Baseline versus cumulative cholesterol, high-density lipoprotein, and glucose index and stroke risk in individuals with cardiovascular-kidney-metabolic syndrome stages 0–3: evidence from the CHARLS study Authors: Chen Chen, Xiao Yang, Cheng Li, Qiqi Xi, Yi Ding Published: 2026 ABSTRACT BackgroundThe cholesterol, high-density lipoprotein, and glucose (CHG) index has been linked to stroke risk. However, evidence is largely based on single-time-point measurements and has not been compared with cumulative exposure, nor evaluated in individuals with cardiovascular-kidney-metabolic syndrome (CKM) stages 0–3. This study aimed to compare the associations of baseline and cumulative CHG index with incident stroke risk in this population.MethodsData were derived from the China Health and Retirement Longitudinal Study (CHARLS). The exposure measures were the baseline and cumulative levels of the CHG index. Multivariable logistic regression models were employed to examine the association between the baseline and cumulative CHG index and stroke risk. Restricted cubic spline (RCS) analysis was used to test potential dose-response relationships. The receiver operating characteristic (ROC) curve was used to compare the predictive performance of the two measures. Subgroup analyses were conducted to evaluate the consistency of these associations in different populations.ResultsAmong 4,661 participants with CKM stages 0–3, 476 (10.2%) developed incident stroke during follow-up. After adjustment for demographic characteristics, lifestyle factors, anthropometric measurement, laboratory parameters, medication use and comorbidities, both the baseline CHG index (OR = 1.42, 95% CI: 1.12–1.81) and cumulative CHG index (OR = 1.14, 95% CI: 1.03–1.26) were positively associated with stroke risk. Participants in the highest tertile (T3) of the baseline CHG index had a 35% higher stroke risk than those in the lowest tertile (T1) (OR = 1.35, 95% CI: 1.03–1.76), and a 34% elevated risk was observed for the cumulative CHG index (OR = 1.34, 95% CI: 1.03–1.76). RCS analysis revealed linear relationships for both measures (p for non-linear = 0.331 and 0.817, respectively; overall p < 0.05 for both). ROC curve analysis showed no statistically significant difference in the area under the curve (AUC) between the two measures (0.672 vs. 0.671, p = 0.330). Subgroup analyses indicated that the associations might be modified by gender and drinking status.ConclusionsAmong individuals with CKM stages 0–3, both baseline and cumulative CHG indices were independently associated with incident stroke risk. Given its moderate discriminatory capacity, the CHG index—derived from routine blood tests—may serve as a preliminary screening tool to flag high-risk individuals, prompting more comprehensive stroke risk evaluation.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40842-026-00315-2. Introduction Cardiovascular-kidney-metabolic syndrome (CKM), proposed by the American Heart Association (AHA), systematically describes the complex interplay among metabolic disorders, chronic kidney disease and cardiovascular disease [1]. This framework highlights a continuum from metabolic dysfunction to multi-organ damage, particularly cardiovascular events [1, 2]. Epidemiological data indicate that nearly 90% of adults in the United States meet the criteria for CKM stage 1 or higher, and approximately 15% are classified as having advanced stages [3]. Among CKM outcomes, stroke imposes a substantial burden [4]. According to the Global Burden of Disease Study 2021, stroke accounted for 7.3 million deaths, representing 10.7% of total global mortality, and ranked third after ischaemic heart disease and COVID-19 [5]. The stroke-related disease burden is particularly pronounced in regions with a moderate sociodemographic index, such as Southeast Asia and the Western Pacific [6]. Of note, the prevalence of CKM components increased substantially during the COVID-19 pandemic [7]; moreover, depressive symptoms are associated with CKM progression in a sex-specific manner, with significant effects observed only in women [8].Consequently, the AHA emphasises active screening for modifiable stroke risk factors during CKM stages 0–3 as key to early prevention [1]. Multiple studies have reported that insulin resistance (IR) is a fundamental pathophysiological factor driving both CKM progression and stroke occurrence [9, 10]. Under the CKM framework, pulse pressure partially mediates the association between the triglyceride-glucose (TyG) index and renal function deterioration, and hypertension amplifies this adverse effect [11]. The gold standard for assessing IR remains the hyperinsulinaemic-euglycaemic clamp (HEC) [12]; however, its operational complexity and high cost preclude large-scale screening [13], underscoring the urgent need for simple and feasible surrogate markers. The recently proposed cholesterol, high-density lipoprotein, and glucose (CHG) index–calculated by logarithmic transformation of total cholesterol, high-density lipoprotein (HDL), and fasting blood glucose (FBG)–integrates both lipid and glucose metabolism [14]. Studies have shown that, in populations with normal glucose regulation (NGR) and pre-diabetes (Pre-DM), the CHG index is positively associated with stroke risk [15] and outperforms the TyG index in predicting cardiovascular events [16]. Compared with these alternatives, Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) requires measurement of fasting insulin and is costly, whereas the TyG index does not include