1. Authoritative U.S. Federal Data Sources
- NIH / NLM ClinicalTrials.gov REST API v2 (
clinicaltrials.gov/api/v2/studies): Full projection of interventional Phase II and Phase III studies with registeredStartDate >= 2018-01-01(55,377 raw protocols; 54,890 primary-cohort protocols after excluding 487 future/placeholder post-2026 start dates). - U.S. FDA Drugs@FDA Bulk Archive (
fda.gov/media/89850/download): Relational submission and approval tables covering 5,848 original New Drug Applications (NDA) and Biologics License Applications (BLA), including 1,184 original approvals from 2018 through 2026 and 2,830 approvals linked directly to consolidated trial sponsors.
2. Hospital Campus Canonicalization, CRO Masking Isolation & M&A Rollups
Across 948,985 protocol facility location rows, Protocol Ledger separates 155,062 anonymized CRO placeholder slots (such as "Research Site" or numeric site codes used across 40 sponsors) from 793,923 named physical hospital slots (83.66% disclosure). Named facilities are resolved across 49,607 campus merges into 22,436 physical medical centers while keeping multi-state hospital systems (such as Mayo Clinic Rochester, Arizona, and Florida) geographically distinct. On the sponsor axis, 3,327 protocols across 156 acquired biotech subsidiaries are consolidated into 37 corporate parent groups.
| Verification Dimension | Verified Count / Value | Scientific & Epidemiological Integrity Guarantee |
|---|---|---|
| Future / Placeholder Start Dates Excluded | 487 | Prevents post-2026 placeholder dates from distorting historical start vintages |
| Far-Future Completion Horizons (>2040) | 56 | Quarantined from velocity medians so 2042 to 2050 placeholders never skew cycle times |
| Enrollment Outliers (>250,000 Subjects) | 3 | Population-scale registry outliers isolated from clinical trial sample medians |
| Anonymized CRO Site Masking Isolation | 155,062 slots | Zero synthetic hospital entities fabricated from masked commercial site codes |
| Individual Investigator Contact PII | 0 | Personal investigator phone numbers, emails, and individual contact names stripped |
3. Automated Publication Quality Gate Verification
Every Protocol Ledger build runs automated invariant checks against the reconciled DuckDB and SQLite publication artifact prior to publishing static and API views. All 7 release gates below must pass with zero silent fallbacks.
| Quality Gate Check | Actual Value | Expected Threshold | Status |
|---|---|---|---|
| Primary Cohort Protocols Minimum Threshold | 54,890 | >= 45,000 | PASS |
| California Protocols Present | 10,046 | >= 4,000 protocols with California sites | PASS |
| Undisclosed Sites Isolated from Named Entities | 0 | 0 | PASS |
| Mayo Campuses Distinct | 3 | 3 | PASS |
| California Flagship Sites Resolved | Verified (true) | Verified (true) | PASS |
| FDA NDA BLA Linked to Sponsors | 2,830 | >= 1,500 linked NDA/BLA approvals | PASS |
| Scene Constellation Populated | sites: 140, sponsors: 36, arcs: 203 | sites>=100, sponsors>=30, arcs>=120 | PASS |
4. Engineering Architecture & Related Documentation
Protocol Ledger is engineered by Meru AI as a deterministic public-data reference system. Additional notes on institutional campus reconciliation, cryptographic provenance digests, and release gate verification are documented in Cross-Agency Entity Resolution, Citation-Grade Public Data Architecture, Healthcare Price & Patent Cliff Crosswalk, and Meru AI Public Data Platforms.