An innovative prognostic model for predicting diabetes risk in the Thai population.

Abstract

To estimate the prevalence and type 2 diabetes, and to develop a prognostic model for identifying individuals at high risk of undiagnosed type 2 diabetes.
The study was designed as a cross-sectional investigation with 4314 participants of Thai background, aged between 15 and 85 years (mean age: 48). Fasting plasma glucose was initially measured, and repeated if the first measurement was more than 126 mg/dl. Type 2 diabetes was diagnosed using the World Health Organization's criteria. Logistic regression model was used to develop prognostic models for men and women separately. The prognostic performance of the model was assessed by the area under the receiver operating characteristic curve (AUC) and a nomogram was constructed from the logistic regression model.
The overall prevalence of type 2 diabetes was 7.4% (n = 125/1693) in men and 3.4% (n = 98/2621) in women. In either gender, the prevalence increased with age and body mass index (BMI). Gender, age, BMI and systolic blood pressure (SBP) were independently associated with type 2 diabetes risk. Based on the estimated parameters of model, a nomogram was constructed for predicting diabetes separated by gender. The AUC for the model with 3 factors was 0.75.
These data suggest that the combination of age, BMI and systolic blood pressure could help identify Thai individuals at high risk of undiagnosed diabetes.

Full text

Page 1

An innovative prognostic model for predicting diabetes risk
in Thai population
Chatlert Pongchaiyakula, Praew Kotruchina, Ekgaluck Wanothayaroja,*, Tuan V. Nguyenb,c
aDivision of Endocrinology and Metabolism, Department of Medicine, Faculty of Medicine, Khon Kaen University, Khon Kaen 40002, Thailand
bGarvan Institute of Medical Research, University of New South Wales, Sydney, Australia
cSchool of Public Health and Community Medicine, University of New South Wales, Sydney, Australia
1.
Introduction
Type 2 diabetes is increasingly recognized as a public health
burden in developing countries [1,2]. Recent studies have
suggested that the prevalence of type 2 diabetes in Asian
populations ranged between 6 and 12% [1,3–5], which is
comparable to estimates from Caucasian populations in the
US [6]. Indeed, it has been projected that a large increase in the
global prevalence of diabetes will take place in developing
countries with 80% of all new cases of diabetes expected to
appear in the developing countries by 2025 [7]. The Asia-Pacific
region, including Thailand, has been identified as of ‘‘prime
importance to the epidemiology of diabetes’’ [8] because the
prevalence of diabetes in this part of the world is higher than
in other developing countries [9].
One of the most significant current discussions in diabetes
is how to identify individuals at high risk of having the disease.
A large proportion of diabetic cases are undiagnosed (e.g.,
asymptomatic), yet these individuals are at risk of having
serious adverse outcomes, including cardiovascular diseases
and mortality. Therefore, there is an urgent need for predictive
models to identify high-risk individuals for early intervention.
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e x x x ( 2 0 1 1 ) x x x – x x x
a r t i c l e i n f o
Article history:
Received 10 December 2010
Received in revised form
8 July 2011
Accepted 14 July 2011
Keywords:
Diabetes
Risk
Nomogram
Epidemiology
a b s t r a c t
Objective: To estimate the prevalence and type 2 diabetes, and to develop a prognostic model
for identifying individuals at high risk of undiagnosed type 2 diabetes.
Research design and methods: The study was designed as a cross-sectional investigation with
4314 participants of Thai background, aged between 15 and 85 years (mean age: 48). Fasting
plasma glucose was initially measured, and repeated if the first measurement was more
than 126 mg/dl. Type 2 diabetes was diagnosed using the World Health Organization’s
criteria. Logistic regression model was used to develop prognostic models for men and
women separately. The prognostic performance of the model was assessed by the area
under the receiver operating characteristic curve (AUC) and a nomogram was constructed
from the logistic regression model.
Results: The overall prevalence of type 2 diabetes was 7.4% (n = 125/1693) in men and 3.4%
(n = 98/2621) in women. In either gender, the prevalence increased with age and body mass
index (BMI). Gender, age, BMI and systolic blood pressure (SBP) were independently associ-
ated with type 2 diabetes risk. Based on the estimated parameters of model, a nomogram
was constructed for predicting diabetes separated by gender. The AUC for the model with 3
factors was 0.75.
Conclusions: These data suggest that the combination of age, BMI and systolic blood pressure
could help identify Thai individuals at high risk of undiagnosed diabetes.
Crown Copyright # 2011 Published by Elsevier Ireland Ltd. All
rights reserved.
* Corresponding author. Tel.: +866 043 363664/363746; fax: +866 043 347542.
E-mail address: ekgaluck@yahoo.com (E. Wanothayaroj).
0168-8227/$ – see front matter. Crown Copyright # 2011 Published by Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.diabres.2011.07.019
DIAB-5229; No. of Pages 6
Please cite this article in press as: Pongchaiyakul C, et al. An innovative prognostic model for predicting diabetes risk in Thai population. Diab Res
Clin Pract (2011), doi:10.1016/j.diabres.2011.07.019
C o n t e n t s l i s t s a v a i l a b l e a t S c i e n c e D i r e c t
Diabetes Research
and Clinical Practice
journal homepage: www.elsevier.com/locate/diabres

