FDA Develops Framework for the Future of AI Regulation


Posted April 3, 2019

Getting Smarter

The March/April issue of BI&T, AAMI’s peer-reviewed and award-winning journal, features a cover story on artificial intelligence. It details how the burgeoning AI revolution is set to alter the development of medical devices and infuse the healthcare technology field with new tools for optimizing all manner of device management activities. Get the details at www.aami.org/bit.

Working with BSI, AAMI also has released a position paper on artificial intelligence and machine learning, looking at the role of standardization. The paper is complimentary and available to download.

The AAMI Podcast also covers artificial intelligence in episode 29.

And the upcoming International Conference on Medical Device Standards and Regulations will include a session on AI and machine learning.

Faced with the challenges and potential of artificial intelligence (AI) and machine learning (ML) in healthcare, the Food and Drug Administration (FDA) has released a framework that proposes a “reimagining” of the regulation of AI- and ML-based devices.

The proposed framework represents “the foundational first step to developing a total product lifecycle approach to regulating these algorithms that use real-world data to adapt and improve,” FDA Commissioner Scott Gottlieb said in a statement. It is intended to “allow the FDA’s regulatory oversight to embrace the iterative nature of these artificial intelligence products” while also ensuring that the agency’s standards for safety and effectiveness are maintained.

“The goal of the framework is to assure that ongoing algorithm changes follow prespecified performance objectives and change control plans, use a validation process that ensures improvements to the performance, safety, and effectiveness of the artificial intelligence software, and includes real-world monitoring of performance once the device is on the market to ensure safety and effectiveness are maintained,” Gottlieb explained.

Although the FDA published final guidance in 2017 describing when changes to software as a medical device (SaMD) should receive premarket clearance, that framework may not work well for AI-based software that can learn from experience and update or improve itself. Under current regulations, the FDA has only cleared “locked” algorithms that are modified by the manufacturer at intervals and don’t continually adapt or learn every time they are used. However, the FDA sees “a great deal of promise beyond locked algorithms that’s ripe for application in the healthcare space.”

The proposed framework, detailed in discussion paper published yesterday by the FDA, focuses on “adaptive” or “continuously learning” algorithms that don’t need manual modification and can learn from new user data acquired through real-world use. The framework outlines a total product lifecycle approach to regulation and is intended to clarify when a continuously learning AI/ML SaMD may require a premarket submission for an algorithm change.

In the discussion paper, the FDA outlines the need for a “predetermined change control plan” in premarket submissions, which would provide detailed information about anticipated modifications that may occur based on an algorithm’s retraining and update strategy, as well as how those changes will be implemented “in a controlled manner that manages risks to patients.” The FDA also expects “a commitment from manufacturers on transparency and real-world performance monitoring,” as well as periodic updates on the changes implemented as part of the approved algorithm change protocol.

In addition, software developers are expected “to have an established quality system that is geared towards developing, delivering, and maintaining high-quality products throughout the lifecycle that conforms to the agency’s standards and regulations,” according to Gottlieb.

The FDA is seeking feedback on its discussion paper until June 3 at www.regulations.gov (docket FDA-2019-N-1185). The agency says it intends to develop draft guidance based on the feedback it receives.

“Collaboration will be key to developing this appropriate framework. We encourage feedback and welcome a diversity of opinions and thoughtful discourse, which will contribute to building the foundation of this regulatory paradigm,” Gottlieb said.