Multi-cohort validation of a multi-analyte liquid biopsy test for early-stage pancreatic cancer detection.

4139 Background: Pancreatic ductal adenocarcinoma (PDAC) has a 5-year survival rate below 12%, largely due to diagnosis at advanced, noncurative stages. Earlier detection could significantly improve survival outcomes; however, current approaches, including imaging and blood based assays lack sufficient sensitivity and specificity. Liquid biopsy methods that combine genomic, epigenomic, and glycan-based biomarkers may improve the detection accuracy by integrating a multi-analyte approach. We developed an improved prediction model for the Avantect Pancreatic Cancer Test (Avantect) using 5-hydroxymethylcytosine (5hmC) profiling, whole-genome based fragmentomics, together with genotyping and haplotype data, and CA19-9 biomarker levels. Methods: We employed a training cohort consisting of 162 PDAC cases and 983 noncancer controls. Cell-free DNA was analyzed using 5hmC profiling, low-pass whole-genome sequencing (WGS), and genotyping, alongside matched plasma CA19-9 measurements. A logistic regression model integrating 5hmC features, fragment size metrics, haplotype information, and CA19-9 levels was constructed and locked at a specificity of 97.75%. Performance was evaluated in two independent validation cohorts consisting of 1,445 individuals (259 PDAC; 1,186 noncancer participants) and 173 individuals (67 PDAC and 106 non-cancers). Sensitivity, specificity, and 95% confidence intervals (CIs) were computed. Results: In an independent validation cohort of 1,445 individuals with various high-risk features including type 2 diabetes, family history, and genetic predisposition, Avantect achieved an overall sensitivity of 82.6% (95% CI: 77.45%-87.04%) and an early stage (stage I-II) sensitivity of 76.8% (n=138; 95% CI: 68.87%-83.57%). A second validation cohort of 173 individuals, enriched for new onset type 2 diabetes, was evaluated and showed a sensitivity of 74.6% (95% CI: 62.51%-84.47%). Specificity in both cohorts remained high at 97.5% (95% CI: 96.41%-98.29%) and 98.1% (95% CI: 93.33%-99.77%) respectively, consistent with the pre-specified rate of 97.75%. Test specificity was further evaluated in a cohort comprising other cancer types, including breast, colorectal, lung, liver, prostate, ovarian, bladder, and kidney cancers, revealing a consistent high specificity of 97.70 (0.05% reduction). Conclusions: The multi-analyte model shows strong and robust performance across multiple independent cohorts detecting PDAC at high specificity. By integrating orthogonal biological signals, through the incorporation of epigenomic, genomic, and glycan biomarkers, the Avantect Pancreatic Cancer Test showed an improved PDAC detection to provide a tool that could significantly improve survival for patients with pancreatic cancer.