Page 2

A number of prognostic models have been developed for
predicting the risk of diabetes in Caucasians, but few are
available for Asian populations [10–13]. It is not clear whether
these models can be applied to different populations where
prevalence and risk factors might be different from the
populations where the tools were developed. Moreover, most
previous models were based on the concept of risk stratifica-
tion, in which continuous variables were categorized into
subgroups, whereas our model was based on the concept of
individualization, in which the continuous nature of risk
factors were preserved in order to increase the degree of
uniqueness of an individual.
In this study, we sought to estimate the prevalence of
undiagnosed type 2 diabetes, and to develop simple prognostic
models for identifying high-risk individuals in primary care
setting.
2.
Study design and methods
2.1.
Setting and subjects
The study was designed as a cross-sectional investigation,
with the setting being Khon Kaen, a northeast rural province of
Thailand (about 445 km from Bangkok). The province has a
population of 1.8 million, living mostly in rural areas. The
setting was a health check-up clinic of the Srinagarind
Hospital, which is a teaching hospital of the University of
Khon Kaen. Although the clinic was set up to serve the entire
provincial population, the majority of clients who visit the
clinic lived around the capital city of the province.
The study was formally approved by the Ethics Committee
of Khon Kaen University and written informed consent was
obtained from each individual. From 2003 to 2004, all men and
women who came to the clinic for health check-up were
invited to provide basic clinical information for the study. A
total of 4314 participants (1693 men and 2621 women) were
included in the study. All participants were of Thai back-
ground and were excluded from analysis if they were taking
medications for hypertension, diabetes, or dyslipidemia.
2.2.
Measurements
The participants were invited to meet with a research nurse
who completed a questionnaire and an informed consent
form. Body weight (including light indoor clothing) was
measured using an electronic balance (accuracy 0.1 kg) and
standing height (without shoes) using a stadiometer (nearest
0.1 cm). Body mass index (BMI) was calculated the ratio of
weight in kg divided by height in m2. Systolic blood pressure
(SBP) was measured twice in the left arm and recorded after a
participant had been seated and rest for 5 min. The average of
the two measurements was then used for all analyses.
Diastolic blood pressure (DBP) was recorded at the fifth
Korotkoff sound. Hypertension was defined as a systolic blood
pressure of at least 140 mmHg and/or a diastolic blood
pressure of at least 90 mmHg.
Serum samples were collected in the morning after a
participant had fasted for 12 h prior to the clinic visit. Blood
samples were immediately centrifuged. Measurements in-
cluded fasting plasma glucose (FPG), total cholesterol, trigly-
cerides. Fasting plasma glucose levels were measured the
glucose oxidase method. Serum total cholesterol and triglyc-
eride were measured by enzymatic methods using an
automatic autoanalyzer (Cobas Integra 800; Roche Diagnos-
tics, Mannheim, Germany). The diagnosis of diabetes was
based on the WHO’s criteria using FPG ?126 mg/dl and
repeated within 1 week.
2.3.
Statistical analyses
In order to develop an optimal model for predicting diabetes
risk, the linear logistic regression models were considered.
Given many potential risk factors for diabetes, the number of
possible models can be very large. The Bayesian Model
Average (BMA) method [14] was applied to search for the
most parsimonious models with consistent and maximum
discriminatory power. In terms of model consistency and
accuracy, it has been shown that the BMA approach performed
better than traditional algorithms such as stepwise regression
[15,16], because it can account for model uncertainty in both
predictions and parameter estimates [14,17]. The prognostic
performance of model was assessed by the area under the
receiver operating characteristic curve (AUC). Based on the
test, an optimal model was selected, and parameter estimates
of this model were then used for constructing a nomogram
using the Design library [18]. The bootstrap method was
applied to examine the predictive accuracy of the nomogram
in new settings. In this method, subsamples, each with 150
men and women, were repeatedly resampled (with replace-
ment) from the original entire dataset, and parameter
estimates were computed for each subsample and were used
for the calibration of predictive accuracy [19].
3.
Results
A total of 4314 Thai participants, aged between 15 and 85 years
(mean age: 49.4 and 47 years for men and women, respective-
ly) were included in this study. Men had a heavier, taller, and
higher BMI than women. Both SBP and DBP, FPG and
triglyceride levels were also significantly higher in men than
in women (Table 1).
The prevalence of diabetes of all participants was 5.2%
(n = 223/4314), with 7.4% in men and 3.7% in women. As
expected, diabetic men and women were, on average, older,
heavier, higher BMI, higher SBP and DBP than those without
diabetes. Furthermore, total cholesterol and triglyceride levels
were significantly higher in the both men and women with
diabetes compared to the non-diabetes group (data not
shown). The prevalence of diabetes increased with advancing
age and higher BMI in both men and women (Table 2).
Being man, advancing age, higher BMI and SBP were each
significantly and independently associated with an increased
risk of diabetes (Table 3). The odds of having diabetes was
increased by 1.3 (95% CI: 1.2–1.4) for each 5-year increased in
age; 1.1 (95% CI: 1.07–1.14) for each 1-kg/m2increased in BMI,
and 1.1 (95% CI: 1.03–1.20) for each 10-mmHg increased in SBP.
Moreover, men had greater odds of diabetes than women (OR
1.8; 95% CI: 1.4–2.4). The AUC for the model was 0.753 (95%
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e x x x ( 2 0 1 1 ) x x x – x x x
2
DIAB-5229; No. of Pages 6
Please cite this article in press as: Pongchaiyakul C, et al. An innovative prognostic model for predicting diabetes risk in Thai population. Diab Res
Clin Pract (2011), doi:10.1016/j.diabres.2011.07.019

