Nutrition, Metabolism & Cardiovascular Diseases (2012) 22, 42e49
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/nmcd
Validity of self-reported abdominal obesity in Thai
adults: A comparison of waist circumference,
waist-to-hip ratio and waist-to-stature ratio
L.L.-Y. Lim a,*, Sam-ang Seubsman b, A. Sleigh a, C. Bain c
a
National Centre for Epidemiology and Public Health, Mills Road, Australian National University, Acton ACT 0200, Australia
Sukhothai Thammathirat Open University, Thailand
c
University of Queensland, Australia
b
Received 16 September 2009; received in revised form 7 April 2010; accepted 7 April 2010
KEYWORDS
Obesity;
Abdominal adiposity
index;
Epidemiologic survey;
Sensitivity and
specificity;
Self-report
Abstract Background and aims: Waist circumference (WC), waist-to-hip ratio (WHR) and
waist-to-stature ratio (WSR), being common proxy measures of abdominal obesity, are useful
tools in epidemiologic studies, but little is known about their validity when the indices are
derived from self-reported measurements. We determine and compare the validity of self-reported WC, WHR and WSR in order to identify the optimal index for use in epidemiologic surveys.
Methods and Results: Technician- and self-reported measurements of height, waist and hip
circumference were obtained from 613 Thai adults (mean age 35 years). Regarding technicianreported measurements as reference, diagnostic test properties were derived and performances
of the indices compared using receiver-operator-characteristic curves and the area-under-thecurve (AUC) analyses. There was good agreement between technician- and self-reported
measurements for WC and WSR (concordance correlation coefficients ranged from 0.84 to
0.90) but not for WHR (0.50 in men, 0.45 in women). The sensitivity and specificity of self-reported WC and self-reported WSR as measures of abdominal obesity were superior to those of
self-reported WHR in both sexes. AUCs for WC and WSR were comparable (0.93 and 0.92, respectively, in men; 0.88 and 0.87 in women) and significantly higher than for WHR (0.80 in men; 0.76 in
women; p < 0.0001).
Conclusion: WC and WSR derived from self-reported waist and height measurements are valid
methods for determining abdominal obesity. Self-reported measurements should not be used
to derive the WHR. In Asian populations, WSR may be the optimal index of abdominal obesity
when measurements are derived from self-reports in epidemiologic surveys.
ª 2010 Elsevier B.V. All rights reserved.
Abbreviations: AUC, area-under-the-curve; BMI, body mass index; PPV, positive predictive value; ROC, receiver-operator characteristic;
STOU, Sukhothai Thammathirat Open University; TCS, Thai Cohort Study; WC, waist circumference; WHR, waist-to-hip ratio; WSR, waist-tostature ratio.
* Corresponding author. Tel.: þ61 2 61258346; fax: þ61 2 61250240.
E-mail addresses: [email protected] (L.L.-Y. Lim), [email protected], [email protected] (S.-ang Seubsman),
[email protected] (A. Sleigh), [email protected] (C. Bain).
0939-4753/$ - see front matter ª 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.numecd.2010.04.003
Validity of self-reported abdominal obesity in Thai adults
With the alarming rise in obesity worldwide, practical and
economic means of measuring excess body fat have
become increasingly important in routine clinical practice and in epidemiologic studies [1]. Body mass index
(BMI), which reflects overall body fat, is the most
commonly used measure of obesity as it has strong links
to risks of adverse health outcomes [1]. Mounting
evidence suggests that excess deposition of fat in the
abdominal region is more strongly associated with
subsequent adverse cardio-metabolic outcomes than
overall body fat [2]. But the optimal approach for routine
measurement of abdominal obesity remains controversial
[3,4].
Waist circumference (WC) and the ratio of waist to hip
circumference (waist-hip ratio, WHR) are two popular
proxy measures of abdominal obesity [5]. More recently,
the ratio of waist to height (waist-stature ratio, WSR) has
been proposed as an alternative measure of abdominal
obesity [6]. All three indices, WC, WHR and WSR, are
strongly associated with cardio-metabolic risks and
subsequent adverse outcomes [7e9], but no one index is
clearly superior as studies comparing the performances of
these indices have conflicting findings. WHR outperformed WC and WSR as a measure of risk of myocardial infarction in the INTERHEART study across 52
countries [10], but a meta-regression of 15 prospective
studies involving 258,114 participants found little difference between WHR and WC [11], while WC and WSR
predicted cardiovascular risk better than did WHR in the
European DETECT study of over 5500 adults [12]. In one
Chinese study of 15,236 adults, WC and WHR were equally
able to identify diabetes mellitus [13], but in another of
75,788 Chinese adults WC outperformed WHR [14].
Several recent studies have reported some superiority of
the WSR in Asian populations [8,14e20].
