Comment from Columbia University Tulane University and NouStarX

Dear Commissioner and Colleagues, We commend the U.S. Food and Drug Administration (FDA) for launching the AI-Enabled Optimization of Early-Phase Clinical Trials Pilot Program. We believe this initiative represents one of the most important advances in clinical research since the widespread adoption of randomized controlled trials (RCTs) and, more recently, real-world evidence (RWE). Rather than simply introducing another digital technology into clinical development, this initiative has the potential to fundamentally reshape how scientific evidence is generated, continuously evaluated, and translated into better patient care. We submit this response from the perspective of three investigators whose combined experience spans more than five decades in electronic health records (EHRs), clinical phenotyping, interoperability, biomedical informatics, and real-world evidence generation, together with extensive experience implementing artificial intelligence (AI) solutions across pharmaceutical research and clinical development. Our careers have bridged academia, healthcare delivery, technology development, and direct pharmaceutical industry trial experience, giving us first-hand insight into both the scientific foundations and the operational realities of AI-enabled evidence generation and sponsor-run trials. Our central message is straightforward: Real-time clinical trials are fundamentally an evidence engineering challenge rather than solely an artificial intelligence challenge. AI will not make clinical trials real-time unless the underlying evidence ecosystem becomes computable, interoperable, continuously validated, and governed through transparent human oversight. AI alone cannot produce trustworthy clinical evidence. The quality, reproducibility, and regulatory acceptability of AI-generated insights depend upon the integrity of the underlying evidence ecosystem including interoperable EHRs, validated computable phenotypes, transparent data provenance, robust causal inference, continuous quality assurance, and meaningful human oversight. We therefore respectfully propose that the FDA view Real-Time Clinical Trials (RTCT) as the next evolutionary stage of the Learning Health System, where routine clinical care and clinical research continuously inform one another through a trustworthy Computational Evidence Infrastructure. The Agency’s recent real-time trial proofs-of-concept with AstraZeneca and Amgen demonstrate that this vision is not only possible but already underway; our recommendations are intended to help the pilot scale these early successes rigorously and equitably. Our recommendations are organized around six scientific principles and are intended to complement the Agency’s pilot program while establishing a long-term framework for AI-enabled regulatory science. We appreciate the opportunity to contribute to this important initiative. Respectfully submitted, Chunhua Weng, PhD, FACMI, FIAHSI · Xiaoyan Wang, PhD, FAMIA · Mei Yang, PhD Chunhua Weng, PhD, FACMI, FIAHSI Professor, Department of Medical Informatics, Columbia University Irving Medical Center Electronic Medical Records and Genomics (eMERGE) Network Coordinator AMIA Innovation in Informatics Lead Xiaoyan Wang, PhD, FAMIA Professor, Department of Health Policy and Management, Tulane University School of Public Health & CSO, NouStarX LLC. ISPOR Generative AI Working Group AMIA NLP working Group OMOP/OHDSI Rare Disease Working Group lead Mei Yang, PhD Chief Executive Officer, NouStarX LLC. Formerly of Merck, Daiichi Sankyo, and AbbVie