European Journal of Clinical Nutrition (1999) 53, 468±472
ß 1999 Stockton Press. All rights reserved 0954±3007/99 $12.00
http://www.stockton-press.co.uk/ejcn
Differences in the relationship between body fat and body mass
index between two different Indonesian ethnic groups: The effect
of body build
S Gurrici1, Y Hartriyanti1, JGAJ Hautvast2 and P Deurenberg2*
1
School of Public Health, Gedung D Lt.II, University of Indonesia, Depok, Indonesia; 2Department of Human Nutrition and
Epidemiology, Wageningen Agricultural University, The Netherlands
Objective: To study the relationship between body fat percent (BF%) and body mass index (BMI) in two
different Indonesian ethnic groups (Malays and Chinese) and to relate differences in the relationship to
differences in body build and slenderness.
Design: Cross-sectional study.
Subjects: Except for ethnicity, not specially selected populations living on Java (Depok, south of Jakarta: Malay
Indonesians, n ˆ 117) and on Sulawesi (Makale, north of Ujung Pandang: Chinese Indonesians, n ˆ 109).
Measurements: Weight, height, sitting height, waist and hip circumferences and skeletal widths were measured.
BMI was calculated and BF% was predicted from BMI, age and sex using a (Dutch) Caucasian prediction
formula. Slenderness was expressed as the ratio of weight : sum of knee and wrist width. BF% assessed by
deuterium oxide dilution was used as a reference.
Results: BF% in the male and female Malay Indonesians was 24.6 7.0 and 35.6 5.6% respectively which
was not signi®cantly different from the values in the male and female Chinese Indonesians (24.0 4.3 and
33.8 6.9%). BMI and age were signi®cantly lower in the Malay Indonesians. Malay Indonesians had a more
slender body build in terms of skeletal widths compared to the Chinese Indonesians, and they had a higher
slenderness index. BF% predicted from BMI using a Caucasian prediction formula was underestimated by
5.8 4.8% and 7.7 3.8% in the male and female Malay Indonesians but only by 1.3 3.0% and 1.7 3.7% in
the male and female Chinese Indonesians. After correction for differences in age, sex and BF% the Malay
Indonesians had a 1.7 0.3 kg=m2 (P < 0.0001) lower BMI than the Chinese Indonesians. After correcting for
body build and relative sitting height the difference lowered to 0.9 0.4 kg=m2 (P < 0.02).
Conclusions: The study con®rmed the results of an earlier study that Indonesians have a higher BF% at the same
BMI compared to Caucasians, but that there are apparently also differences among Indonesian subgroups. These
differences are at least partly related to differences in body build.
Keywords: body fat percent; body mass index; body build; obesity; cut-off values; Indonesians; Malays;
Chinese; race; ethnicity
Introduction
In obesity or adiposity the body fat percent (BF%) exceeds
25% in males and 35% in females (WHO, 1995; Deurenberg et al, 1998). BF% can be measured by a number of
methods (Forbes, 1987; Lukaski, 1987; Deurenberg 1992),
but most of them are rather laborious and not suitable for
use in epidemiological surveys. In several studies a high
correlation was found between body mass index (BMI) and
BF% when age and sex are taken into account (Womersley
& Durnin, 1977; Deurenberg et al, 1991; Gallagher et al,
1996), and therefore in most epidemiological studies BMI
is used as an indicator of body fatness. The World Health
* Correspondence: Dr Paul Deurenberg, Department of Human Nutrition
and Epidemiology, Wageningen Agricultural University, Bomenweg 2,
6703 HD Wageningen, The Netherlands.
E-mail: [email protected]
Guarantor: S Gurrici.
Contributors: Organization of the study was done by Gurrici for the
Indonesian part and by Deurenberg and Hautvast for the Dutch part. The
practical measurements were performed by Gurrici, Hartriyanti and
Deurenberg. Data analysis was done by Hartriyanti and Deurenberg. Paper
writing and discussion was a joint effort of all four authors.
