The efficacy of a relatively cheap regimen of selective decontamination (SDD) was evaluated in a diverse population of ICU patients. Patients requiring prolonged ICU stay (greater than 5 days) were randomly allocated to a treatment group or control group. Control patients (n = 52) received perioperative antimicrobial prophylaxis and antibiotic treatment was instituted only on sound clinical and bacteriological criteria. Treated patients (n = 48) received gastro-intestinal and oro-pharyngeal decontamination with polymyxin E, norfloxacin, amphotericin B and systemic antibiotic prophylaxis with trimethoprim until decontamination was achieved. The rate of gram-positive infections was not altered by SDD. The incidence of gram-negative respiratory tract, urinary tract and line infections was significantly reduced from 44%, 27% and 15% respectively in the control group to 6%, 4% and 0% in the treatment group. Mortality from nosocomial sepsis and overall mortality were also significantly reduced from 15% and 54% to 0% and 31% respectively. The ICU stay was not reduced by SDD, nor was time on the ventilator or use of therapeutic antibiotics. The reduction in morbidity and mortality was achieved at a relatively low cost.