Research 2026-08-20

The FDA has cleared 1,357 AI medical devices. Three of them have been tested on whether patients live longer or better.

A team led by Rawan Abulibdeh at the University of Toronto went through every AI-enabled medical device the US Food and Drug Administration had authorised as of 5 December 2025 — 1,357 of them — and looked for evidence that the device helped a patient. They published the count in PLOS Digital Health on 19 August. Thirty-four devices appeared in a registered clinical trial. Twelve of those trials had posted results. Twelve had produced a peer-reviewed paper. And three had been evaluated against patient-centred outcomes, meaning things a patient would recognise as mattering: death, stroke, hospitalisation, quality of life. The evidence that does exist is also narrow. Most of the studies ran inside well-resourced health systems, and most systematically excluded pregnant women, adults over 75 and people who do not speak English. 'We expected the evidence base to be thin, but not this thin,' Abulibdeh said.

Why it mattersThis is not a claim that the devices do not work. Most were cleared through a pathway that asks whether a device performs comparably to something already on the market — whether the algorithm flags the nodule the radiologist would have flagged — and many pass that bar honestly. The finding is that almost nobody took the next step and checked whether flagging it changed what happened to the patient. Those are different questions, and medicine has learned the hard way that the first does not imply the second: a test can be accurate and still lead to treatment that does not help. The exclusions matter for the same reason. A device validated on English-speaking patients under 75 in a teaching hospital has not been shown to work on the people most likely to encounter it in an under-resourced clinic. Regulatory clearance is being read by hospitals and patients as evidence of benefit, and on this count it mostly is not.

#Health AI #Science & Research

✓ Verified · 4 sources

Machine learning read the shape of sick brain cells and picked out nine already-approved drugs that calmed them down
2026-08-21 Given four hours and a GPU to improve the way AI is trained, the best agent scored 0.25 out of 1 — and most never tried
2026-08-21 The agent invented a second person to vouch for its code. A 24-year-old in Texas refused to believe either of them.
2026-08-21 Pew put half a million web pages through a detector: a third of everything published since ChatGPT carries its marks
2026-08-21 So many machine-written proofs arrived that mathematicians built a filing office — and one of its two checks is done by a language model
2026-08-20

The FDA has cleared 1,357 AI medical devices. Three of them have been tested on whether patients live longer or better.

Related stories