Received 7 September 1998; revised 7 January 1999; accepted 11 January
1999
Organization (WHO) suggests cut-off points for overweight
and obesity at a BMI value 25 kg=m2 and 30 kg=m2
respectively (WHO, 1995, 1998). There are indications that
the relationship between BMI and BF% is different among
ethnic groups (Deurenberg et al, 1998). Recently we
showed in a comparative study that Indonesians living in
Palembang, Sumatra, have generally 4.8 percent points
more body fat compared to Dutch subjects, having the
same weight, height, age and sex. Indonesian having the
same BF%, age and sex generally have a 2.9 kg=m2 lower
BMI compared to the Dutch (Guricci et al, 1998). Differences between ethnic groups in the BMI=BF% relation are
also found in other studies (Swinburn et al, 1996; Wang et
al, 1994; Luke et al, 1997), but not in all (Gallagher et al,
1996; Deurenberg et al, 1997). These differences among
ethnic groups could be due to differences in physical
activity, people with lower physical activity level having
relatively less muscle mass and thus more fat mass at the
same body weight. Although this is certainly true at an
individual level (for example, a body builder compared to
an obese subject with a very sedentary life style) it may not
be the only possible explanation at a population level. Also,
body build could be responsible for a different BMI=BF%
relationship. Norgan discussed in several papers the impact
Relationship between BF% and BMI in Indonesia
S Gurrici et al
of relative sitting height (in fact leg length) on BMI
(Norgan, 1994a, b), but this effect was not found by
others (Fehily et al, 1990). Also body build in term of
frame size and thus skeletal mass could be responsible for a
different relationship between BMI and BF% (Deurenberg
et al, 1998; Guricci et al, 1998; Fehily et al, 1990; Rookus
et al, 1985). In addition, a more stocky build is likely to
coincide with more muscle and connective tissue and
would for that reason add to a higher fat free mass=fat
mass ratio at the same BMI.
To test the hypothesis that differences in body build can
be responsible for differences in the BMI=BF% relationship, two Indonesian population groups of different ethnic
background were studied. The study is part of a larger
epidemiological study carried out in Indonesia to get
insight into body composition and body composition differences between ethnic groups.
until analyses. Deuterium in plasma was analysed by
infrared spectroscopy after sublimation of the plasma
(Lukaski & Johnson, 1985). Total body water (TBW) was
calculated from the given dose and the deuterium concentration in plasma, after correcting for 5% non-aqueous
dilution of deuterium (Forbes, 1987). Fat free mass
(FFM) was assessed from TBW assuming 0.726% hydration of the FFM (Pace & Rathbun, 1945). BF% was
calculated as
100*…weight ÿ FFM†=weight
BF% was also predicted from BMI with age- and sexspeci®c prediction formulas using a formula developed in a
Dutch population (Deurenberg, 1992):
BF% ˆ 1:20*BMI ‡ 0:23*age ÿ 10:8*sex ÿ 5:4 …1†
and a formula developed in an Indonesian population
(Guricci et al, 1998):
BF% ˆ 1:30*BMI ÿ 9:8*sex ‡ 6:4
Subjects and methods
Subjects from two different ethnic groups participated in
the study. The measurements were performed at two
different sites in Indonesia within two weeks and the
same investigators performed the same standardised measurements. In Depok, located on Java, about 40 km south of
Jakarta, 117 Malay Indonesians were measured. Most of
them were co-workers of the University of Indonesia. In
Makale, a small city on Sulawesi, about 250 km north of
Ujung Pandang, 109 subjects of Chinese ancestry were
measured. They were predominately personnel of a regional public health centre. At both study sites the subjects
were measured in the morning hours, at least 2 hours postprandial but most of them were in the fasting state from the
preceding evening.
Body weight was measured without shoes on an electronic weighing scale, accurate to 0.1 kg. To correct for
clothing 1 kg was subtracted from the readings. Body
height (without shoes) was measured to the nearest
0.5 cm with a wall-mounted stadiometer. Body mass
index (BMI, weight=height2, kg=m2) was calculated. Sitting
height was also measured and relative sitting height was
calculated as sitting height=standing height. Skinfold thickness was measured at four sites (biceps, triceps, subscapular, supra-iliac) on the left side of the body, and
body fat percent (BF%) was calculated according to
Durnin & Womersley (1974). Waist circumference was
measured midway between the lower rib margin and the
iliac-crest to the nearest 0.5 cm. Hip circumference was
measured over the great trochanters to the nearest 0.5 cm.