HDL, rendering it less comprehensive. This distinction is critical, as the role of HDL is not linear: in hypertensive patients, very high HDL levels (> 80 mg/dL) paradoxically increase cardiovascular risk, and this U-shaped association is more pronounced in men [17], highlighting the need to incorporate HDL when assessing metabolic risk. Furthermore, unlike comprehensive tools such as the Framingham Risk Score or China-PAR model, which require blood pressure, waist circumference, and lifestyle information, the CHG index relies only on routine lipid and glucose measurements, making it more suitable for large-scale population screening or primary healthcare settings. Although multi-omics and artificial intelligence are now being used to build multidimensional biomarker systems covering multiple pathological axes [18], the CHG index–a simple test based on routine blood assays–still offers distinct value for initial screening in the above settings. A recent editorial has noted that CKM management is moving from a conceptual model toward actionable clinical pathways, with structured lifestyle and risk-factor management and emerging pharmacotherapies (e.g., SGLT2 inhibitors, GLP-1 receptor agonists) offering additional options for early-stage patients [19]. Therefore, using data from CHARLS, the present study aimed to evaluate the associations of the CHG index, both baseline and cumulative exposure levels, with incident stroke risk among individuals with CKM stages 0–3. The findings are expected to provide a low-cost and accessible haematological biomarker for identifying individuals at high risk of stroke during the early CKM stages, thereby facilitating targeted interventions and reducing the stroke burden. Materials and methods Data source and study population Data were obtained from the CHARLS database (http://charls.pku.edu.cn/), a nationally representative, ongoing prospective cohort study of community-dwelling Chinese adults aged 45 years and older, designed to monitor their health, socioeconomic status, and demographic characteristics. The survey employed a multistage stratified probability sampling design and, by 2020, had completed five waves of data collection. CHARLS was approved by the Institutional Review Board of Peking University (IRB00001052–11015). All field staff received systematic professional training and conducted face-to-face interviews using a standardised questionnaire. Written informed consent was obtained from all participants before data collection [20]. This secondary analysis was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The 2011–2012 wave served as the baseline. The cumulative CHG index was calculated as the mean of the CHG values from 2012 to 2015, multiplied by the time interval (2015 − 2012) [21]. Incident stroke events occurring during the 2018 and 2020 follow-up waves were the study outcomes. The detailed participant selection process is shown in Fig. 1. Fig. 1Flowchart of the study population. This figure illustrates the stepwise selection of participants from the 2011–2012 CHARLS survey. To avoid reverse causality and focus on the early prevention window, participants with CKM stage 4 (which already includes clinical cardiovascular events such as stroke) were excluded. Exclusion criteria included: age under 45 years, absence of CKM staging data, CKM stage 4, missing follow‑up stroke data, missing CHG data for both 2011–2012 and 2015, and missing covariate data. The numbers shown under each box represent the sample size remaining at each stage of inclusion and exclusion Flowchart of the study population. This figure illustrates the stepwise selection of participants from the 2011–2012 CHARLS survey. To avoid reverse causality and focus on the early prevention window, participants with CKM stage 4 (which already includes clinical cardiovascular events such as stroke) were excluded. Exclusion criteria included: age under 45 years, absence of CKM staging data, CKM stage 4, missing follow‑up stroke data, missing CHG data for both 2011–2012 and 2015, and missing covariate data. The numbers shown under each box represent the sample size remaining at each stage of inclusion and exclusion Calculation formulas for CHG index and cumulative CHG index \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CHG{\rm{ }}index\, = \,Ln{\rm{ }}\left[ \matrix{ TC{\rm{ }}\left({mg/dL} \right){\rm{ }} \hfill \cr \times {\rm{ }}FBG{\rm{ }}\left({mg/dL} \right){\rm{ }} \hfill \cr /{\rm{ }}\left({2{\rm{ }} \times {\rm{ }}HDL{\rm{ }}\left({mg/dL} \right)} \right) \hfill \cr} \right]$$\end{document} Cumulative CHG index: (CHG2012 + CHG2015) / 2 × 3 years. While this method assumes a linear change, it may not capture long-term dynamic variations; this limitation is discussed in the Discussion section. Definition of CKM syndrome stages 0–4 According to the AHA diagnostic criteria [1], CKM syndrome is classified into stages 0 to 4, where stage 0 represents no CKM-related risk factors, stage 1 indicates excess or dysfunctional adiposity, stage 2 denotes metabolic risk factors or chronic kidney disease, stage 3 indicates subclinical cardiovascular abnormalities, and stage 4 represents clinical cardiovascular disease (including stroke) [1]. Detailed staging criteria are provided in the supplementary material (Supplementary Table 10). Stroke diagnosis Stroke was the primary outcome of this study. It was identified based on participants’ self-reports during the 2018 and 2020 follow-up waves. To minimise bias, individuals who had been diagnosed with stroke before 2015 were