Comparison of Several Adiposity Indexes in Predicting Hypertension among Chinese Adults: Data from China Nutrition and Health Surveillance (2015–2017)

The current study is to explore the association of the Chinese visceral adiposity index (CVAI) with hypertension, and to compare the predictive power of different adiposity indexes regarding hypertension among Chinese adults aged over 45 years. A total of 99,201 participants aged over 45 years from the China Nutrition and Health Surveillance 2015–2017 were included in this study. Multivariate adjusted logistic regression was used to calculate the odds ratio (OR) and 95% confidence interval (CI) of hypertension. Multivariate adjusted restricted cubic spline analyses were applied to explore the association of adiposity indexes with hypertension. Receiver operating characteristic (ROC) analyses were used to compare the predictive powers of different adiposity indexes of hypertension. All eight adiposity indexes included in this study were positively associated with hypertension. Compared with those in the lowest quartile of the CVAI, the participants in the highest quartile showed a significantly higher risk of hypertension (OR = 3.70, 95% CI = 3.54–3.86) after multiple adjustments. The ROC analyses suggested that the CVAI was the strongest predictor of hypertension compared to other adiposity indexes in both genders. The findings supported that the CVAI could serve as a reliable and cost-effective method for early identifying hypertension risk.


Introduction
During the last three decades, the number of hypertensive patients aged from 30 to 79 has increased from 650 million in 1990 to 1.28 billion in 2019 globally. Among those hypertensive patients, nearly half did not report having been previously diagnosed with hypertension [1]. In China, hypertension is the primary cause of cardiovascular diseases. The prevalence of hypertension has been increasing rapidly, which has further imposed heavy burdens on the medical care system [2]. Therefore, an earlier diagnosis of hypertension is of great necessity to reduce the risk of adverse health outcomes.
Among a wide spectrum of hypertension-related risk factors, adiposity is less controversial [3]. Previous evidence has indicated that visceral adipose tissue, rather than subcutaneous adipose tissue, was more strongly associated with the risk of hypertension [4,5]. Although visceral adipose tissue can be evaluated using computed tomography (CT) and magnetic resonance imaging (MRI), they are not suitable for large population-based studies because of the high cost and potential exposure risk. Body mass index (BMI) and waist circumference (WC), generally used to reflect adiposity, are incapable of identifying adipose tissue in the abdominal region. Several adiposity indexes developed to facilitate the screening of hypertension have emerged recently, including the visceral adiposity index (VAI), lipid accumulation product index (LAP), body roundness index (BRI), body shape index (ABSI), and conicity index (CI) [6][7][8][9][10]. However, previous studies have indicated that these indexes may have limitations in predicting hypertension among Chinese populations [11][12][13].
The Chinese visceral adiposity index (CVAI) was developed by Xia et al. in 465 Chinese adults to diagnose visceral adiposity, and its predictive power has been validated in a cross-sectional study [14]. Meanwhile, this index has been proven to be associated with some diseases, including metabolic-associated fatty liver disease [15], diabetes [16], diabetic complications [17], carotid plaque [18], and renal damage [19]. Nevertheless, little attention is paid to the connection between hypertension and the CVAI among middle-aged and elderly Chinese people [20]. The accuracy of different adiposity indexes in identifying hypertension remains to be verified.
The current study aims to explore the association of the CVAI with hypertension and to compare the predictive powers of different adiposity indexes (CVAI, VAI, LAP, BRI, ABSI, CI, WC, and BMI) regarding hypertension among Chinese middle-aged and elderly people, based on a nationally representative survey.

Participants
All the data in the study were obtained from the China Nutrition and Health Surveillance (2015-2017) (CNHS 2015-2017). The multistage stratified random sampling method was adopted to obtain a representative national sample in 2015-2017. All the participants were 18 years old and above and were from 31 provinces/municipalities/autonomous regions in mainland China. The details of the design and sampling method of the CNHS are available in the previous publication [21]. Participants without complete blood pressure or anthropometric measurement information were excluded. Finally, a total of 99,201 participants aged over 45 years were included in this study.
The CNHS 2015-2017 was approved by the Ethics Review Committee of the Chinese Center for Disease Control and Prevention -19B, in June 2015, and all the subjects had signed the informed consent before the survey.