The rise in recent decades of mortality from cardiovascular disease in Thailand has been attributed to the
increased prevalence of obesity among Thai adults [21].
This study was motivated by the desire to assess
abdominal obesity in the Thai Cohort Study (TCS), a large
prospective study of health risk transitions in over
80,000 men and women from the Sukhothai Thammathirat Open University (STOU) in Thailand [22]. We
needed to determine which of WC, WHR or WSR would be
optimal for use in a self-administered questionnaire as
information in the TCS is sought through mail-out
surveys. However there is a dearth of information about
the validity of WC, WHR and WSR when the underlying
anthropometric measurements are self-reported. We
found just ten studies examining the accuracy of selfreported WC or self-reported WHR [23e32], and none of
self-reported WSR. Further these studies were all conducted in the US and Europe and may not be applicable
to an Asian population.
We collected technician- and self-reported measurements
of height, waist and hip circumference on a representative
sample of 613 adult Thai subjects. Our purpose is to determine
the validity of WC, WHR and WSR when these indices are
derived from self-reported measurements and, by comparing
their performances, to determine the optimal abdominal
obesity index for routine use in epidemiologic studies.
43
Methods
Study population and data collection
Study subjects were a convenience sample of 613 STOU
students who attended the main campus in Bangkok during
May 2008 for special courses on professional ethics before
graduation. Participation was voluntary. Participants were
given written Ministry of Public Health instructions for
measuring waist and hip circumference along with
a diagram and photo illustration.1 They were asked to self
report their weight and height and then use a graduated
tape to measure their waist and hip circumferences. The
waist was measured at the level of the umbilicus and hip
measurement defined as the widest part of the hip. The
participants wore light clothing and these were not
removed. After reporting their results on a form they were
invited to allow the same measurements to be made by
a trained technician and all agreed. Weight and height
measurements were made by one technician and waist and
hip measurements by two technicians working together.
Height was measured without shoes with a SECA stadiometer. Height, waist and hip were measured to the
nearest centimetre.
Waist, hip, height and the derived WHR and WSR
variables
WHR was calculated as WC divided by hip circumference;
WSR as WC divided by height. Technician-reported indices
are calculated from technician-reported measurements; and
self-reported indices from self-reported measurements.
Discrepancies were calculated as self-reported minus technician-reported measurements. Discrepancies in WC or hip
circumference of over 5 cm, height discrepancies of over
10 cm, and percentage discrepancies in WHR or WSR over 5%
were checked for data entry errors.
Statistical methods
Agreement between self- and technician-reported
measurements was assessed using the concordance correlation coefficient [33]. The coefficient evaluates the degree
to which pairs of measurements fall on the 45 line through
the origin and is an appropriate measure of agreement
between two continuous variables. The Pearson correlation
coefficient is reported here for comparison with existing
studies because it is frequently, although inappropriately,
used as a measure of agreement [34].
By regarding the self-reported indices as diagnostic tests
and the technician-derived indices as reference (gold
standard), validity of the self-reported indices was assessed
1
English translation of the instructions for waist and hip
measurements were “Measure in a standing position. Breathe
gently. Place the tape measure around your waist at the level of
your navel touching your body but not too tight. Place the tape
measure parallel to the waist line of your pants or skirt.” and
“Follow the same process used in measuring the waist except you
place the tape measure around the widest part of the hip area just
touching your body as described above.”
44
by the standard diagnostic test criteria of sensitivity,
specificity, positive and negative predictive values [35].
Sensitivity is the proportion of obese individuals who are
correctly identified as obese from the self-reported obesity
index; specificity the proportion of non-obese individuals
who are correctly identified as non-obese from the selfreported index. Positive predictive value (PPV) is the
probability of truly being obese when the result from the
self-reported obesity index is positive; negative predictive
value the probability that the self-reported index will
correctly identify a non-obese person. Youden’s Index,
defined as (sensitivity þ specificity 1), combines sensitivity and specificity into a single number and gives a simple
measure of the overall performance of a test which is
particularly useful for comparing test performance in situations where sensitivity and specificity are equally important [35]. Values are close to 1 when both sensitivity and
specificity are high.
These methods rely on a (binary) gold standard to
dichotomise subjects into normal or abnormal, where
abnormal in our context is excess abdominal obesity. The
technician-reported WC, WHR and WSR are gold standards
on a continuous-scale which can be converted to binary
gold standards using cut-off points. We use cut-off points
for WC, WHR and WSR derived from a study of 5305 Thai
adults [36]. The seven cardiovascular risk factors2 examined in the study reported optimal cut-off points for WC
ranging from 80 to 85 cm in men and 81 to 85 cm in women,
for WHR from 0.89 to 0.91 in men and 0.85 to 0.88 in women
and for WSR from 0.49 to 0.52 for men and 0.53 to 0.56 in
women [36]. The smallest and the largest cut-off points are
used here. The largest cut-off points were associated with
Type II diabetes and the smallest with dyslipidemia in both
sexes for all three indices.
