FDA’s AI ML Action Plan will Boost More Investment in AI ML Productsby Priya Dialani February 21, 2021
Experts view that more investment in AI and ML can provide access to safe and effective medical devices
Last month the U.S. Food and Drug Administration introduced the association’s first Artificial Intelligence/Machine Learning (AI/ML)- Based Software as a Medical Device (SaMD) Action Plan. This ‘Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device Action Plan’ was delivered by the Digital Health Center of Excellence (DCE). The DCE introduced this on 22 September 2020 and exists within the FDA’s Center for Devices and Radiological Health. The DCE’s point is to push the FDA’s general commitment to the headway of digital health technology.
Dr. Adam Saltman Chief Medical Officer at Eko, having previously served as a medical officer for the FDA for 12 years, co-authored the FDA’s initial whitepaper on AI back in 2019 points out that there will be more companies to invest in AI and ML products.
“This action plan comes at the perfect time, as interest in adaptive learning AI is rapidly increasing, and all stakeholders are invested in maintaining the building momentum of innovation while assuring that patients and providers have access to safe and effective medical devices. This is an especially daunting task, as adaptive AI/ML represents a radically new kind of device: Something that may look very different from what it was at the time of its initial premarket evaluation, and something that will keep changing and improving autonomously with continued use.”
The FDA is doing its best to fulfill its mandate to ensure that Americans have access to safe and effective medical products, while also encouraging technological progress. That is a difficult balancing act at any time, but when a medical device can change on its own, and may change frequently, it’s much too burdensome and simply unrealistic to expect that the manufacturer and the FDA will have to submit and review each change as it occurs
He further adds, “At a high level, the Action Plan outlines five discrete points of continued regulatory development, the cornerstone of which is an entirely new framework that permits an adaptive AI/ML SaMD to change autonomously, but only within guardrails “pre-specified” at the time of initial review. This out of the box thinking is welcomed, but as pointed out in the Plan, will only succeed if the other Plan provisions are well formulated and executed: Continued performance surveillance and feedback through real world evidence collection; adherence to yet-to-be-developed “good machine learning practices;” maximal transparency regarding a SaMD’s inputs, outputs, and underlying mechanics; and understanding and managing potential algorithm bias.
The concerted effort to involve patient input at every stage of AI/ML development and regulation is much needed and to be applauded. Understanding patient needs and experiences is critical to keeping medical devices safe and effective, and if these devices are learning and changing on their own, it is the patients who will see those changes first and be most affected by them
“This is a tall order, both for industry and FDA. But the ensemble Plan will work to foster innovation while maintaining the reasonable assurance of device safety and effectiveness we all expect. FDA is to be commended in taking this innovative, interactive, proactive, and inclusive approach”, says Dr. Saltman.