To compare the efficacy of a selective parenteral and enteral antimicrobial regimen in patients with fulminant liver failure, we classified 104 patients on reaching grade II encephalopathy as infected or non-infected. Patients who were infected were randomly assigned to receive IV cefuroxime (group 1) or selective parenteral and enteral antimicrobial regimen (group 2). Noninfected patients were randomly selected to receive either selective parenteral and enteral antimicrobial regimen (group 3) or no initial antimicrobials until clinically indicated (group 4). The four groups were comparable regarding age, sex, cause of disease, coma grade, international normalization ratio, presence of kidney failure and indicators of poor prognosis on admission to the study. Clinical parameters such as white cell count, temperature or changes in the chest radiograph, which were used to stratify patients into those infected or not, were not good predictors of infection because early infection rates were similar in the two groups. Three patients died within 24 hr and were excluded from the analysis. We found 42 microbiologically confirmed infections: group 1, 6 of 21; group 2, 8 of 21; group 3, 9 of 28; and group 4, 19 of 31. A reduction in infection was seen between groups 3 and 4 (p < 0.05). Patients receiving the selective parenteral and enteral antimicrobial regimen (groups 2 and 3) had fewer infections than the control group (group 4) (p < 0.005). Groups receiving early antimicrobial therapy (groups 1, 2 and 3) had a lower incidence of infection compared with group 4 (p < 0.0005). Overall, 55.5% survived, with no significant difference between the four groups.(ABSTRACT TRUNCATED AT 250 WORDS)