The relationship between cholesterol,high-density lipoprotein,and glucose index and hypertension: A study based on two national cohorts

Abstract

Background

Hypertension remains a major global public health concern and is strongly linked to metabolic dysregulation. The cholesterol–high-density lipoprotein–glucose index, a recently proposed composite marker that integrates lipid and glycemic parameters, has not been comprehensively investigated in relation to hypertension.

Methods

Data from two nationally representative cohorts were analyzed: the China Health and Retirement Longitudinal Study (CHARLS) for longitudinal evaluation and the US National Health and Nutrition Examination Survey (NHANES) for cross-sectional assessment. Logistic regression analyses were used to examine the association between the cholesterol–high-density lipoprotein–glucose index and hypertension. Restricted cubic spline models were applied to assess potential non-linear relationships, while receiver operating characteristic curves were employed to evaluate the discriminative capacity of the cholesterol–high-density lipoprotein–glucose index. Sensitivity analyses were conducted to confirm the robustness of the findings.

Results

A total of 4031 participants from CHARLS and 10,355 participants from NHANES were included. Higher cholesterol–high-density lipoprotein–glucose index values were significantly and independently associated with an increased risk of hypertension in both cohorts. In the CHARLS cohort, each one-unit increase in cholesterol–high-density lipoprotein–glucose corresponded to a 1.60-fold higher risk of hypertension after full adjustment, while in the NHANES cohort, the risk was 1.73-fold higher. Restricted cubic spline analysis indicated a linear association between cholesterol–high-density lipoprotein–glucose and hypertension. Receiver operating characteristic curve analyses demonstrated modest discriminative ability (AUC: 0.566 in CHARLS; 0.612 in NHANES). Sensitivity analyses supported the consistency and stability of these results.

Conclusion

The cholesterol–high-density lipoprotein–glucose index shows an independent and positive association with hypertension across diverse populations. However, its discriminative power as a standalone marker remains modest and should be interpreted with caution. Further prospective and mechanistic studies are needed to validate its clinical applicability and to explore the potential value of integrating cholesterol–high-density lipoprotein–glucose with other cardiometabolic risk indicators.

Keywords

CHG index hypertension cross-sectional study NHANES CHARLS

Introduction

Hypertension (HTN) continues to be one of the most critical global public health concerns, affecting more than one billion adults worldwide. It represents the leading modifiable risk factor for cardiovascular disease (CVD) and premature mortality.^1–3 Despite extensive public health initiatives and advances in medical management, HTN control rates remain far from optimal. According to the World Health Organization's 2023 report, only about half of hypertensive adults are aware of their condition, and an even smaller proportion achieve adequate blood pressure control.¹ Mounting evidence indicates that metabolic abnormalities, such as hyperglycemia and dyslipidemia, play central roles in the onset and progression of HTN. These metabolic disturbances contribute to elevated blood pressure through several interrelated mechanisms, including sympathetic nervous system activation, endothelial dysfunction, and vascular remodeling.^4–10 As a result, identifying integrative metabolic markers that capture these disturbances may help improve the early detection of risk and the development of preventive strategies for HTN.

The cholesterol–high-density lipoprotein–glucose (CHG) index, defined as Ln [TC (mg/dL) × FBG (mg/dL)/(2×HDL (mg/dL)], is a composite biomarker derived from total cholesterol, HDL cholesterol (HDL-C), and fasting blood glucose. Originally developed for type 2 diabetes screening, the CHG index has since been investigated in relation to cardiovascular risk, diabetic microvascular complications, metabolic syndrome, and all-cause mortality.^11–14 However, its potential association with HTN has not yet been clearly established, nor is it known whether this relationship is independent of traditional metabolic risk factors. Understanding this association may provide valuable insights into the metabolic mechanisms underlying blood pressure regulation and could help refine the stratification of HTN risk.

To address this knowledge gap, the present study employed data from two nationally representative cohorts, the China Health and Retirement Longitudinal Study (CHARLS) and the US National Health and Nutrition Examination Survey (NHANES), to explore the relationship between the CHG index and HTN. We hypothesized that higher CHG index values would be independently and positively associated with an increased risk of HTN. Clarifying this relationship may improve our understanding of the metabolic pathways that influence HTN and inform the development of more precise prevention and risk assessment strategies in clinical practice.

Materials and methods

Study design, data sources, and preprocessing

This study utilized data from two nationally representative cohorts: the China Health and Retirement Longitudinal Study (CHARLS; https://charls.pku.edu.cn/) and the US National Health and Nutrition Examination Survey (NHANES; https://www.cdc.gov/nchs/nhanes/). All procedures adhered to the ethical principles outlined in the Declaration of Helsinki (1975; revised 2024) and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.¹⁵ As both datasets are publicly accessible and fully de-identified, ethical approval was waived by the Ethics Committee of the Affiliated People's Hospital of Fujian University of Traditional Chinese Medicine. The respective institutional review boards approved the original survey protocols, and all participants provided written informed consent (CHARLS: Institutional Review Board of Peking University¹⁶; NHANES: Institutional Review Board of the National Center for Health Statistics).¹⁷ Since this investigation represents a secondary analysis of existing survey data, no formal sample size calculation was performed. Instead, all eligible participants meeting the predefined inclusion criteria were included to enhance the precision and generalizability of the findings.

