Retrospective evaluation of commonly used equations to predict energy expenditure in mechanically ventilated, critically ill patients

Study Objective. To determine which of four commonly used equations to estimate energy expenditure is precise and unbiased compared with energy expenditure as measured by indirect calorimetry. Design. Retrospective, observational study. Setting. Adult medical intensive care unit in a research hospital of the National Institutes of Health Clinical Center. Patients. Seventy-six adult, mechanically ventilated, critically ill patients. Intervention. Indirect calorimetry reports generated by the National Institutes of Health Critical Care Medicine Department's Metabolic Cart Consult Service were reviewed. Bias and precision of resting energy expenditure (REE) estimated by equations were computed using mean prediction error (ME) and root mean squared prediction error (MSE). Equations were considered precise if the 95% confidence interval for MSE was within 15% of the measured energy expenditure (MEE) determined by indirect calorimetry. Equations were considered unbiased if the 95% confidence interval for ME included zero. Paired t tests were used to compare estimated REE values for each predictive equation with MEE values determined by indirect calorimetry. Data were stratified into regions of bias using classification and regression tree analysis, as well as visual inspection of estimated REE-versus-MEE curves for each equation. Measurements and Main Results. The Harris-Benedict equation multiplied by an activity factor of 1.2 was unbiased and precise. The Ireton-Jones equation was precise but biased. The American College of Chest Physicians' consensus recommendation was biased and imprecise. The Harris-Benedict equation without an activity factor also demonstrated bias and imprecision. Conclusions. The Harris-Benedict equation multiplied by an activity factor of 1.2 is suitable for predicting REE and may be used in the absence of indirect calorimetry.