Introduction
It has been shown that athletes who are undergoing multiple intense training sessions throughout the week or are larger in stature likely have higher total daily energy expenditures (TDEEs) than sedentary individuals (3). Nutritional interventions are often implemented by sport nutrition practitioners to help athletes consume adequate energy to provide sufficient fuel for training and maintain body mass, particularly lean body mass, in an attempt to mitigate performance decrements throughout the season while maximizing training adaptations. To do so, practitioners must be aware of the overall state of energy balance (energy in vs. energy out) for the athlete (17). When levels of expenditure exceed those of intake, an individual is in a negative state of energy balance. This negative energy balance may lead to a reduction in overall body mass. In addition, dependent on the consumption of other nutrients, such as protein, losses in lean body mass and subsequently strength may also occur (17). Therefore, to preserve lean body mass and optimize training adaptations such as strength and power, it is imperative for an athlete to consume an appropriate amount of energy (caloric intake) based on their level of training and basal metabolic requirements. To provide individualized calorie recommendations, a practitioner must know the TDEE of the athlete to better understand their current state of energy balance. Evidence suggests that basal metabolic rate, often measured or expressed as resting energy expenditure or resting metabolic rate (RMR), comprises approximately 60–70% TDEE (7). Therefore, identification of RMR can serve as a valuable tool in developing nutritional interventions to enhance sport performance or help with weight management strategies in athletes. Specifically, establishing an athlete's RMR could help establish daily caloric requirements to ensure adequate energy intake. This can also help to avoid any long-term imbalance between energy intake and energy output, which could lead to negative health outcomes, including metabolic dysfunction or ultimately make the athlete more susceptible to weight (mostly body fat (BF)) gain or performance decrements (27).
Indirect calorimetry is currently accepted as a valid and reliable measure of RMR. This process, however, requires expensive laboratory equipment and thus, may not be a viable option for practitioners. As a result, several prediction equations have been developed to offer a low-cost alternative method for estimation of RMR (8,10,13,18). Most equations currently in use today were developed based on populations comprised predominantly older, overweight or obese, and minimally active adults and may not be applicable within athletic populations (1,13,19). For example, Mifflin et al. (18) developed an equation to predict RMR using various physical characteristics (i.e., height, body mass, and age) of obese individuals. Similarly, the RMR prediction equation of Nelson et al. (19) was derived from values from previous reports that included obese and nonobese, nonactive adults. Because of the differences in activity levels and body composition, these RMR prediction equations may not be appropriate for younger athletic populations. As evidenced, when select equations were used in active populations and compared with direct RMR assessment to validate their accuracy, several of these equations were found to over or underestimate RMR by as much as 300 kcals (±10%) (26). These differences could have a substantial impact on weight management strategies and performance long term (15). A potential reason for the inaccuracies of these prediction equations could be a result of certain variables used within the prediction model; specifically, the inclusion, or exclusion, of lean body mass. For example, athletes tend to have a higher amount of lean body mass than the general population, which may invalidate some of the traditionally used RMR prediction equations that do not account for varying amounts of lean body mass and the subsequent increase in relative metabolic activity (8). It has been shown that lean body mass can have a strong impact on energy requirements (9,28). Specifically, Webb (28) observed a strong correlation between 24 hours energy expenditure (under resting conditions) and fat-free mass (FFM) in both men and women. Cunningham (8) found that lean body mass accounts for approximately 70% of the variability in the prediction of RMR and as expected, when FFM is used to predict RMR, equations are much more accurate in certain populations (25). Recently, Flack et al. (12) examined the accuracy of several RMR prediction equations in healthy adults and found several of them to underestimate RMR. In addition, the authors concluded that the prediction bias of the RMR equations was inversely associated with the magnitude of RMR and lean body mass suggesting that RMR prediction equations may become less accurate for individuals with higher amounts of lean body mass.
Some prediction equations that have been developed for more healthy or active populations are associated with certain limitations as well. For example, De Lorenzo et al. (10) developed a prediction equation for athletes without the inclusion of FFM in a male-only sample. Furthermore, the RMR prediction equation of Cunningham (8) derived from healthy, nonactive adults, used an estimation of lean body mass value as opposed to direct assessments. Based on these findings, there are inherent gaps in the literature regarding the efficacy and appropriateness of estimating RMR in athletes, particularly with team sports and strength-/power-based athletes. In addition, no previous studies have identified the best prediction equation to use with female athletes and whether or not they are appropriate to use for determination of RMR. Therefore, the purpose of this study was to compare the accuracy of commonly used RMR prediction equations with values obtained from direct measurement in an athletic sample of men and women.