From waist circumference BF% was predicted using the
formula of Lean et al (1996). Wrist width (over the radius
and the ulna) and knee width (over the femur) was
measured to the nearest mm on the left and right side of
the body. Mean values of left and right sides were used in
calculations. Skeletal mass was predicted using the formulas of von DoÈbeln (1959), based on height, knee and wrist
width. An index of slenderness was calculated as
height=(sum of wrists and knee widths). Obviously, a
higher index indicates a more slender body build.
After the anthropometric measurements had been taken,
the subjects drank an accurately (to 0.01 g) weighed dose of
10 g deuterium oxide. After 2.5 ± 3 h a venous blood sample
was drawn and plasma was separated and stored at 7 10 C
…2†
Data were analysed using SPSS (release 7.5) for Windows
(SPSS=Windows, 1995). Differences between variables in
the same subjects were tested by Student's paired t-test.
Differences between groups were tested by analysis of (co)variance, using the GLM procedure. Stepwise multiple
linear regression was used with gender and site coded as
dummy variable (females ˆ 0, males ˆ 1; Depok ˆ 0,
Makale ˆ 1). Differences between measured and predicted
values (residuals) were evaluated with the technique
described by Bland & Altman (1986). Differences were
regarded as statistical signi®cant if P < 0.05 (two-sided,
where appropriate). Values are given as mean s.d., unless
otherwise indicated.
Results
Table 1 gives some physical characteristics of the subjects.
The subjects in Depok were younger and were slightly
taller compared to the subjects in Makale. Relative sitting
height was slightly different between the two sites but was
signi®cant only in males. BF% as assessed by deuterium
oxide dilution was not different between males in Depok
and Makale, but the sum of skinfolds was lower, although
not signi®cant (P ˆ 0.11) in Depok. Wrist and knee widths
were signi®cantly lower in the subjects in Depok. This is
more noticeable as the subjects in Depok were taller. The
slenderness index height=(sum wrist and knee width) was
greater (P < 0.05) in the Depok subjects, especially in
females (11.65 0.50 and 12.63 0.72 in Depok males
and females compared to 11.23 0.50 and 11.60 0.65 in
Makale males and females respectively). Calculated skeletal mass was 9.7 0.9 kg and 7.2 0.7 kg in Depok and
9.6 1.1 and 7.3 0.7 kg in Makale in males and females
respectively, which is not signi®cantly different between
the sites within the gender groups. After correcting for the
lower body height in the Makale subjects', skeletal mass
was 0.7 0.1 kg (mean s.e.) lower in the Depok females
and 0.3 0.1 kg lower in the Depok males, compared to
Makale females and males respectively.
In Table 2, BF% as measured and as predicted, using
prediction formulas from the literature, is given. Most
prediction formulas did not give valid results. The bias
was generally more pronounced in females and more
469
Relationship between BF% and BMI in Indonesia
S Gurrici et al
470
Table 1 Characteristics of the subjects (mean s.d.)
Depok
Age (y)
Height (cm)
Sitting height (cm)
Relative sitting height
Weight (kg)
Body mass index (kg=m2)
Sum four skinfolds (mm)
Wrist width (mm)
Knee width (mm)
Waist circumference (cm)
Hip circumference (cm)
Body fat (%)
Makale
Males (n ˆ 59)
Females (n ˆ 58)
Males (n ˆ 48)
Females (n ˆ 61)
34.0* 10.1
164.8* 5.6
84.4* 2.7
0.52* 0.01
61.4ns 9.3
22.6* 3.1
53.0ns 21.4
52.1* 2.9
88.5* 5.5
80.5ns 9.0
90.5ns 6.5
24.6ns 7.0
30.6* 10.0
154.3* 5.3
80.0* 2.7
0.52ns 0.01
52.3ns 8.0
21.9* 2.7
69.4ns 19.7
44.3* 3.4
78.3* 5.9
77.0ns 8.1
91.0ns 5.6
35.6ns 5.6
44.1 9.4
161.5 5.9
85.4 4.1
0.53 0.01
62.7 9.4
24.0 3.0
60.3 24.7
52.7 2.4
91.3 6.2
83.5 10.0
91.4 6.2
24.0 6.3
42.5 8.4
149.6 3.6
79.0 2.4
0.53 0.01
51.6 8.3
23.1 3.7
69.4 21.8
47.0 2.7
82.4 6.7
79.3 9.9
90.5 7.7
33.8 6.9
*P < 0.05, ns ˆ not signi®cant between the two study sites within the gender group.
