Artificial Intelligence Software as a Medical Device: Regulatory Approvals and Post-Deployment Evaluation Frameworks in North America and Europe
Author(s)
Remziye Zaim, PhD, MBA, MSc, MS1, Isaac Corro Ramos, PhD, MSc2, Frederick W. Thielen, PhD, MSc3, James Andrew Shaw, PhD, MPT4;
1University of Toronto, Dalla Lana School of Public Health, Toronto, ON, Canada, 2Erasmus University Rotterdam, Institute for Medical Technology Assessment, Rotterdam, Netherlands, 3Erasmus University Rotterdam, Erasmus School of Health Policy & Management, Rotterdam, Netherlands, 4University of Toronto, Temerty Faculty of Medicine, Toronto, ON, Canada
1University of Toronto, Dalla Lana School of Public Health, Toronto, ON, Canada, 2Erasmus University Rotterdam, Institute for Medical Technology Assessment, Rotterdam, Netherlands, 3Erasmus University Rotterdam, Erasmus School of Health Policy & Management, Rotterdam, Netherlands, 4University of Toronto, Temerty Faculty of Medicine, Toronto, ON, Canada
Presentation Documents
OBJECTIVES: Artificial Intelligence (AI) software classified as a medical device (SaMD) is governed by regulatory agencies globally. This study aims to compare the regulatory approvals of SaMDs across North America and Europe and explore Health Technology Assessment (HTA) frameworks related to the post-deployment evaluation of SaMDs within healthcare systems.
METHODS: We searched SaMD regulatory approvals using databases from the United States Food and Drug Administration (FDA), the European Medicines Agency (EMA), and the United Kingdom (U.K.) Medicines and Healthcare products Regulatory Agency (MHRA). We then examined HTA guidance from the U.K. National Institute for Health and Care Excellence (NICE) and Canada’s Drug Agency (CDA) databases, specifically for the SaMD classification.
RESULTS: The FDA's Digital Health Center of Excellence authorized a total of 1016 SaMDs. The most common medical specialties were radiology (777 [76.48%]), cardiovascular use (104 [10.24%]), and neurology (42 [4.13%]). Unlike the FDA's searchable database, the MHRA, Health Canada, and the EMA do not maintain a publicly accessible registry dedicated to approved SaMDs. While the MHRA, FDA, and Health Canada have jointly established guiding principles for the transparency of AI-enabled devices, EU regulations require manufacturers to undergo a conformity assessment process to demonstrate that their devices meet safety and performance requirements. Final approval and CE ("Conformite Europeenne") markings are granted by Notified Bodies—independent organizations designated by EU member states—not by the EMA. HTA frameworks from the NICE and CDA do not yet provide specific guidance on the post-deployment evaluation processes of SaMDs.
CONCLUSIONS: This study examines the regulation of AI-enabled medical devices in North America and Europe, focusing on SaMDs. It underscores the need for publicly accessible registries of regulated AI-enabled medical devices and adaptive HTA frameworks to guide continuous monitoring and ethical governance of AI for responsible integration into health systems.
METHODS: We searched SaMD regulatory approvals using databases from the United States Food and Drug Administration (FDA), the European Medicines Agency (EMA), and the United Kingdom (U.K.) Medicines and Healthcare products Regulatory Agency (MHRA). We then examined HTA guidance from the U.K. National Institute for Health and Care Excellence (NICE) and Canada’s Drug Agency (CDA) databases, specifically for the SaMD classification.
RESULTS: The FDA's Digital Health Center of Excellence authorized a total of 1016 SaMDs. The most common medical specialties were radiology (777 [76.48%]), cardiovascular use (104 [10.24%]), and neurology (42 [4.13%]). Unlike the FDA's searchable database, the MHRA, Health Canada, and the EMA do not maintain a publicly accessible registry dedicated to approved SaMDs. While the MHRA, FDA, and Health Canada have jointly established guiding principles for the transparency of AI-enabled devices, EU regulations require manufacturers to undergo a conformity assessment process to demonstrate that their devices meet safety and performance requirements. Final approval and CE ("Conformite Europeenne") markings are granted by Notified Bodies—independent organizations designated by EU member states—not by the EMA. HTA frameworks from the NICE and CDA do not yet provide specific guidance on the post-deployment evaluation processes of SaMDs.
CONCLUSIONS: This study examines the regulation of AI-enabled medical devices in North America and Europe, focusing on SaMDs. It underscores the need for publicly accessible registries of regulated AI-enabled medical devices and adaptive HTA frameworks to guide continuous monitoring and ethical governance of AI for responsible integration into health systems.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
HPR25
Topic
Health Policy & Regulatory
Topic Subcategory
Approval & Labeling, Coverage with Evidence Development & Adaptive Pathways
Disease
No Additional Disease & Conditions/Specialized Treatment Areas