OBJECTIVES: We measured the energy and protein needs in 50 sequential, critically ill, ventilated patients requiring continuous renal replacement therapy (CRRT) for renal failure by using indirect calorimetry and three sequential isocaloric protein-feeding regimes of 1.5, 2.0, and 2.5 g. kg(-1). d(-1). We also assessed the compliance of actual feeding with target feeding and correlated the predictive energy requirements of the formulae with the actual energy expenditure (EE) measured by indirect calorimetry. We also determined whether these feeding regimes affected patient outcome. METHODS: The energy and protein needs of 50 consecutive, critically ill patients (31 male; age 53.3 +/- 17.4 y; Acute Physiology and Chronic Health Evaluation (APACHE II) score: 26.0 +/- 8.0; Acute Physiology and Chronic Health Evaluation score predicted risk of death: 50.0 +/- 25.0%) were assessed by using indirect calorimetry and ultrafiltrate nitrogen loss. Entry into this study was on commencement of CRRT. To eliminate any beneficial effect from the passage of time on nitrogen balance, 10 of the 50 patients were randomized to receive 2.0 g. kg(-1). d(-1) throughout the study, and the others received an escalating isocaloric feeding regime (1.5, 2.0, and 2.5 g. kg(-1). d(-1)) at 48-h intervals. Enteral feeding was preferred, but if this was not tolerated or unable to meet target, it was supplemented or replaced by a continuous infusion of total parenteral nutrition. Energy was given to meet caloric requirements as predicted by the Schofield equation corrected by stress factors or based on the metabolic cart readings of EE and was kept constant for all patients throughout the trial. Patients were stabilized on each feeding regime for at least 24 h before samples of dialysate were taken for nitrogen analysis at 8-h intervals on the second day. CRRT was performed by using a blood pump with a blood flow of 100 to 175 mL/min. Dialysate was pumped in and out counter-currently to the blood flow at 2 L/h. A biocompatible polyacrylonitrile hemofilter was used in all cases. RESULTS: EE was 2153 +/- 380 cal/d and increased by 56 +/- 24 cal/d (P < 0.0001) throughout the 6-d study period to 2431 +/- 498 cal/d. At study entry, the mean predicted (Schofield) caloric requirement was 2101 +/- 410. Patients received 99% of the predicted energy requirements. However, the mean EE was 11% higher at 2336 +/- 482 calories. This difference was not uniform. If the predicted caloric requirement was less than 2500, the EE exceeded the predicted by an average of 19%. If the predicted caloric requirement was greater than 2500, the EE on average was 6% less than predicted. This relation was significant (P = 0.025) and has not been described previously. Nitrogen balance was inversely related to EE (P = 0.05), positively related to protein intake (P = 0.0075), and more likely to be attained with protein intakes larger than 2 g. kg(-1). d(-1) (P = 0.0001). Nitrogen balance became positive in trial patients over time but were negative in control patients over time (P = 0.0001). Nitrogen balance was directly associated with hospital outcome (P = 0.03) and intensive care unit outcome (P = 0.02). For every 1-g/d increase in nitrogen balance, the probability of survival increased by 21% (P = 0.03; odds ratio, 1.211; 95% confidence limits, 1.017,1.443). Further, although enterally and parenterally fed patients had lower mortalities than predicted, the presence of enteral feeding, even after adjusting for predicted risk of death, had a statistically significant benefit to patient outcome (P = 0.04). CONCLUSIONS: This study found that a metabolic cart can improve the accuracy of energy provision and that a protein intake of 2.5 g. kg(-1). d(-1) in these patients increases the likelihood of achieving a positive nitrogen balance and improving survival. Enteral feeding is preferable, but if this is not possible or does not achieve the target, then it should be supplemented by parenteral feeding.