Methods
Experimental Approach to the Problem
The current study used a cross-validation design to determine the accuracy of 5 commonly used prediction equations (8,10,13,18,19) compared with indirect calorimetry for determination of RMR in male and female athletes. Athletes completed a single day of testing to determine body composition and RMR. Physical characteristics and descriptive information from the subjects were then used to compute predicted RMR using 5 different prediction equations. The predicted RMR values from the equations were then statistically compared with measured RMR values.
Subjects
Twenty-two female (19.7 ± 1.4 years; 166.2 ± 5.5 cm; 63.5 ± 7.3 kg; 49.2 ± 4.3 kg of FFM; 23.4 ± 4.4 BF%) and 28 male (20.2 ± 1.6 years; 181.9 ± 6.1 cm; 94.5 ± 16.2 kg; 79.1 ± 7.2 kg of FFM; 15.1 ± 8.5 BF%;) National Collegiate Athletic Association (NCAA) Division III athletes participated in this observational study as presented in Table 1. Subjects were recruited from a variety of male (football, n = 21; track and field, n = 4; baseball, n = 3) and female (soccer, n = 15; swimming/diving, n = 4, track and field, n = 3) team sports. Before testing, all subjects were informed of the details of their participation and provided written informed consent in compliance with the Protection of Human Subjects Guidelines put forth by the Institutional Review Board at the University of Wisconsin—La Crosse.
;)
Table 1.: Participant demographics.*
Procedures
Resting Metabolic Rate and Body Composition
Participants were asked to visit the Human Performance Laboratory in a rested (>24 hours) and fasted (>12 hours before testing) state. On arrival to the laboratory, participants were assessed for height and weight using a SECA physicians scale (SECA, Hamburg, Germany). Participants then completed a RMR analysis using indirect calorimetry (ParvoMedics True One Metabolic System; ParvoMedics, Utah, Sandy, USA). This is a nonexertional test with participants lying supine on an examination table. A clear, hard plastic hood and soft, clear plastic drape was placed over the participants' neck and head to determine resting oxygen uptake and metabolic rate. All participants remained motionless without falling asleep for approximately 20 minutes. Data were recorded after the first 10 minutes of testing during a 5-minute period in which criterion variables (e.g., V̇o2 L·min−1) changed less than 5%. After the RMR assessment, all participants had their body composition assessed using air displacement plethysmography (BODPOD; Cosmed, Chicago, IL, USA). The BodPod has been shown to be a valid tool for the measurement of body composition and is highly correlated with hydrostatic weighing (2). Calibration procedures were completed based on manufacturer guidelines using an empty chamber and calibration cylinder of a standard volume (49.55 L) before testing. Subjects were instructed to wear tight-fitting clothing, remove all jewelry, and wear a swim cap before entering the testing chamber. Subjects were also instructed to sit motionless in the chamber, utilizing normal breathing patterns. Lung volume was directly assessed to correct for relative body volume. Fat and FFM were determined based on body mass and body volume obtained from the BODPOD using the Brozek equation (4). Test-to-test reliability of performing this body composition assessment in our laboratory with athletic populations has yielded high reliability for body mass (R = 0.999), BF percent (R = 0.994), and FFM (R = 0.998).
Prediction Equations
Resting metabolic rate values for each participant were then estimated using 5 commonly used prediction equations (8,10,13,18,19) as summarized in Table 2.
Table 2.: Summary of RMR prediction equations.*
Statistical Analyses
A 1-way analysis of variance (ANOVA) was used to assess sex differences in baseline demographics. Linear regression analysis was used to assess the accuracy of each RMR prediction method by calculating the standard error of estimate (SEE), constant error (CE = mean difference for measured RMR − predicted RMR), and root-mean-squared prediction error . Two separate 1-way repeated-measures ANOVAs were used to compare differences among measured RMR and the predicted RMR values of each equation for men and women. Follow-up analyses included paired-sample t tests with Bonferroni correction (0.05/5 = 0.01). Bland-Altman plots were created to assess the accuracy and precision of the prediction equations. Statistical significance was set at an alpha of p ≤ 0.05. All data were analyzed using the Statistical Package for the Social Sciences (SPSS, Version 21.0; SPSS Inc., Chicago, IL, USA).
