Computer-aided polyp detection multicenter international randomized controlled study with a focus on community clinics: Geriatric Assessment-Driven Intervention clinical trial.

BACKGROUND AND AIMS: Computer-aided polyp detection (CADe) systems can enhance polyp detection, although their performance has largely been evaluated at academic rather than community practices. This randomized controlled trial evaluated the impact of the DEtection of Elusive Polyps-2 (DEEP2) CADe system on colonoscopy quality metrics. METHODS: Participants undergoing routine colonoscopies were randomized to a control arm, standard-of-care high-definition white-light endoscopy (HD-WLE), or an intervention arm, HD-WLE with real-time DEEP2 system assistance. Coprimary end points: adenomas per colonoscopy (APC; superiority evaluation at a 0.025 level of significance), and percent positive agreement (PPA, defined [per regulatory request] as the proportion of histologically confirmed clinically significant excised lesions relative to total lesions; noninferiority evaluation with a lower 95% CI threshold of -10%). Secondary end points included adenoma detection rate (ADR), polyp detection rate, polyps per colonoscopy, false positive rate, and false alarm rate. RESULTS: This study enrolled 1407 participants, most in community clinics (n = 1208, 85.9%). Use of the DEEP2 system resulted in significantly higher APC (0.74 vs 0.60; difference, 0.14; P = .0002). The PPA difference was -0.06 (95% CI, -0.15 to 0.03; P = .09). There was a numerical, but not statistically significant, ADR improvement (42.9% vs 38.9%; P = .031). A post hoc analysis revealed significant benefit in APC and ADR with the DEEP2 system only within community clinics. CONCLUSIONS: These results support the incorporation of CADe devices into colonoscopy practice for improving efficacy and quality over standard-of-care, particularly in community practices.

Duke Scholars

Author David Asher Leiman Medicine, Gastroenterology