Validity of bioelectric impedance for body
composition assessment in children
LINDA
B. HOUTKOOPER,
TIMOTHY
G. LOHMAN,
SCOTT B. GOING,
AND MATT
C. HALL
Departments
of Exercise and Sport Sciences, Body Composition Laboratory, and Nutrition
and Food Science, Cooperative Extension, University of Arizona, Tucson, Arizona 85721
HOUTKOOPER,LINDAB.,TIMOTHY G.LOHMAN,SCOTT B.
GOING, AND MATT C. HALL. Validity of bioelectric impedance
for body composition assessment in children. J. Appl. Physiol.
66(2): 814-821,
1989.-Whole-body
bioelectrical
impedance
analysis (BIA) was evaluated for its reliability
and accuracy in
estimating body composition
in children. The hypothesis that
the index, body height2 divided by resistance (RI), can accurately predict fat-free body mass (FFB) and percent fat (%FAT)
in children was tested on 94 Caucasian children lo-14 yr old.
Criterion variables were FFB and %FAT estimated using multicomponent
equations developed for children. BIA measurements (resistance and reactance) were found to be reliable.
Prediction
accuracy (standard error of the estimate, SEE) for
FFB from RI alone was 2.6 kg and for %FAT from RI and body
weight was 4.2%. For RI, anthropometric
variables and reactance, the SEE improved to 1.9 kg FFB. For RI and anthropometric variables, the SEE was 3.3% FAT. For anthropometric
variables alone, the SEE’s were 2.1 kg FFB and 3.2% FAT.
Adult FFB and %FAT prediction
equations cross-validated
with this sample resulted in SEE’s similar to those for adult
samples. We conclude that RI together with anthropometry
is
a reliable and an acceptably accurate method of estimating FFB
mass and %FAT in children.
whole-body bioelectrical
resistance; body composition methods;
human body composition;
nutritional
assessment; whole-body
impedance
WHOLE-BODY bioelectrical impedance (BIA) is a relatively new, noninvasive technique for estimating body
composition that has been tested in the adult human
population (14, H-20, 27). Bioelectrical impedance (2)
is defined as the hindrance to the flow of an alternating current and includes two components, resistance (R)
and reactance (Xc). Impedance is expressed as 2 =
D.
R is the sum of the in-phase vectors, and Xc
is the sum of the out-of-phase vectors (18). BIA has been
used to estimate the volume of total body water (TBW),
fat-free body mass (FFB), and percent body fat (%FAT)
(12, 14, 18, 19, 21, 27). The theoretical relationship
between BIA and TBW volume was proposed by Hoffer
et al. (12) based on earlier work by Thomasset (31, 32)
and has been reviewed by several investigators (14, 18,
19, 21, 22, 24, 27). The current published validation
studies (14, 18, 19, 27) using BIA in adults have shown
that the resistance and reactance components of impedance measurements are highly reproducible and offer a
814
practical method of estimating TBW and FFB in adults.
However, validation studies for children or youth have
not been published.
Most researchers have used body density (D) or TBW
and the two-component body composition model as the
criterion methods for estimating FFB and %FAT in
validation studies. The validity of prediction equations
developed using a two-component model is based on
meeting the underlying assumptions of an unchanging
FFB composition within and among individuals (28).
When these assumptions of the two-component model
are violated, significant errors in estimation of body fat
will be made (16, 33).
For children, because of violations of the assumptions
underlying the two-component model, there are major
problems associated with using this model to determine
criterion variables for use in validation of new indirect
body composition assessment methods. Recent investigations (15) have shown that in children there is a
relatively large variability in proportions of the various
components of the FFB mass, particularly the TBW and
bone mineral content (BMC). The three- and four-component body composition models address the underlying
assumptions regarding the FFB inherent in the twocomponent model by including measurements of TBW
(3) and BMC (5, 17). Thus using these multiple-component models for body composition assessment in populations with changing proportions of FFB components
improves the validity of the indirect methods of body
composition assessment (15, 16). Therefore, in children,
multiple-component body composition models should be
used to determine criterion variables.
This study was designed to determine the reliability
and accuracy of BIA measurements for estimating FFB
and %FAT in Caucasian children and to cross-validate
in children prediction equations for FFB and %FAT
developed from adult samples using BIA. Multiple-component models based on body density (D) and body water
based on deuterium dilution space (W) were used as the
criterion methods for measuring FFB-DW and %FATDW in this study. Prediction equations were developed
for estimating FFB-DW and %FAT-DW from BIA alone,
BIA plus selected anthropometric variables, and selected
anthropometric variables alone.
METHODS
The subjects in this study were 94 boys and girls aged
lo-14 yr. This study was approved by the University
0161-7567/89 $1.50 Copyright 0 1989 the American Physiological Society
IMPEDANCE
BODY COMPOSITION
Human Subjects Review Committee,
and before study
participation
informed consent was obtained from all
subjects and their parents or guardians. Subjects with a
sum of triceps and subscapular skinfolds equaling or
exceeding the 25th percentile of the National Children
and Youth Fitness Study norms for their respective age
and sex were classified as obese (higher skinfolds correspond to lower percentiles) (26). Subject age distribution
was as follows: 10 yr of age, n = 17, 11 yr, n = 22; 12 yr,
n= 27; 13 yr, n = 18; and 14 yr, n = 10. Subject
characteristics by sex are reported in Table 1.