Results
Male athletes exhibited significantly higher amounts of height, body mass, and FFM (p < 0.001) along with a significantly lower-BF percent (p < 0.001) compared with female athletes. Table 3 presents a summary of the mean differences in measured vs. predicted RMR in male athletes. All prediction equations yielded values that were statistically different from the measured RMR value. All the prediction equations significantly underestimated RMR while the Cunningham equation (8) had the smallest mean difference (165 kcals).
Table 3.: Comparison of measured and predicted RMR values in male athletes (paired-samples t tests).*†
Table 4 presents a summary of the mean differences in measured vs. predicted RMR in female athletes. All prediction equations yielded values that were statistically different from the measured RMR value with the exception of the Cunningham equation, which had the smallest mean difference of 39 kcals (8). The Nelson, Mifflin-St. Jeor, and Harris-Benedict equations (13,18,19) underestimated RMR, whereas the De Lorenzo and Cunningham equations (8,10) overestimated RMR.
Table 4.: Comparison of measured and predicted RMR values in female athletes (paired-samples t tests).*†
Table 5 presents a summary of the results from the regression analysis between the measured RMR and predicted RMR values in the male athletes. The variance in predicted RMR values from the various equations ranged from an SEE = 206 kcal·d−1 (Harris and Benedict (13)) to an SEE = 240 kcal·d−1 (Cunningham (8)), accounting for 51 and 34% of the variance in male athletes, respectively. In male athletes, the Harris-Benedict equation (13) was found to be the most accurate prediction formula with the lowest root-mean-square prediction error (RMSPE) value of 284 kcals, whereas the Mifflin equation (18) yielded the highest RMSPE value of 466 kcals.
Table 5.: Regression coefficients of the differences (measured-predicted) and variances for RMR prediction equations in male athletes.*
Table 6 presents a summary of the results from the regression analysis between the measured RMR and predicted RMR values in the female athletes. In the female athletes, the variances in predicted RMR values from the different equations ranged from an SEE = 95 kcal·d−1 from the Harris-Benedict equation (13) to an SEE = 121 kcal·d−1 from the De Lorenzo equation (10), accounting for 64 and 43% of the variance, respectively. The Cunningham (8) equation was found to be the most accurate prediction equation with the lowest RMSPE value of 110 kcals, whereas the Nelson (19) equation yielded a RMSPE value of 228 kcals and therefore, was the least accurate prediction equation in the female athletes.
Table 6.: Regression coefficients of the differences (measured-predicted) and variances for RMR prediction equations in female athletes.*
The results of the Bland-Altman analysis are presented in Figure 1. The relationships between the CE and the average of measured RMR and predicted RMR were significant for the Nelson (r = 0.39), Mifflin (r = 0.70), Cunningham (r = 0.62), and De Lorenzo (r = 0.52) equations for male athletes. The relationships between the CE and the average of measured RMR and predicted RMR were significant for the Mifflin (r = 0.60), Cunningham (r = 0.59), and Harris-Benedict (r = 0.79) for female athletes, respectively.
Figure 1.: Relationships between constant error (actual − predicted RMR) and average RMR ([measured + predicted]/2) for men (A) (n = 28) and women (B) (n = 21) in the current sample of athletes. Solid line represents mean of constant error. Dashed lines represent ± 1.96 SD of constant error. RMR = resting metabolic rate. *p ≤ 0.05.