Basic Information Collection
Information related to gender, age, geographic region (urban/rural area), educational level, marital status, income, family history, lifestyle factors (smoking and alcohol consumption, sleep duration, physical activity, sedentary behavior, medical examination within one year), and previous disease diagnoses was collected through a standardized questionnaire by well-trained investigators in a face-to-face manner.

Anthropometric Measurement and Definition of Hypertension
To reduce measurement bias, height, weight, waist circumference, and blood pressure were measured by well-trained local CDC staff. Anthropometric measurements were carried out under a fasting condition in the morning in a fixed place. Height was measured with a metal column height meter with an accuracy of 0.1 cm while the participants were standing in an upright position. Weight was measured with an electronic weight scale with an accuracy of 0.1 kg. Waist circumference was measured twice using a tape with an accuracy of 0.1 cm, and the average value was calculated.
Blood pressure was measured three times to the nearest 1 mmHg with the cuff positioned equivalent to the heart level on the left arm by an electronic sphygmomanometer (OMRON HBP-1300) after the participants had rested in a seated position for at least 5 min. The average value of the three-times systolic blood pressure (SBP) and diastolic blood pressure (DBP) was used for the data analysis. Hypertension was defined as SBP ≥ 140 mmHg and (or) DBP ≥ 90 mmHg and (or) receiving antihypertensive medications within two weeks [22].

Laboratory Test
An 8 mL fasting blood sample was collected to measure fasting glucose, total cholesterol, triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and glycosylated hemoglobin (HbA1c). All the above measurements were taken by professionals following strict quality control standards in the laboratory.

Adiposity Indexes Calculations
The specific formula for calculating the Adiposity Indexes is as follows: where BMI is the body mass index; VAI is the visceral adiposity index; WC is the waist circumference; TG is triglyceride; HDL-C is high-density lipoprotein cholesterol; CVAI is the Chinese visceral adiposity index; LAP is the lipid accumulation product index; ABSI is a body shape index; BRI is the body roundness index; and CI is the conicity index.

Covariates
The covariates in the sample description and logistic regression analysis were as follows. (1) The geographic region was categorized as an urban area or rural area. (2) The education level was categorized as primary school and below, junior middle school, or senior high school and above. (3) Marital status was divided into married or other. (4) Income levels were divided into low (<10,000 CNY), medium (10,000-25,000 CNY), or high (>25,000 CNY), based on the annual household income per capita. (5) According to the WC, the participants were grouped as normal weight or central obesity [23]. (6) According to the BMI, the participants were divided into underweight (BMI < 18.5), normal weight (18.5 ≤ BMI < 24), overweight (24 ≤ BMI < 28), or obese (BMI ≥ 28) [23]. (7) A family history of hypertension (Yes/No) was defined as any one of the lineal relatives (including grandparents, parents, or siblings) who had been diagnosed with hypertension. (8) Smoking status was separated into current smoker or non-smoker. (9) Alcohol drinking was divided into alcohol intake within one month or not. (10) Sleep duration was categorized  (11) Physical activity was divided into low (MET < 600), medium (600 ≤ MET ≤ 3000), or high (MET > 3000), according to the weekly total metabolic equivalent (MET-min/week) and total duration of different physical activities [24]. (12) Sedentary behavior was grouped as <2 h, 2~3 h, or ≥4 h per day. (13) A medical examination within one year was categorized as yes or no.

