Enhancing colonoscopy outcomes: A systematic review and meta-analysis of randomized controlled trials evaluating artificial intelligence–assisted adenoma and polyp detection.

e15519 Background: Colorectal cancer (CRC) remains a leading cause of cancer-related mortality worldwide. Colonoscopy is the cornerstone of CRC screening; however, adenoma detection is operator dependent and susceptible to human error. Artificial intelligence (AI)–assisted colonoscopy has emerged as a real-time adjunct to enhance lesion recognition and procedural quality. We conducted a systematic review and meta-analysis of randomized controlled trials (RCTs) to evaluate the impact of AI-assisted colonoscopy on adenoma detection rate (ADR) and polyp detection rate (PDR). Methods: This systematic review and meta-analysis was conducted in accordance with PRISMA guidelines. PubMed, Google Scholar, and the Cochrane Library were searched from inception through August 2025 for RCTs comparing AI-assisted colonoscopy with standard colonoscopy. ADR was defined as the proportion of patients with ≥1 histologically confirmed adenoma per colonoscopy, and PDR as the proportion with ≥1 detected polyp. Pooled odds ratios (ORs) with 95% confidence intervals (CIs) were calculated using a DerSimonian–Laird random-effects model. Statistical heterogeneity was assessed using the I² statistic, and statistical significance was defined as p < 0.05. Results: Twenty-six RCTs comprising 14,056 patients were included (AI-assisted: n = 7,026; standard colonoscopy: n = 7,030). AI-assisted colonoscopy was associated with significantly higher detection rates compared with standard colonoscopy, with pooled ORs of 1.37 (95% CI 1.25–1.51; p < 0.001) for ADR and 1.36 (95% CI 1.19–1.56; p < 0.001) for PDR. Moderate heterogeneity was observed for ADR (I² = 60%) and substantial heterogeneity for PDR (I² = 75%), likely reflecting variability in AI platforms, baseline detection rates, and operator experience. Sensitivity analyses confirmed the robustness of the findings. Funnel plot assessment did not suggest significant publication bias. Risk of bias assessment demonstrated low risk for random sequence generation and outcome assessment blinding, with performance bias primarily related to the inability to blind endoscopists to AI assistance. Conclusions: AI-assisted colonoscopy significantly improves adenoma and polyp detection compared with standard colonoscopy, supporting its role as a tool to enhance screening quality and procedural performance. These improvements may reduce missed lesions and potentially lower interval CRC risk. Further large-scale implementation and real-world effectiveness studies are warranted to optimize integration across diverse practice settings.