S650 AI-Assisted vs Traditional Colonoscopy in Colorectal Neoplasia Detection: A Systematic Review and Meta-Analysis of RCTs and Observational Studies

Introduction: In the United States, Colorectal cancer (CRC) remains a leading cause of cancer-related mortality. This systematic review and meta-analysis compare AI-assisted colonoscopy (AI-AC) to traditional colonoscopy (TC), integrating both randomized controlled trials (RCTs) and observational studies (OBS) to evaluate AI’s impact not only in controlled research settings but also in clinical practice. Methods: A comprehensive systematic search was conducted on PubMed, Embase, Scopus, and the Cochrane Library to identify RCTs and OBS comparing AI-AC with TC for adenoma and polyp detection till March 2025. Meta-analysis was conducted using RevMan 5.4.1. For dichotomous outcomes, pooled risk ratios (RRs) with 95% confidence intervals (CIs) were calculated using the Mantel-Haenszel method. For continuous outcomes, mean differences (MDs) with 95% CIs were used. Random or fixed-effects models were applied based on heterogeneity, assessed using Higgins I² statistic. Statistical significance was set at P-value < 0.05. The risk of bias was assessed using the RoB 2.0 tool for RCTs and the Newcastle-Ottawa Scale for OBS. Results: A total of 14 studies were included: 11 RCTs with 10,481 patients (5,252 AI-AC, 5,229 TC) and 3 OBS with 4,544 patients (1,756 AI-AC, 2,788 TC). Adenoma Detection Rate (ADR) was significantly higher with AI-AC in both RCTs (RR 1.41, CI 1.25–1.60, P < 0.00001) and OBS (RR 1.15, CI 1.07–1.23, P < 0.0001), with a pooled RR of 1.34 (CI 1.21–1.48, P < 0.00001). Adenomas per Colonoscopy (APC) was also higher with AI-AC, with a mean difference of 0.19 (CI 0.16–0.22, P < 0.00001) in RCTs, 0.14 (CI 0.07–0.21, P = 0.0002) in OBS, and a pooled mean difference of 0.18 (CI 0.15–0.21, P < 0.00001). Polyp Detection Rate (PDR) was significantly higher in the AI-AC group in RCTs (RR 1.40, CI 1.27–1.55, P < 0.00001) but not in OBS (RR 1.02, CI 0.82–1.28, P = 0.84), with a pooled RR of 1.36 (CI 1.22–1.51, P < 0.00001). AADR was higher with AI-AC in RCTs (RR 1.45, CI 0.94–2.23, P = 0.09) and OBS (RR 1.56, CI 0.84–2.91, P = 0.16) without statistical significance, though pooled analysis showed significant improvement (RR 1.44, CI 1.06–1.95, P = 0.02). Conclusion: Results of RCTs and OBS consistently show that AI-AC significantly improves ADR, APC, PDR, and AADR compared to TC. These findings support that AI-AC enhances neoplasia detection across diverse settings and may improve screening efficacy in routine practice. Variability among studies likely reflects differences in patient populations, endoscopist experience, and AI systems.