Safe and consistently reliable parenteral hyperalimentation in the care of the critically ill or injured remains an unrealized expectation. Our experience with critically ill patients managed with standard 25% dextrose/4.25% protein hyperalimentation solution (S.H.A.) delivered through a centrally placed catheter demonstrated that S.H.A. was associated with a high catheter infection rate, fluctuations in hyperosmolar tolerance, unstable insulin requirements, and a high discontinuation rate. In retrospect this unsatisfactory experience resulted from the inappropriate assumption that a single type of nutritional support and delivery system could meet the varying metabolic requirements of the critically ill or injured. In an effort to reduce these complications, we have recently utilized a system of selective hyperalimentation in managing 25 consecutive patients. We now identify specific risk factors for groups of patients and use these factors to select the appropriate route and solution. Selective hyperalimentation has resulted in a lowering of infusion complications and an improvement in completion rate, with satisfactory weight gain and protein response. These results suggest that our approach is successful in meeting the nutritional requirements of differing groups of critically ill patients without exposing unsuitable candidates to the potential risks of standard hyperalimentation.