In critically ill patients, the inaccuracy of predictive formulas for nutritional assessment often leads to inappropriate and potentially detrimental feeding regimens. This study evaluates the clinical utility of the metabolic cart in an urban university hospital setting. Twenty-six studies were performed on each of 26 patients (18 surgical, 8 medical) using an MMC Horizon metabolic cart. Although 58% of patients were overweight, 42% were still shown to have a kwashiorkorlike pattern of malnutrition. Three patients demonstrated a marasmic-like pattern. Fifteen percent of studies showed patients to be hypometabolic and 62% hypermetabolic. Harris-Benedict resting energy expenditure, based on actual or ideal body weight, underestimated needs; however, addition of a metabolic activity factor overestimated needs. Only 32% of patients were fed appropriately; 41% were underfed, and 27% were overfed. Urine area nitrogen correlated poorly with energy expenditure. Measured RQ appropriately reflected substrate utilization in 77% of studies; multiple factors may have caused differences between measured and predicted RQ in 23%. Use of the metabolic cart determines precisely the metabolic state, identifies problems with substrate utilization, and enables the physician to design the most efficacious nutritional regimen.