CHARLS is a nationally representative longitudinal study of Chinese adults aged 45 years and older, designed to capture detailed information on health status, socioeconomic factors, and family structure.¹⁸ For the present analysis, data from Wave 1 (2011, baseline) and Wave 3 (2015, follow-up) were used, as these waves included all necessary biochemical measurements. Participants with HTN at baseline were excluded, and normotensive individuals were followed for approximately 4 years to assess incident HTN. Initially, 8489 individuals were identified. Exclusions were made for participants aged <45 years (n = 268), those missing CHG index components (n = 3172), or lacking HTN data (n = 1274). Additional exclusions were applied to records missing information on sex (n = 2), education (n = 3), smoking status (n = 5), or drinking status (n = 2). The final analytic sample included 4031 participants (Figure 1a).

Figure 1.

The flowchart of subject screening in CHARLS (a) and NHANES (b). CHARLS: China Health and Retirement Longitudinal Study; NHANES: National Health and Nutrition Examination Survey; CHG index: cholesterol, high-density lipoprotein, and glucose index; HTN: hypertension; T2DM: type 2 diabetes mellitus; PIR: poverty income ratio.

NHANES is a cross-sectional, nationally representative survey conducted by the Centers for Disease Control and Prevention (CDC) to evaluate the health and nutritional status of the US population.¹⁹ For this study, data from five continuous survey cycles (2009–2018) were combined. Among 49,693 participants, individuals with missing CHG index components (n = 34,426), education (n = 2707), marital status (n = 5), poverty income ratio (PIR) (n = 1209), drinking (n = 984), smoking (n = 6), or type 2 diabetes mellitus (T2DM) status (n = 1) were excluded. The final analytic sample comprised 10,355 participants (Figure 1b).

Definition of primary variables

The (CHG) index¹⁴ was developed by integrating TC, HDL, and FBG into the formula: Ln [TC (mg/dL) * FBG (mg/dL)/2 * HDL (mg/dL)].

HTN was defined by meeting any of the following criteria²⁰: (1) a diagnosis by a physician or other healthcare professional; (2) systolic blood pressure (SBP) ≥ 130 mmHg or diastolic blood pressure (DBP) ≥ 80 mmHg; or (3) current use of antihypertensive medication.

Definition of covariates

Both cohorts included the following covariates: sex, age, educational level (junior high school or below, senior high school, and tertiary education), marital status (married/cohabiting, separated/divorced/widowed, and never married), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), smoking status (yes/no), drinking status (yes/no), and T2DM (yes/no).

For the NHANES cohort, additional covariates were included: race/ethnicity, poverty income ratio (PIR), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), and alkaline phosphatase (ALP). Socioeconomic status was assessed based on income level relative to the poverty threshold. Participants were categorized as low income (PIR ≤ 1), middle income (PIR between 1 and < 4), or high income (PIR ≥ 4).²¹

T2DM was defined according to the following criteria²²:

self-reported physician diagnosis of diabetes;

glycated hemoglobin (HbA1c) ≥ 6.5%;

fasting plasma glucose ≥ 126 mg/dL;

current use of insulin; or

current use of oral antidiabetic medications.

Statistical analysis

Standard, unweighted methods were applied to the analyses of the CHARLS dataset.²³ Given the complex, multistage sampling design of NHANES, all analyses for that cohort incorporated sampling weights to ensure population representativeness. Following NHANES analytical guidelines, the original two-year sample weights were adjusted by multiplying them by one-fifth (1/5) to correspond to the combined 10-year analytical period.^24,25

Baseline characteristics were summarized as means ± standard deviations (SD) for continuous variables and as frequencies with corresponding percentages (n, %) for categorical variables. Between-group differences were assessed using Student's t-tests for continuous variables and chi-square tests for categorical variables. To evaluate the relationship between the CHG index and HTN, three weighted logistic regression models were developed.

Model 1: Unadjusted (crude model).

Model 2: Adjusted for age, sex, education level, and marital status.

Model 3: Fully adjusted, including all covariates from Model 2 plus TG, LDL-C, smoking status, and alcohol consumption. For the NHANES dataset, Model 3 was further adjusted for race/ethnicity, PIR, ALT, AST, ALP, and GGT.

Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to quantify associations across models. Restricted cubic spline (RCS) regression was performed to explore potential non-linear associations between the CHG index and HTN. Subgroup analyses were further conducted to determine whether demographic or clinical variables modified the observed associations. Subgroup factors included age, sex, race, education level, marital status, PIR, ALT, AST, ALP, GGT, TG, insulin levels, alcohol consumption, and smoking status. Multicollinearity was evaluated using variance inflation factors (VIFs), with values <10 indicating acceptable levels of collinearity.^26,27

To address missing data, multiple imputation was conducted using the “MICE” package in R, consistent with previous studies.^28–30 All multivariable analyses were performed using the imputed datasets.³¹ To assess the robustness of the findings, two sensitivity analyses were carried out: (1) analyses based on the original (non-imputed) datasets in both CHARLS and NHANES; and (2) analyses using unweighted NHANES data.

The discriminative performance of the CHG index for HTN was examined using receiver operating characteristic (ROC) curve analysis, with the area under the curve (AUC) representing diagnostic accuracy. The optimal cutoff value was identified by maximizing the Youden index.³² Model calibration was further assessed using the Brier score and calibration plots generated by bootstrap resampling (1000 iterations). The Brier score reflects the overall accuracy of probabilistic predictions, with lower values indicating better performance. Scores below 0.25 are considered indicative of acceptable predictive ability. Calibration plots were used to visually assess the agreement between predicted and observed outcomes.^33,³⁴ All statistical analyses were performed using R software (version 4.4.1), and a two-tailed P-value < 0.05 was considered statistically significant.

