Volume 11 • Supplement 1 • 2008
VA L U E
I N
H E A LT H
The Association of Body Mass Index with Health-Related
Quality of Life: An Exploratory Study in a Multiethnic
Asian Population
Hwee-Lin Wee, PhD,1 Yin-Bun Cheung, PhD,2 Wai-Chiong Loke, FCFP (Singapore),3
Chee-Beng Tan, FCFP (Singapore),3 Mun-Hong Chow, FCFP (Singapore),3 Shu-Chuen Li, PhD,4,5
Kok-Yong Fong, FRCP (Edin),1,6 David Feeny, PhD,7,8 David Machin, PhD,9,10 Nan Luo, PhD,11
Julian Thumboo, FRCP (Edin)1,6
1
Department of Rheumatology and Immunology, Singapore General Hospital, Singapore; 2MRC Tropical Epidemiology Group, London School
of Hygiene and Tropical Medicine, London, UK; 3SingHealth Polyclinics, Singapore; 4Department of Pharmacy, National University of Singapore,
Singapore; 5Discipline of Pharmacy & Experimental Pharmacology, School of Biomedical Sciences, University of Newcastle, Callaghan, NSW,
Australia; 6Department of Medicine,Yong Loo Lin School of Medicine, National University of Singapore, Singapore; 7Kaiser Permanente
North-west Center for Health Research, Portland, OR, USA; 8Health Utilities Incorporated, Dundas, ON, Canada; 9National Cancer Center
Singapore, Singapore; 10Clinical Trials & Epidemiology Research Unit, Singapore; 11Centre for Health Services Research,Yong Loo Lin School
of Medicine, National University of Singapore, Singapore
A B S T R AC T
Objectives: To evaluate the association between body mass
index (BMI) and health-related quality of life (HRQoL) in a
multiethnic Asian population in Singapore, and to explore if
the World Health Organization (WHO) recommendation of
alternative BMI cutoffs for Asians could be further strengthened by evidence of higher risk of impaired HRQoL using
these criteria.
Methods: Consenting English, Chinese, Malay and Tamilspeaking primary care patients (age ⱖ 21 years) were interviewed using English/their respective mother tongue versions
of the EQ-5D/EQ-VAS, Health Utilities Index (HUI2 &
HUI3) and the SF-6D. We first evaluated the relationship
between BMI and HRQoL (overall and individual attributes
for each instrument) using multiple linear/logistic regression
(where appropriate) to adjust for factors known to affect
HRQoL. We next reorganized BMI into five categories
(reflecting the differences in cutoffs between International/
Asian classifications) and evaluated if median HRQoL scores
were significantly different across these categories.
Results: Among 411 participants [response rate: 87%;
median age: 51 years; obese: 19% (International); 33%
(Asian)], after adjusting for sociodemographic and other
factors, a tendency for underweight and obese subjects to
report lower overall HRQoL scores was observed for most
instruments. At the individual attribute level, obese subjects
reported significantly lower HUI2 pain scores (regression
coefficient: -0.035, P = 0.029) and greater odds of reporting
problems for SF-6D role-limitations (odds ratio: 2.9, P =
0.005). Median overall HRQoL scores were not significantly
different across the five BMI categories.
Conclusion: Consistent with available studies, obese subjects
reported worse HRQoL than normal-weight subjects. That
underweight subjects also reported worse HRQoL is interesting and requires confirmation. HRQoL was similar in Asians
using either WHO criteria.
Keywords: Asia, body mass index, obesity, primary, healthcare, quality of life, Singapore.
Introduction
of obesity in the 15 member states of the European
Union amounted to 32.8 trillion euros in 2002 [2]. In
the United States, in 1995, the total cost attributable to
obesity amounted to $99.2 billion, with approximately
$51.64 billion being direct medical costs [7]. To the
best of our knowledge, similar data are not available
for Asia.
There is an increasing recognition of the association
between high or low body mass index (BMI) and healthrelated quality of life (HRQoL) [8,9]. HRQoL refers to
the overall impact of a medical condition on the physical, mental, and social well-being of an individual [10].
In a cross-sectional study involving 5,817 people aged
14 to 61 from the general population, being overweight
The global prevalence of obesity has been increasing
and is fast approaching epidemic proportions in many
countries [1–4]. Obesity is associated with a significant
disease burden and costs and is therefore an important
public health concern. For example, obesity is an
important independent risk factor for cardiovascular
disease [5,6]. The total direct and indirect annual costs
Address correspondence to: Julian Thumboo, Department
of Rheumatology and Immunology, Singapore General
Hospital, Outram Road, Singapore 169608. E-mail: julian.
[email protected]
10.1111/j.1524-4733.2008.00374.x
© 2008, International Society for Pharmacoeconomics and Outcomes Research (ISPOR)
1098-3015/08/S105
S105–S114
S105
Wee et al.
S106
was associated with poorer functional status, considerable pain, worry, and restricted activity [11]. In another
cross-sectional study involving 5633 people aged 16 to
64 years from the general population, obesity was similarly found to be associated with worse HRQoL as
measured by the SF-36, and this association was reportedly modified by age and sex [12]. Besides crosssectional studies, other studies have evaluated the
impact of weight loss interventions [13,14] on HRQoL
with mixed results. In a meta-analysis of 34 randomized
controlled trials with behavioral, surgical or pharmacological interventions, nine showed improvements in
HRQoL measured using generic instruments [15]. Nevertheless, overall quality of these trials was judged to be
poor. Mixed results were also obtained from trials
evaluating the impact of bariatric surgery on HRQoL
[16]. A review of the impact of LAP-BAND placement
reported all aspects of HRQoL improved substantially,
especially physical disability, and post–weight-loss
HRQoL measures approximated those of the general
population [17]. Although several studies have shown
that obese subjects are more likely than nonobese
subjects to suffer from poorer physical health [18,19]
and in some studies, poorer mental health [20,21],
the relationship between obesity and HRQoL among
Asians is poorly understood. To the best of our knowledge, there was only one such published study to date
[22], which studied only one ethnic group (Chinese) and
showed that excess weight was associated with worse
physical health, but not mental health. Given that Asia
is a culturally diverse region, a better understanding of
the relationship between obesity and HRQoL in a
multiethnic Asian population would be important.
