Real-Time Clinical Data Flow: Why the FDA’s New Pilot Validates the Data Flow Modernization Agenda
Date: April 30, 2026
Tamer Chowdhury, on behalf of ACDM Technology DMEG.
On April 28, FDA Commissioner Dr. Marty Makary unveiled an operational pilot in which FDA reviewers can see pre-agreed safety signals and clinical endpoints in near real time, streaming from investigator sites through cloud-based dashboards as trials run. The pilot is live today with MD Anderson, the University of Pennsylvania, Paradigm Health, and sponsors including Amgen and AstraZeneca. Dr. Makary noted that roughly 45% of clinical trial time is currently consumed by paperwork and administrative process — a number the pilot is designed to attack directly.
For ACDM members, the substantive question is not whether real-time review is coming, but what technology foundation makes it possible — and that is exactly the territory our DMEG Modernizing Clinical Data Flow: Leveraging CDISC 360i Standards white paper covers.
Read the paper here
From periodic SAS transports to API-based JSON exchange. The pilot’s signal-streaming model is incompatible with traditional one-time, end-of-study SAS v5 XPT submissions. It requires a data exchange substrate that is lightweight, machine-readable, and continuously updatable. Dataset-JSON — now a CDISC standard and accepted by FDA in pilot submissions — fits this need precisely. JSON is API-native, schema-rich, and supports incremental delivery in a way that SAS transport files never could.
Protocol as structured data: USDM. Real-time signals only have meaning against a known protocol context — endpoints, visits, eligibility, schedule of activities. The Unified Study Definitions Model (USDM) turns the protocol itself into a structured, machine-consumable artifact, so that what the FDA reviewer sees in a dashboard is anchored to the same digital protocol the sponsor and site are executing against. Without USDM, “real-time review” collapses back into manual reconciliation.
Semantics over mappings: Biomedical Concepts. The pilot’s pre-agreed safety and efficacy signals are, in CDISC terms, Biomedical Concepts (BCs) — concept-level definitions of what a measurement is, independent of any single CRF or SDTM domain. BCs (with their associated DECs and Dataset Specializations) provide the shared semantic layer that allows a signal defined once to be computed identically at the site, the sponsor, and the agency. This is the core architectural shift that 360i enables and that the DMEG paper argues for: stop mapping data downstream, start defining it upstream.
Signal-based reporting, not full-dataset dumps. Jeremy Walsh, FDA’s Chief AI Officer, was explicit in the webinar that the pilot is not about pushing every byte to the agency. It is about pre-defining reportable safety and efficacy signals and streaming those, in parallel with — not replacing — existing regulatory processes. This is signal-as-contract, and it presupposes the standards stack above.
What this means for data management. The skill set the pilot demands — protocol digitization, concept-driven specification, API-based exchange, computable submission artifacts — is a data management skill set. The role shifts from late-stage data cleaning toward upstream design of the digital protocol, the BCs, and the data contracts that govern flow from EDC and EHR to sponsor to agency. This is the future state our white paper sketches, and the FDA pilot is, in effect, a concrete reference implementation of where the industry is heading.
The full FDA session is available on YouTube; the Modernizing Clinical Data Flow white paper is available through the DMEG.
ACDM members interested in contributing to the next phase of DMEG’s work on real-time data architectures are warmly invited to get in touch: https://acdmglobal.org/technology-dmeg/