excluded from the analysis. Participants were classified as having experienced a stroke event if they responded affirmatively to the question, “Have you been diagnosed with a stroke by a doctor?” or if they answered “yes” to any of the questions regarding the methods used to manage their stroke, including traditional Chinese medicine, Western medicine, physiotherapy, acupuncture, and occupational therapy. This method of outcome ascertainment has been used in previous population-based studies and is considered valid for stroke surveillance in large-scale cohort settings [22, 23]. Covariates Covariates were selected from the CHARLS database and included the following: demographic characteristics (age, gender, hukou (household registration system: agricultural/non-agricultural), education (below high school/high school or above), marital status (married/other)); anthropometric measurements (body mass index, height, weight); lifestyle factors (smoking status (never/former/current), drinking status (never/former/current)); laboratory parameters (triglycerides, total cholesterol, fasting blood glucose, high-density lipoprotein cholesterol, C-reactive protein); medication use (antihypertensive drugs, antihyperlipidemic drugs, and antidiabetic drugs); and comorbidities (hypertension, diabetes, cancer, liver disease, lung disease, kidney disease). Hypertension and diabetes were defined on the basis of self-reported physician diagnosis, medication use, and physical examination/laboratory tests; see Supplementary Table 10 for details [24]. Statistical analysis All statistical analyses were performed using R software (version 4.5.2). Continuous variables were described according to their distributions. Normally distributed variables are presented as mean ± standard deviation, with between-group comparisons conducted using analysis of variance (ANOVA). Non-normally distributed variables are expressed as medians (interquartile range), and between-group comparisons were performed using the Kruskal-Wallis H test. Categorical variables are presented as frequencies (percentages), and between-group comparisons were conducted using the χ²-test. All hypothesis tests were two-sided, and p-values < 0.05 were considered statistically significant. Multivariable logistic regression models were used to evaluate the associations of the baseline and cumulative CHG index with incident stroke risk, treating stroke as a binary outcome over the fixed follow-up period. Results are presented as odds ratios (ORs) with 95% confidence intervals (CIs). Three sequentially adjusted models were constructed to control for confounding factors: a crude model with no adjustment; Model 1, adjusted for age, gender, hukou, education, and marital status; and Model 2, which further adjusted for smoking status, drinking status, C-reactive protein, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, hypertension, diabetes, cancer, liver disease, lung disease, and kidney disease. Restricted cubic splines (RCS) with three knots were fitted to examine potential dose-response relationships. Receiver operating characteristic (ROC) curve analysis was performed to compare the predictive performance of the baseline and cumulative CHG index, and DeLong’s test was used to assess differences in the area under the curve (AUC). Subgroup analyses were carried out by repeating the fully adjusted model across strata of age, gender, hukou, marital status, education, smoking status, drinking status, hypertension, diabetes, and CKM stages 0–3 (categorised as CKM 0–1, CKM 2, and CKM 3). All primary analyses were based on complete-case data, excluding participants with missing values for any covariate. To assess the robustness of our findings, several sensitivity analyses were conducted. First, multiple imputation by chained equations (MICE) was performed to handle missing data (n = 551). Five imputed datasets were generated with 20 iterations, and the results were pooled using Rubin’s rules. Second, the CHG index (both baseline and cumulative) was modelled separately as a continuous variable and as tertiles in the regression models. Third, Cox proportional hazards models were fitted instead of logistic regression to account for time-to-event, and the proportional hazards assumption was tested using Schoenfeld residuals. Fourth, competing risk regression (Fine-Gray model) was performed, treating all-cause death as a competing event. All sensitivity analyses were conducted for both baseline and cumulative CHG indices, and the results were compared with those from the primary complete-case logistic regression analysis. Results Baseline characteristics of study participants A total of 4,661 participants with CKM stages 0–3 were included in this study, of whom 476 (10.2%) developed incident stroke during follow-up. In the overall population, the mean baseline CHG index was 5.33 (SD = 0.44), with a range of 2.67 to 8.94, and the mean cumulative CHG index was 15.80 (SD = 1.09), with a range of 11.86 to 23.25. Based on the overall distribution, the tertile cutoffs for the baseline CHG index were < 5.119, 5.119–5.459, and > 5.459; for the cumulative CHG index, the corresponding cutoffs were < 15.285, 15.285–16.125, and > 16.125. As shown in Table 1, participants in higher CHG tertiles tended to have higher BMI, weight, CRP, FBG, TC, and TG, and lower HDL levels, along with a higher prevalence of hypertension, diabetes, and advanced CKM stage (stage 3). Table 1Baseline characteristics of participantscharacteristicTotal(n = 4661)T1(n = 1555)T2(n = 