Additionally, few studies (in the East or West) have
evaluated the association between being underweight
and HRQoL. Using three measures of HRQoL (the
EQ-5D, EQ-VAS, and SF-6D) [23], Sach et al. found
that underweight subjects reported significantly poorer
HRQoL compared to normal-weight subjects as measured by the SF-6D. This result was not confirmed by
the scores from the EQ-5D and EQ-VAS. Huang et al.
[22] reported that compared to normal-weight subjects, underweight subjects had similar or slightly
lower scores on the physical health subscales of the
SF-36 and clearly lower scores on the mental health
subscales of the SF-36 domains, although the differences were not clinically important (i.e., less than 5
points on a 100-point scale). Nevertheless, these
results were not adjusted for the presence of medical
conditions which themselves might lead to being
underweight and/or might directly affect HRQoL.
Given that being underweight could be associated
with higher morbidity [24–27] and mortality [28–30],
and poorer health-seeking behavior [31], it is thus
important to study the association between being
underweight and HRQoL, with adjustment for other
potential confounders.
In 2004, the World Health Organization (WHO)
revised its recommendation for BMI cutoffs for Asians
[32] (see Methods) based on evidence that compared
to Caucasians, the risks of Type 2 diabetes and cardiovascular diseases among Asians is already substantial
even at BMI lower than the existing WHO cutoffs for
overweight (ⱖ25 kg/m2). The evidence base for this
recommendation could be further strengthened if there
are other areas of health (e.g., HRQoL) in which risk
of impairment is greater for Asians compared to Caucasians of comparable BMI.
Aims
Thus, the aims of this exploratory study were 1) to
evaluate the association between BMI and HRQoL (at
both overall and individual attribute levels) in a multiethnic Asian population in Singapore, and 2) to
explore if HRQoL impairment is greater at BMI
lower than the existing WHO cutoffs in this same
population.
Hypotheses
Based on the literature, we hypothesized that 1) compared to normal-weight subjects, underweight, preobese and obese subjects would experience decrements
in HRQoL, and 2) decrements in HRQoL would be
larger for obese compared to pre-obese and smaller for
underweight compared to obese subjects. To provide
further evidence to support the revised BMI cutoffs, we
would expect to observe a gradation in HRQoL across
the five BMI categories defined by the Asian and International BMI classifications, with successively lower
HRQoL being observed in successively higher BMI
categories.
Methods
Subjects and Study Design
Chinese, Malay and Indian patients were recruited as
part of an Institutional Review Board approved study
of HRQoL performed at the SingHealth Geylang Polyclinic, a public health-care institution facility with 15
doctors responsible for delivering primary care. Inclusion criteria were aged 21 and above, ability to comprehend one of the four survey languages (English,
Chinese, Malay, and Tamil) and absence of cognitive
impairment as assessed by the recruiters.
Consenting subjects were interviewed by interviewers of the same ethnicity using a questionnaire containing the Singapore English, Chinese, Malay, and Tamil
versions of the EQ-5D/EQ-VAS, HUI2, HUI3 and
SF-6D. Validity of several of these instruments in this
study sample was previously reported [33–36], while
validity for others is currently being reported (Wee
et al. unpublished data, 2006). Scores for the EQ-5D,
HUI2, HUI3, and SF-6D are all on the conventional
Body Mass Index and Health-Related Quality of Life among Asians
scale in which dead = 0.00 and perfect health = 1.00.
Sociodemographic and other factors known to influence HRQoL were also assessed. Physician-reported
acute and chronic medical conditions were obtained
using a standardized, pretested form.
Instruments
EuroQoL 5-Dimensions (EQ-5D). The EQ-5D is a
generic, preference-based instrument comprising a
health classification system with five dimensions
(mobility, self-care, usual activities, pain, and anxiety/
depression), each with three response levels (no
problem, some problems, and severe problems) and a
visual analog scale (EQ-VAS). The health classification
system describes a total of 243 health states, each of
which is assigned a utility weight, range -0.594 to 1,
using a utility scoring function derived from the UK
general population using the time trade-off method
[37]. We performed similar analyses using the US
scoring function [38] but did not present the results
here as they were very similar to those using the UK
scoring function. Respondents classified and rated
their health on the day of the survey. A difference of
0.07 or more in EQ-5D utility scores has been reported
to be clinically important [39] while a difference of 5
or more in EQ-VAS scores has been proposed to be
clinically important [40].
Health Utilities Index (HUI2 and HUI3). The HUI2
and HUI3 consist of two independent but complementary systems, which together describe almost 1 million
unique health states. Both instruments include a
generic comprehensive health status classification
system and a generic HRQoL utility scoring system,
using a utility scoring function derived from a representative sample of the Canadian general population
using the standard gamble (SG) and visual analog scale
(VAS) methods [41]. The scores for HUI2 range from
-0.03 (worst health state) to 1 (perfect health) and
scores for HUI3 range from -0.36 to 1. The questionnaire used in this study was the self-assessment, selfcompleted version of the HUI 15Q (4-week recall
period) that includes items sufficient to classify respondents in both the HUI2 and HUI3 systems [1,2]. A
difference of 0.03 or more for overall HUI2 and HUI3
scores is clearly clinically important [41–43].
SF-6D. The SF-6D is a six-dimensional health classification system comprising physical functioning, social
functioning, role-limitations, vitality, pain and mental
functioning, with four to six levels per dimension, thus
defining a total of 18,000 health states. The scores for
SF-6D range from 0.29 to 1 and were obtained using a
utility scoring function derived from a representative
sample of the UK general population using the SG
S107
technique [44]. A difference of 0.04 or more in SF-6D
scores is considered clinically important [45].
Family functioning measure (FFM). The FFM is a
three-item scale (five-level response options, poor to
excellent) previously validated in Singapore [46],
assessing the quality of interactions among family
members [47], with higher scores (range, 0–100)
reflecting better family functioning. High FFM scores
have previously been found to be associated with
better HRQoL [48].