Statistical Analysis
Given that the abnormal distribution of our data, the continuous variables in this study are shown as the median with an IQR (interquartile range) and were compared by Wilcoxon tests between the different groups. The categorical variables were described by count (n, %) and compared by chi-square tests between the different groups.
Three multivariate logistic regression models were built to identify the connections between the adiposity indexes and hypertension. The results are presented as the odds ratio (OR) and 95% confidence interval (95% CI). Each adiposity index was divided into quartiles, with the lowest quartile (Q1) as the reference group. Meanwhile, each adiposity index as a continuous variable was examined per SD increment. Model 1 was an unadjusted model. Model 2 was adjusted for age and gender. Model 3 was further adjusted for the geographic region, education, marital status, income, family history, smoking, alcohol drinking, sleeping time, physical activity, sedentary behavior, medical examination within one year, diabetes, and dyslipidemia.
To further analyze the dose-response relationships between hypertension and the adiposity indexes, multivariate-adjusted restricted cubic spline (RCS) analyses were used, and the four knots were set at the 5th, 35th, 65th, and 95th percentiles, with the 10th percentile as the reference knot [25]. Furthermore, receiver operating characteristic (ROC) analyses were used to determine the area under the ROC curves (AUC) between hypertension and the adiposity indexes to compare the predictive powers of the different adiposity indexes of hypertension in males and females. The differences between the different adiposity indexes for hypertension in both genders were determined according to DeLong et al.'s nonparametric approach [26]. The definition of statistical significance was two-sided, p < 0.05.
In the current study, SAS version 9.4 software (SAS Institute, Inc., Cary, NC, USA) was used for all the statistical data cleaning. R software version 4.2.2 was used to perform the ROC curve and dose-response analyses.

Basic Characteristics
The basic characteristics, anthropometric measurements, and laboratory test data of the participants aged 45 years and above in CNHS 2015-2017 are shown in Tables 1 and 2. As Table 1 shows, 52% of the total participants had hypertension. Of the total, 51.9% of males and 52% of females had hypertension. As Table 2 shows, the average systolic and diastolic blood pressures of patients with hypertension were 152.0 mmHg and 86.33 mmHg, respectively. The hypertensive participants tended to be older and had greater weight, FPG, TC, TG, LDL-C, HbA1c, and lower HDL-C concentrations. Meanwhile, the participants with the following characteristics were more likely to develop hypertension: lower education and income, married, family history of hypertension, central obesity or general obesity, smoking at that time, inadequate sleep duration, longer sedentary behaviors, receiving medical examination within one year, diabetes, and dyslipidemia.
As the results show in Table 3, the hypertensive participants had higher adiposity indexes compared with the non-hypertensive participants (p < 0.0001).  Abbreviations: SBP-systolic blood pressure, DBP-diastolic blood pressure, FPG-fasting plasma glucose, TC-total cholesterol, TG-triglyceride, LDL-C-low-density lipoprotein cholesterol, HDL-C-high-density lipoprotein cholesterol, HbA1c-glycosylated hemoglobin.  Table 4 shows the associations between these adiposity indexes and hypertension by quartile and per standard deviation (SD) increase, respectively. In all three models, the ORs for hypertension increased significantly with the increasing quartile number for all the adiposity indexes (p for trend < 0.0001), suggesting that those in the highest quartile had a significantly higher risk of hypertension compared with participants in the lowest quartile of the adiposity indexes. With Q1 of the CVAI as the reference, the ORs (95% CIs) for hypertension in quartiles 2, 3, and 4, were 1. 55 (1.50, 1.61  The dose-response relationship between the ORs (95% CIs) of hypertension and the CVAI is shown in Figure 1. Further information on RCS and the other adiposity indexes and hypertension is available in Supplemental Figure S1. After an adjustment for several confounding factors, a nonlinear association (p for nonlinearity < 0.05) was detected between the risk of hypertension and these indexes for both men and women, except for the WC of males (p for nonlinearity = 0.058). At the initial increase stage of the CVAI score, the change rate of the OR tended to slightly and then rapidly increase.

Comparison of the Association of CVAI, VAI, LAP, ABSI, BRI, CI, WC, and BMI with Hypertension Risk
The ROC curves for hypertension and the adiposity indexes for males and females are shown in Figure 2. The ROC curves of all the indexes are above the reference line, indicating that these indexes could be used to identify hypertension. However, among them, the predictive power for hypertension of the CVAI showed to be best for both genders.

Comparison of the Association of CVAI, VAI, LAP, ABSI, BRI, CI, WC, and BMI with Hypertension Risk
The ROC curves for hypertension and the adiposity indexes for males and females are shown in Figure 2. The ROC curves of all the indexes are above the reference line, indicating that these indexes could be used to identify hypertension. However, among them, the predictive power for hypertension of the CVAI showed to be best for both genders.

