New Predictive Resting Metabolic Rate Equations for... : Medicine & Science in Sports & Exercise (original) (raw)

The 24-h energy expenditure (24hEE) is defined by the sum of the resting metabolic rate (RMR), the thermic effect of the food, and the energy expended during nonvolitional activities (i.e., spontaneous and nonexercise physical activity) besides the cold-induced and the exercise activity thermogenesis (1,2). RMR is the largest component of 24hEE and, depending on the population, commonly accounts for 50%–70% of 24hEE (3). In a laboratory setting, the indirect calorimetry (IC) technique is the gold standard method to assess the RMR. However, it requires sophisticated and high-cost equipment and personnel. As an alternative method, predictive equations have been broadly used for estimating the RMR in different populations such as patients (4,5), children and adolescent athletes (6–8), and adult athletes (9–11). Nevertheless, available equations rely on sex, age, and anthropometric/body composition variables, which may induce underestimations or overestimations in the predicted RMR values when applied in different groups from those included in the equations’ development. Altogether, this may limit the predicted RMR accuracy and, thus, influence the dietary prescription (12,13).

High-level athletes have specific energy requirements to satisfy the demands for training sessions, suitable posttraining recovery, and maintaining adequate body composition. Therefore, daily energy intake (macronutrients and micronutrients) levels should be accurately prescribed to maintain their full metabolic functions and homeostasis (1,10). An imbalance between energy intake and expenditure may result in difficulty to reach suitable body composition, increased risk of muscle damage and fatigue, impaired recovery between sessions, menstrual cycle disorders, and even Relative Energy Deficiency in Sport, which could negatively impact athletes’ performance (14–16). However, some difficulties may arise when using IC assessments to quantify high-level athletes’ “true” energy expenditure. For example, the RMR assessment should be performed under well-controlled conditions (e.g., 8- to 12-h fasting, refrain from physical activity during the last 24/48 h, etc.) (17,18), which might influence their training programming. Taken together, for an accurate athletes’ RMR determination and daily energy intake prescription, predictive equations could be an interesting and feasible option to develop proper nutritional plans. Indeed, the RMR component accounts for approximately 50%–70% of 24hEE (3). Thus, this is an important parameter to guarantee a suitable nutritional plan that satisfies the athletes’ energy needs and minimizes the occurrence of Relative Energy Deficiency in Sport syndrome (1,14,16,19,20).

Unfortunately, few predictive equations have been developed from and/or for athletes—especially for high-level athletes. De Lorenzo et al. (9) developed, more than 20 yr ago, a predictive equation using an athletes’ cohort, which included only three sports (i.e., water polo, judo, and karate). More recently, ten Haaf and Weijs (11) proposed new equations based on recreational athletes’ cohorts. Although the ten Haaf and Weijs study advanced in predicting the RMR in this population, possibly the RMR may be influenced by the performance level of the athletes, once it is expected that high-level athletes have more pronounced physical attributes than recreational ones (e.g., lean body mass) (21). Therefore, the literature still lacks a predictive equation developed for different sports and high-level athletes. Thus, the present study aims a) to assess the agreement between the measured RMR using a reference method (i.e., IC) and different previously proposed predictive equations to estimate the RMR, and b) to propose and cross-validate two new predictive equations for estimating RMR in high-level athletes from different sports.

METHODS

Participants

One hundred and two high-level athletes (44 female), comprising 21 different Olympic sports (archery, 2; artistic swimming, 2; athletics (track and field), 31; badminton, 2; beach volleyball, 9; boxing, 6; canoeing, 4; cycling, 3; diving, 2; judo, 4; karate, 2; marathon swimming, 1; rowing, 5; sailing, 1; surfing, 1; swimming, 9; taekwondo, 5; triathlon, 1; water polo, 8; weightlifting, 2; and wrestling, 2), took part in the study. The inclusion criteria were as follows: 1) athletes who compete at the Brazilian national level in their respective sports. However, most of them compete at higher levels (World Championship, 87%; Olympic Games, 45%); 2) a valid RMR assessment that accomplished the following gas exchange criteria (17,18): a) respiratory exchange ratio ranging from 0.7 to 1.0 (i.e., physiological range) (18), b) coefficient of variation (as a percentage) ≤10% for both oxygen consumption (V̇O2) and carbon dioxide production (V̇CO2), and c) a coefficient of variation ≤5% for respiratory exchange ratio (21); and 3) not to present injuries, diseases, or being taking any drug that could interfere in the subjects’ resting metabolism. The study was approved by the Research Ethics Committee of the Rio de Janeiro Municipal Health Secretariat and conducted following the Declaration of Helsinki (revision of 2013). All volunteers provided written informed consent before their participation.

Anthropometry

Height (Sanny 2020, São Bernardo do Campo, SP, Brazil), body weight (BW; Welmy W200, Santa Bárbara d’Oeste, SP, Brazil), and skinfolds were evaluated by two experienced (≥5 yr) researchers (R. F. and A. d. s. I.). Importantly, both researchers are certified by the International Society for the Advancement of Kinanthropometry, and each one has performed more than 200 body composition assessments. In brief, to evaluate the skinfolds, one researcher performed the measurements, whereas the other supervised the procedures, took notes, and guaranteed the quality of the measures. The skinfolds (chest, triceps, subscapular, axillary, suprailiac, thigh, and abdominal) were taken on the right side of the body, at least twice, using a Harpenden caliper (Baty International, Burgess Hill, England). If the first two measurements differed more than 5% from each other, additional measurements were taken until the desired range (i.e., ≤5%) was achieved. Then, the average of the two closest measurements was considered the final skinfold value. First, fat mass (FM) was estimated using the two-compartment Siri’s equation (22) from body density estimative, using the seven-skinfold formulas for men (23) and women (24), respectively. Then, the fat-free mass (FFM) was calculated by subtracting FM from BW. Lastly, the calculation of the somatotype was performed using Heath and Carter’s method (25).

RMR Data Collection and Selection

Two days before the test, all athletes were instructed about the study details and recommendations to be followed before the RMR assessment as described: a) 8- to 12-h fasting, b) abstention from moderate (24 h) or intense (48 h) physical exercise, and c) no use of alcohol, soft drinks, or caffeine 24 h before the test (17,18). Besides, the subjects were oriented to keep their usual sleep routine before the evaluation. Furthermore, the subjects commuted by car or stayed in a hotel near the sports center facility, performing the minimum physical activity on their way to the laboratory. When arriving at the laboratory, the subjects were placed in a quiet and thermoneutral room (20°C–23°C ± 1°C, 50%–70% relative humidity). After that, the RMR assessment was performed for 30 min while subjects were lying in a supine position, using a plastic canopy hood attached to the Quark CPET (Cosmed, Rome, Italy) metabolic cart. The V̇O2 and V̇CO2 gas exchange was continuously registered in a 10-s rolling average sampling mode, and the fraction of expired CO2 was manually adjusted (between 0.7% and 1.3%) during the whole test. Furthermore, before each measurement, gas analyzers (using reference concentration gases: 16.00% O2 and 5.00% CO2) and flow (using a standard 3-L syringe) were calibrated following the manufacturer’s instructions.