Body mass index. We calculated subjects’ BMI (kg/m2)
by dividing their weight (in kilograms) by the square of
their height (in meters). Weight and height for all subjects were measured using the same instruments. We
evaluated the relationship between BMI and HRQoL
by treating BMI as a categorical variable according
to 1) WHO International BMI classification, and 2)
WHO revised cutoffs for Asians. The International
WHO BMI classifications are: 1) <18.5 kg/m2: underweight; 2) 18.5 to <25 kg/m2: normal-weight; 3) 25 to
<30 kg/m2: pre-obese and 4) ⱖ30 kg/m2: obese. The
WHO Asian BMI classifications are: 1) <18.5 kg/m2:
underweight; 2) 18.5 to <23 kg/m2: normal-weight;
3) 23 to <27.5 kg/m2: pre-obese and 4) ⱖ27.5 kg/m2:
obese. Using the new Asian classifications, subjects
with BMI between 23 to <25 kg/m2 would be reclassified as pre-obese and subjects with BMI between 27.5
to <30 kg/m2 would be reclassified as obese. Hence, we
additionally reorganized BMI (kg/m2) into five categories reflecting the differences in cutoffs between International and Asian classifications: 18.5 to <23, 23 to
<25, 25 to <27.5, 27.5 to <30, ⱖ30, for the purpose of
comparison between International and Asian cutoffs.
As the definition for underweight (<18.5 kg/m2) was
not different in the revised BMI definition, this category was not included in the reorganization.
Statistical Analyses
We compared differences in subject characteristics by
ethnicity using chi-square tests for categorical variables and Kruskal–Wallis test for continuous variables.
To evaluate the influence of BMI on overall HRQoL
scores, we performed multiple linear regression (MLR)
analyses with overall HRQoL scores as dependent
variables in separate models, while adjusting for other
covariates that may potentially influence HRQoL
including sociodemographic (i.e., age, sex, ethnicity,
and years of education) and other factors including
marital status (yes/no), smoking status (yes/no), work
status (yes/no), presence of chronic medical conditions,
and FFM scores. To allow for comparisons with other
published studies, we used BMI categories defined
according to WHO International classification (i.e.,
non-Asian classification) in the MLR analyses.
Wee et al.
S108
Table 1
Characteristics of subjects, mean body mass index and health-related quality of life (HRQoL) scores by ethnicity
Ethnicity
N (%) unless otherwise stated
Median age (IQR) (years)
Female
Median education (IQR) (years)
Married
Working
Smoking cigarettes
Presence of chronic medical conditions*
Median family functioning scores (IQR)
Body mass index (International classification) (kg/m2)
Underweight (<18.5)
Normal weight (18.5 to <25)
Pre-obese (2.5 to <30)
Obese (ⱖ30)
Body mass index (Asian classification) (kg/m2)
Underweight (<18.5)
Normal weight (18.5 to <27.5)
Pre-obese (23 to <27.5)
Obese (ⱖ27.5)
Median HRQoL scores (IQR)
EQ-5D (possible score range: -0.59 to 1)
EQ-VAS (possible score range: 0 to 100)
HUI2 (possible score range: -0.03 to 1)
HUI3 (possible score range: -0.36 to 1)
SF-6D (possible score range: 0.29 to 1)
All (n = 411)
Chinese (n = 164)
Malays (n = 127)
Indians (n = 120)
P-values
51.0 (38.0, 61.0)
216 (52.6)
8.0 (3.0, 10.0)
288 (70.1)
247 (60.1)
77 (18.7)
259 (63.0)
66.7 (50.0, 75.0)
51.0 (39.0, 61.0)
78 (47.6)
10.0 (3.0, 12.0)
109 (66.5)
102 (62.2)
29 (17.7)
109 (66.5)
58.3 (41.7, 75.0)
52.0 (39.0, 64.0)
67 (52.8)
8.0 (3.0, 10.0)
88 (69.3)
73 (57.5)
28 (22.1)
86 (67.7)
62.5 (50.0, 75.0)
49.0 (37.0, 63.0)
71 (59.2)
8.0 (3.0, 10.0)
91 (75.8)
72 (60.0)
20 (16.7)
64 (53.3)
75.0 (50.0, 83.3)
0.66
0.16
0.002
0.23
0.72
0.50
0.032
0.013
<0.001
17 (4.1)
166 (40.4)
150 (36.5)
78 (19.0)
14 (8.5)
92 (56.1)
49 (29.9)
9 (5.5)
2 (1.6)
41 (32.3)
44 (34.7)
40 (31.5)
1 (0.8)
33 (27.5)
57 (47.5)
29 (24.2)
17 (4.1)
95 (23.1)
163 (39.7)
136 (33.1)
14 (8.5)
56 (34.2)
69 (42.1)
25 (15.2)
2 (1.6)
24 (18.9)
45 (35.4)
56 (44.1)
1 (0.8)
15 (12.5)
49 (40.8)
55 (45.8)
0.80 (0.73, 1)
70.0 (60.0, 80.0)
0.88 (0.80, 0.95)
0.85 (0.68, 0.92)
0.89 (0.79, 0.95)
0.80 (0.73, 1)
70.0 (50.0, 80.0)
0.87 (0.80, 0.92)
0.83 (0.67, 0.91)
0.87 (0.76, 0.94)
0.80 (0.73, 1)
70.0 (60.0, 80.0)
0.90 (0.80, 0.95)
0.88 (0.79, 0.95)
0.94 (0.81, 1)
0.80 (0.73, 0.97)
70.0 (60.0, 80.0)
0.88 (0.80, 0.95)
0.82 (0.62, 0.93)
0.88 (0.76, 0.94)
<0.001
0.18
0.20
0.069
0.002
<0.001
*List of physician-reported chronic medical conditions include hypertension, diabetes mellitus, ischemic heart disease, cerebrovascular accident, hyperlipidaemia, osteoarthritis,
other type of arthritis, rheumatism and asthma.
IQR, interquartile range.
