We compared the clinical efficacy and safety of large-volume paracentesis and dialytic ultrafiltration in the treatment of refractory ascites in cirrhotic patients. A group of cirrhotic subjects (age 49–80 years) were randomly allocated to either continuous paracentesis (1–1.5 l/hour) or dialytic ultrafiltration until disappearance of ascites. Each patient was maintained on bed rest, fluid restriction (1 l/day) and a low (25 mmol/day) sodium diet for 14 days. Five patients (three in the paracentesis group and two in dialytic ultrafiltration group) developed massive ascites 3–5 months later, and received the crossover treatment. The average volume of fluid removed was similar in the two groups (4.70 ± 1.47 l for dialytic ultrafiltration versus 4.69 ± 1.84 l for paracentesis), but the treatment period was significantly shorter with dialytic ultrafiltration. The plasma creatinine significantly increased three days after paracentesis but did not increase in patients treated with dialytic ultrafiltration. There was an initial fall in mean arterial pressure during the first two hours of either treatment; a further fall in blood pressure was observed with paracentesis but not with dialytic ultrafiltration. Pretreatment plasma renin activity was elevated, but was not altered by either treatment. Plasma atrial natriuretic peptide levels were in the high-normal range before treatment. Paracentesis was associated with a delayed fall in plasma atrial natriuretic peptide, while dialytic ultrafiltration induced a modest but significant rise. No complication was experienced with dialytic ultrafiltration in the two weeks following treatment, but four of the eight patients who underwent paracentesis had developed severe complications. Dialytic ultrafiltration of ascitic fluid is a safe procedure in cirrhotic patients. Large-volume paracentesis without intravenous colloid reinfusion causes complications and carries the potential risk of reducing the effective intravascular volume.