Results

Baseline characteristics of participants

The baseline characteristics of participants from both cohorts are summarized in Tables 1 and 2. In the CHARLS cohort, a total of 4145 participants were included, comprising 1309 individuals with HTN and 2836 without HTN. Compared with the non-HTN group, participants with HTN were significantly older. They showed higher levels of TG, LDL-C, and CHG index values (all P < 0.05). Significant group differences were also observed in sex distribution, smoking and alcohol consumption behaviors, prevalence of T2DM, and marital status (all P < 0.05). However, educational level did not differ significantly between the two groups (P = 0.607).

Table 1.

Baseline characteristics of CHARLS participants.

Variables	All(N = 4145)	Non-HNT group (N = 2836)	HNT group(N = 1309)	P
Age, years, mean (SD)	57.45 (8.53)	56.76 (8.14)	58.93 (9.14)	<0.001
TG, mg/L, mean (SD)	121.45 (91.39)	117.77 (92.02)	129.31 (89.56)	<0.001
LDL, mg/L, mean (SD)	113.91 (32.69)	112.86 (32.00)	116.13 (34.00)	0.003
CHG index, mean (SD)	5.25 (0.41)	5.23 (0.40)	5.32 (0.42)	<0.001
Gender, N (%)				<0.001
Female	2174 (53.9)	1539 (56.1)	635 (49.3)
Male	1857 (46.1)	1204 (43.9)	653 (50.7)
Smoking, N (%)				0.003
No	2482 (61.6)	1732 (63.1)	750 (58.2)
Yes	1549 (38.4)	1011 (36.9)	538 (41.8)
Drinking, N (%)				0.008
No	2677 (66.4)	1859 (67.8)	818 (63.5)
Yes	1354 (33.6)	884 (32.2)	470 (36.5)
T2DM, N (%)				0.012
No	3542 (87.9)	2435 (88.8)	1107 (85.9)
Yes	489 (12.1)	308 (11.2)	181 (14.1)
Education, N (%)				0.607
Junior high school and below	3623 (89.9)	2460 (89.7)	1163 (90.3)
Senior high school	361 (9.0)	248 (9.0)	113 (8.8)
Tertiary	47 (1.2)	35 (1.3)	12 (0.9)
Marital status, N (%)				0.003
Married/cohabiting	3681 (91.3)	2533 (92.3)	1148 (89.1)
Never married	33 (0.8)	18 (0.7)	15 (1.2)
Separated/divorced/widowed	317 (7.9)	192 (7.0)	125 (9.7)

CHARLS: China Health and Retirement Longitudinal Study; CHG index: cholesterol–high-density lipoprotein–glucose index; LDL: low-density lipoprotein; TGs: triglycerides; T2DM: type 2 diabetes mellitus.

Table 2.

Baseline characteristics of NHANES participants.

Variables	All(N = 10,355)	Non-HNT group (N = 4986)	HNT group(N = 5369)	P
Age, years, mean (SD)	49.94 (17.66)	41.49 (15.84)	57.79 (15.51)	<0.001
ALT, IU/L, mean (SD)	24.86 (18.17)	23.73 (17.21)	25.91 (18.96)	<0.001
AST, IU/L, mean (SD)	25.26 (20.28)	24.06 (17.65)	26.37 (22.40)	<0.001
ALP, IU/L, mean (SD)	70.33 (25.79)	66.74 (21.54)	73.67 (28.80)	<0.001
GGT, IU/L, mean (SD)	29.47 (40.81)	24.07 (27.27)	34.49 (49.69)	<0.001
TG, mg/L, mean (SD)	129.61(121.16)	117.02 (130.64)	141.31(110.36)	<0.001
LDL, mg/L, mean (SD)	112.84 (35.75)	112.55 (34.44)	113.11 (36.93)	0.058
CHG index, mean (SD)	5.26 (0.44)	5.17 (0.40)	5.34 (0.45)	<0.001
Gender, N (%)				<0.001
Female	5258 (50.8)	2722 (54.6)	2536 (47.2)
Male	5097 (49.2)	2264 (45.4)	2833 (52.8)
PIR, N (%)				0.066
High income	2599 (25.1)	1314 (26.4)	1285 (23.9)
Middle income	5477 (52.9)	2551 (51.2)	2926 (54.5)
Low income	2279 (22.0)	1121 (22.5)	1158 (21.6)
Smoking, N (%)				<0.001
No	5728 (55.3)	3002 (60.2)	2726 (50.8)
Yes	4627 (44.7)	1984 (39.8)	2643 (49.2)
Drinking, N (%)				<0.001
No	2480 (23.9)	1093 (21.9)	1387 (25.8)
Yes	7875 (76.1)	3893 (78.1)	3982 (74.2)
T2DM, N (%)				<0.001
No	7922 (76.5)	4333 (86.9)	3589 (66.8)
Yes	2433 (23.5)	653 (13.1)	1780 (33.2)
Race, N (%)				<0.001
Mexican American	1480 (14.3)	842 (16.9)	638 (11.9)
Non-Hispanic Black	2042 (19.7)	735 (14.7)	1307 (24.3)
Non-Hispanic White	4460 (43.1)	2118 (42.5)	2342 (43.6)
Other Hispanic	1046 (10.1)	550 (11.0)	496 (9.2)
Other Race – including multi-racial	1327 (12.8)	741 (14.9)	586 (10.9)
Education, N (%)				<0.001
Junior high school and below	2285 (22.1)	982 (19.7)	1303 (24.3)
Senior high school	5508 (53.2)	2544 (51.0)	2964 (55.2)
Tertiary	2562 (24.7)	1460 (29.3)	1102 (20.5)
Marital status, N (%)				<0.001
Married/cohabiting	6159 (59.5)	2958 (59.3)	3201 (59.6)
Never married	1888 (18.2)	1250 (25.1)	638 (11.9)
Separated/divorced/widowed	2308 (22.3)	778 (15.6)	1530 (28.5)

