Development of a serum-based miRNA signature for early detection of pancreatic cancer: a multicenter cohort study

Abstract

Abstract Background A grim prognosis of pancreatic cancer (PCa) was attributed to the difficulty in early diagnosis of the disease. Identifying novel biomarkers for early detection of PCa is thus urgent to improve the overall survival rates of patients. Methods The study was performed firstly by identification of candidate microRNAs (miRNAs) in formalin-fixed, paraffin-embedded tissues at either early (n = 100) or advanced (n = 100) stages, to that in benign tissues (n = 100) using microarray profiles, and followed by validation in a serum-based cohort study to assess clinical utility of the candidates as a noninvasive biomarker. In the cohorts, a total of 1273 participants including 571 patients with pancreatic ductal adenocarcinoma, 90 patients with chronic pancreatitis, 217 patients with other pancreatic diseases, and 395 healthy controls from four centers were retrospectively recruited as two cohorts including training and validation cohort. The collected serum specimens were analyzed by real-time polymerase chain reaction. Results We identified 27 miRNAs that were expressed differentially in both early and advanced stages of PCa tissues as compared to the benign. Of which, the top-four was selected by the criteria of log2(fold change) > 4 and FDR < 0.05 as a panel whose diagnostic efficacy was fully assessed in the serum-based cohorts. Patients with PCa at early-stage were significantly discriminated from healthy controls by the panel with AUCs of 0.971 (95%CI: 0.956–0.987) and 0.933 (95%CI: 0.892–0.974) in the training and validation cohorts, respectively. Furthermore, the panel distinguished early-stage PCa from non-PCa including chronic pancreatitis as well as pancreatic cystic neoplasms with AUCs of 0.924 (95%CI: 0.899–0.949) and 0.861 (95%CI: 0.818–0.903) in the training and validation cohorts, respectively. Moreover, the panel eliminated interference from other digestive tumors with a specificity of 90.2%. Strikingly, this panel exhibited superior to four biomarkers routinely used in clinic, including CA19-9, CA125, CEA and CA242. Conclusions A serum-based panel of four miRNAs was developed showing remarkably discriminative ability of early-stage PCa from either healthy controls or other pancreatic diseases, suggesting it may be developed as a novel, noninvasive approach for early screening of PCa in clinic.