Discussion
The purpose of the current study was to evaluate the accuracy of several commonly used RMR prediction equations among athletic populations. A secondary aim was to identify any difference in RMR prediction accuracy between sexes. Based on the results of the current study, the Harris-Benedict (13) equation was most accurately able to predict RMR values in men yielding the lowest RMSPE value of 284 kcals·d−1 and accounting for 51% of the variance in measured RMR; however, the mean difference was still significant and therefore, the equation did not accurately predict measured RMR. In the female athletes, the Cunningham equation was best able to predict RMR as it yielded the lowest RMSPE value of 110 kcals·d−1 and accounted for 53% of the variance in measured RMR. Furthermore, the Cunningham equation yielded a mean difference value that was nonsignificant and therefore in female athletes, it seems as although the Cunningham equation is able to reasonably predict RMR. Our results indicating that the Cunningham equation (8) was best in predicting RMR in female athletes is in agreement with other reports that also found this equation to accurately predict RMR in athletic populations within a range of 160 kcals·d−1 (10,26). The reason the Harris-Benedict equation yielded the lowest RMSPE value and was considered the most accurate prediction equation, although still significantly different from measured RMR, in this study for men is unclear. This is contrary to previous findings which examined the accuracy of various RMR prediction equations in male athletes (10,25,26). The Harris-Benedict (13) equation was originally developed in 1918 based on a predominately young population consisting of men and women who were of a healthy weight. This prediction model incorporates multiple components of body stature such as height, body mass, and age, all of which have been shown to influence RMR (9,10,20). Conversely, other predictive models may only utilize an individual's weight or FFM to estimate RMR. Therefore, the inclusion of multiple variables in this case may have improved the accuracy of the predicted RMR value in men. To our surprise, a FFM-based equation was not the best predictor in male athletes as anticipated. Athletes tend to have greater amounts of FFM compared with the general population and because FFM has been shown to be highly correlated with RMR, we hypothesized that this equation would be the most accurate across both sexes. Other reports have found that FFM accounts for a large variability in measured RMR and is likely the reason why the Cunningham (8) equation has been shown to accurately predict RMR in athletic populations as the model is primarily based on FFM (5,7,8,25,26). The Cunningham (8) equation was developed using a cohort of 19 individuals who were classified as being more “active” from the original Harris-Benedict population (13).
It is possible that the higher measured RMR in the men of the current study may have influenced the outcome of the regression analysis and ultimately the accuracy of each equation. For example, the mean measured RMR of the male athletes in the current study was 2,405 kcal·d−1 which is far above those reported in other studies using male athletes or active individuals (10,21,25,26). When examining the Bland-Altman plots, it is clear that most prediction equations were more accurate at lower ranges of measured RMR and less accurate with the higher RMR values, particularly in the male athletes. Therefore, it is possible that prediction equations may be less likely to accurately predict RMR for athletes with a higher RMR value, as would be the case with larger athletes. Specifically it seems as although most prediction equations will underestimate RMR, with the exception of the De Lorenzo and Cunningham equations in female athletes (8,10). This leads to the presumption that certain RMR prediction models may be population specific. For example, de Oliveira et al. (11) found the World Health Organization (FAO/WHO/UNU) and Harris-Benedict equations to best predict RMR (−180 kcals·d−1) in overweight and obese adults (1,13). Similarly, Weijs et al. (29) found the FAO/WHO/UNU prediction equation (1) to best predict RMR of adult patients in a clinical setting (RMSPE of 233 kcal·d−1) for outpatients (RMSPE of 182 kcal·d−1), inpatients (RMSPE of 277 kcal·d−1), and underweight patients (RMSPE of 219 kcal·d−1). This may also hold true for athletic populations. As evidenced, Haaf and Weijs (25) found the Cunningham equation (8) to best predict RMR (within 10%) in a population of recreationally active Dutch male and female athletes of various sports. Furthermore, Thompson and Manore (26) found the Cunningham (8) equation to best predict RMR in male (RMSPE 158 kcals·d−1) and female (RMSPE 103 kcals·d−1) endurance athletes. In addition, De Lorenzo et al. (10) evaluated several RMR prediction equations and found the Harris-Benedict (RMSPE 49 kcals·d−1) (13) and Cunningham (RMSPE −59 kcals·d−1) (8) equation to best predict RMR in 51 male athletes training +3 hours per day. The authors (10) also developed their own prediction equation which was found to have a RMSPE of 1 kcal·d−1. This equation, however, was not found to be very accurate in its ability to predict RMR for the subjects included in the current study.