To evaluate the influence of BMI on individual
attributes of HRQoL, we performed MLR analyses
with HUI2 and HUI3 individual single-attribute utility
scores as dependent variables in separate models, while
adjusting for the potential influence of sociodemographic and other covariates (as specified above). As
individual attribute scores are not available for EQ-5D
and SF-6D, we collapsed responses to EQ-5D and
SF-6D items into two levels—with or without problems, and performed multiple logistic regression
(LOGIT) with these dichotomous responses as dependent variables in separate models. This approach of
collapsing responses has also been used in other studies
[49,50]. We similarly adjusted for the potential influence of sociodemographic factors and other covariates.
To evaluate if HRQoL impairment is greater at BMI
lower than the existing WHO cutoffs, we compared,
using Kruskal–Wallis tests, median HRQoL scores
across the five predefined BMI categories that contrast
the International and Asian cutoffs as aforementioned.
Statistical significance was defined at P < 0.05. We did
not adjust for multiple comparisons as based on the
approach taken by several respected researchers that
the presumption (of “universal” null hypothesis)
underlying the theory of adjustment for multiple comparisons does not hold [51–53].
Results
Subjects
Of 660 subjects approached, 574 participated (response rate: 87%), 108 did not complete the study,
another 55 subjects were excluded from this analysis
because 23 had missing sociodemographic or clinical
information, 29 had missing HRQoL data, and three
had BMI above 50.0 kg/m2 (identified as outliers by
visual inspection of scatter plots). Thus, a total of 411
subjects (40% Chinese, 31% Malays and 29%
Indians) provided complete data for analysis. The percentage of obese subjects was 19% by the WHO
International classification and 33% by the WHO
Asian classification. Details of subject characteristics
and HRQoL scores are summarized in Table 1. As
compared with other ethnic groups, Chinese subjects
reported significantly more years of education and
lower FFM scores while Malay subjects reported a
significantly higher prevalence of chronic medical
conditions, and higher HUI3 and SF-6D utility scores.
Influence of BMI on Overall HRQoL Scores (Table 2)
With or without adjustment for sociodemographic
factors and other covariates, compared to normalweight subjects, a trend of underweight and obese
subjects reporting lower overall HRQoL scores (i.e.,
negative regression coefficients) was observed on most
instruments, with some of these differences being statistically significant but not clinically important. Compared to normal-weight subjects, pre-obese subjects
generally reported better (albeit marginally) overall
HRQoL scores on all instruments except the SF-6D.
Using the SF-6D, pre-obese subjects reported worse
(albeit marginally) overall HRQoL scores compared to
normal-weight subjects.
-0.096
(-0.18, -0.012)
0
0.018
(-0.019, 0.055)
-0.036
(-0.082, 0.011)
-0.082
(-0.17, 0.004)
0
0.0003
(-0.038, 0.038)
-0.057
(-0.10, -0.01)
0.13
0.35
0.026*
0.015*
0.99
0.06
P-value
-11
(-18, -3.4)
0
0.3
(-3.0, 3.6)
-3.6
(-7.8, 0.5)
-9.4
(-17, 2.1)
0
0.7
(-2.6, 3.9)
-2.2
(-6.2, 1.7)
EQ-VAS
0.085
0.85
0.005*
0.27
0.68
0.012*
P-value
-0.034
(-0.010, 0.032)
0
0.011
(-0.018, 0.040)
-0.013
(-0.050, 0.024)
-0.019
(-0.085, 0.047)
0
0.007
(-0.022, 0.037)
-0.011
(-0.047, 0.024)
HUI2
utility scores
0.48
0.46
0.32
0.54
0.62
0.57
P-value
-0.097
(-0.21, 0.013)
0
0.029
(-0.020, 0.077)
-0.011
(-0.072, 0.050)
-0.067
(-0.018, 0.046)
0
0.024
(-0.026, 0.074)
-0.002
(-0.064, 0.059)
HUI3
utility scores
Regression coefficients, 95% confidence interval
0.71
0.25
0.085
0.94
0.35
0.25
P-value
-0.025
(-0.082, 0.032)
0
-0.001
(-0.026, 0.025)
-0.027
(-0.059, 0.004)
-0.017
(-0.076, 0.042)
0
-0.006
(-0.032, 0.020)
-0.018
(-0.050, 0.014)
SF-6D
utility scores
*P < 0.05, showing subjects in this category had health-related quality of life scores significantly different from subjects with normal weight.
†
Based on WHO International classification: 1) <18.5 kg/m2: underweight; 2) 18.5 to <25 kg/m2: normal-weight; 3) 25 to <30 kg/m2: pre-obese and 4) ⱖ30 kg/m2: obese.
‡Reference category: normal-weight subjects.
§
Other covariates include marital status (yes/no), smoking status (yes/no), work status (yes/no), presence of chronic medical conditions (yes/no) and Family Functioning Measures scores (continuous variable, score range 0–100).
HUI, Health Utilities Index.
Obese (n = 78)
Normal weight‡ (n = 166)
Pre-obese (n = 150)
Adjusting for age, sex,
ethnicity, education and
other covariates§
BMI† (kg/m2)
Underweight (n = 17)
Obese (n = 78)
Normal weight (n = 166)
Pre-obese (n = 150)
‡
Not adjusting for covariates
BMI† (kg/m2)
Underweight (n = 17)
EQ-5D
utility scores
0.09
0.96
0.39
0.27
0.64
0.58
P-value
Table 2 Results of five linear regression models relating five health related-quality of life scores with body mass index (BMI) categories (with normal-weight as reference group), with and
without adjustment for covariates
Body Mass Index and Health-Related Quality of Life among Asians
S109
Wee et al.