Once the RMR assessment finished, the V̇O2 and V̇CO2 data were downloaded and averaged every 1 min using an MS Office Excel spreadsheet (Microsoft Corporation, Redmond, WA). Then, following the recommendations, the first 5-min data of the measurement were discarded (17,18). In the remaining 25-min data, the high-filter method was used as proposed by Alcantara et al. (48) for young and middle-age adults and Freire et al. (26) for high-level athletes. In this study by Freire et al. (26), other methods (e.g., time interval or steady-state methods) for gas exchange data selection were compared to determine the “best” method to estimate the RMR in high-level athletes. In brief, when studying the association between RMR across the different methods and its classical predictors (e.g., BW or body composition), the high-filter method yielded the highest coefficient of determination (_R_2 = 0.88) and the lowest estimate standard error (SEE = 133 kcal·d−1) compared with the rest of methods studied. Thus, to use the high-filter method, after discarding the first 5-min data, the mean 25-min RMR is calculated (using Weir’s equation (27); see discussion hereinafter). Then, in a second step, from the remaining 25-min data, those 1-min RMR values that are either 95% lower or 105% higher than the mean 25-min RMR are excluded for further processing. Therefore, for the data that accomplished the aforementioned criteria, the RMR was estimated using Weir’s equation (27) (considering no urinary nitrogen excretion) and expressed as kcal·d−1.

Lastly, it should be mentioned that the RMR assessment was the first assessment among a more comprehensive physical evaluation. Therefore, the athletes did not lay on the bed before the RMR assessment (i.e., there was no acclimation period) before the gas exchange measurement.

RMR Predictive Equations

Predictive RMR equations (hereinafter “_estimated RMR_”) were compared with measured RMR using IC (hereinafter “_measured RMR_”). The inclusion criteria to select the equations were as follows: a) recommended by the American College of Sports Medicine (14) for use in athletes or by the Food and Agriculture Organization/World Health Organization (28), b) equations developed based on athletes’ cohort, and c) equations designed for both sexes. The exclusion criteria were as follows: a) equations that included patients or obese subjects in the sample, b) equations obtained from a small sample size (considered as n < 50), and c) equations that did not accomplish all inclusion criteria. Thus, nine equations that accomplished all criteria were included: Harris and Benedict (29) (H&B), Cunningham 1980 (30) (Cunningham1), Cunningham 1991 (31) (Cunningham2), FAO/WHO/UNU 1985 (28) (FAO/WHO/UNU), Jagim et al. (32) (Jagim), ten Haaf and Wejis based on BW (11) (tH&W1) and FFM (11) (tH&W2), De Lorenzo et al. (9) (De Lorenzo), and Wong et al. (10) (Wong).

Statistical Analysis

Comparison between measured and estimated RMR

The unpaired Student’s _t_-tests were used to compare the anthropometric and body composition variables between males and females and between modeling and cross-validation groups.

A one-way repeated-measures ANOVA with Bonferroni post hoc comparisons was used to compare the means of measured RMR and estimated RMR values. The post hoc analysis was fixed in the differences between the measured and the predicted RMR values. A simple linear regression analysis was used to test the agreement between measured and estimated RMR values by calculating the slope, intercept, and coefficient of determination (predicted _R_2, hereinafter called “_R_2”). Also, MD and root mean square error (RMSE) (33) were calculated. MD was calculated as follows:

MD=∑i=1nŶi−Yin

where n is the number of individuals, Ŷi is the measured value of _i_th individual, and Yi is the predicted value for the _i_th individual. Later, the MD was also expressed as a percentage (i.e., %MD). The RMSE was obtained using the following equation:

RMSE=∑i=1nŶi−Yi2n

The %RMSE was calculated by dividing the RMSE by the mean of measured RMR for each predictive equation (34). Moreover, it was calculated the percentage of subjects that presented the estimated RMR value within ±5% and ±10% of the measured RMR. The literature proposed these thresholds as an acceptable error comparing the measured and estimated RMR (35–37). To be considered as a “good equation,” all the three following criteria must be accomplished: a) no statistical difference (post hoc test, P ≥ 0.05) between measured and predicted RMR values, b) %MD ≤ ±10%, and c) %RMSE ≤ 10%.

Finally, the Bland–Altman method (38) was used to identify the bias and 95% limits of agreement between the measured and estimated RMR values for all equations. The data homoscedasticity was tested by simple linear regressions using slope and intercept of bias (measured − predicted RMR, _y-_axis) as a function of the mean of measured and predicted RMR (_x_-axis). A nonsignificant statistical difference means that slope and intercept are not different from zero. Thereby, the amount of error does not change significantly across the different RMR ranges.

New predictive equations proposal

Two different multiple linear regression models were designed to propose the two new RMR equations. First, 70% of the sample was randomly selected to develop the predictive models. After that, the proposed equations were tested (i.e., cross-validated) in the remaining 30% of the sample. Linear regression was used to calculate the R_2 and P value from measured (dependent) and estimated RMR values (independent) for both predictive (70%) and cross-validation (30%) groups. MD and RMSE were also calculated for both groups as aforementioned. The first equation (hereinafter “equation 1_”) was designed based on the classical predictors of RMR (i.e., sex, age, BW, and height) (39):

equation1=β0+β1sex+β2age+β3BW+β4height

where _β_0 is the intercept and _β_1 : _β_4 are the coefficients of each predictive variable. Sex was considered as a categorical (dummy) variable (male, 0; female, 1).

The second equation (complete model (CM)) was initially developed using the same classical predictors adding FM (kg), FFM (kg), and somatotype domains (u.a.) (endomorphy (Endo), mesomorphy (Meso), and ectomorphy (Ecto)). Besides, two-way interactions between BW and the other predictive variables were also included in the model:

CM=β0+β1sex+β2age+β3BW+β4height+β5FM+β6LBM+β7Ecto+β8Meso+β9Endo+β10BW·sex+β111BW·age+β12BW·height+β13BW·FM+β14BW·LBM+β15BW·Ecto+β16BW·Meso+β17BW·Endo

The CM was submitted to the stepwise method to determine the “ideal model” (hereinafter “equation 2_”_), or in other words, the model including the fewest number of variables and interactions. Both CM and equation 2 were compared using one-way ANOVA to confirm if the stepwise suggestion was the best fit (33). The adjusted coefficient of determination (adjusted-_R_2) was used to compare models with different predictors and confirm the most parsimonious model. Similarly, the ANOVA was also employed to compare equations 1 and 2. The residuals of estimated and measured RMR values from the modeling group (70% of the sample) were compared with residuals of validating group (30%) using the unpaired Student’s _t_-test. Finally, we compared the results of the new (i.e., equations 1 and 2) and the existing predictive equations based on a subsample of 31 subjects. Hence, we performed the same analysis used in the first step of the study, in which we compared the validity and the accuracy of the new proposed and existing equations (i.e., estimated RMR) versus the measured RMR.