Receiver-operator-characteristic (ROC) curves and areaunder-the-curve (AUC) analyses were used to compare the
overall performances of the self-reported indices. This
approach was necessary because of the absence of
universally accepted cut-off points for the three indices
since comparing performances of the indices using arbitrary choices of cut-off points could be misleading. For
example, with proposed cut-off points for WC in women
ranging from 80 cm to 90 cm [37], the sensitivity (or
specificity) of self-reported WC could be artifactually
varied with choice of cut-off point. Consequently
comparison of the performances of the self-reported
indices could be manipulated through judicious choices of
cut-off points for each index. The ROC-AUC method used
here overcomes this potential bias by giving equal
consideration to all possible cut-off points for each index
when computing its AUC.
Derivation of the ROC curves was based on a method by
Obuchowski [38] for continuous-scale gold standards. An
ROC curve is a trace of sensitivity against “1-specificity”.
Instead of relying on a single cut-off point, the ROC-AUC
method here creates the ROC trace over the entire range of
potential cut-off points. For each of WC, WHR and WSR, 50
values between the minimum and maximum of the
2
Hypertension, Type II diabetes, high total cholesterol, low highdensity lipoprotein cholesterol, high low-density lipoprotein
cholesterol, high triglyceride and dyslipidemia.
L.L.-Y. Lim et al.
observed values of the index were defined, and sensitivity
and “1-specificity” computed using these values as cut-off
points. For all three indices, sensitivity and “1-specificity”
increase with increasing magnitude of the cut-off point. For
example, for WC in men, 50 values between 71.6 cm
(minimum) and 101 cm (maximum) were defined, and
sensitivity and “1-specificity” computed using the 50 values
in turn as cut-off point. For clarity in graphing the ROC
curves, the x-axis (“1-specificity”) was partitioned into
seven sectors (0e.01, .01e.05, .05e.1, .1e.2, .2e.3,
.3e.4, .4e1), and the sensitivities and “1-specificities”
averaged within each sector for plotting. Non-parametric
AUC estimates were derived using the trapezoidal method
[39] and variances estimated by bootstrapping [40]. All
analyses and computations were performed using Stata
Version 9.
Results
The sample of 613 Thai adults ranged in age from 21 to 62
years, with a mean of 35 years. Mean BMI was 24.6 kg/m2
(SD 3.8) in men and 22.3 kg/m2 (SD 3.7) in women.
Agreement between technician- and self-reported
measurements
Except for hip circumference in women, all other selfreported values were statistically significantly different
from technician-reported values (Table 1). Both men and
women over-estimated their height and under-estimated
waist circumference (WC), while men also under-estimated
their hip circumference. Correlations between technicianand self-reported values were lower for WHR than for WC
and WSR, for both men and women (Table 1).
Self-reported values of the three abdominal obesity
indices were under-estimated by both men and women,
except for WHR which was over-estimated by men (Table
2). When examined within BMI tertiles, the percentage
differences between technician- and self-reported values
increased with increasing body size (Table 2). The correlations between technician and self-reported values were
also higher as body size increased, indicating greater
systematic bias in self-reported values with increasing body
size.
Validity
Tables 3 and 4 show, for men and women respectively,
diagnostic test values for the three abdominal obesity
indices using two sets of cut-off points to define excess
body fat. In men, cut-off points used were 80 cm and 85 cm
for WC, 0.89 and 0.91 for WHR and 0.49 and 0.52 for WSR
(Table 3). Sensitivity was similar for the three indices and
specificity was similar for WC and WSR but lower for WHR
than for WC and WSR (Table 3). In women, cut-off points
used were 81 cm and 85 cm for WC, 0.85 and 0.88 for WHR
and 0.53 and 0.56 for WSR (Table 4). Both sensitivity and
specificity were similar for WC and WSR, and higher than for
WHR (Table 4). Youden’s Index was consistently lowest for
WHR (Tables 3 and 4).
Validity of self-reported abdominal obesity in Thai adults
45
Table 1 Technician-reported and self- reported measurements, discrepancies and correlations for waist circumference (WC),
hip circumference, height and the derived indices waist-hip ratio (WHR) and waist-stature ratio (WSR) separately for men
(n Z 313) and women (n Z 300).
