Prediction equations for resting energy expenditure in overweight and normal-weight black and white children

¹ From the Unit on Growth and Obesity, Developmental Endocrinology Branch, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD (JRM, DCA-W, JE, ENS, EMF, and JAY); the Children’s Hospital of Philadelphia (AMT); Children’s Hospital of Pittsburgh (SAA); and Pennington Biomedical Research Center (JPD and GAB)

² Supported by RO1 HD27503 (SAA) K24 HD01357 (SAA), MO1-RR00084 (SAA), HD-28020 (JPD), ZO1 HD-00641 (JAY) and by the National Center for Minority Health and Health Disparities (JAY).

³ Reprints not available. Address correspondence to JR McDuffie, Department of Nutrition, School of Public Health, Campus Box #7461, McGavran-Greenberg, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599. E-mail: mcduffj{at}unc.edu.

ABSTRACT  
Background: Accurate estimation of children’s resting energy expenditure (REE) is important for planning dietary therapy.

Objective: Our objective was to compare the utility of 5 REE prediction equations in a diverse sample of young children.

Design: REE was obtained in 502 black and white girls and boys aged 6–11 y by using indirect calorimetry at 4 US sites. Measured REE and REE predicted from the equations were compared.

Results: None of the equations provided both accurate and unbiased estimates of REE. Two new sets of sex-specific equations including race as a factor were generated and evaluated. One set used easily measured variables—females: REE = 0.046 x weight – 4.492 x 1/height² – 0.151 x race + 5.841; males: REE = 0.037 x weight – 4.67 x 1/height² – 0.159 x race + 6.792—and accounted for 72% and 69%, respectively, of REE variance. The other set used body-composition variables—females: REE = 0.101 x fat-free mass + 0.025 x fat mass + 0.293 x height³ – 0.185 x race + 1.643; males: REE = 0.078 x fat-free mass + 0.026 x fat mass – 2.646 x 1/height² – 0.244 x race + 4.8—and accounted for 75% and 71%, respectively, of REE variance. When split by race and adiposity, the small bias generated could be corrected to within 0.25 MJ (60 kcal) of the mean measured value.

Conclusion: Sex-specific equations must take race into account to predict REE adequately in children.

Key Words: Child • prediction equation • energy expenditure • metabolic rate • race • obesity

INTRODUCTION  
Resting energy expenditure (REE) is the largest component of total daily energy expenditure, accounting for 60–70% of total daily expenditure (1). Therefore, the ability to estimate REE accurately is of the utmost importance for adequate dietary therapy. The gold standard for the measurement of REE is indirect calorimetry, but the equipment required to measure respiratory exchange makes this procedure time-consuming, costly, and often unavailable. To avoid both the need to determine REE directly and the problems caused by the great variability between measurement conditions (2), several prediction equations were developed in the first half of the 20th century (3–5). However, studies suggest that they generally overestimate children’s REE by as much as 20% (6, 7).

In response to the inaccuracy of these equations, new equations for predicting REE in children were developed. Using direct measurements of REE from >5000 generally healthy, normal-weight children and adolescents, the FAO/WHO/UNU published revised equations in 1985 (8). Schofield et al (9) published a different set of equations from a larger meta-analysis. Maffeis et al (6) compared the new FAO/WHO/UNU equations with older equations and concluded that the FAO/WHO/UNU equations also overestimated REE; thus, Maffeis et al developed a new equation based on a sample of 130 obese and nonobese children aged 6–10 y. Similarly, Molnar et al (7) measured the REE of 371 obese and nonobese Eastern European children aged 10–16 y, Tverskaya et al (10) measured the REE of 110 obese children aged 3–18 y, and then both groups proposed their own equations for estimation of REE.