The statistical analyses were performed using the R Core Team software (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA). The significance level was fixed at P ≤ 0.05. All data were presented as mean ± SD unless otherwise stated.

RESULTS

Comparison between measured and estimated RMR

Male athletes presented greater BW, height, and FFM (all P < 0.0001), endomorphy (_P_ = 0.03), and mesomorphy (_P_ < 0.001) compared with female athletes. Besides, there were no significant differences between the modeling and cross-validation groups for all subjects’ characteristics (_P_ > 0.05; Table 1).

TABLE 1 - Subjects’ characteristics by sex and group.

| | Female (n = 44), Mean ± SD (95% CI) | Male (n = 58), Mean ± SD (95% CI) | P | Modeling Group (n = 71), Mean ± SD (95% CI) | Validation Group (n = 31), Mean ± SD (95% CI) | P | | | ---------------------------------------- | ------------------------------------ | ------------------------- | ---------------------------------------------- | ------------------------------------------------ | -------------------------- | ---- | | Age (yr) | 25.7 ± 4.7 (24.3–27.2) | 24.6 ± 3.8 (23.6–25.0) | 0.18 | 25.4 ± 4.0 (24.5–26.4) | 24.2 ± 4.7 (22.5–26.0) | 0.19 | | BW (kg) | 64.3 ± 9.7 (61.4–67.3) | 85.3 ± 20.1 (80.0–90.6) | <0.0001 | 77.9 ± 21.0 (72.9–82.2) | 72.5 ± 15.0 (67.1–78.0) | 0.21 | | Height (cm) | 167.6 ± 8.0 (165.2–170.0) | 182.6 ± 8.1 (180.4–184.7) | <0.0001 | 176.1 ± 10.8 (173.6–178.7) | 176.1 ± 11.5 (171.9–180.3) | 0.98 | | FFM (kg) | 53.2 ± 6.1 (51.4–55.1) | 77.2 ± 14.3 (73.5–81.0) | <0.0001 | 68.1 ± 17.8 (63.9–72.3) | 64.0 ± 13.2 (59.2–68.9) | 0.25 | | FM (kg) | 11.1 ± 4.8 (9.6–12.5) | 8.0 ± 7.2 (6.1–9.9) | 0.08 | 9.7 ± 6.4 (8.2–11.2) | 8.5 ± 6.0 (6.3–10.7) | 0.37 | | Endomorphy (a.u.) | 3.2 ± 0.8 (2.9–3.4) | 2.4 ± 1.3 (2.0–2.7) | 0.03 | 2.8 ± 1.2 (2.6–3.1) | 2.4 ± 1.1 (2.0–2.8) | 0.09 | | Mesomorphy (a.u.) | 4.1 ± 1.2 (3.7–4.4) | 5.1 ± 1.8 (4.6–5.6) | 0.001 | 4.8 ± 1.8 (4.3–5.2) | 4.3 ± 1.2 (3.9–4.7) | 0.15 | | Ectomorphy (a.u.) | 2.2 ± 1.1 (1.9–2.5) | 2.3 ± 1.3 (1.9–2.6) | 0.91 | 2.1 ± 1.2 (1.8–2.4) | 2.6 ± 1.1 (2.2–3.0) | 0.06 |

Data presented as mean ± SD and 95% confidence interval (CI). P values come from unpaired Student’s _t_-test. Values in bold represent statistically significant differences (P ≤ 0.05).

The modeling and validation groups were composed by 52% and 45% of females, respectively.

For females (Table 2), from a group perspective, only tH&W1, Jagim, and Cunningham2 equations did not differ between measured and estimated RMR (P > 0.05). All RMR estimations presented a low MD (<148 kcal·d−1 or ≈9%) and RMSE (147–237 kcal·d−1 or 9%–15%). However, all the _R_2 values yielded by the different equations were relatively low (_R_2 = ≈35%–41%). From the individual perspective, the Cunningham2 predictive equation yielded the best accuracy within the ±5% and ±10% of measured RMR (43% and 73%, respectively).

TABLE 2 - Validity and accuracy of RMR equations in females (n = 44).

| | Mean ± SD (95% CI) | P Value (Post Hoc) | MD (SD) (kcal·d−1) | MD (SD) (%) | Linear Regression | RMSE (kcal·d−1) | RMSE (%) | ±5% Accurate (%) | ±10% Accurate (%) | | | | -------------------- | ---------------------- | ------------------ | ----------- | ----------------- | --------------- | ----------- | ---------------- | ----------------- | -- | -- | | R 2 | P | | | | | | | | | | | Measured RMR | 1577 ± 170 (1525–1629) | — | — | — | — | — | — | — | — | — | | Estimated RMR | | | | | | | | | | | | H&B (29) | 1490 ± 104 (1428–1491) | <0.0001 | 118 (135) | 7 (8) | 0.37 | <0.0001 | 180 | 12 | 30 | 59 | | tH&W1 (11) | 1573 ± 155 (1526–1620) | 0.10 | 34 (146) | 2 (9) | 0.36 | <0.0001 | 150 | 9 | 39 | 66 | | FAO/WHO/UNU (28) | 1429 ± 132 (1389–1469) | <0.0001 | 148 (132) | 9 (8) | 0.41 | <0.0001 | 200 | 13 | 25 | 57 | | De Lorenzo (9) | 1683 ± 165 (1633–1733) | <0.001 | −106 (151) | −7 (10) | 0.35 | <0.0001 | 185 | 12 | 36 | 75 | | Wong (10) | 1505 ± 127 (1466–1543) | 0.001 | 72 (138) | 4 (9) | 0.36 | <0.0001 | 156 | 10 | 36 | 68 | | Jagim (32) | 1645 ± 205 (1141–1706) | 0.10 | −68 (171) | −5 (11) | 0.36 | <0.0001 | 182 | 12 | 39 | 59 | | Cunningham1 (30) | 1650 ± 132 (1610–1690) | <0.001 | −73 (134) | −5 (9) | 0.40 | <0.0001 | 153 | 10 | 41 | 70 | | Cunningham2 (31) | 1519 ± 132 (1478–1559) | 0.06 | 59 (134) | 3 (8) | 0.40 | <0.0001 | 147 | 9 | 43 | 73 | | tH&W2 (11) | 1695 ± 139 (1653–1737) | <0.0001 | −118 (136) | −8 (9) | 0.40 | <0.0001 | 181 | 11 | 25 | 59 |

Data are presented as mean ± SD. H&B represents the equation developed by Harris and Benedict in 1918. tH&W1 and tH&W2 represents the equation designed by ten Haaf and Weijs in 2015, based on BW and FFM, respectively. Cunningham1 and Cunningham2 mean the equations developed by Cunningham in 1980 and 1991, respectively. The ±5% and ±10% accurate columns represent the thresholds that have been proposed in the literature as an acceptable error comparing the measured RMR (using IC) and the estimated RMR (using predictive equations). Values in bold represent significant differences (P ≤ 0.05).