1552)T3(n = 1554)P-value Age 58.18 ± 8.4057.75 ± 8.5458.43 ± 8.6158.37 ± 8.040.040 Gender 0.010 Female2581(55.37)813(52.28)877(56.51)891(57.34) Male2080(44.63)742(47.72)675(43.49)663(42.66) Hukou < 0.001 Agriculture4031(86.48)1388(89.26)1345(86.66)1298(83.53) Non-agriculture630(13.52)167(10.74)207(13.34)256(16.47) Education 0.400 Below high school4303(92.32)1424(91.58)1440(92.78)1439(92.60) High school or above358(7.68)131(8.42)112(7.22)115(7.40) Marital status 0.890 Married4209(90.30)1400(90.03)1402(90.34)1407(90.54) Others452(9.70)155(9.97)150(9.66)147(9.46) BMI 23.65 ± 3.7922.13 ± 3.2623.73 ± 3.6825.09 ± 3.80< 0.001 Height 157.88 ± 8.35157.72 ± 8.46157.64 ± 8.25158.29 ± 8.320.060 Weight 59.09 ± 11.1355.17 ± 9.8759.05 ± 10.4363.04 ± 11.59< 0.001 Smoking status < 0.001 Current1376(29.52)519(33.38)438(28.22)419(26.96) Former352(7.55)96(6.17)119(7.67)137(8.82) Never2933(62.93)940(60.45)995(64.11)998(64.22) Drinking status < 0.001 Current1184(25.40)469(30.16)372(23.97)343(22.07) Former374(8.02)125(8.04)128(8.25)121(7.79) Never3103(66.57)961(61.80)1052(67.78)1090(70.14) TG 134.71 ± 114.6085.85 ± 39.15116.04 ± 56.03202.25 ± 165.64< 0.001 TC 194.37 ± 38.42178.64 ± 32.10192.53 ± 33.16211.96 ± 41.70< 0.001 FBG 109.84 ± 34.6796.97 ± 12.74104.07 ± 17.18128.47 ± 51.03< 0.001 HDL 51.31 ± 15.2462.33 ± 14.9850.74 ± 11.2240.86 ± 10.73< 0.001 CRP 2.38 ± 6.641.87 ± 4.762.22 ± 6.633.05 ± 8.06< 0.001 Antihypertensive drugs < 0.001 No3906(83.80)1404(90.29)1337(86.15)1165(74.97) Yes755(16.20)151(9.71)215(13.85)389(25.03) Antihyperlipidemic drugs < 0.001 No4455(95.58)1526(98.14)1487(95.81)1442(92.79) Yes206(4.42)29(1.86)65(4.19)112(7.21) Antidiabetic drugs < 0.001 No4527(97.13)1541(99.10)1541(99.29)1445(92.99) Yes134(2.87)14(0.90)11(0.71)109(7.01) Hypertension < 0.001 No2925(62.75)1102(70.87)1009(65.01)814(52.38) Yes1736(37.25)453(29.13)543(34.99)740(47.62) Diabetes < 0.001 No4037(86.61)1502(96.59)1443(92.98)1092(70.27) Yes624(13.39)53(3.41)109(7.02)462(29.73) Cancer 0.360 No4633(99.40)1549(99.61)1542(99.36)1542(99.23) Yes28(0.60)6(0.39)10(0.64)12(0.77) Lung diseases 0.070 No4245(91.07)1401(90.10)1408(90.72)1436(92.41) Yes416(8.93)154(9.90)144(9.28)118(7.59) Liver disease 0.390 No4538(97.36)1511(97.17)1507(97.10)1520(97.81) Yes123(2.64)44(2.83)45(2.90)34(2.19) Kidney disease 0.690 No4449(95.45)1479(95.11)1482(95.49)1488(95.75) Yes212(4.55)76(4.89)70(4.51)66(4.25) CHG 2012 5.33 ± 0.444.94 ± 0.215.28 ± 0.175.78 ± 0.41< 0.001 CHG 2015 5.20 ± 0.364.88 ± 0.195.18 ± 0.165.55 ± 0.33< 0.001 cumulative CHG 15.80 ± 1.0914.72 ± 0.4215.69 ± 0.2416.98 ± 0.87< 0.001 CKM < 0.001 0332(7.12)259(16.66)61(3.93)12(0.77) 1882(18.92)428(27.52)362(23.32)92(5.92) 21849(39.67)581(37.36)688(44.33)580(37.32) 31598(34.28)287(18.46)441(28.41)870(55.98)BMI: body mass index; TG: triglyceride; TC: total cholesterol; FBG: fasting blood glucose; HDL: high-density lipoprotein cholesterol; CRP: C-reactive protein; CHG: Cholesterol, High-Density Lipoprotein, and Glucose; CHG2012: CHG index in 2012; CHG2015: CHG index in 2015; CKM: cardiovascular-kidney-metabolic syndromeNote: Because all continuous variables were non‑normally distributed (Shapiro‑Wilk test, all p < 0.001), their median (IQR) values are provided in Supplementary Table 8 Table 1Baseline characteristics of participantscharacteristicTotal(n = 4661)T1(n = 1555)T2(n = 1552)T3(n = 1554)P-value Age 58.18 ± 8.4057.75 ± 8.5458.43 ± 8.6158.37 ± 8.040.040 Gender 0.010 Female2581(55.37)813(52.28)877(56.51)891(57.34) Male2080(44.63)742(47.72)675(43.49)663(42.66) Hukou < 0.001 Agriculture4031(86.48)1388(89.26)1345(86.66)1298(83.53) Non-agriculture630(13.52)167(10.74)207(13.34)256(16.47) Education 0.400 Below high school4303(92.32)1424(91.58)1440(92.78)1439(92.60) High school or above358(7.68)131(8.42)112(7.22)115(7.40) Marital status 0.890 Married4209(90.30)1400(90.03)1402(90.34)1407(90.54) Others452(9.70)155(9.97)150(9.66)147(9.46) BMI 23.65 ± 3.7922.13 ± 3.2623.73 ± 3.6825.09 ± 3.80< 0.001 Height 157.88 ± 8.35157.72 ± 8.46157.64 ± 8.25158.29 ± 8.320.060 Weight 59.09 ± 11.1355.17 ± 9.8759.05 ± 10.4363.04 ± 11.59< 0.001 Smoking status < 0.001 Current1376(29.52)519(33.38)438(28.22)419(26.96) Former352(7.55)96(6.17)119(7.67)137(8.82) Never2933(62.93)940(60.45)995(64.11)998(64.22) Drinking status < 0.001 Current1184(25.40)469(30.16)372(23.97)343(22.07) Former374(8.02)125(8.04)128(8.25)121(7.79) Never3103(66.57)961(61.80)1052(67.78)1090(70.14) TG 134.71 ± 114.6085.85 ± 39.15116.04 ± 56.03202.25 ± 165.64< 0.001 TC 194.37 ± 38.42178.64 ± 32.10192.53 ± 33.16211.96 ± 41.70< 0.001 FBG 109.84 ± 34.6796.97 ± 12.74104.07 ± 17.18128.47 ± 51.03< 0.001 HDL 51.31 ± 15.2462.33 ± 14.9850.74 ± 11.2240.86 ± 10.73< 0.001 CRP 2.38 ± 6.641.87 ± 4.762.22 ± 6.633.05 ± 8.06< 0.001 Antihypertensive drugs < 0.001 No3906(83.80)1404(90.29)1337(86.15)1165(74.97) Yes755(16.20)151(9.71)215(13.85)389(25.03) Antihyperlipidemic drugs < 0.001 No4455(95.58)1526(98.14)1487(95.81)1442(92.79) Yes206(4.42)29(1.86)65(4.19)112(7.21) Antidiabetic drugs < 0.001 No4527(97.13)1541(99.10)1541(99.29)1445(92.99) Yes134(2.87)14(0.90)11(0.71)109(7.01) Hypertension < 0.001 No2925(62.75)1102(70.87)1009(65.01)814(52.38) Yes1736(37.25)453(29.13)543(34.99)740(47.62) Diabetes < 0.001 No4037(86.61)1502(96.59)1443(92.98)1092(70.27) Yes624(13.39)53(3.41)109(7.02)462(29.73) Cancer 0.360 No4633(99.40)1549(99.61)1542(99.36)1542(99.23) Yes28(0.60)6(0.39)10(0.64)12(0.77) Lung diseases 0.070 No4245(91.07)1401(90.10)1408(90.72)1436(92.41) Yes416(8.93)154(9.90)144(9.28)118(7.59) Liver disease 0.390 No4538(97.36)1511(97.17)1507(97.10)1520(97.81) Yes123(2.64)44(2.83)45(2.90)34(2.19) Kidney