The criterion variables, FFB-DW
and %FAT-DW,
were determined using equations developed from multiple-component
models based on body density and body
water from deuterium
dilution
space combined. The
equations used to calculate each of these dependent
variables were previously developed for children and
youth (2, 15) are given below
%FAT-DW
= [(2.057/D)
FFB-DW
- (0.786 x W) - 1.2861
x 100
= Wt - (Wt x %FAT-DW/lOO)
where D is body density (g/cm3), Wt is body weight (kg),
and W (water fraction of body weight) = body water
from deuterium dilution (kg)/Wt (kg).
Separate prediction equations were developed using
the independent variables: resistance index (RI) [RI =
Ht (cm)“/resistance
(Q), where Ht is height], impedance
index (II) [II = Ht (cm”)/impedance
( Q)], RI and reactance, RI and weight, body mass index [Wt (kg)/Ht (m)2],
and selected anthropometric
dimensions.
All measurements
and a medical examination
were
completed in one 4-hour session conducted during the
same time period on each test day. All subjects were
measured 3 h postprandially
and at least 12 h after
Age, Yr
Height,
Weight,
cm
kg
Resistance, Q
Reactance,
Q
RI, Q-cm
Body density, g/ml
Body water,*
kg
% FAT-DW
FFB-DW
BMI
ASSESSMENT
vigorous exercise. Medical examinations
by a pediatrician screened subjects for health problems that could
affect body composition measurements and assessed sexual maturation
using a modified Tanner scale (29, 30).
Subjects in this sample were classified as prepubescent
(n = 13), pubescent (n = 49), or postpubescent (n = 32).
Resistance and reactance were measured with subject
reclining using an impedance analyzer with four body
surface electrodes and a standard conduction current of
800 PA and 50 kHz (BIA model 101, RJL Systems,
Detroit, MI). Two surface self-adhesive spot electrodes
were placed on the dorsal surface of the right hand and
on the dorsal surface of the right foot as recommended
by RJL Systems (13). Resistance and reactance were
recorded twice, 1 min intervening,
at hour 0 (initial
reading), hour 2, and hour 3. Within each hour, repeated
measurements were made- without removing electrodes
between measurements. The mean of the hour 2 and 3
values was used for subsequent data analysis.
Resistance and reactance measurements were made on
both the right and left side of the body at hour 2 with
identical protocol on a subsample of 73 subjects. The
correlation coefficients between the left and right side
measurements were 0.98 for resistance, impedance, and
RI and 0.89 for reactance. The relatively small differences between the right and left side resistance (12.1 Q)
and reactance (0.3 Q) measurements were significant for
resistance (P 5 0.05) and nonsignificant
for reactance.
Body water (W) was estimated from respiratory water
samples using a modified deuterium oxide (D20) dilution
technique reported by Boileau et al (3). After the D20
dose was consumed, respiratory water samples were collected from each subject at hours 0,2, and 3. The subject
breathed continuously into a tube submerged in a methanol-dry ice bath. Subjects inhaled through the nose and
exhaled through the mouth and the tube for 8-10 min.
Means&SD
Range
Means+SD
12.3kl.l
155.9t9.9
50.6k14.6
643.2t76.6
60.2t6.6
38.727.9
1.033t0.017
28.3t6.0
23.5zk6.9
35.9t7.4
19.7k3.3
10.2-14.2
132.6-177.2
29.9-99.4
458.5-835.3
49.5-80.5
26.5-56.5
0.994-1.065
19.4-41.6
12.2-41.4
25.3-53.9
13.5-26.7
12.321.4
153.1tll.l
47.5k12.9
612.7t79.5
59.Ok7.4
39.5t9.8
1.043kO.020
28.2t6.2
20.4t9.1
36.6k7.8
19.8k3.5
10.3-14.8
132.3-175.1
27.2-96.2
445.0-802.3
40.5-77.3
26.1-62.2
1.000-1.081
18.4-45.0
4.5-36.1
23.0-56.9
15.0-29.2
16.1k6.5
10.7k5.7
18.1t8.6
23.0t9.6
6.2-33.5
4.6-31.3
5.9-48.3
8.7-50.0
13.9t6.8
lO.lk7.2
15.5klO.2
17.6t8.3
4.9-34.4
4.2-36.8
3.9-34.1
2.3-37.9
79.6t8.8
67.0k9.4
65.5-106.8
52.8-96.2
78.6k9.4
69.7k10.4
61.5-115.2
54.6-105.2
Skinfolds, mm
Triceps
Subscapular
Abdomen
Thigh
Circumferences,
Chest
Girls,
widths,
cm
23.9k3.1
n = 41; boys, n = 53. RI, height2/resistance;
space (W); FFB-DW,
Range
cm
Abdomen
Skeletal
Hip
815
IN CHILDREN
%FAT-DW,
18.4-32.7
%fat
estimated
fat-free body estimated from D and W; BMI, weight/height2.
23.0t2.5
from
body
density
(D) and body water
* Deuterium dilution space.
16.2-30.5
from
deuterium
dilution
816
IMPEDANCE
BODY
COMPOSITION
After sample collection, the tubing with frozen sample
inside was closed on each end, allowed to thaw, then
transferred
into a labeled plastic vial, and stored frozen
until analysis. The respiratory
water samples were analyzed for D20 concentration
by infrared spectrophotometry (3) using a Wilks Miran lA-FF infrared spectrophotometer in series with a Doric Digital Trendicator
(series
410A) to measure the absorbance of D20 in the samples.
