The decision to initiate total parenteral nutrition (TPN) in hospitalised patients should be based on the presence of clinically significant starvation and dysfunction of the gastrointestinal tract. It must also take into account the clinical status of the patient, considering major treatment strategies and the need for prolonged hospitalisation, the benefits of feeding and the attendant risks of central venous alimentation. Recent evidence in surgical patients in intensive care provides the impetus for early parenteral feeding; withholding TPN and inducing a cumulative caloric deficit of ≥ 10 000 calories has been associated with a survival disadvantage compared to those patients with a positive caloric balance. Moreover, the incidence of serious organ failure was consistently higher in the group with cumulative caloric deficits. Additional evidence favouring the provision of TPN exists, but the axiom ‘if the gut works, use it’ still prevails. Exceptions to this precept do exist, however, particularly in critically ill patients. The metabolic derangements encountered in these patients could be so severe that it may be impossible to correct the electrolyte and acid-base abnormalities via the enteral route. For example, such patients may have large potassium requirements and/or severe alkalaemia necessitating systemic acidification with hydrochloric acid, precluding enteral delivery due to gastrointestinal intolerance. In this setting, combined enteral feeding with 10 to 20 ml/h to maintain gut integrity (via a post-pyloric feeding tube) and TPN during the acute phases of illness is an exciting possibility.Once the decision to feed is made, the amount of nutrition prescribed may assume equal importance with respect to patient outcome. The frequent use of the Harris-Benedict equation, plus a multiplying factor for stress, may overestimate caloric requirements; this is particularly true during critical illness. The dangers of overfeeding may be just as harmful as not feeding at all. The use of indirect calorimetry provides the most accurate measurement of resting energy expenditure. However, in the absence of indirect calorimetry, modified equations to estimate caloric needs are available. Caution must be observed as caloric intakes exceeding the range of 25 to 35 kcal/kg may be dangerous, particularly in the severely ill patient with preexisting organ failure.The amount of protein and the ‘calorie-mix’ necessary for optimal nutritional support is open to debate. Recent evidence has demonstrated no additional benefit to nitrogen balance in severely septic patients when protein was given at a level exceeding 1.5 g/kg/day. Similarly, protein intakes > 1.75 g/kg/day in patients with advanced gastrointestinal cancer did not achieve a state of net protein synthesis. Therefore, for most patients with moderate to severe degrees of stress, a level of protein intake up to 1.75 g/kg/day is reasonable, since levels above this offer no additional benefits and most likely lead to ureagenesis.Carbohydrate intake is important for a number of vital physiological functions, and it is an essential macronutrient. However, excessive glucose administration is associated with a number of adverse effects. In general, glucose infusion rates should not exceed 4 mg/kg/min for 2 reasons: first, this rate is equal to its optimal infusion rate, and dosages above this level increase the rate of lipogenesis; and second, even doubling its optimal infusion rate has not been shown to improve protein-sparing in severely stressed patients. Hence, the carbohydrate content should generally not exceed 4 mg/kg/min (i.e. approximately 400g in the reference 70kg man) with the balance of the calories provided as lipids. When given as long chain triglycerides (LCT), lipids should preferably be given continuously as a 3-in-1 or total nutrient admixture.Finally, we recommend that nutrition support teams take an aggressive approach to the management of severely ill patients. Manipulations that reduce volume burdens, such as concentrating all separate (i.e. piggyback) infusions, as well as using the TPN as a drug vehicle (where appropriate), will afford greater likelihood of providing the necessary protein and calories. Furthermore, managing the plethora of metabolic derangements frequently encountered in the intensive care unit (i.e. acid-base and electrolyte disturbances) via the TPN should reduce the dangers of protracted imbalances. It is obvious that professional training of pharmacists and physicians specialising in clinical nutrition is needed to achieve optimal care of these patients.