Previous studies that tested the accuracy of REE equations concluded that the equations of the FAO/WHO/UNU or of Schofield et al were the best of the available equations for calculating REE in children and adolescents (11–13). However, those studies were limited by their use of small (n < 120) and racially homogeneous samples. The ability of these relatively new equations to predict REE has not been systematically compared in a large, racially diverse sample of overweight and normal-weight children. Sample diversity is important because the REE of blacks is, on average, significantly lower than that of whites of similar body weight (14–19), although one investigator proposed that this difference may be explained by the higher REE of white boys (20). In addition, an FAO/WHO/UNU Expert Committee concluded that a single equation cannot be used to estimate the REE of boys and girls across all racial groups and that, to avoid making clinically significant errors in the estimations of energy requirements, new equations should be derived for some races (21). The purpose of the present study was to evaluate the accuracy of the abovementioned equations to estimate actual REE in 502 black and white 6–11-y-old children whose REE was measured according to stringent procedural guidelines. After finding that none of the existing equations were acceptably accurate and unbiased in all race and sex subgroups, we developed new, sex-specific equations.

SUBJECTS AND METHODS  
Subjects
We studied 212 black and 290 white children aged 6–11 y, of whom 311 were girls and 191 were boys (Table 1). Race was self-reported. At the National Institutes of Health (NIH), all subjects reported all 4 grandparents to be of the same race. At other sites, the interviewer confirmed race to the generation of grandparents or great-grandparents. Of the 502 subjects, 189 (37.6%) were normal-weight [body mass index (BMI; in kg/m²) between the 5th and 84th percentile], 55 (10.9%) were at risk of overweight (BMI between the 85th and 95th percentile), and 258 (51.4%) were overweight (BMI above the 95th percentile) for age and sex. The data on 176 NIH subjects were collected from children recruited from the metropolitan Washington, DC, area. The data on 325 subjects were obtained from 4 of the authors (AT, SA, JD, GB) who performed their measurements of REE at 3 other sites (136 subjects from Philadelphia, 69 from Pittsburgh, and 121 from Baton Rouge, LA) under essentially the same rigorous conditions as were used at the NIH. All subjects underwent a general history and physical examination to rule out major medical illnesses. Subjects were asked to follow their usual diet during the week preceding the study. All subjects were studied in the early morning after an overnight fast and were instructed to void before measurements were obtained. The institutional review boards of all participating centers approved the REE protocols. Written informed consent was obtained from subjects and their parents, as required.

View this table:
TABLE 1. Subject characteristics¹

 
Body composition
Heights and weights were obtained by using calibrated stadiometers and digital scales. These measurements were made while the subject wore minimal clothing and no shoes. Body composition was assessed with the use of dual-energy X-ray absorptiometry (DXA) on the Hologic QDR-2000 (Hologic, Waltham, MA) in the pencil-beam mode at 3 of the sites—NIH (Bethesda, MD), The Children’s Hospital of Philadelphia, and the Pennington Biomedical Research Center—and with the use of the Lunar #LNR0528 (General Electric, Madison, WI) at the Children’s Hospital of Pittsburgh. Cross-calibration between instruments was not performed; however, the instruments were periodically calibrated in the same way at each institution by using a commercially available anthropomorphic phantom or other block of tissue-equivalent material. This procedure included testing of the high-voltage setting, the longitudinal and transverse movements of the scan arm, the operation of the shutter mechanism, and the accuracy and precision of the detector system. A trained observer determined and reported Tanner stage as a composite average score. In girls, this was the average of the Tanner stage from each of the 2 breasts and the pubic hair Tanner stage. In boys, this was the average of the Tanner stages for public hair and genital development.

Resting energy expenditure
REE was assessed by using open-circuit indirect calorimetry performed with the use of a respiratory metabolic cart (SensorMedics 2900 or Deltatrac; SensorMedics Corp, Yorba Linda, CA). Before each test, each of the calorimeters was calibrated with the use of a reference gas mixture (96% oxygen and 4% carbon dioxide) obtained from the manufacturer. At each institution, the metabolic cart was also checked by using alcohol burns or a gas mixture composed to simulate the effects of a methanol burn. Measurements on both outpatients and inpatients were completed in the morning after a 12-h fast and a minimum 30-min rest period. The REE was measured in a quiet, thermoneutral room with the subject resting comfortably in a supine position and watching children’s videos. A clear plastic ventilated hood was placed over the subject’s head to allow the sampling of respiratory gases. Measurements were recorded at 1-min intervals for a minimum of 30 min. Results from the first 5–10 min (during which time the children adjusted to the procedural environment), from any minutes during which the machine was adjusted after equilibration, and from any minutes associated with movement or with a respiratory quotient (RQ) exceeding 0.95 were excluded from the analysis. Research assistants monitored the subject throughout the REE measurement at each institution and noted the minutes associated with movement so that the measurements from those minutes could be deleted. The minimum number of usable minutes of data to allow the calculation of REE was 15. On occasion, data from 30 min were collected to obtain 15–20 valid data points. Review of 26 charts (NIH data) showed that the mean number of usable minutes was 17.2 ± 1.9 (median: 17; modal value: 15), and that 1.6 ± 2.1 readings were deleted because of an RQ > 0.95. The remaining measurements were averaged, and REE was calculated by using the equation of de Weir (22).

