Abstract
Background
The Cardiovascular Metabolic Index (CMI) has emerged as a robust indicator for predicting metabolic diseases. However, the extent to which CMI can forecast the future risk of cardiovascular disease among individuals with impaired glucose metabolism—specifically those with diabetes and prediabetes—remains uncertain. This uncertainty is particularly concerning given the substantial burden of cardiovascular complications that these populations face. To address this critical gap in knowledge and to provide actionable insights for targeted prevention and management strategies, we undertook this study. By doing so, we aimed to elucidate whether CMI can serve as a reliable predictor in this high-risk group of individuals with impaired glucose metabolism, thereby informing more effective and personalized approaches to managing cardiovascular risk.
Method
We included patients with impaired glucose metabolism in our baseline cohort and utilized K-means clustering to categorize changes in CMI into three distinct groups. We then applied a multivariate logistic regression model to assess the association between CMI, cumulative CMI, and changes in CMI with the incidence of hypertension and stroke within this population. This statistical approach enabled us to control for multiple confounding variables, thereby providing a more accurate assessment of the predictive power of CMI-related metrics. Furthermore, to explore potential nonlinear relationships between CMI and cumulative CMI with the occurrence of hypertension and stroke events, we employed restricted cubic spline (RCS) regression models. Finally, we utilized ROC curve analysis to evaluate the predictive ability of CMI and cumulative CMI for future risks of hypertension and stroke in individuals with impaired glucose metabolism.
Results
During the follow-up period from 2015 to 2018, we monitored 4,207 participants with impaired glucose metabolism and observed 746 new cases of hypertension and 144 stroke events. After meticulously adjusting for multiple potential confounding factors, our analysis revealed that the CMI significantly increases the risk of hypertension (OR = 1.27, 95% CI: 1.15–1.40) and stroke (OR = 1.31, 95% CI: 1.05–1.67) in this population. Compared with the first quartile of CMI, both the third and fourth quartiles were associated with heightened risks of hypertension and stroke. While cumulative CMI and its quartiles were linked to hypertension, no significant association was found with stroke. A nonlinear relationship was observed between CMI and hypertension risk (p = 0.017). Participants with moderate CMI levels (OR: 1.44, 95% CI: 1.21–1.71) and high CMI levels (OR: 2.69, 95% CI: 1.66–4.31) had significantly higher risks of hypertension than those with stable CMI levels. ROC curves showed AUC values of 0.73 and 0.67 for CMI in predicting hypertension and stroke, respectively.
Conclusion
Elevated levels of the CMI are associated with an increased risk of hypertension and stroke among individuals with impaired glucose metabolism. Maintaining low CMI levels may help mitigate the risk of hypertension in these populations. Given that CMI can be easily measured in community settings, clinicians can utilize it for risk stratification and develop personalized prevention strategies. These strategies may help reduce the burden of future complications such as hypertension and stroke in patients with impaired glucose metabolism.
Keywords: Cardiometabolic index, Hypertension, Diabetes, Stroke, Prospective cohort study
Introduction
Diabetes poses a significant global health challenge, with approximately 529 million people affected worldwide [1]. Cardiovascular and cerebrovascular complications remain the leading causes of morbidity and mortality in this population—about 68% of older adults with diabetes die from cardiovascular diseases, and 16% are at risk of stroke [2–4]. Early identification of individuals at high risk is crucial for effective prevention.
The Cardiovascular Metabolic Index (CMI), proposed in 2015, is a composite indicator derived from the triglyceride-to-high-density lipoprotein cholesterol ratio (TG/HDL-C) and waist-to-height ratio (WHtR) [5]. As both lipid imbalance and central obesity are key contributors to cardiometabolic disease, CMI provides a convenient and integrative reflection of metabolic health that can be easily assessed in clinical practice.
Previous studies have linked higher CMI to several metabolic and cardiovascular outcomes, including diabetes, hypertension, and cardiovascular disease [6–11]. For example, U.S. population studies have shown that each unit increase in CMI corresponds to a higher risk of hypertension [12], while Chinese data indicate an association between elevated CMI and stroke incidence [13]. However, evidence on the prospective predictive value of CMI for hypertension and stroke in individuals with impaired glucose metabolism remains scarce.
Therefore, this study aims to elucidate the relationship between baseline and cumulative CMI and the future risk of hypertension and stroke in adults with diabetes and prediabetes, using a prospective cohort design. This work seeks to provide new evidence to support early intervention strategies for this high-risk population.
Methods
Study design and population information
This study leveraged data from the China Health and Retirement Longitudinal Study (CHARLS), a national survey spanning 28 provinces in China [14]. Utilizing multistage probability sampling, the study recruited a substantial number of participants from 450 villages and urban communities across 150 counties and regions. The baseline survey commenced in 2011, with follow-ups occurring every 2–3 years.
For our analysis, we utilized the 2015 survey data as the baseline, focusing on participants with impaired glucose metabolism. We calculated the CMI and cumulative CMI using measurements of TG, HDL-C, and WHtR obtained from blood samples collected in 2011 and 2015, as well as subsequent measurements. Following cluster analysis, we examined the relationship between these indicators and the risk of new-onset hypertension and stroke at the 2018 follow-up endpoint.
In this study, we included individuals with impaired glucose metabolism, specifically those diagnosed with diabetes or prediabetes. The diagnostic criteria for diabetes were as follows: a self-reported physician diagnosis of diabetes, combined with a history of diabetes medication use, fasting blood glucose ≥ 7.0 mmol/L, hemoglobin A1c (HbA1c) ≥ 6.5%, or 2-hour blood glucose/random blood glucose ≥ 11.1 mmol/L. Prediabetes was defined as fasting blood glucose between 6.1 and 6.9 mmol/L, 2-hour oral glucose tolerance test blood glucose between 7.8 and 11.0 mmol/L, or HbA1c between 5.7% and 6.4%. We included participants who had complete information on fasting blood glucose, HbA1c, history of diabetes medication use, TG, HDL-C, and WHtR.The exclusion criteria were: (1) presence of hypertension or stroke at baseline; (2) missing information on fasting blood glucose, HbA1c, TG, HDL-C, etc., in either 2011 or 2015; (3) missing or extremely abnormal values for height or waist circumference; (4) age below 45 years; (5) participants with extensive missing information (see Fig. 1).
Fig. 1.
Flow chart of the population included in this study
The CHARLS cohort study has been approved by the Biomedical Ethics Review Committee of Peking University. All participants provided written informed consent, or their legal guardians provided consent on their behalf. As the data for this study were derived from published sources and the necessary informed consent and ethical approval had already been obtained, there was no need to seek additional informed consent or ethical approval.
Assessment of CMI changes and cumulative CMI
Levels of HDL-C and TG were obtained from blood samples collected during the first and third waves of the survey. For height measurements, any change exceeding 10 centimeters between the first and third waves was considered an abnormal measurement. Similarly, any waist circumference measurement deviating by three standard deviations from the overall average (whether higher or lower) was classified as abnormal.
The following calculations were performed using the collected data:
(1)The WHtR was calculated by dividing the waist circumference by height.
(2) CMI was calculated using the formula (TG/HDL-C) × WHtR.
(3) Cumulative CMI was calculated as (2012 CMI + 2015 CMI)/2 × time (2015 − 2012).
Hypertension and stroke assessment
To assess hypertension, we utilized a standardized question: “Have you been diagnosed with hypertension by a doctor?” Participants who responded “yes” were initially categorized as having hypertension. This initial classification was subsequently confirmed by verifying a history of antihypertensive medication use or blood pressure readings (systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg).
Similarly, the occurrence of stroke was ascertained using the question: “Have you been diagnosed with a stroke by a doctor?” The definitions of hypertension and stroke were based on self-reported diagnoses. If participants had previously reported hypertension or stroke in earlier survey waves, these conditions needed to be re-diagnosed in subsequent waves. In cases where participants denied their previously self-reported diagnoses of hypertension or stroke, these inconsistencies were retrospectively corrected.Our methods for determining hypertension and stroke align with those employed in previous studies utilizing CHARLS data [15].
