Efficacy of artificial intelligence for adenoma detection in water exchange colonoscopy: A two-center randomized controlled trial.

Background Water exchange (WE) and artificial intelligence-based computer-aided detection (CADe) separately improve the adenoma detection rate (ADR) and adenomas per colonoscopy (APC). We aimed to determine whether combining WE with CADe enhanced APC compared with WE alone. Methods This randomized controlled trial was conducted at hospitals in Italy (ENDO-AID; Olympus) and Taiwan (CAD-EYE; Fujifilm) using different CADe devices. Patients aged 45─75 years undergoing colonoscopy for screening, surveillance, and positive fecal blood test results were randomized to either WE with CADe assistance (WE-CADe) or WE alone. The primary outcome was APC, with 752 patients planned for enrollment. Results An interim analysis was conducted on 560 patients (75% of the enrollment target; mean age: 59.4 years; men: 299; WE-CADe group: 279), with similar baseline characteristics between the two groups. APC was significantly higher in the WE-CADe group compared with the WE-alone group (1.39 [95% CI 1.06─1.72] vs. 1.05 [95% CI 0.87─1.23]) with an incidence rate ratio of 1.32 (95% CI 1.14─1.54), representing an absolute increase of 0.34. The observed significant difference led to early trial termination. No significant differences were found in ADR and sessile serrated lesion detection rates between the groups (54.1% vs. 50.2%, P = 0.350 and 3.6% vs. 3.6%, P = 0.987, respectively), but this study was not powered to detect such differences. Withdrawal times and the mean number of non-neoplastic lesions per colonoscopy were comparable. Conclusions For WE colonoscopy, integrating CADe statistically increased APC without prolonging the withdrawal time or causing a concomitant increase in non-neoplastic lesion resection.