Comparison of Different Anthropometric Indices to Predict Prediabetes in North China, Beijing: a 7-year Prospective Study

Obesity is an increasing problem worldwide and is one of the underlying risk factors for prediabetes. Although WHtR, BMI, WC and WHR were found to be associated with dysglycemia, in view of signicant differences in basic characteristics, glycemic metabolism, and ethnicity, it is of practical value to investigate which indicators are the most useful measures to predict the incidence of prediabetes in Chinese population. The aim of this study was to compare the value of different anthropometric measures of obesity in the detection of prediabetes in a cohort study in China and to identify the best cutoff point of predicting prediabetes in this population. tertiles. anthropometric the most useful measures to predict the incidence of prediabetes in Chinese population. In this study, we compared the value of different anthropometric measures of obesity in the detection of prediabetes in a cohort study in China, aiming to explore the ability of these anthropometric indices to predict prediabetes and to identify the best cut-off point in this population. anthropometric with their and sensitivity validity for varies and regions. the effective for of prediabetes in followed by BMI. WC and were weak predictors. While in followed and

In a Chinese study population, WHtR was the best predictor of the development of prediabetes in the general population and in females 7 years in advance, while WHR could predict the development of prediabetes in males. Early identi cation of prediabetes can better prevent diabetes.

Background
The prevalence of prediabetes is rapidly increasing worldwide. Among Chinese adults, the overall prevalence of diabetes and prediabetes was estimated to be 11.6% and 50.1%, respectively [<link rid="bib1">1</link>] . Prediabetes is a warning sign of diabetes. Approximately 75%-80% of patients with diabetes ultimately develop cardiovascular disease, and patients with prediabetes have also been shown to have an increased risk of heart attack and stroke [2 − 4] . Measures to prevent prediabetes are critical.
Obesity is an increasing problem worldwide and is one of the underlying risk factors for metabolic diseases, such as dysglycemia, dyslipidemia, hypertension, and cardiovascular disease. Obesity can be identi ed by many indicators to identify obesity, such as the body mass index (BMI), waist-to-height ratio (WHtR), waist-to-hip ratio (WHR) and waist circumference (WC). BMI was the most commonly used indictor for de ning obesity, which is recommended by World Health Organization (WHO). While studies conducted subsequently found that BMI can neither distinguish between muscle and fat body mass [5] nor re ect abdominal fat. Some studies also have shown that increased BMI was not associated with increased mortality [6][7][8][9][10] . The studies available to date still have no conclusion as to which anthropometric parameter best predicts mortality. WC or WHR is encouraged for the additional use of BMI [11] . Recently, new anthropometric indices such as WHtR have been proposed as alternative indicators of obesity and can predict health risks and substantial health risks.
Although WHtR, BMI, WC and WHR were found to be associated with dysglycemia, in view of signi cant differences in basic characteristics, glycemic metabolism, and ethnicity [12][13][14][15] , it is of practical value to investigate which indicators are the most useful measures to predict the incidence of prediabetes in Chinese population. In this study, we compared the value of different anthropometric measures of obesity in the detection of prediabetes in a cohort study in China, aiming to explore the ability of these anthropometric indices to predict prediabetes and to identify the best cut-off point in this population.

Study participants
The present work was a part of the survey from the REACTION study. All subjects participated in a community longitudinal survey, which was conducted in the communities of Shijingshan district, Beijing, China. A total of 19443 residents were recruited and invited to participate in a questionnaire survey in 2011. The inclusion criteria were as follows: a, patients with normal blood glucose levels; and b, patients without diabetes, stroke, chronic heart disease or hyperlipidemia. The exclusion criteria were as follows: a, the presence of primary diseases at baseline, such as diabetes, heart infarction, stroke or chronic heart tolerance (IGT): FBG < 6.1 mmol/L and 7.8 ≤ BPG < 11.1 mmol/L, impaired fasting glucose (IFG): 6.1 ≤ FBG < 7.0 mmol/L and BPG < 7.8 mmol/L, and IFG + IGT: 6.1 ≤ FBG < 7.0 mmol/L and 7.8 ≤ BPG < 11.1 mmol/L; and for nondiabetes: FBG < 6.1 and PBG < 7.8 mmol/L simultaneously.

Statistical analysis
Statistical analysis was performed using SPSS software (version 21.0.0.0). Continuous variables with normal distribution are presented as the mean ± standard deviation (SD), while data skewed distribution are presented as the median (Q1, Q3). Categorical data are recorded as n and proportions. Student's t test, ANOVA and chi-squared tests were performed to test for signi cant differences in the general characteristics between the groups. Pearson correlation analysis was used to analyze the correlation coe cient. The odds ratio values (OR values) and the corresponding 95% con dence intervals (CIs) after adjusting confounding factors in different models were calculated using multivariate logistic regression. Finally, receiver operating characteristic (ROC) curve was used to analyze the discrimination ability and to determine the best cut-off value. Sensitivity and speci city were calculated based on cut-off values, which were estimated using the maximized Youden index.