The body water values determined by this D20 dilution
technique slightly overestimate
total body water values
(35)
The procedures and instrumentation
for measurement
of the underwater
weight were modified from those suggested by Akers and Buskirk
(l), Slaughter et al. (29)
and Wilmore (34). The residual volume was determined
simultaneously
with underwater
weight (34).
Anthropometric
measurements
included body weight,
skinfolds,
circumferences,
skeletal
widths,
standing
height, and sitting height. Skinfold thickness
measurements were made with a Harpenden
caliper three times
on the right side of the body at nine sites: triceps, biceps,
subscapular,
midaxillary,
waist suprailiac,
anterior suprailiac, abdomen, thigh, and medial calf. The seven body
circumference
measurements
included upper arm (relaxed), upper arm (contracted),
forearm, chest (end expiration), abdomen, middle thigh, and calf. Three measurements were made at each site on the right side of the
body using a narrow retractable
steel tape (Lufkin
146
ME). The 10 skeletal width measurements
included
shoulder, hip, chest, standing height, sitting height, elbow, wrist, knee, ankle, and forearm length on the right
side of the body. All skeletal width measurements
were
made by applying moderate tension compressing the soft
tissue against the bones using a rigid-triangle
narrowblade anthropometer
or a bow caliper. All anthropometric dimensions
are described in detail by Houtkooper
(13)
Analysis of variance of within-day
repeated measurements over a 3-h period for resistance
and reactance
indicated there were small but significant
(P 5 0.05)
increases in the mean resistance values between hours 0
and 2 (A = 8.2 Q) and between hours 2 and 3 (A = 2.6
Q). The standard error of measurement
for resistance
across the three trials was 7.2 Q. The directions
of
changes of resistance values across the 3 h varied somewhat from subject to subject.
Reactance measurements
increased slightly across the
3-h period. The average magnitudes of increase between
hoursOand2(A=
1.6 Q) and between hours 2 and 3 (A
= 0.8 Q) were statistically
significant
(P 5 0.05). The
standard error of measurement
for reactance was 2.3 Q.
Hour 2 and 3 values were slightly more stable for both
resistance and reactance, and the mean of hour 2 and 3
mean resistance and reactance values were used for data
analysis.
Measurement
reliability. Between-day
reliability of resistance, reactance, hydrometry,
densitometry,
and anthropometry
measurements
was assessed by dependent t
tests on a subsample of 14 subjects from repeat measurements obtained 4-5 wk after the initial measurements.
The correlations
of between-day
measurements
were
ASSESSMENT
IN
CHILDREN
very high, ranging from r = 0.96 to 0.999 for height,
weight, resistance, and impedance. The standard error
of measurement
for between-day
resistance values was
11.6 Q. Between-day
correlations
were also high (0.820.92) for reactance, body density, and body water. Between-day correlations
for skinfolds, circumferences,
and
skeletal widths were very high (0.96-0.99) for 13 measurement sites and were high (0.89-0.95) for the remaining 10 measurement
sites.
Data analysis. The data were analyzed using leastsquares multiple regression techniques (23). In the initial
analysis to determine the best power of height for calculation of RI [Ht (cm)“/resistance
(St)], the independent
variables, log height and log resistance, were regressed
on the logarithmic
transformed
dependent
variable,
FFB-DW.
The regression coefficient for log height was
1.970. For other exponents
of height, larger standard
errors of the estimate (SEE’s) were found. The RI variable was also tested for a curvilinear
relation with FFBDW, and the relation was found to be linear.
Further multiple regression analyses were done to identify the best prediction equations for estimating %FATDW and FFB-D W from categorical variables of fatness
category (obese = -1, nonobese = +l), sex group (male
= 1, female = -l), sex by fatness interaction,
age, and
sexual maturation
level (1 = prepubescent,
0 = pubescent, -1 = postpubescent)
and from continuous variables
of weight, body mass index, impedance index, resistance
index, reactance, skinfolds, skeletal dimensions, and limb
and trunk girths. Equations were developed for predicting %FAT-DW
and FFB-DW
from 1) RI (body height2/
resistance)
only, 2) RI combined with selected anthropometric variables, and 3) selected anthropometric
variables only. The significance
level was set at P 5 0.05.
For the final equations, the following stepwise regression
analyses were performed-FFB-DW:
RI, chest circumference, hip skeletal width, and reactance; %FAT-DW:
RI, abdominal circumference,
sum of triceps, abdomen,
and thigh skinfolds.
RESULTS
The means t SD for body size, composition, bioelectric
values (resistance,
reactance, and impedance), and anthropometric
dimensions
of the sample are given in
Table 1. Average body density, body water, height, and
weight were similar to results from other studies in
children (6-11, 16, 25).
Regression analysis indicated that body-height2/resistante (RI) and body weight were significant predictors of
the criterion variable FFB-DW.
For RI alone the adjusted R2 = 0.88 and SEE was 2.6 kg and for body weight
alone the adjusted R2 = 0.77 and SEE was 3.6 kg. The
difference
between these two SEE’s was statistically
significant at the 0.05 level of probability.
When RI and
body weight were used to predict FFB-DW,
the adjusted
R2 increased to 0.93 and the SEE decreased to 2.0 kg.
When RI and body weight were used to predict %FATDW, the adjusted R2 = 0.74 and SEE was 4.2% (Table
2). For body weight alone the adjusted R2 = 0.26 and
SEE was 7.1% FAT-DW.
