We compared the diagnostic utility of DU-PAN-2 and CAR-3 with that of CA 19-9 in differentiating pancreatic cancer (23 patients) from chronic pancreatitis (16 patients) and various extra-pancreatic diseases (28 patients) mainly of the upper gastrointestinal and biliary tract. The influence of some pathophysiologic variables on the three markers was also assessed. The sensitivities of the three markers in detecting pancreatic cancer were: CA 19-9, 83%; DU-PAN-2, 56%; and CAR-3, 39%. In patients with chronic pancreatitis and extra-pancreatic diseases, CA 19-9 gave the highest number of false positives. Receiver-operating characteristic curves showed that the ability of CAR-3 to discriminate between pancreatic cancer and other diseases was similar to that of CA 19-9, whereas DU-PAN-2 was a less reliable discriminator. Correlations were found between the behavior of all three markers and that of the cholestasis indices (ALP and GGT). Our findings indicate that DU-PAN-2 and CAR-3 serum determinations do not provide any more information than does CA 19-9 alone. The latter remains the marker of choice in the differential diagnosis of pancreatic cancer, even though it cannot be considered a definitive aid. Serum levels of all three markers increase in the presence of extra-hepatic cholestasis, possibly due to interference with the hepatic clearance of glycoproteins and destruction of ductal biliary epithelium.