S110
Table 3 Results of three linear regression models relating three single attribute utility scores with body mass index (BMI) categories
(with normal weight as reference group), with and without adjustment for covariates
Regression coefficients, 95% confidence interval
HUI2 pain
Not adjusting for covariates
BMI† (kg/m2)
Underweight (n = 17)
Normal weight‡ (n = 166)
Pre-obese (n = 150)
Obese (n = 78)
Adjusting for age, sex,
ethnicity, education and
other covariates§
BMI† (kg/m2)
Underweight (n = 17)
Normal weight‡ (n = 166)
Pre-obese (n = 150)
Obese (n = 78)
-0.015 (-0.070, 0.040)
0
-0.013 (-0.037, 0.011)
-0.044 (-0.074, 0.015)
-0.026 (-0.083, 0.030)
0
-0.007 (-0.033, 0.018)
-0.035 (-0.066, -0.0035)
P-value
0.59
0.29
0.004*
0.36
0.56
0.029*
HUI3 dexterity
P-value
HUI3 pain
P-value
0.003 (-0.030, 0.036)
0
0.002 (-0.013, 0.017)
-0.019 (-0.036, -0.0007)
0.85
-0.10 (-0.16, -0.042)
0
-0.006 (-0.032, 0.018)
-0.020 (-0.053, 0.009)
<0.001*
-0.10 (-0.16, 0.041)
0
-0.006 (-0.032, 0.020)
-0.020 (-0.054, 0.012)
0.001*
-0.001 (-0.036, 0.034)
0
0.004 (-0.012, 0.019)
-0.015 (-0.034, 0.004)
0.79
0.04*
0.96
0.63
0.12
0.66
0.23
0.63
0.23
*P < 0.05, showing subjects in this category had single attribute utility scores significantly different from subjects with normal weight.
†
Based on WHO International classification: 1) <18.5 kg/m2: underweight; 2) 18.5 to <25 kg/m2: normal weight; 3) 25 to <30 kg/m2: pre-obese; and 4) ⱖ30 kg/m2: obese.
‡
Reference category: normal weight subjects.
§Other covariates include marital status (yes/no), smoking status (yes/no), work status (yes/no), presence of chronic medical conditions (yes/no) and Family Functioning Measures
scores (continuous variable, score range 0–100).
HUI, Health Utilities Index.
Without adjustment for sociodemographic factors
and other covariates, compared to normal weight
subjects, obese subjects reported significantly lower
EQ-5D utility scores (regression coefficient: -0.057,
P = 0.015). Nevertheless, with adjustment for sociodemographic factors and other covariates, the impact of
obesity on EQ-5D utility scores (-0.036, P = 0.13) was
no longer statistically significant. Without adjustment
for sociodemographic factors and other covariates,
compared to normal weight subjects, underweight subjects reported lower EQ-VAS scores (regression coefficient: -9.4, P = 0.012), which were both statistically
significant and clinically important. With adjustment
for sociodemographic factors and other covariates,
underweight subjects reported lower EQ-5D utility
(-0.096, P = 0.026) and EQ-VAS scores (-10.8,
P = 0.005), the magnitude of which was both statistically significant and clinically important, as well as
lower HUI2 (-0.034, P = 0.32) and HUI3 (-0.097,
P = 0.085) utility scores; the differences were clinically
important but did not reach statistical significance.
Interestingly, among all BMI categories, underweight rather than obese subjects reported the largest
decrements in EQ-5D utility, EQ-VAS, HUI2 and HUI3
utility scores while obese subjects reported the largest
decrements in SF-6D utility scores. As coexisting
chronic medical condition was a potential confounder
in this situation, we evaluated the prevalence of any
chronic medical conditions across BMI categories and
found that they were not significantly different. The
prevalence of chronic medical conditions in the various
BMI categories were 82% (underweight), 57%
(normal), 64% (pre-obese) and 69% (obese), respectively (P = 0.091). Prevalence of specific chronic
medical conditions (e.g., diabetes, hypertension, etc.)
was also similar across BMI categories (results not
shown).
Influence of BMI on Individual HRQoL Attributes
(Tables 3 and 4)
Without adjustment for sociodemographic factors
and other covariates, the influence of BMI on individual HRQoL attributes was statistically significant
for HUI2 pain, HUI3 dexterity and pain, EQ-5D
mobility and usual activities, as well as SF-6D role
limitations and pain attributes (Tables 3 and 4, results
shown only for those attributes reaching statistical
significance). With adjustment for sociodemographic
factors and other covariates, compared to normalweight subjects, underweight subjects reported significantly lower HUI3 pain scores (regression coefficient:
-0.10, P = 0.001, Table 3) and greater odds of reporting problems on EQ-5D usual activities (odds ratio:
8.3, P = 0.011, Table 4).
With adjustment for sociodemographic factors and
other covariates, compared to normal-weight subjects,
obese subjects reported significantly lower HUI2
pain scores (regression coefficient: -0.035, P = 0.029,
Table 3) and greater odds of reporting problems on
SF-6D role limitations (odds ratio: 2.9, P = 0.005,
Table 4).
Overall HRQoL Scores Across Five BMI Categories
Median overall HRQoL scores were not significantly
different across the five BMI categories (Table 5), suggesting that unlike risk of Type 2 diabetes and cardiovascular diseases, use of the different BMI cutoffs
would have no impact as far as HRQoL is concerned.
S111
Body Mass Index and Health-Related Quality of Life among Asians
Table 4 Results of four logistic regression models relating four dimensions of health with body mass index (BMI) categories (with
normal weight as reference group), with and without adjustment for covariates
Odds ratio for reporting problems, 95% confidence interval
EQ-5D mobility
Not adjusting for covariates
BMI† (kg/m2)
Underweight (n = 17)
Normal weight‡ (n = 166)
Pre-obese (n = 150)
Obese (n = 78)
Adjusting for age, sex,
ethnicity, education and
other covariates§
BMI† (kg/m2)
Underweight (n = 17)
Normal weight‡ (n = 166)
Pre-obese (n = 150)
Obese (n = 78)
0.8 (0.2, 3.7)
1.0
1.5 (0.8, 2.7)
2.5 (1.3, 4.7)
2.1 (0.4, 12.0)
1.0
1.1 (0.5, 2.2)
2.0 (0.9, 4.5)
P-value
EQ-5D usual activities
0.76
2.3 (0.6, 9.1)
1.0
1.5 (0.7, 3.1)
2.6 (1.2, 5.7)
0.19
0.006*
0.40
8.3 (1.6, 41.9)
1.0
1.1 (0.5, 2.5)
1.8 (0.7, 4.4)
0.83
0.08
P-value
SF-6D role limitations
0.22
0.30
0.018*
0.011*
0.86
0.19
1.6 (0.5, 5.2)
1.0
1.1 (0.6, 2.0)
2.3 (1.2, 4.3)
2.2 (0.6, 7.9)
1.0
1.1 (0.6, 2.1)
2.9 (1.4, 6.1)
P-value
0.45
0.68
0.01*
0.25
0.82
0.005*
SF-6D pain
0.8 (0.3, 2.3)
1.0
1.7 (1.1, 2.7)
1.3 (0.8, 2.3)
0.9 (0.3, 2.7)
1.0
1.4 (0.9, 2.3)
1.1 (0.6, 2.1)
P-value
0.73
0.027*
0.31
0.86
0.18
0.73
*P < 0.05, showing that the odds of subjects in this category reporting problems on any health dimensions were significantly different from subjects with normal weight.