Multiple regression analyses to assess the significance
IMPEDANCE
BODY COMPOSITION
2. Regression analysis of fat-free body with RI
and RI plus weight and percent body fat
with RI and weight
TABLE
Dependent
Variables
Regression
Coefficients
RI
R
Adjusted
R2
SEE
Intercept
Weight
FFB-DW
o.t33*
0.94
0.88
2.6
4.43
FFB-DW
0.%3* 0.24*
0.96
0.93
2.0
2.69
%FAT-DW
-1.11"
1.04"
0.87
0.74
4.2
15.16
RI, resistance index ( height2/resistance); FFB-DW, fat-free body
mass estimated from body density (D) and body water from deuterium
dilution space (W). %FAT-DW, %fat estimated from body density (D)
and body water from deuterium dilution space (W). R, correlation
coefficient; SEE, standard error of the estimate. * P I 0.05.
of RI combined with the categorical variables age, sex,
sexual maturity,
fatness category (obese or nonobese),
and sex by fatness interaction
for predicting FFB-DW
showed significant (P 5 0.05) regression coefficients for
all independent variables except sex by fatness interaction. For %FAT-DW, all of the above independent variables, except sex, had significant regression coefficients.
Two methods were used to determine the significant
anthropometric
measurements for predicting FFB-DW
and %FAT-DW.
In one method, a priori, we selected
anthropometric
variables that represented the limbs and
trunk and included these variables with RI in the multiple regression analyses. In the other method step-up
multiple regression analysis was used to select significant
anthropometric
variables from all the measured anthropometric variables. These two methods identified the
same significant predictor anthropometric
variables that
when combined with RI resulted in the best-fitting equations for predicting FFB-DW and %FAT-DW.
The identified
significant
anthropometric
measurements that improved prediction accuracy for FFB-DW
when combined with RI (adjusted R2 increased 0.01; SEE
decreased 0.5 kg) were hip skeletal width, forearm length,
and chest circumference. When RI was combined with
these significant anthropometric
v ariables and the categorical variables fatness, sex, and sex by fatness i.nteraction, the categorical variables were no longer significant predictors of FFB-DW. When reactance was combined with RI and the significant
anthropometric
variables in a regression analysis, all of these independent variables except forearm length continued to be
significant predictors of FFB-DW. The best-fitting prediction equation for FFB-DW for the total sample had
an adjusted R2 = 0.94 and SEE of 1.9 kg and was
FFB-DW
= 0.713 (RI) + 0.150 (chest circumference)
+ 0.493 (hip skeletal width)
+ 0.121 (reactance)
- 21.41
If reactance is not measured, the intercept in the equation becomes -12.20 (Table 3).
Regression analyses predicting %FAT-DW
indicated
that RI, weight, and contracted upper arm and abdomen
circumferences were significant predictors. When these
two anthropometric
measurements were combined with
the independent variables RI, weight, fatness category,
ASSESSMENT
TABLE
817
IN CHILDREN
3. Regression analyses predicting FFB-D W
Independent
Variables
FFB-DW
R
Adjusted
R2
SEE
Body weight
0.88
0.77
3.6
II
0.94
0.88
2.6
RI
0.94
0.88
2.6
RI plus reactance
0.95
0.89
2.5
RI plus body weight
0.96
0.93
2.0
Anthropometry alone
0.96
0.92
2.1
RI plus anthropometry*
0.97
0.93
2.0
RI plus anthropometry and
0.97
0.94
1.9
reactance?
n = 94. FFB-DW, fat-free body mass estimated from body density
(D) and body water from deuterium water space (W). R, correlation
coefficient; SEE, standard error of estimate; II, impedance index; RI,
resistance index. * FFB-DW = 0.652 (RI) + 0.155 (chest circumference)
+ 0.494 (hip skeletal width) - 12.20. t Best-fitting equation: FFB-DW
= 0.713 (RI) + 0.150 (chest circumference) + 0.493 (hip skeletal width)
+ 0.121 (reactance) - 21.41.
sex, sex by fatness interaction,
and the sum of three
skinfolds (triceps, abdomen, and thigh), only RI, sum of
three skinfolds, and abdomen circumference were significant predictors. When RI was combined with these two
significant anthropometric
variables, fatness, sex, and
sex by fatness interaction were no longer significant. The
best-fitting prediction equation for %FAT-DW for the
total sample (adjusted R2 = 0.85 and SEE of 3.3% FAT)
was as follows
%FAT-DW
- - 0.235 (RI) + 0.252 (abdomen
circumference)
+ 0.281 (sum of triceps, abdomen,
and thigh skinfolds)
- 0.044
All variables in this equation were statistically
significant.
When only anthropometric
variables (skinfolds, circumferences, skeletal widths, and lengths) were used in
multiple
regression analyses to predict the dependent
variable FFB-DW, the best-fitting prediction equation
had an adjusted R2 = 0.92 and SEE of 2.1 kg. These
adjusted R2 and SEE were about equal to those of the
best-fitting
prediction equation for FFB-DW that included RI and the significant anthropometric
variables
(Table 3). The best-fitting
equation for predicting
%FAT-DW including only anthropometric
variables had
an adjusted R2 = 0.85 and SEE of 3.2% FAT. These
values were equal to the adjusted R2 and SEE for the
best-fitting equation for %FAT that included RI and the
significant anthropometric
variables (Table 4).
