Prevalence of type 2 diabetes mellitus and impaired fasting glucose, and their associated lifestyle factors among teachers in the CLUSTer cohort

CLUSTer cohort

CLUSTer is a cohort study of Malaysian teachers with the primary goal of understanding the clustering of lifestyle risk factors on teachers’ health and well-being, including type 2 diabetes (Moy et al., 2014). Using a multi-staged sampling technique, approximately 14,500 primary and secondary schoolteachers were recruited between 2013 and 2015. They were recruited from 534 schools in six study areas (Penang, Selangor, Terengganu, Malacca, Johor, and Kuala Lumpur) in Peninsular Malaysia.

Study design

This is a cross-sectional study from the CLUSTer cohort. During baseline recruitment, all eligible teachers free from psychiatric illnesses were asked to answer a questionnaire containing demographic information, health status, and lifestyle. The investigators also collected anthropometric parameters such as waist circumference. Teachers’ blood pressure and fasting blood glucose levels were measured using a portable blood pressure device and a laboratory venous glucose test, respectively.

Study variables

The demographic variables were age, gender, ethnicity, education level, marital status, and family history of T2DM. Lifestyle variables included daily fruit and vegetable consumption (servings per day), physical activity measured by the International Physical Activity Questionnaire short version (IPAQ-short), smoking status, alcohol consumption, waist circumference, sitting duration, sleep duration, and acute mental health status (depression, anxiety, and stress rated using the Depression, Anxiety, and Stress Scale 21 questions version (DASS-21)). According to the Ministry of Health guidelines, at least three daily vegetable and two fruit servings are recommended (Ministry of Health Malaysia, 2020). As for physical activity, at least five days of moderate-intensity physical activity (30 min) is required to maintain health (Ministry of Health Malaysia, 2020).

The outcomes of this study included self-reported T2DM, undiagnosed T2DM and IFG status. Teachers’ blood samples were collected and stored in a −80 Celsius freezer before being analysed for fasting blood glucose (FBG) at the University of Malaya Medical Centre laboratory. To make the results comparable with those of the NHMS survey, the authors adopted the outcome definition by the Institute of Public Health 2015 for undiagnosed T2DM and IFG. Participants who previously did not report having T2DM but had FBG laboratory readings ≥ 6.1 mmol/L were considered to have undiagnosed T2DM. Conversely, teachers who were free of T2DM yet their FBG levels exceeded 5.6 mmol/L but were below 6.1 mmol/L were considered to be co-living with IFG (Institute for Public Health, 2015).

The operational definitions of all the variables are described in Table S1.

Statistical analysis

In descriptive analysis, a complex analysis was conducted using a “survey” package in R software (Lumley, 2004). A complex analysis incorporated an assigned weight throughout the data analysis (Lohr, 1999). The complex sample weight was computed based on the multiplication of the two-staged sampling units, known as school and teacher (Fig. S1). This total weight was applied to compute the weighted means with standard deviation for all scaled variables whereas frequencies with weighted percentages were used for all categorical variables. The weight was used to account for the multi-level data, including the unequal probability of selection and non-respondents. A matched age group comparison was performed between teachers and the Malaysian general population on the prevalence of T2DM, undiagnosed DM, and IFG. Briefly, the Malaysian general population in the NHMS survey had approximately equal proportions of males and females. Approximately 60% of the investigated Malaysian population was married and two-thirds of them were employed. However, ethnicity information was not disclosed. Further details of the survey can be assessed from the Institute for Public Health (2015).

Teachers who answered the majority of the survey questionnaires were included in the inferential analysis. A multiple imputation chained equation (MICE) technique was applied to deal with all missing data (Van Buuren et al., 2021). In bivariate and multivariable factor association analyses, univariable and multivariable logistic regressions were performed using the same “survey” package. All non-lifestyle covariates with p value <0.25 were included in the multivariable model, along with all lifestyle factors. The multicollinearity test was used to detect high correlations between variables included in the multivariable regression model. Variables had variance inflation factors below 5.0 suggesting no multicollinearity issue (Heiberger & Holland, 2015). Lifestyle variables with p values <0.05 determined via multivariable analysis were considered significantly associated with T2DM and IFG, respectively. Additional analyses on interaction were conducted, and significant findings were incorporated into the final models. A “ggeffect” package in R (Lüdecke, 2018) was used to illustrate the multivariable regression findings. Finally, additional stratified analyses were conducted, both descriptively and inferentially among male and female teachers.