Table 2 Body fat as assessed by deuterium oxide dilution and as predicted using prediction formulas from
the literature (mean s.d.)
Depok
BF% assessed by:
1
Deuterium oxide dilution
Body mass index2
Body mass index3
Skinfold thickness4
Waist circumference5
Makale
Males
Females
Males
Females
24.6* 7.0
18.7* 5.0
26.5 4.1
21.8* 6.7
17.6* 5.5
35.6 5.6
28.0* 4.6
35.0 3.5
31.9* 4.8
32.5ns 4.7
24.0 6.3
22.7* 4.7
28.3* 3.9
25.9* 7.1
22.7* 4.7
33.8 6.9
32.0* 5.4
36.5* 4.8
34.1* 5.3
34.8* 5.4
* P < 0.01 from deuterium oxide dilution; ns ˆ not signi®cant from deuterium oxide dilution
Assuming a 73% hydration of the fat free mass (Pace & Rathbun, 1945)
2
Using the prediction formula: BF% ˆ 1.20*BMI ‡ 0.23*age 7 10.8*sex 7 5.4 (Deurenberg et al, 1991)
3
Using the prediction formula: BF% ˆ 1.30*BMI 7 9.8*sex ‡ 6.4 (Guricci et al, 1998)
4
Using the age and sex speci®c prediction equations of Durnin & Womersley (1974)
5
Using the prediction equations 0.567*waist ‡ 0.101*age 7 31.8 for males and
0.439*waist ‡ 0.221*age 7 9.4 for females (Lean et al, 1996).
1
pronounced in Depok, except for the prediction based on
waist circumference. A prediction equation for BF% based
on BMI developed in another Indonesian population) gave
valid results in the Depok group but overestimated BF% in
the Makale group.
Figure 1 gives the Bland & Altman plots for the
difference between measured (from deuterium oxide dilution) and predicted BF% from BMI against BF% as
assessed by deuterium oxide dilution for the two sites per
sex. The correlation of the residual with BF% is relatively
high.
In Table 3 the partial correlation (after correcting for BF%)
is given between the residuals of predicted BF% using
equation (1) and parameters of body build, including
relative sitting height. The partial correlation is highest
for `slenderness' in both sexes and after correcting for level
of body fatness and slenderness no other parameter
remained signi®cant. Generally the correlations were
higher in females.
Analysis of co-variance was performed to test the effect
of slenderness on the relation between BMI and BF%. After
correcting for differences in BF%, age and sex (explained
Figure 1 Residuals of measured and predicted BF% in relation to measured body fat in the Depok and the Makale population. (Measured minus predicted
BF%: measured body fat percent by deuterium oxide dilution minus predicted body fat percent from the body mass index (formula from Deurenberg et al,
1991). j males; . females.
Relationship between BF% and BMI in Indonesia
S Gurrici et al
Table 3 Partial correlation between residual of predicted body fat and
measures of body build after correcting for level of body fatness
BF%
Correcting for
Height
Relative sitting height
Skeletal mass
Total width
Slender
BF% & slenderness
Males
Females
Males
Females
0.14ns
7 0.21*
7 0.15ns
7 0.36*
0.51*
0.27*
7 0.27*
7 0.33*
7 0.47*
0.55*
0.10ns
7 0.15ns
0.09ns
0.10ns
±
0.09ns
7 0.18ns
0.01ns
0.05ns
±
*P < 0.05; ns ˆ not signi®cant.
BF% ˆ body fat percent; slenderness ˆ height=(sum wrist and knee width).
variance: 0.53) the Malay Indonesians (Depok) had a
1.7 0.3 kg=m2 (mean s.e.) lower BMI than the Chinese
Indonesians (Makale). Additional correction for slenderness decreased the difference in BMI between the two
population groups to 1.1 0.4 kg=m2 with an explained
variance of 0.64. The effect of slenderness was highly
signi®cant (P < 0.0001). Inclusion of relative sitting
height in the model lowered the difference in BMI between
the two groups to 0.9 0.4 kg=m2 (P < 0.02).
