The role of selective decontamination of the digestive tract (SDD) for the prevention of nosocomial infection in critically ill patients remains controversial, and the efficacy of this technique in patients who are already infected on presentation to the intensive care unit has not previously been assessed. We performed a double-blind randomized placebo controlled trial of SDD (parenteral cefotaxime, six-hourly oral and enteral polymyxin E, tobramycin, and amphotericin B vs placebo) for all infected patients presenting to the ICU requiring mechanical ventilation for more than 48 hours and ICU stay of more than 5 days. Daily clinical and microbiological monitoring for secondary infection was undertaken until hospital discharge. In all, 59 selective decontamination and 76 placebo fully comparable patients fulfilled criteria for enrollment and analysis (APACHE II 15.2 vs 15.1). The number of patients receiving SDD who developed nosocomial infections was significantly reduced (P = 0.048), and there were no infections caused by the enterobacteriaceae or Candida spp in this group. No difference in ICU (17.5 vs 18.8 days) or hospital stay (32.7 vs 34.2 days) or mortality (17% vs 22.3%) was shown. Critically ill, primarily infected patients are protected from nosocomial infection by the use of SDD.