Chinnapong - stock.adobe.com
Less than 1% of FDA-cleared AI devices tested for clinical benefits
A recent study revealed that only three of 1,300+ FDA-cleared AI medical devices were evaluated for patient outcomes, exposing gaps in clinical evidence requirements.
The FDA has cleared more than 1,300 AI-enabled medical devices, but the vast majority have not been evaluated for patient-centered outcomes, a new report shows.
This comes as the FDA continues to grapple with how best to regulate AI and its impact on patients.
Published in PLOS Digital Health, the study aimed to examine the gap between the number of FDA-cleared AI-enabled devices and the clinical evidence supporting them. Researchers from the U.S. and Canada aimed to quantify the gap among tools that received the FDA's 510(k) clearance, which does not require the vendor to demonstrate independent evidence of safety and effectiveness. The 510(k) pathway only asks that vendors show a tool has "substantial equivalence" to an existing one.
Very few FDA-cleared AI devices evaluated for patient outcomes
For the study, the researchers evaluated all 1,357 AI- and machine learning-enabled medical devices cleared by the U.S. FDA through Dec. 5, 2025. They identified published studies on the devices' outcomes via PubMed searches, as well as prospective trial registrations from ClinicalTrials.gov links in FDA 510(k) summary pages.
They found that only 34 (2.5%) devices were linked to registered prospective trials, of which 12 (0.9%) posted results and 12 (0.9%) progressed to peer-reviewed publication. This means, of the few devices being studied, very few published early results or more comprehensive results evaluated by independent experts. Notably, only three devices (0.2%) had been evaluated for patient-centered outcomes such as mortality, morbidity or readmissions.
Not only that, but industry sponsorship dominated this clinical evidence. Of the 34 trials, 32 (94%) were industry-led, with designs heavily focused on accuracy metrics rather than patient outcomes.
"For clinicians and health systems, this creates an evidence base that is incomplete, systematically biased, and tilted toward optimism," the study authors noted.
The small number of devices linked to clinical evidence also highlighted alarming specialty-specific insights. For instance, radiology accounts for 78% of cleared devices, but only 1% were linked to prospective trials. Cardiovascular and neurology devices fared better, but still less than 10% of devices in each specialty had a prospective clinical trial.
The study authors underscored that closing this clinical evidence gap will require evidence that measures patient outcomes rather than algorithmic accuracy. They suggested that the FDA mandate pre-registration of prospective trials for all Class II and III AI devices before clearance.
Further, they recommended that devices be validated in populations where the tools are used and that CMS and other insurers link reimbursement to demonstrated clinical benefit.
FDA's broad stance on AI innovation
The research comes as the FDA released a discussion paper on the regulation of GenAI-enabled devices. The paper offers a framework for risk assessment as well as potential approaches to premarket evaluation and postmarket monitoring.
"Generative AI-enabled medical devices are poised to reshape the health technology landscape, and the FDA has an important responsibility to provide thoughtful leadership for this new era," said Rick Abramson, M.D., director of the FDA's Digital Health Center of Excellence, in the press release.
The FDA is also requesting stakeholder feedback on its proposed approach to regulating genAI devices. The agency asked device manufacturers, clinicians, consumers, researchers and the public to submit their feedback on Regulations.gov by Oct. 19, 2026.
However, even as the FDA seeks feedback on AI regulation, it has begun selecting companies for its Technology-Enabled Meaningful Patient Outcomes for Digital Health Devices Pilot, including those that use AI.
Among the four companies currently participating in the pilot are Candence, which offers an AI‑powered Proactive Care Engine, and Dexcom, which added genAI to its over-the-counter continuous glucose monitors two years ago.
Anuja Vaidya has covered the healthcare industry since 2012. She currently covers healthcare IT and innovation, including artificial intelligence, digital healthcare, EHRs and interoperability.
