Continuous veno-venous haemofiltration is increasingly used to treat acute renal failure in critically ill patients, but a clear definition of an adequate treatment dose has not been established. We undertook a prospective randomised study of the impact different ultrafiltration doses in continuous renal replacement therapy on survival. We enrolled 425 patients, with a mean age of 61 years, in intensive care who had acute renal failure. Patients were randomly assigned ultrafiltration at 20 mL h(-1) kg(-1) (group 1, n=146), 35 mL h(-1) kg(-1) (group 2, n=139), or 45 mL h(-1) kg(-1) (group 3, n=140). The primary endpoint was survival at 15 days after stopping haemofiltration. We also assessed recovery of renal function and frequency of complications during treatment. Analysis was by intention to treat. Survival in group 1 was significantly lower than in groups 2 (p=0.0007) and 3 (p=0.0013). Survival in groups 2 and 3 did not differ significantly (p=0.87). Adjustment for possible confounding factors did not change the pattern of differences among the groups. Survivors in all groups had lower concentrations of blood urea nitrogen before continuous haemofiltration was started than non-survivors. 95%, 92%, and 90% of survivors in groups 1, 2, and 3, respectively, had full recovery of renal function. The frequency of complications was similarly low in all groups. Mortality among these critically ill patients was high, but increase in the rate of ultrafiltration improved survival significantly. We recommend that ultrafiltration should be prescribed according to patient's bodyweight and should reach at least 35 mL h(-1) kg(-1).