- More than 1,400 AI medical devices are currently on the market, with most lacking robust independent validation studies.
- About 97% of AI and machine learning medical devices reach the market through the FDA’s 510(k) pathway.
- The 510(k) pathway requires substantial equivalence to existing devices rather than independent proof of safety or effectiveness.
- Predicate device chains can include older, recalled, or even non-AI devices.
- Premarket studies for AI tools are rarely randomized, prospective, multisite, or representative of diverse patient populations.
- Existing postmarket surveillance systems, including the MAUDE database, may capture as little as 0.5% of adverse events.
- The authors urged clinicians and procurement teams to evaluate training data, subgroup performance, and validation populations rather than relying solely on FDA clearance.
Daily News
Stay up to date with the latest clinical headlines and other information tailored to your specialty.