NHANES: National Health and Nutrition Examination Survey; CHG index: cholesterol–high-density lipoprotein–glucose index; TGs: triglycerides; T2DM: type 2 diabetes mellitus; ALT: alanine aminotransferase; ALP: alkaline phosphatase; AST: aspartate aminotransferase; GGT: gamma-glutamyl transferase; PIR: poverty income ratio.

In the NHANES cohort, 10,355 participants were analyzed, including 5369 with HTN and 4986 without HTN. Individuals with HTN had significantly higher mean levels of age, ALT, AST, ALP, GGT, TG, LDL-C, and CHG index compared with their non-hypertensive counterparts (all P < 0.05). Similarly, significant differences were observed in sex, race, marital status, smoking and drinking behaviors, and T2DM prevalence (all P < 0.05). In comparison, no significant difference was found in the PIR between the two groups (P > 0.05). Unlike the CHARLS findings, the NHANES dataset revealed a statistically significant difference in educational level between participants with HTN and those without HTN (P < 0.05).

Moreover, the VIF analysis indicated that all covariates had VIF values below 5, suggesting no evidence of multicollinearity. The detailed VIF results are presented in Supplemental Tables 1 and 2. Overall, across both cohorts, individuals with HTN consistently showed higher CHG index values, aligning with the primary hypothesis of this study.

Association between CHG index and HTN

In the CHARLS cohort, the CHG index, analyzed as a continuous variable, demonstrated a significant positive association with the risk of HTN. In the unadjusted model, each one-unit increase in the CHG index corresponded to a 1.69-fold higher risk of HTN (OR = 1.69, 95% CI: 1.44–1.98, P < 0.001). After adjusting for age, sex, education level, and marital status (basic model), the association remained essentially unchanged (OR = 1.70, 95% CI: 1.45–2.00, P < 0.001). Further adjustment for T2DM, LDL-C, TG, alcohol consumption, and smoking (fully adjusted model) continued to show a significant relationship, with a 1.60-fold higher risk of HTN (OR = 1.60, 95% CI: 1.22–2.09, P < 0.001).

Quartile-based analyses of the CHG index confirmed this trend. Compared with participants in the lowest quartile (Q1), those in the highest quartile (Q4) showed significantly elevated HTN risks of 1.88-fold (95% CI: 1.55–2.27), 1.89-fold (95% CI: 1.56–2.29), and 1.68-fold (95% CI: 1.30–2.18) in the unadjusted, basic, and fully adjusted models, respectively (all P < 0.001) (Table 3).

Table 3.

Logistic regression between CHG index and HTN (CHARLS).

CHARLS	Model 1OR (95% CI)[P value]	Model 2OR (95% CI)[P value]	Model 3OR (95% CI)[P value]
Crude model	1.69 (1.44–1.98)[P < 0.001]	1.70 (1.45–2.00)[P < 0.001]	1.60 (1.22–2.09) [P < 0.001]
Q1	Reference	Reference	Reference
Q2	1.26 (1.04–1.54)[P = 0.019]	1.26 (1.03–1.53)[P = 0.023]	1.22 (0.99–1.50) [P = 0.058]
Q3	1.47 (1.21–1.78)[P < 0.001]	1.52 (1.25–1.85)[P < 0.001]	1.43 (1.15–1.78) [P = 0.001]
Q4	1.88 (1.55–2.27)[P < 0.001]	1.89 (1.56–2.29)[P < 0.001]	1.68 (1.30–2.18) [P < 0.001]

CHG index: cholesterol, high-density lipoprotein, and glucose index; HTN: hypertension; CHARLS: China Health and Retirement Longitudinal Study; T2DM: type 2 diabetes mellitus; LDL: low-density lipoprotein; TGs: triglycerides; OR: odds ratio; CI: confidence interval.

Model 1: Non-adjusted model.

Model 2: Adjusted for age, gender, education, and marital status.

Model 3: Adjusted for age, gender, education, marital status, T2DM, LDL, TG, drinking, and smoking.

Similarly, in the NHANES cohort, a consistent and significant positive association was observed between the CHG index and HTN risk. In the unadjusted model, each one-unit increase in CHG was associated with a 3.06-fold higher likelihood of HTN (OR = 3.06, 95% CI: 2.62–3.58, P < 0.0001). After adjusting for age, sex, education, and marital status, the risk remained elevated (OR = 2.40, 95% CI: 2.06–2.79, P < 0.0001). Following full adjustment for race, T2DM, LDL-C, TG, PIR, liver function markers (ALT, AST, ALP, and GGT), smoking, and alcohol consumption, the association persisted, with a 1.73-fold higher risk of HTN (OR = 1.73, 95% CI: 1.35–2.21, P < 0.0001).

Quartile comparisons yielded comparable findings: relative to Q1, participants in Q4 demonstrated significantly higher risks of HTN, 3.41-fold (95% CI: 2.88–4.03), 2.65-fold (95% CI: 2.20–3.18), and 2.42-fold (95% CI: 1.77–3.30), across the unadjusted, basic, and fully adjusted models, respectively (all P < 0.001) (Table 4).

