FDA's stated goal: view safety signals and endpoints in real time
The U.S. Food and Drug Administration announced it will pilot the use of artificial intelligence (AI) and data science to enable "real-time" clinical trials, with the explicit aim of accelerating development of drug therapies that the agency says are slowed by data and procedural bottlenecks. In his statement, the commissioner said the agency is "boldly advancing a modern approach whereby FDA scientists can view safety signals and endpoints in real time as a trial progresses," and framed the effort as a step toward "running real-time, continuous trials across all phases of drug development."
AstraZeneca's Traverse and Amgen's Stream‑SCLC proofs-of-concept
The agency said two proof-of-concept trials have already seen "successful initiation." AstraZeneca is running a phase 2, multi-site trial called Traverse that enrolls patients with mantle cell lymphoma who have not yet received treatment; the trial involves the University of Texas MD Anderson Cancer Center and the University of Pennsylvania. Amgen is conducting a Phase 1b trial called Stream‑SCLC with patients who have limited-stage small cell lung carcinoma, and the FDA said final site selection for that trial is in process.
The RFI: questions, participation, and schedule
To expand beyond the two pilots, the FDA published a request for information (RFI) to gather public comment on how AI-enabled technologies "can improve efficiency, speed and quality of decision-making in early phase clinical trials." The RFI, which the agency opened for comment, asks for input on potential pilot design and implementation, evaluation metrics, success criteria, the types of trials that might benefit most, the infrastructure needed, and how a pilot could accommodate participants with varying levels of AI capacity. The RFI is open for public comment until May 29. The FDA said it intends to publish final selection criteria for a wider pilot program in July and complete pilot selections in August.
What the FDA disclosed — and what it did not
The agency confirmed it met with the sponsors of both proof-of-concept trials to establish criteria for reporting key data signals such as efficacy outcomes and safety concerns. Beyond that, the FDA did not detail the specific AI or data science technologies used in the AstraZeneca and Amgen projects. That omission leaves the announcement focused on procedural aims and timelines rather than technical architecture or vendor choices.
How trial sponsors, academic centers, and biotech firms are positioned
- Trial sponsors (AstraZeneca, Amgen): The two companies are already engaged in pilots the FDA describes as "successfully initiated," and have worked with the agency on criteria for reporting efficacy and safety signals.
- Academic medical centers (MD Anderson, University of Pennsylvania): These centers are named participants in the Traverse trial, placing them at the operational front line of the FDA's experiment with near‑real‑time data reporting.
- Biotech firms and AI vendors: Industry voices framed the pilot as promising. Jim Foote, founder and CEO of First Ascent Biomedical — a biotech firm using AI-enabled tools for advanced oncology treatment discovery — said, "AI can compress timelines by automating tasks the agency has historically done manually: adverse-event signal detection, protocol-deviation monitoring, document review and pattern recognition across the millions of pages submitted."
The FDA framed the pilot as corrective: most clinical development is still run in discrete, phase-by-phase studies, and the agency says the "hiatus" between phases can slow product development. By soliciting public input and setting a compact calendar (RFI comments due May 29; selection criteria in July; pilot selections in August), the agency has established a clear near-term roadmap for widening its experiment.
The announcement leaves several operational details unspecified — most notably the specific AI systems or data pipelines that will be used — but it sets a definitive timetable and names concrete trials and institutions already participating. The next knot in the story is procedural: what the public comments recommend and which trial designs and infrastructures the FDA selects this summer.