disease 0.690 No4449(95.45)1479(95.11)1482(95.49)1488(95.75) Yes212(4.55)76(4.89)70(4.51)66(4.25) CHG 2012 5.33 ± 0.444.94 ± 0.215.28 ± 0.175.78 ± 0.41< 0.001 CHG 2015 5.20 ± 0.364.88 ± 0.195.18 ± 0.165.55 ± 0.33< 0.001 cumulative CHG 15.80 ± 1.0914.72 ± 0.4215.69 ± 0.2416.98 ± 0.87< 0.001 CKM < 0.001 0332(7.12)259(16.66)61(3.93)12(0.77) 1882(18.92)428(27.52)362(23.32)92(5.92) 21849(39.67)581(37.36)688(44.33)580(37.32) 31598(34.28)287(18.46)441(28.41)870(55.98)BMI: body mass index; TG: triglyceride; TC: total cholesterol; FBG: fasting blood glucose; HDL: high-density lipoprotein cholesterol; CRP: C-reactive protein; CHG: Cholesterol, High-Density Lipoprotein, and Glucose; CHG2012: CHG index in 2012; CHG2015: CHG index in 2015; CKM: cardiovascular-kidney-metabolic syndromeNote: Because all continuous variables were non‑normally distributed (Shapiro‑Wilk test, all p < 0.001), their median (IQR) values are provided in Supplementary Table 8 Baseline characteristics of participants BMI: body mass index; TG: triglyceride; TC: total cholesterol; FBG: fasting blood glucose; HDL: high-density lipoprotein cholesterol; CRP: C-reactive protein; CHG: Cholesterol, High-Density Lipoprotein, and Glucose; CHG2012: CHG index in 2012; CHG2015: CHG index in 2015; CKM: cardiovascular-kidney-metabolic syndrome Note: Because all continuous variables were non‑normally distributed (Shapiro‑Wilk test, all p < 0.001), their median (IQR) values are provided in Supplementary Table 8 The relationship between CHG index, cumulative CHG index, and stroke in individuals with CKM stages 0–3 Multivariable logistic regression analyses revealed that, after full adjustment for confounders, each 1-unit increase in the baseline CHG index was associated with a 42% higher risk of stroke (OR = 1.42, 95% CI: 1.12–1.81), while each 1-unit increase in the cumulative CHG index was associated with a 14% higher risk (OR = 1.14, 95% CI: 1.03–1.26). For both measures, stroke risk increased progressively across tertiles. Compared with the lowest tertile (T1), participants in the highest tertile (T3) of the baseline CHG index had a 35% higher risk (OR = 1.35, 95% CI: 1.03–1.76), and those in the highest tertile of the cumulative CHG index had a 34% higher risk (OR = 1.34, 95% CI: 1.03–1.76). The trend tests were statistically significant for both measures (p for trend = 0.030) (Table 2). Table 2Logistic regression results of CHG index (baseline and cumulative levels) with stroke in the population with CKM stages 0–3characterCrude modelModel 1Model 2OR (95% CI)P-valueOR (95% CI)P-valueOR (95% CI)P-value Baseline CHG index 1.75 (1.45–2.12)< 0.0011.77 (1.46–2.15)< 0.0011.42 (1.12–1.81)0.004 Tertiles of the baseline CHG index T1(< 5.119)refrefref T2(5.119–5.459)1.36 (1.06–1.75)0.0101.36 (1.06–1.75)0.0201.25 (0.97–1.61)0.090 T3(> 5.459)1.74 (1.37–2.21)< 0.0011.73 (1.36–2.20)< 0.0011.35 (1.03–1.76)0.030 p for trend< 0.001< 0.0010.030 Cumulative CHG index 1.26 (1.17–1.37)< 0.0011.27 (1.17–1.38)< 0.0011.14 (1.03–1.26)0.010 Tertiles of the cumulative CHG index T1(< 15.285)refrefref T2(15.285–16.125)1.33 (1.03–1.70)0.0301.30 (1.01–1.68)0.0401.20 (0.93–1.56)0.160 T3(> 16.125)1.78 (1.40–2.25)< 0.0011.78 (1.40–2.26)< 0.0011.34 (1.03–1.76)0.030 p for trend< 0.001< 0.0010.030Crude model was unadjusted;Model 1 was adjusted for age, gender, marital status, hukou and education;Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Table 2Logistic regression results of CHG index (baseline and cumulative levels) with stroke in the population with CKM stages 0–3characterCrude modelModel 1Model 2OR (95% CI)P-valueOR (95% CI)P-valueOR (95% CI)P-value Baseline CHG index 1.75 (1.45–2.12)< 0.0011.77 (1.46–2.15)< 0.0011.42 (1.12–1.81)0.004 Tertiles of the baseline CHG index T1(< 5.119)refrefref T2(5.119–5.459)1.36 (1.06–1.75)0.0101.36 (1.06–1.75)0.0201.25 (0.97–1.61)0.090 T3(> 5.459)1.74 (1.37–2.21)< 0.0011.73 (1.36–2.20)< 0.0011.35 (1.03–1.76)0.030 p for trend< 0.001< 0.0010.030 Cumulative CHG index 1.26 (1.17–1.37)< 0.0011.27 (1.17–1.38)< 0.0011.14 (1.03–1.26)0.010 Tertiles of the cumulative CHG index T1(< 15.285)refrefref T2(15.285–16.125)1.33 (1.03–1.70)0.0301.30 (1.01–1.68)0.0401.20 (0.93–1.56)0.160 T3(> 16.125)1.78 (1.40–2.25)< 0.0011.78 (1.40–2.26)< 0.0011.34 (1.03–1.76)0.030 p for trend< 0.001< 0.0010.030Crude model was unadjusted;Model 1 was adjusted for age, gender, marital status, hukou and education;Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Logistic regression results of CHG index (baseline and cumulative levels) with stroke in the population with CKM stages 0–3 Crude model was unadjusted; Model 1 was adjusted for age, gender, marital status, hukou and education; Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Restricted cubic spline analyses Among individuals with CKM stages 0–3, both the baseline CHG index and the cumulative CHG index showed linear positive associations with stroke occurrence. For the baseline CHG index, the p-value for non-linearity was 0.331 and the overall p-value was 0.012; for the cumulative CHG index, the corresponding values were 0.817 and 0.037, respectively (Fig. 2). Fig. 2Restricted cubic spline curves illustrating the associations of baseline (A) and cumulative (B) CHG index with the risk of new-onset stroke. Crude model was unadjusted, Model 1 was adjusted for age, gender, marital status, hukou and education, Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Restricted cubic spline curves illustrating the associations of baseline (A) and cumulative (B) CHG