The U.S. Food and Drug Administration has announced new guidance to support the development and marketing of safe and effective medical devices enhanced by artificial intelligence. This guidance includes recommendations for marketing submissions, documentation, and information needed throughout the total product life cycle for regulatory oversight of safety and efficacy.
Why It Matters
Last month, the FDA released the final predetermined change control plan guidance for AI and machine learning submissions. This guidance defines what is required to maintain AI/ML components and submit them for regulatory review without triggering a new marketing submission. The new guidance provides medical device developers with key recommendations for product design, development, and documentation for initial submissions.
The guidance, set to be published in the Federal Register on January 7, will be the first to offer total product life cycle recommendations for AI-enabled devices. It will tie together all design, development, maintenance, and documentation recommendations, providing a comprehensive approach for developers and innovators to follow throughout the device life cycle.
The FDA encourages early and ongoing engagement from developers to guide activities throughout the device life cycle, including planning, development, testing, and monitoring. With over 1,000 AI-enabled devices authorized through premarket pathways, the FDA has compiled requirements and shared learnings to serve as a valuable resource for developers.
The new guidance will address strategies for transparency and bias, offering advice on bias risk management and suggestions for thoughtful AI design and evaluation. The FDA will accept public comments on the draft guidelines through April 7, specifically seeking feedback on AI life cycle alignment, generative AI recommendations, performance monitoring, and information for AI medical device users.
The Larger Trend
In a blog co-written by FDA officials last year, life cycle management principles were emphasized as crucial for navigating the complexities and risks associated with AI software in healthcare. The continuous learning and adaptability of AI pose risks such as biases in data or algorithms, which could harm patients and disadvantage certain populations.
To address these risks, the FDA first established predetermined change control plans for AI/ML devices. This approach ensures that performance considerations, including race, ethnicity, disease severity, gender, age, and geographical factors, are addressed throughout the development, validation, implementation, and monitoring of AI/ML-enabled devices.
On the Record
Troy Tazbaz, director of the Digital Health Center of Excellence at the FDA, highlighted the importance of recognizing the unique considerations of AI-enabled devices. As exciting developments continue in this field, it is crucial to address specific risks and challenges associated with AI-enhanced medical devices.
In Conclusion
The FDA’s new guidance on AI-enabled medical devices underscores the agency’s commitment to ensuring the safety and efficacy of these innovative technologies. By providing comprehensive recommendations for developers and addressing transparency, bias, and performance monitoring, the FDA aims to support the responsible development and marketing of AI-enhanced medical devices. Stay tuned for further updates on this evolving regulatory landscape.