Covariate
The covariates included in the analysis encompass sociodemographic characteristics and health status indicators. Sociodemographic characteristics include age, gender, marital status, household registration status, educational attainment, smoking status, drinking status, household income level, and body mass index. Health status indicators include heart disease, dyslipidemia, cancer, use of lipid-lowering drugs, anticancer drugs, antidiabetic drugs, fasting blood glucose, and glycated hemoglobin.
Dyslipidemia is defined as meeting any of the following criteria: total cholesterol (TC) ≥ 240 mg/dl, TG ≥ 150 mg/dl, low-density lipoprotein cholesterol (LDL-C) ≥ 160 mg/dl, HDL-C < 40 mg/dl, self-reported dyslipidemia, or use of lipid-lowering therapy.
Statistical analysis
Continuous variables are presented as mean (standard deviation) or median (interquartile range, IQR), while categorical variables are presented as frequency (proportion). A two-sided p-value of less than 0.05 is considered to indicate statistical significance. All statistical analyses were conducted using R software version 4.4.3 and MATLAB R2024a.
K-means clustering analysis
We utilized K-means clustering analysis, an unsupervised machine learning technique, to categorize the CMI indices [16]. The optimal number of clusters was determined using the elbow method, as illustrated in Fig. 2, resulting in three distinct categories. Based on the CMI measurements from 2012 to 2015, participants were divided into the following three groups:
Fig. 2.
Cluster Analysis Classification and Results Presentation Diagram of CMI Changes. (A) Elbow Method to Determine the Optimal Clustering Point for K-means Clustering. (B) Optimal Clustering Point Based on Comprehensive Evaluation of Clustering Score Coefficient and Silhouette Coefficient. (C) K-means Clustering Results Presentation Diagram. (D) CMI Clustering Changes from 2012 to 2015
Group 1 (n = 2,281): CMI remained stable at a low level, changing from 2.54 ± 0.46 in 2012 to 2.51 ± 0.48 in 2015, with a cumulative CMI of 7.57 ± 0.98.
Group 2 (n = 1,839): This group exhibited moderate CMI levels with a slight increase over time, from 3.28 ± 0.46 in 2012 to 3.50 ± 0.55 in 2015, and a cumulative CMI of 10.17 ± 1.09.
Group 3 (n = 87): Characterized by a high initial CMI level that decreased over time, from 5.65 ± 3.48 in 2012 to 4.04 ± 0.73 in 2015, with a cumulative CMI of 14.54 ± 4.96.
These classifications offer valuable insights into the varying trajectories of CMI among the study participants (see Fig. 2).
Multivariate logistic regression analysis
We employed multivariate logistic regression analysis to examine the relationship between different CMI clusters and the incidence of new-onset hypertension and stroke. The results are presented as odds ratios along with their 95% confidence intervals. We constructed three models to progressively adjust for potential confounding factors: (1) Model 1: Adjusted for age and sex. (2) Model 2: Adjusted for age, sex, marital status, education level, household registration information, household consumption level, BMI, smoking, and alcohol consumption.(3) Model 3: Further adjusted for heart disease, dyslipidemia, cancer, lipid-lowering drugs, anticancer drugs, antidiabetic drugs, fasting blood glucose, and glycated hemoglobin. These models allowed us to comprehensively assess the impact of different CMI clusters on the risk of hypertension and stroke while accounting for a wide range of potential confounding factors.
Receiver operating characteristic (ROC) curve analysis
We conducted ROC curve analysis to evaluate the predictive ability of CMI for future hypertension and stroke in individuals with abnormal glucose metabolism. We compared the predictive performance of CMI with that of CMI at wave 1 (CMI_w1), CMI at wave 3 (CMI_w3), and cumulative CMI. Additionally, we constructed ROC curves for three logistic regression models to assess their predictive performance.
Restricted cubic spline (RCS) model
To explore the nonlinear relationship between cumulative CMI and the future risk of hypertension and stroke in individuals with impaired glucose metabolism, we developed RCS models. Model selection was guided by the Bayesian Information Criterion. We employed the likelihood ratio test to evaluate the presence of nonlinear relationships within these RCS models. The covariates adjusted for in the RCS models were consistent with those used in Model 3 of the logistic regression analysis.
Subgroup analysis and interactions
To further substantiate the relationship between clustered CMI and cumulative CMI and the incidence of hypertension and stroke among individuals with abnormal glucose metabolism, we undertook subgroup analyses. These analyses were conducted separately for patients with hypertension and those with stroke. The subgroups examined encompassed gender, household registration status, marital status, educational level, smoking status, drinking status, heart disease, dyslipidemia, and cancer.
Sensitivity analysis
The dataset utilized in this study contains some missing values. To address this issue, we employed multiple imputation using the “mice” package in R to fill in the missing covariates. Additionally, to ensure the robustness of our findings, we conducted a sensitivity analysis by repeating the analysis using the complete dataset without multiple imputation.
Results
Baseline information for 4,207 participants
Baseline data were collected during the third wave of the survey in 2015, encompassing a total of 4,207 participants with impaired glucose metabolism for analysis. As illustrated in Table 1, the average age of participants at baseline was 61.7 ± 8.9 years, with 54.5% being female. The majority of participants had an educational level of high school or below (90.0%), were married or had a partner (86.4%), and were rural residents (83.4%). The proportions of non-smokers (56.2%) and non-drinkers (65.3%) exceeded those of smokers and drinkers. Additionally, 98.5% of participants had no history of cancer, and 93.5% had no dyslipidemia. Table 1 also presents the baseline characteristics of participants in each group based on cluster analysis. Furthermore, we detailed the baseline characteristics of participants according to cumulative CMI classification in Table 2.
Table 1.
