European Journal of Clinical Nutrition (1999) 53, 134±143
ß 1999 Stockton Press. All rights reserved 0954±3007/99 $12.00
http://www.stockton-press.co.uk/ejcn
New equations to estimate basal metabolic rate in children aged
10 ± 15 years
CJK Henry1*, S Dyer1 and A Ghusain-Choueiri,1
1
School of Biological and Molecular Sciences, Oxford Brookes University, Oxford, UK
Objective: To develop new equations for the estimation of basal metabolic rate in children aged 10 ± 15 years,
and to evaluate the impact of including pubertal stage into the equations.
Design: Mixed longitudinal.
Setting: The children were recruited from schools in Oxford, and the measurements were made in the schools.
Subjects: 195 school children, aged 10 ± 15 years, were recruited in three cohorts. The gender distribution of the
subjects was 40% boys and 60% girls.
Methods: Basal metabolic rate (BMR) was measured, by indirect calorimetry, at 6-monthly intervals for 3 years.
Anthropometric data, height, weight, body breadths and skinfold measurements (biceps, triceps, subscapular,
suprailiac and medial calf) were collected on each occasion. Fat and fat-free mass was calculated from the
skinfold measurements. Pubertal development was also assessed on annually by paediatricians. Pubic hair (PH)
and gonad (G) development was assessed in boys and breast (B) development in girls. The girls were questioned
about menarche. Stepwise multiple regression analysis was used to develop and assess new formulae for BMR
that also incorporate pubertal development.
Results: The mean BMR measured was 5.754 (s.d. 0.933) MJ=day (138 (s.d. 22) kJ=kg body wt=day) in the boys
(n ˆ 351) and 5.476 (s.d. 0.725) MJ=day (121 (s.d. 20) kJ=kg body wt=day) in the girls (n ˆ 554). Weight was the
most important factor in developing the regression equations for the calculation of BMR in both sexes (R2 ˆ 0.61
and 0.52 for boys and girls, respectively). Stepwise multiple regression analyses, with independent variables such
as gender, weight, height, puberty stage and skinfolds, allowed several BMR regression equations to be
developed. The inclusion of the menarche status in the regression equations signi®cantly (P < 0.05) improved
BMR estimation in the pre-menarche girls. Boys, pubertal stage as assessed by Pubic Hair (PH) and Gonadal
Stage (G) did not contribute to a signi®cant improvement in BMR estimation, except for 11-year-olds.
Conclusions: The inclusion of pubertal stage afforded only minor improvements in the derivation of regression
equations for the estimation of BMR of children aged between 10 and 15 years.
Sponsorship: Nestle Foundation, Lausanne, Switzerland.
Descriptors: adolescence; anthropometry; BMR; prediction equations; pubertal stage
Introduction
The measurement of basal metabolism requires specialized
equipment, some expertise on the part of the practitioner
and a high level of subject cooperation. In many situations,
particularly in ®eld conditions, these requirements cannot
be met and it is then necessary to use prediction formulae to
estimate basal metabolic rate (BMR).
In 1981, the FAO=WHO=UNU `Committee on Protein
and Energy Requirements' recommended that daily energy
requirements should be estimated using measurements of
energy expenditure. This recommendation dramatically
enhanced the importance of measuring BMR, which represents the largest component of total energy expenditure
(TEE). The estimation of TEE from BMR is based on the
factorial technique whereby the basal metabolic rate is
multiplied by the physical activity level (PAL) appropriate
*Correspondence: Professor CJK Henry, School of Biological and
Molecular Sciences, Oxford Brookes University, Gipsy Lane, Oxford OX3
OBP, UK.
Received 10 July 1998; revised 9 September 1998; accepted
20 September 1998
to an individual or a group of subjects (James & Scho®eld,
1990).
The publication of the 1985 FAO=WHO=UNU report on
protein and energy requirements provided several new
prediction equations developed speci®cally for children.
A collation of the worldwide data on basal metabolism
(Scho®eld et al, 1985) indicated that, relative to adults
(n ˆ 4726 aged 18 ± 60 y), there was a paucity of data from
children (n ˆ 751 aged 3 ± 10 y) and adolescents (n ˆ 883
aged 10 ± 15 y). The paucity of BMR data from children
was previously highlighted by Hayter & Henry (1994).
Since 1985, many studies have investigated BMR in
children (Bandini et al, 1990; Katch et al, 1990 Goran et
al, 1994, 1995). Unfortunately, most of these have concentrated on subgroups of the general population, in particular the obese. Studies on normal-weight healthy children
have remained relatively few. In fact, recent data on BMR
in normal-weight children are almost exclusively from
small control groups in obesity studies (Elliot et al, 1989;
Dietz et al, 1991; Maffeis et al, 1993a, b).