REE prediction equations
We compared measured REE with REE predicted by each of 5 equations developed with the use of pediatric populations (Table 2). The equations of interest were the 1985 versions of the equations of the FAO/WHO/UNU (8) and of Schofield et al (9) that were based on both weight and height and the equations proposed by Molnar et al in 1995 (7), Tverskaya et al in 1998 (10), and Maffeis et al in 1993 (6). These prediction models will subsequently be referred to as the FAO/WHO/UNU, Schofield, Molnar [both 1 (ie, general) and 2 (ie, sex-specific)], Tverskaya, and Maffeis equations, respectively. The Molnar 1, Molnar 2, Tverskaya, and Maffeis equations were developed specifically to improve estimation of REE in an overweight population. The FAO/WHO/UNU and Schofield equations were developed from predominantly normal-weight samples. The Harris-Benedict equation was also included because it is the most widely used of the earlier equations (4). Although often recommended for children, the Fleisch equation (23) was not included because it was specifically designed for use in hospitalized (ill) pediatric populations.

View this table:
TABLE 2. Equations for resting energy expenditure (REE)¹

 
Statistical analysis
Parametric data were analyzed on a Macintosh G3 Power PC (Apple Computers, Cupertino, CA) by using STATVIEW and SUPERANOVA software (versions 5.0.1 and 1.11, respectively; Abacus Concepts Inc, Berkeley, CA). All results were expressed as means ± SDs. Methods employed to assess agreement were: Bland-Altman comparisons (24), simple regression, and analysis of variance (ANOVA) with repeated measures using sex, race, and BMI percentile group (<95th and >95th) as between-group factors. The Bland-Altman approach plots the difference between 2 methods against the average value for the 2 methods. For Bland-Altman comparisons, equations showing significant correlation (or slope) between the difference (actual REE-predicted REE) and the mean of the actual and predicted REE were described as showing magnitude bias, because such correlations indicate that the error increases when the best estimates of the individual REEs are increasingly different from actual measured mean REE for the sample. Significant ANOVAs were followed by paired t tests. The significance of t test results was adjusted by using the Bonferroni-Holm correction for multiple comparisons. A value of P 0.05 was defined as significant.

Two sets of prediction equations for males and females, one using height and weight and one using body composition via DXA, were developed through multiple regression on the entire sample with actual measured REE as the dependent variable and age, Tanner stage, race, height, height², height³, 1/height, 1/height², 1/height³, weight, weight², weight³, BMI, BMI², BMI³, fat mass, and fat-free mass as the tested, potential independent variables. In all models, neither Tanner stage nor age significantly improved prediction (P > 0.3). The prediction equations developed from the present data were cross-validated to obtain a truer estimate of the error by using a 90%/10% split procedure (25–27). For the 90%/10% split procedure, a prediction equation was developed in a randomly selected 90% of the sample and tested in the remaining 10% of the sample that had been excluded from the equation development. This procedure was carried out >25 times for each of the 4 equations developed; this degree of repetition within the sample approaches the degree of specificity provided by the bootstrap method of error estimation (28). The new prediction equations developed were also tested by using the Bland-Altman method to ascertain the degrees of systematic and magnitude bias present when the REE predicted by the new equations was compared with the measured REE.