Technician-reported
Self-reported
Discrepancya
Correlation coefficients
Mean
Mean
(SD)
Mean
(95%CI)
Concordanceb
Pearson
(SD)
Waist circumference (WC), cm
Men
84.3
(10.0)
Women
74.4
(8.8)
83.2
73.2
(8.8)
(8.3)
1.09
1.17
(1.54, 0.64)
(1.69, 0.65)
0.90
0.85
0.92
0.86
Hip circumference, cm
Men
96.2
Women
91.4
(7.0)
(7.5)
93.7
91.6
(7.8)
(8.8)
2.45
0.19
(2.90, 2.01)
(0.62, 1.01)
0.81
0.62
0.86
0.62
Height, cm
Men
Women
(6.0)
(5.2)
168.7
157.6
(6.4)
(5.3)
1.71
1.32
(1.46, 1.95)
(1.1, 1.53)
0.87
0.91
0.94
0.94
167.0
156.3
Waist-Hip Ratio (WHR)
Men
0.87
Women
0.81
(0.06)
(0.05)
0.89
0.80
(0.06)
(0.06)
0.013
0.013
(0.007, 0.019)
(0.019, 0.006)
0.50
0.45
0.51
0.47
Waist-Stature Ratio (WSR)
Men
0.50
Women
0.48
(0.06)
(0.06)
0.49
0.46
(0.05)
(0.05)
0.012
0.011
(0.014, 0.009)
(0.015, 0.008)
0.88
0.84
0.91
0.85
a
b
Discrepancy Z Self-reported values e Technician-reported values.
Concordance correlation coefficient Z Bias correction factor Pearson correlation coefficient (Lin 1986).
PPV was also lower for WHR. The PPVs corresponding to
the smallest and largest cut-off points for WHR were 56.8%
to 63.3%, respectively, in men and 34.5% to 52.5%,
respectively, in women (Tables 3 and 4). PPVs for WC and
WSR were similar, and higher in men than in women.
confidence interval 0.91e0.95) and 0.92 (0.90e0.95) in
men, and 0.88 (0.86e0.91) and 0.87 (0.84e0.90) in women.
The AUCs for WHR were 0.80 (0.76e0.84) in men and 0.76
(0.71e0.80) in women, which were significantly lower
(p < 0.0001) than the AUCs of the other two indices.
Relative performance of the indices
Discussion
The ROC curves of WC and WSR were similar for both sexes
and consistently above the ROC curve of WHR (Fig. 1). The
AUCs for WC and WSR were, respectively, 0.93 (95%
We sought to determine and compare the validity of WC,
WHR and WSR when these indices are derived from selfreported measurements to identify the best index for use in
Table 2 Percentage differences between self-reported and technician-reported values for waist circumference (WC),
waist-hip ratio (WHR) and waist-stature ratio (WSR) overall and within body size tertiles.
Men
% difference,a (95% CI)
n
Waist Circumference
(WC)
Waist-Hip Ratio
(WHR)
Waist-Stature Ratio
(WSR)
313
1.1 (1.6, 0.5)
1.7 (1.0, 2.4)
2.0 (2.6, 1.5)
105
104
104
1.0 (0.1, 1.9)
1.5 (2.4, 0.6)
2.6 (3.4, 1.8)
4.4 (3.0, 5.8)
1.5 (0.4, 2.6)
0.8 (1.8, 0.1)
0.2 (0.6, 1.1)
2.7 (3.6, 1.7)
3.7 (4.6, 2.9)
300
1.3 (2.0, 0.6)
1.4 (2.2, 0.6)
2.1 (2.8, 1.4)
100
100
100
0.0 (1.2, 1.3)
1.5 (2.6, 0.3)
2.5 (3.7, 1.4)
1.0 (2.7, 0.6)
1.3 (2.4, 0.2)
1.8 (3.2, 0.5)
0.8 (2.1, 0.5)
2.2 (3.3, 1.1)
3.4 (4.6, 2.2)
b
By BMI tertiles
Bottom third
Middle third
Top third (Highest BMI)
Women
% difference,a (95% CI)
By BMI tertilesb
Bottom third
Middle third
Top third (Highest BMI)
a
b
Percentage difference Z (Self-reported value Technician-reported value)/Technician-reported value 100.
Tertiles calculated from BMI values derived from technician measurements of weight and height.
46
L.L.-Y. Lim et al.
Table 3 Test values for self-reported waist circumference (WC), self-reported waist-hip ratio (WHR) and self-reported
waist-stature ratio (WSR) using two sets of cut-off pointsa to define excess body fat in men (n Z 313).
Sensitivity
Specificity
Positive predictive
value (PPV)
Negative predictive
value (NPV)
Youden’s
Indexb
Prevalence, %
Technician
Self
Waist circumference, WC
Smallest cut-off (80 cm)
Largest cut-off (85 cm)
0.86
0.70
0.86
0.95
0.89
0.90
0.82
0.82
0.72
0.64
57
40
55
31
Waist-hip ratio, WHR
Smallest cut-off (0.89)
Largest cut-off (0.91)
0.82
0.63
0.73
0.82
0.63
0.57
0.87
0.85
0.55
0.45
36
27
47
30
Waist-stature ratio, WSR
Smallest cut-off (0.49)
Largest cut-off (0.52)
0.80
0.68
0.89
0.98
0.90
0.95
0.80
0.84
0.69
0.66
54
36
49
26
a
b
The smallest and the largest of seven cut-off points reported in Aekplakorn (2006).