During the early half of this century, some investigators
proposed that puberty affected BMR (Topper & Muller,
1932; Bruen, 1933; Lewis et al, 1943a). Others found
evidence only for a continued decline of BMR=kg through-
Basal metabolic rate in children
CJK Henry et al
out puberty (Blunt et al, 1926). Talbot (1925) suggested
that any pubertal changes in BMR were due either to
pathological thyroid activity (refuted by Bierring, 1931)
or simply to the greatly increased rate of growth during
puberty. The evidence for a true rise or fall in BMR during
puberty thus remained controversial.
Results from recent studies to predict BMR in children
using various variables are still equivocal. Some authors
(Bandini et al, 1990; Brown et al, 1996) claimed that the
inclusion of pubertal status improved the prediction of
BMR; others, however, were unable to con®rm such a
®nding (MolnaÂr et al, 1995). The in¯uence of puberty on
BMR is believed to be the result of changes in body
composition occurring during adolescence (Talbot, 1925;
Katch et al, 1985).
The aims of the present study were to generate new
equations that can be used to estimate BMR for 10 ± 15
year-old children, using an exceptionally large database of
BMR measurements and to assess the inclusion of pubertal
status on improving these equations in adolescent children.
Subjects and methods
Subjects
One hundred and ninety-®ve Oxford school children (40%
boys, 60% girls) were recruited in three cohorts. The ®rst
cohort, recruited in 1992, consisted of 68 children, aged
10 ± 13 y. Subsequently, cohorts of 79 and 48 children aged
10 ± 11 years were recruited. The characteristics of the
subjects are shown in Table 1. The protocol was approved
by the Oxford Health Authority ethics committee. Informed
consent was obtained from the parents of each child.
Measurement of BMR
BMR was measured by indirect calorimetry using a ventilated hood system (Datex Deltatrac, Datex Instrumentation
Corp., Helsinki, Finland). All measurements were made in
the post-absorptive state (12 ± 14 h fast), in a thermoneutral
environment (24 ± 26 C) and with no external stimulation.
The subjects were rested for at least 30 min before the
measurement was made while the subjects were awake and
supine. The measurement conditions were similar to those
used by previous investigators measuring BMR. The calorimeter was calibrated with a reference gas mixture (95%
oxygen and 5% carbon dioxide (s.d. 0.003%)) prior to each
session's of measurements. Approximately 30 min of
respiratory gas exchange data were collected. The ®rst
5 ± 10 min of data were discarded, as recommended by
Isbell (1991). This allowed the subject time to acclimatise
Table 1 Characteristics of subjects
Variable
Age (y)
Weight (kg)
Height (cm)
Body mass index (kg=m2)
% Fat
Fat-free mass (FFM) (kg)
BMR (MJ=day)
BMR (kJ=kg wt)
BMR (kJ=kg FFM)
Boys (n ˆ 78)
(mean s.d.)
Girls (n ˆ 117)
(mean s.d.)
12.2 1.08
43.6 10.32
150.6 9.76
18.8 3.30
21.4 5.53
33.5 7.02
5.75 0.933**
138 22**
174 19**
12.2 1.09
47.0 11.0
151.4 8.14
20.1 3.78
26.3 7.31
33.7 6.00
5.47 0.725
121 20
164 20
**Gender difference (by Student's t-test), P < 0.001.
to the canopy and instrument noise. The average of the last
20 min of measurements was used to determine 24 h BMR.
Serial, 6-monthly measurements of BMR were taken over a
period of 3 years. Seventy nine per cent of the children had
their BMR measured between 4 and 7 times. Using the
method described by Rieper et al (1993), the within-subject
coef®cient of variation was 2.0%.
Anthropometry
Height was measured using a portable, freestanding stadiometer (CMS Weighing Equipment, London NW1 0JH,
UK). The children were measured without shoes according
to the procedure detailed by Gordon et al (1988). Weight
was measured using an electronic balance accurate to 100 g
(Soehnle model 7300, CMS Weighing Equipment). The
children were weighed in indoor clothing. All measurements were made in the fasting state. Skinfold measurements were made at ®ve sites (biceps, triceps, subscapular,
suprailiac and medial calf); limb lengths (Martin et al,
1988) and body breadths (Cameron, 1978) were also
assessed during the same period. Skinfold measurements
were used to derive percentage body fat (PaÃrõÂzkova & Roth,
1972), and fat-free mass (FFM).
Assessment of pubertal status
Once a year, for the duration of the study, two doctors
working in Child Health assessed the children's sexual
development against pictures and written descriptions of
Tanner's puberty ratings (Tanner, 1978). A male doctor
assessed the appearance of pubic hair (PH) and gonad (G)
development in the boys. A female doctor assessed the
girls' breast (B) development. In addition, at 6-monthly
intervals girls were asked whether they had begun to
menstruate.
Boys, PH and G stages were analysed separately
although, owing to low numbers, the G stages 4 and 5
and PH stages 3 and 4 were combined (see Tables 4 and 7).