†
Based on World Health Organization International classification: i) <18.5 kg/m2: underweight; ii) 18.5 to <25 kg/m2: normal weight; iii) 25 to <30 kg/m2: pre-obese and iv)
ⱖ30 kg/m2: obese.
‡Reference category: normal weight subjects.
§
Other covariates include marital status (yes/no), smoking status (yes/no), work status (yes/no), presence of chronic medical conditions (yes/no) and Family Functioning Measures
scores (continuous variable, score range 0–100).
Discussion
In this exploratory study, the first of its kind in a
multiethnic Asian population and one of only two
published Asian studies, we evaluated the association
between BMI and HRQoL. As hypothesized, the magnitude of decrements in overall HRQoL scores was
larger for obese compared to pre-obese subjects. Nevertheless, contrary to our a priori hypothesis, compared to normal weight subjects, pre-obese subjects
reported marginally better overall HRQoL scores on
five of six HRQoL instruments. Furthermore, we were
surprised to find that underweight, rather than obese
subjects, reported the largest decrements in overall
HRQoL scores for four of five instruments. We regard
these findings as interesting but preliminary given the
small number of underweight subjects in this study. In
addition, it is perhaps noteworthy to find that median
HRQoL scores across five BMI categories (defined to
reflect the different cutoffs between International and
Asian BMI classifications) were not significantly different, thus suggesting that the use of these different BMI
cutoffs would have no impact as far as HRQoL is
concerned. If this result is confirmed in additional
Table 5
Comparison of median health-related quality of life (HRQoL) scores across body mass index (BMI) categories
Median overall HRQoL scores
(IQR)
EQ-5D
EQ-VAS
HUI2
HUI3
SF-6D
studies, it might imply that the same WHO International classification may be used in both Asian and
Caucasian studies evaluating the association between
BMI and HRQoL, thus enhancing the comparability of
findings from such studies.
Our observation that underweight subjects reported
worst HRQoL compared to pre-obese and obese subjects differs from published studies [22,23] deserves
comment. First, the discrepancies in our findings were
unlikely to be explained by choice of instrument as
identical instruments (EQ-5D, EQ-VAS, and the
SF-6D) were used by Sach et al. [23]. Second, the discrepancies were unlikely to be attributed to poor construct validity of these instruments as the validity
of these instruments in this study population have
been previously or are currently being reported (see
Methods). Third, error in BMI measurement is
unlikely to explain the discrepancies as we used actual
measurements of weight and height, rather than figures
based on self-report [22,23]. In fact, among all identified population-based studies evaluating the relationship between BMI and HRQoL [18–23], ours was
the only study to use actual measurements of weight
and height. Discrepancies in actual and self-reported
BMI (kg/m2)
18.5 to <23
0.85
70.0
0.87
0.85
0.88
(0.73,
(60.0,
(0.80,
(0.69,
(0.80,
IQR, interquartile range; HUI, Health Utilities Index.
1)
80.0)
0.92)
0.92)
0.95)
23 to <25
0.80
70.0
0.87
0.82
0.88
(0.73,
(60.0,
(0.80,
(0.69,
(0.78,
1)
80.0)
0.95)
0.92)
0.95)
25 to <27.5
27.5 to <30
ⱖ30
P-value
0.80 (0.73, 1)
72.5 (60.0, 80.0)
0.88 (0.80, 0.95)
0.88 (0.71, 0.93)
0.89 (0.79, 0.95)
0.80 (0.73, 1)
70.0 (60.0, 80.0)
0.88 (0.80, 0.95)
0.85 (0.64, 0.95)
0.89 (0.80, 0.95)
0.80 (0.73, 1)
70.0 (60.0, 80.0)
0.90 (0.78, 0.95)
0.85 (0.66, 0.93)
0.88 (0.74, 0.95)
0.34
0.47
0.93
0.70
0.98
Wee et al.
S112
weight and height have been previously reported [54].
Fourth, cultural differences were also unlikely to fully
explain the discrepancies in findings as a study among
Chinese in Taiwan [22] also found that higher BMI
was associated with worse physical (but not mental)
HRQoL measured using the SF-36. Having ruled out
several possible sources of errors, we shall attempt to
provide some plausible explanations for the observed
discrepancies, although these need to be confirmed in
future studies. First, BMI may not be an appropriate
marker for obesity even though it is most easily measured [27]. In view of this, further studies are required
to study the association between obesity and HRQoL
using other markers of obesity such as waist-to-hip
ratio or percent body fat. Second, another possible
explanation is that the effect size of the influence of
BMI on HRQoL is generally small. In the only published study that directly reported the effect sizes of the
influence of BMI on SF-36 subscales and summary
scales [22], the effect size did not exceed 0.5 (Cohen’s
definition of medium effect size) [55] for any of the
subscales or summary scales, with the largest effect size
of 0.34 being reported for obese subjects on the physical functioning subscale without adjusting for any
covariates. Third, it should be noted that BMI is a risk
factor of poor health. Being obese increases the probability of developing chronic conditions such as type 2
diabetes and cardiovascular disease. Thus, the absence
of a stronger relationship between current HRQoL
and current BMI in a cross-sectional analysis may be
because of the effects of BMI on health status and
therefore HRQoL have not yet occurred. It will be
important to explore the relationship between BMI
and HRQoL in longitudinal studies.
