The successful use of orthotopic liver transplantation in fulminant hepatic failure has created a need for early prognostic indicators to select the patients most likely to benefit at a time when liver transplantation is still feasible. Univariate and multivariate analysis was performed on 588 patients with acute liver failure managed medically during 1973-1985, to identify the factors most likely to indicate a poor prognosis. In acetaminophen-induced fulminant hepatic failure, survival correlated with arterial blood pH, peak prothrombin time, and serum creatinine--a pH less than 7.30, prothrombin time greater than 100 s, and creatinine greater than 300 mumol/L indicating a poor prognosis. In patients with viral hepatitis and drug reactions three static variables [etiology (non A, non B hepatitis or drug reactions), age less than 11 and greater than 40 yr, duration of jaundice before the onset of encephalopathy greater than 7 days] and two dynamic variables (serum bilirubin greater than 300 mumol/L and prothrombin time greater than 50 s) indicated a poor prognosis. The value of these indicators in determining outcome was tested retrospectively in a further 175 patients admitted during 1986-1987, leading to the construction of models for the selection of patients for liver transplantation.