Baseline characteristics of individuals classified by class of the change in the CMI
| Characteristics | Level | Overall | Change in the CMI | P-value | ||
|---|---|---|---|---|---|---|
| Class 1 | Class 2 | Class 3 | ||||
| Number | 4207 | 2281 | 1839 | 87 | ||
| Age (mean (SD)) | 61.7(8.9) | 62.2(9.0) | 61.0(8.8) | 62.8(9.1) | < 0.001 | |
| Total_household_consumption(mean (SD)) | 13054.4(20247.6) | 12617.7(21954.8) | 13600.0(18050.2) | 12972.9(17218.6) | 0.302 | |
| BMI (mean (SD)) | 24.1(17.3) | 21.8(4.9) | 26.8(25.4) | 27.0(4.7) | < 0.001 | |
| Sex (%) | Female | 2293(54.5) | 1062(46.6) | 1171(63.7) | 60(69.0) | < 0.001 |
| Male | 1914(45.5) | 1219(53.4) | 668(36.3) | 27(31.0) | ||
| Education (%) | Below high school | 3788(90.0) | 2088(91.5) | 1620(88.1) | 80(92.0) | 0.007 |
| College or above | 45(1.1) | 20(0.9) | 24(1.3) | 1(1.1) | ||
| High school | 374(8.9) | 173(7.6) | 195(10.6) | 6(6.9) | ||
| Marital_status (%) | Married/Partnered | 3636(86.4) | 1973(86.5) | 1587(86.3) | 76(87.4) | 0.579 |
| Never married | 26(0.6) | 18(0.8) | 8(0.4) | 0(0.0) | ||
| widowed/separated/divorced | 545(13.0) | 290(12.7) | 244(13.3) | 11(12.6) | ||
| Hukou (%) | Rural | 3508(83.4) | 1976(86.6) | 1455(79.1) | 77(88.5) | < 0.001 |
| Urban | 699(16.6) | 305(13.4) | 384(20.9) | 10(11.5) | ||
| Smoking (%) | Ever smokers | 1844(43.8) | 1139(49.9) | 671(36.5) | 34(39.1) | < 0.001 |
| Never smokers | 2363(56.2) | 1142(50.1) | 1168(63.5) | 53(60.9) | ||
| Drinks (%) | Ever drinkers | 1460(34.7) | 885(38.8) | 550(29.9) | 25(28.7) | < 0.001 |
| Never drinkers | 2747(65.3) | 1396(61.2) | 1289(70.1) | 62(71.3) | ||
| Heart_problem(%) | No | 3665(87.1) | 2020(88.6) | 1580(85.9) | 65(74.7) | < 0.001 |
| Yes | 542(12.9) | 261(11.4) | 259(14.1) | 22(25.3) | ||
| Cancer(%) | No | 4144(98.5) | 2247(98.5) | 1812(98.5) | 85(97.7) | 0.823 |
| Yes | 63(1.5) | 34(1.5) | 27(1.5) | 2(2.3) | ||
| Dyslipidemia(%) | No | 3935(93.5) | 2186(95.8) | 1681(91.4) | 68(78.2) | < 0.001 |
| Yes | 272(6.5) | 95(4.2) | 158(8.6) | 19(21.8) | ||
| Meds_diabetesw3 (%) | No | 3914(93.0) | 2203(96.6) | 1653(89.9) | 58(66.7) | < 0.001 |
| Yes | 293(7.0) | 78(3.4) | 186(10.1) | 29(33.3) | ||
| Meds_heartproblems(%) | No | 3801(90.3) | 2090(91.6) | 1635(88.9) | 76(87.4) | 0.008 |
| Yes | 406(9.7) | 191(8.4) | 204(11.1) | 11(12.6) | ||
| Meds_stroke(%) | No | 4123(98.0) | 2242(98.3) | 1803(98.0) | 78(89.7) | < 0.001 |
| Yes | 84(2.0) | 39(1.7) | 36(2.0) | 9(10.3) | ||
| Meds_cancer(%) | No | 4159(98.9) | 2257(98.9) | 1817(98.8) | 85(97.7) | 0.537 |
| Yes | 48(1.1) | 24(1.1) | 22(1.2) | 2(2.3) | ||
| Meds_HBP(%) | No | 4034(95.9) | 2214(97.1) | 1740(94.6) | 80(92.0) | < 0.001 |
| Yes | 173(4.1) | 67(2.9) | 99(5.4) | 7(8.0) | ||
| Meds_dyslipidemia(%) | No | 4089(97.2) | 2243(98.3) | 1768(96.1) | 78(89.7) | < 0.001 |
| Yes | 118(2.8) | 38(1.7) | 71(3.9) | 9(10.3) | ||
| HDL(mean (SD)) | 52.0(12.0) | 56.0(12.6) | 47.5(9.3) | 43.4(8.7) | < 0.001 | |
| LDL(mean (SD)) | 105.3(29.0) | 103.9(28.0) | 107.3(30.2) | 98.2(26.5) | < 0.001 | |
| TC(mean (SD)) | 187.5(37.0) | 183.0(34.6) | 192.5(38.9) | 199.8(40.8) | < 0.001 | |
| TG(mean (SD)) | 140.2(89.1) | 98.6(46.7) | 185.2(97.8) | 278.8(131.3) | < 0.001 | |
| SBP(mean (SD)) | 123.0(17.3) | 120.9(17.2) | 125.2(17.2) | 128.5(15.9) | < 0.001 | |
| DBP(mean (SD)) | 72.6(10.8) | 71.2(10.4) | 74.4(10.9) | 73.7(10.9) | < 0.001 | |
| Glucose(mean (SD)) | 107.2(38.7) | 99.0(24.7) | 115.5(47.2) | 145.1(70.2) | < 0.001 | |
| FBG(mean (SD)) | 6.0(2.2) | 5.5(1.4) | 6.4(2.6) | 8.1(3.9) | < 0.001 | |
| HbA1C(mean (SD)) | 6.2(1.0) | 6.0(0.7) | 6.4(1.2) | 7.2(1.7) | < 0.001 | |
| CMI_w1(mean (SD)) | 2.9(0.9) | 2.5(0.5) | 3.3(0.5) | 5.7(3.5) | < 0.001 | |
| CMI_w3(mean (SD)) | 3.0(0.7) | 2.5(0.5) | 3.5(0.6) | 4.0(0.7) | < 0.001 | |
| C_CMI (mean (SD)) | 8.9(2.0) | 7.6(1.0) | 10.2(1.1) | 14.5(5.0) | < 0.001 | |
Table 2.
Baseline characteristics of individuals classified by quartiles of the cumulative average CMI
| Characteristics | Level | Overall | Quartiles of the cumulative average CMI | P-value | |||
|---|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | ||||
| Number | 4207 | 1052 | 1051 | 1052 | 1052 | ||
| Age(mean (SD)) | 61.7 (8.9) | 62.7 (9.0) | 61.8 (9.0) | 61.1 (8.7) | 61.2 (8.9) | < 0.001 | |
| Total_household_consumption (mean(SD)) | 13054.4 (20247.6) | 12177.5 (15651.6) | 13241.8 (27706.6) | 13505.9 (17616.6) | 13292.6 (17813.8) | 0.435 | |
| BMI(mean (SD)) | 24.1 (17.3) | 20.8 (3.1) | 22.6 (6.2) | 24.1 (2.8) | 28.8 (33.3) | < 0.001 | |
| Sex(%) | Female | 2293 (54.5) | 400 (38.0) | 556 (52.9) | 643 (61.1) | 694 (66.0) | < 0.001 |
| Male | 1914 (45.5) | 652 (62.0) | 495 (47.1) | 409 (38.9) | 358 (34.0) | ||
| Education(%) | Below high school | 3788 (90.0) | 965 (91.7) | 962 (91.5) | 925 (87.9) | 936 (89.0) | 0.044 |
| College or above | 45 (1.1) | 8 (0.8) | 11 (1.0) | 14 (1.3) | 12 (1.1) | ||
| High school | 374 (8.9) | 79 (7.5) | 78 (7.4) | 113 (10.7) | 104 (9.9) | ||
| Marital_status(%) | Married/Partnered | 3636 (86.4) | 918 (87.3) | 907 (86.3) | 907 (86.2) | 904 (85.9) | 0.162 |
| Never married | 26 (0.6) | 12 (1.1) | 5 (0.5) | 6 (0.6) | 3 (0.3) | ||
| Widowed/separated/divorced | 545 (13.0) | 122 (11.6) | 139 (13.2) | 139 (13.2) | 145 (13.8) | ||
| Hukou(%) | Rural | 3508 (83.4) | 934 (88.8) | 899 (85.5) | 831 (79.0) | 844 (80.2) | < 0.001 |
| Urban | 699 (16.6) | 118 (11.2) | 152 (14.5) | 221 (21.0) | 208 (19.8) | ||
| Smoking(%) | Ever smokers | 1844 (43.8) | 587 (55.8) | 480 (45.7) | 409 (38.9) | 368 (35.0) | < 0.001 |
| Never smokers | 2363(56.2) | 465(44.2) | 571 (54.3) | 643 (61.1) | 684 (65.0) | ||
| Drinks(%) | Ever drinkers | 1460(34.7) | 475(45.2) | 351(33.4) | 332 (31.6) | 302 (28.7) | < 0.001 |
| Never drinkers | 2747(65.3) | 577(54.8) | 700 (66.6) | 720(68.4) | 750 (71.3) | ||
| Meds_diabetes(%) | No | 3914(93.0) | 1022 (97.1) | 1009 (96.0) | 993 (94.4) | 890 (84.6) | < 0.001 |
| Yes | 293(7.0) | 30(2.9) | 42(4.0) | 59(5.6) | 162(15.4) | ||
| Meds_heartproblems(%) | No | 3801(90.3) | 969(92.1) | 959(91.2) | 950(90.3) | 923(87.7) | 0.005 |
| Yes | 406(9.7) | 83(7.9) | 92(8.8) | 102(9.7) | 129(12.3) | ||