Youden’s Index Z (Sensitivity þ Specificity 1).
large epidemiologic surveys. Self-reported WHR performed
significantly worse than the other two indices. Self-reported WC and self-reported WSR both demonstrated good
validity and were comparable on all diagnostic testing
performance criteria.
We therefore considered two other criteria to choose
between WC and WSR.
(1) Predictive ability. For the Thai population of the TCS,
WSR may be preferable to WC as several studies on
Asian populations have demonstrated superiority of
WSR over other abdominal obesity indices [8,14e20]. In
Western populations, although a few studies have
shown superiority of WSR [9,12,41], the conflicting
findings over an optimal abdominal obesity index have
led some authors to conclude that the differences in
predictive ability between the indices are clinically
irrelevant [7,9].
(2) Simplicity of cut-off points. With WSR, a universal cutoff point of 0.5 has been recommended for all
subgroups e men and women, children and adults, as
well as Asians and Caucasians [6]. This is an advantage
over WC for which universally applicable cut-off points
cannot be developed because populations differ in the
level of risk associated with a particular waist circumference [1,42]. For example, at least seven pairs of cutoff points have been suggested for WC for men and
women in different ethnic populations [37].
With these considerations, we believe WSR is the most
appropriate index of abdominal obesity for use in epidemiologic surveys, particularly in Asian populations.
Previous work on self-reported abdominal obesity
indices
We are not aware of any other study of self-reported WSR,
but have identified a few of self-reported WC or selfreported WHR. These studies focused primarily on agreement between self- and technician-reported measurements
rather than on validity of the self-reported indices.
Nevertheless there were some interesting observations
between our findings and results from these studies.
Table 4 Test values for self-reported waist circumference (WC), self-reported waist-hip ratio (WHR) and self-reported waiststature ratio (WSR) using two sets of cut-off pointsa to define excess body fat in women (n Z 300).
Sensitivity
Specificity
Positive predictive
value (PPV)
Negative predictive
value (NPV)
Youden’s
Indexb
Prevalence, %
Technician
Self
Waist circumference, WC
Smallest cut-off (81 cm)
Largest cut-off (85 cm)
0.63
0.60
0.96
0.99
0.81
0.88
0.91
0.95
0.59
0.59
21
12
16
8
Waist-hip ratio, WHR
Smallest cut-off (0.85)
Largest cut-off (0.88)
0.44
0.40
0.88
0.93
0.53
0.35
0.84
0.95
0.32
0.33
23
8
19
10
Waist-stature ratio, WSR
Smallest cut-off (0.53)
Largest cut-off (0.56)
0.67
0.60
0.97
0.99
0.84
0.79
0.93
0.97
0.57
0.66
15
8
11
6
a
b
The smallest and the largest of seven cut-off points reported in Aekplakorn (2006).
Youden’s Index Z (Sensitivity þ Specificity 1).
Validity of self-reported abdominal obesity in Thai adults
47
Women
.6
.5
.4
0
.1
.2
.3
Sensitivity
.7
.8
.9
1
Men
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
1-Specificity
Figure 1 Receiver-operator characteristic (ROC) curves for men and women for waist circumference, WC (dCd), waist-hip
ratio, WHR ($$e-$$e) and waist-stature ratio, WSR ($$$$:$$$$).
First, the majority of studies had cautioned against the
use of self-reported WHR as we have. Four were relatively
small studies, with less than 200 subjects each
[25,26,28,29]. The British study of 4492 participants found
unacceptably high misclassification rates for self-reported
WHR (66.9% and 69.8% of men and women, respectively)
leading the authors to conclude that “using WHR derived
from self-reported measurements may be more likely to
obscure true associations than using either waist or hip
measurement alone” [32].
One study [30] concluded WHR would be a useful tool in
epidemiologic studies. This study, comprising 111
employees from a medical institution, found high sensitivity
(77% in men; 81% in women) and specificity (93% in both) for
self-reported WHR. The high levels of accuracy may be
because participants and technicians were instructed to
take each measurement twice, plus a third if there was
excessive difference between the first two, and to report
the average measurement.
Second, as in our study, all previous studies found the
correlations between self- and technician-reported
measurements for WHR to be lower than the correlations
for WC [25,29e32]. In men, WHR correlations ranged from
0.44 to 0.78 compared with 0.80 to 0.95 for WC; in women
WHR correlations ranged from 0.62 to 0.83 compared with
0.83 to 0.99 for WC. Although high correlations do not imply
that the self-reported measurements are valid, low correlations do indicate that the self-reported measurements
are not acceptable substitutes for technician-reported
measurements.