RESULTS  
REE was measured in all 502 subjects by indirect calorimetry. Analysis of variance of the measured REE by site found no systematic differences attributable to the center where the data were collected; therefore, results from all 4 sites were combined for subsequent analyses. Contemporaneously obtained anthropometric and demographic data were available from which to predict REE in all subjects but the 6 for whom DXA results were not available. Therefore, the Tverskaya equation (10) was evaluated in 496 subjects. The measured REE of black subjects was significantly less than that of white subjects after analysis of covariance adjustment for total fat mass, total fat-free mass, and sex (5.57 ± 1.13 compared with 5.85 ± 1.12; P < 0.0001).

In the sample as a whole, compared with measured REE, only the REE estimated by the FAO/WHO/UNU equation (Figure 1) appeared not to show a significant increase in slope (ie, the measurement error by which actual REE departed from the mean measured REE, subsequently referred to as magnitude bias). However, the FAO/WHO/UNU equation showed systematic bias, significantly underestimating the mean of the measured REE by –0.215 ± 0.825 MJ (–51 ± 197 kcal, or 3.7%; P < 0.0001). All of the other equations exhibited a significant magnitude bias (Figure 1). Most of these other equations also significantly underestimated (all: P < 0.0001) the mean REE of the total sample—Harris-Benedict equation: –0.19 ± 1.51 MJ (3.4%); Molnar 1 equation: –0.40 ± 1.41 MJ (7.0%); Molnar 2 equation: –0.28 ± 1.50 MJ (4.8%); Tverskaya equation: –0.24 ± 0.77 MJ (4.3%); and Maffeis equation: –0.65 ± 1.49 MJ (11.2%). The Schofield equation significantly overestimated the mean REE (mean difference: 0.28 ± 1.73 MJ, or 4.8%; P < 0.0001).

View larger version (52K):
FIGURE 1.. Bland-Altman comparisons of equations examined in the total sample. REE, resting energy expenditure; FAO, FAO/WHO/UNU; avg, average; HB, Harris-Benedict. Mean differences: FAO/WHO/UNU equation (8), –0.215 ± 0.825 MJ; Harris-Benedict equation (4), –0.195 ± 1.506 MJ; Molnar 1 equation (7), –0.397 ± 1.412 MJ; Schofield equation (9), 0.275 ± 1.730 MJ; Tverskaya equation (10), –0.242 ± 1.546 MJ; and Maffeis equation (6), –0.652 ± 1.492 MJ (all P < 0.0001). Magnitude biases: FAO/WHO/UNU equation, –0.047 (P = 0.1418); Harris-Benedict equation, –0.288 (P < 0.0001); Molnar 1 equation, –0.096 (P = 0.0005); Schofield equation, 0.210 (P < 0.0001); Tverskaya equation, –0.394 (P < 0.0001); and Maffeis equation, –0.559 (P < 0.0001) All R² values were significant (P < 0.0001) except that from the FAO/WHO/UNU equation.

 
When the data were split by both race and sex, none of the equations accurately predicted REE for all 4 subgroups without significant magnitude bias, and the error increased as the individual data points departed from each mean (Table 3). The FAO/WHO/UNU equation provided an accurate estimate of the mean REE for white males (P > 0.5), but it showed a small, but significant magnitude bias in this group (r = 0.225, P < 0.05). FAO/WHO/UNU underestimated the REE of black and white females, overestimated the REE of black males, and showed a strong magnitude bias in all 3 of these groups. The Harris-Benedict equation accurately estimated the mean REE of black females but underestimated that of white females and showed a significant magnitude bias in both groups. It also underestimated the mean REE of both black and white males, although there was a smaller, but still significant magnitude bias among all males. The Molnar 1 and Molnar 2 equations both underestimated the mean REE of males without magnitude bias. The Molnar 2 equation accurately estimated the mean REE of white females but not that of black females, and it showed magnitude bias in both groups. The Molnar 1 equation showed both types of error among females. Conversely, the Schofield equation overestimated the mean REE of females with less magnitude bias, but it showed both types of error among males. The Tverskaya equation provided an adequate estimate of mean REE for white females only, but this measurement also exhibited magnitude bias. The Maffeis equation did not accurately predict the REE of any subgroup.