| Meds_stroke(%) | No | 4123(98.0) | 1036(98.5) | 1030(98.0) | 1040(98.9) | 1017(96.7) | 0.002 |
| Yes | 84(2.0) | 16(1.5) | 21(2.0) | 12 (1.1) | 35 (3.3) | ||
| Meds_cancer(%) | No | 4159(98.9) | 1040(98.9) | 1041(99.0) | 1040(98.9) | 1038(98.7) | 0.88 |
| Yes | 48(1.1) | 12(1.1) | 10(1.0) | 12(1.1) | 14(1.3) | ||
| Meds_HBP(%) | No | 4034(95.9) | 1028(97.7) | 1015(96.6) | 1007(95.7) | 984(93.5) | < 0.001 |
| Yes | 173 (4.1) | 24(2.3) | 36(3.4) | 45(4.3) | 68(6.5) | ||
| Meds_dyslipidemia(%) | No | 4089(97.2) | 1036(98.5) | 1032(98.2) | 1029(97.8) | 992(94.3) | < 0.001 |
| Yes | 118(2.8) | 16(1.5) | 19(1.8) | 23(2.2) | 60(5.7) | ||
| Heart_problem(%) | No | 3665(87.1) | 938(89.2) | 922(87.7) | 930(88.4) | 875(83.2) | < 0.001 |
| Yes | 542 (12.9) | 114(10.8) | 129 (12.3) | 122(11.6) | 177(16.8) | ||
| Cancer(%) | No | 4144(98.5) | 1035(98.4) | 1036(98.6) | 1039(98.8) | 1034(98.3) | 0.814 |
| Yes | 63(1.5) | 17(1.6) | 15(1.4) | 13(1.2) | 18(1.7) | ||
| Dyslipidemia(%) | No | 3935(93.5) | 1015(96.5) | 1003(95.4) | 985(93.6) | 932(88.6) | < 0.001 |
| Yes | 272(6.5) | 37(3.5) | 48(4.6) | 67(6.4) | 120(11.4) | ||
| Stroke(%) | No | 4113(97.8) | 1034(98.3) | 1025(97.5) | 1037(98.6) | 1017(96.7) | 0.015 |
| Yes | 94(2.2) | 18(1.7) | 26(2.5) | 15(1.4) | 35(3.3) | ||
| HDL(mean (SD)) | 52.0(12.0) | 59.2(13.9) | 52.9(10.7) | 50.0(9.6) | 45.8(9.3) | < 0.001 | |
| LDL(mean (SD)) | 105.3(29.0) | 98.8(27.0) | 107.6(27.9) | 109.4(28.8) | 105.3(31.1) | < 0.001 | |
| TC(mean (SD)) | 187.5(37.0) | 179.4(34.8) | 185.3(34.1) | 189.8(34.7) | 195.5(42.1) | < 0.001 | |
| TG(mean (SD)) | 140.2(89.1) | 86.9(44.6) | 109.6(47.5) | 145.5(70.7) | 218.9(111.8) | < 0.001 | |
| SBP(mean (SD)) | 123.0(17.3) | 120.2(17.3) | 121.2(17.0) | 123.7(17.1) | 126.7(17.3) | < 0.001 | |
| DBP(mean (SD)) | 72.6(10.8) | 70.6(10.3) | 71.6(10.5) | 73.5(11.1) | 74.8(10.6) | < 0.001 | |
| Glucose(mean (SD)) | 107.2(38.7) | 96.6(22.9) | 101.6(26.5) | 105.5(32.5) | 125.1(57.0) | < 0.001 | |
| FBG(mean (SD)) | 6.0(2.2) | 5.4(1.3) | 5.6(1.5) | 5.9(1.8) | 7.0(3.2) | < 0.001 | |
| HbA1C(mean (SD)) | 6.2(1.0) | 6.0(0.6) | 6.1(0.8) | 6.1(0.8) | 6.6(1.4) | < 0.001 | |
| CMI_w1(mean (SD)) | 2.9(0.9) | 2.3(0.4) | 2.7(0.3) | 3.0(0.3) | 3.7(1.2) | < 0.001 | |
| CMI_w3(mean (SD)) | 3.0(0.7) | 2.2(0.5) | 2.7(0.3) | 3.1(0.3) | 3.8(0.6) | < 0.001 | |
| C_CMI(mean (SD)) | 8.9(2.0) | 6.7(0.8) | 8.2(0.3) | 9.2(0.3) | 11.2(2.0) | < 0.001 | |
The CMI, cumulative CMI, and changes in these metrics are associated with an increased risk of future hypertension and stroke among individuals with abnormal glucose metabolism
During the follow-up period from 2015 to 2018, among the 4,207 individuals with impaired glucose metabolism, 746 new cases of hypertension (17.7%) and 94 new cases of stroke (2.2%) were identified. For each unit increase in CMI, there was a 27% increase in the risk of hypertension (OR = 1.27, 95% CI: 1.15–1.40, p = 4.46E-06) and a 31% increase in the risk of stroke (OR = 1.31, 95% CI: 1.05–1.67, p = 2.32E-02). When using the first quartile as the reference, both the third (Q3) and fourth (Q4) quartiles of CMI were positively associated with the risk of hypertension and stroke. Additionally, we observed an association between cumulative CMI and the risk of hypertension, but no statistically significant association with the risk of stroke.
Participants were classified based on cumulative CMI quartiles. In the fully adjusted model, compared with the first quartile, the third quartile (OR = 1.48, 95% CI: 1.19–1.85, p = 5.44E-04) and the fourth quartile (OR = 1.84, 95% CI: 1.47–2.31, p = 9.23E-08) were associated with an increased risk of hypertension. However, the relationship between the second quartile (OR = 1.23, 95% CI: 0.98–1.54, p = 7.81E-02) and hypertension risk was not significant (Table 4). Additionally, no statistically significant association was found between cumulative CMI quartiles and stroke risk.
Table 4.
Association between the CMI related indeces and stroke incidence in diabetes and prediabetets populations
| Label | Model1 | Model2 | Model3 | |||
|---|---|---|---|---|---|---|
| OR(95%CI) | P Value | OR(95%CI) | P Value | OR(95%CI) | P Value | |
| CMI Continuous | ||||||
| CMI | 1.20(1.03–1.39) | 1.69E-02 | 1.28(1.05–1.59) | 1.64E-02 | 1.31(1.05–1.67) | 2.32E-02 |
| CMI Quartiles | ||||||
| CMI_Q1 | 1.00(Reference) | 1.00(Reference) | 1.00(Reference) | |||
| CMI_Q2 | 1.70(1.01–2.95) | 5.02E-02 | 1.73(1.02–3.00.02.00) | 4.55E-02 | 1.74(1.02–3.04) | 4.54E-02 |
| CMI_Q3 | 2.24(1.35–3.83) | 2.36E-03 | 2.25(1.34–3.90) | 2.75E-03 | 2.32(1.36–4.09) | 2.50E-03 |
| CMI_Q4 | 1.99(1.18–3.43) | 1.11E-02 | 2.02(1.18–3.63) | 1.29E-02 | 2.06(1.15–3.88) | 1.81E-02 |
| C_CMI Continuous | ||||||
| C_CMI | 1.15(1.00–1.30.00.30) | 2.60E-02 | 1.14(0.99–1.30) | 4.06E-02 | 1.12(0.96–1.28) | 1.05E-01 |
| C_CMI Quartiles | ||||||
| C_CMI_Q1 | 1.00(Reference) | 1.00(Reference) | 1.00(Reference) | |||
| C_CMI_Q2 | 1.09(0.64–1.86) | 7.41E-01 | 1.10(0.65–1.88) | 7.21E-01 | 1.12(0.65–1.92) | 6.85E-01 |
| C_CMI_Q3 | 1.71(1.05–2.82) | 3.34E-02 | 1.69(1.03–2.83) | 4.01E-02 | 1.66(1.00–2.79.00.79) | 5.22E-02 |
| C_CMI_Q4 | 1.76(1.08–2.92) | 2.46E-02 | 1.78(1.07–3.05) | 2.84E-02 | 1.69(0.99–2.99) | 5.73E-02 |
| CMI Cluster | ||||||
| Cluster_1 | 1.00(Reference) | 1.00(Reference) | 1.00(Reference) | |||
| Cluster_2 | 1.43(1.01–2.03) | 4.19E-02 | 1.42(1.00–2.04.00.04) | 5.40E-02 | 1.38(0.95–2.00.95.00) | 9.00E-02 |
| Cluster_3 | 2.07(0.71–4.83) | 1.30E-01 | 2.13(0.73–5.03) | 1.17E-01 | 1.74(0.56–4.40) | 2.83E-01 |
Table 3 presents the results of the logistic regression analysis, which examined the relationship between changes in CMI and the incidence of hypertension and stroke. After adjusting for various risk factors in Model 3, compared with Class 1, Class 2 (OR = 1.39, 95% CI: 1.19–1.62, p = 3.73E-05) and Class 3 (OR = 2.38, 95% CI: 1.62–3.48, p = 8.94E-06) exhibited a higher risk of hypertension. However, compared with Category 1, there was no statistically significant difference in the increased risk of stroke for Category 2 (OR = 1.37, 95% CI: 0.96–1.97, p = 8.44E-02) and Category 3 (OR = 1.99, 95% CI: 0.77–5.15, p = 1.57E-01).
