At the MedTech Conference in Toronto, discussions centered around the pivotal topic of artificial intelligence (AI) in medical devices, involving key regulators from the U.S. and Canada. The Food and Drug Administration (FDA) and Health Canada highlighted their ongoing and future regulatory strategies in view of the rapid integration of AI technologies in medical settings.
The FDA has approved nearly 1,000 AI-enabled medical devices, with Health Canada also authorizing hundreds. Despite this significant adoption, several critical issues remain, particularly concerning the management and regulation of AI’s adaptability over time and the oversight of generative AI tools.
During the conference, it was pointed out that one of the FDA’s top priorities for the fiscal year 2025 is to develop draft guidance focusing on lifecycle management and premarket submission recommendations for AI-driven device software functions. Troy Tazbaz, director of the FDA’s Digital Health Center of Excellence, emphasized the necessity of this guidance by noting the widespread enthusiasm for AI’s capabilities without a proportional focus on post-development integration and management.
This need for comprehensive lifecycle oversight aligns with concerns published in a JAMA article, which stressed the importance of monitoring AI performance in real-world environments due to its potential evolution and sensitivity to contextual changes. Addressing these concerns, the FDA and Health Canada are exploring how best to ensure the long-term safety and effectiveness of AI healthcare technologies. One approach discussed involved health systems playing a larger role in monitoring AI algorithms, although current clinical information systems may not be fully equipped for this task.
An example of how post-market oversight is currently handled can be seen with the FDA’s authorization of the first sepsis detection tool. The FDA mandated that the manufacturer, Prenosis, submit a post-market performance management plan. This plan is crucial to continually assess the product’s performance and address any potential bias, which is essential given the high stakes of medical applications of AI.
Significantly, AI/machine learning devices presently approved by the FDA are in a “locked” state, meaning their algorithms cannot be altered without re-approval from the FDA. However, changes are on the horizon. Following new authority granted by Congress in 2022, the FDA can now approve certain, predefined modifications to medical devices via predetermined change control plans (PCCP). Such documentation outlines permissible modifications to a device and the methods for evaluation.
Further integration of PCCP was echoed by Canada’s approach to AI in medical devices. Marc Lamoureux, manager of the Digital Health Division at Canada’s Medical Devices Directorate, indicated that Canada is set to issue guidance on machine learning-enabled medical devices after considering feedback on a draft released earlier in 2023. He noted that the guidance would focus on issues such as bias, data representativeness, and transparency, incorporating PCCP into its framework. Moreover, Canadian regulators might impose conditions on the licenses of certain devices to ensure continuous performance checks and adaptability to the Canadian demographic.
This conversation between U.S., Canada, and U.K. regulatory bodies underscores a collaborative effort to establish good machine learning practices and effective transparency measures suited for the progressive nature of medical AI technologies.
Reforming AI’s application in medical technology requires a balance between innovation and safety, a challenge recognized at the MedTech conference. The implementation of PCCPs and guidance for lifecycle management are steps towards harmonizing the rapid advancements in AI with the robust regulatory oversight necessary to secure patient safety and device efficacy. The impending final guidance from the FDA on PCCPs, anticipated to be released soon, represents a significant milestone in achieving this balance, signaling a proactive regulatory stance in harnessing the benefits of AI while mitigating associated risks.
#FDA #Health #Canada #give #sneak #peek #future #regs
