	`ct_norm.nct_interventions`	`nct_id`, `nct_version`, `intervention_id`
`outcomes`	One row per primary/secondary outcome per version	`ct_norm.nct_primary_outcomes`, `nct_secondary_outcomes`	`nct_id`, `nct_version`, `ord`
`eligibility`	Per-version eligibility text + structured age/sex fields	`ct_norm.nct_core` (eligibility subset)	`nct_id`, `nct_version`
`locations`	Per-version trial site locations	`ct_norm.nct_locations`	`nct_id`, `nct_version`, `ord`
`sponsors`	Per-version lead sponsor + collaborator rows	`ct_norm.nct_collaborators`	`nct_id`, `nct_version`, `name`
`results`	Outcome measurements + adverse events for trials with results	`ct_norm.nct_results_*` family	varies by sub-table

The first three rows above (history_metadata, version_history, history_original_outcomes) are the per-trial cumulative-history configs introduced as a replacement for the -1 sentinel rows removed in v2026.05.15. The underlying tables are populated for every trial in our normalized pipeline (active and ended alike) — a coverage improvement of ~215K trials over the prior sentinel-only state.

The core config will not break — additive changes only (new columns may appear; no renames or type changes within the published schema). Breaking changes will be announced and gated behind a major version bump.

Source

Constructed from the ClinicalTrials.gov internal History API (/api/int/studies/{NCT}?history=true and /api/int/studies/{NCT}/history/{version}). Note that this is an undocumented internal endpoint; the public API v2 (/api/v2/studies/{NCT}?history=true) does not return version history.

License

This derivative dataset is published under CC-BY-NC-4.0 (Creative Commons Attribution-NonCommercial 4.0).

You may — for non-commercial purposes — use, copy, redistribute, and build derivative works from this dataset, with attribution.

You may not — without obtaining a separate commercial license — use this dataset, or any derivative of it, for purposes "primarily intended for or directed toward commercial advantage or monetary compensation" (CC-BY-NC's own language). This includes: redistributing it as part of a commercial product or service, using it as training data for commercial machine-learning models, incorporating it into proprietary databases offered for sale, or using it for quantitative trading or commercial market analysis.

For commercial licensing inquiries, contact bret.gaulin@gmail.com.

About the underlying data

The underlying ClinicalTrials.gov data is a U.S. Government work in the public domain — the license here covers the derivative work: the normalization pipeline, version-tracking schema, JSON Patch decomposition, and documentation. Anyone is free to re-derive equivalent data from ClinicalTrials.gov directly and license it however they wish; the NC restriction applies to this particular derivative, not to the raw facts.

Citation

If you use this dataset in research, please cite:

@misc{clinicaltrials_history_2026,
  title  = {ClinicalTrials.gov Version History: A longitudinal mirror},
  author = {Bret Gaulin},
  year   = {2026},
  url    = {https://huggingface.co/datasets/brbk/clinical_trials_history}
}

Acknowledgments

ClinicalTrials.gov is operated by the National Library of Medicine at the National Institutes of Health. Schema design references the AACT (Aggregate Analysis of ClinicalTrials.gov) project from the Clinical Trials Transformation Initiative.

Downloads last month: 74

Total file size:

981 MB

You need to agree to share your contact information to access this dataset