Analysis of girls' pubertal data was approached in two
ways, either by using B stage as the group variable (stages
4 and 5 were combined) or by dividing the data into preand post-menarchal (Tables 4 and 6)
Assessment of predictive power of existing equations
Existing predictive equations were classi®ed into two broad
categories. Category 1 (CAT 1) represented the more
commonly used equations such as the Scho®eld (Scho®eld
et al, 1985) and FAO=WHO=UNU (1985) equations. Category 2 (CAT 2) represented the less well-known and more
recent equations (e.g. equations from Johnson, 1968; Bandini et al, 1990; Tounian et al, 1993 MolnaÂr et al, 1995).
All the equations chosen for the analysis were derived from
BMR data obtained from children of similar age ranges to
those of the present study.
The measured BMR values grouped by pubertal stages
were expressed as a percentage of the estimated values in
order to facilitate comparison:
(Measured BMR/Estimated BMR) 100
Thus, values above 100 indicated that the estimated
value was an underestimation while values below 100
indicated an overestimation of measured BMR.
135
Basal metabolic rate in children
CJK Henry et al
136
Statistical analyses
The BMR values collected from the Oxford children were
used to develop a series of regression equations. All the
BMR measurements from each child were used in these
regressions. The data collected for each sex were analysed
separately and across the whole age range available. In
addition, regressions were also developed for mixed gender
and gender groups classi®ed by chronological or pubertal
stage.
The means and standard deviations (s.d.) of the BMR
data collected from either gender were calculated for
distinct year bands and for the entire age range. BMR
was also scaled by unit weight (BMR=kg) and by FFM
(BMR=kg FFM). Student's t-test was used to test for mean
differences between genders and ANOVA for differences
between estimated and measured BMR values.
The distributions of measured values were tested for
normality using a Komolgorov ± Smirnov Goodness-of-Fit
test (SPSS for Windows, Version 6.1). Simple regression
and multiple stepwise regression (SPSS for Windows,
version 6.1) determined which variables explained the
most variation in measured BMR. When using stepwise
regression, any variables that did not contribute signi®cantly (P < 0.05) to the explained variance in BMR were
removed from the regression.
Several combinations of variables (see Table 2) were
incorporated in the regression formulae for estimating
BMR in groups divided by age, puberty or menarche
status. The goodness of ®t of the equation to the source
data was assessed according to recommended methods by
Scho®eld et al (1985) and Altman (1995). The following
graphical methods were employed as an additional test of
the ability of the equations to estimate BMR.
(a) Plots of estimated BMR (from the new equations)
against measured BMR were used to assess the general
agreement between paired estimated and measured
values (e.g. Figure 2a, b).
(b) Plots of the residuals against estimated BMR indicated
the constancy of estimation across the range of values
(e.g. Figure 3a).
(c) The distribution of the residuals was tested by plotting a
histogram of the standardized residuals (e.g. Figure 3b).
Table 2
Correlation of subject characteristics with measured BMR
Observations
Weight (kg)
Height (cm)
Fat mass (kg)
Puberty
Age (y)
Figure 1 Regression of BMR (MJ=day) against weight for (a) girls, (b) boys.
Boys (n ˆ 351)
Girls (n ˆ 554)
Boys and girls
(n ˆ 906)
0.78
0.66
0.62
0.4
0.4
0.72
0.47
0.62
0.33
0.18
0.69
0.55
0.52
0.30
0.27
Basal metabolic rate in children
CJK Henry et al
These forms of graphical assessment have been carried out
for each new equation developed in the present analysis.
Results
New regression equations to estimate BMR in 10 ± 15-year
old girls and boys
The present database was used to derive new equations for
BMR in 10 ± 15-year-old boys and girls. Table 2 shows that
the correlation between measured BMR and any group
characteristic was highest with weight, followed in a descending order by height, fat mass, pubertal stage and age.
To examine the quality of the new regression models,
plots were made of the BMR estimated from the new
equations against the measured BMR values (Figure 2a,
b). These indicated that the majority of measured values
were spaced evenly around the estimated values. However,
at low values of measured BMR there was a slight tendency
for the estimated values to be overestimated. The reverse
was true for the highest measured BMRs.
Plots were also made of the residuals against the estimated BMR values in both boys and girls, using the
equations based on weight given in Table 3. For both
genders, the values were found to be distributed in a roughly
horizontal band along zero and the dispersion was fairly
even. This is illustrated in Figure 3a. Since the plots were
similar for boys and girls, for clarity only the plot for girls is
shown. This indicated that the estimated BMR values
showed similar individual variation throughout the range
of data. The new regression equations were also tested for
normal distribution of the standardised residuals (residuals
for each subject divided by the residual standard deviation
(r.s.d.)) about the mean. The standardised residuals showed
normal distribution; for clarity only the plot for the girls is
shown (Figure 3b).