Characteristics of participants at baseline
Totally, 2568 normoglycemic participants (810 males and 1758 females) were recruited in 2011 ( Fig. 1), with an average age of 62.68 ± 6.97 years. After 7 years of follow-up, 52 subjects had developed newly diagnosed diabetes, and 423 individuals had developed prediabetes. Among participants with prediabetes, 278 had IGT, 105 had IFG, and 40 had IGT + IFG. The incidences of prediabetes and diabetes were 115.3 per 1000 person-years and 14.2 per 1000 person-years, respectively. As can be seen in Table 1 and Table 2, the average age of subjects who developed diabetes (64.06 ± 5.23 years old) or prediabetes (64.06 ± 6.94 years old) was older than that of participants who did not develop diabetes (62.65 ± 7.04 years old and 62.36 ± 6.98 years old). As expected, the levels of systolic pressure, WHtR, weight, BMI, WC, WHR, TG, FBG, PBG and HbA1c among newly diagnosed diabetes patients were more likely to be higher than the levels among those who did not develop diabetes (p = 0.019, p = 0.001, p = 0.002, p = 0.001, p = 0.001, p = 0.004, p = 0.002, p < 0.001, p = 0.014 and p < 0.001). While the result of HDL-c was the opposite (p < 0.001). Moreover, the family history of diabetes was signi cantly different among different groups (p = 0.046). The rate of current employment was much lower while the rate of regular exercise and diabetes family history were much higher in the group of people who had developed prediabetes than in the other groups ( Table 2). The levels of WHtR (p < 0.001), weight (p < 0.001), BMI (p < 0.001), WC (p < 0.001), HC (p = 0.016), WHR (p < 0.001), systolic pressure (p < 0.001), diastolic pressure (p = 0.001), TG ( < = 0.001), LDL-c (p = 0.008), FBG (p < 0.001), PBG (p < 0.001) and HbA1c (p < 0.001) among those with prediabetes were more likely to be higher than those among people who did not develop prediabetes. The result of HDL-c was also the opposite (p = 0.002).

Association of anthropometric values with incident diabetes risk
To further analyze the function of these anthropometrics, we used strati ed analyses for different sexes.
To better discuss the relationship, we divided the WHtR, BMI, WC and WHR into tertiles. The results are shown in Table 3. Although all of these anthropometric values are meaningful in the general population, the results change after strati cation by gender. The results showed that these 4 indices were signi cantly different between diabetic and nondiabetic participants in the female group, but the difference disappeared in the male group.

Association of different anthropometric values with incident prediabetes risk
In order to better discuss the relationship between different anthropometric values and blood glucose status, we further divided people into prediabetic and normal blood glucose groups. The results are shown in  Receiver operating characteristic (ROC) curve analysis ROC curve analyses was performed to compare the predictive values of the anthropometric measures.
The respective areas under the ROC curves (AUC) was calculated for the prediction of prediabetes. The results are presented in Table 5 and Fig. 2. In the general population, the AUC of WHtR, BMI, WC and WHR were 0.598, 0.593, 0.589 and 0.588, respectively, and the WHtR was the strongest predictor of prediabetes  Optimal cut-off points of anthropometric indices in the prediction of prediabetes Table 6 shows the best cut-off points and corresponding sensitivity and speci city of each index in the prediction of prediabetes. For females, the optimal cut-off point of WHtR was 0.48, with the largest Youden index and highest sensitivity, BMI was 23.5 kg/m 2 , WC was 75.6 cm, WHR was 0.84; for males, the best cut-off point of WHtR was 0.50, BMI was 25.5 kg/m 2 , WC was 88.9 cm, WHR was 0.93.

