Missing Clinical Data Raises AI Medical Device Recall Risk

Missing Clinical Data Raises AI Medical Device Recall Risk

Artificial intelligence-powered medical devices lacking complete clinical study documentation face significantly higher recall rates, according to research examining nearly three decades of FDA clearances.

A study published June 11 in JAMA Network Open analyzed 903 healthcare AI devices cleared by the U.S. Food and Drug Administration between 1995 and 2024. Researchers found that approximately 5% of these devices—43 products total—were recalled at a median interval of 458 days after clearance.

The analysis, led by Daniel Windecker, MD, of the University of Bern's Department of Diagnostic and Interventional Neuroradiology, identified several factors that increase recall likelihood. Devices with missing information from supporting clinical studies showed elevated risk compared to those with published analyses. Products flagged by market surveillance systems, including Europe's CORE-MD tool, similarly demonstrated higher recall rates.

Why it matters

As healthcare organizations accelerate AI adoption, understanding which validation gaps predict device failures becomes critical for procurement decisions and patient safety protocols. The findings suggest current premarket evaluation frameworks may inadequately assess how AI performs across diverse real-world clinical settings—a gap that could expose health systems to both regulatory and patient safety risks.

Radiology bears disproportionate burden

Of the 903 devices studied, 692 were radiology-related, with 4.3% (30 products) recalled during the study period. The research found radiology devices are approximately 52% more likely to face recall than AI products in other medical specialties, though this elevated rate likely reflects radiology's dominant share of the total device population analyzed.

Devices ineligible for FDA's third-party review process also showed higher recall hazard, as did products lacking documented clinical performance data.

Misuse drives most recalls

The most common trigger for FDA recalls involved deviations from a device's intended use—what the researchers characterized as "unique safety challenges" specific to AI systems. Incorrect utilization of AI devices carried a higher hazard of market withdrawal compared to other failure modes.

By contrast, issues related to temperature (occurring in roughly one of nine recalls) and labeling problems (about two of nine recalls) were associated with lower recall risk.

The researchers warn against deploying AI in clinical contexts that differ from the original intended setting or without robust supporting evidence. "Although this limitation is not unique to AI-enabled devices, their data-driven and context-dependent nature may amplify the consequences of inadequate validation," the authors wrote.

Call for stronger oversight

Windecker and colleagues believe their work represents the first systematic exploration of associations between AI device characteristics and recall risk. They advocate for more rigorous and transparent clinical performance evaluation, along with strengthened postmarket surveillance.

"These findings highlight the importance of robust clinical validation and strengthened postmarket oversight for AI-enabled devices," the authors concluded.

The findings were first reported in JAMA Network Open, using data from FDA adverse event reports and manufacturer databases through August 31, 2024.