Table 4.

Logistic regression between CHG index and HTN (NHANES).

NHANES	Model 1OR (95% CI)[P value]	Model 2OR (95% CI)[P value]	Model 3OR (95% CI)[P value]
Crude model	3.06 (2.62–3.58)[P < 0.001]	2.40 (2.06–2.79)[P < 0.001]	1.73 (1.35–2.21)[P < 0.001]
Q1	Reference	Reference	Reference
Q2	1.57 (1.37–1.79)[P < 0.001]	1.39 (1.18–1.63)[P < 0.001]	1.36 (1.14–1.62)[P = 0.001]
Q3	2.22 (1.90–2.59)[ P < 0.001]	1.83 (1.54–2.18)[ P < 0.001]	1.79 (1.48–2.16)[P < 0.001]
Q4	3.41 (2.88–4.03)[P < 0.001]	2.65 (2.20–3.18)[P < 0.001]	2.42 (1.77–3.30)[P < 0.001]

CHG index: cholesterol, high-density lipoprotein, and glucose index; HTN: hypertension; NHANES: National Health and Nutrition Examination Survey; TGs: triglycerides; T2DM: type 2 diabetes mellitus; LDL: low-density lipoprotein; PIR: poverty income ratio; ALT: alanine aminotransferase; AST: aspartate aminotransferase; ALP: alkaline phosphatase; GGT: gamma-glutamyl transferase; OR: odds ratio; CI: confidence interval.

Model 1: Non-adjusted model.

Model 2: Adjusted for age, gender, education, and marital status.

Model 3: Adjusted for age, gender, race, education, marital status, T2DM, LDL, TG, PIR, ALT, AST, ALP, GGT, drinking, and smoking.

Furthermore, the RCS analyses further confirmed a statistically significant overall association between CHG and HTN in both cohorts (P_overall < 0.001 for CHARLS and NHANES; Figure 2a and b). Tests for nonlinearity were non-significant (CHARLS: P_nonlinear = 0.712; NHANES: P_nonlinear = 0.083), supporting a predominantly linear positive relationship between the CHG index and HTN risk.

Figure 2.

RCS curves for the association of the CHG index and HTN. Panels (a) RCS curves from the CHARLS data set (adjusted for age, gender, education, marital status, T2DM, LDL, TG, drinking, and smoking), and panels (b) RCS curves from the NHANES data set (adjusted for age, gender, race, education, marital status, T2DM, LDL, TG, PIR, ALT, AST, ALP, GGT, drinking, and smoking). CHARLS: China Health and Retirement Longitudinal Study; NHANES: National Health and Nutrition Examination Survey; RCS: restricted cubic spline; CHG index: cholesterol, high-density lipoprotein, and glucose index; HTN: hypertension; T2DM: type 2 diabetes mellitus; LDL: low-density lipoprotein; TGs: triglycerides; PIR: poverty income ratio; ALT: alanine aminotransferase; AST: aspartate aminotransferase; GGT: gamma-glutamyl transferase; ALP: alkaline phosphatase.

Sensitivity analysis

Sensitivity analyses using the unweighted NHANES dataset confirmed that the positive association between the CHG index and HTN remained robust (Supplemental Tables S3 and S4). Similarly, studies conducted on the original, non-imputed datasets from both CHARLS and NHANES yielded consistent results, further supporting the reliability of the observed association (Supplemental Table S5). These findings demonstrate the overall stability and robustness of the study outcomes.

ROC curve analysis and calibration performance

In the CHARLS cohort, the CHG index demonstrated an area under the receiver operating characteristic (ROC) curve (AUC) of 0.566 (95% CI: 0.547–0.585), indicating modest discriminatory ability for identifying HTN. The optimal cutoff value was 5.20, yielding a sensitivity of 0.52 and a specificity of 0.59 (Figure 3a). In the NHANES cohort, the AUC was slightly higher at 0.612 (95% CI: 0.602–0.623), also reflecting limited discriminative performance. The optimal cutoff value was 5.16, with corresponding sensitivity and specificity values of 0.53 and 0.64, respectively (Figure 3b).

Figure 3.

ROC curve analysis of the CHG index for predicting HTN in CHARLS (a) and NHANES (b). ROC: receiver operating characteristic; CHG index: cholesterol, high-density lipoprotein, and glucose index; HTN: hypertension; CHARLS: China Health and Retirement Longitudinal Study; NHANES: National Health and Nutrition Examination Survey.

Calibration plots demonstrated satisfactory model fit, showing close agreement between predicted and observed probabilities of HTN (Figure 4a and b). The Brier scores were 0.215 for CHARLS and 0.240 for NHANES, both below the 0.25 threshold, indicating good calibration and consistent predictive reliability across models.

Figure 4.

Calibration plots of the CHG-based logistic regression model in two cohorts: (a) CHARLS and (b) NHANES. Apparent performance (final model fitted on the full data) and bootstrap bias-corrected performance are shown. The 45° dashed line indicates perfect calibration. Apparent performance = final model on the full data; bias-corrected performance = apparent performance minus the expected optimism estimated from the bootstrap samples.

