The CHARLS project aimed to provide a high-quality microdata set representing households and individuals, facilitating research on population aging and interdisciplinary studies related to aging. The survey covered a wide range of topics, including demographic information, family structure, intergenerational transfers, health status, medical care and insurance, employment, income, expenditures, and assets. The national baseline survey of CHARLS was launched in 2011, employing a multi-stage probability proportional to size (PPS) sampling method. CHARLS was an ongoing longitudinal survey, with follow-up waves conducted every 2–3 years. The survey sample included 450 villages, 150 counties, and 28 provinces, covering over 17,000 individuals from approximately 10,000 households. Data collection involved face-to-face interviews conducted in participants' homes using computer-assisted personal interviewing (CAPI) technology. The datasets were publicly available and could be accessed from the CHARLS homepage at http://charls.pku.edu. cn/en. This study utilized data from the wave 3 in 2015 of CHARLS, which includes detailed health examination indicators and lifestyle information. We initially included 21,095 individuals. The exclusion criteria were as follows: (1) individuals aged below 45 (n = 497); (2) individuals with incomplete data on CVAI (n = 7,178); (3) individuals with incomplete data on hypertension (n = 4,633); and eventually, 8,787 participants were included, and the screening process is illustrated in Figure 1.

CVAI was a comprehensive indicator that integrates anthropometric measurements and functional parameters to assess visceral adiposity distribution. Following the studies of Amato (14) and Han (15), a specific formula incorporating waist circumference (WC), body mass index (BMI), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-C) was used to calculated CVAI. Due to differences in adiposity distribution and metabolism between men and women, the CVAI formula was tailored for each gender to improve accuracy. CVAI was calculated using the following formulas:

CVAI (male) = −267.93 + 0.68 × age (years) + 0.03 × BMI (kg/m²) + 4.00 × WC (cm) + 22.00 × LgTG (mmol/L) - 16.32 × HDL-C (mmol/L)

CVAI (female) = −187.32 + 1.71 × age (years) + 4.23 × BMI (kg/m²) + 1.12 × WC (cm) + 39.76 × LgTG (mmol/L) - 11.66 × HDL-C (mmol/L)

BMI is expressed in kg/m², calculated as weight (kg) divided by height squared (m²).

Hypertension was determined using both self-reported medical history and blood pressure measurements. Hypertension was defined as either: (1) a self-reported physician's diagnosis of hypertension; or (2) measured systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg, based on the average of three readings; or (3) current use of antihypertensive medication.

The CHARLS working group used computer-assisted personal interviewing (CAPI) technology, and all physical and biochemical measurements were conducted by trained professionals using standardized protocols. Demographic and lifestyle information were collected through structured household interviews conducted by trained surveyors, including age, sex (female, male), residence (rural, urban), education level (illiteracy, primary school, middle school, high school and above), smoking status, drinking status. Health conditions and medical history were obtained through self-reported questionnaires and medical history interviews. Participants were asked if they had ever been diagnosed with dyslipidemia, diabetes mellitus (DM), heart diseases by a doctor. Participants also reported whether they were currently taking lipid-lowering medications. Blood samples were collected by trained healthcare professionals at participants' homes, and the samples were analyzed in certified laboratories. Trained medical professionals conducted standardized physical examinations to collect anthropometric and blood pressure measurements. Detailed methodologies are accessible on the CHARLS website (http://charls.pku.edu.cn/).

The baseline characteristics of the study population were calculated, including the mean and standard deviation for continuous variables and the frequency and percentage for categorical variables. The differences in CVAI and other covariates between the hypertension and non-hypertension groups were compared using t-tests or chi-square tests. A 3-knot restricted cubic spline (RCS) plot was utilized to explore the nonlinear relationship between CVAI and hypertension. Specifically, we used three knots placed at the 10th, 50th, and 90th percentiles of the CVAI distribution, which is a commonly recommended approach in epidemiological studies to balance model flexibility and stability. The aim was to establish the cutoff point using the RCS curve, followed by conducting a logistic regression analysis anchored on this threshold. A logistic regression model was conducted to analyze the association between CVAI and hypertension, adjusting for potential confounding factors. The results were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Subgroup analyses were conducted to explore the heterogeneity in the association between CVAI and hypertension. While our subgroup analyses suggested potential effect modifications by factors such as age, sex, and smoking status, it is important to acknowledge the issue of multiple testing. Conducting numerous statistical comparisons increases the probability of Type I error, where an association is falsely deemed significant. The CHARLS study was approved by the Ethics Committee of Peking University, and all participants provided informed consent. This study used publicly available anonymized data, which did not require additional ethical review. All statistical analyses were conducted using spss27.0, Stata 17.0, and R statistical software, with a significance level set at P < 0.05.

