Where this data comes from

What we hold today

Studies

17,264

Recruiting

731

With posted results

605

Last synced

2 hours ago

Registries

There are 18 WHO Primary Registries plus ClinicalTrials.gov as the dominant data provider. The table below states plainly which of them we currently ingest — the rest are listed because they exist and are on the roadmap, not because we hold their records.

The Netherlands Trial Register (NTR) is not listed: it is no longer a WHO primary registry. Historical NTR records reachable through the WHO ICTRP are treated as an archived source.

Known gaps

• One registry so far. This deployment ingests ClinicalTrials.gov only. Records registered exclusively in a national registry — a study registered only in ChiCTR, DRKS or CTRI, for example — are not here yet. WHO ICTRP bulk ingestion is the route to that breadth and is not yet enabled.
• Cross-registry deduplication is not running. A study registered in several places would currently appear once per registry. With a single source ingested this has no present effect, but it means the “registered in N registries” panel on a trial page reflects only what ClinicalTrials.gov itself records.
• Studies that ended long ago are excluded by design. We skip a study whose completion date has passed unless we were already tracking it, or unless it ended within the last 540 days without posting results. Studies with no completion date recorded are always included. So these counts describe what we hold, not what the registries hold.
• Documents are linked, not stored. We record the registry link for every protocol, statistical analysis plan and consent form a sponsor posted, and send you to the registry to read it. We do not host copies.
• Eligibility criteria are not yet parsed into structured statements. Trial pages show the sponsor’s criteria text verbatim, which is the authoritative version. The structured breakdown requires the enrichment layer, which is built but not yet enabled.
• Registry records go stale. A study can read “recruiting” for years after it stopped, because nobody updated it. We show the date the sponsor last confirmed the record on every trial page, and flag records unconfirmed for more than two years. We cannot fix the underlying data — only make its age visible.

How syncing works

We read the ClinicalTrials.gov API v2 on a schedule, requesting only records whose last-update date is newer than our watermark. Each record’s raw payload is stored before it is normalised, so a mapping error can be replayed without re-fetching. Unchanged records are detected by hash and skipped. When a tracked field changes we write a snapshot and derive a human-readable event from it, which is what the status history on each trial page is built from.

Recent ingestion runs

How studies are grouped into areas

Therapeutic areas are derived from the MeSH terms ClinicalTrials.gov assigns to each study’s conditions, by matching the MeSH descriptor chain against the root descriptor of each area. The assignment is reproducible from the record rather than curated per study. A study may touch several areas; exactly one is marked primary.

Counts on the homepage and the areas list use the primary assignment only, so they sum to the corpus total. Filtered browse uses every assignment, so a study appears under each area it touches. Both are correct, and they will not agree — that is expected, not a fault.