1
Caucasian children’s fat mass: routine anthropometry versus airdisplacement plethysmography Nathalie Michels1, Inge Huybrechts1, Karin Bammann2,3, Lauren Lissner4, Luis Moreno5, Maarten Peeters6,7, Isabelle Sioen1,7, Barbara Vanaelst1,7, Krishna Vyncke1,7, Stefaan De Henauw1,8 1
Department of Public Health, Faculty of Medicine and Health Sciences, Ghent University, De
Pintelaan 185, 2 Blok A, B-9000 Ghent, Belgium 2
Institute for Public Health and Nursing Care Research (IPP), University of Bremen, Grazer Str. 2
D-28359 Bremen, Germany 3
BIPS Institute for Epidemiology and Prevention Research, Achterstrasse 30, D-28359 Bremen,
Germany 4
Department of Public Health and Community Medicine, Sahlgrenska Academy, University of
Gothenburg, 405 30 Göteborg, Sweden 5
GENUD (Growth, Exercise, Nutrition and Development) research group, School of Health
Sciences, University of Zaragoza, Domingo Miral s/n, 50.009 Zaragoza, Spain 6
Research Centre for Exercise and Health, Department of Biomedical Kinesiology, Faculty of
Kinesiology and Rehabilitation Sciences K.U. Leuven, Belgium 7
Research Foundation – Flanders, Egmontstraat 5, B-1000 Brussels, Belgium
8
Department of Health Sciences, Vesalius, Hogeschool Gent, Keramiekstraat 80, B-9000 Ghent,
Belgium
Running title: Anthropometry versus BOD POD in children Keywords: air-displacement plethysmography, anthropometry, bioelectrical impedance, child, skinfold thickness
*
Corresponding Author:
Nathalie Michels - Ghent University De Pintelaan 185 - 2 Blok A 9000 Gent, Belgium
[email protected] Tel: 0032 9 332 36 85, Fax: 0032 9 332 49 94
2
1
ABSTRACT
2 3
This paper will use fat mass percentage obtained via BOD POD® air-displacement
4
plethysmography
5
measurements/indices and 2) of fat mass percentage assessed with equations (FMeq%) based on
6
skinfold thickness and bioelectrical impedance. In 480 Belgian children (5 to 11 years old) weight,
7
height, skinfold thickness (triceps and subscapular), body circumferences (mid-upper arm, waist
8
and hip), foot-to-foot bioelectrical impedance (Tanita®) and FMADP% were measured.
9
Anthropometric measurements and calculated indices were compared with FMADP%. Next,
10
published equations were used to calculate FMeq% using impedance (equations of Tanita®, Tyrrell,
11
Shaefer and Deurenberg) or skinfold thickness (equations of Slaughter, Goran, Dezenberg and
12
Deurenberg). Both indices and equations performed better in girls than in boys. For both sexes, the
13
sum of skinfold thicknesses resulted in the highest correlation with FMADP%, followed by triceps
14
skinfold, arm fat area and subscapular skinfold. In general, comparing FMeq% with FMADP%
15
indicated mostly an age and sex effect and an increasing underestimation but less dispersion with
16
increasing FM%. The Tanita® impedance equation and the Deurenberg skinfold equation performed
17
the best, although none of the used equations were interchangeable with FMADP%. In conclusion,
18
the sum of triceps and subscapular skinfold thickness is recommended as marker of FM% in the
19
absence of specialized technologies. Nevertheless, the higher workload, cost and survey
20
management of an immobile device like the BOD POD® remains justified.
21 22
(FMADP%)
to
examine
the
relative
validity
of
1)
anthropometric
3
23
1. Introduction
24
Air-displacement plethysmography (ADP), integrated in the commercially available system BOD
25
POD® is a validated technique to assess body composition
26
compartment model of including a quick, comfortable, automated, non-invasive and safe
27
measurement process, making it feasible for children. As the best performing two-compartment
28
model, ADP is more reliable for body composition than routine anthropometric measurements
29
Nevertheless, the conversion from body density obtained by ADP measurement to fat mass
30
percentage (FMADP%) needs consideration in children. As the chemical maturation of lean tissue
31
changes with age and proceeds differently in males and females, age- and sex-adjusted factors need
32
to be considered for children as was done by Lohman and more up-to-date by Wells (3,4).
(1)
. It has the advantage over the four-
(2)
.
33
Although ADP is considered as a more feasible method for large scale surveys in
34
comparison with the four-compartment model, the immobile aspect of ADP could also be a
35
limitation for large scale surveys. Due to these logistic and budgetary constraints, examinations in
36
large scale epidemiological studies including the assessment of body composition are often
37
restricted to only routine measurements (weight, height, circumferences, skinfold thickness and
38
bioelectrical impedance (BIA)). However, the accuracy of routine anthropometric measurements in
39
children is still doubted, although inter- and intra-observer error can be controlled when thorough
40
training is carried out (5). Furthermore, a variety of techniques are used (e.g. whole body vs foot-to-
41
foot BIA) and equations to convert the measurement in FM% are population-specific. Especially in
42
children, the assessment of body composition remains a challenging task (6,7).