Limitations
We recognize several limitations of this study. First,
our subjects were recruited from a primary care setting
and our findings may therefore not be generalizable to
the general population. Nevertheless, our results are
interesting in suggesting that the association of BMI
with HRQoL among Asians could be different from
that among Caucasians and provides a basis for future
research. Second, as the number of underweight subjects is small (n = 17), and confidence interval is wide
for some results, we advise caution in interpreting the
findings. Further studies with disproportionate sampling to include a larger number of underweight subjects may be useful to confirm our findings. Third, we
have computed health utility scores using population
preference weights from the UK and Canada and these
may not fully reflect local preferences. Nevertheless,
this is a pragmatic compromise as a Singapore utility
function was not available at the time of this study.
Fourth, the cross-sectional nature of this study does
not allow us to make any inferences about the causal
associations between BMI and HRQoL. Future pro-
spective longitudinal studies are required to better
understand the causal association, if any, between
BMI and HRQoL. Finally, although obesity-specific
HRQoL instruments could be more sensitive than the
generic instruments that we had used in our study, the
reliability and validity of such instruments have not
been previously evaluated in this multiethnic Asian
population and are thus not available for use in our
study.
Conclusions
In conclusion, in this multiethnic Asian population,
underweight subjects unexpectedly reported poorer
HRQoL than normal weight, pre-obese, and obese
subjects on five of six HRQoL instruments. If further
studies confirm that the use of the different BMI
cutoffs would have no impact as far as HRQoL is
concerned, then the WHO International classification
may be used in both Asian and Caucasian studies
evaluating the association between BMI and HRQoL,
thus improving the comparability of findings from
such studies.
The authors wish to thank Dr. Tai Ee Shyong for his valuable
comments on an earlier draft of this manuscript and Ms.
Coralie Ang, Ms. Gladys Yap, Ms. Malini and other staff of
SingHealth Polyclinics (Geylang), Republic of Singapore, for
their assistance with the logistics of this study.
Source of financial support: This study was funded by programme grant 03/1/27/18/226 from the Biomedical Research
Council of Singapore.
Conflict of Interest: It should be noted that David Feeny has
a proprietary interest in Health Utilities Incorporated,
Dundas, Ontario, Canada. HUInc. distributes copyrighted
Health Utilities Index (HUI) materials and provides methodological advice on the use of HUI.
References
1 Dal Grande E, Gill T, Taylor AW, et al. Obesity in
South Australian adults—prevalence, projections and
generational assessment over 13 years. Aust N Z J
Public Health 2005;29:343–8.
2 Fry J, Finley W. The prevalence and costs of obesity in
the EU. Proc Nutr Soc 2005;64:359–62.
3 Wang HS, Zhai F, Popkin BM. Trends in the distribution of body mass index among Chinese adults,
aged 20–45 years (1989–2000). Int J Obes (Lond)
2007;31:272–8.
4 Ogden CL, Carroll MD, Curtin LR, in the United
States, 1999–2004. JAMA 2006;295:1549–55.
5 Zhou B, Wu Y, Yang J, et al. Overweight is an independent risk factor for cardiovascular disease in
Chinese populations. Obes Rev 2002;3:147–56.
Body Mass Index and Health-Related Quality of Life among Asians
6 Nicklas BJ, Cesari M, Penninx BW, et al. Abdominal
obesity is an independent risk factor for chronic heart
failure in older people. J Am Geriatr Soc 2006;54:
413–20.
7 Wolf AM, Colditz GA. Current estimates of the economic cost of obesity in the United States. Obes Res
1998;6:97–106.
8 Katz DA, McHorney CA, Atkinson RL. Impact of
obesity on health-related quality of life in patients with
chronic illness. J Gen Intern Med 2000;15:789–96.
9 Jensen GL. Obesity and functional decline: epidemiology and geriatric consequences. Clin Geriatr Med
2005;21:677–87.
10 Guyatt GH, Feeny DH, Patrick DL. Measuring
health-related quality of life. Ann Intern Med 1993;
118:622–9.
11 Stewart AL, Brook RH. Effects of being overweight.
Am J Public Health 1983;73:171–8.
12 Larsson U, Karlsson J, Sullivan M. Impact of overweight and obesity on health-related quality of life—a
Swedish population study. Int J Obes Relat Metab
Disord 2002;26:417–24.
13 Torquati A, Lutfi RE, Richards WO. Predictors of
early quality-of-life improvement after laparoscopic
gastric bypass surgery. Am J Surg 2007;193:471–5.
14 Karlsson J, Taft C, Ryden A, et al. Ten-year trends in
health-related quality of life after surgical and conventional treatment for severe obesity: the SOS intervention study. Int J Obes (Lond) 2007;31:1248–61.
15 Maciejewski ML, Patrick DL, Williamson DF. A
structured review of randomized controlled trials of
weight loss showed little improvement in healthrelated quality of life. J Clin Epidemiol 2005;58:568–
78.
16 Ballantyne GH. Measuring outcomes following bariatric surgery: weight loss parameters, improvement in
co-morbid conditions, change in quality of life and
patient satisfaction. Obes Surg 2003;13:954–64.
17 Dixon JB, O’Brien PE. Changes in comorbidities and
improvements in quality of life after LAP-BAND
placement. Am J Surg 2002;184(Suppl.):S51–4.
18 Doll HA, Petersen SE, Stewart-Brown SL. Obesity and
physical and emotional well-being: associations
between body mass index, chronic illness, and the
physical and mental components of the SF-36 questionnaire. Obes Res 2000;8:160–70.
19 Jia H, Lubetkin EI. The impact of obesity on healthrelated quality-of-life in the general adult US population. J Public Health (Oxf) 2005;27:156–64.
20 Hassan MK, Joshi AV, Madhavan SS, Amonkar MM.
Obesity and health-related quality of life: a crosssectional analysis of the US population. Int J Obes
Relat Metab Disord 2003;27:1227–32.
21 Trakas K, Oh PI, Singh S, et al. The health status of
obese individuals in Canada. Int J Obes Relat Metab
Disord 2001;25:662–8.