Page 3

confidence interval: 0.712–0.775). The Hosmer–Lemeshow
calibration test result was 8.04 (P = 0.43).
Based on the parameter estimates of model, the probability
of having diabetes (denoted by p) was estimated by the following
equation: p = 1/(1 + exp(? x)), where x = ?8.857 + 0.0506 ? Age
+ 0.0948 ? BMI + 0.0106 ? SBP for men, and x = ?9.437 + 0.0506
? Age + 0.0948 ? BMI + 0.0106 ? SBP for women. If a probability
of diabetesof atleast 0.10 is consideredhigh-risk, then according
Table 1 – Characteristics of study subjects.
Variable
Women
Men
P-value
N
Age (year)
Weight (kg)
Height (cm)
BMI (kg/m2)
BMI ?25 kg/m2(%)
BMI > 30 kg/m2(%)
Systolic blood
pressure (mmHg)
Diastolic blood
pressure (mmHg)
Hypertension (%)
Fasting blood
glucose (mg/dl)
Total cholesterol
(mg/dl)
Triglyceride (mg/dl)
2621
47.0 ? 10.4
56.9 ? 9.3
155.1 ? 5.6
23.7 ? 3.8
31.2
6.1
117.3 ? 16.4
1693
49.4 ? 11.0
67.4 ? 10.4
165.5 ? 5.8
24.6 ? 3.5
42.3
5.9
121.9 ? 16.3
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.001
<0.0001
76.2 ? 10.9
80.4 ? 11.0
<0.0001
21.7
88.8 ? 20.6
32.3
96.1 ? 31.0
<0.0001
<0.0001
211.2 ? 42.5
212.6 ? 43.6
0.3151
115.7 ? 76.1
163.2 ? 118.9
<0.0001
Notes: Unless otherwise stated, data are shown as mean ? stan-
standard deviation. P-values were derived from t-test (for con-
tinuous data) or Chi-square test (for categorical data) for difference
between men and women.
Table 2 – Prevalence of DM by age, sex and BMI category.
Strata
Women
Men
P-value
Frequencya
%
Frequencya
%
Age group
14–39
40–49
50–59
60+
Total
BMI group
<23
23–25
25–30
>30
4/616
24/1038
38/631
32/336
98/2621
0.7
2.3
6.0
9.5
3.7
7/303
35/622
41/429
42/339
125/1693
2.3
5.6
9.6
12.4
7.4
0.0295
0.0004
0.0315
0.2335
<0.0001
25/1280
23/530
39/650
11/161
1.9
4.3
6.0
6.8
25/547
31/435
52/612
17/99
4.6
7.1
8.5
17.2
0.0017
0.0609
0.0866
0.0090
P-values were derived from Chi-square test (for categorical data)
for difference between men and women.
aFrequency is represented by the number of DM cases over total
sample size.
Fig. 1 – Nomogram for predicting diabetes in Thai men and women. Instruction for usage: Mark an individual’s age on the
‘‘Age’’ axis, and draw a vertical line to the ‘‘Point’’ axis to determine the number of points the individual receives for her
age. Repeat this process for the BMI and SBP. Add the number of points from each predictor. Mark this sum on the ‘‘Total
Points’’ axis, and draw a vertical line down to meet the ‘‘Risk of Diabetes’’ axis, to find the woman’s probability of having
diabetes. Example: Mrs. X, 50 years old, BMI 30 kg/m2and has SBP 160 mmHg; her points for age is approximately 37, her
BMI points is 50; and SBP points is 33 Her total points is therefore 37 + 50 + 33 = 120, and the probability of having diabetes
is around 0.09. In other words, in 100 women like her, 9 will have diabetes.
Table 3 – Risk factors for undiagnosed type 2 diabetes.
Risk factors
Odds ratio and 95%
confidence interval
P-value
Gender (men)
Age (per 5 years)
Body mass index
(per 1 kg/m2)
Systolic blood pressure
(per 10 mmHg)
1.79 (1.35–2.36)
1.29 (1.20–1.38)
1.10 (1.07–1.14)
<0.0001
<0.0001
<0.0001
1.11 (1.03–1.20)
0.009
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e x x x ( 2 0 1 1 ) x x x – x x x
3
DIAB-5229; No. of Pages 6
Please cite this article in press as: Pongchaiyakul C, et al. An innovative prognostic model for predicting diabetes risk in Thai population. Diab Res
Clin Pract (2011), doi:10.1016/j.diabres.2011.07.019

