Body mass index and type 2 diabetes in Thai adults: defining risk thresholds and population impacts

In recent decades, many low and middle-income countries (LMICs) have achieved substantial economic growth and this has led to increased urbanization, the adoption of new health behaviours (e.g. smoking and a ‘western’ diet), and an epidemiological shift from infectious diseases towards non-communicable diseases including type 2 diabetes mellitus (T2DM). These changes in health-risk behaviours, environment, and health outcomes, termed the ‘health-risk transition’, have been occurring in Thailand and T2DM now affects over four million adults [1].

Overweight and obesity significantly increase the risk of T2DM [2, 3] and their prevalence is increasing in Thailand [4]. Of concern, the association between body mass index (BMI) and T2DM risk is modified by ethnicity, with Asian populations having an increased risk of T2DM at BMI levels considered to be in the healthy weight range for Caucasian populations (<25 kg/m²) [5]. Accordingly, in 2000 the World Health Organization (WHO) recommended that lower BMI cut-off points should be used to define overweight (23- < 25 kg/m²) and obesity (>25 kg/m²) in Asian populations [6]. These recommendations were based on the few prevalence studies at the time [7‐10]. In 2004, the WHO released an additional statement indicating that a range of BMI cut-off points may be necessary for guiding public health action in different Asian populations since the available data were inconsistent [11].

Since publication of these two recommendations, many studies assessed the validity of lower Asian-specific BMI cut-points for assessing diabetes, cardiovascular and mortality risk. Results were inconclusive, potentially due to confounders or exclusion of women [12], cross-sectional design [5, 13‐15] or small samples of Asian participants [5, 16]. The prevalence of overweight and obesity are projected to continue to rise in Thailand. Understanding the optimal Southeast Asian BMI cut-off values associated with T2DM risk and the contribution of excess weight to the development of T2DM will have major implications for public health planning in Thailand. Therefore we examined the BMI threshold associated with T2DM risk in a large prospective cohort of adult Thais. We also estimated population attributable fractions and number of T2DM cases attributable to overweight and obesity in Thailand.

Of the 87,151 initial TCS participants, 60,569 were followed-up in 2009. Of these, 706 had prevalent diabetes in 2005 and 28 did not have a response for the diabetes question in 2009 and were excluded. Of the remaining 59,835, a total of 39,507 eligible participants (without missing diabetes responses) were followed-up in 2013. Of these, 486 did not have credible height or weight data. The final study sample included 39,021 participants of whom 688 reported a new diagnosis of diabetes (see flow chart in Additional file 1: Figure S1).

The baseline characteristics of participants by sex and diabetes status are shown in Table 1. Among both men and women, T2DM incidence increased with increasing age, BMI, and income (p < 0.001). T2DM incidence was double among those who lived in a city rather than a rural area as a child (p < 0.001). Among women, T2DM incidence was highest in those who consumed sugar-sweetened beverages (SSBs) daily (p < 0.01). Among men, T2DM incidence was highest in current smokers and those who consumed alcohol regularly (p < 0.001).

The age and multivariable adjusted sex-specific associations between baseline BMI and T2DM incidence by 2013 are shown in Table 2 and in Fig. 1. A BMI of 20.75- < 23.00 (compared to 18.5- < 20.75) was associated with higher T2DM risk in women (OR =3.0, 95% CI 1.6–5.7). In men, a BMI of 23 < 25 (compared to 18.5- < 20.75) was associated with higher T2DM risk (OR = 2.3, 95% CI 1.3–3.9). The cubic spline non-linear modelling of BMI’s effect on T2DM showed that the points of inflection where the association became statistically significant compared to a reference of 20.00 kg/m² for men and women were 21.60 (OR = 1.27, 95% CI 1.00–1.61) and 20.03 (OR = 1.02, 95% CI 1.02–1.03), respectively. In women, BMI levels <20 kg/m² appeared to be inversely associated with risk suggesting thresholds might be even lower for women, however, few participants fell into these categories thus definite conclusions are difficult to draw. At a BMI of 25.0 kg/m², T2DM risk was exponentially increased in both men (OR = 4.3, 95% CI 3.0–6.1) and women (OR = 11.6, 95% CI 6.8–19.8.), respectively. These findings suggest that the exponential relationship between BMI and T2DM is greater among women than men and that excess adiposity in Southeast Asian women has major implications for T2DM risk even at BMI levels considered to be in the healthy weight range for Asian populations (<23 kg/m²) (Fig. 1).

