Performance of a blood-based screening test for the early detection of colorectal cancer.

Abstract

e15627 Background: While the clinical benefit of highly sensitive screening tools for colorectal cancer (CRC) has been established, adherence rates remain low. The convenience and familiarity of blood-based testing may reduce disease burden through greater access and compliance to regular and recommended asymptomatic screening. We report on preliminary findings from a minimally invasive blood-based early cancer detection (ECD) screening test for CRC. Methods: An assay to interrogate differentially methylated genomic regions was developed and used in this study. Using an independent cohort of plasma samples, a machine-learning model was trained to classify patient samples into cancer-free and CRC based on observed differential methylation patterns in circulating cell-free DNA (cfDNA). A test set of plasma samples from 432 individuals was used to evaluate performance. Healthy controls (N = 305) were sourced from biobanks, including the ongoing PROCEED observational study (NCT06620627). CRC patients (N = 127) were sourced from the CIRCULATE-Japan GALAXY observational study (UMIN000039205). The assay performance was calculated based on concordance with clinical diagnosis (CRC, cancer-free). Stage-adjusted sensitivity was estimated based on stage distributions that are expected to be representative of the final pivotal study. Results: Among the 127 CRC cases, 47% had stage I/II disease, mean age was 65±12 years, and 46% were male. Among the 305 healthy individuals, mean age was 56±10 years and 42% were male. Overall sensitivity was 95% (95% CI: 92-99%), and specificity was 91% (95% CI: 88-94%). Overall stage-adjusted sensitivity estimates remained at 95%. Stage I sensitivity was 92% overall. Screen-detected sensitivity, adjusted for stage, was 91%. Sensitivity was 100% (n = 14/14) in microsatellite-high tumors and 95% (n = 107/113) in microsatellite-stable tumors. Multivariate analysis demonstrated associations between methylation signal and tumor size, cancer stage, and sex. Compared with a tumor-informed, variant-based circulating tumor DNA (ctDNA) test (SignateraTM), we observed a PPA (positive percent agreement) of 100%. Moreover, the differentially methylated allele fraction (DMAF) across target regions strongly correlated with variant allele frequencies (VAFs) from the tumor-informed test (Spearman’s correlation coefficient, 0.85). Conclusions: Using epigenetic markers specific for CRC, we developed a sensitive and specific blood-based test to detect cfDNA from early-stage and asymptomatic disease. The methylation signal strongly correlated with ctDNA VAFs from a clinically validated tumor-informed test and increased with cancer stage and disease burden. This test could serve as a means to increase patient adherence rates for CRC screening, thereby improving patient outcomes.