Design: Pragmatic cluster-randomised quality improvement study
Population: 816 endoscopists at 139 US veterans' facilities, over 330,000 colonoscopies
Primary outcome: Adenoma detection rate
Effect: +4.2 points (50.7% to 54.9%) at CADe sites; adjusted OR 1.22 (1.15 to 1.28)

This pragmatic cluster-randomised study, published in June 2026, equipped randomly selected US veterans' health facilities with computer-aided detection devices and compared them with control facilities. It covered more than 330,000 colonoscopies by 816 endoscopists across 139 facilities.

Adenoma detection rose from 50.7% to 54.9% at facilities with the devices (an increase of 4.2 percentage points), while it fell slightly at control sites. After adjustment, device availability was associated with higher odds of adenoma detection (odds ratio 1.22, 95% CI 1.15 to 1.28). Detection of other lesions and withdrawal time did not change, and the effect was similar regardless of the endoscopist's baseline detection rate.

Trials had shown benefit, but real-world studies had been inconsistent; this is one of the larger pragmatic confirmations. Whether more adenomas found translates into fewer cancers, and whether reliance on the device dulls skills, are both unknown.

Consider computer-aided detection as an aid to adenoma detection in colonoscopy units.
Keep monitoring each endoscopist's adenoma detection rate.
Maintain withdrawal time and bowel preparation standards; the device does not replace them.
Expect a modest gain of a few percentage points.

Why it matters

Real-world benefit from these devices had been uncertain; this pragmatic design answers that question for detection, not for cancer.

The statistics, in plain English

An odds ratio of 1.22 for adenoma detection reflects an absolute gain of about four percentage points. Adenoma detection is a surrogate linked to lower interval cancer risk, but the study did not measure cancers.