Background
The co-morbidity of diabetes mellitus (DM) and pulmonary tuberculosis (PTB) represents a double burden with significant public health implications [
1,
2]. Globally, although the incidence of PTB is slowly decreasing, an increase is seen in the proportion of PTB cases with DM. Moreover, the prevalence of DM steadily increases, especially in developing countries where PTB is highly endemic [
3,
4]. Baker conducted a meta-analysis of observational studies about the association between DM and TB disease outcomes, and showed that DM was associated with worse treatment outcomes, and increased the risk of failure, death, and relapse among patients with PTB. Therefore, more attention should be paid to the control and prevention of PTB patients with DM [
5].
Also, patients with impaired fasting glucose (IFG) are more prone to progress to the DM stage -approximately 8.8% per year in China [
6,
7]- in the absence of interventional measures. In developing countries, the nutritional status has improved along with the economic growth. A number of studies have provided strong evidence of an association between patients who are overweight or obese and risk of DM [
8‐
10]. The effect of BMI on DM has primarily focused on patients without other diseases. A low BMI is a significant individual risk factor for development of recent active TB [
11], and patients who are overweight have a decreased incidence of TB [
12]. Studies involving the association between BMI and DM or IFG in PTB patients are limited [
13], and no data are available on using BMI cut-offs to predict DM or IFG in PTB patients.
We designed this cross-sectional study using primary data to explore the association of BMI with DM or IFG in PTB patients in China, and to determine the optimal BMI cut off value for prediction of DM or IFG in Chinese adult patients with PTB.
Methods
Study population
A total of 3,505 PTB patients, 18-45 years of age, were selected from counties of Linyi in Shandong province, China. All the PTB patients were newly-diagnosed and registered for Directly Observed Treatment, Short Course (DOTS) in PTB clinics between September 2010 and March 2013, and also were diagnosed in the PTB clinic of each county by X-ray and sputum smear examination. All participants have written informed consent.
Detection of indices
DM was diagnosed by World Health Organization criteria based on the fasting plasma glucose (FPG) level (World Health Organization, 1999). The diagnosis of IFG was based on criteria for the classification of glucose tolerance based on the FPG level, which was defined as a FPG range from 5.6–6.9 mmol/L (American Diabetes Association, 2014). The participants fasted for 8–10 h before the blood testing.
Basic information and anthropometric indices of all patients were collected, such as age, gender, educational level, weight, and height. The BMI (kg/m
2) was calculated using the following formula: BMI = weight (kg)/height (m)
2. Underweight, normal weight, overweight, and obese categories were defined using the modified criteria for the Chinese population; the BMI cut-off values were 18.5 kg/m
2, 24.0 kg/m
2, and 28.0 kg/m
2, respectively [
14].
Blood lipids were also determined in fasting venous blood samples using an automatic biochemical analyzer in each clinic, and included triglycerides (TG), total cholesterol (TCHO), and high-density lipoprotein (HDL) levels.
Statistical analysis
SPSS version 17.0 (SPSS, Inc. Chicago, IL) was used for statistical analyses. The characteristics of PTB patients in the DM, IFG and normal FPG groups and inter-quartile range (IQR) of BMI were compared and analyzed. The mean and standard deviation for continuous variables, such as age, BMI, and blood lipid content, and proportions for categorical variables, including the prevalence of DM or IFG, the percentage of age group, gender, and educational level, are reported. One-way analysis of variance was used to test continuous variables. A chi-square test was used to compare categorical variables. Multinomial logistic regression analyses were performed, and the variables for inclusion in the multivariate model were chosen based on plausibility and variables with P values <0.05 on univariate analysis were entered into the multivariate analysis. Receiver operating characteristic (ROC) analysis was used to predict the ability of diagnosis of BMI for DM and IFG in PTB patients. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), negative likelihood ratio (NLR) and Youden’s index were used for ROC analysis. Youden’s index is a single statistic that captures the performance of a dichotomous diagnostic test. It was defined for all points of the ROC curve, and the maximum value of the index was used as a criterion for selecting the optimum cut-off point. Youden’s index = Sensitivity + Specificity – 1. All variables were checked for collinearity. Independent variables included age (divided into two categories [18–30 and 31–45 years]), gender, educational level, BMI (divided into three categories [<18.5 kg/m2 as underweight, 18.5–23.9 kg/m2 as normal weight, and ≥ 24.0 kg/m2 as overweight and obese]). A P value <0.05 was considered statistically significant.