Third, the trend of decreasing accuracy with increasing
body size in self-reported WC and self-reported WHR were
reported in three studies [26,28,30e32]. This trend, also
observed in the present study, has been consistently
documented in validation studies of self-reported BMI, for
example [43e45], and in our parallel validation study of
self-reported BMI [46].
In summary, key findings from this study are consistent
with conclusions of earlier studies, that (1) WHR derived
from self-reported measurements is not valid, (2) selfreported WC performs better than self-reported WHR, and
(3) the validity of self-reported obesity indices decreases
with increasing body size.
Comparing validity of self-reported WSR to validity
of self-reported BMI
BMI derived from self-reported weight and height is widely
used in epidemiologic studies as a measure of obesity. In nonAsian populations, the sensitivity and specificity of selfreported BMI range from 68% to 86% and 96% to 99%, respectively, in men and from 69% to 89% and 97% to 99% in women
[43e45,47]. Our parallel validation study of self-reported BMI
in Thailand using the Thai BMI cut-off point of 25 kg/m2 for
obesity found sensitivity and specificity of 74.2% and 97.3%,
respectively, in men and 71.9% and 100% in women [46].
The sensitivity and specificity of self-reported BMI are
similar to the sensitivity and specificity of self-reported
WSR using a cut-off point of 0.5 (76% and 97.6%, respectively, in men; 69% and 97.3% in women). The comparability
in validity of self-reported WSR with validity of the widely
used self-reported BMI supports the case for use of WSR
derived from self-reported measurements in epidemiologic
studies.
Positive predictive values (PPVs)
PPV is arguably the most important test property for the
purpose of determining obesity from self-reported data
48
because it reflects the probability that a positive test result
correctly identifies obesity. In this study, PPVs for WC and
WSR were generally high, close to 80% or greater, but PPVs
for WHR were low, ranging from 35% and 63%. The low PPVs
for WHR mean that, in our study population, individuals
whose self-reported WHR indicate excess abdominal fat
have low probabilities of truly being obese.
Generalisability
Our sample was designed to represent Thais who were
educated enough to self-report measurements. They were
drawn from the STOU student population who were all Thai
citizens (72% have Chinese ethnic links), enrolled for
distance-learning university degrees and almost all had
completed high school and the central 90% of the age
distribution was between 25 and 46 years. The median
income of the TCS sample was the equivalent of US$2550
and the income distribution was quite similar to that of the
Thai population in general [22]. Overall they were sufficiently homogeneous in relation to factors that could affect
self-reporting and sufficiently representative to allow
generalisation to educated adult Thais.
Limitations
Care should be taken when generalising to non-Asian populations because of different body frames and possibly
different culturally desirable body images which may
influence the bias in self-reported body size measurements.
Confirmation of our findings in other settings is needed.
Another limitation was waist measurement only at the
level of the umbilicus which may restrict the clinical utility
of our findings. This restriction may not however be
significant as authors of the Oxford study had concluded
that acceptable results were obtained in their study despite
no instructions being given to participants on how to make
waist and hip measurements [32].
We did not examine or control for variability in the technician-reported measurements. As Panoulos [48] found
significant inter-technician variability in waist measurement,
more reliable technician measurements using, for example,
replications could affect our findings. However, we expect
minimal differences from replicated measurements in the
present study because in every instance of more than 50 cases
where participants had requested the technicians repeat
their measurements, the repeated results were the same.
Accuracy of self-reported waist measurements could be
improved by requesting respondents to have another person
measure them [28] or by instructing respondents to take at
least two separate measurements and to report the average
[30,31].
Conclusions
WC and WSR derived from self-reported waist and height
measurements are valid methods for measuring abdominal
obesity in epidemiologic studies. Use of WHR should be
restricted to measurement by trained technicians because
of poor validity when derived from self-reported measurements. In Asian populations, WSR may be the optimal
L.L.-Y. Lim et al.
measure of abdominal obesity when measurements are
derived from self-reports in epidemiologic surveys.
Funding
This study was supported by the International Collaborative
Grants Scheme with joint grants from the Wellcome Trust
UK (GR0587MA) and the Australian NHMRC (268055).
Appendix. Supplementary data
Supplementary data associated with this article can be
found in the online version, at doi:10.1016/j.numecd.2010.
04.003.
References
[1] WHO Consultation. Obesity: preventing and managing the
global epidemic (1999). Geneva: World Health Organization;
2000.
[2] Despres JP, Lemieux I. Abdominal obesity and the metabolic
syndrome. Nature 2006;444:881e7.
[3] Litwin SE. Which measures of obesity best predict cardiovascular risk? J Am Coll Cardiol 2008;52(8):616e9.
