Correlation between Chinese visceral adiposity index and incidence of hypertension across different blood pressure status: a cohort study

Abstract

Background

Hypertension is a significant global public health issue and its pathogenesis is strongly associated with visceral obesity. The Chinese Visceral Adiposity Index (CVAI), a well-established indicator based on the metabolic profile of Asian populations, has been linked to an increased risk of hypertension. However, there is a lack of comprehensive evidence regarding the differential relationship between the CVAI and the incidence of hypertension in participants with normal vs. elevated blood pressure (BP). This study aimed to utilize longitudinal data to examine the association between the CVAI and the incidence of hypertension across populations with different baseline BP levels.

Methods

Data from the China Health and Retirement Longitudinal Study (CHARLS) database were used for this retrospective cohort analysis. Participants aged ≥45 years without hypertension at baseline (2011) were included and followed up until 2020 to assess the incidence of hypertension. The CVAI was calculated using a sex-specific formula incorporating age, waist circumference (WC), body mass index (BMI), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-C). Multivariate logistic regression and restricted cubic spline (RCS) models were used to evaluate the nonlinear association between CVAI and hypertension risk. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the predictive ability of CVAI. For the sensitivity analyses, univariate RCS models were employed to assess the robustness of the findings.

Results

Among 5311 participants, 1,819 (34.25%) developed hypertension during the 9-year follow-up period. CVAI emerged as an independent risk factor for incident hypertension. Comparing the highest to the lowest quartile of CVAI, the adjusted odds ratio (OR) was 1.77 (95% CI: 1.32–2.38) in the normal BP group and 3.21 (95% CI: 2.26–4.61) in the elevated BP group.

Conclusions

A linear dose-response relationship was observed between CVAI and hypertension risk in both groups. These associations remained robust and linear in sensitivity analyses after excluding participants with diabetes, heart disease, or stroke. Clinically, CVAI offers a simple, low-cost tool to improve hypertension risk stratification and enable earlier targeted prevention, particularly among individuals with elevated BP.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05617-y.

Background

Although significant progress has been made in the treatment and prevention of hypertension, it remains a leading risk factor for cardiovascular morbidity and mortality worldwide, posing a major public health challenge [1]. Global epidemiological data from 2010 reported that 1.39 billion adults have been diagnosed with hypertension, with a continuing upward trend over the past decade [2]. Furthermore, BP exhibits a continuous, graded association with cardiovascular risk rather than a threshold effect. Reflecting this evidence, the 2024 ESC guidelines formally introduced the distinct category of elevated BP (120–139/70–89 mmHg) to elucidate significant cardiometabolic risk, while maintaining the conventional hypertension threshold (≥140/90 mmHg) [3].

Beyond traditional risk factors such as high sodium intake, alcohol use, and genetic predisposition, visceral adiposity, a core component of metabolic syndrome, has emerged as a key driver in hypertension pathogenesis [4]. Data from a Chinese population revealed that 35–64-year-old adults had overweight and obesity rates of 38.8% and 20.2%, respectively, and these groups faced a 16–28%-higher hypertension risk than normal-weight individuals [5]. Prolonged hypertension induces multi-organ damage through endothelial dysfunction, myocardial remodeling, and renal injury, increasing cardiovascular events and mortality [6]. Alarmingly, WHO 2019 data indicate suboptimal global hypertension control rates [7]. Therefore, the early identification of high-risk populations and targeted preventive measures are crucial to reduce the incidence of hypertension.

In the context of early hypertension risk stratification, obesity-related indicators are crucial targets owing to their measurability and metabolic relevance. Given the well-established association between obesity and hypertension [4, 8], studies focusing on obesity-related indicators to investigate their mechanistic links with the development of hypertension are required. Visceral fat accumulation plays a more critical pathogenic role in metabolic disorders than does subcutaneous fat accumulation. In older adults, visceral fat has been identified as an independent predictor of cardiovascular events and all-cause mortality [9]. This finding has driven the development of precise methods for assessing visceral adiposity.

Although WC outperforms BMI in reflecting abdominal obesity, its capacity to differentiate visceral fat from subcutaneous fat remained limited [10]. Furthermore, although computed tomography (CT)/magnetic resonance imaging represent the gold standard for visceral fat measurement, the high cost thereof restricts clinical utility [11, 12]. To address this, researchers have developed a visceral adiposity index (VAI) that integrates WC, body mass index (BMI), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-C) to estimate visceral fat [13]. Studies have confirmed significant associations between the VAI and metabolic syndrome components. However, owing to ethnic variations in fat distribution, the predictive performance of the index shows population heterogeneity [14]. Particularly in Asian populations, visceral fat accumulation occurs at relatively low BMI, highlighting the need for ethnicity-specific assessment tools.

