Many FDA-approved AI devices lack validation data

A recent article in *Nature Medicine* reveals that over 40% of artificial intelligence tools approved by the FDA lack clinical evidence. Researchers from the University of North Carolina School of Medicine found that 226 out of 521 AI-enabled medical devices authorized between 1995 and 2022 did not include clinical validation data, which accounts for about 43%.

The study's authors urged the FDA and AI developers to publish more clinical validation data and prioritize prospective studies. Most AI devices were cleared through the FDA's 510(k) process, which is less rigorous and requires manufacturers to demonstrate safety and effectiveness compared to existing devices.

Among 292 devices with validation data, only 22 were subjected to randomized controlled trials. The researchers argued that prospective studies are essential for assessing AI tools' effectiveness, as they can reveal confounding factors and offer clearer insights into clinical utility.