View this table:
TABLE 3. Differences between measured and predicted resting energy expenditure (REE) by equation

 
When the degree of adiposity (overweight compared with normal weight) was included as a factor along with race and sex, the FAO/WHO/UNU equation, previously the most promising equation overall, was without magnitude bias only in white males with BMI 95th percentile and black males with BMI < 95th percentile. The Harris-Benedict equation, previously with only a small magnitude bias for all males, showed both systematic and magnitude biases for all males with BMIs < 95th percentile, but remained unbiased for black and white males with BMIs 95th percentile. The Molnar 1 and 2 equations, both of which previously were without magnitude bias for the REE of all males, diverged when adiposity was added as a factor. The Molnar 1 equation remained without magnitude or systematic bias only in black and white males with BMIs 95th percentile. The Molnar 2 equation showed magnitude bias in all of the males with BMIs < 95th percentile and had a systemic bias around the mean for both normal-weight and overweight males. None of these 4 equations was without magnitude bias for any of the female subgroups. The Schofield equation, previously without magnitude bias for all females, was without magnitude bias only for females with BMIs 95th percentile and black males with BMIs < 95th percentile after adiposity was added as a factor. The Tverskaya and Maffeis equations showed magnitude or systematic bias, or both, for all groups. Log transformation, as recommended by Bland and Altman (29), did not correct the magnitude bias of any equation in >2 of the 8 subgroups.

In response to these results showing that none of the equations adequately predicted the REE of all race, sex, and weight subgroups, 4 new sex-specific equations were generated and evaluated by cross-validation. One set of sex-specific equations, containing only easily measured variables such as height and weight, was generated for general, clinical use. The other set, containing body-composition variables such as fat mass, was generated for research purposes.

In females, the percentage of the variance explained by the equation generated from the total sample was 72% for the equation using weight and height and 75% for the equation using body composition (Table 4; Figure 2A and B). The average r² of the equations generated by the twenty-seven 90% cross-validation procedures performed on the female subjects was 0.75 for the weight and height equation and 0.71 for the body-composition equation. The average of the 27 correlations between the actual REE and the 10% test samples excluded from the equation generation sets was 0.86 for the weight and height equation and 0.84 for the body-composition equation. When the sample was split by race and BMI, the mean measured REE did not differ from the mean predicted REE of any of the 4 subgroups (all P > 0.37), and the Bland-Altman slopes also did not differ significantly.

View this table:
TABLE 4. New equations for resting energy expenditure (REE) in MJ¹

 

View larger version (40K):
FIGURE 2.. Correlation of resting energy expenditure (REE) from new REE prediction equations with measured REE. A: Plot of actual REE versus predicted REE in females obtained with the weight and height equation, R² = 0.72. B: Plot of actual REE versus predicted REE in females obtained with the body-composition equation, R² = 0.75. C: Plot of actual REE versus predicted REE in males obtained with the weight and height equation, R² = 0.69. D: Plot of actual REE versus predicted REE in males obtained with the body-composition equation, R² = 0.71.

 
In males, the percentage of the variance explained by the equations generated from the total sample was 69% for the weight and height equation and 71% for the body-composition equation (Table 4; Figure 2 C and D). The average r² of the equations generated by the twenty-six 90% cross-validation procedures performed on the male subjects was 0.69 for the weight and height equation and 0.72 for the body-composition equation. The average of the 26 correlations between the actual REE and the 10% test samples excluded from the equation-generation sets was 0.81 for the weight and height equation and 0.82 for the body-composition equation. When the sample was split by race and BMI, the mean measured REE did not differ from the mean predicted REE of any of the 4 subgroups (all P > 0.26), and the Bland-Altman slopes also did not differ significantly.

