One hundred thirty-seven patients with fulminant hepatic failure were entered into two controlled trials of charcoal hemoperfusion carried out concurrently. In trial A, 75 patients with grade 3 encephalopathy were randomized to receive 5 or 10 h of hemoperfusion daily. Overall survival rates for the two groups were similar (51.3% vs. 50.0%) as was the frequency of major complications including cerebral edema and renal failure. In trial B, in which 62 patients with established grade 4 encephalopathy on admission were randomized to a no-perfusion group or to have 10 h of hemoperfusion daily, overall survival rates for the two groups were again similar (39.3% and 34.5%, respectively). There was in both trials a significant relationship between survival and etiology quite independent of the use or duration of hemoperfusion. Thus, percentage survival for the acetaminophen-overdose cases was 52.9%, for hepatitis A 66.7%, for hepatitis B 38.9%, for presumed non-A, non-B hepatitis 20%, and for halothane or drug reaction 12.5%. Within the etiologic subgroups survival was also influenced by the three major complications that developed, being inversely related to their frequency and combination, except in the non-A, non-B hepatitis and halothane or drug reaction subgroups, which had a high mortality throughout. In the latter cases particularly, orthotopic liver transplantation merits early consideration and in the group with better "intrinsic" survival (acetaminophen, hepatitis A and B) intensive management of complications (rather than charcoal hemoperfusion) would appear to be of major importance.