Computer-Aided Detection Systems for Colonoscopy and Colorectal Cancer Prevention
Mené à Taïwan auprès de 1 356 participants (âge moyen : 60 ans ; 50 % de femmes et 50 % d'hommes), cet essai multicentrique randomisé évalue la non infériorité, par rapport à une coloscopie standard et du point de vue du taux de détection d'adénomes colorectaux, d'une coloscopie avec détection assistée par ordinateur
The adenoma detection rate (ADR) is the most widely endorsed quality indicator for colonoscopy, and computer-aided detection (CAD) systems have consistently increased ADRs across randomized clinical trials (RCTs), with a pooled relative risk of approximately 1.24. Yet whether each additional percentage point in ADR translates into proportional colorectal cancer (CRC) prevention remains uncertain, especially for patients with high baseline ADR. Hsu et al report a multicenter RCT from 4 centers in Taiwan that randomized 1457 adults to CAD-assisted or standard high-definition colonoscopy. The trial enrolled patients referred for colonoscopy on the basis of a positive fecal immunochemical test (FIT), symptoms, routine screening, or surveillance. CAD achieved noninferiority for ADR compared with standard colonoscopy (58.5% vs 53.3%; prespecified noninferiority margin, −10%). Superiority in the overall population was not statistically significant (P = .05). In an exploratory subgroup of FIT-positive patients, CAD showed a higher ADR (65.3% vs 57.4%; adjusted odds ratio, 1.39; 95% CI, 1.05-1.86; P = .02) and higher adenomas per colonoscopy (1.64 vs 1.39; P = .01). The nonneoplastic polypectomy rate did not differ between groups (4.6% vs 4.1%; P = .66), addressing concerns that CAD without concurrent computer-aided diagnosis might prompt unnecessary resections. The strengths of this RCT are notable. By evaluating CAD in a pathway aligned with FIT-based screening—while also including patients undergoing colonoscopy for symptoms, routine screening, or surveillance—Hsu et al tested the technology in a clinically relevant, high-prevalence cohort. This pragmatic design provides direct evidence for screening programs weighing CAD implementation. The comparable nonneoplastic polypectomy rate between groups also removes a practical barrier to adoption.
JAMA Network Open , éditorial en libre accès, 2026