Original Paper

Comparison of Several Equations and Derivation of a New Equation for Calculating Basal Metabolic Rate in Obese Children
Rozalia Tverskaya, BS, Russell Rising, PhD, Debra Brown, MPH, RD, and Fima Lifshitz, MD, FACN Maimonides Medical Center Department of Pediatrics, Nutrition & Body Composition Laboratory, Brooklyn, New York Key words: basal metabolic rate, obesity, body composition, prediction equations
Objectives: To compare basal metabolic rate (BMR) calculated by the Harris-Benedict, Ravussin, Cunningham, World Health Organization (WHO) and Schofield equations to BMR determined in an obese pediatric population. The second objective is to derive a new equation, based on measured BMR in obese children, for calculating BMR in obese pediatric patients. Methods: The study included 110 (50 male/60 female) healthy obese subjects (BMI 28) (11.7 2.8 years, 73 27 kg, 152 14 cm and 38 6% fat) who had preprandial BMR determined by indirect calorimetry. These results were compared to BMR calculated with the five above mentioned equations. Fat-free mass was determined by bioelectrical impedance and body composition was calculated using the appropriate equation. The age groups analyzed were as follows: males 3 to 10 and 11 to 18 years old; females 3 to 10 and 11 to 18 years old. A new equation was derived by stepwise multiple regression analysis using 100 randomly selected subjects from our test group and tested using the remaining 10 subjects. Results: Basal metabolic rate calculated by the Ravussin and Cunningham equations in all subgroups was lower (p 0.05) than measured BMR. Basal metabolic rate calculated by the Harris-Benedict equation was lower (p 0.05) than measured BMR in male populations ages 3 to 10, 11 to 18, and in the entire cohort. Measured BMR was overestimated by the Harris-Benedict equation (p 0.05) in females 11 to 18 years old; by the WHO equation (p 0.05) in both male and females 3 to 10 years old and by the Schofield equation (p 0.05) in males 11 to 18 years old. In comparison to measured BMR, the WHO equation appeared to be the most accurate for estimating BMR in males and females 11 to 18 years old. However, BMR calculated using our new equation in the 10 test subjects was similar to measured BMR. Conclusions: The WHO equation was the most accurate of the prediction equations studied. However, our new prediction equation may be more appropriate for calculating BMR in an obese pediatric population.

INTRODUCTION
Recent studies have been conducted in an attempt to determine the most accurate prediction equation for calculating basal metabolic rate (BMR) in children. Kaplan and coworkers [1] found a significant difference between measured BMR and BMR calculated by the World Health Organization (WHO) equation in 19 obese children. However, these investigators also found no difference between the measured BMR and that predicted by the Harris-Benedict (HB) and Schofield (SCF) height weight equations. In contrast, Dietz and coworkers [2]

found no significant difference between measured BMR and that estimated by the WHO equation in the mixed group of obese and non-obese children. However, these investigators found that BMR, calculated using the Cunningham equation, overestimated measured BMR. Furthermore, Molnar [3] stated that all but the Cunningham equation overestimated BMR in both obese and non-obese children in comparison with measured values. In our analysis we determined the most accurate equation for use in the obese pediatric population by comparing BMR, measured by indirect calorimetry, to BMR calculated from the

Address reprint requests to: Russell Rising, PhD, Miami Childrens Hospital Research Institute, 3100 S.W. 62nd Avenue, Miami, FL 33155.

Journal of the American College of Nutrition, Vol. 17, No. 4, 333336 (1998) Published by the American College of Nutrition 333

Calculating BMR in Obese Children

five equations. Furthermore, we derived a new BMR equation, based on accurate measurements of BMR in obese children, specific for calculating BMR in an obese pediatric population. by each equation according to the age groups listed in Table 1. Paired t-test was used to determine differences between measured BMR and BMR calculated by each equation.

Derivation of a New Prediction Equation

METHODS
Subjects
One hundred and ten obese (BMI 28) pediatric patients (50 male/60 female) who had measurements of preprandial BMR (BMR) between 19921996 were used for analysis (Table 1). Eighty-one percent of the patients were Caucasian, 11% Hispanic and 8% African-American. All subjects were healthy other than obesity.

Measurements
Preprandial BMR was measured for 30 minutes early in the morning using a Deltatrac ventilated hood indirect colorimeter (Model MBM-100, Sensor Medics Corporation, Yorba Linda, CA). Indirect calorimetry is a method by which respiratory gas exchange is measured. Oxygen from inspired air is assumed to be used for bio oxidation, while carbon dioxide from expired air is assumed to be from combustion of substrates. This allows the estimate of the type and amount of substrate oxidized and the amount of energy produced. Burning ethyl alcohol in a canopy system similar to ours yielded measurements of the ventilation rate of oxygen with a precision of 3.1% and accuracy of 4.3%, whereas the ventilation rate measures for carbon dioxide had a precision of 3.9% and an accuracy of 4.7% [4]. To ensure a resting awake state, all subjects were instructed to avoid unnecessary movements and watch non-violent programs or videos. A technician monitored the patient during the entire test. Body composition was assessed by measuring resistance and reactance with a Xitron 4000B impedance analyzer (Xitron Technologies, San Diego, CA) set at 50 kHz and fat-free mass and percent fat were calculated using Weight Manager (RJL systems, version 2.05).