Collectively, the results of the current study provide evidence that most RMR prediction equations being examined underestimated RMR with the exception of the De Lorenzo (10) and Cunningham (8) equations in female athletes. Therefore, when relying on these RMR prediction equations for energy recommendations, it is important to understand that they may underestimate an athlete's actual energy needs. Similar findings have been reported in other studies examining the accuracy of RMR prediction equations in athletic populations, particularly endurance trained athletes (10,25,26). For example, De Lorenzo et al. (10) found that the Cunningham equation slightly overestimated (+59 kcal·d−1) measured RMR, whereas the Harris-Benedict and Mifflin equation underestimated RMR in 51 advanced-trained male athletes across multiple sports. Similarly, Thompson and Manore (26) reported that all mean predicted RMR values were lower than measured RMR with the exception of the Cunningham equation in male and female endurance athletes. The reason these RMR prediction equations consistently underestimate RMR is unclear; however, several factors may play a role. Athletes are more active through their training and sport-specific activities, which may have elevated RMR values above what may be observed with lesser active or sedentary individuals as increases in physical activity, particularly recent activity, have been shown to influence RMR values (6,23,24). For example, Bullough et al. (6) found that RMR is influenced by the recent state of energy balance within an individual. Specifically, the authors found the recent exercise activity and an adequate energy intake was associated with elevated rates of RMR. Recent physical activity may even influence RMR up to 72–96 hours postexercise (6,14,22), which may also explain why certain RMR prediction equations seem to underestimate measured RMR and reiterates the need for more sport-specific RMR prediction equations to accurately reflect the fluctuations in metabolic activity after sport activities. Furthermore, increases in physical activity, particularly resistance training, will likely increase body mass and FFM, both of which have been shown to account for a large proportion of the variability in measured RMR and to be positively correlated with RMR (8–10,26). Athletes who have greater amounts of FFM may have an elevated RMR value compared with age-matched controls. However, in this particularly study, the FFM-based equations were only found to be more accurate in female athletes. It is also possible that the 2-compartment model of body composition determination used in the current study influenced the predictive accuracy of the FFM used in the equations, as previous results have shown variations in RMR when indexing RMR against FFM using different body composition models which is a limitation of the current study (16). Another limitation is that most athletes in the current study were younger collegiate athletes, and therefore, the results may be different in younger or older athletic populations. Regardless, when the currently used RMR prediction equations are used to predict RMR in athletes, the resultant values will likely be lower than measured RMR.
Practical Applications
If direct access to metabolic equipment is unavailable, RMR prediction equations can serve to estimate RMR; however, the equations used in the current study are likely to underestimate RMR in collegiate athletes. Therefore, practitioners should be cautious when using these equations to determine energy requirements for athletes and may want to conservatively increase the predicted RMR value by 100–300 kcals depending on the equation used. Further, based on the results of the current study it seems as although the Harris-Benedict equation is best able to predict RMR in male athletes, whereas the Cunningham equation is best able to predict RMR in female athletes. More research is needed to determine an athlete specific RMR prediction equation, particular one that is sport specific to account for differences in body type and activity levels.
References
1. Energy and protein requirements. Report of a joint FAO/WHO/UNU Expert Consultation. World Health Organ Tech Rep Ser 724: 1–206, 1985.
2. Biaggi RR, Vollman MW, Nies MA, Brener CE, Flakoll PJ, Levenhagen DK, Sun M, Karabulut Z, Chen KY. Comparison of air-displacement plethysmography with hydrostatic weighing and bioelectrical impedance analysis for the assessment of body composition in healthy adults. Am J Clin Nutr 69: 898–903, 1999.
3. Black AE, Coward WA, Cole TJ, Prentice AM. Human energy expenditure in affluent societies: An analysis of 574 doubly-labelled water measurements. Eur J Clin Nutr 50: 72–92, 1996.
4. Brozek J, Grande F, Anderson JT, Keys A. Densitometric analysis of body composition: Revision of some quantitative assumptions. Ann N Y Acad Sci 110: 113–140, 1963.
5. Buchholz AC, Rafii M, Pencharz PB. Is resting metabolic rate different between men and women? Br J Nutr 86: 641–646, 2001.
6. Bullough RC, Gillette CA, Harris MA, Melby CL. Interaction of acute changes in exercise energy expenditure and energy intake on resting metabolic rate. Am J Clin Nutr 61: 473–481, 1995.
7. Carpenter WH, Poehlman ET, O'Connell M, Goran MI. Influence of body composition and resting metabolic rate on variation in total energy expenditure: A meta-analysis. Am J Clin Nutr 61: 4–10, 1995.
