Pancreatic ductal adenocarcinoma (PDAC) is one of the deadliest cancers, with an overall five-year survival rate of 11%. Potential curative resection is possible if the tumor is detected at an early stage, with a five-year survival rate of 42%. The only current FDA-cleared biomarker for PDAC is the carbohydrate antigen 19-9 (CA19-9), which is intended for monitoring response to therapy but not for early detection. CA19-9 blood tests have varying sensitivity to detect PDAC and are prone to false positives in the presence of other underlying pancreatic conditions and to false negatives in subpopulations unable to express CA19-9. The goal of our pilot study was to determine if a multiomics approach using cell-free DNA and CA19-9 would be better than CA19-9 alone in detecting PDAC. In this retrospective study of 75 participants, we performed targeted methylation profiling of circulating cell-free DNA and quantitation of plasma CA19-9 abundance. Participants with PDAC (n=39) were 51% male with a mean age of 74.9 years, and consisted of stage II (n=9), stage III (n=11) and stage IV (n=19) pancreatic cancer. Controls (n=36) were 33% male with a mean age of 74.2 years, and included both healthy control/normal pancreas (n=17) and various benign abnormalities of the pancreas or biliary system (n=19). We developed a novel machine learning model that combines CA19-9 and methylation signals to build a joint multiomics prediction. We compared the joint predictions to those based on methylation or CA19-9 alone. Five resamplings of three-fold cross-validation were performed, and sensitivity was calculated for decision thresholds that achieved the desired test set specificity. Across all stages, the multiomics approach achieved a sensitivity of 93% at a specificity of 96%, which was greater than methylation or CA19-9 alone. At 96% specificity, methylation alone achieved a sensitivity of 74% while CA19-9 alone achieved a sensitivity of 87%. In stages II, III and IV, the multiomics approach achieved a sensitivity of 82%, 89%, and 100%, respectively and was also more sensitive than either methylation or CA19-9 alone. These proof-of-concept data demonstrate the promise of using a multiomics approach to develop a highly sensitive and specific test for the early detection of pancreatic cancer. Additional studies are underway, focusing on early-stage disease (stage I/II), to validate these results.

Citation Format: Teng-Kuei Hsu, Tzu-Yu Liu, Billie Gould, Christine Decapite, Amer Zureikat, Alessandro Paniccia, Eric Ariazi, Marvin Bertin, Richard Bourgon, Kaitlyn Coil, Hayley Donnella, Adam Drake, Julie M. Granka, Preet Kaur, Maggie C. Louie, Shivani Mahajan, Amit Pasupathy, Ofer Shapira, Peter Ulz, Chun Yang, C. Jimmy Lin, Randall Brand. Plasma-based detection of pancreatic cancer: A multiomics approach [abstract]. In: Proceedings of the AACR Virtual Special Conference on Pancreatic Cancer; 2021 Sep 29-30. Philadelphia (PA): AACR; Cancer Res 2021;81(22 Suppl):Abstract nr PO-007.