Bland-Altman plots constructed on the 4 new equations showed that significant magnitude bias existed in the REES from both the weight and height equation and the body-composition equation for all 4 of the female subgroups tested (P < 0.05). Significant magnitude bias also existed in the REE from the weight and height equation for 3 of the 4 male subgroups tested (P < 0.02), but it existed in the REE from the body-composition equation for only 1 of the 4 male subgroups tested (P < 0.0001). This bias (slope) was not corrected by using the log transformation suggested by Bland and Altman (29). Therefore, we adjusted the equations mathematically to remove the magnitude bias (Table 4; Figure 3). Systematic bias was also absent from both equations used in females and males when race and BMI were not included as factors. When race and BMI were included as factors, the one adjustment for each sex-specific equation still removed any significant magnitude bias when all 4 subgroups were examined (Table 5). However, these adjustments introduced a small, systematic bias (0.25 MJ, or 60 kcal) into the REE of 1 of the male and 3 of the female subgroups with the use of the height and weight equation, but into the REE of only 1 of the female subgroups with the use of the body-composition equation (Table 5).

View larger version (50K):
FIGURE 3.. Bland-Altman plots of new equations. A: Height and weight equation for females: slope = –0.048 (P = 0.177), mean difference = –0.013 ± 0.729 (P = 0.750), 95% CI = –1.47, 1.45. B: Body-composition equation for females: slope = –0.042 (P = 0.190), mean difference = –0.040 ± 0.659 (P = 0.297), 95% CI = –1.36, 1.28. C: Height and weight equation for males: slope = +0.001 (P = 0.976), mean difference = –0.083 ± 0.695 (P = 0.100), 95% CI = –1.47, 1.31. D: Body-composition equation for males: slope = –0.013 (P = 0.754), mean difference = –0.022 ± 0.676 (P = 0.661), 95% CI = –1.37, 1.33.

 

View this table:
TABLE 5. Amount and significance of bias in newly generated equations

 

DISCUSSION  
We found that there currently are no adequately accurate and unbiased equations for estimating REE among black and white normal-weight and overweight children that predict both the mean and extreme values. This limitation may be due to the facts that only height, weight, age, body fat, and sex were previously considered as factors (4, 6–10); that the samples were restricted or imbalanced in some way [eg, only normal-weight (8), only white (7), or only overweight (10) subjects were included]; or that the sample size was inadequate (6, 10).

To the best of our knowledge, this report presents the first equations that include race as a factor for predicting the REE of 6–11-y-old children. It is well substantiated in the literature that blacks generally have lower REEs than do whites (14, 18–20, 30–33), and it was in the subgroup analyses that the weaknesses of many currently available equations were uncovered. Some investigators (15, 18) propose that differences in the constituents that comprise lean body mass account for the differences in REE of black and white persons. Lean body mass contains skeletal muscle, bone mass, and what is termed residual mass (the active tissue of the gastrointestinal tract, brain, liver, kidneys, heart, and other organs). Skeletal muscle, bone mass, and residual mass have been found to have different tissue-specific metabolic rates at rest: 54.3, 9.6, and 225.7 kJ · kg^–1 · d^–1, respectively (34). It has been suggested that black women have a greater skeletal muscle and bone mass than do white women of similar weight, height, and age (33, 35). Some researchers have found similar data in children (36). Because skeletal muscle and bone mass have lower metabolic rates at rest than do other tissues (37), the greater musculoskeletal mass of blacks, even if accompanied by no changes in residual organ mass, might help to explain the 3–7% lower REE found among blacks than among whites (38). This lower REE among blacks than among whites may partially explain the 3–7% underestimations of the mean measured REE in most of the existing equations evaluated.

This report also presents equations generated under a more stringent data-collection protocol (with respect to both subject inclusion and method) and statistical analysis than was used for previous equations. Subjects met strict criteria for ethnicity and were monitored throughout the test period to ensure the validity of the inclusion of each data point in the determination of measured REE. Statistical analyses tested for both systematic (mean difference) and magnitude (slope) bias, not only in the total sample, but also among the most common demographic subgroups. Rather than the degree of bias that existed being simply reported, bias was minimized mathematically to the point that, on average, REE is predicted within 0.25 MJ (or 60 kcal) regardless of sex, race, or BMI percentile.

