The aim of this study was to evaluate fluorine-18 fluorodeoxyglucose positron emission tomography ([18F]FDG PET) as a tool for the differential diagnosis of pancreatic carcinoma while taking into account serum glucose level. A group of 106 patients with unclear pancreatic masses were recruited for the study. PET was performed following intravenous administration of an average of 190 MBq [18F]FDG. Focally increased glucose utilisation was used as the criterion of malignancy. In addition, the "standardised uptake value" (SUV) was determined 45 min after injection. Carcinoma of the pancreas was demonstrated histologically in 74 cases, and chronic pancreatitis in 32 cases. Employing visual evaluation, 63 of the 74 (85%) pancreatic carcinomas were identified by PET. In 27 of the 32 cases (84%) of chronic pancreatitis it was possible to exclude malignancy. False-negative results (n=11) were obtained mostly in patients with raised serum glucose levels (10 out of 11), and false-positives (n=5) in patients with inflammatory processes of the pancreas. Thus PET showed an overall sensitivity of 85%, a specificity of 84%, a negative predictive value of 71%, and a positive predictive value of 93%. In a subgroup of patients with normal serum glucose levels (n=72), the results were 98%, 84%, 96% and 93%, respectively. Quantitative assessment yielded a mean SUV of 6.4+/-3.6 for pancreatic carcinoma as against a value of 3.6+/-1.7 for chronic pancreatitis (P<0.001), without increasing the diagnostic accuracy. This shows PET to be of value in assessing unclear pancreatic masses. The diagnostic accuracy of PET examinations is very dependent on serum glucose levels.