Adapting FDA Regulation for AI and ML in Medical Devices

The FDA is exploring how to regulate medical devices that incorporate artificial intelligence (AI) and machine learning (ML), acknowledging that their traditional regulatory framework is not fully equipped for the adaptive nature of these technologies. Traditionally, the FDA regulates medical devices through pathways like premarket clearance (510(k)), De Novo classification, or premarket approval, and also reviews significant modifications to existing devices. However, AI and ML-driven software changes often necessitate a premarket review due to their dynamic nature.

On April 2, 2019, the FDA published a discussion paper proposing a regulatory framework for modifications to AI/ML-based Software as a Medical Device (SaMD). This framework suggests a “predetermined change control plan” for premarket submissions, which includes detailed plans for anticipated modifications (Software as a Medical Device Pre-Specifications) and the methodologies for implementing these changes safely (Algorithm Change Protocol).

The proposed framework aims to balance the rapid evolution of AI and ML in medical devices with patient safety. It suggests that manufacturers commit to transparency and real-world performance monitoring for AI/ML-based SaMD, and provide periodic updates to the FDA about the changes implemented under the approved pre-specifications and algorithm change protocol. This approach would allow the FDA to oversee the entire lifecycle of a software product, from premarket development to postmarket performance, leveraging the continuous improvement potential of AI and ML while ensuring patient safety. Additionally, as part of its AI/ML Action Plan, the FDA intends to further develop this regulatory framework, including issuing draft guidance on the predetermined change control plan to refine and adapt its regulatory oversight to the unique challenges posed by AI and ML technologies in medical devices.