The adjusted R2 and SEE’s obtained from regression
analyses with FFB-DW as the dependent variable predicted from the independent variables of body weight,
impedance index, RI alone, and RI plus reactance, RI
plus weight, anthropometry
alone, and the equations
including RI, anthropometry,
and reactance combined
and are summarized in Table 3. Body weight had the
lowest prediction accuracy. Impedance index alone, RI
alone, and RI plus reactance has similar adjusted R2
values and SEE’s. RI plus weight and the two equations
including RI, anthropometry,
and reactance combined or
IMPEDANCE
818
TABLE
BODY COMPOSITION
ASSESSMENT
60-
4. Regression analyses predicting %FAT-D W
Independent
Variables
55 -
%FAT-DW
R
Adjusted
IN CHILDREN
R2
50 -
SEE
BMI
0.73
0.53
5.7
II plus body weight
0.86
0.74
4.2
RI plus body weight
0.87
0.74
4.2
RI plus body weight and
0.87
0.75
4.2
reactance
Anthropometry alone
0.92
0.85
3.2
RI plus anthropometry*
0.92
0.85
3.3
n = 94. %FAT-DW, %fat estimated from body density (D) and body
water from deuterium dilution space (W). R, correlation coefficient;
SEE, standard error of estimate; BMI, body mass index [weight (kg)/
height2 (m)]; II, impedance index; RI, resistance index. * Best-fitting
equation: %FAT-DW = -0.235 (RI) + 0.252 (abdomen circumference)
+ 0.281 (sum of triceps + abdomen + thigh skinfolds) - 0.044.
anthropometry
alone had similar SEE’s that were lower
than those of the other independent
variables tested.
The relationships between FFB-DW and 1) RI alone, 2)
significant anthropometric
measurements alone, and 3)
RI plus the significant anthropometric
measurements
and reactance are presented in Fig. 1.
The results of the regression analyses with %FAT-DW
as the dependent variables are summarized in Table 4.
Body mass index had the lowest prediction accuracy.
Impedance index plus weight, RI plus weight, and RI
plus weight and reactance had the same adjusted R2 and
prediction accuracy. The SEE’s of the two equations
including only anthropometry
or RI plus anthropometry
were similar and were lower than for the other tested
independent variables. The equation that included RI
plus anthropometric
dimensions was similar to anthropometric dimensions alone with an adjusted R2 = 0.85
and SEE’s of 3.3 and 3.2% FAT, respectively. The relationships between %FAT-DW and 1) RI plus weight, 2)
significant anthropometric
measurements
only and 3)
RI plus the significant anthropometric
measurements,
are presented in Fig. 2.
In a cross-validation
study, three prediction equations
for FFB mass that were developed by Lukaski et al. (18,
19) on data from samples of young adults which included
RI alone and RI, body weight, and reactance combined
were used to predict FFB-DW values measured for our
sample of children. These predicted FFB-DW values are
summarized in Table 5. The predicted mean FFB-DW
values from the equations of Lukaski et al. (18, 19) were
all within 1.5 kg of our measured values. The SEE for
prediction of FFB-DW for the total sample, using the
equations published by Lukaski et al. that included only
RI (18,19), had the same SEE’s as the FFB-DW prediction equation that also included only RI and was developed using data from the children (SEE of 2.6 kg). When
the equation Luk3 reported by Lukaski et al. (18) was
used to estimate FFB-DW for the total sample of children the best prediction of FFB-DW was obtained (R2 =
0.93 and SEE of 2.0 kg). The Luk3 equation also resulted
in the best prediction of FFB-DW for boys (R2 = 0.95
and SEE of 1.73 kg) and girls (adjusted R2 = 0.91 and
SEE of 2.21 kg) when analvzed senaratelv.
si
g
45-
g
40-
cb 35t
30 25 -
AdjustedR* x 100 = 88.4
SEE = 2.6 kg
202530354045505560
FFB Estimatedfrom ResistanceIndex
25
AdjustedR* x 100= 92.2
SEE = 2.1kg
2ob
%’
’
’
’
’
202530354045505560
’
’
’
’
FFB Estimatedfrom Anthropometry
60~
55 50 s
s
45-
g
40-
ih
t=
3530 25 -
AdjustedR* x 100= 93.8
SEE= 1.9 kg
A
2ob
$0’
2025
’
’
’
30354045
’
’
’
’
505560
’
FFB Estimatedfrom ResistanceIndex,
Anthropometry,and Reactance
FIG. 1. Relationship
between fat-free body estimated from density
and total body water (FFB-DW) and predicted from resistance index
(RI) alone, anthropometry alone, and RI plus anthropometry and
reactance (best-fitting equation).
DISCUSSION
Results of the regression analyses indicated that the
height exponent of 2.0 used to compute the adult body
height2/resistance
(RI) is also the best exponent for
height to use to commute RI for children. Also, when
IMPEDANCE
BODY COMPOSITION
45r
IN CHILDREN
819
(18, 19), and Segel et al. (27) reported in adult samples.
0
5 1015202530354045
% Fat Estimatedfrom
ResistanceIndex and Weight
45
l
40 F
359 300
i 25-
;
ASSESSMENT
20-
8.
l
15-
‘e
.a:y
.8
lo-
0
‘I(
@t
4.
.
AdjustedR2 x 100 = 85.1
SEE = 3.2%
.f
5-
0’ ’ ’ ’ ’ ’ ’ ’ ’ c
0 5 1015202530354045
% Fat Estimatedfrom Anthropometry
45401
351
B
Q
&
30-
;
201
251
0.4
0%
. CI 89
151Or
--
5
0’
0.
l
:G
AdjustedR2 x 100= 84.6
SEE = 3.3%
0
t
se
0
’
’
’
’
’
’
’
’
’
0 5 1015202530354045
% Fat Estimatedfrom
ResistanceIndex and Anthropometry
FIG. 2. Relationship
between percent fat estimated from density
and total body water (%FFB-DW) and predicted from resistance index
(RI) and body weight, anthropometry alone, and RI plus anthropometry
(best-fitting equation).
resistance, reactance, impedance, or RI was used to predict FFB-DW for children in this study, the highest
correlations were found for RI. These results agree with
those of Hoffer et al. (12), Nyboer (21), Lukaski et al.
