Although providing appropriate levels of energy and nutrients is essential for healing, there is no consensus in the literature regarding the optimal method to predict resting energy expenditure (REE) in seriously injured individuals, and no equation has previously been developed or validated solely with that population. This study sought to determine which of five previously developed equations provided the most accurate estimate of REE when compared to measurement by indirect calorimetry. In addition, the study attempted to adjust those existing equations to improve their adequacy, and to develop new equations specifically from and for seriously injured individuals.
Using a retrospective design, the data from 106 measurements on 83 subjects from a single trauma center in the Midwest were collected, and REE was estimated using the Harris-Benedict, Mifflin-St. Jeor, Ireton-Jones 1992, Ireton-Jones 2002, and Penn State equations. Bias, precision, and limits of agreement were determined using the method of Bland and Altman, and biases were compared using ANOVA by linear mixed model. The best performer was the Harris-Benedict equation, with a bias of -248.1 kcal; it predicted REE within 10% of the measured REE in 39.6% of the subjects and within 500 kcal/day of measured REE in 70.5% of the subjects. An omnibus F-test determined a significant difference in the Bland-Altman bias of the five equations, with F (4, 328) = 603.77, p < 0.0001. Post-hoc analysis using the Scheffé correction demonstrated all pairwise comparisons were significant at p < 0.001. Also using the linear mixed model, bias-predicting equations were developed for all five equations by multiple regression. Finally, two completely new equations were developed using the same data and regression method with one equation containing weight, height, age, sex, and an intercept, and the second equation containing BMI, age, sex, and an intercept. Only one of the five terms in the first (weight/height) equation was statistically significant, but three of the four terms in the second (BMI) equation were significant.
Finding the Harris-Benedict equation most adequate in this population agrees with several previous studies, although none had previously used ANOVA to determine differences between the equations tested for adequacy. Using regression to predict Bland-Altman bias, and then using the equation generated as a correction factor to decrease bias in an existing estimation equation, also had not previously been attempted. The method of correction by bias estimation, as well as the two new equations generated by this study, need to be validated in independent samples.