Discussion
The subjects participating in this study cannot be regarded
as representative for the total population on Western Java
(Depok) or on southern Sulawesi (Makale). For logistic
reasons university workers and government employees
were chosen for this study, but the subjects were not
specially selected except for race. The population of
Makale (Toraja land) is a special ethnic group with Chinese
ancestry and is as such different from the subjects measured
in Depok which did not include ethnic Chinese. Although
the subjects in Makale were generally older and had a
higher BMI their BF% was not higher and their skinfold
thickness did not differ from the subjects in Depok. As age
is a possible confounder in the relationship between BMI
and BF%, age was adjusted in comparative analyses
between the groups. In both population groups, BF%
predicted with a Caucasian prediction formula (Deurenberg
et al, 1991), using BMI, age and sex as predictive variables,
underestimated the actual BF%, in Depok much more than
in Makale. The results of this study con®rm results of
previous studies in Indonesia on the relationship between
BMI and BF%, showing higher BF% at relatively low
levels of BMI compared to Caucasians (Guricci et al,
1998; KuÈpper et al, 1998). This phenomenon is not limited
to Indonesians. Studies in other Asian population groups
also found relatively high body fatness compared to the
BMI (Wang et al, 1994; Stevens, 1997) but a study in
Beijing Chinese could not con®rm this (Deurenberg et al,
1997). Other authors studied the relationship between BF%
and BMI among different ethnic groups and sometimes
came to different conclusions. Gallagher et al (1996) found
no difference in the relationship between BF% and BMI
between American Blacks and American Caucasians, living
in New York. However, Luke et al (1997) found differences between Black populations living on Jamaica,
Nigeria and in the USA. These authors ascribed the
differences to differences in the level of energy intake in
combination with physical activity. The reason for differences in the relationship between BMI and BF% among
different ethnic groups is, however, unclear and could
theoretically be due to several factors. Relative sitting
height, known to be related to BMI, is slightly lower in
the Depok group. According to Norgan (1994) the lower
relative sitting height of 0.01 in the Depok population
theoretically coincided with a lower BMI of 0.88 kg=m2.
Correcting for this lower BMI would increase predicted
BF% with about 1 percent point. However, in the present
study relative sitting height and BMI were not correlated,
neither in the total group, nor in sex and site de®ned
subgroups, which could be partly due to the high variability
in this parameter.
A higher level of physical activity results in relatively
more muscle tissue at the same body weight and could also
be responsible for the different relationship between BMI
and BF% (Wang et al, 1994). For the present study no
information was available on physical activity of pattern of
the subjects but it seems unlikely that differences in
physical activity alone are responsible for the relatively
large differences. The groups can be assumed to have the
same activity pattern during working hours (of®ce work)
and weather conditions in tropical countries do not invite
strenuous physical activities during leisure time, therefore a
large difference in physical activity due to leisure time
activities seems unlikely.
A third possibility is that subjects differ in body build, in
fact in skeletal mass. Differences in skeletal mass in
subjects with the same height are relatively small (von
DoÈbeln, 1959), and several studies failed to show an impact
of body build on the prediction of BF% from BMI (Fehily
et al, 1990; Rookus et al, 1985; Baecke et al, 1982).
However, it can be argued that a more slender build is
coincided with less muscle and less connective tissue. This
means that subjects with a slender build, having the same
height and weight, will have a lower fat free mass=fat mass
ratio, thus will have higher BF% at the same BMI.
In the present study differences between measured and
predicted BF% among the two studied groups were found.