22 Huang IC, Frangakis C, Wu AW. The relationship of
excess body weight and health-related quality of life:
evidence from a population study in Taiwan. Int J
Obes (Lond) 2006;30:1250–9.
23 Sach TH, Barton GR, Doherty M, et al. The relationship between body mass index and health-related
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
S113
quality of life: comparing the EQ-5D, EuroQol VAS
and SF-6D. Int J Obes (Lond) 2007;31:189–96.
Samanic C, Chow WH, Gridley G, et al. Relation of
body mass index to cancer risk in 362,552 Swedish
men. Cancer Causes Control 2006;17:901–9.
Meacham LR, Gurney JG, Mertens AC, et al. Body
mass index in long-term adult survivors of childhood
cancer: a report of the Childhood Cancer Survivor
Study. Cancer 2005;103:1730–9.
Reeves BC, Ascione R, Chamberlain MH, Angelini
GD. Effect of body mass index on early outcomes in
patients undergoing coronary artery bypass surgery.
J Am Coll Cardiol 2003;42:668–76.
Romero-Corral A, Montori VM, Somers VK, et al.
Association of bodyweight with total mortality and
with cardiovascular events in coronary artery disease:
a systematic review of cohort studies. Lancet 2006;
368:666–78.
Jee SH, Sull JW, Park J, et al. Body-mass index and
mortality in Korean men and women. N Engl J Med
2006;355:779–87.
Gu D, He J, Duan X, et al. Body weight and mortality
among men and women in China. JAMA 2006;295:
776–83.
Navarro WH, Loberiza FR Jr, Bajorunaite R, et al.
Effect of body mass index on mortality of patients
with lymphoma undergoing autologous hematopoietic cell transplantation. Biol Blood Marrow Transplant 2006;12:541–51.
Zhu K, Wu H, Jatoi I, et al. Body mass index and use
of mammography screening in the United States. Prev
Med 2006;42:381–5.
WHO Expert Consultation. Appropriate body-mass
index for Asian populations and its implications for
policy and intervention strategies. Lancet 2004;363:
157–63.
Luo N, Seng BK, Thumboo J, et al. A Study of the
Construct Validity of the Health Utilities Index Mark
3 (HUI3) in Patients with Schizophrenia. Qual Life
Res 2006;15:889–98.
Wang Q, Chen G. The health status of the Singaporean population as measured by a multi-attribute
health status system. Singapore Med J 1999;40:389–
96.
Luo N, Chew LH, Fong KY, et al. Validity and reliability of the EQ-5D self-report questionnaire in
Chinese-speaking patients with rheumatic diseases in
Singapore. Ann Acad Med Singapore 2003;32:685–
90.
Luo N, Chew LH, Fong KY, et al. Validity and reliability of the EQ-5D self-report questionnaire in
English-speaking Asian patients with rheumatic diseases in Singapore. Qual Life Res 2003;12:87–92.
Dolan P. Modeling valuations for EuroQol health
states. Med Care 1997;35:1095–108.
Shaw JW, Johnson JA, Coons SJ. US valuation of
the EQ-5D health states: development and testing of
the D1 valuation model. Med Care 2005;43:203–
20.
Brazier J, Roberts J, Tsuchiya A, Busschbach J. A
comparison of the EQ-5D and SF-6D across seven
patient groups. Health Econ 2004;13:873–84.
Wee et al.
S114
40 Luo N, Chew LH, Fong KY, et al. Do English and
Chinese EQ-5D versions demonstrate measurement
equivalence? an exploratory study. Health Qual Life
Outcomes 2003;1:7.
41 Horsman J, Furlong W, Feeny D, Torrance G. The
Health Utilities Index (HUI(R)): concepts, measurement properties and applications. Health Qual Life
Outcomes 2003;1:54.
42 Feeny D, Furlong W, Barr RD, et al. A comprehensive
multiattribute system for classifying the health status
of survivors of childhood cancer. J Clin Oncol 1992;
10:923–8.
43 Feeny D, Furlong W, Torrance GW, et al. Multiattribute and single-attribute utility functions for the
health utilities index mark 3 system. Med Care 2002;
40:113–28.
44 Brazier J, Roberts J, Deverill M. The estimation of a
preference-based measure of health from the SF-36.
J Health Econ 2002;21:271–92.
45 Walters SJ, Brazier JE. Comparison of the minimally
important difference for two health state utility measures: EQ-5D and SF-6D. Qual Life Res 2005;14:
1523–32.
46 Thumboo J, Fong KY, Chan SP, et al. Validation of
the medical outcomes study family and marital functioning measures in SLE patients in Singapore. Lupus
1999;8:514–20.
47 Sherbourne CD, Kamberg CJ. Social functioning:
family and marital functioning measures. In: Stewart
48
49
50
51
52
53
54
55
AL, Ware JE Jr, eds. Measuring Functioning and
Well-Being: The Medical Outcomes Study Approach.
Durham, NC: Duke University Press, 1992.
Thumboo J, Fong KY, Machin D, et al. Quality of life
in an urban Asian population: the impact of ethnicity
and socio-economic status. Soc Sci Med 2003;56:
1761–72.
Roset M, Badia X, Mayo NE. Sample size calculations
in studies using the EuroQol 5D. Qual Life Res 1999;
8:539–49.
Pallant JF, Misajon R, Bennett E, Manderson L. Measuring the impact and distress of health problems
from the individual’s perspective: development of the
Perceived Impact of Problem Profile (PIPP). Health
Qual Life Outcomes 2006;4:36.
Rothman KJ. No adjustments are needed for multiple
comparisons. Epidemiology 1990;1:43–6.
Feise RJ. Do multiple outcome measures require
p-value adjustment? BMC Med Res Methodol 2002;
2:8.
Schulz KF, Grimes DA. Multiplicity in randomised
trials I: endpoints and treatments. Lancet 2005;365:
1591–5.
Palta M, Prineas RJ, Berman R, Hannan P. Comparison of self-reported and measured height and weight.
Am J Epidemiol 1982;115:223–30.
Cohen J. Statistical Power Analysis for the Behavioral
Sciences (2nd ed.). Hillsdale, NJ: L. Erlbaum Associates, 1988.