To address this need for ethnicity-specific assessment, Xia et al. developed the CVAI by incorporating population-specific parameters (age, WC, BMI, TG, and HDL-C), which significantly improves visceral fat estimation accuracy in Chinese populations [14]. Although cross-sectional studies have confirmed an association between CVAI and hypertension prevalence [15], longitudinal evidence regarding the relationship between CVAI, incident hypertension, and BP progression remains to be elucidated.

This study therefore aimed to utilize China Health and Retirement Longitudinal Study (CHARLS) longitudinal data (2011–2020) to systematically evaluate the predictive value of the CVAI across normal and elevated BP subgroups for the first time, providing evidence for BP-stratified precision prevention strategies. Considering the continuously increasing prevalence of hypertension in China, this study has significant implications for early prevention and control strategies.

Methods

Data source and participants

Participants were selected from the CHARLS database. The CHARLS project recruited participants exclusively from the population aged 45 years or older and collected comprehensive data on their lifestyle, health status, economic support, and social engagement. Launched in 2011, this nationally representative study utilized a multistage stratified probability-proportional-to-size sampling, recruiting approximately 19,000 participants from more than 12,000 households. Data were collected every 2 years, resulting in five waves of surveys by 2020. The study protocol was reviewed and approved by the IRB of Peking University, and all participants provided written informed consent prior to participation [16].

The participants from the 2011 CHARLS baseline survey who completed the 2020 follow-up assessment comprised the study population. Among the initial 17,708 individuals, those who met the following baseline exclusion criteria were excluded: (1) age < 45 years, (2) prevalent hypertension at baseline, and (3) missing baseline data for CVAI or covariates. Participants who were lost to follow-up or had missing hypertension diagnostic data by 2020 were excluded from the final analytical cohort (n = 5,311). Hypertension was defined by a comprehensive assessment that included any of the following: (1) a self-reported previous physician diagnosis of hypertension, (2) current use of antihypertensive medication, or (3) recorded SBP ≥140 mmHg or DBP ≥90 mmHg. For baseline stratification and alignment with the 2024 ESC criteria, participants without hypertension at baseline were categorized into two groups: an Elevated BP group (SBP 120–139 mmHg or DBP 70–89 mmHg) and a normal BP group [3]. A detailed flowchart of participant inclusion and exclusion is shown in Fig. 1. Baseline characteristics were compared between the included and excluded participants to assess potential selection bias (Supplementary Material 1).

Fig. 1.

Flowchart of the study

Covariates

The study covariates were as follows: (1) demographic characteristics (age, sex, residential area); (2) behavioral factors (current smoking status, alcohol consumption); (3) clinical parameters (baseline SBP and DBP); (4) socioeconomic factors (education level); and (5) health-related variables (daily sleep duration and comorbidities). Educational level was divided into four categories: below primary school, primary school, middle school, and high school or higher. Thirteen comorbidities were assessed: diabetes mellitus, malignant neoplasms, chronic pulmonary diseases, cardiovascular diseases, stroke, mental disorders, arthritis, dyslipidemia, hepatic diseases, renal diseases, gastrointestinal diseases, asthma, and memory-related disorders. Given that the CVAI calculation already incorporated BMI and WC, these variables were not additionally adjusted in the regression models to avoid multicollinearity and overadjustment. All blood specimens (including TG and HDL-C measurements) were analyzed at the You'anmen Clinical Laboratory Center of Capital Medical University, which holds both College of American Pathologists accreditation and ISO 15189 certification. Frozen plasma and whole blood samples were processed using enzymatic colorimetric methods [17].

CVAI assessment

CVAI was calculated using sex-specific formulas incorporating age, WC, BMI, TG, and HDL-C, as previously validated [18]:

Outcome and follow-up

All participants completed the follow-up by 2020. The primary outcome was 9-year cumulative incidence of hypertension, which was defined as the first occurrence of the condition at any point during the follow-up period (2013–2020). To maximize case ascertainment, we used longitudinal data from the 2013, 2015, 2018, and 2020 waves of the CHARLS. Participants were classified as having incident hypertension if they met any of the following criteria: (1) a self-reported physician diagnosis of hypertension in any follow-up wave; (2) self-reported current use of antihypertensive medication; or (3) an on-site measured SBP ≥ 140 mmHg or DBP ≥ 90 mmHg during the 2013 wave. As physical examination data were not consistently available for the 2015, 2018, and 2020 waves in the accessible dataset, outcome ascertainment for these years relied on validated questionnaire responses regarding previous diagnoses and medication use.

