S511 Blood-Based MCED Testing During Routine Colonoscopy to Enhance Early Detection of High Mortality Gastrointestinal Cancers

Introduction: Colonoscopy is recommended for United States adults aged 45–75, driving 15 million procedures annually. However, non-colorectal gastrointestinal (GI) malignancies with stage-dependent survival lack established screening protocols. Additionally, suboptimal adherence to existing screening guidelines for other cancers, such as lung, represents missed opportunities for early cancer detection, particularly for high-mortality cancer. We evaluated the performance of a blood-based multi-cancer early detection (MCED) test intended for opportunistic screening within routine colonoscopy workflows to leverage established healthcare touchpoints and expand cancer interception beyond current paradigms. Methods: Using peripheral blood samples (NCT05435066), we trained predictive models on over 5,000 samples and performed a blinded validation on a mutually exclusive cohort of adults aged 45–75 (412 treatment-naïve cancer patients and 946 non-cancer controls). The test is intended to detect hepatobiliary (liver, biliary tract), pancreatobiliary (pancreas, gallbladder), upper GI (esophagus, stomach, esophagogastric junction), head and neck, and lung cancers using an initial low-cost assay optimized for sensitivity, followed by reflex testing with a larger panel to enable high-specificity tissue-of-origin (TOO) classification. We report both crude sensitivity (the empirical probability of correctly detecting a cancer signal regardless of TOO), and intrinsic accuracy (the empirical probability of correct TOO classification per cancer type). Results: In this cohort (median age, 64; IQR, 58-69 years; 51% female; 13% BIPOC; 74% White), the test demonstrated 98.7% specificity (95% CI, 97.9-99.3) with overall, stage I, and stage II crude sensitivities of 68.7%, 31.4% and 59.7%, respectively. Intrinsic accuracy was 77.8% for HB, 57.1% for PB, 55.2% for upper GI, 53.5% for lung with early-stage (I or II) intrinsic accuracies of 80%, 40%, and 23%, 19.3%, respectively. Using SEER22-derived incidence estimates, a modeled cohort of 100,000 tested individuals would detect the correct TOO for 28/49 pancreaticobiliary, 24/31 hepatobiliary, and 20/36 upper GI cancers, including 3/7, 11/14, and 2/10 stage I–II cases, respectively. Conclusion: Integration of MCED testing into routine colonoscopy visits enables scalable detection of non-colorectal GI and under-screened high-mortality cancers. This approach may enhance early detection efforts and merits prospective evaluation in opportunistic screening settings.