To investigate whether selective intestinal decontamination from oral administration of poorly absorbable antibiotics is effective in preventing bacterial infection in patients with acute liver failure, the incidence of nosocomial infection in 34 patients consecutively admitted to hospital between 1985-1990 and treated with either neomycin + colistin + nystatin or norfloxacin + nystatin (group I) was compared to the incidence of infection in 57 patients who did not receive oral, poorly absorbable antibiotics and who were consecutively admitted between 1972-1984 (group II). Patients from groups I and II had similar clinical and laboratory data at hospital admission. Twelve patients from group I and 33 from group II developed bacterial infection during the study period. The probability of infection was significantly different (p = 0.0083) in the two groups: 19% vs. 39% at the 3rd day of admission, 33% vs. 74% at the 7th day, and 48% vs. 78% at the 14th day, respectively. This difference was due to a different rate of infection from enterobacteria. Enterobacteria caused one infectious episode in group I and 24 in group II (p less than 0.001). The incidence of infections caused by other organisms, however, was similar in both groups (15 and 19 episodes, respectively). These results suggest that selective intestinal decontamination is useful in reducing the risk of infection from enterobacteria in patients with acute liver failure.