Comment from RQM+

RQM+ submits these comments in response to the Food and Drug Administration's Request for Information on the AI-Enabled Optimization of Early-Phase Clinical Trials Pilot Program (Docket No. FDA-2026-N-4390). RQM+ is a medical technology-focused contract research organization with deep operational experience in cardiovascular device and drug development, spanning IDE-regulated pivotal trials, post-market clinical follow-up studies, and early-phase oncology and cardiac programs. Our comments are informed by direct experience managing the inefficiencies this pilot program is designed to address delayed safety signal detection, lagged go/no-go decision-making, and the absence of a real-time data infrastructure capable of supporting continuous regulatory oversight. Our full written response addresses all question categories across Sections A and B of the RFI. We offer the following summary of our principal recommendations. Pilot design and AI use cases. We recommend prioritizing three near-term AI use cases: real-time safety signal detection and adverse event characterization, Bayesian adaptive dose escalation modeling, and biomarker-based patient stratification. Safety monitoring should be the top priority given the severity of consequences if AI underperforms, and the FDA's existing reporting framework provides a natural benchmark against which performance can be measured. Infrastructure prerequisites. Four foundational elements must be in place before launch: a CDISC-compliant data exchange standard, a secure FDA-maintained data environment with 21 CFR Part 11-compliant audit trails, a shared AI monitoring dashboard accessible to both sponsors and FDA, and critically, a data quality validation layer sitting between the EDC and the AI decision-support environment. Real-time data is pre-cleaning by definition, and the AI system should act on quality-validated signals, not raw unverified entries. Defining a minimum completeness and consistency threshold as a transmission prerequisite is non-negotiable, and maintaining it at the site level is a core CRO operational responsibility. Participant selection. We recommend a tiered AI maturity model for participant selection, assessing organizations across governance, validation, explainability, and infrastructure dimensions. Excluding lower-maturity organizations would be a strategic error: their participation generates the most valuable learning about how to bring less mature participants along, which is essential to the program's broader adoption goals. Evaluation metrics. Adverse event rates and protocol deviation rates represent two distinct hypotheses and should not be conflated. Protocol deviations are directly addressable by AI through real-time site-level flagging; a meaningful reduction against concurrent controls is a reasonable expectation. Adverse event rates are driven by the investigational product's biological or mechanical effects, not by trial conduct. AI's genuine contribution on the AE dimension is earlier detection and faster escalation to sponsor and FDA, not a reduction in the underlying event rate. The RTCT Learning Consortium. We strongly recommend establishing a formal RTCT Learning Consortium under FDA coordination, with participation as a condition of pilot selection. Pre-competitive sharing of infrastructure standards, AI validation frameworks, and operational learnings would materially reduce duplicated investment and accelerate adoption. When differential AI performance across demographic subgroups is observed, findings should be submitted to the Consortium as structured learnings (not failure investigations) contributing to the field's evidence base on equitable AI performance. Extension to CDRH and medical devices. This is the recommendation the response pool is most likely to underrepresent, and we urge the FDA to begin parallel planning now. The continuous data monitoring logic underpinning RTCT applies directly to pivotal IDE trials and post-market clinical follow-up obligations device sponsors already carry. EU MDR PMCF requirements have already created a regulatory environment in which real-time monitoring is expected. Cardiovascular device registries (including the STS database, the ACC NCDR, and the TVT Registry) offer existing structured data infrastructure that could support an RTCT extension track without waiting for the drug pilot to conclude. RQM+'s Cardiovascular Center of Excellence is prepared to support the FDA in designing this framework. RQM+ appreciates the FDA's leadership in advancing this pilot program. We believe AI-enabled continuous monitoring, implemented with appropriate rigor and governance, represents a meaningful step forward for trial efficiency, patient safety, and regulatory transparency. We welcome follow-up discussion on any aspect of these comments.