The new prediction equation was derived using SPSS software (version 7.0, SPSS inc, Chicago, IL) utilizing 100 randomly selected obese subjects from our study group (11 2.8 years, 72 26 kg, 38 5.4% fat). Ten obese subjects were reserved for testing of the equation (13 2.9 years, 88 32 kg, 39 6.3% fat). Stepwise multiple regression analysis (forward selection technique) was used with measured BMR (kcal/day), as the dependent variable, and body weight (kg), fat-free mass (kg), fat mass (kg), age (years), and height (cm) entered as independent variables. Variables were selected if they met the 5% level of significance. All data were expressed as mean standard deviation.

RESULTS
The differences between measured BMR and BMR calculated from each equation for all of our groups is shown in Table 2. In comparison to measured BMR, both the Ravussin and Cunningham equations underestimated BMR (p 0.05) in all subgroups as well as in the entire cohort (p 0.05). The HarrisBenedict equation underestimated (p 0.05) BMR in 3 to 10 and 11 to 18 year old males and in the entire cohort, while it overestimated (p 0.05) BMR in the 11- to 18-year-old female group. The Schofield equation overestimated (p 0.05) BMR in males 3 to 10 and 11 to 18 years old while underestimating (p 0.05) BMR in females 11 to 18 years old. In comparison to measured BMR, the WHO equation showed no significant differences in both 11- to 18-year-old male and female groups as well as in the entire cohort. However, the WHO equation overestimated BMR in 3- to 10-year-old males and females. Our new prediction equation specific for an obese pediatric population from 6 to 18 years old is as follows:

Statistics
Statistical analysis was performed with Microsoft Excel (version 4.0). The measured BMR of each of the following groups were compared separately to BMRs (kcal/d) estimated Table 1. Physical Characteristics of the Entire Cohort
Age group (years) N Height (cm) 136 163 139 158 152 6 10 8 9 14 Weight (Kg) 52 87 55 83 73 10 35 13 21 27 Age (years) 8.2 13.4 9.0 13.0 11.7 1.5 1.7 0.9 1.9 2.8 Fat (%) 34 38 36 40 38 7 4 4 5 6

BMR

775 3.3

28.4 FM

FFM SEX

37 82

age

Males 3 to 10 22 Males 11 to 18 28 Females 3 to 10 16 Females 11 to 18 44 Entire cohort 110

1

Data shown as mean

standard deviation.

where fat-free mass (FFM) is expressed in kg, age in years, fat mass (FM) in kg; and for SEX a value of 1 is given for males and 0 for females. This equation had an R-square of 0.84 and a standard error of 153.0 kcals. Table 3 shows the results of the comparison between measured BMR (males) and BMR calculated using the Ravussin (R), Cunningham (C) and Harris-Benedict (HB) equations, along with our new equation (D), in the 10 test patients. No differences existed between BMR, calculated using our new equation, and BMR determined by indirect calorimetry. The summary of the derivation of the new equation is shown in

334

VOL. 17, NO. 4

Calculating BMR in Obese Children

Table 2. Basal Metabolic Rate by Indirect Calorimetry (Measured) and by Each of Five Prediction Equations
Measured Age group (y) IC1 (kcal/day) 1569 2165 1493 1831 1814 229 495 252 243 401 Ravussin (kcal/day) 1144 1697 1148 1477 1419 116* 321* 161* 203* 309* Cunningham (kcal/day) 1247 1795 1251 1577 1519 156* 318* 159* 201* 306* Equation Harris-Benedict (kcal/day) 1411 1989 1452 1901 1760 156* 505* 212 303* 413* WHO (kcal/day) 1688 1856 1729 1780 1856 241* 617 292* 259 425 Schofield (kcal/day) 1621 2348 1523 1628 1795 211 580* 231 194* 470

Males 3 to 10 Males 11 to 18 Females 3 to 10 Females 11 to 18 Entire cohort

1

Indirect calorimetry. Data shown as mean standard deviation. * Significant difference from measured BMR (p 0.05).

Table 3. Comparison of Basal Metabolic Rate Determined by the New Equation to Basal Metabolic Rate Calculated by Three Prediction Equations for the 10 Test Subjects1,2
Measured 1996 622 New equation 1914 534 Ravussin 1499 406* Cunningham 1599 402* Harris-Benedict 1350 440*

* Significant difference from measured BMR (p 0.05). 1 Data shown as mean standard deviation. 2 All values are kcal/day.

Table 4. All of the variables used in the equation were significant predictors of BMR. This means they were kept if they met the exclusion criteria of the model.