The best single variable for estimating BMR in girls was
body weight. It explained 52% of the variation in BMR
(R2 ˆ 0.52) in all the girls (10 ± 15 years) and gave a r.s.d. of
519 kJ (Table 3 and Figure 1a). In boys, weight explained
61% of the variation in BMR (Table 3 and Figure 1b). This
is an increase of almost 10% over the same variable in girls
(52%); however, the r.s.d. was considerably higher at
575 kJ.
Once the best single-variable regression equation for
BMR was identi®ed (weight), regressions were developed
using multiple anthropometric variables. These regressions
were not considered useful unless they improved upon the
equations based on weight alone (i.e. increased the
explained variance above 52% and 61%, or lowered the
respective r.s.d. below 519 kJ and 575 kJ, for girls and boys
respectively). A selection of such equations can be seen in
Tables 3 and 4 for both genders.
Figure 2 BMR (MJ=day) estimated from weight against measured BMR in (a) girls, (b) boys.
137
Basal metabolic rate in children
CJK Henry et al
138
Figure 3 (a) Girls' weight equation. Estimated BMR (using the regression equation for girls' BMR (kJ=day) ˆ 47.9 weight (kg) ‡ 3230) against residuals.
(b) Standard residuals from estimating girls' BMR using the regression equation for girls: BMR (kJ=day) ˆ 47.9 weight (kg) ‡ 3230.
One of the best multiple regression equations for 10 ± 15year-old girls included age, height and FFM alongside
weight. However, this increased the explained variance by
only 5% and the r.s.d. was lowered by 39 kJ compared to the
equation for weight alone (Table 3). For 10 ± 15-year-old
boys, of the many combinations of the anthropometric and
body composition variables that were tested, one of the best
equations in terms of explained variance involved weight
and the three skinfolds (suprailiac, triceps and subscapular)
(R2 ˆ 0.67, r.s.d. ˆ 527; Table 3).
Equations for separate chronological and developmental
age groups
Dividing the boys' data into groups corresponding to pubertal stage produced some improvements over the `all boys'
equations. The PH stage 1 equation for weight, age and
wrist breadth did not improve on the `all boys' weight R2,
but the r.s.d. was reduced greatly to 471 kJ (compared to
575 kJ). A similar pattern was seen for separate G stage
equations. The best equation (Table 4) was derived for G 3
from mid-upper-arm muscle circumference (MUAMC) and
the log of the sum of ®ve skinfolds (69% explained
variance, r.s.d. ˆ 519 kJ). Development of equations for
separate year bands in the boys rendered very little
improvement over the `all boys' equation derived from
the whole age range, except for the 14-year age band
(Table 5).
Separate regression equations for pre-menarche girls led
to greater improvements in the ability of the equation to
estimate BMR. The pre-menarche weight equation
explained 57% of the BMR variance (Table 4). This is an
improvement of 5% on the `all girls' weight equation and
Table 3 A selection of BMR regression equations explaining at least 60% or 50% of the variance in BMR of 10 ± 15
year-old boys or girls respectively
Regression formula for estimating BMR (kJ=day)
Boys
Weight (kg) 66.9 ‡ 2876
FFM (kg) 105.4 ‡ 2230
Weight (kg) 54.6 ‡ height (cm) 18.8 ‡ 576
FFM (kg) 91.1 ‡ FM (kg) 29.4 ‡ 2422
Weight (kg) 78.5 ‡ suprailiac (mm) 45.3 7 triceps (mm) 54.99 7
subscapular (mm) 38.3 ‡ 294
Girls
Weight (kg) 47.9 ‡ 3230
Weight (kg) 21.0 7 height (cm) 11.0 ‡ FFM (kg) 80.7 7 age (y) 154.6 ‡ 5319
FFM (kg) 96.7 7 gender 383.9 ‡ FM (kg) 21.4 7 age (y) 136.0 ‡ 3949
Gender: girls ˆ 0; boys ˆ 1.
R2
r.s.d.
0.61
0.62
0.62
0.63
575
567
563
558
0.67
527
0.52
0.57
0.60
519
480
522
Basal metabolic rate in children
CJK Henry et al
139
Table 4 A selection of BMR regression equations including a measure of developmental age for 10 ± 15 year old boys and girls
R2
r.s.d.
Weight(kg) 60.0 7 age(y)194 ‡ wrist breadth (mm)50.7 ‡ 2892
MUAMC(cm) 270 ‡ log of the sum of 5 skinfolds (mm)1450 7 1803
0.61
0.69
471
519
Weight
Weight
Weight
Weight
0.52
0.52
0.57
0.61
416
516
485
462
Pubertal stage
Regression formula for estimating BMR (kJ=day)
a
Boys
PH1
G3
Girls
Breast stage 1
Girls 10 ± 15 y
Pre-menarche
Pre-menarche
(kg) 69.9 7 5230
(kg) 50.6 7 menarche statusb 170.9 ‡ 3161
53.6 ‡ 3031
97.07 7 fat mass 74.6 7 age 121.2 ‡ 3452
a
When the addition of G or PH stages did not signi®cantly improve the equations, they were rejected.
