Clinical Validation of a Non-Invasive Multi-Omics Method for Multi-Cancer Early Detection in Retrospective and Prospective Cohorts

Recent studies highlight the promise of blood-based multi-cancer early detection (MCED) tests for identifying asymptomatic cancer patients. However, most focus on a single cancer hallmark thus limiting effectiveness due to cancer's heterogeneity. Here, a blood-based multi-omics test named SeekInCare for MCED is reported. SeekInCare incorporates multiple genomic and epigenetic hallmarks, including copy number aberration, fragment size, end motif, and oncogenic virus, via shallow whole-genome sequencing from cell-free DNA, alongside seven protein tumor markers in one tube of blood. Artificial intelligence algorithms were developed to distinguish cancer patients from non-cancer individuals and to predict the likely affected organ. The retrospective study included 617 cancer patients and 580 non-cancer individuals, covering 27 cancer types. SeekInCare achieved 60.0% sensitivity at 98.3% specificity, resulting in an AUC of 0.899. Sensitivities were 37.7%, 50.4%, 66.7%, and 78.1% in stage I, II, III, and IV patients, respectively. Additionally, SeekInCare was evaluated in a prospective cohort consisting of 1203 individuals who received the test as a laboratory-developed test (median follow-up time: 753 days) in which it achieved 70.0% sensitivity at 95.2% specificity. The performances of SeekInCare in both retrospective and prospective studies demonstrate that SeekInCare is a blood-based MCED test, showing comparable performance to the other tests currently in development. These findings support its potential clinical utility as a cancer screening test in high-risk populations.