DISCUSSION
The results of this study demonstrated that, among the BMR equations, the World Health Organization (WHO) equation provided BMR results similar to our measured values in obese males and females ages 11 to 18 years old. Such results are probably due to the fact that the World Health Organization used more than 7000 people, including some obese pediatric subjects [5]. Our findings are consistent with those of Dietz and coworkers [2] who found that the WHO equation predicts BMR adequately in a group of obese/non-obese adolescent males and females. The Ravussin and Cunningham equations underestimated BMR in all subgroups as well as in the entire cohort. Such results are not surprising since these equations were derived from adult, non obese populations [6,7]. The same explanation applies to the inadequacy of the Harris-Benedict Table 4. Model Summary for the New Equation for Calculating BMR
Step 1 2 3 4
1

Variable entered1 FFM AGE FM SEX

R-square 0.749 0.811 0.833 0.843

SE 191 166 157 153

Variables were excluded if they did not meet the probability of F to enter 0.05).

equation [8]. We found that the five equations evaluated were not entirely appropriate for an obese pediatric population. The Harris-Benedict, Ravussin, Cunningham, World Health Organization and Schofield equations are used by many clinicians in the treatment of obesity [59]. However, the validity of these equations becomes questionable when used in the pediatric obese population [10]. Such findings are not surprising since many of these equations were not derived using measurements of BMR utilizing indirect calorimetry, or were specific for the pediatric population. For example, Harris and Benedict derived their equation using data from healthy, non-obese infants and other subjects in the age range 20 to 70 years old, thus excluding a large group including the pediatric obese population [8]. When BMR values were not available, the respiratory quotient was estimated to be 0.85 and BMR calculated [8]. Furthermore, the same data from Harris and Benedict were included in the derivation of the Cunningham equation [7]. Using lean body mass as a predictive variable, the Ravussin equation was derived based on the data from 249 Pima Indians between the ages of 18 and 41 years old [6]. Although indirect calorimetry was used, this equation did not include an ethnically heterogeneous, pediatric population. The Schofield equation used data from studies which were done at the turn of the century [11]. Not all of the subjects used for this derivation were preprandial prior to the measurement of BMR [11]. Since none of the tested equations were appropriate for an obese pediatric population, we derived a new prediction equation specific for an obese pediatric population between the ages of 6 and 18 years. This new equation was derived based on accurate measurements of preprandial BMR by a ventilated hood indirect calorimeter in obese children within our laboratory. In a random sample of test subjects, the new equation

JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION

335

Calculating BMR in Obese Children

estimated BMR to within 4% of the measured BMR. For clinicians who have no access to indirect calorimetry equipment, the new equation will provide an accurate estimate of BMR in an obese pediatric population. It is important to use prediction equations in populations from which they were derived. This is due to individuals not within a particular population falling outside the parameters set fourth in the equation. For example, using our new equation for calculating BMR for an unusually slender adult would yield an inaccurate result. This is due to this individual not being part of our population that was used to derive the equation. Statistically, this individual was not within the scope of the data. Due to our large influx of obese children from our weight reduction program, a new equation for calculating BMR was derived that was specific for an obese population. This equation will provide health care professionals, who have limited access to indirect calorimetry equipment, a means for calculating BMR accurately in obese children.
2. Dietz H, Bandini L, Dale A, Schoeller A: Estimates of metabolic rate in obese and non-obese adolescents. J Pediatr 118:146149, 1991. 3. Molnar D: Estimates of metabolic rate in obese and non-obese adolescents. J Pediatr 120:660661, 1992. 4. Kinney J, Tucker HN: Energy metabolism. In Tissue Determinants and Cellular Corollaries New York: Raven, pp 119, 1992. 5. FAO/WHO/UNU Expert Consultation: Energy and protein requirements. Geneva: World Health Organization, 1985. 6. Ravussin E, Bogardus C: Relationship of genetics, age, and physical fitness to daily energy expenditure and fuel utilization. Am J Clin Nutr 49:968975, 1989. 7. Cunningham JJ: A reanalysis of the factors influencing basal metabolic rate in normal adults. Am J Clin Nutr 33:23722374, 1980. 8. Harris J, Benedict G: A biometric study of basal metabolism in man. Washington DC: Carnegue Institution, publ. 279, pp 1226, 1919. 9. Schofield WN: Predicting basal metabolic rate, new standards and review of previous work. Hum Nutr: Clin Nutr 39C:540, 1985. 10. Maffeis C, Schutz Y, Micciolo R, Zoccante L, Pinelli L: Resting metabolic rate in six to ten-year-old obese and non-obese children. J Pediatr 122:556562, 1993. 11. Du Bois E: The metabolism of boys 12 and 13 years old compared with the metabolism at other ages. Arch Int Med 17:887901, 1916.

ACKNOWLEDGMENTS
We thank the Maimonides Research and Development Foundation for their generous support for this research.

REFERENCES
1. Kaplan A, Zemel B, Neiswender K, Stallings V: Resting energy expenditure in clinical pediatrics: Measured versus prediction equations. J Pediatr 127:200205, 1995.

Received October 1997; revision accepted March 1998.

336

VOL. 17, NO. 4