43 44
In this paper, the validity of anthropometric measurements/indices and of FM% equations (FMeq%)
45
based on anthropometric measurements (skinfolds and BIA) was investigated with FMADP%, using
46
as reference method the BOD POD® device with the up-to-date Wells adjusting factors for children.
47
As such, research groups without specialized technologies such as ADP can make a well-founded
48
choice on the routine anthropometric measurements, indices and equations to be used in obtaining a
49
good indication of children’s body fat. Furthermore, it is interesting to examine whether the bigger
50
workload, cost and the more complicated survey management with an immobile device like the
51
BOD POD® is justified in large-scale surveys (i.e. are the anthropometric measurements
52
interchangeable with ADP). This study will help in these two decisions using a large child
53
population (n=480) in which a large battery of anthropometric measurements was performed.
54
4
55
2. Methods and materials
56
2.1.
Population
57
Participating children were part of the Belgian control region (i.e. Aalter, a city in Flanders, the
58
northern, Dutch-speaking part of Belgium) of the IDEFICS study, funded within the European Sixth
59
Framework Programme. The aim of the IDEFICS study was to identify and prevent dietary- and
60
lifestyle-induced health effects in infants and children
61
random cluster sampling (all children from a selection of schools in the control city) and 750
62
children participated in the routine anthropometric measurements at school (participation rate of
63
41.3%, 11% overweight). Of these 750, 480 children (52.3% male, 6.7% overweight) between 5-11
64
years old agreed to undergo the extra ADP measurement at the survey center within maximal 3
65
weeks of the school measurement. Data was collected from February to May 2010. All children
66
were Caucasian except for five girls of African-American origin. This study was conducted
67
according to the guidelines laid down in the Declaration of Helsinki and all procedures were
68
approved by the ethics committee of Ghent University. Written informed consent was obtained
69
from all parents.
(8)
. In Belgium, children were selected by
70 71
2.2.
Body composition measurements
72
Routine anthropometric measurements
73
All field work was carried out by two observers and intra- and inter-observer reliability was
74
enhanced by extensive training (5). The children were measured barefooted in underwear and /or T-
75
shirt.
76
Weight was measured in fasting status with an electronic scale (TANITA® BC 420 SMA, Germany)
77
to the nearest 0.1 kg. Height was measured with a telescopic height measuring instrument (SECA
78
225, UK) to the nearest 0.1 cm. The body mass index (BMI) z-score was obtained by calculating the
79
BMI (BMI=weight(kg)/height(m)²) and adjusted for age and sex using British 1990 growth
80
reference data (9). Overweight was determined by the International Obesity Task Force classification
81
(10)
82
Leg-to-leg impedance (ohm) was measured with the electronic TANITA® BC 420 SMA scale
83
(prototype adapted to the small foot size of children). As impedance is dependent on length of the
84
conductor, an impedance index reflecting the fat-free mass was defined as impedance
85
index=height²/impedance. To reflect fat-mass, ‘weight minus impedance index’ was calculated.
86
Skinfold thicknesses (mm) were measured twice on the right side of the body to the nearest 0.2 mm
87
with a skinfold calliper (Holtain, UK, range 0-40 mm) according to the international standards for
.
5
(11)
88
anthropometric assessment
89
skinfold (TSF) was taken halfway between the acromion process and the olecranon process at the
90
back side of the arm. The subscapular skinfold (SSF) was measured 20 mm below the tip of the
91
scapula, at an angle of 45° to the lateral side of the body. If the first and second measurement of the
92
skinfolds differed more than 2 mm, a third measurement was performed.
93
Circumferences (cm) were measured once with an inelastic tape (Seca 200, precision 0.1 cm, range
94
0-150 cm) with the subject in a standing position. Circumference measurements were performed at
95
the following three sites: 1) mid-upper arm (MUAC) relaxed arm, halfway between the acromion
96
process and the olecranon process; 2) waist, halfway between the top of the iliac crest and the lower
97
coastal border (10th rib); 3) hip, at the maximum extension of the buttocks.