Discussion
In this survey of Chinese adults (18–45 years of age), we showed that overweight and obesity were positively associated with DM and IFG in PTB patients. The prevalence of DM and IFG was higher when the BMI was ≥ 23.41 kg/m2 (Q4). Until now the optimal BMI cut off point for IFG and DM screening among PTB patients has not been estimated in China. Based on the current study, the optimal BMI cut-offs for predicting DM and IFG were 22.22 kg/m2 and 22.34 kg/m2, respectively.
After adjusted for the associated factors, including young age, high educational level, and excess weight, BMI was still a risk factor for DM or IFG in PTB patients. In a previous cohort study [
15], the incidence of DM was mainly associated with overweight and obesity in China (28.3% among men and 31.3% among women). However, a inverse association and dose–response relationship between the incidence of TB and BMI up to 30.0 kg/m
2 was shown in a previous meta-analysis [
16]. The relative risk of TB in underweight patients (BMI < 18.5 kg/m
2) compared to normal weight patients was estimated to be 3.2 (95% CI = 3.1–3.3) [
17]. Meanwhile, an association between abnormal weight and elevated glucose (DM and IFG) may still exist. A survey administered in 49 developing countries showed that not only overweight, but also underweight might be involved in the pathogenesis of diabetes [
18]. Thus, overweight and obesity in PTB patients should not be overlooked, whether or not it regards only a small proportion of PTB patients.
In the model adjusted by all of the relevant risk factors, a BMI in the 19.82–21.45 kg/m
2 range was negatively associated with DM in PTB patients, and a BMI > 23.41 kg/m
2 was positively associated with IFG. Moreover, for a BMI ≥ 23.41 kg/m
2
, the prevalence of DM and IFG in PTB patients was much higher in our study. A cross-sectional study carried out in Spain showed that the prevalence of DM in overweight or obese patients was 23.6%, and the higher the BMI, the higher the prevalence of DM [
19]. In the current study, some differences in the BMI associations still existed. Specifically, when the BMI was < 19.82 kg/m
2 (Q1), the prevalence of DM in PTB patients presented slightly decreased, although the differences were not significant. This finding might be attributable to the impact of PTB. Abundant epidemiologic evidence has indicated that low BMI is a risk factor for PTB, and the biological mechanism underlying the relationship has been well-described by induced impairment of cellular immunity [
20]. In addition, underweight patients might also be involved in the pathogenesis of DM through direct and indirect mechanisms [
21]. Therefore, we should pay more attention to the factors associated with underweight, overweight, and obesity, who favor high prevalence of DM and IFG in PTB patients.
We were able to establish BMI cut-off values for DM and IFG in PTB patients in the current study. Although BMI might not be the strongest risk factor to screen undiagnosed DM compared with age, waist circumference, and a family history of DM [
22], it is an easy to acquire anthropometric index for the prediction of DM. With respect to adults, the optimal BMI cut-off value for predicting the presence of DM was 23.3 kg/m
2 for men and 24.0 kg/m
2 for women [
23]. A cross-sectional study conducted in northeast Chinese adults showed that combined with the waist-to-height ratio, a maximal BMI ≥ 23.0 kg/m
2 for DM and ≥ 22.0 kg/m
2 for glucose tolerance abnormalities were better anthropometric indices [
24]. Also, the San Antonio Heart Cohort Study reported that BMI and waist circumference had equal power in predicting development of metabolic syndrome in non-Hispanic Whites and Mexican Americans [
25]; thus, BMI is possibly an appropriate predictor for pathoglycemia. Of note, the optimal BMI cut-off points in our study were lower than other studies, which might because PTB patients were more likely to have a low BMI [
11], and in young patients (18–45 years old), a lower BMI was more common [
26].
There were some limitations in our study. First, there were some limitations in the generalizability of our findings: the age range and only Chinese participants. The age range in our survey was 18–45 years, which comprised young and middle-aged adults in China. Because there were no elderly in our study, which comprise a high-risk group and have a higher prevalence than young and middle-aged adults [
27], the prevalence of DM and IFG would be low. And then, the results of this study were applied to Chinese population, so there might be different results in other countries. Second, Youden’s index was low, which meant the false-positive and false-negative values were high for BMI in predicting DM and IFG in PTB patients.
Acknowledgements
We thank all the co-investigators in Linyi area, Shandong province. We sincerely thank all the study participants.