Performance evaluation of a multi-cancer early detection assay using NMR-based molecular signatures in a multi-center study.

e15079 Background: Cancer is the second leading cause of death worldwide, accounting for over 10 million deaths annually. Despite the proven impact of early detection on patient outcomes, current screening strategies often detect cancers at advanced stages. Wholomics has developed a Multi-Cancer Early Detection (MCED) assay utilizing proprietary molecular signatures derived from human serum, assessed via Nuclear Magnetic Resonance (NMR)-based technology. This study evaluates the performance of the MCED assay in detecting multiple cancer types at early stages. Methods: A retrospective analysis was conducted using 1,211 serum samples sourced from five European centers. The cohort comprised 669 cancer samples (colorectal, lung, pancreatic, cervical, uterine, bladder, breast, gastric, renal, ovarian, esophageal, GIST, liver, and gallbladder cancer), 50 precancerous samples, 50 benign samples, and 442 control samples (healthy individuals and patients with various non-cancerous pathologies such as heart disease, arthritis and type 2 diabetes). Two machine learning models—one optimized for sensitivity and one for specificity—were trained using 10-fold cross-validation and subsequently tested on independent cohorts. Primary endpoints included sensitivity, specificity, and tumor location prediction accuracy. Results: The sensitivity-tuned model achieved a specificity of 97.1% and a sensitivity of 91.8% with a stage I sensitivity of 91.9%, while the specificity-tuned model reached a specificity of 99.1% and a sensitivity of 76.6% with a stage I sensitivity of 78.8%. Additionally, tumor location prediction was highly accurate, with a top-1 accuracy of 77.54%, a top-2 accuracy of 89.13% and a top-3 accuracy of 92.75%. Conclusions: Wholomics’ MCED assay demonstrates strong potential for detecting multiple cancer types using a single blood draw. The robust specificity, sensitivity, and tumor localization capabilities support its clinical utility in early-stage cancer detection. We aim to validate these findings in larger prospective settings and to further assess the assay’s feasibility for average risk cancer screening and risk stratification of high-risk patients, with the goal of improving patient outcomes and reducing healthcare costs.