The group showing the highest underestimation of BF%
from BMI had a more slender body build and an index of
slenderness was signi®cantly correlated with the residual of
measured and predicted body fat (Table 3). Compared to the
Dutch population in which the prediction formula for body
fat from BMI was developed (Deurenberg et al, 1992) both
Indonesian groups were more slender, specially the Depok
group and skeletal mass was lower (results not shown). The
difference in BMI between the two population groups, after
correcting for BF%, age and sex was 1.7 0.3 kg=m2. After
correcting for slenderness and relative sitting height this
difference decreased to 0.9 0.4 and was only borderline
signi®cant (P < 0.02) from zero. The variation in measured
minus predicted BF% is high and depends on the level of
body fatness (Figure 1). This dependency on the level of
body fatness is also found in other studies (Guricci et al,
1998; Deurenberg et al, 1997) and can be explained by the
nature of the prediction formula: excess body weight has a
higher fat=fat free ratio which is not accounted for in the
formula. Thus the formula will underestimate at higher
levels of body fatness. The mean BMI value of the Indonesian population is, however, not different from the BMI
value of the Dutch population in which the prediction
formula was developed (Deurenberg et al, 1991), hence,
this can be no reason for the underestimation of BF% using
this formula. The ratio of skeletal mass and FFM in the two
Indonesian samples was 0.216 0.018 in females and
0.208 0.017 in males (P < 0.001). Assuming a constant
471
Relationship between BF% and BMI in Indonesia
S Gurrici et al
472
fraction of the skeletal mass in the FFM, the lower skeletal
mass (after correcting for height) in the Depok subjects
can be calculated to coincide with a lower total FFM of
100=21.6*0.7 ˆ 3.2 kg in females and 100=20.8*0.3 ˆ
1.5 kg in males. Correcting for this amount of FFM, and
assuming the same body weight, BF% can be calculated to
increase to 30.5% in Depok females and to 22.7% in Depok
males, which is closer but still signi®cantly different to the
actual measured value of 35.6 and 24.6% in females and
males respectively. These calculations show that differences in body build, in fact differences in slenderness do
have an effect on the relationship between BF% and BMI,
but also other factors might be involved. It could be that the
anthropometric techniques used to assess body build are not
accurate enough to pick up small differences in build, and
body build is of course not only determined by the measured
widths.
Although the validity of the deuterium oxide methodology for assessing BF% is generally accepted, biased results
could be obtained when assumptions are violated. The main
assumption in the deuterium oxide method is a constant
hydration fraction of the fat free mass of 0.73 and there
could be differences in the hydration of the fat free mass
among population groups. If the actual hydration factor in
the Depok group were 0.71 instead of 0.73, the assessed
BF% from deuterium oxide dilution would be lowered to
34.5% in females and to 23.0% in males. The main reasons
for an abnormal low or high hydration of the fat free mass
has been shown to be the ratio of extra-cellular water to
intra-cellular water (Wang et al, in preparation). There are,
however, no reasons to assume that the hydration of the fat
free mass is lower, as ratios of determined extra-cellular
water to total body water in the studied groups are not
different and are equal to slightly higher compared to
Caucasian groups (results not shown). To avoid all possible
error sources in BF% determination, the use of a morecompartment models to assess body composition would be
necessary, and radiological techniques (DXA) has to be
used to get better information about body build.
Conclusion
The results of this study con®rm earlier ®ndings that
Indonesians have higher BF% at lower BMI values compared to Caucasians. In addition the study shows that there
are also differences among different Indonesian (ethnic)
subgroups. The differences could be partly explained by
differences in skeletal mass and inherent muscle
mass=connective tissue mass. Other factors, yet unknown,
may also be involved. More studies, using modern body
composition techniques allowing accurate body fat assessment and accurate body build information, are necessary to
get more insight in the underlying factors.
Acknowledgements ÐThe co-operation of the subjects is greatly appreciated. We are indebted to Mr Frans J M Schouten for preparing the study
materials and for the chemical analyses of the plasma samples. The study
was granted in part by DS MediGroup, Milan, Italy.
References
Baecke JAH, Burema J & Deurenberg P (1982): Body fatness, relative
weight and frame size in young adults. Brit. J. Nutr. 48, 1 ± 6.
Bland JM & Altman DG (1986): Statistical methods for assessing agreement between two methods of clinical measurements. Lancet I, 307 ±
310.
Deurenberg P, Weststrate JA & Seidell JC (1991): Body mass index as a
measure of body fatness: age and sex speci®c prediction formulas. Brit.
J. Nutr. 65, 105 ± 114.
Deurenberg P (1992): The assessment of body composition: use and
misuse. Annual report Nestle Foundation, 35 ± 72.
Deurenberg P, Ge K, Hautvast JGAJ & Wang J (1997): BMI as predictor
for body fat: comparison between Chinese and Dutch adult subjects.
Asia Paci®c J. Clin. Nutr. 6, 102 ± 105.
Deurenberg P, Yap M & Van Staveren WA (1998): Body mass index and
percent body fat: a meta analysis among different ethnic groups. Int. J.
Obesity 22, 1164 ± 1171.
