A new CDOTrends analysis examines the gap between the growing number of AI-enabled medical devices and the much smaller body of evidence demonstrating that these tools improve patient outcomes. The article cites FDA data showing 1,524 authorized AI-enabled medical devices as of March 2026, with 76% in radiology. It also references a PLOS Digital Health analysis finding that only 34 of 1,357 FDA-cleared AI devices reviewed had links to registered clinical trials, and only three had been evaluated using patient-centered outcomes.
The article uses gastroenterology as a particularly relevant example. Mark Lambrecht of SAS describes a gastroenterologist who keeps AI polyp detection turned off because of concerns about being replaced, despite the technology’s accuracy. The broader argument is that adoption depends not only on technical performance but on prospective evidence showing meaningful clinical benefit.