A double blind, placebo-controlled trial was performed to test the efficacy of prevention of nosocomial infections by selective digestive decontamination. Placebo or tobramycin (80 mg) and colistin (100 mg) was given four times daily via the gastric tube. Amphotericin B (500 mg/6 h) was administered to all patients. As our ICU is divided into two separate subunits, intestinal decontamination or placebo was administered alternatively to patients of the two subunits during two 3-month periods, separated by a 2-month period without prevention. The decontamination (n = 97) and placebo groups (n = 84) were similar with respect to age, sex, severity score and diagnostic categories on admission. Intestinal decontamination alone failed to significantly reduce the number of infected patients (26% vs 34.5%, p = 0.20), but was effective on ICU-acquired infections (0.33 vs 0.60, p = 0.02) especially gram-negative infection rates (0.17 vs 0.43, p = 0.01). The onset of the first ICU-acquired infection was delayed (9 vs 13 days, p less than 0.001) and incidence of pneumonia (2 vs 13 cases, p less than 0.01) including bacterial pneumonia (0 vs 8 cases, p less than 0.01) was significantly decreased. However, mean ICU stay and mortality were not significantly modified by intestinal decontamination.