A prospective trial was conducted in 20 surgical intensive care unit patients receiving mechanical ventilation to compare estimates of resting energy expenditure (REE) with the measured values. The 12 women and 8 men (mean age 61.7 +/- 2.8 (s.e.m.) years) underwent a full nutritional assessment before measurement of their REE by indirect calorimetry using the MGM II metabolic cart (Utah Medical). Their REE was estimated by the Harris-Benedict formula (mean 1324 +/- 53 (s.e.m.) kcal/day) as well as an empirical formula (where empirical formula = 22 x body weight in kg) (1370 +/- 68 (s.e.m.) kcal/day). Results by either estimate were not statistically different from the measured resting energy expenditure (MREE) (1382 +/- 130 (s.e.m.) kcal/day), by one-way analysis of variance. No multiplication factors were needed to relate the Harris-Benedict formula to MREE in this critically ill population and the estimate by the empirical formula was as good and simpler to perform. Although values by the Harris-Benedict formula and empirical formula correlated significantly with MREE (P less than 0.05) the correlation coefficients were low: 0.48 and 0.45 respectively. Thus, mean estimates were excellent for the group but less predictive for the individual. Multiple linear regressions did reveal that body composition and metabolic stress were of greater predictive value, since an equation relating weight (P less than 0.003), sex (P less than 0.003), white blood cell count (P less than 0.003) and 24-h urinary creatinine excretion (P less than 0.05) could predict 76 per cent of the variation (r = 0.87) in the MREE with an overall significance of P = 0.0002. Given the importance of matching energy intake to needs in many critically ill patients who are mechanically ventilated, accurate measurement of the REE is recommended now that instrumentation is equal to the task.