Discussion

In this study, the association between the CHG index and HTN was investigated using two large, nationally representative cohorts, CHARLS and NHANES. Our findings demonstrated that higher CHG index levels were consistently and independently associated with an increased likelihood of HTN, even after adjusting for a comprehensive range of potential confounders. Restricted cubic spline analyses revealed a linear relationship between the CHG index and the risk of HTN, indicating that incremental increases in CHG were proportionally associated with higher blood pressure risk. However, ROC analyses suggested only modest discriminative performance (AUC < 0.65), implying that while the CHG index reflects relevant pathophysiological mechanisms, its standalone diagnostic utility for identifying HTN remains limited. Sensitivity analyses confirmed the robustness of these findings across various model specifications and datasets.

These results align with the notion that the CHG index captures essential metabolic disturbances contributing to HTN. Elevated CHG values reflect a combined state of dyslipidemia, characterized by increased total cholesterol and reduced HDL-C, as well as hyperglycemia, both of which are hallmarks of insulin resistance. Insulin resistance has been widely recognized as a central driver of HTN through multiple interrelated mechanisms: it enhances renal sodium retention, activates the renin–angiotensin–aldosterone system and the sympathetic nervous system, impairs endothelial nitric oxide–mediated vasodilation, and promotes oxidative stress.^6,^35,³⁶ Experimental and clinical studies have shown that insulin resistance associated with insulin states leads to exaggerated sympathetic activation and diminished resulting capacity, contributing to increased peripheral vascular resistance.³⁵ Moreover, low HDL-C may impair reverse cholesterol transport and endothelial repair processes,³⁷ while elevated total cholesterol and chronic hypertriglyceridemia contribute to atherogenesis, vascular stiffness, and endothelial dysfunction.³⁸^–40 Thus, by integrating lipid and glycemic parameters, the CHG index likely serves as a composite surrogate marker of underlying cardiometabolic dysfunction. These mechanistic pathways provide strong biological plausibility for the observed linear association between the CHG index and HTN risk.

Our findings are consistent with previous research examining metabolic indices linked to HTN. Among these, the triglyceride–glucose (TyG) index, a well-established surrogate marker of insulin resistance, has been repeatedly associated with elevated blood pressure in numerous studies. For instance, Putra et al.⁴¹ and Lukito et al.⁴² reported that individuals in the highest TyG quartile had a significantly greater risk of HTN compared with those in the lowest quartile (adjusted OR ≈ 1.6ared with those in the lowest quartile (adjusted OR d a significantly greater risk of sociatd increase in HTN risk beyond specific TyG thresholds. However, our analyses demonstrated a linear association between the CHG index and HTN across its range, suggesting that the CHG index may serve as a more continuous and sensitive indicator of overall metabolic burden.

Existing literature also supports the broader applicability of the CHG index in assessing metabolic and cardiovascular risk. Mansoori et al.¹⁴ first introduced the CHG index as a novel screening biomarker for type 2 diabetes in an Iranian cohort, demonstrating superior diagnostic performance (AUC = 0.864) compared with the TyG index (AUC = 0.825), along with higher sensitivity and specificity. Subsequently, Mo et al.,¹² using prospective data from the CHARLS cohort, demonstrated a linear association between CHG index levels and cardiovascular events, with predictive ability comparable to the TyG index, which showed a non-linear pattern of risk. Based on this evidence, our study extends the clinical relevance of the CHG index by evaluating its relationship with blood pressure and HTN risk in two large and ethnically distinct populations.

Several limitations of this study should be acknowledged. First, the NHANES analysis was cross-sectional in design, which precludes causal inference and limits the ability to establish temporal relationships. However, the CHARLS dataset included follow-up blood pressure measurements, allowing for a limited prospective evaluation of the association between baseline CHG index and subsequent HTN. However, the relatively short follow-up duration constrains the strength of these longitudinal conclusions.

Second, population characteristics differ between the two cohorts. CHARLS primarily represents middle-aged and older community-dwelling adults in China, whereas NHANES reflects a broader but predominantly US-based population. These demographic and regional differences may affect the generalizability of our findings to other populations or ethnic groups.

Third, some variables, such as self-reported HTN status, medication use, smoking, and alcohol consumption, may be subject to recall bias or misclassification. Although extensive covariate adjustments were made, the potential for residual confounding from unmeasured factors, including dietary intake, specific pharmacologic treatments, physical activity, socioeconomic disparities, and comorbid conditions, cannot be entirely ruled out.

Fourth, the CHG index demonstrated only modest discriminatory power for HTN (AUC < 0.65), indicating limited ability to distinguish between hypertensive and non-hypertensive individuals when used in isolation. Lastly, although this study extends previous research on the CHG index beyond diabetes and cardiovascular outcomes to HTN, direct comparative or incremental value analyses were not performed against other established indices (e.g. TyG index and TG/HDL-C ratio) within the same cohorts. Therefore, the relative predictive advantage of the CHG index remains to be determined.

Future research should aim to address these gaps. First, large-scale prospective cohort studies are needed to confirm whether elevated CHG index levels predict the onset and progression of HTN and to identify optimal cutoff values for clinical risk stratification. Second, interventional studies should explore whether reducing CHG index levels through lipid-lowering or glycemic control strategies can lead to measurable improvements in blood pressure regulation. Third, mechanistic investigations, such as experimental models that manipulate HDL-C or glucose metabolism under insulin-resistant conditions, may help elucidate the biological pathways linking CHG to HTN. Fourth, within-cohort comparative analyses between the CHG index and other metabolic indices (e.g. TyG or TG/HDL-C) are warranted to determine the most effective marker for HTN prediction. Lastly, future studies encompassing diverse ethnic populations and integrating the CHG index with genetic or inflammatory biomarkers could further improve understanding of its clinical relevance and potential applications.

