Biomarkers for early diagnosis of patients with pancreatic cancer (PC) are needed. Our aim was to identify panels of miRNAs in serum in combination with CA 19-9 for use in the diagnosis of PC. Four hundred seventeen patients with PC were included prospectively from Denmark (n = 306) and Germany (n = 111). Controls included 59 patients with chronic pancreatitis (CP) and 248 healthy subjects (HS). MiRNAs were analyzed in pretreatment serum samples from 3 cohorts: discovery cohort (754 human miRNAs, TaqMan® Human MicroRNA assay, Applied Biosystem; PC n = 133, controls n = 72); training cohort (34 miRNAs, real-time qPCR using the Fluidigm BioMark™ System; PC n = 198, controls n = 184); validation cohort (13 miRNAs, real-time qPCR using the Fluidigm BioMark™ System; PC n = 86, controls n = 51). We found that 34 miRNAs in serum from PC patients in the discovery cohort were expressed differently than in controls. These miRNAs were tested in the training cohort, and four diagnostic panels were constructed that included 5 or 12 miRNAs (miR-16, −18a, −20a, −24, −25, −27a, −29c, −30a.5p, −191, −323.3p, −345 and −483.5p). Diagnostic accuracy of detecting PC in the training cohort was AUC (Index I 0.85; II 0.87; III 0.85; IV 0.95; CA 19-9 0.93); specificity (I 0.71; II 0.76; III 0.66; IV 0.90 (fixed sensitivity at 0.85); CA 19-9 0.93). Combining serum CA 19-9 and Index II best discriminated Stages I and II PC from HS [AUC 0.93 (0.90–0.96), sensitivity 0.77 (0.69–0.84), specificity 0.94 (0.90–0.96) and accuracy 0.88 (0.84–0.91)]. In conclusion, we identified four diagnostic panels based on 5 or 12 miRNAs in serum that could distinguish patients with PC from HS and CP.