8. Cunningham JJ. A reanalysis of the factors influencing basal metabolic rate in normal adults. Am J Clin Nutr 33: 2372–2374, 1980.
9. Cunningham JJ. Body composition as a determinant of energy expenditure: A synthetic review and a proposed general prediction equation. Am J Clin Nutr 54: 963–969, 1991.
10. De Lorenzo A, Bertini I, Candeloro N, Piccinelli R, Innocente I, Brancati A. A new predictive equation to calculate resting metabolic rate in athletes. J Sports Med Phys Fitness 39: 213–219, 1999.
11. de Oliveira EP, Orsatti FL, Teixeira O, Maesta N, Burini RC. Comparison of predictive equations for resting energy expenditure in overweight and obese adults. J Obes 2011: 534714, 2011.
12. Flack KD, Siders WA, Johnson L, Roemmich JN. Cross-validation of resting metabolic rate prediction equations. J Acad Nutr Diet 116: 1413–1422, 2016.
13. Harris JA, Benedict FG. A biometric study of human basal metabolism. Proc Natl Acad Sci USA 4: 370–373, 1918.
14. Herring JL, Mole PA, Meredith CN, Stern JS. Effect of suspending exercise training on resting metabolic rate in women. Med Sci Sports Exerc 24: 59–65, 1992.
15. Kisiolek J, Schultz K, Luedke J, Jones MT, Oliver J, Jagim AR. A comparison of resting energy expenditure prediction equations in young recreationally active women. Presented at International Society of Sports Nutrition National Conference, Austin, TX, September 21, 2015.
16. LaForgia J, van der Ploeg GE, Withers RT, Gunn SM, Brooks AG, Chatterton BE. Impact of indexing resting metabolic rate against fat-free mass determined by different body composition models. Eur J Clin Nutr 58: 1132–1141, 2004.
17. Loucks AB. Energy balance and body composition in sports and exercise. J Sports Sci 22: 1–14, 2004.
18. Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO. A new predictive equation for resting energy expenditure in healthy individuals. Am J Clin Nutr 51: 241–247, 1990.
19. Nelson KM, Weinsier RL, Long CL, Schutz Y. Prediction of resting energy expenditure from fat-free mass and fat mass. Am J Clin Nutr 56: 848–856, 1992.
20. Poehlman ET, Berke EM, Joseph JR, Gardner AW, Katzman-Rooks SM, Goran MI. Influence of aerobic capacity, body composition, and thyroid hormones on the age-related decline in resting metabolic rate. Metabolism 41: 915–921, 1992.
21. Poehlman ET, Melby CL, Badylak SF. Resting metabolic rate and postprandial thermogenesis in highly trained and untrained males. Am J Clin Nutr 47: 793–798, 1988.
22. Schulz LO, Nyomba BL, Alger S, Anderson TE, Ravussin E. Effect of endurance training on sedentary energy expenditure measured in a respiratory chamber. Am J Physiol 260: E257–E261, 1991.
23. Speakman JR, Selman C. Physical activity and resting metabolic rate. Proc Nutr Soc 62: 621–634, 2003.
24. Speakman JR, Westerterp KR. Associations between energy demands, physical activity, and body composition in adult humans between 18 and 96 y of age. Am J Clin Nutr 92: 826–834, 2010.
25. ten Haaf T, Weijs PJ. Resting energy expenditure prediction in recreational athletes of 18-35 years: Confirmation of Cunningham equation and an improved weight-based alternative. PLoS One 9: e108460, 2014.
26. Thompson J, Manore MM. Predicted and measured resting metabolic rate of male and female endurance athletes. J Am Diet Assoc 96: 30–34, 1996.
27. Trexler ET, Smith-Ryan AE, Norton LE. Metabolic adaptation to weight loss: Implications for the athlete. J Int Soc Sports Nutr 11: 7, 2014.
28. Webb P. Energy expenditure and fat-free mass in men and women. Am J Clin Nutr 34: 1816–1826, 1981.
29. Weijs PJ, Kruizenga HM, van Dijk AE, van der Meij BS, Langius JA, Knol DL, Strack van Schijndel RJ, van Bokhorst-de van der Schueren MA. Validation of predictive equations for resting energy expenditure in adult outpatients and inpatients. Clin Nutr 27: 150–157, 2008.