CI, confidence interval; _R_2, coefficient of determination.

For males (Table 3), from a group perspective, estimated RMR values using the tH&W1, De Lorenzo, Cunningham1, and Cunningham2 predictive equations were not different from measured RMR (all P ≥ 0.1). Furthermore, these predictive equations also provided the lowest MD (<72 kcal·d−1 or ≈3%) and relatively low RMSE (212–240 kcal·d−1 or 10%–12%). On the other hand, Jagim’s equation presented the greater MD at the group level (−335 kcal·d−1 or ≈−17%; RMSE = 386 kcal·d−1) and at the individual level (7% and 24% of accuracy for ±5% and ±10% of measured RMR, respectively).

TABLE 3 - Validity and accuracy of RMR equations in males (n = 58).

| | Mean ± SD (95% CI) | P Value (Post Hoc) | MD (SD) (kcal·d−1) | MD (SD) (%) | Linear Regression | RMSE (kcal·d−1) | RMSE (%) | ±5% Accurate (%) | ±10% Accurate (%) | | | | -------------------- | ---------------------- | ------------------ | ----------- | ----------------- | --------------- | ----------- | ---------------- | ----------------- | -- | -- | | R 2 | P | | | | | | | | | | | Measured RMR | 2099 ± 400 (1994–2204) | — | — | — | — | — | — | — | — | — | | Estimated RMR | | | | | | | | | | | | H&B (29) | 1896 ± 291 (1909–2062) | <0.01 | 114 (199) | 4 (9) | 0.78 | <0.0001 | 227 | 11 | 36 | 67 | | tH&W1 (11) | 2082 ± 258 (2014–2150) | 0.10 | 17 (213) | −1 (10) | 0.77 | <0.0001 | 212 | 10 | 45 | 72 | | FAO/WHO/UNU (28) | 1975 ± 302 (1895–2054) | <0.01 | 125 (194) | 5 (9) | 0.78 | <0.0001 | 229 | 11 | 33 | 64 | | De Lorenzo (9) | 2046 ± 242 (1983–2110) | 0.84 | 53 (236) | 1 (10) | 0.71 | <0.0001 | 240 | 11 | 38 | 69 | | Wong (10) | 1969 ± 262 (1901–2038) | 0.0001 | 130 (210) | 5 (9) | 0.76 | <0.0001 | 245 | 12 | 29 | 64 | | Jagim (32) | 2435 ± 392 (2332–2538) | <0.0001 | −335 (194) | −17 (0) | 0.78 | <0.0001 | 386 | 18 | 7 | 24 | | Cunningham1 (30) | 2170 ± 309 (2089–2251) | 0.11 | −71 (209) | −5 (10) | 0.73 | <0.0001 | 219 | 10 | 50 | 71 | | Cunningham2 (31) | 2039 ± 309 (1957–2120) | 0.28 | 61 (209) | 2 (9) | 0.73 | <0.0001 | 216 | 10 | 50 | 78 | | tH&W2 (11) | 2243 ± 326 (2157–2329) | <0.0001 | −144 (207) | −8 (10) | 0.73 | <0.0001 | 251 | 12 | 29 | 59 |

Data are presented as mean ± SD. H&B represents the equation developed by Harris and Benedict in 1918. tH&W1 and tH&W2 represent the equation designed by ten Haaf and Weijs in 2015, based on BW and FFM, respectively. Cunningham1 and Cunningham2 mean the equations developed by Cunningham in 1980 and 1991, respectively. The ±5% and ±10% accurate columns represent the thresholds that have been proposed in the literature as an acceptable error comparing the measured RMR (using IC) and the estimated RMR (using predictive equations). Values in bold represent significant differences (P ≤ 0.05).

CI, confidence interval; _R_2, coefficient of determination.

According to the Bland–Altman plots, most of the available equations underestimated the measured RMR (Figs. 1, 2). Finally, predictive equations showed homoscedasticity data distribution for female athletes, except tH&W1, Jagim, and De Lorenzo equations (Fig. 1). On the other hand, both Cunningham’s and tH&W2’s equations presented data homoscedasticity for males (Fig. 2). We observed wider limits of agreement comparing the estimated with measured RMR values, ranging in average of 550 kcal·d−1 for females and 893 kcal·d−1 for males. Concretely, the De Lorenzo equation showed the widest range for females (from −402 to 191 kcal·d−1; Fig. 1), whereas the FAO/WHO/UNU equation showed the widest range for males (from −426 to 716 kcal·d−1; Fig. 2).

FIGURE 1:

Bland–Altman plots for measured and each predicted RMR in high-level female athletes. The solid line represents the MD between measured and predicted RMR values. Dashed lines show the upper and lower limits of agreement (95% confidence interval).

FIGURE 2:

Bland–Altman plots for measured and each predicted RMR in high-level male athletes. The solid line represents the MD between measured and predicted RMR values. Dashed lines show the upper and lower limits of agreement (95% confidence interval).

New predictive equations proposal

Following the aforementioned procedures (see the Statistical Analysis section), the multiple linear regressions models returned the following two new equations:

Equation1=729.50+175.64sex−7.23age+15.87BW+1.08height

Equation2=−2688.12+521.08sex+42.86age+18.98BW+16.76height+85.47Meso+140.54Endo−8.24BW·sex+1.53BW·Endo−0.65BW·age

The equation 1 explained the 85% of the variance in measured RMR (adjusted-_R_2 = 0.84; RMSE = 158 kcal·d−1). The CM presented a better performance, explaining the 89% of the variance in measured RMR (adjusted-_R_2 = 0.86; RMSE = 133 kcal·d−1). Besides, equation 2 presented a similar performance, with _R_2 = 0.89 and adjusted-_R_2 = 0.87, respectively (RMSE = 138 kcal·d−1). Equation 1 was statistically different from the CM (P = 0.01) and from equation 2 (P = 0.002). Conversely, equation 2 was not different from the CM (P = 0.91), confirming the stepwise analysis.

