The prognosis of acute renal failure in patients with preexisting liver decompensation is poor, and hemodialysis is considered futile, especially for hepatorenal syndrome (HRS). Since we observed a more favorable outcome in some patients, we retrospectively evaluated 107 patients with decompensated liver disease and acute renal failure (serum creatinine > 200 mumol/L) treated at the medical department of a university hospital in a 10-year period (1980-1990). HRS in the strict sense (urine-Na < 20 mmol/L while on furosemide) was diagnosed in 26 of 107 patients (24%). Renal function remained compensated in 25 patients, while 82 patients fulfilled the criteria for dialysis treatment (creatinine > 500 mumol/L and/or diuresis < 500 mL/day). In contrast to the current doctrine, 38 of the 82 patients were given hemodialysis (46%). Using the Cox proportional hazard model, the relative risk (presence vs. absence of a risk factor) of dying was increased 8.2-fold (3.9-17.2) in patients with thrombocytopenia < 100/nL, 3.9-fold (1.4-11.3) in those with hepatic encephalopathy and prothrombin time < 30%, 2.8-fold (1.6-4.8) in patients with malignoma, and 2.7-fold (1.5-4.8) in patients not submitted to dialysis despite its indication. In the CART statistics (classification and regression trees), the 33 patients with the poorest outcome were characterized exclusively by thrombocytopenia < 100/nL. HRS in the strict sense was not an independent risk factor. The CART group of 43 patients with favorable prognosis (compensated renal failure or treatment by hemodialysis, absent malignancy) had a 1-year survival rate of 38%. We conclude that thrombocytopenia, encephalopathy, and malignoma, but not HRS per se, are fatal signs that make hemodialysis futile in patients with acute renal failure and decompensated liver disease.