Statistical analysis

The normality of the data distribution was assessed using the Shapiro–Wilk test. Given that continuous variables were non-normally distributed, they are expressed as medians (interquartile range [IQR]: Q1, Q3), whereas categorical variables are presented as frequencies and percentages (%). Differences in continuous variables were analyzed using the Mann–Whitney U test, and categorical variables were compared using the chi-square test or Fisher’s exact test. All data processing and statistical analyses were performed using R software (version 4.4.0). Statistical significance was defined as a two-sided p-value < 0.05.

Logistic regression and restricted cubic spline (RCS) models

To assess the relationship between CVAI and incident hypertension, three hierarchically adjusted logistic regression models were constructed: Model 1 was unadjusted; Model 2 accounted for demographic variables (age and sex) and lifestyle factors (smoking and alcohol use); and Model 3 was additionally adjusted for socioeconomic status (education), comorbidities, and sleep duration. To explore potential nonlinear dose-response associations, restricted cubic spline (RCS) models were constructed based on the fully adjusted framework of Model 3. Separate analyses were conducted for the overall population, normal BP subgroup, and elevated BP subgroup to visualize the relationships across different BP states.

Predictive performance evaluation

Receiver operating characteristic (ROC) curve analysis was used to quantify the predictive performance of the CVAI. DeLong’s test was used to compare the area under the curve (AUC) differences between the CVAI and conventional adiposity indices (BMI and WC) or lipid parameters (TG and HDL-C). Three logistic regression prediction models were developed based on Model 1–3 structures for a comprehensive evaluation.

Subgroup and sensitivity analyses

To validate the generalizability of the predictive utility of the CVAI, stratified analyses were carried out by sex (male/female), age (<60/≥60 years), BMI category (underweight/normal/overweight/obese), diabetes status, drink, smokes, sleep time, and urban/rural residence, with likelihood ratio tests assessing effect heterogeneity across subgroups.

Sensitivity analyses

To verify the robustness of our primary findings and to minimize reverse causality, we performed a sensitivity analysis by excluding participants with self-reported histories of diabetes, heart disease, or stroke at baseline. In this subcohort, univariate RCS models were reconstructed to evaluate the dose-response relationship between baseline CVAI and the risk of incident hypertension without the interference of potential confounders.

Results

Participant characteristics

We included 5,311 participants from the baseline survey of the CHARLS conducted in 2011 as the initial cohort. Baseline characteristic analysis revealed significant differences in demographic and clinical profiles between the elevated BP group (n = 2734, 51.48%) and the normal BP group (n = 2577, 48.52%) (Table 1). Compared with normal BP participants, those with elevated BP exhibited the following: (1) older mean age (58.00 years vs 55.00 years); (2) higher proportion of males (48.2% vs 42.1%); (3) significantly increased BMI (23.28 kg/m² vs 22.29 kg/m²); (4) higher prevalence of drinking and smoking; and (5) higher SBP and DBP. Notably, the elevated BP group demonstrated significantly lower proportions of rural residents (P < 0.05) and a lower prevalence of kidney disease, digestive system disease, and asthma (P < 0.05).

Table 1.

Baseline data of the study participants from the 2011 survey cohort

Characteristics	Total (n = 5311)	Normal BP (n =2577)	Elevated BP (n = 2734)	P-value
Age, year	57.00 (50.00, 63.00)	55.00 (49.00, 61.00)	58.00 (52.00, 64.00)	<0.001
Gender, (%)				<0.001
Female	54.7	57.9	51.8
Male	45.3	42.1	48.2
BMI, (kg/m2)	22.77 (20.67, 25.13)	22.29 (20.35, 24.43)	23.28 (21.07, 25.79)	<0.001
SBP, mmHg	121.00 (110.50, 133.50)	110.00 (104.50, 115.00)	133.00 (126.00, 143.00)	<0.001
DBP, mmHg	72.00 (65.50, 80.00)	66.00 (61.00, 71.00)	79.00 (73.00, 85.00)	<0.001
Smoking, (%)				0.002
No	69.0	71.1	67.0
Yes	31.0	28.9	33.0
Drinking, (%)				0.013
No	66.1	67.8	64.6
Yes	33.9	32.2	35.4
Residential area, (%)				<0.001
Urban	32.3	30.0	34.6
Rural	67.7	70.0	65.4
Education Level, (%)				0.146
Below primary school	48.7	47.9	49.5
Primary school	22.0	21.3	22.6
Middle school	19.3	20.3	18.4
High school and above	10.0	10.5	9.5
Daily sleep time (h)	7.00 (5.00, 8.00)	7.00 (5.00, 8.00)	7.00 (5.00, 8.00)	0.598
Comorbidities, (%)
Diabetes	3.6	3.1	4.0	0.070
Cancer	0.7	0.9	0.6	0.319
Chronic lung disease	8.1	7.6	8.6	0.202
Heart disease	7.4	7.5	7.2	0.729
Stroke	1.1	1.2	1.1	0.819
Psychiatric disorders	1.2	1.3	1.1	0.537
Arthritis	33.2	33.8	32.7	0.412
Dyslipidemia	5.7	5.5	5.9	0.589
Hepatic disease	3.2	3.6	2.8	0.086
Kidney disease	5.1	5.9	4.4	0.018
Digestive system disease	23.1	26.8	19.6	<0.001
Asthma	3.7	3.0	4.4	0.011
Memory-related disease	0.6	0.7	0.5	0.599
CVAI	82.27 (62.73, 114.63)	79.26(56.75, 101.84)	96.61 (70.05, 124.68)	<0.001