Pre-menarche ˆ 0; post-menarche ˆ 1.
Gender; girls ˆ 0; boys ˆ 1.
b
the r.s.d. decreased by 31 kJ. Inclusion of fat mass and age
in the pre-menarche weight equation explained 61% of the
variance and reduced the r.s.d. by 54 kJ to 462 kJ (Table 4).
Notably, none of the regression equations for postmenarche girls improved upon the original calculations
for all the girls together. Separate equations for individual
breast stages showed no advantage over the `all girls'
equations.
Multiple regression analysis for individual year bands
produced several equations with improved R2 and reduced
r.s.d. values (when compared to the weight equation for
10 ± 15-year-old girls). However, analysis of the estimated
versus measured plots, residual plots and the distribution of
the standardised residuals led to the exclusion of several of
these equations. The remaining equations are presented in
Table 5.
Combined sex regression equations were developed and
tested. Although the R2 values were good (a combination of
FFM, gender, age and fat mass explained 60% of the
variation in BMR and gave a r.s.d. of 522 kJ; Tables 3
and 5), the skewed residuals indicated that BMR in girls
was generally overestimated.
Predictive power of existing equations
Existing predictive equations varied considerably in their
ability to predict BMR accurately for the present subjects.
The CAT 1 equations were very accurate at predicting
BMR in girls at the various breast stages, while the CAT 2
equations either substantially under-estimated (Bandini F)
or over-estimated (Tounian) those values (Table 6). When
the measured BMR data were grouped by the menarche
stage, the CAT 1 equations were again better at predicting
BMR in the pre- and post-menarche stages compared to the
CAT 2 equations.
Most of the equations tested were better at estimating
BMR in boys at the earlier stages of puberty as identi®ed by
the G and PH stages (Table 7). The Johnson and the
Bandini `mixed gender' equations gave better predictions
of the BMR values at the earlier stages of puberty, while
the majority of the analysed equations tended to signi®cantly over-estimate BMR values in the later stages of
puberty, except for MolnaÂr `male' and Scho®eld 2.
Discussion
The paucity of information on BMR in adolescent children
has prompted us to develop a series of regression equations
for estimating BMR that take into account the changes in
body composition that occur during this time. A combination of single and multiple variables was considered. These
included some traditional variables such as weight and
height, and others that are less frequently used such as
pubertal stage.
Weight explained 61% of the variation in boys BMR in
the present study. Fat-free mass, which is often cited
(Salas-Salvado et al, 1993) as the better predictor of
BMR in boys, explained 62% of the variation. Such a
negligible improvement in R2, alongside the increased
uncertainty of indirectly measuring FFM, makes weight a
more appropriate single variable to estimate BMR in boys.
Moreover, one of our best multiple regression equations,
which explained 71% of the variance in boys' BMR, was
achieved using weight and three skinfolds.
Table 5 Boys' and girls' regression equations in separate age groupsa
Age bands(y)
Boys
11
14
14
Girls
10
10
12
12
13
13
13
14
a
n
R2
r.s.d.
Weight (kg) 54.0 ‡ height (cm) 17.8 ‡ 775.7
Weight (kg) 95.5 ‡ 1579.5
FFM (kg) 113.5 7 gender 732.2 ‡ 2403.4
122
18
44
0.59
0.71
0.67
493
559
582
Weight (kg) 61.2 ‡ 2743
Weight (kg) 49.0 ‡ height (cm) 24.8 7 331
Weight (kg) 113.2 7 height (cm) 27.8 7 fat mass (kg) 89.3 ‡ 5577
Weight (kg) 82.1 7 height (cm) 25.2 7 triceps skinfold (mm) 50.6 ‡ 6286
Weight (kg) 52.7 ‡ 2865
Weight (kg) 56.1 7 breast stage 150.3 ‡ 3249
Weight (kg) 82.4 7 breast stage 162.5 7 triceps skinfold
(mm) 36.7 7 wrist breadth (mm) 76.1 ‡ 6308
Weight (kg) 41.2 7 wrist breadth (mm) 54.9 ‡ 5821
70
70
178
145
98
72
74
0.59
0.61
0.60
0.60
0.62
0.67
0.73
443
429
472
470
482
453
408
16
0.62
300
Regression formula for estimating BMR (kJ=day)
When dividing data into year bands, equations that did not signi®cantly improve the estimation were rejected.
Basal metabolic rate in children
CJK Henry et al
140
Table 6 Predicted values as percentage of measured values in girls grouped by pubertal stages and menarche status (mean s.d.)