Conclusion

In summary, the CHG index shows a significant association with HTN, reflecting its potential as an integrative marker of metabolic risk. However, its standalone discriminative ability remains modest (AUC < 0.65), suggesting that its clinical application should currently be regarded as exploratory. Future prospective studies are warranted to validate these associations and to assess the performance of models that incorporate the CHG index alongside established cardiovascular risk factors. Moreover, mechanistic investigations are necessary to elucidate further the biological pathways linking the CHG index to the development of HTN.

Supplemental Material

sj-doc-1-sci-10.1177_00368504251396781 - Supplemental material for The relationship between cholesterol, high-density lipoprotein, and glucose index and hypertension: A study based on two national cohorts

Supplemental material, sj-doc-1-sci-10.1177_00368504251396781 for The relationship between cholesterol, high-density lipoprotein, and glucose index and hypertension: A study based on two national cohorts by Zhe Zhang, Shushen Weng, Dong Cai, Lingling Zhao and Mengting Chen in Science Progress

Supplemental Material

sj-doc-2-sci-10.1177_00368504251396781 - Supplemental material for The relationship between cholesterol, high-density lipoprotein, and glucose index and hypertension: A study based on two national cohorts

Supplemental material, sj-doc-2-sci-10.1177_00368504251396781 for The relationship between cholesterol, high-density lipoprotein, and glucose index and hypertension: A study based on two national cohorts by Zhe Zhang, Shushen Weng, Dong Cai, Lingling Zhao and Mengting Chen in Science Progress

Footnotes

Abbreviations

Acknowledgments

We thank all the authors who participated in this study. We thank the NHANES database (https://www.cdc.gov/nchs/nhanes/) and the CHARLS database (）for providing summary-level data and the efforts of all the researchers.

ORCID iD

Zhe Zhang

Ethics approval and consent to participate

This study involved human participants and received ethical approval from the Research Ethics Review Board (ERB) of the National Health and Nutrition Examination Survey (NHANES), with detailed information available at . The CHARLS protocol was approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052-11015). Written informed consent was obtained from all participants in both NHANES and CHARLS prior to data collection. The present study was based entirely on de-identified, publicly available data, and thus, no additional ethical approval or participant consent was required for this secondary analysis.

Authors’ contributions

ZZ (corresponding author) contributed to the study design, conceptualization, manuscript drafting, and interpretation of results and provided supervision, and reviewed and edited the manuscript; SW, DC, LZ, and MTC were responsible for data collection and analysis; all authors participated in critical revisions and approved the final version of the manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.

Data availability statement

Data from the NHANES are publicly accessible at https://www.cdc.gov/nchs/nhanes/. Data from the CHARLS are publicly accessible at .

Supplemental material

Supplemental material for this article is available online: The supplementary materials include additional figures and data tables, as detailed in the supplementary file.

References

Kario

Okura

Hoshide

, et al. The WHO global report 2023 on hypertension warning the emerging hypertension burden in globe and its treatment strategy. Hypertens Res 2024; 47: 1099–1102.

Sakr

Sirasanagandla

Das

, et al. Insulin resistance and hypertension: mechanisms involved and modifying factors for effective glucose control. Biomedicines 2023; 11: 2271.

, et al. Global, regional, and national disease burden attributable to high systolic blood pressure in youth and young adults: 2021 global burden of disease study analysis. BMC Med 2025; 23: 74.

Morse

Zhang

Thakur

, et al. Hypertension and the metabolic syndrome. Am J Med Sci 2005; 330: 303–310.

Mendizabal

Llorens

Nava

. Hypertension in metabolic syndrome: vascular pathophysiology. Int J Hypertens 2013; 2013: 230868.

Horita

Seki

Yamada

, et al. Insulin resistance, obesity, hypertension, and renal sodium transport. Int J Hypertens 2011; 2011: 391762.

Zhang

Guo

, et al. The association of triglyceride and glucose index, and triglyceride to high-density lipoprotein cholesterol ratio with prehypertension and hypertension in normoglycemic subjects: a large cross-sectional population study. J Clin Hypertens (Greenwich) 2021; 23: 1405–1412.

Sinha

Haque

. Insulin resistance is cheerfully hitched with hypertension. Life (Basel) 2022; 12: 564.

Dalal

Padmanabhan

Jain

, et al. LIPITENSION: interplay between dyslipidemia and hypertension. Indian J Endocrinol Metab 2012; 16: 240–245.

10.

Srikanth

Deedwania

. Management of dyslipidemia in patients with hypertension, diabetes, and metabolic syndrome. Curr Hypertens Rep 2016; 18: 76.

11.

Wei

Qian

, et al. Exploring the U-shaped nonlinear relationship of CHG index with metabolic syndrome and mortality risks in metabolic syndrome patients. Lipids Health Dis 2025; 24: 294.

12.

Zhang

, et al. Cholesterol, high-density lipoprotein, and glucose index versus triglyceride–glucose index in predicting cardiovascular disease risk: a cohort study. Cardiovasc Diabetol 2025; 24: 116.

13.

Catak

Konuk

Hepsen

. The cholesterol-HDL-glucose (CHG) index and traditional adiposity markers in predicting diabetic retinopathy and nephropathy. J Diabetes Investig 2025; 16: 1487–1494.

14.

Mansoori

Nosrati

Dorchin

, et al. A novel index for diagnosis of type 2 diabetes mellitus: cholesterol, high density lipoprotein, and glucose (CHG) index. J Diabetes Investig 2025; 16: 309–314.

15.

von Elm

Altman

Egger

, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med 2007; 147: 573–577.

