To determine the clinical value of tumour markers in the diagnosis of malignancy-related ascites (not including hepatocellular carcinoma), serum and ascitic fluid levels of carcinoembryonic antigen, cancer antigen 125, carbohydrate antigen 19-9, tissue polypeptide antigen and serum-ascites albumin gradient were determined in 66 patients with cirrhotic ascites, 28 patients with hepatocellular carcinoma and ascites, and 29 patients with malignancy-related ascites. Three tumour markers and serum-ascites albumin gradient showed significant difference between patients with malignancy-related ascites and those without: serum carcinoembryonic antigen (26.4 +/- 31.5 vs 4.8 +/- 4.6 ng/mL, P < 0.01), ascitic fluid carcinoembryonic antigen (118.4 +/- 196.5 vs 2.0 +/- 1.4 ng/mL, P < 0.01), ascitic fluid carbohydrate antigen 19-9 (12,933 +/- 25,496 vs 23 +/- 67 U/mL, P < 0.01) and serum-ascites albumin gradient (1.1 +/- 0.4 vs 2.0 +/- 0.4 g/dL, P < 0.01). At the best cut-off levels chosen from near 95% of the data in those without malignancy-related ascites, the sensitivity, specificity and accuracy to diagnose malignancy-related ascites were, respectively, 65.5%, 93.6%, 87.0% using serum carcinoembryonic antigen > or = 10 ng/mL; 69.0%, 94.7%, 88.6% using ascitic fluid carcinoembryonic antigen > or = 5 ng/mL; 65.5%, 93.6%, 87.0% using ascitic fluid carbohydrate antigen 19-9 > or = 50 U/mL; 62.1%, 98.9%, 90.2% using serum-ascites albumin gradient < 1.1 g/dL. Although serum-ascites albumin gradient offered the best diagnostic accuracy and specificity, its sensitivity was not good enough. Our study indicates that serum-ascites albumin gradient and tumour markers are not sensitive parameters in the diagnosis of malignancy-related ascites.