Page 4

to this model, virtually all men age 50+ years, BMI 35+ and with
SBP of 100 or more are in the high-risk group whereas all women
age 50+ years, BMI 40+ and with SBP of 100 or more are in the
high-risk group (Table 4).
Based on the regression coefficients, a nomogram for
predicting diabetes risk was constructed (Fig. 1). For example,
a woman aged 50 years, BMI 30 kg/m2, with SBP 160 mmHg, is
predicted to have a 9% chance of having diabetes. However,
with the same BMI and the same SBP, a man aged 70 years
would have a probability of diabetes of approximately 32%.
4.
Discussion
Although public burden of diabetes in Asia is increasingly
recognized, there have been very few studies of diabetes in
Table 4 – Predicted probability of having diabetes for a given age, BMI and SBP by gender.
Age (year)
BMI (kg/m2)
Systolic blood pressure (mmHg)
100
110
120
130
140
150
160
170
180
Men
30
20
25
30
35
40
20
25
30
35
40
20
25
30
35
40
20
25
30
35
40
20
25
30
35
40
0.01
0.02
0.03
0.05
0.08
0.02
0.03
0.05
0.08
0.12
0.03
0.05
0.08
0.12
0.19
0.05
0.08
0.13
0.19
0.28
0.09
0.13
0.20
0.28
0.39
0.01
0.02
0.03
0.05
0.08
0.02
0.04
0.06
0.09
0.13
0.04
0.06
0.09
0.14
0.20
0.06
0.09
0.14
0.21
0.30
0.10
0.14
0.21
0.30
0.41
0.02
0.02
0.04
0.06
0.09
0.02
0.04
0.06
0.10
0.15
0.04
0.06
0.10
0.15
0.22
0.07
0.10
0.15
0.23
0.32
0.10
0.16
0.23
0.33
0.44
0.02
0.03
0.04
0.07
0.10
0.03
0.04
0.07
0.11
0.16
0.05
0.07
0.11
0.16
0.24
0.07
0.11
0.17
0.25
0.34
0.11
0.17
0.25
0.35
0.46
0.02
0.03
0.05
0.07
0.11
0.03
0.05
0.08
0.12
0.17
0.05
0.08
0.12
0.18
0.26
0.08
0.12
0.18
0.27
0.37
0.13
0.19
0.27
0.37
0.49
0.02
0.03
0.05
0.08
0.12
0.03
0.05
0.08
0.13
0.19
0.06
0.09
0.13
0.19
0.28
0.09
0.13
0.20
0.29
0.39
0.14
0.21
0.29
0.40
0.52
0.02
0.04
0.06
0.09
0.14
0.04
0.06
0.09
0.14
0.21
0.06
0.09
0.14
0.21
0.30
0.10
0.15
0.22
0.31
0.42
0.15