[4] Schneider HJ, Wittchen HU, Wallaschofski H. Obesity and risk
of death. N Engl J Med 2009;360(10):1043e4.
[5] Snijder MB, van Dam RM, Visser M, Seidell JC. What aspects of
body fat are particularly hazardous and how do we measure
them? Int J Epidemiol 2006;35(1):83e92.
[6] Ashwell M, Gibson S. Waist to height ratio is a simple and
effective obesity screening tool for cardiovascular risk
factors: analysis of data from the British National diet and
nutrition survey of adults aged 19e64 years. Obesity Facts
2009;2(2):97e103.
[7] Gelber RP, Gaziano JM, Orav EJ, Manson JE, Buring JE,
Kurth T. Measures of obesity and cardiovascular risk among
men and women. J Am Coll Cardiol 2008;52(8):605e15.
[8] Lee CM, Huxley RR, Wildman RP, Woodward M. Indices of
abdominal obesity are better discriminators of cardiovascular
risk factors than BMI: a meta-analysis. J Clin Epidemiol 2008;
61(7):646e53.
[9] Page JH, Rexrode KM, Hu F, Albert CM, Chae CU, Manson JE.
Waist-height ratio as a predictor of coronary heart disease
among women. Epidemiology 2009;20(3):361e6.
[10] Yusuf S, Hawken S, Ounpuu S, Bautista L, Franzosi MG,
Commerford P, et al. Obesity and the risk of myocardial
infarction in 27,000 participants from 52 countries: a casecontrol study. Lancet 2005;366(9497):1640e9.
[11] de Koning L, Merchant AT, Pogue J, Anand SS. Waist circumference and waist-to-hip ratio as predictors of cardiovascular
events: meta-regression analysis of prospective studies. Eur
Heart J 2007;28(7):850e6.
[12] Schneider HJ, Glaesmer H, Klotsche J, Bohler S, Lehnert H,
Zeiher AM, et al. Accuracy of anthropometric indicators of
obesity to predict cardiovascular risk. J Clin Endocrinol Metab
2007;92(2):589e94.
[13] Hu D, Xie J, Fu P, Zhou J, Yu D, Whelton PK, et al. Central
rather than overall obesity is related to diabetes in the
Chinese population: the InterASIA study. Obesity 2007;15(11):
2809e16.
[14] Wang F, Wu S, Song Y, Tang X, Marshall R, Liang M, et al. Waist
circumference, body mass index and waist to hip ratio for
prediction of the metabolic syndrome in Chinese. Nutr Metab
Cardiovasc Dis; 2009 Jan 31.
Validity of self-reported abdominal obesity in Thai adults
[15] Aekplakorn W, Pakpeankitwatana V, Lee CMY, Woodward M,
Barzi F, Yamwong S, et al. Abdominal obesity and coronary
heart disease in Thai Men. Obesity 2007;15(4):1036e42.
[16] He Y, Zhai F, Ma G, Feskens EJM, Zhang J, Fu P, et al.
Abdominal obesity and the prevalence of diabetes and intermediate hyperglycaemia in Chinese adults. Public Health
Nutrition; 2008:1e7.
[17] Ho SY, Lam TH, Janus ED. Waist to stature ratio is more
strongly associated with cardiovascular risk factors than other
simple anthropometric indices. Ann Epidemiol 2003 Nov;
13(10):683e91.
[18] Hsieh SD, Muto T. Metabolic syndrome in Japanese men and
women with special reference to the anthropometric criteria
for the assessment of obesity: proposal to use the waist-toheight ratio. Prev Med 2006 Feb;42(2):135e9.
[19] Wu H-Y, Chen L-L, Zheng J, Liao Y-F, Zhou M. Simple anthropometric indices in relation to cardiovascular risk factors in
Chinese type 2 diabetic patients. Chin J Physiol 2007;50(3):
135e42.
[20] Tseng C-H. Waist-to-height ratio and coronary artery disease
in Taiwanese Type 2 diabetic patients. Obesity 2008;16(12):
2754e9.
[21] Bureau of Health Policy and Planning, Office of Permanent
Secretary, Ministry of Public Health. Thailand health profile
1997e1998. Bangkok: Printing Press, Express Transportation
Organization; 2000.
[22] Sleigh AC, Seubsman SA, Bain C. Cohort profile: the Thai
Cohort of 87,134 Open University students. Int J Epidemiol
2008 Apr;37(2):266e72.
[23] Bigaard J, Spanggaard I, Thomsen BL, Overvad K,
Tjonneland A. Self-reported and technician-measured waist
circumferences differ in middle-aged men and women. J Nutr
2005 Sep;135(9):2263e70.
[24] Dekkers JC, van Wier MF, Hendriksen IJ, Twisk JW, van
Mechelen W. Accuracy of self-reported body weight, height
and waist circumference in a Dutch overweight working population. BMC Med Res Methodol 2008;8:69.