98
Equations found in literature were used to calculate FMeq% as long as they were at least partly
99
based upon a population sample that included children and if all needed parameters were measured
100
in this study. For BIA, only equations using the impedance (as we have measured), and not the
101
resistance or reactance were selected. The selected equations based on impedance or skinfold
102
thickness and their characteristics are listed in Table 1. Apart from the built-in Tanita® equation (12),
103
the equations of Schaefer
104
impedance measurements and the equations of Slaughter (20)
and the mean of both measurements was calculated. The triceps
(13)
, Deurenberg 1 and 2
(14,15)
and Tyrrell (17)
, Goran
(16)
were selected for the
(18)
, Deurenberg 3
(19)
and
105
Dezenberg
106
Indices. Apart from BMI, also Rohrer’s Index (RI= weight[kg]/height[m]³), arm fat area (AFA=
107
(MUAC[cm]²/4pi)
108
waist[cm]/hip[cm]), waist-to-height ratio (WHtR= waist[cm]/height[cm]) and the Conicity Index
109
(ConI= waist[cm]/(0.109* weight [kg] / height [cm] ) were calculated to assess the correlation
110
between these indices and the FMADP%.
for skinfold measurements.
-
(MUAC[cm]-(pi*TSF[cm]))²/4pi)),
waist-to-hip
ratio
(WHR=
111 112
Reference method: ADP
113
Body volume was measured by ADP (BOD POD®, Software version 4.2.4, Life Measurement Inc,
114
Cranlea and Co, Birmingham, United Kingdom) using standardized procedures (21). Children had to
115
refrain from physical activity and food two hours before the measurement. The BOD POD® was
116
calibrated daily and at each measurement according to the manufacturer's guidelines. Children were
117
assessed in tight-fitting bathing suit with swim cap to rule out air trapped in clothes and hair. If the
118
first two readings for body volume differed by more than 150 ml, a third measurement was taken
119
and the two values that were closest and within the criteria for agreement were averaged. Thoracic
6
120
gas volume was predicted by the software with a validated child-specific equation (22). FMADP% was
121
calculated using the up-to-date child-specific conversion factors reported by Wells (4).
122 123
2.3.
Statistical analyses
124
Means and standard deviations were given for all measurements and the association with sex and
125
age was tested. As indices were not normally distributed, Mann-Whitney U tests and Spearman’s
126
correlation coefficients were employed.
127
For the first objective, the correlation between anthropometric measurements/indices and FMADP%
128
was determined using age-adjusted Spearman correlations.
129
The second objective was to compare FMeq% using skinfold measurements or BIA (see Table 1)
130
with FMADP%. First of all, mean difference +/- SD between FMADP% and FMeq% was given; its
131
significance was determined with paired t-test using Bonferroni adjustment. As the difference is
132
calculated as ‘FMADP% minus FMeq%’, a positive mean difference indicates an underestimation by
133
FMeq%. To compare these results with those of the anthropometric indices, age adjusted
134
correlations were calculated.
135
Accuracy and precision were examined using regression analysis. As nine equations were tested
136
separately for boys and girls, the results of 18 regression analyses were noted. If age was a
137
significant predictor, the results of the multiple regression were given (both FMeq% and age as
138
predictors), otherwise of the simple regression. The FMeq% was considered accurate when the
139
regression between FMADP% and FMeq% did not differ significantly from the line of identity (slope
140
not significantly different from 1 and intercept not significantly different from 0). The precision of
141
FMeq% was assessed by the R² and the standard error of the estimate (SEE).
142
The presence of a sex-effect on the FM% difference was separately examined by multiple
143
regression analyses with each time one of the nine FMeq% and sex as predictors. To examine
144
whether there was a real sex effect or whether this was induced due to the FM% difference between
145
sexes, a FMeq%-sex interaction term was included as an extra variable in this regression.
146
Bland-Altman analysis was used to examine the agreement between FMeq% and FMADP%. Apart
147
from the 95% limits of agreement (LOA), also the presence of a direction (=heteroscedacity:
148
upwards/downwards) or dispersion (convergent/divergent) trend across the range of fatness were
149
examined. A downwards or upwards direction trend was present if there was a significant univariate
150
regression line between “FMADP%-FMeq%” and the mean values of FMADP% and FMeq%. A
151
potential dispersion trend (convergent/divergent from the zero point) was visually evaluated. As the
152
difference was calculated as ‘FMADP% minus FMeq%’, an upwards Bland-Altman line indicates an
7
153
underestimation by FMeq% in children with a higher fat mass (if the Bland-Altman regression line
154
goes through the zero point). A convergent dispersion trend demonstrates more FM% agreement
155
between both methods with increasing FM%, while a divergent trend shows more disagreement
156
with increasing FM%.
157
Finally, single measure intraclass correlation (ICC) was calculated as interchangeability measure
158
using a two-way mixed model with absolute agreement. Interchangeability has been suggested to be
159
excellent when the ICC coefficient is higher than 0.75 (23).
160
All statistical analyses were performed using SPSS/PASW version 19 (IBM Corp, USA). To correct
161
for multiple testing when using the FM% equations, a Bonferroni correction was applied with a p-
162
value of 0.003 (0.05/18; 9 equations*2 sexes= 18 tests) as the threshold of significance. For all
163
other analyses, the two-sided level of significance was set at