0.22
0.32
0.43
0.54
0.03
0.04
0.06
0.10
0.15
0.04
0.07
0.10
0.15
0.22
0.07
0.10
0.16
0.23
0.32
0.11
0.16
0.24
0.33
0.44
0.17
0.24
0.34
0.45
0.57
0.03
0.04
0.07
0.11
0.16
0.05
0.07
0.11
0.17
0.24
0.07
0.11
0.17
0.25
0.35
0.12
0.18
0.26
0.36
0.47
0.18
0.26
0.36
0.48
0.60
40
50
60
70
Women
30
20
25
30
35
40
20
25
30
35
40
20
25
30
35
40
20
25
30
35
40
20
25
30
35
40
0.01
0.01
0.02
0.03
0.04
0.01
0.02
0.03
0.05
0.07
0.02
0.03
0.05
0.07
0.11
0.03
0.05
0.08
0.12
0.18
0.05
0.08
0.12
0.18
0.26
0.01
0.01
0.02
0.03
0.05
0.01
0.02
0.03
0.05
0.08
0.02
0.03
0.05
0.08
0.12
0.03
0.05
0.08
0.13
0.19
0.06
0.09
0.13
0.20
0.28
0.01
0.01
0.02
0.03
0.05
0.01
0.02
0.04
0.06
0.09
0.02
0.04
0.06
0.09
0.14
0.04
0.06
0.09
0.14
0.21
0.06
0.10
0.14
0.21
0.30
0.01
0.02
0.02
0.04
0.06
0.02
0.02
0.04
0.06
0.10
0.03
0.04
0.06
0.10
0.15
0.04
0.07
0.10
0.15
0.23
0.07
0.10
0.16
0.23
0.33
0.01
0.02
0.03
0.04
0.07
0.02
0.03
0.04
0.07
0.11
0.03
0.05
0.07
0.11
0.16
0.05
0.07
0.11
0.17
0.25
0.07
0.11
0.17
0.25
0.35
0.01
0.02
0.03
0.05
0.07
0.02
0.03
0.05
0.08
0.12
0.03
0.05
0.08
0.12
0.18
0.05
0.08
0.12
0.18
0.27
0.08
0.13
0.19
0.27
0.37
0.01
0.02
0.03
0.05
0.08
0.02
0.03
0.05
0.08
0.13
0.04
0.06
0.09
0.13
0.19
0.06
0.09
0.13
0.20
0.29
0.09
0.14
0.21
0.29
0.40
0.01
0.02
0.04
0.06
0.09
0.02
0.04
0.06
0.09
0.14
0.04
0.06
0.09
0.14
0.21
0.06
0.10
0.15
0.22
0.31
0.10
0.15
0.22
0.32
0.43
0.02
0.03
0.04
0.06
0.10
0.03
0.04
0.07
0.10
0.15
0.04
0.07
0.10
0.16
0.23
0.07
0.11
0.16
0.24
0.33
0.11
0.17
0.24
0.34
0.45
40
50
60
70
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e x x x ( 2 0 1 1 ) x x x – x x x
4
DIAB-5229; No. of Pages 6
Please cite this article in press as: Pongchaiyakul C, et al. An innovative prognostic model for predicting diabetes risk in Thai population. Diab Res
Clin Pract (2011), doi:10.1016/j.diabres.2011.07.019