BP Blood pressure, BMI Body mass index, SBP Systolic blood pressure, DBP Diastolic blood pressure, CVAI Chinese Visceral Adiposity Index

Characteristics of incident hypertension during follow-up

In the 2020 survey wave, 1,819 participants (34.25%) had incident hypertension. Compared to the non-hypertension group, hypertensive participants were significantly older (58 years vs 56 years, P < 0.001), with higher BMI (23.63 kg/m² vs 22.35 kg/m²), SBP (133.00 mmHg vs 117.00 mmHg), and DBP (76.00 mmHg vs 66.00 mmHg) (all P < 0.001). The hypertensive group included more residents from rural areas (69.9% vs. 66.5%) and participants with lower educational levels. Moreover, the prevalences of diabetes, heart disease, arthritis, dyslipidemia, and asthma was higher in this group (all P < 0.05). Notably, CVAI was markedly elevated in the hypertensive group (100.77 vs 80.91%, P < 0.001) (Table 2).

Table 2.

Characteristics of populations that developed hypertension and those that did not, based on data from 2011

Characteristics	Total (n = 5311)	No occurrence of hypertension (n = 3492)	Occurrence of hypertension (n = 1819)	P-value
Age, year	57.00 (50.00, 63.00)	56.00 (50.00, 62.00)	58.00 (52.50, 64.00)	<0.001
Gender, (%)				0.822
Female	54.7	54.6	55.0
Male	45.3	45.4	45.0
BMI, (kg/m2)	22.77 (20.67, 25.13)	22.35 (20.37, 24.57)	23.63 (21.41, 26.09)	<0.001
SBP, mmHg	121.00 (110.50, 133.50)	117.00 (108.00, 127.00)	131.00 (119.00, 143.50)	<0.001
DBP, mmHg	72.00 (65.50, 80.00)	66.00 (64.00, 77.00)	76.00 (69.50, 85.00)	<0.001
Smoking, (%)				0.482
No	69.0	68.7	69.7
Yes	31.0	31.3	30.3
Drinking, (%)				0.638
No	66.1	66.4	65.7
Yes	33.9	33.6	34.3
Residential area, (%)				0.014
Urban	32.3	33.5	30.1
Rural	67.7	66.5	69.9
Education Level, (%)				<0.001
Below primary school	48.7	46.1	53.8
Primary school	22.0	21.9	22.0
Middle school	19.3	20.9	16.3
High school and above	10.0	11.1	7.9
Daily sleep time (h)	7.00 (5.00, 8.00)	7.00 (5.00, 8.00)	6.00 (5.00, 8.00)	0.094
Comorbidities, (%)
Diabetes	3.6	3.0	4.6	0.003
Cancer	0.7	0.7	0.8	0.868
Chronic lung disease	8.1	7.8	8.6	0.310
Heart disease	7.4	6.4	9.2	<0.001
Stroke	1.1	0.9	1.5	0.083
Psychiatric disorders	1.2	1.1	1.3	0.733
Arthritis	33.2	31.6	36.3	0.001
Dyslipidemia	5.7	5.0	6.9	0.007
Hepatic disease	3.2	3.3	3.0	0.654
Kidney disease	5.1	4.8	5.9	0.089
Digestive system disease	23.1	23.5	22.4	0.395
Asthma	3.7	3.0	5.2	<0.001
Memory-related disease	0.6	0.5	0.7	0.475
CVAI	87.27 (62.73, 114.63)	80.91 (58.07, 106.05)	100.77 (74.91,127.34)	<0.001