[25] Freudenheim JL, Darrow SL. Accuracy of self-measurement of
body fat distribution by waist, hip, and thigh circumferences.
Nutr Cancer 1991;15(3e4):179e86.
[26] Hall TR, Young TB. A validation study of body fat distribution
as determined by self-measurement of waist and hip circumference. Int J Obes 1989;13(6):801e7.
[27] Han TS, Lean ME. Self-reported waist circumference
compared with the ‘Waist Watcher’ tape-measure to identify
individuals at increased health risk through intra-abdominal
fat accumulation. Br J Nutr 1998 Jul;80(1):81e8.
[28] Kushi LH, Kaye SA, Folsom AR, Soler JT, Prineas RJ. Accuracy
and reliability of self-measurement of body girths. Am J Epidemiol 1988;128(4):740e8.
[29] Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB,
Willett WC. Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990 Nov;1(6):466e73.
[30] Roberts CA, Wilder LB, Jackson RT, Moy TF, Becker DM.
Accuracy of self-measurement of waist and hip circumference
in men and women. J Am Diet Assoc 1997 May;97(5):534e6.
[31] Weaver TW, Kushi LH, McGovern PG, Potter JD, Rich SS,
King RA, et al. Validation study of self-reported measures of
fat distribution. Int J Obes Relat Metab Disord 1996 Jul;20(7):
644e50.
49
[32] Spencer EA, Roddam AW, Key TJ. Accuracy of self-reported
waist and hip measurements in 4492 EPIC-Oxford participants.
Public Health Nutr 2004;7(6):723e7.
[33] Lin LI. A concordance correlation coefficient to evaluate
reproducibility. Biometrics 1989 Mar;45(1):255e68.
[34] Bland JM, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measurement.
Lancet 1986 Feb 8;1(8476):307e10.
[35] Bewick V, Cheek L, Ball J. Statistics review 13: receiver
operating characteristic curves. Crit Care 2004 Dec;8(6):
508e12.
[36] Aekplakorn W, Kosulwat V, Suriyawongpaisal P. Obesity indices
and cardiovascular risk factors in Thai adults. Int J Obes 2006;
30(12):1782e90.
[37] Alberti KG, Zimmet P, Shaw J. The metabolic syndrome e
a new worldwide definition. Lancet 2005 Sep 24e30;
366(9491):1059e62.
[38] Obuchowski NA. An ROC-type measure of diagnostic accuracy
when the gold standard is continuous-scale. Stat Med 2006 Feb
15;25(3):481e93.
[39] DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the
areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988
Sep;44(3):837e45.
[40] Efron B, Tibshirani RJ. An introduction to the bootstrap.
London: Chapman & Hall; 1993.
[41] Lamacchia O, Pinnelli S, Camarchio D, Fariello S, Gesualdo L,
Stallone G, et al. Waist-to-height ratio is the best anthropometric
index in association with adverse cardiorenal outcomes in Type 2
diabetes mellitus patients. Am J Nephrol 2009;29(6):615e9.
[42] Misra A, Wasir JS, Vikram NK. Waist circumference criteria for
the diagnosis of abdominal obesity are not applicable
uniformly to all populations and ethnic groups. Nutrition 2005
Sep;21(9):969e76.
[43] Nyholm M, Gullberg B, Merlo J, Lundqvist-Persson C, Rastam L,
Lindblad U. The validity of obesity based on self-reported
weight and height: implications for population studies.
Obesity (Silver Spring) 2007 Jan;15(1):197e208.
[44] Shields M, Gorber SC, Tremblay MS. Estimates of obesity based
on self-report versus direct measures. Health Rep 2008 Jun;
19(2):61e76.
[45] Taylor AW, Dal Grande E, Gill TK, Chittleborough CR,
Wilson DH, Adams RJ, et al. How valid are self-reported height
and weight? A comparison between CATI self-report and clinic
measurements using a large cohort study. Aust N Z J Public
Health 2006 Jun;30(3):238e46.
[46] Lim LL-Y, Seubsman SA, Sleigh AC. Validity of self-reported
weight, height and body mass index among adult open
university students in Thailand: Implications for population
studies of obesity in developing countries. Population Health
Metrics, accepted for publication.
[47] Fonseca Mde J, Faerstein E, Chor D, Lopes CS. Validity of selfreported weight and height and the body mass index within
the “Pro-saude” study. Rev Saude Publica 2004 Jun;38(3):
392e8.
[48] Panoulas VF, Ahmad N, Fazal AA, Kassamali RH, Nightingale P,
Kitas GD, et al. The inter-operator variability in measuring
waist circumference and its potential impact on the diagnosis
of the metabolic syndrome. Postgrad Med J 2008;84(993):
344e7.