index with the risk of new-onset stroke. Crude model was unadjusted, Model 1 was adjusted for age, gender, marital status, hukou and education, Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Comparison of the predictive capabilities of the CHG index and cumulative CHG index The AUC for the baseline CHG index was 0.672 (95% CI: 0.65–0.70), and that for the cumulative CHG index was 0.671 (95% CI: 0.65–0.70). DeLong’s test showed no statistically significant difference in predictive performance between the two indices (p = 0.330) (Fig. 3). Fig. 3ROC curves showing the predictive capabilities of the baseline CHG index and cumulative CHG index. Crude model was unadjusted, Model 1 was adjusted for age, gender, marital status, hukou and education, Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease ROC curves showing the predictive capabilities of the baseline CHG index and cumulative CHG index. Crude model was unadjusted, Model 1 was adjusted for age, gender, marital status, hukou and education, Model 2 was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Subgroup and sensitivity analyses Subgroup analyses showed that the associations of both baseline and cumulative CHG index with stroke risk were generally consistent across strata defined by age (< 60 vs. ≥60 years), hukou, marital status, education, smoking status, hypertension, diabetes, and CKM stage (0–1, 2, and 3) (Fig. 4). However, significant interactions were observed for gender and drinking status (p for interaction = 0.012 for gender and 0.020 for drinking status for the baseline CHG index; 0.029 and 0.047, respectively, for the cumulative CHG index). Detailed subgroup estimates, including ORs and 95% CIs, are provided in Supplementary Table 9. Fig. 4Subgroup analyses of the associations of baseline (A) and cumulative (B) CHG index with new-onset stroke. The model was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Subgroup analyses of the associations of baseline (A) and cumulative (B) CHG index with new-onset stroke. The model was adjusted for age, gender, marital status, hukou, education, smoking status, drinking status, BMI, hypertension, diabetes, CRP, antihypertensive drugs, antihyperlipidemic drugs, antidiabetic drugs, cancer, lung diseases, liver disease and kidney disease Sensitivity analyses confirmed the robustness of our findings. Multiple imputation for missing data yielded effect estimates nearly identical to the complete-case analysis. Modelling CHG indices as continuous variables or tertiles, using Cox proportional hazards models (with the proportional hazards assumption satisfied; global p = 0.362 for baseline CHG and 0.268 for cumulative CHG), and applying competing risk regression (Fine-Gray model) all supported the same conclusions. Detailed results are presented in Supplementary Tables 5, 6, and 7. Discussion This is the first study to systematically evaluate the associations of both baseline and cumulative CHG index with incident stroke risk, and to compare their predictive performance, specifically in a middle-aged and older Chinese population with CKM stages 0–3. The key finding was that although both baseline and cumulative CHG index were independently and linearly associated with higher stroke risk, the cumulative index did not outperform a single baseline measurement (AUC: 0.671 vs. 0.672, p = 0.330). This suggests that, in this population, a single CHG measurement may serve as a simple, low-cost indicator for initial risk assessment, helping to identify individuals who may benefit from further clinical evaluation in large-scale population screening or primary healthcare settings. Previous studies have confirmed that an elevated CHG index is closely associated with the occurrence of stroke and cardiovascular disease. For instance, in a prospective cohort study, Xiao et al. demonstrated that in populations with normal glucose regulation (NGR) and pre-diabetes (Pre-DM), an elevated CHG index increases the risk of stroke [15]. Mo et al. similarly reported a positive association between the CHG index and cardiovascular disease [16]. Zeng et al. further combined the CHG index with obesity-related indicators, finding that the CHG-CVAI composite index exhibited the strongest association with stroke in the general population, with predictive performance superior to other composite indicators [25]. Collectively, this evidence supports the favourable predictive capability of the CHG index for stroke and cardiovascular disease. The present study focused on the CKM stage 0–3 population, who are in a critical transitional phase from the accumulation of metabolic risk factors to the occurrence of clinical cardiovascular events [1, 2]. Our results indicate that even after excluding individuals with CKM stage 4 who have already experienced clinical cardiovascular events, an elevated CHG index remains significantly associated with stroke risk. This association underscores that metabolic dysregulation itself is a core driver of early vascular injury and stroke risk [9, 10], aligning with the AHA’s concept of early identification and management of CKM [1, 26]. This study further examined the comparative value of “cumulative exposure” versus “current state” in risk prediction. Partially consistent with Tian et al. [27], who found that the cumulative CHG index predicted cardiovascular outcomes in the general population, our study also confirmed a positive association between cumulative CHG index and stroke risk. However, ROC curve comparison showed that the CHG index had only modest predictive