Loading Curie Briefs...
Getting the latest healthcare news for you
Getting the latest healthcare news for you
Over 1,300 AI medical devices have been cleared by the FDA — but only 3 were tested on patient outcomes like death or morbidity. A new systematic review reveals a striking evidence gap in how these tools are validated, raising serious questions about whether AI in healthcare is actually helping patients or just clearing a regulatory hurdle.
More than 1,300 AI medical devices have received FDA clearance, yet a sweeping new systematic review published in PLOS Digital Health found that only three of those tools were ever tested on patient-centered outcomes like death, morbidity, hospitalization, or quality of life. Researchers from University Health Network in Canada analyzed the FDA device database and found that the agency's 510(k) clearance pathway — which requires only "substantial equivalence" to an existing device, not proof of clinical effectiveness — is allowing major evidence gaps to go unaddressed.
The studies that did exist were often small, single-center, and excluded vulnerable populations such as pregnant patients, older adults, youth, and non-English speakers. This raises equity concerns: imaging and cardiac monitoring tools, for example, may be used in obstetric emergencies despite never being validated in pregnant women.
By the Numbers:
Why it matters: As AI becomes embedded in clinical decision-making, the lack of rigorous outcome-based validation means clinicians — and patients — may be relying on tools that have never been proven to improve health. Medical societies like the AMA are already calling for stronger regulatory standards to close this gap.