Page 5

Asian populations. The present study represents an attempt to
assess the prevalence of, and risk factors for, undiagnosed
diabetes in a rural Thai population. We found that approxi-
mately 4% of women and 7% of men in this population had
undiagnosed type 2 diabetes. We also found that individuals at
high risk of diabetes could be identified by a nomogram with 3
simple risk factors, namely, advancing age, high BMI, and high
systolic blood pressure.
The prevalence of undiagnosed diabetes in this population
is relatively lower than previous estimates in other Asian
populations [3–5]. However, it should be noted that the present
estimates were derived from a rural population, whereas
previous studies have largely been based on urban popula-
tions. Rural populations are known to have lower risk of type 2
diabetes than urban populations [20,21]. Taken together, these
data seem to suggest that there is a considerable difference in
the prevalence of diabetes between urban and rural popula-
tions. As in other developing countries, Thailand has gone
through rapid economic transition, with urbanization taking
place in the entire country. In parallel with the economic
change, the population health is in a transitional state with
dramatic shift in lifestyle. These data suggest that diabetes
and obesity and associated diseases are now considered a
public health concern in Thailand, replacing the ‘‘traditional’’
concerns such as under-nutrition and infectious diseases.
The three factors that were identified in this study to be
associated with type 2 diabetes have biologic basis. Although
hypertension has long been known to be a consequence of
type 2 diabetes, recent studies seem to suggest that may be an
antecedent of diabetes. Our cross-sectional finding is consis-
tent with the finding from a prospective study, in which Conen
et al. [22] showed that baseline blood pressure was a strong
predictor of type 2 diabetes. A prospective relationship
between hypertension and diabetes may have a biologic basis.
Increased central sympathetic drive is hypothesized to cause
both hypertension, obesity, and in particular central obesity
[23]. There are long-term studies linking early measures of
increased sympathetic activity and blood pressure response to
later obesity and diabetes [24,25]. Nevertheless, whether
hypertension is a risk factor for, or a consequence of, diabetes
is probably less relevant to the purpose of identifying high-risk
individuals. A somewhat similar model has also been
developed for the Vietnamese population [3], which uses
waist-to-hip ratio and systolic blood pressure as predictors. In
our study, we did not measure waist-to-hip ratio, but instead
used body mass index as a measure of fatness. In the
Vietnamese study, the prevalence of diabetes was 10–13%,
which is considerably higher than our estimates. Therefore
the predicted risk of diabetes by the Ta et al.’s model is higher
than the present model. However, there is no significant
difference in AUC values between the present model and Ta
et al.’s model.
Obesity is strongly related to type 2 diabetes, especially
central fat, has long been recognized as a risk factor for type 2
diabetes [26]. In this study, we found that obesity, as
represented by increased BMI, is a risk factor for diabetes
independent of age and systolic blood pressure. While it could
be argued that BMI is a less-than-optimal measure of central
obesity, which is known to be a better predictor of diabetes,
BMI is a measure that can be taken in any busy primary care
setting. Moreover, BMI is also highly correlated with waist to
hip ratio [27]. Taken together, result of this study confirms that