Breast stage
Equation
1
CAT 1
FAO 1a
FAO 2 b
Scho®eld 1a
Scho®eld 2b
CAT 2
Bandini
mixedc
Bandini
Female
Johnson
Molnar Female
Molnar mixedc
Tounian
2
97 10
97 10
98 10
98 10
Menarche status
3
100 10
102 11
101 10
102 11
4 and 5
100 9
102 9
100 9
102 9
99 10
100 10
98 10
100 10
94 9**
Pre
Post
100 9
102 10*
101 9
102 9*
98 10*
99 10
97 10*
98 10
97 9**
92 9**
92 9**
97 10**
97 9*
120 11**
118 13**
109 10**
102 12
109 10**
104 10*
100 10
105 10*
110 10**
83 8**
99 10
107 11**
110 12**
88 9**
95 9**
106 10**
107 10**
88 8**
92 11**
104 11*
105 11**
85 8**
114 12**
98 10**
106 10**
88 8**
101 11
91 10**
103 10*
84 8*
a
Equations based on weight.
Equations based on weight and height.
Equations that include both genders.
Mean values signi®cantly different from measured (by one-way ANOVA): *P < 0.05, **P < 0.001.
b
c
Table 7
Predicted values as percentage of measured values in boys grouped by pubertal stages (mean s.d.)
Gonad stage
Equation
CAT 1
FAO 1a
FAO 2b
Scho®eld 1a
Scho®eld 2b
CAT 2
Bandini
mixedc
Bandini
male
Johnson
Molnar
male
Molnar
mixedc
Pubic hair stage
1
2
3
4 and 5
1
2
3 and 4
100 15
100 15
99 15
115 18**
98 17*
98 17*
97 17*
111 20**
99 19
99 19
98 19
112 22*
90 17**
90 16**
89 16**
102 19
100 17
100 16
98 16
114 19**
100 16
100 16
99 16
114 19**
91 18**
91 18**
90 18**
103 21
103 14
100 17
102 17
90 16*
102 16
102 16
96 14**
92 16**
94 16*
93 18**
82 15**
94 15**
95 15*
85 16**
103 16
110 17**
100 18
106 19*
101 20
107 20
91 17*
95 17
102 17
108 18**
102 17
109 18**
92 19**
97 19
108 17**
104 18
107 20
95 18
107 1**
108 18**
97 19
a
Equations based on weight.
Equations based on weight and height.c Equations that include both genders.
Mean values signi®cantly different from measured (by one-way ANOVA): *P < 0.05, ** P < 0.001.
b
It has been suggested (Bandini et al, 1990; Brown et
al, 1996) that the inclusion of pubertal status improves
the estimation of BMR. In the present study, however,
the inclusion of Tanner's gonad stages did not signi®cantly improve any equation for 10 ± 15-year-old boys.
Similarly, the inclusion of the PH stage in the new
equations only rarely improved the ability of the equation
to estimate BMR in boys, as has also been noted by
MolnaÂr et al (1995). The inclusion of several anthropometric variables and pubertal status in developing regression equations very rarely improved the ability of the
equation to estimate BMR without an unacceptable
increase in the complexity of the equations. Furthermore,
very little advantage was gained by creating separate
regressions for each age band over the equations for
the whole 10 ± 15-year-old sample. Even when values of
R2 and r.s.d. were improved, the residuals for the 10-,
12- or 13-year-old boys were skewed and the equations
had to be rejected. It was apparent that measures of
developmental age or anthropometric variables afforded
very little improvement over the simplest equations using
weight or weight and height.
Weight explained 51% of the variation in BMR for girls
in the present sample. This is in line with other studies that
had also reported that weight explained the largest variation
in girls' BMR (Katch et al, 1985; Fontvieille et al 1993).
Classifying the girls into pre- or post-menarche did
contribute to a signi®cant improvement in the equation's
ability to estimate BMR. The pre-menarche weight equation increased the regression coef®cient by 5% and reduced
r.s.d. by 31 kJ compared to `all girls' equations. A similar
improvement was attained with FFM replacing weight, and
even better results (61% of variance explained) were
achieved when fat mass and age were added to weight in
the pre-menarche equation. In contrast to the boys, developing equations for separate year bands in girls afforded
signi®cant improvements over the values derived from the
whole age range.
Basal metabolic rate in children
CJK Henry et al
Regression equations derived by combining the two
sexes offered little advantage over the use of single-sex
equations. BMR of the girls in particular was generally
overestimated. This was not surprising as it is probably a
result of the gender difference in the BMR found in our
data (Table 2) as well as by others (Bandini et al, 1990;
Spurr et al, 1992). Consequently, single-sex equations are
recommended for this age group. In all cases, using
relatively simple combinations of anthropometric variables,
the estimated values for BMR in boys were better than
those developed for the girls.
The accuracy of existing BMR prediction equations
tested on our sample of children has shown that, on
almost every occasion, the equations from CAT 1 (FAO=WHO=UNU and Scho®eld equations) performed best in
predicting BMR in 10 ± 15-year-old children. This observation has also been noted in other studies (Bandini et al,
1990, and co-workers 1995; Dietz et al, 1991; Firouzbaksh
et al, 1993; Kaplan et al 1995). However, (MolnaÂr 1992,
1995) found that the FAO=WHO=UNU equations overestimated BMR in their sample of adolescents, while
Brown et al (1996) showed that in 9- and 10-year-old
boys predictions were inaccurate in pre-pubertal boys.
