Verdatic turns CDISC's published standard — or your own study artifacts — into realistic, coherent synthetic SDTM data with ready-to-run derivation code. Point it at a schedule and a whole Phase 2/3 study comes out: every structural domain, conformant and coherent. Sponsor data optional.
Most tools do a single slice: a data generator, a mapping helper, or a CDISC library. Verdatic chains them into one loop — and learns from every coverage gap, so it maps more next time.
A Rave ALS build or USDM protocol — or nothing but the CDISC standard.
Fields link to Biomedical Concepts & specializations, confidence-tiered.
Coherent data that obeys the study's own rules across fields, visits & subjects.
Runnable SDTM derivation in SAS, SQL or Python.
Unmatched fields become authoring targets — coverage grows with use.
The constraints come from real edit-checks and the published standard — not hardcoded clinical assumptions. That's what makes the data defensible.
Generated values stay consistent across fields, visits and subjects — because constraints are mined from the study's actual quality rules. Episodic events keep their identity across visits; con-meds couple to the subject's own history.
Patent #1A violation-feedback loop drives a synthetic dataset to pass the study's conformance rules — from 185% violations to zero — with no manual parameter-fiddling. Convergence becomes a headline trust metric.
Patent #2No sponsor study, no real data. Point Verdatic at the CDISC catalog and a schedule; one call yields DM, VS/LB/EG, MH/AE/CM and TV/TA — every SDTM structural class, coherent and conformant.
Sponsor removedA confidence-tiered matcher links source fields to the right Biomedical Concept and specialization — value-level-metadata and specimen aware. High-confidence auto-confirms; the rest are one click.
Confidence-scoredCompiles the mapping into real SDTM derivation code — SAS, SQL or Python — with method-position orchestration and library-function preludes. The authored logic emits to multiple targets.
SAS · SQL · PythonReal usage surfaces which fields can't auto-map; those become the next authoring targets. Coverage is the moat — and it grows automatically with every project you run.
Compounds with useHand Verdatic a USDM schedule and let the standard do the rest. ScaffoldStudyAsync produces every SDTM structural class — generated, longitudinally coherent, cross-domain coherent, and transposed to SDTM long.
One subject roster. Findings in range. Events MedDRA/WHODrug-coded. Concomitant meds coupled to each subject's own medical history. Conformance measured, not assumed.
══ BROAD PHASE 2/3 STUDY (standards-driven, sponsor removed) ══ 100 subjects (2 sites) · 5 visits (Screening D-14 … EOT D168) DM 100 (one per subject) VS 500 → 2000 SDTM long records LB 500 → 2500 SDTM long records EG 500 → 2500 SDTM long records MH 929 AE 486 CM 1536 (events) ≈ 10,051 SDTM records across the study CM coupled to subject's own MH: 1050/1536 (68%) ══ one roster · findings in-range · events coded · CT-valid ══ ────────────────────────────────────────────── CT validity ........................ 100% CORE conformance (SDTM long) ....... 100% (46 rules) Optimizer ........... violations 185% → 0%
Conformant synthetic data for development, validation, training and demonstration — so teams aren't blocked waiting on real, locked, or restricted study data.
Exercise mapping & derivation code against realistic, conformant data — including deliberate dirty-data to hit the right edit checks.
Stand up a representative study and its edit-check behavior without exposing a single real subject.
Coherent longitudinal trajectories and cross-domain coupling — distributions that behave, for method and pipeline testing.
A conformant test-bed to validate CORE rules, explore study designs, and pressure-test CDISC tooling at scale.
Verdatic is in pre-launch. Request early access and we'll show you a conformant study generated from nothing but a schedule.