Abstract
Pancreatic cancer (PaC), which is characterized by a high mortality rate, is often diagnosed at an advanced stage, significantly limiting treatment effectiveness. Early detection is crucial for improving survival rates, especially for individuals at high risk (HR) for PaC. Traditional diagnostic methods, including ultrasound, computed tomography, and magnetic resonance imaging (MRI), have limited sensitivity, especially for detecting early-stage PaC. We explored the potential of miRNA from urinary extracellular vesicles (EVs) as a noninvasive diagnostic marker for PaC. An exploratory case-control study was conducted across multiple Japanese institutions. The study included 248 samples from patients with pancreatic ductal adenocarcinoma (PDAC), the most common type of PaC, and HR patients. Differential expression analysis revealed significant differences in 16 miRNAs between the PDAC and HR samples. A machine learning-based algorithm was developed based on these miRNAs to distinguish between PDAC and HR. The algorithm exhibited an AUC of 0.89, a sensitivity of 0.80, and a specificity of 0.79. The algorithm detected the early-stage PDAC (stage 0-IIA) with a sensitivity of 0.73. These findings highlight the potential of the urinary miRNA algorithm as a noninvasive tool to aid in the detection of PDAC, including early-stage cases, in high-risk populations.