Table 3.
Association between the CMI related indeces and hypertension incidence in diabetes and prediabetets populations
| Label | Model1 | Model2 | Model3 | |||
|---|---|---|---|---|---|---|
| OR(95%CI) | P Value | OR(95%CI) | P Value | OR(95%CI) | P Value | |
| CMI Continuous | ||||||
| CMI | 1.27(1.17–1.37) | 5.13E-09 | 1.32(1.21–1.45) | 8.07E-10 | 1,27(1.15–1.40) | 4.46E-06 |
| CMI Quartiles | ||||||
| CMI_Q1 | 1.00(Reference) | 1.00(Reference) | 1.00(Reference) | |||
| CMI_Q2 | 1.17(0.91–1.49) | 2.16E-01 | 1.18(0.92–1.51) | 1.89E-01 | 1.15(0.89–1.48) | 2.97E-01 |
| CMI_Q3 | 1.55(1.22–1.96) | 3.32E-04 | 1.61(1.27–2.05) | 1.01E-04 | 1.53(1.19–1.97) | 1.06E-03 |
| CMI_Q4 | 1.95(1.55–2.47) | 1.58E-08 | 2.02(1.60–2.57) | 5.58E-09 | 1.86(1.43–2.41) | 3.53E-06 |
| C_CMI Continuous | ||||||
| C_CMI | 1.28(1.18–1.39) | 1.72E-09 | 1.30(1.20–1.42) | 1.18E-09 | 1.23(1.12–1.35) | 1.16E-05 |
| C_CMI Quartiles | ||||||
| C_CMI_Q1 | 1.00(Reference) | 1.00(Reference) | 1.00(Reference) | |||
| C_CMI_Q2 | 1.24(0.97–1.58) | 8.66E-02 | 1.25(0.98–1.60) | 7.07E-02 | 1.21(0.94–1.56) | 1.45E-01 |
| C_CMI_Q3 | 1.51(1.19–1.92) | 7.56E-04 | 1.57(1.23–2.00.23.00) | 2.88E-04 | 1.47(1.14–1.89) | 3.02E-03 |
| C_CMI_Q4 | 2.07(1.64–2.62) | 9.49E-10 | 2.15(1.69–2.73) | 3.09E-10 | 1.85(1.43–2.41) | 3.21E-06 |
| CMI Cluster | ||||||
| Cluster_1 | 1.00(Reference) | 1.00(Reference) | 1.00(Reference) | |||
| Cluster_2 | 1.47(1.24–1.73) | 5.32E-06 | 1.51(1.27–1.78) | 1.84E-06 | 1.44(1.21–1.71) | 4.20E-05 |
| Cluster_3 | 3.34(2.10–5.23) | 1.86E-07 | 3.33(2.09–5.22) | 2.29E-07 | 2.69(1.66–4.31) | 4.75E-05 |
The nonlinear relationship between cumulative CMI and the incidence of hypertension and stroke in diabetes and prediabetes, and its predictive performance
The RCS regression model revealed a nonlinear relationship between cumulative CMI and the risk of hypertension in individuals with impaired glucose metabolism (nonlinear p-value = 0.024). However, no nonlinear relationship was observed between cumulative CMI and stroke risk (nonlinear p-value = 0.247) (Fig. 3). ROC curves indicated that cumulative CMI had the highest diagnostic performance for hypertension (area under the curve [AUC] = 0.73) and stroke (AUC = 0.67). Baseline CMI at wave 3 (CMI_w3) demonstrated diagnostic performance for hypertension (AUC = 0.73) and stroke (AUC = 0.67) that was consistent with cumulative CMI (Fig. 4).
Fig. 3.
Restricted Cubic Spline Model Illustrating the Association Between Cardiometabolic Index(CMI) and Cumulative Cardiometabolic Index (C_CMI) and the Risk of Hypertension and Stroke. (A) CMI and risk of HBP; (B) C_CMI and risk of HBP; (C) CMI and risk of stroke; (D) C_CMI and risk of stroke
Fig. 4.
Receiver Operating Characteristic (ROC) curves comparing the predictive performance of the Cardiometabolic Index (CMI) and the cumulative CMI index for future hypertension and stroke among individuals with diabetes and prediabetes in different logistic model. A: CMI and predicted the future risk of hypertension in different logistic model. B: C_CMI and predicted the future risk of hypertension in different logistic model. C: CMI and predicted the future risk of stroke in different logistic model. D: C_CMI and predicted the future risk of stroke in different logistic model
Subgroup analysis and sensitivity analysis
Table 5; Fig. 5 illustrate the changes in CMI and cumulative CMI, stratified by various factors, and their associations with the risk of hypertension and stroke. Additionally, these figures depict the changes in cumulative CMI and their corresponding associations with hypertension and stroke risk. No significant interactions were observed among the other subgroups. A complete data analysis was conducted, and the results of the sensitivity analysis were consistent with the main findings of this study, thereby confirming the robustness of the results.
Table 5.
