Does computer-aided detection still add value in high-performing colonoscopy? A systematic review and meta-analysis of randomized trials.

Masmoum Mohd Diya, Shayah Yamen, Masmoum Ghali, Asmew Jabr

Clinical journal of gastroenterology · 2026 · PMID 43050091

PubMed ↗DOI ↗

BACKGROUND: Computer-aided detection (CADe) improves adenoma detection in broad randomized-trial populations, but its incremental value may diminish when baseline colonoscopy performance is already high. We evaluated whether CADe improves advanced neoplasia detection in contemporary high-detection practice.

METHODS: PubMed/MEDLINE, Embase, CENTRAL, Scopus, Google Scholar, ClinicalTrials.gov, and WHO ICTRP were searched from inception through September 2026. Randomized trials were eligible when control-arm adenoma detection rate (ADR) was at least 35%; FIT-only trials were analyzed separately. Advanced adenoma detection rate (AADR) and advanced colorectal neoplasia detection rate (ACNDR) were primary outcomes. Random-effects models used restricted maximum likelihood with modified Hartung-Knapp inference. Risk of bias was assessed using RoB 2 and certainty using GRADE. The protocol was prospectively registered in PROSPERO (CRD420261495680). No external funding was received.

RESULTS: Forty-eight unique randomized trial families were included. In routine high-detection trials, no statistically significant improvement in AADR was demonstrated (17 trials; RR, 1.018; 95% CI, 0.921-1.125; I² = 0%). CADe reduced whole-colon tandem adenoma miss rate (6 trials; RR, 0.473; 95% CI, 0.363-0.618) and increased overall ADR (28 trials; RR, 1.137; 95% CI, 1.082-1.194). The ADR effect was not clear when control ADR was at least 50% (RR, 1.039; 95% CI, 0.964-1.120). Sessile serrated lesion detection increased (RR, 1.284; 95% CI, 1.125-1.465). Advanced-neoplasia estimates were limited by sparse compatible reporting.

CONCLUSIONS: In high-detection colonoscopy settings, CADe substantially reduces adenoma misses and improves overall adenoma and sessile serrated lesion detection, demonstrating benefits beyond advanced neoplasia detection. Although a statistically significant improvement in advanced adenoma detection was not demonstrated, the potential long-term clinical benefits of improved lesion recognition remain uncertain. PROSPERO REGISTRATION: CRD420261495680.

📄 이 논문을 인용한 Paperis 글

이 논문이 근거 목록에 올라 있는 Paperis 글입니다.