Ethical clearance

This study obtained ethical approval from University of Malaya Medical Centre Medical Research Ethics Committee (UMMC MREC) (MREC ID: 950.1). Hardcopies of written consent were obtained from all participants.

Teacher’s characteristics, prevalence of T2DM and IFG among the teachers

A total of 14,144 teachers out of the initial sample of 14,427 teachers were included in the analysis after excluding duplicates. The enrolled teachers were around 40 years old, and four-fifths were female. About two-thirds (n = 10773 ,77%-weighted) of the sample were of Malay ethnicity, followed by Chinese (n = 2291, 14%-weighted), Indian (n = 979, 8%-weighted), and other races (n = 101, 1%-weighted). Most of them had at least a bachelor’s degree and above, and at least 80% were married. Approximately half of the teachers reported a family history of T2DM.

Teachers’ lifestyles and daily fruit and vegetable servings were below the recommended serving size, and only 2.3% of teachers had adequate fruit (two) and vegetable (three) servings daily. Similarly, teachers also lacked sleep, particularly on weekdays, where they only slept for an average of six hours compared to seven hours on weekends. Physical activity and sitting duration varied among the teachers. The metabolic equivalent of tasks (METs), in minutes per week, among teachers had a wide range, with a weighted average of 1,344.58, and a dispersion range of 2,908.50. Likewise, the weighted average sitting duration ranged from one to six hours daily. Only 3.7% and 4.0% of teachers were ever smokers and alcohol consumers, respectively. As for weight management, teachers’ documented waist circumferences ranged from 70 to 90 cm, and 52% of them had abdominal obesity, after adjusting for sex. Finally, for acute mental health status, the average depression, anxiety, and stress scores were six, eight, and 10, respectively, but all scores had a standard deviation of six. The high variations in all acute mental health scores indicated that teachers’ mental health status also varied.

The prevalence of known T2DM among teachers was 4.1% (95% CI: 3.7, 4.5),  whereas the prevalence of undiagnosed T2DM was as high as 5.4% (95% CI: 4.9, 6.0). Among the non-diabetic teachers (n = 12,757), about 5.6% (95% CI: 5.1, 6.2) experienced IFG during the baseline survey. The details of the teachers’ characteristics are summarised in Table 1.

A comparison between the prevalence of known T2DM, undiagnosed DM, and IFG between teachers and the general population from the NHMS 2015 is shown in Fig. 1. The comparison based on age ranged from 25 to 59 years to match the general service age of teachers in schools. Comparatively, the weighted prevalence of known and undiagnosed T2DM among teachers was lower than that of the general population. However, when comparing these two diseases status among teachers, the prevalence of undiagnosed T2DM was higher than that of known T2DM across the observed age groups, particularly from age group 35 onwards. In addition , the prevalence of IFG among teachers was higher than that of the general population, especially among those aged 45 to 59 years.

In the inferential analysis, 11,412 out of 14,144 teachers were included, excluding those who had not answered the majority of the survey questionnaires.

Lifestyle factors associated with T2DM and IFG among teachers

Bivariate analyses of factors associated with T2DM and IFG levels are shown in Table S2. Factors associated with T2DM and IFG, from bivariate to multivariable regression plots, are described in separate subsections.

T2DM and its associated lifestyle factors

Of all variables included in the bivariate analyses associated with T2DM, all sociodemographic variables including age, sex, ethnicity, education level, marital status, and family history of T2DM were found to be significant at p value <0.25. The significant lifestyle variables were waist circumference, sitting duration, sleep duration, depression, anxiety, and stress, with p values at the 0.25 level. No multicollinearity issue (variance inflation factor <5) was found among the 12 variables included in the multivariable logistic model.