BMI can be used as a simple and robust risk factor to identify
individuals at high risk of type 2 diabetes.
The present study’s findings should be interpreted within
the context of strengths and potential weaknesses. A major
strength of this study is that the diagnosis of DM was based on
repeated fasting plasma glucose measurement, which mini-
mized any possibility of misdiagnosis. The sample size was
large enough to ensure that the study could assess modest
effect sizes otherwise not possible in smaller studies.
However, since the study was undertaken in a clinic setting,
the prevalence of diabetes could have been over-estimated the
true prevalence in rural populations. Since the setting was an
Asian rural province, these results may not be generalizable to
urban populations or non-Asian populations among whom
lifestyle and demographic structure could be different from
the present population. It should also be noted that the
present nomogram as well as Ta et al.’s model [3] were
developed based on a cross-sectional study, and the predicted
risk reflects a long-term risk, not a time-specific risk which
would require a prospective study.
There are few prognostic models for predicting diabetes
risk in Asians. Most previous models [3,10–13] used risk factors
such age, sex, family history of diabetes, indicators of
hypertension (e.g., systolic BP, use of anti-hypertensive drugs),
and measures of obesity (e.g., BMI, waist circumference). The
area under the ROC curve of these models varied between 0.54
and 0.83. In this study, the model with 3 factors (e.g., age,
systolic blood pressure and BMI) yielded an AUC of 0.70 for
women and 0.77 for men which are quite comparable with
previous findings. There are, however, a number of differences
between our model and previous models. Most previous
models, such as the Gao et al.’s model [10], were based on the
concept of risk stratification, in which continuous variables
were categorized into subgroups, whereas our model was
based on the concept of individualization, in which the
continuous nature of risk factors were preserved in order to
increase the degree of uniqueness of an individual. With
continuous variables, the more risk factors are considered, the
greater likelihood of uniqueness of an individual’s profile can
be defined. Therefore, by modeling risk factors in their
continuous scale the present models can be uniquely tailored
to an individual.
Assessment of absolute risk of diabetes can have practical
application in treatment allocation, risk communication, and
decision making. Since unselective screening for diabetes is
not cost-effective, the nomogram presented here can be used
as an initial screening tool and then followed by further
confirmatory or diagnostic tests, including assessment of
glycemia, HbA1c, lipids and family history. However, the
predicted risk is a continuous probabilistic variable ranging
from 0 to 1, and this raises the issue of selecting an optimal
cut-off predicted probability to classify an individual into
diabetic or non-diabetic group. This is not easy, because the
cut-off value depends on the complex risk-benefit consider-
ation, and perhaps more importantly, an individual’s percep-
tion of risk, which is beyond the scope of the present study.
The risk threshold by which an individual should be selected
for further testing needs further research to maximize the
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e x x x ( 2 0 1 1 ) x x x – x x x
5
DIAB-5229; No. of Pages 6
Please cite this article in press as: Pongchaiyakul C, et al. An innovative prognostic model for predicting diabetes risk in Thai population. Diab Res
Clin Pract (2011), doi:10.1016/j.diabres.2011.07.019