When both new proposed equations were used in the cross-validation group (n = 31), both yielded a good accuracy (_R_2 = 0.71 and 0.74, and RMSE = 200 and 192 kcal·d−1, for equation 1 and equation 2, respectively; Table 4). The comparison between the modeling and cross-validation groups’ residuals did not present statistical differences for equation 1 (P = 0.97) or equation 2 (P = 0.86). At the individual level, both new predicted equations provided the same accuracy for wider limit (61% of accuracy for ±10% of measured RMR), whereas equation 2 presented higher accuracy in the narrower limit (39% vs 32% for ±5% of measured RMR). Moreover, both equations provided a low MD (<15 kcal·d−1 or 2%) and did not present a proportional bias (increased bias as RMR increases/decreases). Besides, most athletes fell within the range of 10% of error, represented by ≈180 kcal·d−1 for this group (Table 4, Fig. 3).

TABLE 4 - Validity and accuracy of RMR equations in the validation group (n = 31).

| | Mean ± SD (95% CI) | P Value (Post Hoc) | MD (SD) (kcal·d−1) | MD (SD) (%) | Linear Regression | RMSE (kcal·d−1) | RMSE (%) | ±5% Accurate (%) | ±10% Accurate (%) | | | | -------------------- | ---------------------- | ------------------ | ----------- | ----------------- | --------------- | ----------- | ---------------- | ----------------- | -- | -- | | R 2 | P | | | | | | | | | | | Measured RMR | 1820 ± 379 (1681–1959) | — | — | — | — | — | — | — | — | — | | Estimated RMR | | | | | | | | | | | | H&B (29) | 1707 ± 285 (1602–1811) | 0.04 | 113 (205) | 5 (11) | 0.71 | <0.0001 | 231 | 13 | 26 | 52 | | tH&W1 (11) | 1808 ± 302 (1697–1919) | 0.10 | 11 (221) | −1 (12) | 0.66 | <0.0001 | 218 | 12 | 32 | 55 | | FAO/WHO/UNU (28) | 1680 ± 287 (1575–1785) | <0.01 | 139 (205) | 7 (11) | 0.71 | <0.0001 | 245 | 13 | 26 | 39 | | De Lorenzo (9) | 1856 ± 251 (1764–1948) | 0.10 | −36 (239) | −4 (12) | 0.62 | <0.0001 | 238 | 13 | 32 | 68 | | Wong (10) | 1717 ± 255 (1624–1811) | 0.10 | 103 (212) | 4 (11) | 0.72 | <0.0001 | 232 | 13 | 26 | 58 | | Jagim (32) | 2017 ± 436 (1831–2186) | 0.002 | −197 (237) | −11 (14) | 0.69 | <0.0001 | 304 | 17 | 26 | 42 | | Cunningham1 (30) | 1884 ± 286 (1779–1989) | 0.10 | −65 (219) | −5 (12) | 0.67 | <0.0001 | 225 | 12 | 35 | 65 | | Cunningham2 (31) | 1753 ± 286 (1648–1858) | 0.10 | 67 (219) | 2 (11) | 0.67 | <0.0001 | 226 | 12 | 35 | 55 | | tH&W2 (11) | 1942 ± 301 (1831–2052) | 0.04 | −122 (218) | −8 (12) | 0.67 | <0.0001 | 247 | 14 | 23 | 55 | | New equation 1 | 1817 ± 305 (1705–1929) | 0.10 | 4 (203) | −1 (11) | 0.71 | <0.0001 | 200 | 11 | 32 | 61 | | New equation 2 | 1835 ± 306 (1723–1947) | 0.10 | −15 (195) | −2 (11) | 0.74 | <0.0001 | 192 | 11 | 39 | 61 |

CI, confidence interval; _R_2, coefficient of determination.

FIGURE 3:

Relationship between measured and predicted RMR (kcal·d−1) for the cross-validation group and Bland–Altman plot (solid line, MD; dashed lines, upper and lower limits of agreement (95% confidence interval)) for equation 1 (left panel) and equation 2 (right panel).

From a subsample (n = 31) analysis, only H&B, FAO/WHO/UNU, Jagim, and tH&W2 estimated RMR were statistically different from the measured RMR. Overall, equation 2 yielded the best agreement compared with existing equations, which showed low MD (−15 kcal·d−1 or −2%), the highest coefficient of determination (_R_2 = 0.74), the lowest RMSE (192 kcal·d−1 or 11%), and 61% and 39% of accuracy for the wider and narrower limit (10% and 5% of measured RMR, respectively). Moreover, estimated RMR derived by equation 1 presented similar results for validity and accuracy compared with those results observed for equation 2.

DISCUSSION

The present study aimed a) to assess the validity of existing predictive RMR equations in high-level athletes and b) to propose new equations based on the current study group. From the existing equations analyzed, the Cunningham2 yielded the best performance for both male and female high-level and Olympic athletes. The new predictive equations proposed in the present study (equation 1 and 2) showed a good agreement at a group level (i.e., cross-validation group, n = 31; _R_2 = 0.71 and 0.74, RMSE = 200 and 192 kcal·d−1, respectively), and good accuracy (61% for the range of ±10% of measured RMR).

To the authors’ knowledge, this is the first study that underwent the proposed equations to a cross-validation process including males and females in the sample. Watson et al. (8) developed their proposed equations in a group of 44 collegiate female athletes and cross-validated them in a subsample of 22 athletes. The studies of Wong et al. (10), ten Haaf and Weijs (11), and Jagim et al. (32) also proposed new predictive RMR for athletes. However, none of them presented the cross-validation analyses for their equations. In general, robust studies in which the results are clear and presented in detail remain scarce in the literature. Moreover, even the traditional RMR equations (i.e., H&B, Cunningham1, Cunningham2, and FAO/WHO/UNU) did not validate their results against a subsample of similar characteristics (e.g., training status, body composition, etc.). Although equations compared in the present study are widely employed in clinical practice, their use in high-level athletes’ cohorts should be cautious because a significant amount of error has been verified in the literature compared with measured RMR (7,10,11,35,40). The Jagim’s (32) RMR prediction for male athletes obtained in the present study showed poor accuracy and agreement (Table 3, Fig. 2), with regard to MD (−335 kcal·d−1) and RMSE (386 kcal·d−1) values.

It is important to highlight that the RMR assessment is not an easy task in the athletes’ day-to-day routine because many factors may influence the gas exchange measurement, such as the acquisition protocol, the equipment used, and accomplishment with the best practices of measurement (17,18), which are recommended. In this regard, Freire et al. (26) found that the method of gas exchange data selection affects the RMR value in high-level athletes. Moreover, they observed that the measurement time could be reduced (i.e., from the usually 30 min to a “reduced protocol” of 20 min) and that the acclimation period before measurement in high-level athletes could be discarded. These results are especially important to sport center settings because high-level athletes generally do not have much availability to evaluate the RMR following the aforementioned recommendations (17,18). On the other hand, accomplishing the recommendation of 48 h without intense exercise or 24 h of moderate exercise is a challenge to sports scientists and coaches in real practice, as it may interfere with the training schedule and/or competitions.