Figure 2 shows the sex-specific association between baseline BMI and eight-year T2DM incidence stratified by age and urbanization status. Among men, while the association between obesity (BMI > 25) and risk of T2DM was higher among participants aged <30 (OR = 15.7, 95% CI 7.9–31.4) than those aged 40 and older (OR = 4.60, 95% CI 2.98–7.11) the interaction term between BMI and age was not statistically significant (p = 0.47). Among women, there was some indication that the BMI-T2DM association was modified by area of residence as a child such that the association between high BMI categories (> 23 and >25) and T2DM risk was higher in participants who lived in an urban area of residence as a child (BMI > 23: OR = 6.8, 95% CI 3.1–15.0; > 25 OR = 24.6, 95% CI 12.7–47.7) than those who lived in a rural area of residence as a child BMI > 23: OR = 3.3, 95% 1.9–5.5; BMI > 25: OR = 10.8, 95% 7.1–16.3), however, the interaction term between BMI and urbanization status was not statistically significant at the 5% level (p = 0.08).

Our prospective nationwide study of Thai adults provides further evidence that T2DM risk is increased at BMI levels below 25 kg/m² in Asian populations and that the use of lower Asia-specific BMI cut-points are necessary for defining T2DM risk in Asian adults, particularly in women. We found that T2DM risk was associated with a BMI of 22 kg/m² and 20 kg/m² in men and women, respectively. We also found that for the same BMI level the association between BMI and T2DM risk was higher in women and that their T2DM risk was already increased at BMI levels currently considered in the ‘healthy range’ for Caucasian and Asian populations. Over 60% of all T2DM cases occurring in this cohort could be attributed to overweight and obesity and our results suggest that a hypothetical 5% reduction in the prevalence of obesity could result in almost 13,000 fewer cases of T2DM annually in the national Thai population.

There are several potential limitations which should be considered when interpreting our findings. All data on height, weight and diabetes diagnoses were ascertained using self-report. Accordingly, there may be some misclassification error which should be considered when interpreting these findings. However, validation studies of self-reported diabetes [18] and self-reported weight and height [30] in this cohort have shown that the accuracy of these self-reported measures is high. Another potential issue is attrition. Over the eight-year follow-up, 50% of the baseline cohort was lost to follow-up. Slight differential attrition was noted by body size with a higher retention rate for participants with a larger weight at baseline. However, previous work with this cohort showed that ORs for the association between BMI and T2DM incidence in the first 4 years (70% retention of baseline cohort) were similar to those from the total eight-year follow-up, indicating that eight-year ORs are likely to be generalizable [21]. Moreover, non-differential attrition was noted by sex, dietary intake, alcohol intake, and area of residence indicating that these variables would unlikely affect BMI-T2DM risk effect estimates (see Additional file 2: Table S1).

An additional issue to consider is the precision of the PAF and PIF estimates. These measures are dependent on the accuracy and the magnitude of the ORs and the prevalence estimates being used. Accordingly, there is potential for error in these estimates due to variations in the accuracy in the self-report of weight and height, differences in the prevalence of overweight and obesity between our cohort and the national Thai population, as well as any risk estimates affected by attrition in this cohort. There is no agreed method for calculating confidence intervals on composite measures such as PAF and PIF estimates. Accordingly, these were not calculated. Nevertheless, our PIF and PAF estimates highlight the potential magnitude of the effects of overweight and obesity on T2DM incidence in Thai adults.

