A randomized, double-blind, placebo-controlled trial of selective decontamination of the oropharynx and gastrointestinal tract was conducted on 61 intubated patients in a medical-surgical intensive care unit (ICU) to determine the impact on nosocomial pneumonia, other infections, and emergence of colonization or infection with antibiotic-resistant bacteria. Over 8 months, 30 patients received an oral paste and solution containing polymyxin, gentamicin, and nystatin; 31 patients received a placebo paste and solution. At study entry, patients in both groups were seriously ill (mean acute physiologic score, 27.2), frequently had pulmonary infiltrates (73.8%), and were likely to be receiving systemic antibiotics (86.9%). There were no differences between study patients and control patients in these characteristics or in frequency of any nosocomial infection (50% vs. 55%), nosocomial pneumonia (27% vs. 26%), febrile days (2.3 vs. 2.0), duration of antibiotic therapy (14.0 vs. 13.4), or mortality rates (37% vs. 48%). There was no difference in infections caused by antibiotic-resistant gram-negative bacilli, although a trend towards more frequent infection with gentamicin-resistant enterococci was found for study patients. Selective decontamination did not appear to be effective in our very ill medical-surgical ICU patients, although the number of patients in our trial was sufficient to detect only a 50% or greater reduction in pneumonia rates.