Beyond the limitations to assess the RMR, Schofield et al. (40) reported that few predictive equations were designed with an athletes’ cohort, which, to a certain extent, may explain the limitation of the existing equations to predict the RMR in this population accurately. Perhaps De Lorenzo (9), ten Haaf and Weijs (11), and Wong (10) are the most successful studies to develop predictive equations based on an athletic cohort with heterogeneous body characteristics. These variables (e.g., BW, FM, and FFM) play an important role in the RMR value. We observed some differences for these variables from the aforementioned studies compared with our study. Comparing the BW from ten Haaf and Wejis’ (11) study with our cohort, it ranged from 52.8 to 102.3 kg and 49.6 to 156.3 kg, respectively. Success in sport is physical attributes dependent. Similarly, the RMR can also be affected by the specific body composition required for each sport, and as at least in absolute terms, the higher is the BW, the higher is the RMR. For example, Koshimizu et al. (41) compared the RMR values among athletes who participated in endurance sports (n = 24), strength/power/sprint sports (n = 23), and team sports (n = 34). They found statistical differences for the measured RMR among the type of sports, despite the fact that RMR was expressed as absolute terms (kcal·d−1) or as “relative” terms (i.e., RMR relative to BW (kcal·kg BW·d−1) or relative to lean body mass (kcal·kg lean body mass·d−1)). Once it is essential having an RMR predictive equation by sport and sex, it is challenging to obtain it, especially with high-level athletes. Therefore, we think that equations should be developed initially based on the most heterogeneous cohort possible with acceptable accuracy, as described in the present study.

In the present study, we developed two predictive equations based on classical variables (e.g., age, sex, BW, body composition, etc.) presented in the literature (39), followed by a cross-validation process. In the modeling group (n = 71), equation 1 provided RMSE values commonly found in the literature or lower (RMSE = 158 kcal·d−1) and high _R_2 values (0.85), which means that these traditional predictors can explain 85% of the variance in the RMR. The CM, beyond the classic variables, added the somatotype and the interaction between BW and other variables, improving the RMR prediction compared with equation 1, as indicated by ANOVA (P = 0.002), _R_2 value (0.89), and RMSE (135 kcal·d−1). However, it was not the most parsimonious model. Equation 2 variables and interactions also resulted in a significant improvement in the RMR prediction, compared with equation 1, as indicated by the _R_2 values (0.89 vs 0.85) and RMSE (138 vs 158 kcal·d−1) and by models’ ANOVA comparison (P = 0.02). It is important to highlight that the adjusted-_R_2 in the equation 2 (0.87) was almost the same as multiple _R_2 (0.89), which means that the inclusion of several variables in equation 2 kept the prediction power with reduced RMSE value using fewer variables but that played a truth “weight” in the model. Besides, the stepwise suggestion was confirmed by comparing both models (CM vs equation 2) by ANOVA analysis, in which they were not statistically different in between (P = 0.67).

The RMSE and _R_2 values were slightly different at comparing the modeling and cross-validation groups. In equation 1, an increase in the RMSE (158 vs 200 kcal·d−1) and a decrease in the _R_2 values (0.85 vs 0.71) were observed. The same results were found for equation 2 (RMSE = 138 vs 192 kcal·d−1, _R_2 = 0.89 vs 0.75). It can be explained, at least partially, by the sample size used in the different steps of the validation process (71 vs 31 subjects). Because we had a relatively smaller sample size in the cross-validation group, lower precision and accuracy results could be expected. To avoid this sample size effect, we compared the estimated RMR derived by the new two proposed equations and those previously proposed by other authors versus the measured RMR (Table 4). Overall, the new equations 1 and 2 presented similar validity, agreement, and accuracy results than existing equations, showing higher _R_2 values and lower RMSE. Of note, an important issue to be considered is the heterogeneous characteristic of our sample regarding BW (i.e., from 49.6 to 156.3 kg) and the sports in which athletes participated (i.e., 21 different sports). Therefore, this heterogeneity may, to a greater or lesser extent, explain why both new equations provided good and promising results and estimated RMR values compared with the existing equations. The use of the currently proposed equations in future studies with high-level athletes may support this hypothesis.

Regarding the contribution of anthropometric and body composition variables explaining the RMR, Johnstone et al. (39) stated that 63.0% and 6.7% of the between-subject variance in the RMR are explained by FFM and FM, respectively. The age was not significant in the model (contributed with 1.7%). Our study found different results once the BW played the most important role in the model, whereas FFM was not significant because it was removed by the stepwise regression and confirmed by ANOVA’s comparison. On the other hand, the mesomorphic component of the somatotype and interactions with endomorphy accounted for increasing the robustness of the new proposed equation. Unlike Cunningham’s findings (30,31), the FFM itself was not a significant predictor for RMR in our sample of high-level athletes, although the mesomorphy may represent this variable to a certain extent. We believe that using somatotype instead of body composition variables (e.g., FM and FFM) is valuable because it is obtained with simpler measurements than others (e.g., dual-energy x-ray absorptiometry (DXA), skinfolds, bioimpedance, etc.). It does not involve expensive equipment and yields smaller technical measurement errors when assessed by expert researchers/technicians (i.e., breadth, girth, and four skinfolds vs several skinfolds), especially for heavy athletes (42,43).

Furthermore, different techniques were employed to determine the subjects’ body composition in the existing equations selected in the present study. The DXA has been recognized as the standard indirect method to determine body composition in athletes because of its reliability and repeatability (44). However, the International Olympic Committee (45) and Kasper et al. (44) suggest that the skinfold method may be preferable for athletes’ body composition assessment because of its simpler implementation, besides able practitioners to follow up the body composition alongside the season with low cost and time (44). Indeed, using these different methods to determine the body composition in athletes will affect the RMR estimative in equations that use these variables in their predictions. Therefore, it becomes difficult to compare different studies that used diverse samples with different ages, body composition outcomes, athletic level, and/or sport. For instance, Korth et al. (46) investigated the effect of different body composition methods (four-compartment (lipids, water, mineral, and protein) and two-compartment methods (e.g., skinfolds, DXA, etc.)) upon the RMR estimative. Although they found significant differences in the RMR estimating, they considered that the body composition method (four vs two-compartment) had a minor influence on the accuracy of RMR estimative (accuracy determined as the agreement between the measured and the estimated RMR). In addition, they suggested that the two-compartment methods may have some advantages over other methods because they are more feasible and are less time-consuming and thus easier to implement in the clinical context.