BMI Body mass index, SBP Systolic blood pressure, DBP Diastolic blood pressure, CVAI Chinese Visceral Adiposity Index

Prediction of incident hypertension using regression models

Logistic regression models were constructed separately for normal and elevated BP groups (Fig. 2). In fully adjusted Model 3, continuous CVAI was significantly associated with increased hypertension risk in both groups (both P < 0.001). The quartile analysis revealed a progressive increase in risk. In the normal BP group, participants in Q3 (OR = 1.41) and Q4 (OR = 1.77, 95% CI: 1.32–2.38) showed significantly higher risk compared to Q1. This association was markedly stronger in the elevated BP group: the risk for Q4 participants increased threefold (OR = 3.21, 95% CI: 2.26–4.61, P < 0.001), and Q3 showed substantial risk (OR = 2.43), while Q2 showed marginal significance (P = 0.059).

Fig. 2.

Logistic regression models separately constructed for normal BP and elevated BP groups. Model 1 was unadjusted. Model 2 was adjusted for age, sex, smoking status, and alcohol consumption. Model 3 further incorporated educational level, comorbidities, and sleep duration as additional covariates based on Model 2

Nonlinear relationships

Within the fully adjusted framework, RCS analysis was employed to evaluate the dose-response relationship between CVAI and hypertension incidence across the overall population, the normal BP group, and the elevated BP group (Fig. 3). No significant nonlinear association was found between CVAI and hypertension risk (Overall, P nonlinear = 0.130; Elevated BP, P nonlinear = 0.262; normal BP, P nonlinear = 0.164; P overall < 0.001), suggesting a predominantly linear trend.

Fig. 3.

RCS model analysis. Multivariable-adjusted RCS models were employed to analyze the dose-response relationship between hypertension incidence and CVAI in the overall population (A), elevated BP group (B), and normal BP group (C)

Predictive performance of CVAI for hypertension incidence

To assess the ability of the CVAI to predict hypertension incidence, we conducted an ROC analysis and compared it with individual metrics, including BMI, WC, TG, and HDL-C (Fig. 4 and Table 3). The results demonstrated that the CVAI outperformed traditional adiposity markers but remained only moderately accurate (Delong test, Table 4). Furthermore, by constructing three logistic regression models and comparing their predictive performances, ROC curve analysis revealed that Model 3, which incorporated multiple clinical variables, achieved the highest predictive accuracy (Fig. 5).

Fig. 4.

ROC curves comparing predictive performance of CVAI, BMI, WC, TG, and HDL-C for hypertension incidence

Table 3.

ROC curves comparing the predictive performance of CVAI, BMI, WC, TG, and HDL-C for hypertension incidence

Category	AUC	95% CI	Sensitivity	Specificity
CVAI	0.638	0.623–0.654	0.524	0.690
BMI	0.605	0.589–0.621	0.540	0.623
WC	0.615	0.599–0.631	0.542	0.646
TG	0.571	0.555–0.587	0.580	0.535
HDL-C	0.538	0.521–0.554	0.508	0.560

Table 4.

Comparison of ROC curves (Delong test) between CVAI and conventional metrics for predicting hypertension incidence

Comparison	Difference of AUC	Standard error	95% CI	Z-value	P-value
CVAI vs BMI	0.033	0.006	0.021–0.045	5.520	<0.001
CVAI vs WC	0.024	0.004	0.015–0.032	5.721	<0.001
CVAI vs TG	0.068	0.008	0.052–0.084	8.276	<0.001
CVAI vs HDL-C	0.101	0.008	0.084–0.117	12.211	<0.001

Fig. 5.

Comparison of predictive performance among three logistic regression models for hypertension incidence

Subgroup analyses

We performed subgroup analyses in individuals with normal and elevated BP to evaluate the applicability of the CVAI across different populations. The analyses were adjusted for potential confounders including sex, age, residential area, educational level, sleep duration, BMI, and diabetes status. In the normal BP group, a significant interaction was observed between sex and CVAI (P for interaction < 0.001). The association appeared to be slightly stronger in males (OR = 1.02, 95% CI: 1.01–1.02) compared to females (OR = 1.01, 95% CI: 1.00–1.01) (Table 5). In the elevated BP group, age significantly modified the association between CVAI and hypertension risk (P for interaction < 0.001) (Table 6).

Table 5.