It has been shown in children that the proportion of
the water and bone mineral content of the FFB mass
changes during growth. These changes violate the underlying assumptions of the two-component body composition model, i.e., that FFB composition
is unchanging
within and among individuals. As a result of the violation
of these assumptions using a two-component model will
lead to significant errors in the estimation of FFB and
%FAT in children (l&16,33). A model that assesses the
body density (D) and body water (W) content will account for changes in the proportion of water in the FFB
and is a better model than one based only on densitometry to measure the criterion variables for validation
studies with children. In this study, the criterion variables (FFB-DW and %FAT-DW)
were estimated using
equations developed for children (2,X) based on a multicomponent model using measurements of body density
from densitometry and total body water from deuterium
dilution.
Because the exact error in multicomponent
FFB-DW estimates cannot be determined, all prediction
errors include errors associated with our criterion method
and this overestimates the actual prediction errors of the
bioelectric impedance approach.
The prediction accuracies (SEE’s) for FFB-DW from
RI alone or body weight alone were less than the SEE
for both RI and body weight combined. When RI was
combined with the categorical variables of age, sex, sexual maturity, and fatness category (obese or nonobese)
in multiple regression analyses, all of these independent
variables were significant. Thus the relationship between
RI and FFB appears to be influenced by other variables.
When significant anthropometric
measurements were
included with RI and the categorical variables, the categorical variables were no longer significant and there was
an improvement
in prediction accuracy (SEE) and adjusted R’. Thus in children use of anthropometric
dimensions with RI improves the SEE for FFB-DW estimates.
The prediction equation for FFB-DW that was developed from the multiple regression analyses and had the
smallest prediction error included RI, chest circumference, hip skeletal width, and reactance. For the total
sample the adjusted R2 = 0.94 and SEE of 1.9 kg for
FFB-DW. The anthropometric
variables in this equation
indicate that measurements reflecting trunk size are an
important
addition to RI for predicting FFB-DW. The
developed prediction equation with the next smallest
prediction error (adjusted R2 = 0.93 and SEE of 2.0 kg)
included RI, chest circumference, and hip skeletal width.
The addition of reactance to this equation resulted in
only a small improvement
of prediction accuracy (SEE
decreased by 0.1 kg) as it did in the prediction equation
reported by Lukaski et al. (18).
The SEE’s of the best-fitting equations developed for
predicting FFB-DW in this sample are comparable to
those reported for adults by Lukaski et al. (18, 19) and
Segal et al. (27). Lukaski et al. and Segal et al. predicted
FFB from body density. Lukaski et al. (19) reported
SEE’s of 2.61 kg for prediction of FFB using the prediction equations that included only RI and had been de-
820
TABLE
IMPEDANCE
5. Cross-validation
Subj
Group
Total
sample
(n = 94)
analysis
BODY
COMPOSITION
of five published
FFB-DW
Measured
Values for
Children, kg
Predicted
LUKlt
Mean
36.29
35.60
FFB-DW
LUKZ$
Mean
ASSESSMENT
adult fat-free
36.58
36.37
37.04
35.66
Females
(n = 41)
35.90
body prediction
equations
that include RI
R
Adjusted
R2
SEE
Regression
Coef
Intercept
0.94
0.94
0.96
0.96
0.96
0.98
0.93
0.93
0.95
0.88
0.88
0.93
0.92
0.92
0.95
0.86
0.86
0.91
2.57
2.57
2.04
2.15
2.15
1.73
2.72
2.72
2.21
0.979*
0.993”
1.029*
0.918*
0.931*
0.977*
1.159*
1.176*
1.150*
1.45
0.28
0.19
3.19
2.09
1.74
-4.21
-5.60
-3.63
LUK3$
Mean
35.10
(n = 53)
CHILDREN
Values for Children
36.28
Males
IN
34.60
35.29
34.37
RI, resistance
index; FFB-DW,
fat-free
body mass estimated
from body density
(D) and body water from deuterium
water space (W). R,
correlation
coefficient;
SEE, standard
error of estimate.
* P 5 0.05. t LUKl
= (0.85 x RI) + 3.04 (Ref. 19). $ LUK2 = (0.838 X RI) + 4.179 (Ref.
18). 5 LUK3
= (0.756 X RI) + (0.110 X weight)
+ (0.107 X reactance)
- 5.463 (Ref. 18).
veloped from their sample of healthy adult males. Using
prediction equations that included resistance, body mass,
and reactance and had been developed from a sample of
male and female subjects, Lukaski and co-workers (18)
reported a SEE of 2.06 kg for FFB.
When three adult-based equations published by Lukaski et al. (18, 19) for predicting FFB that include RI
alone or RI and body weight were used to predict the
measured FFB-DW values for children in a cross-validation study, the results showed a high degree of prediction accuracy for FFB-DW for all three equations as
indicated by high adjusted R2 values (0.88-0.95) and
similar or smaller SEE’s than those reported when the
equations were used to predict FFB values in the original
samples (1.73-2.72 kg). Also, mean predicted values of
FFB-DW were similar. The regression equations developed for children differ only slightly from those developed for adults, with the equations for children having a
greater weight given to body weight than the equations
for adults reported by Lukaski et al. (18, 19).
