Hinestrosa, J.P., Sears, R.C., Dhani, H. et al. https://doi.org/10.1038/s43856-023-00351-4
Abstract
Background
Pancreatic ductal adenocarcinoma (PDAC) has an overall 5-year survival rate of just 12.5% and thus is among the leading causes of cancer deaths. When detected at early stages, PDAC survival rates improve substantially. Testing high-risk patients can increase early-stage cancer detection; however, currently available liquid biopsy approaches lack high sensitivity and may not be easily accessible.
Methods
Extracellular vesicles (EVs) were isolated from blood plasma that was collected from a training set of 650 patients (105 PDAC stages I and II, 545 controls). EV proteins were analyzed using a machine learning approach to determine which were the most informative to develop a classifier for early-stage PDAC. The classifier was tested on a validation cohort of 113 patients (30 PDAC stages I and II, 83 controls).
Results
The training set demonstrates an AUC of 0.971 (95% CI = 0.953–0.986) with 93.3% sensitivity (95% CI: 86.9–96.7) at 91.0% specificity (95% CI: 88.3–93.1). The trained classifier is validated using an independent cohort (30 stage I and II cases, 83 controls) and achieves a sensitivity of 90.0% and a specificity of 92.8%.
Conclusions
Liquid biopsy using EVs may provide unique or complementary information that improves early PDAC and other cancer detection. EV protein determinations herein demonstrate that the AC Electrokinetics (ACE) method of EV enrichment provides early-stage detection of cancer distinct from normal or pancreatitis controls.