Hypertension Risk
The ROC curves for hypertension and the adiposity indexes for males and females are shown in Figure 2. The ROC curves of all the indexes are above the reference line, indicating that these indexes could be used to identify hypertension. However, among them, the predictive power for hypertension of the CVAI showed to be best for both genders.  Table 5 displays the area under the ROC curves (AUCs) and their 95% CIs for each adiposity index. The AUCs of all the adiposity indexes predicting hypertension were greater than 0.5, and the CVAI (Males: 0.636, 95% CI: 0.631-0.641; Females: 0.706, 95% CI: 0.702-0.710) had the largest AUC among the eight adiposity indexes of hypertension. The differences between the AUC of the CVAI and that of the other indexes for hypertension in both men and women were all significant (p < 0.01).
In men, the cutoff with the highest Youden index of the CVAI was 98.268, with a sensitivity of 56.2% and a specificity of 63.80%. In women, the cutoff with the highest Youden index of the CVAI was 101.165, with a sensitivity of 65.31% and a specificity of 65.12%.  Table 5 displays the area under the ROC curves (AUCs) and their 95% CIs for each adiposity index. The AUCs of all the adiposity indexes predicting hypertension were greater than 0.5, and the CVAI (Males: 0.636, 95% CI: 0.631-0.641; Females: 0.706, 95% CI: 0.702-0.710) had the largest AUC among the eight adiposity indexes of hypertension. The differences between the AUC of the CVAI and that of the other indexes for hypertension in both men and women were all significant (p < 0.01). In men, the cutoff with the highest Youden index of the CVAI was 98.268, with a sensitivity of 56.2% and a specificity of 63.80%. In women, the cutoff with the highest Youden index of the CVAI was 101.165, with a sensitivity of 65.31% and a specificity of 65.12%.

