Although some clinical studies seem to prove the efficacy of nonantialdosteronic potassium-sparing diuretics in the treatment of ascites, no controlled study has compared the efficacy of these drugs with that of antialdosteronic diuretics. Forty nonazotemic cirrhotic patients were randomized to receive amiloride (group A, n = 20) or potassium canrenoate (group B, n = 20). The initial doses of amiloride and potassium canrenoate were 20 mg and 150 mg, respectively. The doses were increased in stepwise fashion to 60 and 500 mg/day, respectively, if no response ensued. Nonresponders to the highest doses of each drug were later treated with potassium canrenoate and amiloride, respectively. Seven of 20 group A patients responded to amiloride, whereas 14 of 20 group B patients responded to potassium canrenoate (p < 0.025). Seven of 13 non-responders to amiloride later responded to potassium canrenoate, whereas only two of the nonresponders to potassium canrenoate later responded to amiloride. The diuretic responses to amiloride and potassium canrenoate were related to the activity of the reninaldosterone system. All responders to amiloride (n = 9) had normal values of plasma aldosterone. All nonresponders to amiloride who later responded to potassium canrenoate (n = 7) had increased levels of plasma aldosterone. Moreover, on comparison of all responders (n = 21) and nonresponders (n = 12) to potassium canrenoate, a higher degree of renal proximal sodium reabsorption (with consequent limitation of sodium delivery to the distal tubule) was found to be the main difference. These results indicate that (a) amiloride is less effective than potassium canrenoate in the treatment of ascites in cirrhotic patients and (b) amiloride may represent an alternative to potassium canrenoate only in patients without increased activity of the renin-aldosterone system. (Hepatology 1994;19:72–79).