AI-enabled medical devices with missing information on supporting clinical studies had a higher estimated recall hazard in a JAMA Network Open cohort study.
The study included 903 FDA-authorized AI-enabled medical devices cleared or approved between August 1995 and August 2024. Devices were followed from FDA authorization to recall or administrative censoring through August 31, 2024.
Overall, 43 devices, or 4.8%, were recalled after a median interval of 458 days. Radiology accounted for 692 devices in the dataset, with 30 recalls, or 4.3%.
Researchers used FDA authorization records, the FDA MAUDE database, and the CORE-MD postmarket surveillance tool. Device problems were mapped to International Medical Device Regulators Forum codes.
Missing clinical-study information was associated with higher recall hazard compared with devices supported by published clinical studies. The reported hazard ratio was 1.39, with a 95% credible interval of 0.84 to 3.52.
Radiology-panel devices also showed higher estimated recall hazard than the reference group. The reported hazard ratio was 1.52, with a 95% credible interval of 0.84 to 5.20.
Use-related problems were the most common recall trigger and were linked to higher recall hazard. The authors reported that incorrect use occurred in 12 of 31 devices with available problem-code data and had a hazard ratio of 3.33.
Postmarket safety signals mattered. Devices flagged by the CORE-MD tool had a hazard ratio of 4.28, while devices flagged in both CORE-MD and MAUDE had a hazard ratio of 2.77.
“These findings highlight the importance of robust clinical validation and strengthened postmarket oversight for AI-enabled devices,” the authors wrote.
The authors said the study’s limits include reliance on the FDA AI-enabled device list and the small number of recalled devices. They also said future work should include multijurisdictional analyses and time-to-recall modeling.
Sources
- Clinical Evidence and FDA Recalls of Artificial Intelligence-Enabled Medical Devices. Peer-reviewed study
- Artificial Intelligence-Enabled Medical Devices. Government document
- Manufacturer and User Facility Device Experience (MAUDE) Database. Government document
About the author
RadiologySignal.com writersEditorial Team
Radiology Signal Staff covers developments across medical imaging, radiology AI, imaging informatics, clinical research, and radiology business. The team monitors primary sources, peer-reviewed studies, company announcements, society updates, and healthcare industry news to deliver concise reporting for imaging professionals.