The prediction errors for %FAT-DW for the children
from RI and body weight were smaller than from body
mass index (BMI). When RI and body weight were
combined with categorical variables of age, sexual maturity, and sex by fatness interaction in multiple regression analyses, all of these independent variables were
significant predictors of %FAT-DW.
When significant
anthropometric
measurements
were included with RI
and body weight, only RI and the anthropometric
measurements
were significant
variables for predicting
%FAT-DW.
Again, use of anthropometric
dimensions
along with RI improved prediction of body composition.
The prediction equation for %FAT-DW that was developed from multiple
regression analyses and had the
smallest prediction error included RI, abdomen circumference, and sum of triceps, abdomen, and thigh skinfolds. For the total sample the best-fitting equation has
an adjusted R2 = 0.85 and SEE of 3.3% for %FAT-DW.
The SEE’s from the best-fitting equation for predicting %FAT-DW
for this study are comparable to those
reported by Lukaski et al. (18) and lower than those
reported by Segal et al. (27). Lukaski and co-workers
(18) estimated %FAT using body density estimated from
densitometry
and the equation from Brozek et al. (4).
Lukaski et al. (18) reported SEE’s of 2.9% fat for males
using prediction equations that included only RI and
developed from their sample of adult males. Using prediction equations developed from a sample of male and
female subjects that included RI, Lukaski and co-workers
(18) reported a SEE of 2.7% fat for males and 3.1% fat
for females. Segal et al. (27) using a sample of adult
males and females reported a SEE of 6.1% fat using
prediction equations for estimating body density provided by RJL Systems and then calculated percent fat
using the equation of Siri (28).
In summary, we conclude that the high level of withinday and between-day reliability of resistance and reactance measurements combined with the relatively small
prediction errors (SEE) for FFB-DW and %FAT-DW
indicate that BIA is a valid method of estimating these
aspects of body composition in lo- to Nyr-old
healthy
Caucasian children. The prediction
of FFB-DW and
%FAT-DW are significantly
improved by including selected anthropometric
variables (for FFB-D W: chest circumference and hip skeletal width; for %FAT-DW: abdomen circumference and sum of triceps, abdomen, and
thigh skinfolds) along with RI in the prediction equation.
However, the prediction
accuracy for FFB-DW and
%FAT-DW from anthropometry
alone (FFB-DW SEE
of 2.1 kg; %FAT-DW
SEE of 3.2%) is very similar to
that from RI and selected anthropometric
variables
(FFB-DW SEE of 1.9 kg; %FAT-DW
SEE of 3.3%).
Before the general utility is established for the equations
developed in this study for predicting
FFB-DW and
%FAT-DW developed from this sample of children, these
equations need to be cross-validated in other populations
of children using a multiple-component
body composition model to measure the criterion variables.
We thank Cheri Carswell,
Gary Feiss, and Michael
Hewitt for their
technical
assistance
and Leah Brown
for manuscript
preparation.
This study was supported
in part by grants from the National
Institutes
of Health
(AM-355186)
and RJL Systems, Detroit,
MI.
Received
11 March
1988; accepted
in final
form
20 September
1988.
REFERENCES
1. AKERS,
utilizing
1969.
R., AND E. R. BUSKIRK.
“force cube” transducer.
An underwater
weighing
system
J. Appl. Physiol. 26: 649-652,
IMPEDANCE
2. BOILEAU,
and body
BODY
COMPOSITION
R. A., T. G. LOHMAN,
AND M. H. SLAUGHTER.
Exercise
composition
in children
and youth. Stand. J. Sport Sci.
7: 17-27,1985.
3. BOILEAU, R. A., T. G. LOHMAN,
4.
5.
ASSESSMENT
M. H. SLAUGHTER,
T. E. BALL,
S. B. GOING, AND M. K. HENDRIX.
Hydration
of the fat-free
body
in children
during maturation.
Hum. Biol. 56: 651-666, 1984.
BROZEK, J., F. GRANDE, J. T. ANDERSON,
AND A. KEYS. Densitometric analysis of body composition:
revision
of some quantitative
assumptions.
Ann. NY Acad. Sci. 110: 113-140, 1963.
CAMERON,
J. R., AND J. A. SORENSON.
Measurement
of bone
mineral
in vivo: an improved
method. Science Wash. DC 140: 230-
1986.
19. LUKASKI,
14.
of Whole-Body
Bioelectric
Impedance
Analysis
for Body Composition
Assessment
in Nonobese
and Obese
Children
and Youth
(Dissertation).
Tucson,
AZ: University
of
Arizona,
1986, p. 240-243.
KUSHNER,
R. F., AND D. A. SCHOELLER.
Estimation
of total body
water by bioelectrical
impedance
analysis.
Am. J. Clin. Nutr.
44:
417-424,1986.
15. LOHMAN, T. G. Applicability
16.
of body composition
techniques
and
constants
for children
and youths. In: Exercise and Sport Sciences
Reviews,
edited by K. B. Pandolph.
New York: Macmillan,
1986,
p. 325-357.
LOHMAN,
T. G. Research
progress
in validation
of laboratory
methods
of assessing body composition.
Med. Sci. Sports Exercise
16:596-603,1984.
17. LOHMAN,
T. G. Research
relating
to assessment
of skeletal status.
In: Body-Composition
Assessment
in Youth and Adults,
Report of
the Sixth Ross Conference
on Medical
Research.