Strengths of the present study include the relatively large cohort of black and white children, the representation of both overweight and normal-weight children in the cohort, and the multisite nature of the data collection. Limitations include the lack of other cohorts that may differ in body composition or REE, such as Asian or Native American children, and the underrepresentation of extremely lean children. In this study, age and Tanner stage did not add to the predictive power of the REE equation. However, the ages of the children studied were restricted to the years of prepuberty and early puberty; therefore, it is not surprising that Tanner stage did not add to the predictive value of the equation. Further studies are needed to establish adequate REE prediction equations for use in older adolescents, in whom age or Tanner stage (or both) may be an important variable (39), as well as for use in children of ethnic groups not represented in this cohort.

We conclude that previously developed prediction equations for the REE of young children do not accurately reflect measured REE of black and white children. To adequately predict the REE of 6–11-y-old black and white children, sex-specific equations should take into account not only height and weight but also race. With stringent statistical evaluation and a small amount of mathematical manipulation, equations can be generated that obtain accurate and unbiased estimates of the REEs of boys and girls aged 6–11 y, both black and white and normal-weight and overweight.

ACKNOWLEDGMENTS  
JRM and JAY contributed to the conception, design, and conduct of the experiment; to the interpretation of the data; and to the writing of the manuscript. JE, ENS, and EMF contributed to the conduct of the experiment. DCA-W contributed to the interpretation of the data. AMT, SAA, JPD, and GAB contributed to the conduct of the experiment, the interpretation of the data, and the editing of the manuscript. None of the authors had a conflict of interest.

