Abstract C125: A cost-effective, blood-based multi-cancer detection test validated in a large-scale study across diverse sample types, platforms, and populations

Abstract Background: Conventional cancer screening methodologies face significant challenges in resource-limited settings, constrained by high costs, technical complexity, and substantial infrastructure dependencies. These limitations critically restrict their implementation in low- and middle-income countries (LMICs). We present OncoSeek, a cost-effective multi-cancer early detection (MCED) blood test (reagent cost: 25), and evaluated its clinical performance across diverse populations. Methods: A multi-center study involving 15, 122 participants (3, 029 cancer/12, 093 non-cancer) from Brazil, China, and the United States was conducted, divided into one training and six validation cohorts. Plasma/serum samples were analyzed for seven protein tumor markers (AFP, CA125, CA15-3, CA19-9, CA72-4, CEA, CYFRA 21-1) using Roche, Abbott, Luminex, and ELISA platforms in both retrospective and prospective cohorts. An artificial intelligence (AI) -driven algorithm integrated biomarker concentrations with age/sex data to differentiate cancer cases from non-cancer cases and predict tissue of origin (TOO). Furthermore, the fifth validation cohort (n=1, 849 symptomatic patients: 1, 031 cancer/818 non-cancer) expanded OncoSeek's diagnostic use, targeting symptomatic individuals requiring biopsy/surgical confirmation. Results: Conventional single-threshold approaches exhibited cumulative false-positive rates with the increasing numbers of protein tumor markers (PTMs). In contrast, OncoSeek reduced false positives, achieving 93. 0% specificity (vs. 54. 3% with conventional methods) and 51. 7% sensitivity in the training cohort. Robust performance persisted across all cohorts (pooled sensitivity: 58. 4%; specificity: 92. 0%; AUC range: 0. 744–0. 912) for 14 common cancer types, with sensitivities ranging from 38. 9% to 83. 3%. Sensitivity was observed to increase with advancing clinical stages: 42. 8% at stage I, 52. 1% at stage II, 61. 9% at stage III, and 79. 7% at stage IV. TOO prediction accuracy reached 65. 4% among true positives. For symptomatic patients, OncoSeek demonstrated 73. 1% sensitivity, 90. 6% specificity, and an AUC of 0. 845 in cancer diagnosis. Conclusions: OncoSeek outperforms traditional methods, showing robust performance across ethnicities, sample types, and platforms. Retrospective validation in symptomatic population supports its clinical utility for aiding decision-making. The accuracy of TOO prediction facilitates diagnostic workup. Citation Format: Mao Mao, Shen Yong, Xia Yong, Chang yin. yin, Xing ping. ping, Li Shi. yong, Wu Wei, Zhu Rui. dan, Zhong Guo. lin, Zhu Dan. dan, Raphael Brandão, Xu Qing. xia, Ji Ling. A cost-effective, blood-based multi-cancer detection test validated in a large-scale study across diverse sample types, platforms, and populations abstract. In: Proceedings of the 18th AACR Conference on the Science of Cancer Health Disparities; 2025 Sep 18-21; Baltimore, MD. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2025;34 (9 Suppl): Abstract nr C125.