ClinicalTrials.gov

ClinicalTrials.gov is the U.S. National Library of Medicine's registry and results database of publicly and privately supported clinical studies, and Datadory delivers it to biotechnology teams as structured, analysis-ready rows. As of August 2026 the registry holds 599,549 studies, and each one is organized into named protocol modules: identification (registry ID, short and official titles, responsible organization), status, sponsors and collaborators, conditions, design, arms and interventions, outcome measures, eligibility, contacts and locations, oversight, description and references.

Two derived layers ride along on every record: MeSH condition-browse terms give each study a position in the disease hierarchy, and a has-results flag separates studies that posted result data from those carrying protocol documentation only. That combination makes this the canonical ledger of who studied what, on how many participants, at which sites, under which phase label - the reference point every drug-development dashboard quietly depends on. Get a sample of this dataset and inspect real records before committing pipeline time.

Three real records, straight from the corpus:

nctId      : NCT00787735
briefTitle : Improving Delivery of Mental Health Services for Patients With
             Drug Use and Other Behavioral and Emotional Problems
status     : COMPLETED
leadSponsor: Johns Hopkins University
phases     : NA
enrollment : 380
startDate  : 2004-01

nctId      : NCT00429442
briefTitle : Simvastatin as an add-on Treatment to Copaxone for the
             Treatment of Relapsing Multiple Sclerosis
status     : WITHDRAWN
leadSponsor: Anna Tsakiri
phases     : PHASE3
conditions : Relapsing Remitting Multiple Sclerosis

nctId      : NCT07641842
briefTitle : Shoulder RA Survey Study
status     : RECRUITING

Three rows, three different life-cycle positions: one behavioral study completed years ago with an actual enrollment of 380, one Phase 3 drug trial withdrawn before it started, one survey currently recruiting. That spread is the point - status, phase and enrollment arrive on the same row shape whether a study is two decades old or filed last week, so a single query covers the whole history of the registry rather than one snapshot of it.

Sixteen core fields, every one with a verified definition. The dictionary below is the working schema for a study row: identifiers and titles to anchor the record, status and dates to place it in time, sponsor and phase to classify it, enrollment to size it, conditions and MeSH terms to route it to a disease area, and the interventions, eligibility rules, outcome measures and site list that describe its execution.

Geography - global. Studies are registered from more than 200 countries, and each record's site list carries facility, city, state and country, with geo-point coordinates where provided.

Temporal - the registry began accepting records in 2000, and the corpus spans 1999 to present, so the early registration era and today's pipeline live in one table. Some start dates carry month precision only (the worked sample above starts 2004-01).

Granularity - one record per registered study, with protocol, arms and interventions, outcomes, site locations and results nested beneath that single key. The row shape stays constant across all 599,549 records, which is what lets a filter written against 2004 data run unchanged against 2026 data.

API, files, or your warehouse. Daily, weekly, or hourly.

• Competitive intelligence on sponsor pipelines - lead sponsor joined with phase labels and overall status turns 599,549 records into a per-company view of who is running what, where it sits between RECRUITING and COMPLETED, and where trials were WITHDRAWN or SUSPENDED. The competitive intel product teams use cases profile maps more of these workflows.
• Trial feasibility and site strategy - the site module lists facility, city, state and country per study, so site-density questions ('how many rheumatoid-arthritis studies ran in the last five years, and in which cities?') answer from one field rather than a geocoding project.
• Enrollment benchmarking - enrollment counts ship with a type of ACTUAL or ANTICIPATED, which is exactly the pair needed to compare what sponsors planned against what they recruited.
• Condition mapping at scale - derived MeSH terms arrive with identifier, term and ancestor tree, so free-text condition strings roll up the disease hierarchy without a terminology service.

Data scientists get a labeled, schema-stable corpus large enough to train on - enrollment prediction, success-factor models and trial-duration estimates all start from the same sixteen fields. Their data scientists use cases briefing goes deeper. Developers and platform builders get stable row shapes keyed on a permanent registry ID - a low-friction backend for clinical-trial search and monitoring products; see the developers builders use cases page. Competitive-intel product teams get the sponsor-by-phase-by-status cube out of the box. Start from the biotechnology data hub, then read vs ChEMBL for where a study registry beats a compound database.