In post-menarchal girls, height (used in the second
equation of the FAO=WHO=UNU) signi®cantly improved
BMR estimation, indicating that it may be an important
additional predictor in this group of adolescents. A possible
explanation for this observation may be the larger amounts
of fat mass in post-menarchal girls. Talbot (1925) commented that height was a better reference than weight for
BMR in girls. He reasoned that height confounded the
`calorie diluting effect of fat' as height was more independent of the inactive fat mass than was weight. Two studies
of obese female subjects (Dietz et al 1991; MolnaÂr et al,
1992, 1995) lend support to this idea, as the authors note
that the FAO=WHO=UNU equation including height and
weight (FAO 2) proved a better predictor of BMR for their
obese subjects. In the present study, however, including
height in the regression equation did not signi®cantly
improve its ability to estimate BMR in the post-menarche
girls over the use of weight alone.
Conclusions
In summary, the BMR data collected in the present study
were used to produce a large number of regression equations for the estimation of BMR in 10 ± 15-year-old boys
and girls. Weight was the single most important factor in
the equations in either gender. The addition of other growth
or body composition proxies (pubertal status) marginally
improved the estimated BMR, although this was mainly for
subgroups of the 10 ± 15-year-old sample, such as premenarche girls or 11-year-old boys.
Acknowledgements ÐThe authors are grateful for the Nestle Foundation,
Switzerland, for funding this research. We gratefully acknowledge the
cooperation of the Oxford schools and children who took part in the study,
and Joan Webster for assistance in preparation of this manuscript.
References
Altman DG (1995): Practical Statistics for Medical Research. London:
Chapman & Hall.
Bandini LG, Schoeller DA & Dietz WH (1990): Energy expenditure in
obese and non obese adolescents. Pediatr. Res. 27, 198 ± 203.
Bandini LG, Morelli JA, Must A & Dietz WH (1995): Accuracy of
standardized equations for predicting metabolic rate in premenarcheal
girls. Am. J. Clin. Nutr. 62, 711 ± 714.
Bierring E (1931): The Standard Metabolism of Boys (Seven to Eighteen
Years, Inclusive). Copenhagen: Levin & Munksgaard.
Blunt K, Tilt J, McLaughlin L & Gunn KB (1926): The basal metabolism
of girls. J. Biol. Chem. 67, 491 ± 503.
Brown, D.C., Kelnar, C.J.H. and Wu, F.C.W. (1996): Energy metabolism
during male human puberty. II. Use of testicular size in predictive
equations for basal metabolic rate. Ann. Hum. Biol. 23, 281 ± 284.
Bruen C (1933): Variation of basal metabolic rate per unit surface area
with age. II. The pubertal acceleration. J. Nutr. 6, 383 ± 395.
Cameron N. (1978): The Measurement of Human Growth. London: Croon
Helm.
Dietz WH, Bandini LG & Schoeller DA (1991): Estimates of metabolic
rate in obese and nonobese adolescents. J. Pediatr. 118, 146 ± 149.
Durnin JVGA (1981): Basal metabolic rate in man. In Joint FAO=WHO=UNU Expert Consultation on Energy and Protein Requirements,
Rome, 5 to 17 October.
Elliot DL, Goldberg L, Kuehl KS & Hanna C (1989): Metabolic evaluation
of obese and non obese siblings. J. Pediatr. 114, 957 ± 962.
FAO=WHO=UNU (1981): Basal metabolic rate in man. In Joint FAO=WHO=UNU Expert Consultation on Energy and Protein Requirements,
Rome, 5 to 17 October.
FAO=WHO=UNU (1985): Energy and Protein Requirements. Report of a
joint FAO=WHO=UNU Expert Consultation. Geneva: WHO, Technical
Report Series 724.
Firouzbakhsh S, Mathis RK, Dorchester WL, Oseas RS, Groncy PK, Grant
KE & Finklestein JZ (1993): Measured resting energy expenditure in
children. J. Pediatr. Gastroenterol. Nutr. 16, 136 ± 142.
Fontvieille A-M & Ravussin E (1993): Metabolic rate and body composition of Pima Indian and Caucasian children. Crit. Rev. Food. Sci. Nutr.
33, 363 ± 368.
Goran MI, Kaskoun M & Johnson R (1994): Determinations of resting
energy expenditure in young children. J. Pediatr. 125, 362 ± 367.
Goran MI, Kaskoun M, Johnson R, Martinez C, Kelly B & Hood V (1995):
Energy expenditure and body fat distribution in Mohawk children.
Pediatrics 95, 89 ± 95.