Association of CMI and cumulative CMI with the risk of stroke stratified by different factors
| Variable | N | Percent | OR | OR_LL | OR_UL | P value | P for interaction |
|---|---|---|---|---|---|---|---|
| CMI | |||||||
| Sex | 0.603 | ||||||
| Female | 2293 | 54.5 | 1.22 | 1 | 1.49 | 0.051 | |
| Male | 1914 | 45.5 | 1.13 | 0.92 | 1.39 | 0.244 | |
| Education | 0.879 | ||||||
| Below high school | 3788 | 90 | 1.15 | 0.98 | 1.35 | 0.079 | |
| High school or above | 419 | 10 | 1.19 | 0.85 | 1.65 | 0.308 | |
| Smoking | 0.646 | ||||||
| Ever smokers | 1844 | 43.8 | 1.12 | 0.89 | 1.41 | 0.326 | |
| Never smokers | 2363 | 56.2 | 1.2 | 1 | 1.44 | 0.044 | |
| Drinks | 0.541 | ||||||
| Ever drinkers | 1460 | 34.7 | 1.23 | 1 | 1.5 | 0.047 | |
| Never drinkers | 2747 | 65.3 | 1.12 | 0.92 | 1.37 | 0.253 | |
| Hukou | 0.835 | ||||||
| Rural | 3508 | 83.4 | 1.14 | 0.97 | 1.35 | 0.105 | |
| Urban | 699 | 16.6 | 1.2 | 0.82 | 1.74 | 0.356 | |
| Marital_status | 0.87 | ||||||
| Partnered | 3636 | 86.4 | 1.16 | 0.99 | 1.35 | 0.071 | |
| Separated | 571 | 13.6 | 1.2 | 0.8 | 1.78 | 0.377 | |
| Heart_problem | 0.869 | ||||||
| No | 3665 | 87.1 | 1.14 | 0.96 | 1.35 | 0.139 | |
| Yes | 542 | 12.9 | 1.17 | 0.89 | 1.53 | 0.261 | |
| Dyslipidemia | 0.778 | ||||||
| No | 3935 | 93.5 | 1.15 | 0.98 | 1.34 | 0.082 | |
| Yes | 272 | 6.5 | 1.06 | 0.61 | 1.82 | 0.839 | |
| BMI | 0.059 | ||||||
| <18.5 | 296 | 7 | 4.29 | 1.81 | 10.14 | 0.001 | |
| 18.5–24.9 | 2617 | 62.2 | 1.25 | 0.96 | 1.62 | 0.097 | |
| 25–30 | 1101 | 26.2 | 1.03 | 0.7 | 1.52 | 0.869 | |
| >30 | 193 | 4.6 | 1.03 | 0.66 | 1.62 | 0.881 | |
| Age | 0.876 | ||||||
| <60 | 2020 | 48 | 1.18 | 0.96 | 1.45 | 0.117 | |
| >60 | 2187 | 52 | 1.15 | 0.95 | 1.41 | 0.157 | |
| C_CMI | |||||||
| Sex | 0.54 | ||||||
| Female | 2293 | 54.5 | 1.1 | 0.96 | 1.27 | 0.171 | |
| Male | 1914 | 45.5 | 1.19 | 0.96 | 1.48 | 0.106 | |
| Education | 0.211 | ||||||
| Below high school | 3788 | 90 | 1.1 | 0.97 | 1.24 | 0.129 | |
| High school or above | 419 | 10 | 1.48 | 0.94 | 2.32 | 0.088 | |
| Smoking | 0.665 | ||||||
| Ever smokers | 1844 | 43.8 | 1.17 | 0.95 | 1.44 | 0.149 | |
| Never smokers | 2363 | 56.2 | 1.11 | 0.96 | 1.27 | 0.152 | |
| Drinks | 0.557 | ||||||
| Ever drinkers | 1460 | 34.7 | 1.09 | 0.95 | 1.26 | 0.208 | |
| Never drinkers | 2747 | 65.3 | 1.18 | 0.96 | 1.45 | 0.12 | |
| Hukou | 0.385 | ||||||
| Rural | 3508 | 83.4 | 1.17 | 1 | 1.38 | 0.057 | |
| Urban | 699 | 16.6 | 1.03 | 0.8 | 1.32 | 0.829 | |
| Marital_status | 0.916 | ||||||
| Partnered | 3636 | 86.4 | 1.11 | 1 | 1.25 | 0.061 | |
| Separated | 571 | 13.6 | 1.14 | 0.72 | 1.8 | 0.565 | |
| Heart_problem | 0.526 | ||||||
| No | 3665 | 87.1 | 1.15 | 0.97 | 1.36 | 0.109 | |
| Yes | 542 | 12.9 | 1.06 | 0.88 | 1.27 | 0.533 | |
| Dyslipidemia | 0.831 | ||||||
| No | 3935 | 93.5 | 1.11 | 0.99 | 1.25 | 0.083 | |
| Yes | 272 | 6.5 | 1.05 | 0.62 | 1.76 | 0.863 | |
| BMI | 0.066 | ||||||
| <18.5 | 296 | 7 | 4.57 | 1.73 | 12.08 | 0.002 | |
| 18.5–24.9 | 2617 | 62.2 | 1.17 | 0.94 | 1.45 | 0.165 | |
| 25–30 | 1101 | 26.2 | 1.04 | 0.82 | 1.32 | 0.735 | |
| >30 | 193 | 4.6 | 1.13 | 0.58 | 2.2 | 0.726 | |
| Age | 0.871 | ||||||
| <60 | 2020 | 48 | 1.11 | 0.97 | 1.27 | 0.113 | |
| >60 | 2187 | 52 | 1.14 | 0.92 | 1.4 | 0.225 |
Fig. 5.
Subgroup Analysis Results: Association Between Cardiometabolic Index (CMI) and Hypertension Risk in Individuals with Diabetes and Prediabetes
Discussion
This prospective cohort study included 4,207 participants with impaired glucose metabolism and aimed to explore the relationship between changes in CMI, cumulative CMI, and clustered CMI and the risk of hypertension and stroke. The study identified three distinct CMI trajectory models and four cumulative CMI quartiles. Compared with the low-level stable group, both the moderate-level stable group and the high-level declining group exhibited a significant association between CMI and hypertension risk. However, no significant statistical association was found with stroke risk. While CMI was associated with both hypertension and stroke risk, cumulative CMI was only associated with hypertension risk. Similarly, compared with the first quartile, the third and fourth quartiles of cumulative CMI were significantly associated with hypertension risk. Subgroup analysis and sensitivity analysis results were consistent, thereby confirming the robustness of the study findings.
In this study, K-means clustering analysis was employed to categorize individuals based on changes in CMI and cumulative CMI. Compared with the first group of participants, who maintained a consistently low and stable CMI, the second group (characterized by moderate and stable CMI) and the third group (with high CMI and a slow downward trend) exhibited a higher risk of hypertension. Similar findings were observed when participants were grouped according to cumulative CMI quartiles. Individuals with higher cumulative CMI were found to have a higher risk of hypertension. This study represents the first investigation into the association between the combined index CMI of TG/HDL-C and WHtR and the risk of hypertension and stroke in individuals with impaired glucose metabolism. Our results suggest that CMI may increase the risk of hypertension in individuals with impaired glucose metabolism.
These findings may be closely related to several underlying mechanisms. The TG/HDL-C ratio serves as an indicator of lipid metabolism disorders and insulin resistance, which can increase the risk of atherosclerosis, cardiovascular events, and metabolic syndrome [17–19]. An elevated WHtR is associated with an increased overall mortality rate from coronary heart disease and diabetes [20–23]. First, a higher WHtR is closely linked to increased visceral fat [24]. The accumulation of visceral fat promotes insulin resistance and metabolic syndrome, thereby further increasing the risk of complications associated with impaired glucose metabolism [24–27]. Excessive visceral fat tissue leads to adipocyte hypertrophy and dysfunction, triggering a series of inflammatory responses [28]. Adipocytes release pro-inflammatory factors such as tumor necrosis factor-α and interleukin-6, which can trigger systemic inflammation and impair vascular function, thereby increasing the risk of hypertension [29]. On the other hand, dyslipidemia is closely associated with atherosclerosis. The accumulation of LDL particles on arterial walls increases the formation of atherosclerotic plaques, which is a key factor in the development of hypertension and other cardiovascular diseases [30]. Additionally, lipid metabolism disorders may influence blood pressure regulation through oxidative stress and endothelial dysfunction [31, 32]. Imbalances in the production and clearance of reactive oxygen species can lead to cellular and tissue damage, not only directly impairing vascular endothelial cells but also indirectly affecting blood pressure regulation by promoting inflammatory responses and the progression of atherosclerosis [33–35]. Third, insulin resistance may lead to elevated blood insulin levels, which can increase blood volume by affecting renal sodium retention and directly influence blood pressure by activating the sympathetic nervous system and promoting vascular smooth muscle cell proliferation [36]. Elevated levels of insulin-like growth factor 1 may also affect vascular remodeling by promoting atherosclerosis, thereby increasing the risk of hypertension [37].
The association between CMI and hypertension risk in individuals with abnormal glucose metabolism is likely not attributable to the isolated effect of a single factor, but rather to the potential interactions among these indicators. The cardiometabolic index (CMI) integrates a marker of central obesity (waist-to-height ratio) with an index of atherogenic dyslipidaemia (TG/HDL-C). By combining these two pathophysiologically related dimensions, CMI may better capture the metabolic milieu that predisposes to hypertension and cardiovascular disease than either anthropometric or single lipid measures alone [13, 38, 39]. Empirical studies in large Chinese and international cohorts have shown that CMI is independently associated with hypertension and cardiovascular outcomes and often demonstrates greater discrimination than conventional single measures (e.g., BMI, WC, TG or TG/HDL-C alone) [13, 38, 39]. Central obesity and lipid metabolism disorders work synergistically to promote hypertension through shared metabolic pathways and pathological mechanisms. CMI integrates changes in lipid metabolism, insulin resistance, and other factors, thereby offering a more comprehensive reflection of an individual’s metabolic health status. By consolidating multiple metabolic abnormalities, CMI enhances the accuracy of predicting the onset of hypertension.
