Accelerating Adoption of Digital Health Technologies by Use of Real-World Evidence and Predetermined Change Control Plans

At Cardiovascular Research Technologies (CRT), held in Washington, DC from March 7 to 10, 2026, Matthew Beckwith, Engineer and Senior Lead Reviewer at FDA CDRH, Office of Cardiovascular Devices, outlined updated guidance on real-world evidence (RWE) and introduced Predetermined Change Control Plans (PCCPs) for artificial intelligence (AI)-enabled medical devices. The central message: regulators are expanding pathways for real-world data use while establishing clearer expectations for adaptive AI systems.

The updated FDA RWE guidance, released nearly a decade after its prior version, aims to close the gap between pre-specified investigational device exemption (IDE) studies and retrospective submissions of medical data. The focus, said Beckwith, is on “expanding the use of real-world data (RWD) to support medical device claims.”

For AI and machine learning (AI/ML) applications, Beckwith emphasized 2 pillars: relevance and reliability. Relevance includes data availability, clinical outcomes, longitudinal follow-up, linkage of data sources, timeliness, and generalizability to broader populations. Reliability addresses data accrual methods, quality, integrity, and cybersecurity protections.

Examples of RWE in the guidance include expanded indications for a LASIK device using retrospective database analysis and the Rothman Index, which leverages electronic medical record data to predict clinical deterioration across inpatient and ambulatory settings.

The newly released PCCP guidance for AI-enabled devices allows manufacturers to predefine anticipated post-approval software modifications and associated testing methods (eg, sensitivity metrics). If agreed upon, these updates can be implemented without additional FDA submissions, enabling continuous algorithm improvement.

A separate draft AI/ML lifecycle guidance document—which is open for public comment—establishes baseline expectations for regulatory submissions and ongoing management.

According to Beckwith, AI systems must demonstrate “valid scientific evidence,” particularly regarding dataset integrity and cybersecurity safeguards. The concept of “junk in, junk out” was highlighted as a central regulatory concern in algorithm development.

The PCCP framework was described as a “powerful tool” enabling iterative AI model refinement while preserving oversight through predefined testing and potential auditing.

For developers of AI-enabled medical devices, the updated RWE and PCCP guidance signal increased regulatory flexibility coupled with heightened expectations for data quality, transparency, and lifecycle management. These policies may accelerate innovation while standardizing review criteria.

The FDA’s updated Real-World Evidence guidance and new PCCP framework represent significant steps toward modernizing oversight of AI in healthcare. Ongoing public engagement and implementation will shape how adaptive medical device algorithms evolve within regulatory boundaries.