performance (AUC ≈ 0.67), and the cumulative CHG index did not provide statistically significant incremental predictive information beyond the baseline CHG index. This finding underscores that while the CHG index is associated with risk, its standalone discriminatory power is limited. Several factors may account for this finding. First, the CKM syndrome represents a continuum from metabolic abnormalities to target organ damage. In early stages (0–3), an individual’s metabolic dysregulation may be relatively stable or already well captured by a single measurement; thus, incorporating cumulative exposure offers limited additional benefit. Second, the cumulative CHG index was calculated using only two time points (2012 and 2015), which cannot capture long-term dynamic changes in metabolic risk, and the use of the mean value ignores temporal variability. Therefore, the CHG index should be interpreted as one of several risk indicators, not as a definitive diagnostic or risk stratification tool. Future studies should validate these findings using longer follow-up periods and more frequent measurements (e.g., every six months or annually). Moreover, stroke risk is influenced by multiple factors beyond metabolic indicators, including functional status. A recent CHARLS-based study by Yang et al. demonstrated a significant independent association between functional dependence and cardiovascular disease risk [28]. Future research could combine the CHG index with other risk factors, such as functional status, or use it as a preliminary screening tool, followed by comprehensive risk assessment using established instruments like the Framingham Risk Score or the China-PAR model in those identified as high-risk. From a pathophysiological perspective, the CHG index, as a composite marker integrating total cholesterol, high-density lipoprotein, and fasting blood glucose [14], exhibits an essential association with stroke risk that can be comprehensively interpreted within the continuous pathological framework of CKM. This process begins with overweight/obesity or adipose tissue dysfunction. Dysfunctional visceral adipose tissue excessively secretes pro-inflammatory cytokines (e.g., TNF-α and IL-6) and free fatty acids (FFA), thereby impairing insulin receptor function and inducing IR [29]. Persistent IR leads to decreased insulin sensitivity in peripheral tissues [30], subsequently causing hyperinsulinaemia and chronic hyperglycaemia [31]. Hyperinsulinaemia and hyperglycaemia directly damage the vascular endothelium and promote the formation and progression of atherosclerotic plaques by activating inflammatory signalling pathways (e.g., NF-κB) in endothelial cells and macrophages and by exacerbating oxidative stress [32, 33]. Concurrently, this metabolic dysregulation is frequently accompanied by characteristic pro-atherogenic dyslipidaemia, featuring elevated triglycerides, reduced high-density lipoprotein levels, and impaired cholesterol reverse transport function, along with the production of more pro-atherogenic oxidised low-density lipoprotein [34, 35]. During the early stages of CKM (e.g., stages 0–3), a mutually reinforcing vicious cycle forms between visceral fat accumulation and IR, continuously amplifying the effect of vascular damage [36, 37]. The combined effects of glucose metabolism disorders, dyslipidaemia, and chronic inflammation progressively erode the vascular endothelium, promoting atherosclerosis; simultaneously, by activating pathways such as the renin-angiotensin-aldosterone system, they influence haemodynamics, ultimately culminating in clinical stroke events [38]. In subgroup analyses, significant interaction effects for this risk association were observed in women and never-drinkers. On the one hand, the stroke risk was higher in women than in men, consistent with the findings of Zeng et al. [25]. This may be attributable to the decline in oestrogen’s protective effects after menopause, which exacerbates insulin resistance and amplifies the negative impact of metabolic abnormalities on the vascular system [39, 40]. On the other hand, the higher risk observed in never-drinkers presents a more complex picture: it may indicate that the CHG index can identify individuals with “concealed high risk”—those who lack the traditional behavioural risk factor of alcohol consumption but have inherently high metabolic risk due to genetic factors, dietary patterns, or insufficient physical activity [41]. These differences alert clinicians to pay close attention to women and never-drinkers with elevated CHG index levels. Nevertheless, these subgroup findings should be interpreted with caution: subgroup analyses are often underpowered due to smaller sample sizes, and the observed interactions (p for interaction = 0.012 for gender and 0.047 for drinking status) may be susceptible to type I error and spurious associations. Therefore, these results are exploratory in nature and require independent validation in future prospective studies before firm conclusions can be drawn. The findings of this study have certain clinical implications. As a composite index derived from routine, low-cost laboratory tests, the CHG index may play an adjunctive role in primary healthcare and large-scale population screening. For individuals with CKM stages 0–3, a single baseline measurement provides predictive information similar to cumulative exposure. However, given its moderate discriminatory ability (AUC ≈ 0.67), the CHG index should not be used as a standalone diagnostic or risk stratification