Importance: Biomarkers for the early diagnosis of patients with pancreatic cancer are needed to improve prognosis.

Objectives: To describe differences in microRNA expression in whole blood between patients with pancreatic cancer, chronic pancreatitis, and healthy participants and to identify panels of microRNAs for use in diagnosis of pancreatic cancer compared with the cancer antigen 19-9 (CA19-9).

Design, setting, and participants: A case-control study that included 409 patients with pancreatic cancer and 25 with chronic pancreatitis who had been included prospectively in the Danish BIOPAC (Biomarkers in Patients with Pancreatic Cancer) study (July 2008-October 2012) plus 312 blood donors as healthy participants. The microRNA expressions in pretreatment whole blood RNA samples were collected and analyzed in 3 randomly determined subcohorts: discovery cohort (143 patients with pancreatic cancer, 18 patients with chronic pancreatitis, and 69 healthy participants), training cohort (180 patients with pancreatic cancer, 1 patient with chronic pancreatitis, and 199 healthy participants), and validation cohort (86 patients with pancreatic cancer, 7 patients with chronic pancreatitis, and 44 healthy participants); 754 microRNAs were screened in the discovery cohort and 38 microRNAs in the training cohort and 13 microRNAs in the validation cohort.

Main outcomes and measures: Identification of microRNA panels (classifiers) for diagnosing pancreatic cancer.

Results: The discovery cohort demonstrated that 38 microRNAs in whole blood were significantly dysregulated in patients with pancreatic cancer compared with controls. These microRNAs were tested in the training cohort and 2 diagnostic panels were constructed comprising 4 microRNAs in index I (miR-145, miR-150, miR-223, miR-636) and 10 in index II (miR-26b, miR-34a, miR-122, miR-126*, miR-145, miR-150, miR-223, miR-505, miR-636, miR-885.5p). The test characteristics for the training cohort were index I area under the curve (AUC) of 0.86 (95% CI, 0.82-0.90), sensitivity of 0.85 (95% CI, 0.79-0.90), and specificity of 0.64 (95% CI, 0.57-0.71); index II AUC of 0.93 (95% CI, 0.90-0.96), sensitivity of 0.85 (95% CI, 0.79-0.90), and specificity of 0.85 (95% CI, 0.80-0.85); and CA19-9 AUC of 0.90 (95% CI, 0.87-0.94), sensitivity of 0.86 (95% CI, 0.80-0.90), and specificity of 0.99 (95% CI, 0.96-1.00). Performances were strengthened in the validation cohort by combining panels and CA19-9 (index I AUC of 0.94 [95% CI, 0.90-0.98] and index II AUC of 0.93 [95% CI, 0.89-0.97]). Compared with CA19-9 alone, the AUC for the combination of index I and CA19-9 was significantly higher (P = .01). The performance of the panels in patients with stage IA-IIB pancreatic cancer was index I AUC of 0.80 (95% CI, 0.73-0.87); index I and CA19-9 AUC of 0.83 (95% CI, 0.76-0.90); index II AUC of 0.91 (95% CI, 0.87-0.94); and index II and CA19-9 AUC of 0.91 (95% CI, 0.86-0.95).

Conclusions and relevance: This study identified 2 diagnostic panels based on microRNA expression in whole blood with the potential to distinguish patients with pancreatic cancer from healthy controls. Further research is necessary to understand whether these have clinical implications for early detection of pancreatic cancer and how much this information adds to serum CA19-9.