Association between CVAI and hypertension risk by subgroup in individuals with normal BP

Variables	n (%)	OR (95% CI)	P	P for interaction
Normal BP group	2577 (100.00)	1.01 (1.01~1.01)	<0.001
Age				0.860
45–60	1908 (74.04)	1.01 (1.01~1.01)	<0.001
>60	669 (25.96)	1.01 (1.01~1.02)	<0.001
Sleep time				0.829
<7	1278 (49.59)	1.01 (1.01~1.02)	<0.001
7–9	1189 (46.14)	1.01 (1.01~1.01)	<0.001
>9	110 (4.27)	1.01 (1.00~1.03)	0.036
BMI				0.749
<18.5	203 (7.88)	1.01 (0.99~1.03)	0.218
18.5–24.9	1850 (71.79)	1.01 (1.01~1.01)	<0.001
25–30	469 (18.20)	1.01 (1.00~1.02)	0.003
>30	55 (2.13)	1.00 (0.99~1.02)	0.606
Gender				0.001
Female	1086 (42.14)	1.01 (1.00~1.01)	<0.001
Male	1491 (57.86)	1.02 (1.01~1.02)	<0.001
Diabetes				0.704
NO	2498 (96.93)	1.01 (1.01~1.01)	<0.001
YES	79 (3.07)	1.01 (1.00~1.03)	0.042
Residential area				0.674
Urban	772 (29.96)	1.01 (1.01~1.02)	<0.001
Rural	1805 (70.04)	1.01 (1.01~1.02)	<0.001
Education level				0.395
Below primary school	1235 (47.92)	1.01 (1.01~1.02)	<0.001
Primary school	549 (21.30)	1.01 (1.00~1.01)	0.012
Middle school	522 (20.26)	1.01 (1.00~1.02)	<0.001
High school and above	271 (10.52)	1.01 (1.00~1.02)	0.021

Table 6.

Association between CVAI and hypertension risk by subgroup in individuals with elevated BP

Variables	n (%)	OR (95% CI)	P	P for interaction
Elevated BP group	2734 (100.00)	1.01 (1.01~1.01)	<0.001
Age				0.016
45–60	1674 (61.23)	1.01 (1.01~1.01)	<0.001
>60	1060 (38.77)	1.01 (1.00~1.01)	<0.001
Sleep time				0.736
<7	1360 (49.74)	1.01 (1.01~1.01)	<0.001
7–9	1257 (45.98)	1.01 (1.01~1.01)	<0.001
>9	117 (4.28)	1.01 (1.00~1.02)	0.142
BMI				0.571
<18.5	158 (5.78)	1.01 (1.00~1.02)	0.194
18.5–24.9	1721 (62.95)	1.01 (1.01~1.01)	<0.001
25–30	742 (27.14)	1.01 (1.00~1.02)	<0.001
>30	113 (4.13)	1.02 (1.01~1.03)	0.006
Gender				0.941
Female	1318 (48.21)	1.01 (1.01~1.01)	<0.001
Male	1416 (51.79)	1.01 (1.01~1.01)	<0.001
Diabetes				0.960
NO	2624 (95.98)	1.01 (1.01~1.01)	<0.001
YES	110 (4.02)	1.01 (1.00~1.02)	0.054
Residential area				0.776
Urban	946 (34.60)	1.01 (1.01~1.01)	<0.001
Rural	1788 (65.40)	1.01 (1.01~1.01)	<0.001
Education level				0.224
Below primary school	1354 (49.52)	1.01 (1.01~1.01)	<0.001
Primary school	617 (22.57)	1.01 (1.00~1.01)	<0.001
Middle school	504 (18.43)	1.01 (1.01~1.02)	<0.001
High school and above	259 (9.47)	1.02 (1.01~1.02)	<0.001

Sensitivity analysis

To verify the robustness of our findings, we performed a sensitivity analysis by excluding participants with a baseline history of diabetes, heart disease, or stroke. Univariate RCS models were used to evaluate the dose-response relationship between CVAI and risk of hypertension. The characteristics of the population included in the sensitivity analysis are presented in Supplementary Materials 2 and 3. The results indicated that the linear relationship between CVAI and the incidence of hypertension persisted in both the overall population (Fig. 6A) and the subgroups with elevated and normal BP (Fig. 6B, C).

Fig. 6.

Sensitivity analysis of the dose-response relationship between CVAI and risk of hypertension using RCS models. Participants with a baseline history of diabetes, heart disease, and stroke were excluded to verify the robustness of the results. The models were unadjusted (univariate analysis). A Overall population. B Participants with elevated BP. C Participants with normal BP

Discussion

Among 5,311 participants, 34.25% developed hypertension during the 9-year follow-up period. CVAI emerged as an independent risk factor for incident hypertension. A linear dose-response relationship was observed between CVAI and hypertension risk in both groups. These associations remained robust and linear in sensitivity analyses after excluding participants with diabetes, heart disease, or stroke.