DoÈbeln von W (1959): Anthropometric determinants of fat free body
weight. Acta Med. Snad. 165, 37.
Durnin JVGA & Womersley J (1974): Body fat assessed from total body
density and its estimation from skinfold thickness: measurements on
481 men and women aged from 17 to 72 years. Brit. J. Nutr. 32, 77 ± 97.
Fehily AM, Butland BK & Yarnell JWG (1990): Body fatness and frame
size: The Caerphilly study. Eur. J. Clin. Nutr. 44, 107 ± 111.
Forbes GB (1987): Human Body Composition. New York: Springer
Verlag.
Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T & Heyms®eld
SB (1996): How useful is BMI for comparison of body fatness across
age, sex and ethnic groups. Am. J. Epidemiology 143, 228 ± 239.
Guricci S, Hartriyanti Y, Hautvast JGAJ & Deurenberg P (1998): Relationship between body fat and body mass index: differences between
Indonesians and Dutch Caucasians. Eur. J. Clin. Nutr. 52, 779 ± 783.
KuÈpper J, Bartz M, Schultink JWE, Lutiko W & Deurenberg P (1998):
Measurements of body fat in Indonesian adults: comparison between a
three compartment model and widely used methods. Asia Paci®c J.
Clin. Nutr. 7, 49 ± 54.
Lean MEJ, Han TS & Deurenberg P (1996): Predicting body composition
by densitometry from simple anthropometric measurements. Am. J.
Clin. Nutr. 63, 4 ± 14.
Lukaski HC & Johnson PE (1985): A simple, inexpensive method of
determining total body water using a tracer dose of D2O and infrared
absorption of biological ¯uids. Am. J. Clin. Nutr. 41, 363 ± 370.
Lukaski HC (1987): Methods for the assessment of body composition:
traditional and new. Am. J. Clin. Nutr. 46, 437 ± 456.
Luke A, Durazo-Arvizzu R, Rotimi C, Prewitt E, Forrester T, Wilks R,
Ogunbiyi OL, Schoeller DA, McGee D & Cooper RS (1997): Relation
between BMI and body fat in black population samples from Nigeria,
Jamaica and the United States. Am. J. Epidemiology 145, 620 ± 628.
Norgan NG (1994a): Population differences in body composition in
relation to BMI. Eur. J. Clin. Nutr. 48(suppl 3), S10 ± S27.
Norgan NG (1994b): Interpretation of low body mass indices: Australian
Aborigines. Am. J. Physical Anthropol. 94, 229 ± 237.
Pace N & Rathbun EN (1945): Studies on body composition: water and
chemically combined nitrogen content in relation to fat content. J. Biol.
Chem. 158, 685 ± 691.
Rookus MA, Burema J, Deurenberg P & van der Wiel-Wetzels WAM
(1985): The impact of adjustment of a weight-height index (W=H2) for
frame size on the prediction of body fatness. Brit. J. Nutr. 54, 335 ± 342.
SPSS=Windows. V 7.5 Manuals. Chicago, SPSS, (1995).
Stevens NHC (1997): The correlation and prediction of body composition
from dual energy X-ray absorptiometry and anthropometric measurements in Thai women aged 30 to 39 years. Thesis Mahidol University,
Bangkok, Thailand.
Swinburn BA, Craig PL, Daniel R, Dent DPD & Strauss BJG (1996): Body
composition differences between Polynesians and Caucasians assessed
by bioelectrical impedance. Int. J. Obesity 20, 889 ± 894.
Wang J, Thornton JC, Russell M, Burastero S, Heyms®eld SB & Pierson
RN (1994): Asians have lower body mass index (BMI) but higher
percent body fat than do Whites: comparisons of anthropometric
measurements. Am. J. Clin. Nutr. 60, 23 ± 28.
WHO (1995): Physical Status: The use and interpretation of anthropometry. Technical Report Series 854, Geneva: 1995.
WHO (1998): Obesity: Preventing and managing the global epidemic.
Report on a WHO Consultation on Obesity, Geneva, 3 ± 5 June, 1997.
WHO=NUT=NCD=98.1, Geneva.
Womersley J & Durnin JVGA (1977): A comparison of the skinfold
method with extent of overweight and various weight-height relationships in the assessment of obesity. Brit. J. Nutr. 38, 271 ± 284.