Multivariable analysis showed that, the only lifestyle variable that remained significantly associated (p value <0.05) with T2DM was waist circumference after adjusting for all covariates. This finding concurs with the literature indicating that waist circumference is indeed a good indicator to predict the risk of developing T2DM (Siren, Eriksson & Vanhanen, 2012). Additionally, an ad hoc interaction analysis indicated that there was a significant interaction between age, history of T2DM, and waist circumference. The final model showed that every one unit increase in waist circumference increased the odds of acquiring T2DM by 14%, with 95% CIs from 1.08 to 1.20, p value <0.001 (Table S3).

Two multivariable regressions were plotted to display the interaction effect of age and family history of T2DM on waist circumference in relation to T2DM among teachers (Fig. 2). From the left figure, denoted as A, the risk of developing T2DM differed with waist circumference (70 cm onwards) by age group. However, the risk of T2DM increased exponentially when waist circumference exceeded 80 cm, especially among those aged 45 and 55 years old. Teachers whose waist circumference exceeded 110 cm had an approximately equal risk of developing T2DM regardless of their age. When further stratified by family history of T2DM, there was an exponential increment in T2DM risk as teachers’ waist circumference increased, where those with a family history of T2DM had a consistently higher risk than their counterparts. However, when the waist circumference exceeded 100 cm, there was no obvious difference in the risk of acquiring T2DM regardless of the family history of the disease.

Impaired fasting glucose (IFG) and its associated lifestyle factors

In bivariate analysis (Table S2), covariates such as age, sex, ethnicity, marital status, and family history of T2DM were significantly associated with IFG. All lifestyle factors were found to be significant at p value <0.25 level related to the outcome, with the exception of sitting duration and acute depression and anxiety levels. Multicollinearity was not observed for any of the 12 variables included in the multivariable model.

Multivariable logistic regression (Table S3) showed that only waist circumference and physical activity were significantly (p value <0.05) associated with IFG after accounting for the effect of other covariates. Hence, the final adjusted odds ratio of waist circumference and physical activity was as follows: One centimetre increase in waist circumference increased the odds of IFG by 10.0%, with a 95% confidence interval of 1.05 to 1.15. Conversely, teachers with moderate (aOR = 0.71, 95% CI: 0.52, 0.98) or high (aOR = 0.56, 95% CI: 0.40, 0.80) physical activity were protective against IFG compared to those with low physical activity.

A multivariable regression plot on waist circumference and physical activity in relation to IFG is illustrated in Fig. 3. Three significant interaction terms, namely age with waist circumference, ethnicity with waist circumference, and ethnicity with physical activity, were included. Figure 3A shows the effect of age on waist circumference plotted against the risk of IFG. Similar to T2DM, IFG risk increased exponentially with waist circumference ranging from 60 cm to 120 cm, where older teachers tended to have the highest risk across all waist circumferences. However, the risk was approximately the same when waist circumference was approximately 100 cm.

When observing waist circumference by ethnicity (Fig. 3B), there was a trend of increasing risk among Malay and Chinese teachers when their waist circumference increased. There was no exponential increase among Indian teachers in relation to waist circumference, even though their risk was the highest even with a normal waist circumference below 80 cm. A large confidence interval for Indian teachers indicates a low sample size. Hence, these results must be interpreted carefully.

On the other hand, the plot of waist circumference by race against the outcome in Fig. 3C shows that there was a decreasing risk of impaired glucose from low to high physical activity. This trend is consistent for all races except Chinese. Moderately active Chinese teachers tended to have a slightly higher point estimate of the risk of acquiring impaired glucose than low- and high-active teachers, but the confidence intervals for all physical activity groups among Chinese teachers overlapped. This indicated that the effect of physical activity on the risk of acquiring IFG was minimal among Chinese individuals compared to other races.

When stratified by gender, most sociodemographic, lifestyles, mental health status variables were comparable between male and female teachers. The average physical activity of the male teachers was slightly higher than that of the female teachers. Male teachers tended to smoke and consume more alcohol than their female counterparts. As for the outcomes (known T2DM, undiagnosed T2DM, and IFG), male teachers recorded higher percentages of all these outcomes compared to female teachers. All the details of the weighted descriptive analysis are shown in Table S4. Additional regression models stratified by gender showed that the plots were similar (Figs. S2 & S3). This indicates that the stratified analysis by gender did not provide additional information, probably due to the imbalance in the numbers of male and female teachers in the CLUSTer cohort.