With global shifts in dietary patterns and living standards, the increasing prevalence of obesity has markedly increased the risk of metabolic disorders [19]. Epidemiological studies have shown a strong association between obesity and hypertension [20]. For instance, National Health and Nutrition Examination Survey data revealed that hypertension prevalence was significantly higher in obese individuals (42.5%) compared to the non-obese population (15.3%) [21]. Notably, while overall adolescent hypertension rates declined from 2001 to 2016, the risk remained persistently elevated among obese adolescents [22].

Pathophysiologically, the interaction between obesity and hypertension involves complex multisystem regulation [23]. In addition to genetic and environmental factors, overactivation of the sympathetic nervous system is a central mechanism. Elevated adrenergic activity is observed even in normotensive obese individuals and is synergistically enhanced when obesity coexists with hypertension [23, 24]. Visceral fat accumulation positively correlates with the degree of sympathetic activation [25]. This association drives hypertension progression through multiple pathways, including renal physical compression and metabolic abnormalities such as insulin resistance and inflammatory cytokine release [26]. Collectively, these mechanisms indicate that obesity-related hypertension is a multifactorial process involving neuroendocrine dysregulation, abnormal fat distribution, and target organ damage.

In clinical practice and epidemiological research, BMI and WC have long been regarded as the cornerstone metrics for obesity assessment. However, these conventional indices have inherent limitations. BMI fails to distinguish between adipose and lean tissue masses or reflects fat distribution patterns [27]. Similarly, WC merely provides an estimate of general abdominal adiposity, lacking the capacity to discriminate visceral from subcutaneous fat deposits [28, 29].

In contrast, the CVAI, a novel metric incorporating age, BMI, WC, and lipid profiles, offers superior accuracy in quantifying visceral fat content, specifically in Chinese populations [30–32]. Our results corroborate substantial evidence, such as the longitudinal findings by Niu et al. [33] and a large-scale investigation by Han et al. [34], reaffirming that the CVAI is an independent risk factor for hypertension development. Furthermore, comparative studies have consistently highlighted the advantages of the CVAI. For instance, Gui et al. [35] demonstrated that the CVAI outperformed other obesity indices in identifying hypertension risk, whereas Li et al. [36] revealed that the CVAI exhibited significantly stronger associations with hypertension onset than traditional indicators, such as the VAI, BMI, WC, and waist-to-hip ratio.

Crucially, beyond validating these established associations, our study innovatively stratified participants according to the 2024 ESC of Cardiology guidelines (normal vs. elevated BP). This approach elucidated, for the first time, the differential predictive utility of CVAI across distinct BP statuses. These findings provide novel evidence-based insights that facilitate the shift toward precise prevention and management of obesity-related hypertension.

The predictive superiority of CVAI is likely attributable to its precise reflection of the biological characteristics of visceral adipose tissue (VAT). Anatomically, VAT represents a typical ectopic fat depot, and its pathological accumulation occurs in lean tissues (e.g., liver, heart [including pericardial, epicardial, and intramyocardial fat], and skeletal muscle) [37–39]. This ectopic deposition promotes the progression of hypertension through multiple pathways [40]. First, VAT overproduces proinflammatory cytokines (e.g., IL-6 and TNF-α) [41–43], inducing chronic low-grade inflammation and insulin resistance. This subsequently activates the renin–angiotensin–aldosterone system (RAAS) and sympathetic nervous system, ultimately leading to hypertension [44]. Second, VAT accumulation triggers leptin resistance and hypoadiponectinemia, exacerbating BP elevation via hypothalamic–pituitary–adrenal (HPA) axis hyperactivity and vascular endothelial dysfunction [45–47]. Significant ethnic disparities exist in body fat distribution. Asian populations exhibit pronounced visceral fat accumulation at lower BMI thresholds than Caucasian populations [48, 49], limiting the predictive utility of conventional obesity indices. These mechanistic insights not only explain the superiority of CVAI in hypertension risk prediction but also provide biological plausibility for the robust association between CVAI and hypertension risk observed in our study.