Columbus,
OH:
Ross Labs., 1985, p. 38-41.
18. LUKASKI,
H. C., W. W. BOLONCHUK,
C. B. HALL,
AND W.
A.
bioelectrical
impedance
method
J. Appl. Physiol. 60: 1327-1332,
H. C., P. E. JOHNSON,
W. W. BOLONCHUK,
LYKKEN. Assessment
of fat-free mass using bioelectrical
measurement
of the human body. Am. J. Clin. Nutr.
AND G. I.
impedance
41: 810-817,
1985.
20. MILLS,
21.
6. FORBES,
27:531-534,1969.
13. HOUTKOOPER,
L. B. Validity
821
CHILDREN
SIDERS. Validation
of tetrapolar
to assess human body composition.
232,1963.
G. B. Body
composition
in adolescence.
In: Human
Growth,
edited by F. Falkner
and J. M. Tanner.
New York: Plenum,
1978, vol. 2, p. 239-272.
7. GARN, S. M., D. C. CLARK, AND K. E. GUIRE. Growth,
body
composition,
and development
of obese and lean children.
Cur.
Concepts. Nutr.
3: 23-46, 1975.
8. HARSHA, D. W., R. R. FRERICHS, AND G. S. BERENSON. Densitometry and anthropometry
of black and white children
and youth.
Hum.Biol.
50: 261-280,1978.
9. HASCHKE, F., S. J. FOMON, AND E. E. ZIEGLER. Body composition
of a nine-year-old
reference
boy. Pediatr.
Res. 15: 847-849, 1981.
10. HASCHKE,
F. Part I. Total body water in normal adolescent
males.
Acta Paediatr.
Stand. Suppl. 307: 3-12, 1983.
11. HEALD, F. P., E. E. HUNT, R. SCHWARTZ, C. D. COOK, D. ELLIOTT,
AND B. VAJDA. Measures
of body fat and hydration
in adolescent
boys. Pediatrics
31: 226-239, 1963.
12. HOFFER,
E. C., C. K. MEADOR,
AND D. C. SIMPSON.
Correlation
of whole-body
impedance
with total body water. J. Appl. Physiol.
IN
22.
23.
24.
25.
26.
27.
W. J., AND D. RAU. University
of Alaska,
AnchorageSection of High Latitude
Study, and the Mt. McKinley
Project.
AlaskaMed.
25:21-28,1983.
NYBOER, J. Guiding
concepts
in the science of impedance
plethysmography.
In: Basic Factors in BioeZectricaZ
Impedance
Plethysmography
of Cardiac
Output,
Lung
Volumes,
and the Cerebral
Circulation.
Pittsburgh,
PA: Instrument
Sot. Am., 1970, p. 38-51.
NYBOER, J. Workable
volume and flow concepts of biosegments
by
electrical
impedance
plethysmography.
J. Life Sci. 2: 1-13, 1972.
PEDHAUSER,
E. J. Multiple
Regression
in Behavioral
Research. New
York: Holt Reinhart
& Winston,
1982.
PETHIG, R. Dielectric
and Electronic
Properties
of Biological Materials. Chicester,
UK: Wiley, 1979, p. 225-235.
ROCHE, A. F. Research
progress
in the field of body composition.
Med. Sci. Sports Exercise
16: 579-583,1984.
Ross, J. G., C. 0. DOTSON,
G. G. GILBERT, AND S. J. KATZ. New
standards
for fitness measurement.
The National
Children
and
Youth
Fitness
Study.
J. Phys. Ed. Rec. Dance January,
62-66,
1985.
SEGAL, K. R., B. GUTIN, E. PRESTA, J. WANG, AND T. B. VAN
ITALLIE.
Estimation
of human
body composition
by electrical
impedance
methods:
a comparative
study. J. Appl. Physiol.
58:
1565-1571,1985.
28. SIRI, W. E. Body
composition
from fluid spaces and density:
analysis of methods.
In: Techniques
for Measuring
Body Composition. Washington,
DC: Natl. Acad. Sci., 1961, p. 223-244.
29. SLAUGHTER,
M. H., T. G. LOHMAN,
R. A. BOILEAU,
R. J. STILLMAN, M. VAN LOAN, C. A. HORSWILL,
AND J. H. WILMORE.
Influence
of maturation
on relationship
of skinfolds
to body density: a cross-sectional
study. Hum. Biol. 56: 681-689,
1984.
30. TANNER, J. M. Growth
of Adolescence
(2nd ed.). Oxford,
UK:
Blackwell,
1962.
31. THOMASSET, A. Bio-electrical
properties
of tissue impedance
measurements.
Lyon. Med. 207: 107-118,1962.
A. Bio-electrical
properties
of tissues. Lyon. Med.
32. THOMASSET,
209:1325-1352,1963.
33. WILMORE,
J. H. A reaction
34.
35.
to the manuscripts
of Roche and
Buskirk.
Med. Sci. Sports Exercise 16: 594-595,
1984.
WILMORE,
J. H. A simplified
method for determination
of residual
lung volumes. J. Appl. Physiol. 27: 96-100, 1969.
WONG, W. W., W. J. COCHRAN,
W. J. KLISH, E. O’BRIAN
SMITH,
L. S. LEE, AND P. D. KLIEN. In vivo isotope-fractionation
factors
and the measurement
of deuteriumand oxygen-18dilution
spaces
from plasma, urine, saliva, respiratory
vapor and carbon dioxide.
Am. J. CZin. Nutr. 47: l-6, 1988.