S596 Meta-Analysis of Economic Evaluations of AI-Assisted Colonoscopy in CRC Screening Programs

Introduction: Colorectal cancer (CRC) remains a leading cause of cancer-related death, with colonoscopy being the gold standard for screening. Despite its efficacy, significant adenoma miss rates and cost inefficiencies persist. Artificial intelligence (AI)-assisted colonoscopy, particularly computer-aided detection (CADe), has demonstrated improved adenoma detection rates (ADR). However, the economic viability of AI integration into population-level CRC screening programs remains underexplored. To perform a meta-analysis of published economic evaluations of AI-assisted colonoscopy to determine its cost-effectiveness and impact on healthcare resource utilization in CRC screening programs. Methods: A systematic review was conducted using PubMed, Embase, and Cochrane (2015–2024) in accordance with PRISMA guidelines. Inclusion criteria were cost-effectiveness analyses, budget impact models, or health economic simulations evaluating AI-enhanced colonoscopy versus standard methods. Data were extracted on incremental cost-effectiveness ratio (ICER), quality-adjusted life years (QALYs), and cost per additional adenoma detected. Meta-analysis was performed using R (meta, metafor) and Python (pandas, statsmodels) to pool effect sizes and evaluate heterogeneity. Results: From 312 screened studies, 14 met inclusion criteria. AI-assisted colonoscopy was cost-effective in 12/14 models, with a pooled ICER of $14,730 per QALY gained (95% CI: $9,450–$20,870), well below the conventional $50,000 threshold. AI increased ADR by a mean of 10.2% (95% CI: 7.1–13.3%) and reduced the cost per advanced adenoma detected by 28%. Models incorporating downstream benefits (e.g., reduced interval cancers, fewer late-stage treatments) showed a net savings of $310–$580 per screened individual over a 10-year horizon. Sensitivity analyses identified ADR improvement and device cost as key drivers. However, heterogeneity in healthcare system models and reimbursement assumptions remained a limitation. Conclusion: AI-assisted colonoscopy demonstrates robust cost-effectiveness and potential long-term savings in CRC screening programs. These findings support the strategic implementation of CADe systems to improve diagnostic yield and optimize healthcare resources. As a resident-led meta-analysis using real-world modeling data, this work underscores AI's role not only in clinical improvement but also in cost-containment—critical for national policy decisions.