The primary purpose of this study was to develop and cross-validate new equations for estimating REE in a group of elite male athletes of different sports, and then to compare them with existing formulas. Findings show that the new equations provide the best prediction of REE in the validation group, while the use of BIA-derived PhA improves the prediction power of the equation.
Meeting energy requirements is a priority of athletes. Inadequate energy intake might compromises performance and reduces the benefits of training [1, 2]. Energy needs are usually estimated by REE multiplied by the appropriate activity factor. To date, only a few number of predictive equations for REE have been specifically developed for athletes [13–15]. De Lorenzo formula [13] was derived in a sample of 51 male athletes (22 water polo, 12 judo, 17 karate) who exercised at least 3 hours/day; in that paper REE was underestimated by most of the seven equations selected from the literature. Later, Wong et al. [14] proposed gender-specific predictive formulas for elite Malaysian athletes in most cases practicing combat sports. Of note, Malaysian population seemed to have relatively low body frames and size and, therefore, low REE [14]. They found that mean basal metabolic rate measured by indirect calorimetry were similar in males to those estimated using the HB [5], FAO [7] and De Lorenzo [13] equations, but the accuracy of the predictive formulas was not evaluated. Finally, ten Haaf et al. [15] developed two predictive equations for recreational athletes practicing > 3 hour/day two times a week, the first formula being based on weight, the second one on FFM. Authors pointed out that the weight-based equation had a higher precision (83% for males) compared to the De Lorenzo formula (77.4% for males).
In the present study, we developed an equation based on age and main anthropometric variables (weight, stature, and BMI) (Model 1, Equation A). In addition to age, weight emerged as the only significant predictor. Two of the existing formulas for athletes identified also stature as predictor [13, 15] while in the athletes we studied, REE was correlated to stature in univariate analysis, but not in multiple regression analysis, p = 0.374). Instead of using BIA-derived body composition (strictly dependent on the BIA formula used), we opted for including raw BIA variables (BI-index and PhA) in the regression model (Model 2, Equation B).
BI-index is directly related to FFM and quite always included as predictor in the BIA equations to predict FFM. More recently, attention has been focused on the role of PhA as a biomarker of body cell mass and muscle quality as well as of water distribution (ratio between extracellular water-ECW and intracellular water-ICW)[28]. Thus, high PhA indicates greater cellularity (e.g. more body cell mass relative to FFM), cellular integrity and cell functions [28]. It may represent a proxy parameter of muscle quality in athletes, being significantly associated with physical activity and muscle strength [29, 30]. A recent systematic review showed that PhA was higher in athletes vs controls whereas it was still uncertain to what extent PhA differs among various sports [31]. In addition, PhA may help in detecting low muscle quality and identifying sarcopenia [32]. In previous studies, we also found that both BI-index and PhA improved the prediction power of REE under physiological conditions [18]. The findings of the present paper confirmed that PhA was as a significant predictor along with weight, with R² increasing from 0.637 to 0.675 and SEE decreasing from 150 to 141 kcal/day. On the contrary, BI-index was not recognized as a stronger predictor than weight, possibly because of low body fat percentage and low BMI.
As additional aim, we validated the two new equations and eight formulas selected from the literature (5 for the general population and 3 for athletes), at both population and individual level. On the average, the accuracy was very good for our new formulas, since bias ranged within ± 1%. Similarly, most of the selected equations, except the Mifflin and Owen ones, showed an acceptable prediction accuracy (bias ± 5%).
From a practical point of view, evaluating the accuracy of predictive equations at individual level (within ± 10%) is crucial for the nutritional management of the single athlete. This study shows that precision was high for the new formulas, especially for Equation B (~ 92%) including PhA in the model while it was lower, being close to 75%, for most of the other formulas (with the exception of the Mifflin and Owen ones for which it was much lower). Looking at the Bland-Altman plots, most of the prediction equations were more accurate at lower ranges of MREE and less accurate with the higher REE values. The new formulas gave the narrowest limits of agreement and the lowest bias.
To the best of authors’ knowledge, this is the first study that developed and cross-validate equations for elite athletes to predict REE based not only on anthropometric measures, but also on raw BIA variables. Overall, we conducted this study in a reasonable large sample of individuals, using recognized and well-documented methods and in line with similar previous studies in healthy subjects. Furthermore, the assessment of BIA with the same device has limited the device-related changes in PhA. Nevertheless, these findings are subject by a number of limitations. Since this is a retrospective study, our findings need to be confirmed in larger samples and in different sports disciplines. Additionally, we studied elite athletes mostly practicing endurance sports. Lastly, female athletes were excluded from the analysis due to the small number of potential participants (n = 27); therefore, we have developed new athlete-specific predictive equations for estimating REE in elite male athletes only.