What was studied: Researchers tested whether adding artificial intelligence (AI) to colonoscopy helps doctors find more precancerous growths (adenomas) in people with Lynch syndrome — an inherited condition that greatly raises the risk of colorectal cancer and requires regular colonoscopy screening.
How it was done: About 750 adults with Lynch syndrome at nine specialist centres in Belgium, Germany, the Netherlands and Spain were randomly assigned to have either a standard high-definition colonoscopy or the same procedure with an AI system (CAD EYE) that flags suspicious areas on the screen in real time. The AI was also tested on its ability to tell, on the spot, whether a growth was precancerous or harmless.
What was found:
- Adenomas were found in 34% of patients with AI vs 31% without — a difference too small to be meaningful.
- No advantage for AI in any other measure, including flat growths, advanced adenomas, or cancers detected.
- Procedure times (about 26 minutes) and patient comfort were the same in both groups.
- For identifying growth types on sight, the AI performed slightly worse than the expert doctors, and both struggled with a tricky lesion type called sessile serrated lesions.
- Safety was similar; three minor complications occurred in the AI group, none clearly caused by the AI.
What it means: In expert centres where doctors are already highly skilled and take their time, adding this AI tool didn’t improve results. Careful technique, good bowel preparation and adequate inspection time still matter most.
Important caveats: The study wasn’t designed to prove the two approaches are equivalent — only that AI wasn’t better in this setting. AI could still help in less specialised clinics where detection rates vary more, and future AI trained specifically on the subtle, flat lesions typical of Lynch syndrome might perform better.
Bottom line for patients: A high-quality standard colonoscopy at an experienced centre remains appropriate care — you’re not missing out if your clinic doesn’t use AI.
Published in The Lancet Gastroenterology & Hepatology (CADLY2 trial; first author Dr Robert Hüneburg, University Hospital Bonn).
