CDISC's Expanding Perimeter Is Outrunning Its Implementation Ecosystem
SDTM was built for interventional trials; FDA is now formalising its use for observational data, while Dataset-JSON v1.1 awaits regulatory acceptance it doesn't yet have.
- Data standards & interoperability
- Stat Programming
- Regulatory
CDISC SDTM was designed for interventional trials with protocol-defined visits, controlled data collection, and predictable domain structures. None of that describes an EHR, a claims database, or a patient registry. A peer-reviewed article in JSCDM puts the structural problem plainly: heterogeneous data collection, absent visit frameworks, non-standard variable proliferation, and domain models that strain visibly when fitted to real-world sources. CDISC published Considerations for SDTM in Observational Studies in 2024 — the first formal guidance acknowledging the mismatch — but it translated the problem into official language more than it resolved it. The JSCDM paper’s value is operational: domain selection strategies, non-standard variable handling, and Trial Summary domain adaptations for observational contexts. Regulatory pressure is real and directional; according to FDA’s CBER-CDER Data Standards Program documentation, SDTM for Observational Studies v1.0 is among the agency’s tracked standards, signalling intent to formalise compliance expectations rather than merely encourage them. Teams building RWD pipelines for regulatory submissions would be unwise to treat current flexibility as permanent.
The foundational format layer is also moving. Dataset-JSON v1.1, API v1.0, and a Compressed variant represent the clearest step yet toward retiring SAS XPT. The key caveat: FDA’s Study Data Technical Conformance Guide still names SAS XPT as the required submission format, and no confirmed regulatory acceptance timeline or submission pilot for Dataset-JSON has been published. The plumbing is ready before the regulator has signed off on the connection. Engage with the format now; do not retire XPT workflows on the basis of this release alone.
On the infrastructure side, OpenStudyBuilder (Novo Nordisk / CDISC COSA) and TransCelerate’s Digital Data Flow are both betting on linked-metadata and machine-readable protocols, respectively, to close the automation gap — conceptually sound, but neither offers external adoption benchmarks yet. Three items also remain in public review, with DHT1 Draft v1.0 particularly worth the attention of any team ingesting wearable data into trials. Comment windows close; missing them is a choice too.
Protocol read: The CDISC stack is moving simultaneously at every layer — observational extensions, format modernization, machine-readable metadata — but the regulator’s acceptance timeline lags the standards body’s by enough that retiring old workflows now would be premature. Engage with the new layers; do not switch off the old ones.
What to do now:
- Pilot Dataset-JSON v1.1 internally but keep SAS XPT validation workflows current until FDA publishes a confirmed acceptance timeline.
- Build observational SDTM compliance against the JSCDM operational guidance now; FDA’s tracking of SDTM for Observational Studies v1.0 signals formalization is coming, not optional.
- File comments on DHT1 Draft v1.0 if your team ingests wearable data — the public-review window is the leverage point, and it closes.