Gordon CC, Cameron Chumlea W & Roche AF (1988): Stature, recumbent length and weight. In Anthropometric Standardisation Reference
manual, ed. TG Lohman, AF Roche & R Martorell, 1 ± 3. Illinois:
Human Kinetics.
Hayter JF & Henry CJ (1994): The re-examination of basal metabolic rate
predictive equations: the importance of geographic origin of subjects in
sample selection. Eur. J. Clin. Nutr. 48, 702 ± 707.
Isbell TR, Klesges RC, Meyers AW & Klesges LM (1991): Measurement
reliability and reactivity using repeated measurements of resting energy
expenditure with a face mask, mouthpiece, and ventilated canopy. J.
Parenter. Enteral. Nur. 15, 165 ± 168.
James WPT & Scho®eld EC (1990): Human Energy Requirements. A
Manual for Planners and Nutritionists. New York: Oxford University
Press.
Johnson HL (1968): Energy metabolism at various weights: Man Part 1.
Basal: Infants and children. In Metabolism. Biological Handbook, ed.
PL Altman & DS Dittmer, 344. Bethesda.
Kaplan AS, Zemel BS, Neiswender KM & Stallings VA (1995): Resting
energy expenditure in clinical pediatrics: measured versus prediction
equations. J. Pediatr. 127, 200 ± 205.
Katch V, Rocchini A, Becque D, Marks C & Moorehead C (1985): Basal
metabolism of obese adolescents: age, gender and body composition
effects. Int. J. Obes. 9, 69 ± 76.
Katch VL, Marks CC, Becque MD, Moorehead C & Rocchini A (1990):
Basal metabolism of obese adolescents: evidence for energy conservation compared to normal and lean adolescents. Am. J. Hum. Biol. 2,
543 ± 551.
Lewis RC, Duval AM & Iliff A (1943): Effect of adolescence on basal
metabolism of normal children. Am. J. Dis. Child. 66, 396 ± 403.
Maffeis C, Schutz Y, Micciolo R, Zoccante L & Pinelli L (1993a): Resting
metabolic rate in six to ten year old obese and nonobese children.
J. Pediatr. 122, 556 ± 562.
Maffeis C, Schutz Y, Zoccante R, Micciolo R & Pinelli L (1993b): Meal
induced thermogenesis in lean and obese prepubertal children. Am. J.
Clin. Nutr. 57, 481 ± 485.
Martin AD, Carter JEL, Hendy KC & Malina RM (1988): Segment
lengths. In Anthropometric Standardisation Reference Manual,
ed. TG Lohman, AF Roche & R Martorel, 9 ± 26. Illinois: Human
Kinetics.
MolnaÂr D (1992): Estimates of metabolic rate in obese and nonobese
adolescents. J. Pediatr. 120, 660 ± 661.
141
Basal metabolic rate in children
CJK Henry et al
142
MolnaÂr D, Jeges S, Erhardt E & Schutz. (1995): Measured and predicted
resting metabolic rate in obese and nonobese adolescents. J. Pediatr.
127, 571 ± 577.
PaÃrõÂzkova J & Roth Z (1972): The assessment of depot fat in children from
skinfold thickness measurements by Holtain (Tanner=Whitehouse)
caliper. Hum. Biol. 44, 613 ± 620.
Rieper H, Karst H, Noack R & Johnsen DJ (1993): Intra- and interindividual variations in energy expenditure of 14 ± 15 year old schoolgirls as determined by indirect calorimetry. Br. J. Nutr. 69, 29 ± 36.
Salas-Salvado J, Barenys-Manent M, Recasens Gracia MA & MartõÂHenneberg C (1993): In¯uence of adiposity on the thermic effect of
food and exercise in lean and obese adolescents. Int. J. Obes. 17,
717 ± 722.
Scho®eld WN, Scho®eld C & James WPT (1985): Basal metabolic rate Ð
Review and prediction, together with an annotated bibliography of
source material. Hum. Nutr. Clin. Nutr. 39c (Suppl. 1), S1 ± S96.
Spurr GB, Reina JC & Hoffman RG (1992): Basal metabolic rate of
Colombian children 2 ± 16 years of age: ethnicity and nutritional status.
Am. J. Clin. Nutr. 56, 623 ± 629.
Talbot FB (1925): Basal metabolism of children. Physiol. Rev. 5,
477 ± 517.
Tanner JM (1978): Foetus into Man Ð Physical Growth from Conception
to Maturity. London: Open Books Publishing Ltd=Fletcher & Son Ltd.
Topper A & Muller H (1932): Basal metabolism of normal children. The
puberty reaction. Am. J. Dis. Child. 43, 327 ± 336.
Tounian P, Giradet J P, Carlier L, Frelut M., Veinberg F & Fontaine JL
(1993): Resting energy expenditure and food-induced thermogenesis in
obese children. J. Pediatr. Gastroenterol. Nutr. 16, 451 ± 457.