AI-enabled devices lacking validation data prior to FDA clearance more likely to be recalled
Artificial intelligence-enabled devices that are not prospectively validated prior to receiving clearance from the Food and Drug Administration are more likely to be subject to recalls.
A new paper in JAMA Health Forum details numerous factors that increase the likelihood an AI-enabled device will be recalled within a few years of its official clearance. Notably, a device’s prior validation methods, or lack thereof, significantly influence its odds of requiring FDA interference after becoming commercially available.
“Artificial intelligence-enabled medical devices (AIMDs) are increasingly present in U.S. clinical practice, with nearly all cleared through the U.S. Food and Drug Administration 510(k) pathway,” Tinglong Dai, PhD, with the Carey Business School, Johns Hopkins University, and colleagues noted. “Because 510(k) clearance does not require prospective human testing, many AIMDs enter the market with limited or no clinical evaluation; meanwhile, recalls may undermine clinician and patient confidence in their performance.”
For their work, the team reviewed the information of all recalls issued for AI-enabled devices between Nov. 15 to Nov. 30, 2024. They paid special attention to the devices’ validation methods prior to earning clearance, classifying them as none, retrospective or prospective based on FDA summaries; the recalls themselves were categorized as diagnostic or measurement errors, functionality delay or loss, physical hazards, biochemical hazards, and postmarket changes.
Of the 950 devices included in the analysis, 60 (or 6.3%) were associated with 182 recall events. The majority of recalls were owed to diagnostic- of measurement-related errors, affecting more than 935,000 units total; functionality delay (44 recalls), physical hazards (14 recalls) and biochemical hazards (13 recalls) followed behind.
Nearly half of the recalls occurred within the first 12 months of the devices being cleared by the FDA. Devices without validation methods reported to the agency were almost twice as likely to be recalled compared to those that were accompanied by either retrospective of prospective validation details. Those same devices also were involved in larger recalls. Notably, devices developed by public companies accounted for 91.8% of the recalls and 98.7% of recalled units.
The authors suggested their findings indicate that the FDA might be overlooking performance failures of AI tools during their clearance procedures. But there is a way to address the issue at hand, they advised.
“Requiring prospective evaluation or issuing time-limited clearances that lapse without confirmatory data may reduce these risks,” the team suggested. “Given that over 90% of recalled units were produced by public companies, heightened premarket clinical testing requirements and postmarket surveillance measures may improve identification and reduction of device errors, similar to risk-based strategies in pharmacovigilance.”
Learn more about the findings here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.