Further analysis using restricted cubic splines revealed a nonlinear relationship between CMI and hypertension risk. Specifically, the association between cumulative CMI and hypertension risk exhibited an inverted L-shaped curve. Compared with the first quartile, there was a rapid increase in hypertension risk in the third and fourth quartiles of cumulative CMI. However, no significant association was observed between cumulative CMI and hypertension risk in the second quartile.This study is the first to elucidate the complex relationship between CMI and hypertension in individuals with impaired glucose metabolism. The rapid increase in hypertension risk associated with higher cumulative CMI is closely linked to visceral fat accumulation and worsening atherosclerosis [12, 24–27, 39]. Additional ROC curve analysis indicated that CMI has significant predictive value for the occurrence of hypertension and stroke. Therefore, future research should further investigate the application of CMI across different populations to better assess whether combining CMI with other indicators can enhance predictive accuracy. This would provide stronger scientific evidence for its potential clinical application. Subgroup analyses confirmed that the results are applicable across different populations and align with the main findings.
Additionally, this study did not reveal a significant association between CMI and stroke risk in individuals with impaired glucose metabolism. As shown in Table 3, and 4, the association between CMI and stroke risk in Models 1 and 2 was very close to the statistical significance threshold. Specifically, in Table 4, the results of Model 1 and Model 2 are statistically significant. Therefore, although this study failed to confirm the relationship between CMI and stroke risk, this may be due to the relatively low incidence of stroke in the study population. Increasing the sample size may enhance the likelihood of identifying a positive correlation. In summary, future large-scale prospective studies will further explore and clarify the causal relationship between CMI and stroke risk in individuals with impaired glucose metabolism, which will be more persuasive.
Traditional cardiovascular risk assessment models (such as the Framingham Risk Score (FRS)) mainly consider age, gender, blood pressure, cholesterol levels, smoking and diabetes status as risk factors. These factors do not directly capture central obesity or metabolic disorders, which play a key role in the pathogenesis of hypertension and stroke, especially in individuals with impaired glucose metabolism. In contrast, CMI integrates the triglyceride-high-density lipoprotein-C ratio and waist-to-height ratio, combining indicators of dyslipidemia and visceral fat, both of which are at the core of insulin resistance and vascular dysfunction. Therefore, CMI can more directly reflect the cardiometabolic risk of patients with diabetes and prediabetes. In addition, CMI can be easily calculated using conventional clinical data without the need for complex algorithms or clinical history, making it a practical and cost-effective tool for large-scale screening and follow-up. Future research directly comparing the predictive accuracy of CMI with established scores, such as the predictive accuracy of FRS in different metabolic subgroups, will be valuable to confirm its incremental clinical application and explore whether incorporating CMI into traditional risk models can improve prognostic performance.
While the CMI and metabolic syndrome (MetS) share multiple common indicators—including triglycerides, HDL-C, and central obesity metrics—their clinical interpretation and application differ significantly. MetS, as a binary diagnostic criterion, identifies patients by meeting at least three of five predefined criteria (e.g., elevated blood pressure and fasting blood glucose). In contrast, CMI, as a continuous quantitative metric, integrates lipid profile abnormalities and central obesity data to more accurately assess cumulative metabolic burden. Crucially, CMI requires no blood pressure or fasting glucose data, relying solely on routine lipid measurements and anthropometric data, making it particularly suitable for large-scale screening and early identification of high-risk populations in resource-constrained or community settings. Although MetS covers a broader spectrum of cardiometabolic disorders, CMI quantifies metabolic dysfunction below diagnostic thresholds more effectively. It should be noted that due to incomplete data from MetS indicators, this study did not directly compare CMI with MetS. Future prospective research should evaluate whether CMI offers predictive value for hypertension and stroke risk that surpasses MetS.
The strength of this study lies in its utilization of a prospective cohort design to examine the relationship between CMI and the risk of hypertension and stroke in individuals with impaired glucose metabolism. After adjusting for a variety of risk factors, we identified a significant association between CMI and hypertension risk among those with abnormal blood glucose levels. The robustness of these findings was further confirmed through sensitivity analyses. Additionally, we employed the K-means clustering method to systematically delineate long-term trajectories of CMI changes. This approach yielded reliable and distinct groupings, with individuals within each group demonstrating specific and consistent changes in their CMI indices over the follow-up period. The resulting clusters were not only robust but also endowed with enhanced statistical power.
This study also has several limitations that should be acknowledged. First, all participants in this study were Chinese middle-aged and older adults aged 45 years or older, with an average age of 61.7 years. Therefore, these results are primarily applicable to the middle-aged and older adult population in China, and further research is needed to confirm their applicability to other countries and populations. Second, the cases of hypertension and stroke included in this study were self-reported by patients and diagnosed by physicians, which may introduce classification bias. However, hypertension cases were validated through multiple sources of information, including measured systolic and diastolic blood pressure and antihypertensive medication use, while stroke cases were confirmed through validation with anti-stroke medication use. Additionally, previous studies have shown that self-reported data in this cohort typically align with medical records, and misreporting is non-systematic, indicating minimal misreporting error. Furthermore, the number of newly diagnosed stroke patients included in this study was very small, with only 97 cases. As shown in Tables 3 and 4, there was a statistically significant association between CMI and stroke risk in Models 1 and 2, but no association was observed in Model 3. Increasing the sample size and the number of stroke patients included may better elucidate the relationship between CMI and stroke risk. Another limitation of the present study is that we did not perform direct, head-to-head evaluations of CMI against each of its component variables (TG, TG/HDL-C, and waist-to-height ratio/WC) using formal incremental metrics (e.g., ΔAUC, net reclassification improvement, or integrated discrimination improvement). Such analyses would be valuable to determine whether CMI provides additional predictive information beyond its individual components. We therefore recommend that future prospective studies explicitly compare CMI with its individual components using ROC comparison and reclassification analyses to establish its incremental clinical utility. Besides, although we adjusted for potential confounding factors affecting hypertension and stroke risk in the models, unmeasured confounding factors may still influence the results. Another limitation of this study is the lack of systematic data on mortality during follow-up, which prevented assessment of death as a competing risk. Given the characteristics of the study population, the potential impact of mortality on hypertension and stroke incidence is likely small, but future longitudinal studies should include competing risk analyses to validate our findings Table 5.
Conclusion
In summary, elevated CMI indices in individuals with impaired glucose metabolism are associated with an increased risk of hypertension and stroke. Our findings suggest that maintaining relatively low CMI index levels may help individuals with impaired glucose metabolism prevent the onset of hypertension and stroke. Given that the CMI index is easily accessible in both urban and rural communities, clinicians can utilize changes in this indicator to stratify population risk. This approach enables more personalized strategies for mitigating the risk of hypertension and stroke-related complications in individuals with impaired glucose metabolism.
Acknowledgements
We thank all workers who contributed to the CHARLS database.
Abbreviations
- CMI
Cardiovascular metabolic index
- RCS
Restricted cubic spline
- TG
Triglycerides
- HDL-C
High-density lipoprotein cholesterol ratio
- WHtR
Waist-to-height ratio
- NHANES
National Health and Nutrition Examination Survey
- BMI
Body mass index
- CHARLS
China Health and Retirement Longitudinal Study
- HbA1c
Hemoglobin A1c
- LDL-C
Low-density lipoprotein cholesterol
- CMI_w1
CMI at wave 1
- CMI_w3
CMI at wave 3
- AUC
Area under the curve
Author contributions
YS and YZ conceived and designed the study and outlined the structure of the manuscript. YS, XZ and JX were responsible for selecting the references and contributed to the writing process. YS and XZ undertook the data cleaning of the CHARLS dataset. YS, XZ, CZ and JX performed the statistical analyses for this study. YZ, GC and JX contributed to the revision and finalization of the manuscript. GC guided the revision of the manuscript. All authors participated in the preparation and approved the final version of the manuscript.
Funding
This work was supported by the Natural Science Foundation of Hunan Province, China(Grant No. 2025JJ80441).