Discussion
This national representative cross-sectional study, containing data on nearly 100,000 individuals, showed that the risk of hypertension in Q4 of the CVAI is more than doubled compared with Q1 of the CVAI after adjusting for potential confounding factors. We also found that the AUC of the CVAI for the diagnosis of hypertension was the largest, and the OR per 1 SD increase in the CVAI with hypertension was the highest when adjusting for the same confounders among all eight adiposity indexes in both genders. These results indicated that the CVAI might have the strongest predictive power to identify hypertension in Chinese middle-aged and older people compared with the VAI, LAP, ABSI, BRI, CI, WC, and BMI.
The CVAI, composed of age, physical measurement data (BMI and WC), and biochemical blood indexes (TG and HDL-C levels), was initially used to reflect the amount of visceral fat and has been established as a reliable indicator for evaluating metabolic health among the Chinese population [14]. Previous studies reported that the CVAI or the VAI was superior to the BMI, WC, ABSI, BRI, CI, or LAP in the diagnosis of hypertension [20,27,28], which is consistent with the results of our study. The reason might be that the BMI alone is not enough to differentiate body fat from lean tissues, and it does not differentiate the location of central fat from peripheral fat [29]. WC and LAP can better reflect abdominal obesity than the BMI but have limitations in distinguishing subcutaneous from visceral adipose tissue [14,30]. The ABSI also has also similar limitations to the BMI and WC [8] and was reported to be a weak index for metabolic syndrome and its components in a few cross-sectional studies [31]. Fujita et al. proved that, compared with BMI or WC, the ABSI was not a better predictor of hypertension in Japanese adults [32]. The possible reason for this may be racial diversity. As widely recommended adiposity indexes, WC and BMI are also alternative predictors for hypertension if the data for blood lipid levels are not available. CI, developed by Valdez et al. in 1991, is calculated by weight, height, and WC [9]. Though there is plenty of evidence indicating that CI is a reliable index for predicting hypertension, in our study it is inferior to the CVAI in identifying hypertension [33,34]. These discrepancies might have several explanations, including ethnic differences, and different sizes and characteristics of the study samples. However, the CVAI and VAI can reflect visceral fat mass [14].
It is noteworthy that the BRI, a novel predictor of visceral adipose tissue and body fat percentage, developed by D.M. Thomas et al. in 2013, also outperformed other adiposity indexes, except the CVAI, as shown in the ROC curves ( Figure 2) [7]. Many studies have shown its relationship with hypertension [35][36][37]. However, few studies have compared the predictive ability of the CVAI and the BRI for hypertension simultaneously. Given the excellent performance of the BRI in the current study and the simplicity of its formula, there is a need for further prospective studies to confirm its relationship with hypertension.
It is reported that fat distribution varies by different ethnicities, and Asians could be more prone to have visceral adipose accumulation compared with Caucasians [38,39]. This also reconfirms that the CVAI is more suitable than the VAI for the Chinese population, which is consistent with this study.
Though the positive association between the CVAI and the risk of hypertension has been demonstrated in different populations, there are still considerable studies that reached different conclusions. A cross-sectional study reports that, in adult Chinese, blood pressure is more strongly associated with general adiposity rather than with central adiposity [40]. Our study supports that abdominal obesity is more closely associated with hypertension rather than with generalized obesity. A previous study by Lee et al. supports our results [41]. Han et al., in their prospective cohort study of 10,304 Chinese adults, used the CVAI as a predictor of hypertension and followed up for 6.03 years. Their results indicated that the CVAI could be a reliable index of hypertension in rural Chinese adults [20].
Our study also presents optimal cut-off points for these adiposity indexes. Because it is easy to measure the WC and BMI, the two indexes are commonly used in daily practice to evaluate obesity. The optimal cut-off point for BMI is basically identical to recommendations from the 2018 Chinese Guidelines for Prevention and Treatment of Hypertension [42]. However, the present results show that the optimal cut-off point for WC is inconsistent with the cut-off point according to the guidelines (85 cm for males and 80 cm for females) [42]. Therefore, the WC cut-off points should be reconsidered regarding early detection among Chinese adults with a high risk of hypertension.
Meanwhile, the gender differences in the associations of the CVAI, VAI, LAP, ABSI, BRI, CI, WC, and BMI with hypertension were observed. The AUC values of all the adiposity indexes of the women were higher than those of the men after an adjustment for the same potential confounding factors. Mounting evidence indicates that the predictive powers of adiposity indexes vary by sex [43][44][45]. This could be partly explained by different fat distribution in each gender. It has been discovered that women in Asia tend to carry more fat around their waist than men do, and this may cause more adverse metabolic outcomes in women than men [46]. However, Chen et al. reported that visceral adipose tissue is more prevalent in men rather than women and contributes more to SBP in men rather than in women [44]. Meanwhile, sex hormones and the pathogenesis of hypertension may also cause the differences [47]. Further study needs to be conducted concerning the effects of those potential confounding factors on hypertension. Women may require female-specific and primary prevention and lifestyle management for adiposity and hypertension compared to men.
To the best of our knowledge, this is the first study to compare the CVAI with other obesity indicators, including the VAI, LAP, ABSI, BRI, CI, WC, and BMI, with the risk of hypertension among Chinese people aged 45 years or older recruited by a nationally representative sample. This current study shows the CVAI should be recommended as the best predictor of hypertension in males and females. Furthermore, in the present paper, confounding factors are better controlled.
However, several limitations should be noted in our study. Firstly, the causal relationship between adiposity indexes and hypertension cannot be inferred due to the crosssectional design. Secondly, although a wide range of confounders were adjusted, residual confounding factors are not taken into consideration, such as dietary factors. Thirdly, as our study only focuses on middle-aged and older Chinese adults, the current results might not be applicable to populations from other age groups.

Conclusions
In conclusion, all eight adiposity indexes in this study can be used to predict hypertension. The CVAI showed superior predictive powers to the other indexes in middle-aged and elderly Chinese people. The findings demonstrate that the CVAI could serve as a reliable and cost-effective method for the early identification of a high risk of hypertension. It also makes the primary screening and prevention of hypertension among the Chinese population more convenient, thereby reducing the risk of hypertension. Further prospective studies are necessary to validate the connection between the CVAI and adverse health outcomes in other age groups.

Data Availability Statement:
According to the policy of the National Institute for Nutrition and Health, China CDC, data related to this research are not allowed to be disclosed.