This study utilized large-scale longitudinal data from the CHARLS to systematically investigate the relationship between CVAI and hypertension incidence across populations with diverse baseline BP statuses over a 9-year follow-up period. The results demonstrated that the CVAI was positively associated with hypertension risk, regardless of baseline BP (all P for trend < 0.001), and this association remained robust after adjusting for age, sex, lifestyle, and clinical characteristics. RCS analyses further confirmed a clear linear dose-response relationship (all P for linear < 0.001; P for nonlinear > 0.05). Importantly, the sensitivity analysis reinforced the robustness of these findings. Even after excluding participants with baseline diabetes, heart disease, and stroke, the linear dose-response relationship persisted in both the normal and elevated BP groups. This indicates that the impact of visceral adiposity on BP progression is independent of preexisting cardiometabolic multimorbidity. The association was markedly stronger in the elevated BP group (adjusted OR = 3.21, 95% CI: 2.26–4.61) compared to the normal BP group (adjusted OR = 1.77, 95% CI: 1.32–2.38). This suggests that visceral adiposity accumulation may exert a synergistic effect in individuals with preexisting hemodynamic strain, substantially accelerating the progression to hypertension. Consistent with previous reports by Wen et al. [50] and Gao et al. [51], our study provides new longitudinal evidence supporting a causal link between visceral fat accumulation and the onset of hypertension. Clinically, our analysis highlights the CVAI as a promising instrument for early screening. Although the relative risk was lower in the population with normal BP, applying the CVAI to this group is critical for primary prevention, enabling the identification of at-risk individuals before BP elevation occurs.

Understanding population heterogeneity is crucial for precision prevention. Our subgroup analyses revealed distinct risk patterns based on the baseline BP status. In the normal BP group, sex significantly modified the association between CVAI and incident hypertension. This association was stronger in males than in females. Pathophysiologically, this aligns with the male predisposition to visceral (“android”) obesity, whereas premenopausal females are generally protected by estrogen, favoring subcutaneous fat storage that buffers metabolic disturbances [28]. Clinically, this suggests that males are metabolically more sensitive to hemodynamic stress in the early stages of normotension, warranting prioritized CVAI intervention, even for men with normal BP.

In contrast, in the elevated BP group, age emerged as a significant effect modifier. This indicates that once BP is elevated, the pathogenic impact of visceral adiposity interacts with age-related vascular aging processes [52]. Collectively, these findings support a shift from “one-size-fits-all” prevention to targeted interventions tailored to individual profiles, focusing on males in the normal BP stage and age-stratified management in the elevated BP stage.

This study had some limitations. First, a comparison of baseline characteristics between the included and excluded participants (Supplementary Material 1) revealed that the excluded group was significantly older and had a higher prevalence of comorbidities (e.g., diabetes and heart disease). This indicates a potential selection bias favoring a relatively healthier cohort, which may affect the generalizability of frail populations. Second, the diagnosis of hypertension relied partly on self-reporting, potentially introducing a recall bias. Third, the study was confined to middle-aged and older Chinese adults, limiting the extrapolation to younger cohorts or other ethnicities. Finally, despite multivariate adjustments, residual confounding from unmeasured factors (e.g., genetics and physical activity intensity) remains possible. Notwithstanding these limitations, our findings provide pioneering evidence on the differential predictive utility of the CVAI across BP statuses, addressing a critical gap in prior research. Future studies should validate these findings using large-scale multinational cohorts.

Conclusions

The CVAI serves as a robust predictor of hypertension risk, exhibiting a clear linear dose-response relationship that persists even after excluding cardiometabolic comorbidities. Our findings highlight the differential utility of CVAI; while it signals a synergistic risk in individuals with elevated BP, it is critically valuable for early risk stratification in normotensive populations. Given its simplicity and cost-effectiveness, integrating the CVAI into widespread community screening programs could optimize precision prevention strategies and enable timely interventions, particularly in resource-limited settings.

Abbreviations

BP

Blood pressure

CVAI

Chinese Visceral Adiposity Index

CHARLS

China Health and Retirement Longitudinal Study

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

BP

Blood pressure

TG

Triglycerides

WC

Waist circumference

BMI

Body mass index

HDL-C

High-density lipoprotein cholesterol

RCS

Restricted cubic spline

ROC

Receiver operating characteristic

OR

Odds ratio

Authors’ contributions

Xingyun Yang and Zhengyang Tang contributed to the study conception and design. Zhengyang Tang performed data analysis. Xingyun Yang prepared the draft of the article. Zongyi Jiang performed literature collection and curation. Aihua Fei reviewed and revised the manuscript. All authors contributed to the article and approved the submitted version.

Funding

No financial support was received for this study.

Data availability

The data that support the findings of this study are available from the China Health and Retirement Longitudinal Study (CHARLS) public repository: [http://charls.pku.edu.cn/] (http://charls.pku.edu.cn/). Access is granted upon user registration and agreement to the data use terms.

Declarations

Ethics approval and consent to participate

The CHARLS study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of Peking University (IRB00001052-11015). All participants provided written informed consent before participating in the CHARLS study.

Consent for publication

No applicable

Competing interests

The authors declare no competing interests.