Predicted estimates of resting energy expenditure have limited clinical utility in patients with cirrhosis: an individual patient data analysis

Abstract

Background & Aim: Malnutrition is associated with adverse clinical outcomes in patients with cirrhosis. Accurate assessment of energy requirements is needed to optimize dietary intake. Resting energy expenditure (REE), the major component of total energy expenditure, can be measured using indirect calorimetry (mREE) or estimated using prediction equations (pREE). This study assessed the usefulness of predicted estimates of REE in this patient population. Methods: Individual mREE data were available for 900 patients with cirrhosis (mean [±1SD] age 55.7±11.6 yr; 70% men; 52% south-east Asian) and 282 healthy controls (mean age 36.0±12.8 yr; 52% men; 18% south-east Asian). Metabolic status was classified using thresholds based on the mean±1SD of the mREE in the healthy controls. Comparisons were made between mREE and pREE estimates obtained using the Harris-Benedict, Mifflin, Schofield and Henry equations. Stepwise regression was used to build three new prediction models which included sex, ethnicity, body composition measures, and MELD scores. Results: The mean mREE was significantly higher in patients than controls when referenced to dry body weight (22.43.8 cf. 20.82.6 kcal/kg/24hr; p<0.001); there were no significant sex differences. The mean mREE was significantly higher in Caucasian than Asian patients (23.1±4.4 cf. 21.7±2.9 kcal/kg/24hr; p<0.001). Overall, 37.1% of Caucasians and 25.3% of Asians were classified as hypermetabolic. The differences between mREE and pREE were both statistically and clinically relevant; in the total patient population, pREE estimates ranged from 501 kcal/24hr less to 548 kcal/24hr more than the mREE. Newly-derived prediction equations provided better estimates of mREE but still had limited clinical utility. Conclusions: Prediction equations do not provide useful estimates of REE in patients with cirrhosis. REE should be directly measured.