REFERENCES  

1. Wong WW, Butte NF, Hergenroeder AC, Hill RB, Stuff JE, Smith EO. Are basal metabolic rate prediction equations appropriate for female children and adolescents? J Appl Physiol 1996;81:2407–14.
2. Ventham JC, Reilly JJ. Reproducibility of resting metabolic rate measurement in children. Br J Nutr 1999;81:435–7.
3. Boothby WM, Sandiford RB. Normographic charts for the calculation of the metabolic rate by the gasometer method. Boston Med Surg J 1921;185:337–54.
4. Harris JA, Benedict FG. A biometric study of basal metabolism in man. Washington, DC: Carnegie Institute of Washington, 1919:40–4.
5. Clifford E. Body satisfaction in adolescence. Percept Mot Skills 1971;33:119–25.
6. Maffeis C, Schutz Y, Micciolo R, Zoccante L, Pinelli L. Resting metabolic rate in six-to ten-year-old obese and nonobese children. J Pediatr 1993;122:556–62.
7. Molnar D, Jeges S, Erhardt E, Schutz Y. Measured and predicted resting metabolic rate in obese and nonobese adolescents. J Pediatr 1995;127:571–7.
8. FAO/WHO/UNU Expert Panel. Energy and protein requirements. Report of a Joint FAO/WHO/UNU Expert Consultation. World Health Organ Tech Rep Ser 1985;724:1–206.
9. Schofield WN. Predicting basal metabolic rate, new standards and review of previous work. Hum Nutr Clin Nutr 1985;39:5–41.
10. Tverskaya R, Rising R, Brown D, Lifshitz F. Comparison of several equations and derivation of a new equation for calculating basal metabolic rate in obese children. J Am Coll Nutr 1998;17:333–6.
11. Dietz WH, Bandini LG, Schoeller DA. Estimates of metabolic rate in obese and nonobese adolescents. J Pediatr 1991;118:146–9.
12. Finan K, Larson DE, Goran MI. Cross-validation of prediction equations for resting energy expenditure in young, healthy children. J Am Diet Assoc 1997;97:140–5.
13. Rodriguez G, Moreno LA, Fleta J, Buena M. Resting energy expenditure in children and adolescents: agreement between calorimetry and prediction equations. Clin Nutr 2002;21:255–60.
14. Gannon B, DiPietro L, Poehlman ET. Do African Americans have lower energy expenditure than Caucasians? Int J Obes Relat Metab Disord 2000;24:4–13.
15. Hunter GR, Weinsier RL, Darnell BE, Zuckerman PA, Goran MI. Racial differences in energy expenditure and aerobic fitness in premenopausal women. Am J Clin Nutr 2000;71:500–6.
16. Wong WW, Butte NF, Ellis KJ, et al. Pubertal African-American girls expend less energy at rest and during physical activity than Caucasian girls. J Clin Endocrinol Metab 1999;84:906–11.
17. Yanovski JA. Resting energy expenditure in African American and white children. Am J Clin Nutr 2001;73:149–50.
18. Tershakovec AM, Kuppler KM, Zemel B, Stallings VA. Age, sex, ethnicity, body composition, and resting energy expenditure of obese African American and white children and adolescents. Am J Clin Nutr 2002;75:867–71.
19. Arslanian S, Suproasongsin C, Janosky JE. Insulin secretion and sensitivity in black versus white prepubertal healthy children. J Clin Endocrinol Metab 1997;82:1923–7.
20. DeLany JP, Bray GA, Harsha DW, Volaufova J. Energy expenditure in preadolescent African American and white boys and girls: the Baton Rouge Children’s Study. Am J Clin Nutr 2002;75:705–13.
21. Torun B, Davies PSW, Livingstone MBE, Paolisso M, Sackett R, Spurr GB. Energy requirements and dietary energy recommendations for children and adolescents 1 to 18 years old. Eur J Clin Nutr 1996;50:S37–81.
22. de Weir JB. New methods for calculating metabolic rate with special reference to protein metabolism. J Physiol 1949;109:1–9.
23. Fleisch A. Le metabolisme basal standard et sa determination au moyen du "Metabocalculator". Helv Med Acta 1951;1:23–44(in French).
24. Bland J, Altman D. Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8:135–60.
25. Efron B. Estimating the error rate of a prediction rule: improvement on cross-validation. J Am Statist Assoc 1983;78:316–31.
26. Shao J. Linear model selection by cross-validation. J Am Statist Assoc 1993;88:486–94.
27. Picard RR, Cook D. Cross-validation of regression models. J Am Statist Assoc 1984;79:575–83.
28. Gong G. Cross-validation, the jackknife, and the bootstrap: excess error estimation in forward logistic regression. J Am Statist Assoc 1986;81:108–13.
29. Bland J, Altman D. A note on the use of the intraclass correlation coefficient in the evaluation of agreement between two methods of measurement. Comput Biol Med 1990;20:337–40.
30. Kaplan AS, Zemel BS, Stallings VA. Differences in resting energy expenditure in prepubertal black children and white children. J Pediatr 1996;129:643–7.
31. Morrison JA, Alfaro MP, Khoury P, Thornton BB, Daniels SR. Determinants of resting energy expenditure in young black girls and young white girls. J Pediatr 1996;129:637–42.
32. Yanovski SZ, Reynolds JC, Boyle AJ, Yanovski JA. Resting metabolic rate in African-American and Caucasian girls. Obes Res 1997;5:321–5.
33. Ortiz O, Russell M, Daley TL, et al. Differences in skeletal muscle and bone mineral mass between black and white females and their relevance to estimates of body composition. Am J Clin Nutr 1992;55:8–13.
34. Heymsfield SB, Gallagher D, Kotler DP, Wang Z, Allison DB, Heshka S. Body-size dependence of resting energy expenditure can be attributed to nonenergetic homogeneity of fat-free mass. Am J Physiol Endocrinol Metab 2002;282:E132–8.
35. Gallagher D, Visser M, De Meersman RE, Sepulveda D, Baumgartner RN, Pierson RN. Appendicular skeletal muscle mass: effects of age, gender, and ethnicity. J Appl Physiol 1997;83:229–39.
36. Yanovski JA, Yanovski SZ, Filmer KM, et al. Differences in body composition of black and white girls. Am J Clin Nutr 1996;64:833–9.
37. Elia M. Organ and tissue contribution to metabolic rate. In: Kinney JM, Tucker HN, eds. Energy metabolism: tissue determinants and cellular corollaries. New York: Raven Press, 1992:61–80.
38. Sharp TA, Bell ML, Grunwald GK, et al. Differences in resting metabolic rate between white and African-American young adults. Obes Res 2002;10:726–32.
39. Sun M, Gower BA, Bartolucci AA, Hunter GR, Figueroa-Colon R, Goran MI. A longitudinal study of resting energy expenditure relative to body composition during puberty in African American and white children. Am J Clin Nutr 2001;73:308–15.