tool. Instead, it should be combined with other established risk scoring systems, such as the Framingham Risk Score or the China-PAR model. For individuals in the highest tertile of the CHG index, particularly women and never-drinkers, clinicians may consider the CHG index as an auxiliary reference for stroke risk, recommend intensive lifestyle interventions [26], and assess the need to initiate or intensify pharmacological treatment for hypertension, dyslipidaemia, and diabetes according to current guidelines or the above risk scoring systems, to prevent progression to CKM stage 4 [1, 2]. Nevertheless, these applications require further validation in prospective intervention studies. This study involved multiple related analyses, including comparisons between baseline and cumulative CHG indices, as well as analyses using continuous variables and tertiles. Consequently, borderline significant findings with p-values near 0.05 (e.g., approximately 0.03) should be interpreted with caution, as they may reflect exploratory observations. We did not perform formal multiplicity adjustment (e.g., Bonferroni correction), because such adjustment may increase the risk of type II error in an exploratory study. These findings warrant validation in independent cohorts. This study has several strengths. First, it is a prospective cohort study based on a nationally representative sample and provides the first evidence of an association between the CHG index (both baseline and cumulative levels) and stroke risk in a middle-aged and older Chinese population with CKM stages 0–3. Second, it systematically compares the predictive performance of baseline and cumulative CHG indices, showing that a single measurement offers similar predictive information to long-term cumulative monitoring. Third, the robustness of the findings is reinforced by multiple sensitivity analyses, including multiple imputation, Cox models with proportional hazards testing, and competing risk regression. Fourth, as a composite indicator derived from routine lipid and glucose tests, the CHG index is easy to obtain and low-cost, which may facilitate its use in primary healthcare and population screening, although its moderate discriminatory ability (AUC ≈ 0.67) should be taken into account. However, our study has several limitations. First, stroke events were based on self-reported physician diagnoses, which may introduce recall and misclassification bias. The CHARLS database lacks objective medical records or imaging data for validation. If the misclassification is non-differential (unrelated to the CHG index), any bias would be towards the null, meaning our observed effect is likely an underestimate. Future studies should validate strokes using objective medical records or neuroimaging. Second, the cumulative CHG index was calculated from only two time points (2012 and 2015), which cannot capture the temporal variability of CHG levels. Future studies should use at least three repeated measurements to identify latent trajectory classes and examine their associations with stroke risk. Third, despite adjusting for many confounders, we cannot rule out residual confounding from unmeasured factors such as diet quality, physical activity intensity, socioeconomic variables, and medication adherence. The lack of detailed medication information (e.g., dosage, duration) may affect effect estimation. Future studies should collect more detailed medication data. Fourth, our data come from the Chinese elderly population, so conclusions should be extrapolated cautiously to other races or younger populations. Fifth, we did not directly compare the CHG index with established risk tools like the Framingham Risk Score or the China-PAR model, so we cannot conclude whether it offers incremental predictive value. Future head-to-head comparisons are needed. Finally, due to data limitations, we could not distinguish between ischaemic and haemorrhagic stroke, and future studies should explore associations by stroke subtype. In addition, calibration metrics (e.g., calibration plot or Brier score) were not provided, as the primary aim was to evaluate etiologic associations rather than to develop a clinical prediction model. Conclusions In this prospective analysis of a nationally representative CHARLS cohort, both baseline and cumulative CHG indices were independently and positively associated with incident stroke risk among middle-aged and older adults with cardiovascular-kidney-metabolic syndrome stages 0–3, exhibiting a linear dose-response trend. However, ROC analysis yielded an area under the curve of approximately 0.67, indicating only moderate discriminatory capacity. Therefore, although the CHG index—a simple measure derived from routine blood tests—may serve as a preliminary risk alert in large-scale population screening or primary healthcare settings, its limited predictive accuracy precludes its use as a standalone stroke risk stratification tool in clinical practice. Instead, it should be used in conjunction with other established clinical risk factors and validated risk scoring systems, such as the Framingham Risk Score or the China-PAR model. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4 Supplementary Material 5 Supplementary Material 5 Supplementary Material 6 Supplementary Material 6 Supplementary Material 7 Supplementary Material 7 Supplementary Material 8 Supplementary Material 8 Supplementary Material 9 Supplementary Material 9 Supplementary Material 10 Supplementary Material 10 Supplementary Material 11 Supplementary Material 11