Data availability
The original contributions presented in the study are included in the manuscript. Further inquiries can be directed to the corresponding author.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Gaoyuan Cui, Yi Zeng and Jian Xia are contributed equally to this work.
References
- 1.Global (2023) regional, and National burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the global burden of disease study 2021. Lancet 402(10397):203–234 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.AhnSK et al (2021) Incidence hypertension and fasting blood glucose from real-world data: retrospective cohort for 7-years follow-up. Int J Environ Res Public Health 18(4):2085. 10.3390/ijerph18042085. Accessed 21 Fed 2021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Li S et al (2025) Association of hypertension and long-term blood pressure changes with new-onset diabetes in the elderly: A 10-year cohort study. Diabetes Obes Metab 27(1):92–101 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Doumas M et al (2013) Gender differences in hypertension: myths and reality. Curr Hypertens Rep 15(4):321–330 [DOI] [PubMed] [Google Scholar]
- 5.Shi WR et al (2018) Estimate of prevalent diabetes from cardiometabolic index in general Chinese population: a community-based study. Lipids Health Dis 17(1):236 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Cheng H, He X, Jin X (2024) The relationship between cardiometabolic index and infertility in American adults: a population-based study. Front Endocrinol (Lausanne) 15:1424033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wu X et al (2024) Cardiometabolic index is associated with increased bone mineral density: a population-based cross-sectional study. Front Public Health 12:1403450 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Zhu XM, Xu Y, Zhang J (2024) Cardiometabolic index is associated with heart failure: a cross-sectional study based on NHANES. Front Med (Lausanne) 11:1507100 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Sun Q et al (2023) Cardiometabolic index (CMI), lipid accumulation products (LAP), waist triglyceride index (WTI) and the risk of acute pancreatitis: a prospective study in adults of North China. Lipids Health Dis 22(1):190 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Chen M et al (2025) Association between cardiometabolic index and gestational diabetes mellitus: a cross-sectional study. Endocrine 87(2):569–577 [DOI] [PubMed] [Google Scholar]
- 11.Song J et al (2024) Non-linear associations of cardiometabolic index with insulin resistance, impaired fasting glucose, and type 2 diabetes among US adults: a cross-sectional study. Front Endocrinol (Lausanne) 15:1341828 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Zhou H et al (2025) Exploring the nonlinear association between cardiometabolic index and hypertension in U.S. Adults: an NHANES-based study. BMC Public Health 25(1):1092 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wang H et al (2018) Validity of cardiometabolic index, lipid accumulation product, and body adiposity index in predicting the risk of hypertension in Chinese population. Postgrad Med 130(3):325–333 [DOI] [PubMed] [Google Scholar]
- 14.Zhao Y et al (2014) Cohort profile: the China health and retirement longitudinal study (CHARLS). Int J Epidemiol 43(1):61–68 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.He D et al (2024) Changes in frailty and incident cardiovascular disease in three prospective cohorts. Eur Heart J 45(12):1058–1068 [DOI] [PubMed] [Google Scholar]
- 16.Ren Q et al (2024) Association between triglyceride glucose-waist height ratio index and cardiovascular disease in middle-aged and older Chinese individuals: a nationwide cohort study. Cardiovasc Diabetol 23(1):247 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Perla FM et al (2017) The role of lipid and lipoprotein metabolism in non-alcoholic fatty liver disease. Children (Basel) 4(6):46. 10.3390/children4060046. Accessed 6 Jun 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Sultani R et al (2020) Elevated triglycerides to High-Density lipoprotein cholesterol (TG/HDL-C) ratio predicts Long-Term mortality in High-Risk patients. Heart Lung Circ 29(3):414–421 [DOI] [PubMed] [Google Scholar]
- 19.Che B et al (2023) Triglyceride-glucose index and triglyceride to high-density lipoprotein cholesterol ratio as potential cardiovascular disease risk factors: an analysis of UK biobank data. Cardiovasc Diabetol 22(1):34 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chandramouli C et al (2019) Association of obesity with heart failure outcomes in 11 Asian regions: a cohort study. PLoS Med 16(9):e1002916 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Liu Y et al (2021) Association of anthropometric indices with the development of diabetes among hypertensive patients in china: A cohort study. Front Endocrinol (Lausanne) 12:736077 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.General and abdominal adiposity and hypertension in eight world regions: a pooled analysis of 837 population-based studies with 7·5 million participants. Lancet, (2024) 404(10455): pp. 851–863 [DOI] [PMC free article] [PubMed]
- 23.Quetglas-Llabrés MM et al (2022) Inflammatory and oxidative stress markers related to adherence to the Mediterranean diet in patients with metabolic syndrome. Antioxidants 11(5):901. 10.3390/antiox11050901. Accessed 1 May 2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Li X et al (2022) Predictive effect of triglyceride Glucose-Related Parameters, obesity Indices, and lipid ratios for diabetes in a Chinese population: A prospective cohort study. Front Endocrinol (Lausanne) 13:862919 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Muir LA et al (2016) Adipose tissue fibrosis, hypertrophy, and hyperplasia: correlations with diabetes in human obesity. Obes (Silver Spring) 24(3):597–605 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Verboven K et al (2018) Abdominal subcutaneous and visceral adipocyte size, lipolysis and inflammation relate to insulin resistance in male obese humans. Sci Rep 8(1):4677 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Das R et al (2022) A STAT5-Smad3 dyad regulates adipogenic plasticity of visceral adipose mesenchymal stromal cells during chronic inflammation. NPJ Regen Med 7(1):41 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Van Herck MA et al (2020) Diet reversal and immune modulation show key role for liver and adipose tissue T cells in murine nonalcoholic steatohepatitis. Cell Mol Gastroenterol Hepatol 10(3):467–490 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Aljafary MA, Al-Suhaimi EA (2022) Adiponectin system (rescue hormone): the missing link between metabolic and cardiovascular diseases. Pharmaceutics 14(7):1430. 10.3390/pharmaceutics14071430. Accessed 7 Jul 2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Mundal LJ et al (2022) Association of familial hypercholesterolemia and statin use with risk of dementia in Norway. JAMA Netw Open 5(4):e227715 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Aday AW et al (2018) Lipoprotein particle Profiles, standard Lipids, and peripheral artery disease incidence. Circulation 138(21):2330–2341 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Kimura T et al (2018) Regulatory CD4(+) T cells recognize major histocompatibility complex class II Molecule-Restricted peptide epitopes of Apolipoprotein B. Circulation 138(11):1130–1143 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wang R et al (2023) A TEMPOL and Rapamycin loaded nanofiber-covered stent favors endothelialization and mitigates neointimal hyperplasia and local inflammation. Bioact Mater 19:666–677 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Masenga SK et al (2023) Mechanisms of oxidative stress in metabolic syndrome. Int J Mol Sci 24(9):7898. 10.3390/ijms24097898. Accessed 26 Apr 2023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Min BK et al (2019) Therapeutic effect of dichloroacetate against atherosclerosis via hepatic FGF21 induction mediated by acute AMPK activation. Exp Mol Med 51(10):1–12 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Agbaje AO (2022) Mediating role of body composition and insulin resistance on the association of arterial stiffness with blood pressure among adolescents: the ALSPAC study. Front Cardiovasc Med 9:939125 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.De Giorgi A et al (2022) Insulin-like growth factor-1 (IGF-1) as predictor of cardiovascular mortality in heart failure patients: data from the T.O.S.CA. registry. Intern Emerg Med 17(6):1651–1660 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Liu C et al (2025) The impact of cardiometabolic index on cardiovascular disease risk among diabetic patients: evidence from two National cohorts. Diabetes Metab Res Rev 41(4):e70044 [DOI] [PubMed] [Google Scholar]
- 39.Li F et al (2022) Anthelmintics Nitazoxanide protects against experimental hyperlipidemia and hepatic steatosis in hamsters and mice. Acta Pharm Sin B 12(3):1322–1338 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The original contributions presented in the study are included in the manuscript. Further inquiries can be directed to the corresponding author.