The study included a total of 8,787 individuals, with 2,626 (27.89%) having hypertension and 6,161 (70.11%) without hypertension (Table 1). The results indicated that the incidence of hypertension among the middle-aged and older adults varied based on several factors, including age, gender, education, smoking, drinking, dyslipidemia, diabetes mellitus (DM), use of lipid-lowering drugs, heart diseases, fasting blood glucose levels, creatinine, uric acid, triglycerides (TG), total cholesterol (TC), low-density lipoprotein (LDL), weight, body mass index (BMI), waist circumference, systolic blood pressure (SBP), diastolic blood pressure (DBP), and the Chinese visceral adiposity index (CVAI). The differences among these factors were statistically significant (p < 0.05). The average age of the entire sample was 61.46 ± 9.19 years, with hypertensive individuals being significantly older (63.98 ± 9.26 years) compared to non-hypertensive individuals (60.39 ± 8.95 years, p < 0.001). There were 1,334 females with hypertension (27.91%) and 3,445 without hypertension (72.09%). Among the middle-aged and older adults living in rural areas, 4,763 (70.06%) were in the non-hypertensive group, while 2,035 (29.94%) were in the hypertensive group. Among middle-aged and older adults with a high school education or higher, 263 (27.86%) were diagnosed with hypertension, while 681 (72.14%) were non-hypertension. Of those with hypertension, 1,139 (31.84%) were smokers, and 1,487 (28.54%) were non-smokers. In terms of alcohol consumption, 2,068 (68.57%) of the middle-aged and older adults in the non-hypertensive group consumed alcohol, compared to 948 (31.43%) in the hypertensive group. Regarding comorbidities, among middle-aged and older adults with hypertension, 406 individuals (38.85%) had dyslipidemia, 235 (40.80%) had diabetes mellitus (DM), and 421 (36.64%) had heart diseases. Among participants using lipid-lowering drugs, 600 (61.22%) were in the non-hypertensive group, whereas 380 (38.78%) were in the hypertensive group. Biochemically, hypertensive individuals had significantly higher levels of fasting blood glucose (104.25 ± 33.41 mg/dl), creatinine (0.82 ± 0.25 mg/dl), uric acid (5.10 ± 1.47 mg/dl), total cholesterol (188.57 ± 37.26 mg/dl), triglycerides (148.82 ± 95.03 mg/dl), and low-density lipoprotein (LDL) (105.10 ± 29.56 mg/dl), compared to those without hypertension. Other physical measurements showed that hypertensive individuals had a significantly higher weight (61.90 ± 12.18 kg), BMI (24.79 ± 4.04 kg/m²), and waist circumference (88.72 ± 11.52 cm). Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were also significantly higher in hypertensive individuals (152.38 ± 15.58 mmHg and 86.45 ± 11.18 mmHg, respectively). The Chinese visceral adiposity index (CVAI) was significantly higher in hypertensive individuals (122.30 ± 40.06) compared to non-hypertensive individuals (104.54 ± 38.72, p < 0.001), suggesting a strong association between CVAI and hypertension.

In addition, a 3-knot RCS was employed to model the dose-response relationship between CVAl and elevated hypertension, as depicted in Figure 2. The CVAI was plotted as the horizontal coordinate and the odds ratio (OR) was plotted as the vertical coordinate. The results, after adjusting for potential covariates, indicated that higher levels of CVAI cor-responded with an increasing risk of hypertension (nonlinear P < 0.001). The corresponding inflection point value was 108.88.

The relationship between CVAI and hypertension is displayed in Table 2. After controlled for potential covariates in Model 3, the findings indicated a significant positive relationship with an OR of 1.010 (95% CI: 1.008, 1.011). The RCS was applied to examine the nonlinear relationship between CVAI and hypertension. An inflection point value of 108.88 was derived (see Figure 2C), which was then used as a cutoff point to classify the CVAI as a dichotomous variable for logistic regression. In Model 1, the OR for the upper category in comparison with the lower category was 2.189 (95% CI: 1.993, 2.405). In Model 2, after controlled for age and gender, the OR for the higher category increased to 2.019 (95% CI: 1.834, 2.223). In model 3, multiple confounding factors were controlled, including age, sex, education, residence, smoking, alcohol consumption, diabetes, hyperlipidemia, heart disease, and lipid- lowering medication use. The OR was 1.967 (95% CI: 1.781, 2.172), indicating that the relationship between higher CVAI and hypertension remained statistically significant. The trend analysis confirmed a consistent positive correlation between the two categories (P for trend < 0.001).

Our subgroup analysis utilized stratification variables such as sex (Male vs. Female), residence (Rural vs. Urban), education (Illiteracy vs. Primary school vs. Middle school vs. High school and above), smoke (Yes vs. No), and drinking status (Yes vs. No). The results were visually presented using a forest plot. As presented in Figure 3, these analyses revealed that these factors influenced the relationship between the CVAI and hypertension. Specifically, a positive association between CVAI and hypertension was observed across all subgroups. This association appeared to be stronger within specific subgroups: such as women (P for interaction < 0.001), and non-smokers (P for interaction = 0.001). In contrast, residence, education, drinking status did not appear to have a substantial impact on the relationship (P for interaction >0.05). Overall, the test for interaction between subgroups indicated that the relationship between CVAI and hypertension varied significantly across different subgroups.