16.

Wang

Shi

, et al. Association between the cumulative average triglyceride glucose-body mass index and cardiovascular disease incidence among the middle-aged and older population: a prospective nationwide cohort study in China. Cardiovasc Diabetol 2024; 23: 16.

17.

Yang

. Association between appendicular lean mass and chronic obstructive pulmonary disease: epidemiological cross-sectional study and bidirectional Mendelian randomization analysis. Front Nutr 2023; 10: 1159949.

18.

Wei

Qian

, et al. Estimated glucose disposal rate and its dual role in hyperlipidemia risk and mortality: a secondary analysis of retrospective cohort data from the U.S. and Chinese. Lipids Health Dis 2025; 24: 266.

19.

Peng

, et al. The roles of lipids and inflammation in the association between the triglyceride-glucose index and arterial stiffness: evidence from two large population-based surveys. Lipids Health Dis 2024; 23: 190.

20.

Yuan

, et al. Hypertension and NAFLD risk: insights from the NHANES 2017–2018 and Mendelian randomization analyses. Chin Med J (Engl) 2024; 137: 457–464.

21.

Chen

Ding

, et al. Composite dietary antioxidant index associated with delayed biological aging: a population-based study. Aging (Albany NY) 2024; 16: 15–27.

22.

Hou

, et al. Association of METS-IR index with type 2 diabetes: a cross-sectional analysis of the National Health and Nutrition Examination Survey data from 2009 to 2018. PLoS One 2024; 19: e308597.

23.

Wang

, et al. Remnant cholesterol inflammatory index and its association with all-cause and cause-specific mortality in middle-aged and elderly populations: evidence from US and Chinese national population surveys. Lipids Health Dis 2025; 24: 155.

24.

Sun

Shen

, et al. Abdominal fat accumulation increases the risk of high blood pressure: evidence of 47,037 participants from Chinese and US national population surveys. Nutr J 2024; 23: 153.

25.

Duan

Zhao

, et al. Metabolic score for insulin resistance (METS-IR) predicts all-cause and cardiovascular mortality in the general population: evidence from NHANES 2001–2018. Cardiovasc Diabetol 2024; 23: 243.

26.

Zhang

Dong

, et al. Association between exposure to a mixture of phenols, pesticides, and phthalates and obesity: comparison of three statistical models. Environ Int 2019; 123: 325–336.

27.

Yang

Liu

Zhao

, et al. The association between the triglyceride-glucose index and sarcopenia: data from the NHANES 2011–2018. Lipids Health Dis 2024; 23: 219.

28.

Yan

Chai

Yang

, et al. Association of the atherogenic index of plasma with frailty in U.S. adults: a cross-sectional study based on NHANES. Lipids Health Dis 2025; 24: 84.

29.

Zhang

, et al. Association of the newly proposed dietary index for gut microbiota and constipation: a cross-sectional study from NHANES. Front Nutr 2025; 12: 1529373.

30.

. Serum alpha-1 acid glycoprotein and gallstone risk in US adult women: a cross-sectional analysis of the NHANES. Front Nutr 2025; 12: 1527717.

31.

Chen

Zhou

Yang

, et al. Association between lipoproteins and telomere length in US adults: data from the NHANES 1999–2002. Lipids Health Dis 2019; 18: 80.

32.

Guo

Mutailipu

Wen

, et al. Association between lymphocyte to high-density lipoprotein cholesterol ratio and insulin resistance and metabolic syndrome in US adults: results from NHANES 2007–2018. Lipids Health Dis 2025; 24: 9.

33.

Huang

, et al. Development of a prognostic app (iCanPredict) to predict survival for Chinese women with breast cancer: retrospective study. J Med Internet Res 2022; 24: e35768.

34.

Peng

Zhang

Zhou

. A cross-sectional study comparing machine learning and logistic regression techniques for predicting osteoporosis in a group at high risk of cardiovascular disease among old adults. BMC Geriatr 2025; 25: 6673.

35.

Mancusi

Izzo

di Gioia

, et al. Insulin resistance the hinge between hypertension and type 2 diabetes. High Blood Press Cardiovasc Prev 2020; 27: 515–526.

36.

, et al. Role of hyperinsulinemia and insulin resistance in hypertension: metabolic syndrome revisited. Can J Cardiol 2020; 36: 671–682.

37.

Tran‐Dinh

Diallo

Delbosc

, et al. HDL and endothelial protection. Br J Pharmacol 2013; 169: 493–511.

38.

Funk

Yurdagul

Orr

. Hyperglycemia and endothelial dysfunction in atherosclerosis: lessons from type 1 diabetes. Int J Vasc Med 2012; 2012: 569654.

39.

Mao

Yin

Yang

, et al. Role of advanced glycation end products on vascular smooth muscle cells under diabetic atherosclerosis. Front Endocrinol (Lausanne) 2022; 13: 983723.

40.

Gordin

Groop

. Aspects of hyperglycemia contribution to arterial stiffness and cardiovascular complications in patients with type 1 diabetes. J Diabetes Sci Technol 2016; 10: 1059–1064.

41.

Putra

ICS

Pranata

Pramudyo

, et al. The association between triglyceride-glucose index and incidence of hypertension: a systematic review and dose-response meta-analysis. J Hypertens 2023; 41: e10.

42.

Lukito

Kamarullah

Huang

, et al. Association between triglyceride-glucose index and hypertension: a systematic review and meta-analysis. Narra J 2024; 4: e951.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.16 MB

0.03 MB