A peripheral blood mononuclear cell DNA methylation-based risk stratification tool for esophageal squamous cell carcinoma

Early detection of esophageal squamous cell carcinoma is essential for improving survival, but reliance on endoscopy limits feasibility for population-wide screening. We developed ESCCseeker, a minimally invasive risk stratification tool based on peripheral blood mononuclear cells' DNA methylation. In a two-stage multicenter study in China, genome-wide profiling identifies distinct methylation markers between ESCC patients and controls. In the development and verification cohort, machine learning yields a model integrating eight differentially methylated positions with age, gender, SCC-Ag, and CYFRA21-1. ESCCseeker demonstrates robust performance (AUC: 0.987 training, 0.966 test, 0.943 verification), outperforming conventional biomarkers. In a prospective multicenter cohort, the predefined cutoff stratifies individuals into low- and high-risk groups. Restricting endoscopy to high-risk individuals reduces the number of endoscopies needed from 25.02 to 4.78, avoiding 83% of unnecessary procedures while preserving >90% detection sensitivity. These findings establish ESCCseeker as a reliable pre-endoscopic triage tool that minimizes unnecessary invasive procedures and prioritizes high-risk individuals. This trial was registered with the Chinese Clinical Trial Registry (ChiCTR2500095967).