CADe results from a large international, multi-centre, randomised-controlled trial : More Adenomas Detected, No Increase in Unnecessary Polypectomies

Aims There are multiple commercially available computer-aided detection (CADe) systems which have been shown to improve adenoma detection during colonoscopy. However, all these trials are fraught with significant bias as endoscopists cannot be blinded to the use of AI. We aim to address endoscopist’s bias by powering our study for two co-primary endpoints, adenoma per colonoscopy (APC) as superiority end point and adenoma per extraction (APE) as a non-inferiority endpoint. APE is defined as the total number of histologically confirmed adenomas divided by the total number of polypectomies per colonoscopy.

Methods This is a multi-center, international, randomised study to evaluate Wise Vision CADe system, in patients undergoing a screening or surveillance colonoscopy. Patients were randomised to CADe group or control group without CADe in 1:1 ratio. All endoscopists were experienced with ADR between 25% and 40% and had lifetime procedure count of>1000 colonoscopies. All polyps detected in both groups were removed and extracted for histological analysis.

Results 828 patients were recruited from 5 centres across Europe and USA. 70 patients were excluded from the analysis due to inadequate bowel prep, caecal intubation failure or technical issues. 758 patients were included in the final analysis with 377 patients in CADe group and 381 patients in control group. The mean age of the population was 59.3 years with 194/758 (25.6%) being female. The mean APC in control group was 0.73 and that in CADe group was 0.96 with a mean difference of 0.23 (95% CI: 0.01 to 0.45), corresponding to a 31.5% increase in APC compared to the control group (IRR=1.31; p=0.026). APE rate in CADe group was non-inferior to the control group (47.0% vs. 45.3%, respectively), with a difference of 0.02 (95% CI: -0.04 to 0.08; p=0.683). ADR in the CADe group was significantly higher than the control group (48.3% vs 36.0%, p=<0.01). Sessile Serrated Lesion (SSL) detection rate was numerically higher in the CADe group, at 9.5% compared to 7.9% in the control group but did not reach a statistical significance (p=0.414).

Conclusions Most CADe trials in the past have demonstrated an increase in ADR but failed to evaluate an impact on APC. We believe that APC is a superior endpoint than ADR. Our study demonstrates that CADe can significantly improve adenoma per colonoscopy (APC) as well as adenoma detection rate (ADR). The non-inferiority of APE rates in our study validates our findings as a true AI effect and negates any endoscopist-related bias. It also reassures that CADe did not result in unnecessary polypectomies and associated cost and complications.

Publication History

Article published online:
27 March 2025