Abstract 4075: Enhanced early detection of pancreatic cancer using a multi-analyte liquid biopsy approach

Abstract Background: Pancreatic ductal adenocarcinoma (PDAC) has a 5-year survival rate of less than 12% due to late diagnosis when curative options are limited. Early detection, especially stage I-II disease, could improve survival, but current non-invasive tests lack enough sensitivity. Multi-analyte liquid biopsies that combine genomic, epigenomic, and glycan data show promise for better detection. We developed a prediction model for the early detection of pancreatic cancer by integrating epigenomic and genomic features with haplotype information and CA19-9 biomarker levels, achieved a markedly more sensitive predictor. Methods: We employed a training set of 162 pancreatic cancer patients and 983 non-cancer patients. Cell-free DNA was isolated and employed to perform 5-hydroxymethylcytosine (5hmC) profiling, low pass whole-genome sequencing (WGS) and genotyping. CA19-9 biomarker levels were also measured in plasma. The resulting 5hmC gene-based changes, fragment size differences, and genotyping coupled with CA19-9 levels, were used for logistic regression model building. The prediction model performance was validated in an independent validation cohort of 1,445 individuals, consisting of 259 pancreatic cancers and 1,186 non cancer with several high-risk features including diabetes, family history, genetic predisposition and heavy smokers. Cancer performance was measured through sensitivity and specificity metrics and 95% confidence intervals (CIs). Results: After training the prediction model to optimize feature selection, the final model was locked at a specificity threshold of 97.75% and used to score an independent clinical cohort in a blinded manner. In the independent validation cohort, the multi-analyte model achieved an overall sensitivity of 82.6% (95% CI: 77.45%-87.04%) and importantly, early-stage sensitivity (n=138) was 76.8% (95% CI: 68.87%-83.57%). These measures point to an increase of sensitivity of 8-15% compared to an earlier version of the assay. Test specificity was 97.45% (95% CI: 96.41%-98.29%), consistent with the pre-established 97.75% training specificity, highlighting robust performance across datasets. Conclusions: The multi-analyte model demonstrated strong early-stage PDAC detection while maintaining high specificity. By integrating orthogonal biological features such as epigenomics, genomics and CA19-9 levels, we have been able to improve PDAC detection performance, ensuring that the breadth of oncogenic development can be detected through an early detection blood test. Citation Format: Anna Bergamaschi, David Haan, Verena Friedl, Glenn Oliviera, Michael Cipriano, Tierney Phillips, Yuan Xue, Yuhong Ning, Micah Collins, Michael Kesling, Michael Riviere, Nicky Nguen, Vanessa Lopez, Anna Leighton, Roger Malta, Maryam Nabiyouni, Carolina Fraire, Gulfem Guler, Shivani Dhillon, Ceyda (Jada) Coruh, Melissa Peters, Shimul Chowdhury, Eric Nilson, Stephen Quake, Wayne Volkmuth, Samuel Levy. Enhanced early detection of pancreatic cancer using a multi-analyte liquid biopsy approach abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4075.