The aim of this study was to determine the value and limitations of 18F-fluorodeoxyglucose (FDG)-position-emission tomography (PET) for differentiating benign and malignant pancreatic disease and for staging malignant disease. One hundred fifty-nine patients with 89 malignant and 70 benign pancreatic lesions all received PET, computed tomography (CT), and endoscopic retrograde cholangiopancreatography (ERCP) before pancreatic surgery. The original reports were compared for all patients (group I; N = 159), for a subgroup that neither had fasting plasma glucose levels > or =130 mg/dL or known elevated levels of C-reactive protein ([CRP], group II; n = 123), and for the remaining patients (group III; n = 36). For group I, accuracy values (areas under receiver operating characteristic [ROC] curves) for differentiation of benign/malignant masses were 0.86 (PET), 0.93 (ERCP), 0.82 (CT), and 0.95 for ERCP + PET (N = 159). For group II, ROC areas increased to 0.92 (PET), 0.94 (p < 0.05; n = 123) (ERCP), 0.82 (CT), 0.97 (p < 0.05; n = 123) (ERCP + PET). The results for group III were 0.71 (PET), 0.81 (CT), and 0.93 (ERCP); (n = 36). With 54 patients of group II that either had contradictory or indeterminate/technically unsuccessful CT/ERCP, PET was correct in 43 patients (84%). Sensitivity/specificity for lymph node staging was 49%/63%, respectively. For patients with hepatic metastasis, PET was 70% sensitive and 95% specific, missing some metastasis that were <1 cm. PET detected peritoneal metastasis in 25% of patients, missing poorly localized microscopic spread. For selected patients who have indeterminate pancreatic masses but no hyperglycemia or serologic evidence of active inflammation, FDG-PET is an independent functional assay that significantly adds to the diagnostic accuracy of ERCP and CT in the differentiation of benign and malignant pancreatic disease. PET can reliably detect hepatic, peritoneal, and other distant metastases that are > or =1 cm.