Important strengths of this study include the large sample of Thai adults and our nationwide coverage. Moreover, to our knowledge this is the largest prospective study to assess the BMI cut-points associated with T2DM risk and the number of T2DM cases attributed to overweight and obesity in adults living in Southeast Asia.

This study found that T2DM risk is increased at BMI levels considered to be in the ‘normal’ range for Caucasian populations (<25 kg/m²). Our findings are consistent with previous studies that recommend using a BMI cut-off between 21 and 24 kg/m² to define overweight and obesity in Asian populations based on diabetes and cardiovascular risk [5, 13‐15, 31‐33]. However, studies of body size and mortality could shed light on appropriate cut-offs for BMI on health more generally. For example, a few mortality studies conducted in different Asian populations did not find evidence that mortality risk was increased at BMI levels <25 kg/m² [34]. However, these studies were unable to adjust for potential confounding effects of factors such as infectious diseases and the prevalence of smoking in these different populations. Adjusting for related factors in this study had minimal influence on the BMI-T2DM relationship. However, these factors might modulate more of the BMI-mortality association than the BMI-T2DM relationship [35].

A likely explanation for the increased risk of T2DM at these low levels of BMI is body composition. Evidence suggests that the relationship between BMI and body fat differs by ethnicity, with Asian populations having higher proportions of body fat compared to muscle mass than Caucasian populations when matched by age and BMI [36]. Ethnicity related differences in body fat composition and distribution may be associated with epigenetic programming [37]. Babies that develop in an undernourished intrauterine environment have been shown to experience ‘catch up’ growth later in life that has been associated with preservation of adipose tissue, obesity, and increased insulin resistance [38]. As well, children growing up with under-nutrition and frequent infection who survive to adulthood have low attained height, and a life-long threat of adult obesity, which is widespread in Thailand today [39]. Many Asian populations have experienced generations of under-nutrition prior to experiencing the accelerated nutrition transition currently underway in much of Asia. Accordingly, some Asian populations may be more predisposed to storing visceral fat or are more insulin resistant at lower levels of adiposity [36, 40].

Effects of epigenetic programming may be particularly intensified when there is a strong mismatch between the intrauterine and early life environments [41], which may explain the potentially increased association between obesity and T2DM incidence noted in TCS females who lived in an urban area of residence during childhood. Living in an urban area during childhood has been shown to be associated with increased exposure to an energy-dense western diet, reduced physical activity and a higher prevalence of obesity during adulthood [42]. Therefore, unlike children who are raised in a rural area of residence and may have low exposure to the ‘western diet’, those who are raised in an urban area of residence may have pro-longed exposure to an obesogenic environment and consequently an increased risk of T2DM later in life.

As in previous studies, this study found that for the same BMI T2DM risk was higher in women [22]. Evidence suggests that this sex-specific association may be related to differences in body composition and hormones between men and women. For instance, men have higher concentrations of testosterone than women and testosterone has been shown to inversely associate with adiposity and insulin resistance in men [43]. This effect may relate to testosterone’s role in increasing lean body mass and decreasing inflammatory cytokines, which can increase insulin resistance [43]. These findings highlight the need for public health interventions to target the promotion of healthy weight differently in Asian men and women.

Our findings show that the public health impact of preventing or reducing the prevalence of obesity in Thailand could be profound. Shifting as little as 5% of the population with obesity to a lower BMI of 23- < 25 kg/m² could lead to a substantial decrease in the number of T2DM cases occurring in the Thai population; and this benefit could be even greater than we have calculated considering the high and increasing prevalence of obesity, particularly in women, in the national Thai population [27].

The findings from this prospective study of Thai adults suggest that using lower BMI cut-points is necessary for defining T2DM risk in Southeast Asian populations. Our findings suggest that T2DM risk is already increased at BMI levels <23 kg/m². Therefore, public health action and response may be required at lower BMI levels to help curb the T2DM epidemic currently underway in Southeast Asia. Further research is required to confirm these findings in different Southeast Asian populations.