In contrast to common sense, FFM was not included as a predictor of the RMR in the present model. Indeed, the stepwise analysis selected “BW” and “mesomorphy” as predictors rather than FFM to be included in the model. It could be explained by the high collinearity between BM and FFM (r = 0.95). Nevertheless, the inclusion of BW instead of FFM plays a favorable role in the new proposed equation, mostly because of the feasibility of measuring BM over estimating FFM. Some issues are raised for estimating the FFM, primarily those concerning the method used (e.g., skinfolds, DXA, etc.), as previously discussed. Balci et al. (35) tested the accuracy of most used predictive equations in athletes and sedentary people, and Cunningham’s equation did not present the best performance to predict the RMR on this sample. In the present study, Cunningham2 showed the best performance among all predictive equations for group and individual levels (Tables 2, 3). Moreover, Jagim’s equation did not include the FFM in their model, which is in line with our present study.

Besides, there is no consensus in the literature regarding the best existing equation to predict RMR in high-level athletes. For example, the H&B’s equation provided better performance for a group of Turkish (males and females; RMSE = 252 kcal·d−1; _R_2 = 0.40), female Malaysian Olympic young athletes (10) from different sports (RMSE = 131 kcal·d−1; _R_2 = 0.42), and for male national collegiate athletes (47) (RMSE = 284 kcal·d−1; _R_2 = 0.51). In the present study, this equation showed similar performance (RMSE = 180 kcal·d−1, MD = 118 kcal·d−1 for females; RMSE = 227 kcal·d−1, MD = 114 kcal·d−1 for males). The Cunningham1 equation yielded the best performance in a group of Dutch recreational athletes from several sports (11) (RMSE = 145 kcal·d−1) and female national collegiate athletes (RMSE = 110 kcal·d−1; _R_2 = 0.53). The tH&W2 equation was the best choice for female rugby players (36), whereas the De Lorenzo equation showed the best performance among male Malaysian Olympic young adults (9) (RMSE = 168 kcal·d−1; _R_2 = 0.58). It can be attributed to different levels of athletes’ performance, body composition method used, and metabolic cart and/or method (i.e., gas exchange data selection; e.g., (26,48), used to assess the RMR.

The present study has some limitations. First, the menstrual cycle of female athletes was not controlled. In a systematic review with a meta-analysis study, Benton et al. (49) found a small and not statistically significant effect (effect size = 0.23, P = 0.055) of the menstrual cycle upon RMR analyzing a subgroup of more recent studies (after the year 2000). Beyond, high-level athletes have a strict schedule to perform the assessments, being challenged to adequate the assessment to their menstrual cycle. Nevertheless, and considering the uniqueness and interest of our study sample, we opted to include the female athletes in the analyses, taking into account this possible limitation. Second, the method used to estimate the body composition (skinfolds) has some disadvantages compared with more sophisticated methods, such as underwater weighing, air-displacement plethysmography, and/or DXA, an issue that may explain (at least partially) why we observed a poor accuracy with equations that included body composition parameters. However, the evaluators were certified by the International Society for the Advancement of Kinanthropometry (level 1) with more than 5 yr of experience and more than 200 body composition assessments performed. Beyond this limitation, this study intends to help the athletes’ staff in the field. Hence, RMR prediction must not be dependent on sophisticated equipment, in our view.

CONCLUSIONS

The new predictive equations provided a superior coefficient of determination (_R_2 = 0.71–0.74), lower RMSE (192–200 kcal·d−1), and similar individual accuracy (61% for ±10% of the measured RMR error) compared with the leading equations presented in the literature. It is important to highlight that this is the only known study that cross-validated the proposed predictive equations for high-level athletes of both sexes. Furthermore, from the existing equations, the Cunningham2 is the best option for female and male high-level athletes, but its accuracy may be affected by the choice of body composition method. The Jagim equation should not be used with high-level male athletes. Lastly, the authors suggest that equation 2 should be used to predict the RMR in high-level athletes whose somatotype has been evaluated, whereas equation 1 may be used when not.

We thank all athletes who participated in the study. The study was funded by the Brazil Olympic Committee and the Brazilian Funding Authority for Studies and Projects. J. M. A. A. is supported by the Spanish Ministry of Education (FPU 15/04059) and by the University of Granada, Plan Propio de Investigación 2020 Programa de Contratos Puente.

The authors declare no conflict of interest. The results of the present study do not constitute an endorsement by the American College of Sports Medicine. The authors declare that the results of the study results are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.

REFERENCES

1. Burke LM. Energy needs of athletes. Can J Appl Physiol. 2001;26(suppl):S202–19.

2. Lam YY, Ravussin E. Indirect calorimetry: an indispensable tool to understand and predict obesity. Eur J Clin Nutr. 2017;71(3):318–22.

3. Owen OE. Resting metabolic requirements of men and women. Mayo Clin Proc. 1988;63(5):503–10.

4. Livingston EH, Kohlstadt I. Simplified resting metabolic rate-predicting formulas for normal-sized and obese individuals. Obes Res. 2005;13(7):1255–62.

5. Frankenfield DC. Impact of feeding on resting metabolic rate and gas exchange in critically ill patients. JPEN J Parenter Enteral Nutr. 2019;43(2):226–33.

6. Loureiro LL, Fonseca S Jr, Castro NG, Dos Passos RB, Porto CP, Pierucci AP. Basal metabolic rate of adolescent modern pentathlon athletes: agreement between indirect calorimetry and predictive equations and the correlation with body parameters. PLoS One. 2015;10(11):e0142859.

7. Reale RJ, Roberts TJ, Lee KA, Bonsignore JL, Anderson ML. Metabolic rate in adolescent athletes: the development and validation of new equations, and comparison to previous models. Int J Sport Nutr Exerc Metab. 2020;30(4):249–57.

8. Watson AD, Zabriskie HA, Witherbee KE, Sulavik A, Gieske BT, Kerksick CM. Determining a resting metabolic rate prediction equation for collegiate female athletes. J Strength Cond Res. 2019;33(9):2426–32.

9. 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. 1999;39(3):213–9.

10. Wong JE, Poh BK, Nik Shanita S, et al. Predicting basal metabolic rates in Malaysian adult elite athletes. Singapore Med J. 2012;53(11):744–9.

11. ten Haaf T, Weijs PJM. Resting energy expenditure prediction in recreational athletes of 18–35 years: confirmation of Cunningham equation and an improved weight-based alternative. PLoS One. 2014;9(10):e108460.

12. Compher C, Cato R, Bader J, Kinosian B. Harris–Benedict equations do not adequately predict energy requirements in elderly hospitalized African Americans. J Natl Med Assoc. 2004;96(2):209–14.

13. De Lorenzo A, Tagliabue A, Andreoli A, Testolin G, Comelli M, Deurenberg P. Measured and predicted resting metabolic rate in Italian males and females, age 18–59 y. Eur J Clin Nutr. 2001;55(3):208–14.

14. Rodriguez NR, DiMarco NM, Langley S. Position of the American Dietetic Association, Dietitians of Canada, and the American College of Sports Medicine: nutrition and athletic performance. J Am Diet Assoc. 2009;109(3):509–27.