Main ndings
To the best of our knowledge, this is the rst study to compare the function of these four anthropometric indices to predict prediabetes in a long-term cohort study. The present study found that WHtR is the best predictors of the development of prediabetes in a general population and in females in 7 years in advance compared with WC, BMI and WHR, while the WHR can predict the development of prediabetes in males compared with other indices. We determined the optimal cut-off values of these four variables for the prediction of the development of prediabetes in this speci c population. For females, the optimal cutoff point of WHtR was 0.48, BMI was 23.5 kg/m 2 , WC was 75.6 cm and WHR was 0.84; for males, the best cut-off point of WHtR was 0.50, BMI was 25.5 kg/m 2 , WC was 88.9 cm and WHR was 0.93. Early identi cation of prediabetes can better prevent diabetes.
The ability of WHtR, BMI, WC and WHR to predict prediabetes in general population and female Previous studies comparing WHtR, WC, BMI and WHR have been inconsistent in the detection of diabetes [16][17][18] , while few studies analyzed the relationship between these indicators and prediabetes.
Other studies have shown that more body fat and higher metabolic risk were found for a given BMI value than European individuals [19] , and BMI does not provide details on body fat distribution.
Results in our study were consistent with above opinions. As in other studies, we found that WHtR in females was the best predictor of incident prediabetes. F. Javier Sangro et al [20] studied the association of general and abdominal obesity with prediabetes among 2022 participants in the PREDAPS study, and they found that, compared with BMI among women, the WHtR showed a stronger association with prediabetes, and the results were different between men and women. This can be explained by the following reason, abdominal obesity can induce a state of insulin resistance, which lead to a defective response to insulin in peripheral tissue and result in altered glucose uptake and utilization [20][21][22] . A study from Jordan revealed that the WHtR had a strong association and performed better for the detection of high FBG than the other anthropometric indices [23] . An study conducted among Iranian found that among women, compared with the use of BMI, hip circumference and WC, the WHtR was the only signi cant anthropometric predictor of prediabetes [24] . Additionally, the study on the relationship between anthropometric measurements with their best cut-off values and dysglycemia found that WHtR had the highest sensitivity [25] . The differences in these results indicated that the validity of anthropometric measurements for the prediction of prediabetes development varies among different genders, ethnicities, and regions. In our study, WHtR was the most effective indices for the prediction of prediabetes in females, followed by BMI. WC and WHR were relatively weak predictors. While in males, WHR was the most useful measurement in predicting prediabetes, followed by BMI, WHtR and WC.
Our results in male showed that compared with the function of WHtR, BMI and WC, WHR was the best predictor of incident prediabetes in males. A high WHR is an important risk factor for hyperglycemia. The study of Manju Bala [26] found that WHR is positively correlated with HbA1c, which means that a decrease in WHR is effective for the prevention of prediabetes. The previous ndings regarding the association between the WHR and prediabetes [27,28] were consistent with the results of our study.
The best cut-off points for these four anthropometric indices It is well known that a single set of cut-offs cannot be applied universally, and there is substantial evidence to show that Asian individuals need lower anthropometric cut-offs for the identi cation of obesity and abdominal obesity than European individuals [29] . Interestingly, in the present study, the best cut-off points of these four anthropometric measurements for the detection of prediabetes differ from those used for the diagnosis of obesity. In addition, it is also smaller than those in other studies, especially in WHtR, BMI and WC, which means that if we do not use a speci c value, the frequency of prediabetes may be underestimated. Better control of WHtR, BMI and WC below the diagnosis value of obesity can better prevent the occurrence of prediabetes. While the value of WHR in the present study was bigger than previous. It is maybe because aging can both increase body fat and change the distribution.
Therefore, the need to determine speci c cut-off points of these variables for optimal screening of prediabetes is urgent. An obvious difference was observed in these four indices in different gender, suggesting that sex-speci c values should be recommended for use in practice. This need has also been illustrated in a previous study [30][31][32] . In the present study, the cut-off points of WHtR, BMI and WC were smaller than those that are currently recommended for use in this population, which means that the speci c cut-off points can play a role in early prevention.
The ability of WHtR, BMI, WC and WHR to predict diabetes Although a recent meta-analysis [33] , a prospective study in Iran [34] and some articles [35] indicated that WHtR might be a better predictor of T2D, our study found that there are no signi cant relationship between any obesity indices and incident diabetes in either sex. The results were the same as those of Bozorgmanesh [36] , who studies the relationship between central-obesity indices and incident diabetes. This may be because the follow-up time in most of the literature is not long enough. More long-term follow-up data are needed to further analyze the relationship between traditional anthropometric indices and diabetes.

Limitations
Some limitations should be addressed in our work. First, the majority of the participants were middleaged or elderly individuals, so the applicability of this conclusion to the general population remains to be determined. Second, the size of the sample was limited, especially the population of males. More participants will be followed up in the future. Third, there was a lack of OGTT reproducibility. A limitation of nancial resources and the unwillingness of subjects were the two major causes of this problem. We tried our best to minimize the associated bias and improve the diagnostic accuracy as much as possible by combining these data with the questionnaire data, FBG and HbA1c results.

Conclusion
In summary, WHtR is the best predictors of the development of prediabetes in a general population and in females in 7 years in advance, while the WHR can predict the development of prediabetes in males. We determined the optimal cut-off values of these four variables for the prediction of the development of prediabetes in this speci c population. Early identi cation of prediabetes can better prevent diabetes.

Figure 1
Flow chart of the selection of study participants