To avoid any negative outcomes associated with under- or overfeeding it is essential to estimate nutrient requirements before commencing nutrition support. The energy requirements of an individual vary with current and past nutritional status, clinical condition, physical activity and the goals and likely duration of treatment. The evidence-base for prediction methods in current use, however, is poor and the equations are thus open to misinterpretation. In addition, most methods require an accurate measurement of current weight, which is problematic in some clinical situations. The estimation of energy requirements is so challenging in some conditions, e.g. critical illness, obesity and liver disease, that it is recommended that expenditure be measured on an individual basis by indirect calorimetry. Not only is this technique relatively expensive, but in the clinical setting there are several obstacles that may complicate, and thus affect the accuracy of, any such measurements. A review of relevant disease-specific literature may assist in the determination of energy requirements for some patient groups, but the energy requirements for a number of clinical conditions have yet to be established. Regardless of the method used, estimated energy requirements should be interpreted with care and only used as a starting point. Practitioners should regularly review the patient and reassess requirements to take account of any major changes in clinical condition, nutritional status, activity level and goals of treatment. There is a need for large randomised controlled trials that compare the effects of different levels of feeding on clinical outcomes in different disease states and care settings.