15. Thompson J, Manore MM. Predicted and measured resting metabolic rate of male and female endurance athletes. J Am Diet Assoc. 1996;96(1):30–4.

16. Mountjoy M, Sundgot-Borgen JK, Burke LM, et al. IOC consensus statement on Relative Energy Deficiency in Sport (RED-S): 2018 update. Br J Sports Med. 2018;52(11):687–97.

17. Compher C, Frankenfield D, Keim N, Roth-Yousey L. Best practice methods to apply to measurement of resting metabolic rate in adults: a systematic review. J Am Diet Assoc. 2006;106(6):881–903.

18. Fullmer S, Benson-Davies S, Earthman CP, et al. Evidence analysis library review of best practices for performing indirect calorimetry in healthy and non-critically ill individuals. J Acad Nutr Diet. 2015;15(9):1417–46.

19. Loucks AB. Energy balance and body composition in sports and exercise. J Sports Sci. 2004;22(1):1–14.

20. Capling L, Beck KL, Gifford JA, Slater G, Flood VM, O’Connor H. Validity of dietary assessment in athletes: a systematic review. Nutrients. 2017;9(12):1313.

21. Gissis I, Papadopoulos C, Kalapotharakos VI, Sotiropoulos A, Komsis G, Manolopoulos E. Strength and speed characteristics of elite, subelite, and recreational young soccer players. Res Sports Med. 2006;14(3):205–14.

22. Siri WE. Body composition from fluid spaces and density: analysis of methods. Nutrition. 1993;9(5):480–91.

23. Jackson AS, Pollock ML. Generalized equations for predicting body density of men. Br J Nutr. 1978;40(3):487–504.

24. Jackson AS, Pollock ML, Ward A. Generalized equations for predicting body density of women. Med Sci Sports Exerc. 1980;12(3):175–81.

25. Carter JEL. The Heath–Carter Anthropometric Somatotype. San Diego (CA): San Diego State University; 2002.

Cited Here

26. Freire R, Alcantara JMA, Hausen M, Itaborahy A. The estimation of the resting metabolic rate is affected by the method of gas exchange data selection in high-level athletes. Clin Nutr ESPEN. 2021;41:234–41.

27. Weir JB. New methods for calculating metabolic rate with special reference to protein metabolism. J Physiol. 1949;109(1–2):1–9.

28. FAO/WHO/UNU. Energy and Protein Requirements. Geneva, Switzerland: World Health Organization; 1985.

Cited Here

29. Harris JA, Benedict FG. A biometric study of human basal metabolism. Proc Natl Acad Sci U S A. 1918;4(12):370–3.

30. Cunningham JJ. A reanalysis of the factors influencing basal metabolic rate in normal adults. Am J Clin Nutr. 1980;33(11):2372–4.

31. Cunningham JJ. Body composition as a determinant of energy expenditure: a synthetic review and a proposed general prediction equation. Am J Clin Nutr. 1991;54(6):963–9.

32. Jagim AR, Camic CL, Askow A, et al. Sex differences in resting metabolic rate among athletes. J Strength Cond Res. 2019;33(11):3008–14.

33. Krzanowski W. An Introdution to Statistical Modelling. New York (NY): Oxford University Press Inc.; 1998. p. 252.

Cited Here

34. Malek MH, Berger DE, Housh TJ, Coburn JW, Beck TW. Validity of V̇O2max equations for aerobically trained males and females. Med Sci Sports Exerc. 2004;36(8):1427–32.

35. Balci A, Badem EA, Yılmaz AE, et al. Current predictive resting metabolic rate equations are not sufficient to determine proper resting energy expenditure in Olympic Young Adult National Team Athletes. Front Physiol. 2021;12:625370.

36. O’Neill JERG, Walsh CS, McNulty SJ, et al. Resting metabolic rate in female rugby players: differences in measured versus predicted values. J Strength Cond Res. 2020;36(3):845–50.

37. Amaro-Gahete FJ, Sanchez-Delgado G, Alcantara JMA, et al. Congruent validity of resting energy expenditure predictive equations in young adults. Nutrients. 2019;11(9):223.

38. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–10.

39. Johnstone AM, Murison SD, Duncan JS, Rance KA, Speakman JR. Factors influencing variation in basal metabolic rate include fat-free mass, fat mass, age, and circulating thyroxine but not sex, circulating leptin, or triiodothyronine. Am J Clin Nutr. 2005;82(5):941–8.

40. Schofield KL, Thorpe H, Sims ST. Resting metabolic rate prediction equations and the validity to assess energy deficiency in the athlete population. Exp Physiol. 2019;104(4):469–75.

41. Koshimizu T, Matsushima Y, Yokota Y, et al. Basal metabolic rate and body composition of elite Japanese male athletes. J Med Invest. 2012;59(3–4):253–60.

42. Xiaoyuan L, Wenhui L, Youfeng W, et al. Obesity and Heath–Carter somatotyping of 3438 adults in the Xinjiang Uygur Autonomous Region of China by multivariate analysis. Diabetes Metab Syndr Obes. 2021;15(14):659–70.

43. Campa F, Silva AM, Talluri J, Matias CN, Badicu G, Toselli S. Somatotype and bioimpedance vector analysis: a new target zone for male athletes. Sustainability. 2020;12(11):4365.

44. Kasper AM, Langan-evans C, Hudson JF, et al. Come back skinfolds, all is forgiven: a narrative review of the efficacy of common body composition methods in applied sports practice. Nutrients. 2021;13(4):1075.

45. Ackland TR, Lohman TG, Sundgot-Borgen J, et al. Current status of body composition assessment in sport: review and position statement on behalf of the ad hoc research working group on body composition health and performance, under the auspices of the I.O.C. Medical Commission. Sports Med. 2012;42(3):227–49.

46. Korth O, Bosy-Westphal A, Zschoche P, Glüer CC, Heller M, Müller MJ. Influence of methods used in body composition analysis on the prediction of resting energy expenditure. Eur J Clin Nutr. 2007;61(5):582–9.

47. Jagim AR, Camic CL, Kisiolek J, et al. Accuracy of resting metabolic rate prediction equations in athletes. J Strength Cond Res. 2018;32(7):1875–81.

48. Alcantara JMA, Sanchez-Delgado G, Amaro-Gahete FJ, Galgani JE, Ruiz JR. Impact of the method used to select gas exchange data for estimating the resting metabolic rate, as supplied by breath-by-breath metabolic carts. Nutrients. 2020;12(2):487.

49. Benton MJ, Hutchins AM, Dawes JJ. Effect of menstrual cycle on resting metabolism: a systematic review and meta-analysis. PLoS One. 2020;15(7):e0236025.

Keywords:

RESTING ENERGY EXPENDITURE; REE; ENERGY AVAILABILITY; SPORT; INDIRECT CALORIMETRY; PREDICTIVE EQUATIONS