The small dense LDL cholesterol-glucose index combined with visceral adiposity (sdLG-cVAI) as a novel predictor of incident cardiovascular disease: a prospective cohort study Authors: Xumao Deng, Tao Wu, Qiyuan Liang, Xiunan Liu, Panpan Qin, Simeng Li, Zhuoxing Li, Hongwei Leng, Yun Sun, Kun Zhang, Xiang Xiao Published: 2026 ABSTRACT AbstractBackgroundDyslipidemia, impaired glucose metabolism, and obesity are established cardiovascular disease (CVD) risk factors. Composite markers like the triglyceride-glucose (TyG) index reflect lipid quantity but overlook qualitative features such as small dense low-density lipoprotein cholesterol (sdLDL-C), a highly atherogenic subfraction linked to dysglycemia. This study aimed to develop a novel sdLDL-C-glucose (sdLG) index, combine it with obesity indicators to enhance CVD risk stratification, and evaluate its predictive value for incident CVD.MethodsAdults aged ≥ 45 years without baseline CVD from the China Health and Retirement Longitudinal Study (CHARLS) were included (2011–2020). CVD (heart disease and stroke) events were identified during follow-up. The sdLG index, calculated from sdLDL-C and fasting glucose, was integrated with obesity indicators to form derived indices. The predictive performance of the indices was compared using time-dependent area under the receiver operating characteristic curve (AUC), Concordance index (C-index), and net reclassification improvement/integrated discrimination improvement (NRI/IDI) to select the optimal index. Dose–response relationships were examined via restricted cubic splines (RCS). Associations of the optimal index (baseline, cumulative exposure, and trajectory patterns) with incident CVD were evaluated using Cox proportional hazards models, supplemented by subgroup and sensitivity analyses.ResultsAmong 3,969 participants, 1,029 incident CVD events occurred. The sdLG index outperformed TyG index in long-term discrimination. Among derived indices, sdLG index combined with the Chinese visceral adiposity index (sdLG-cVAI) showed the best discrimination and reclassification performance. Dose–response relationships for baseline (P for overall < 0.001; P for nonlinearity = 0.105) and cumulative sdLG-cVAI (cusdLG-cVAI) (P for overall < 0.001; P for nonlinearity = 0.522) were linear. In fully adjusted models, the highest vs. lowest quartile was associated with increased CVD risk for both baseline sdLG-cVAI (HR 1.77, 95% CI 1.44–2.16, P < 0.001) and cumulative sdLG-cVAI (HR 1.87, 95% CI 1.53–2.29, P < 0.001). Participants in the persistently high trajectory cluster faced higher risk than those in the stable low cluster (HR 1.75, 95% CI 1.46–2.10, P < 0.001). Subgroup analyses revealed some heterogeneity. Sensitivity analyses yielded results consistent with the main regression analyses.ConclusionThe sdLG index outperforms the traditional TyG index for long-term CVD risk prediction. The novel composite sdLG-cVAI independently predicts incident CVD and improves risk stratification in middle-aged and older adults.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12944-026-02987-2. Background Cardiovascular disease (CVD) remains a major global public health challenge. Its pathogenesis is closely linked to the complex interplay between obesity, dysregulated glucose metabolism, and abnormal lipid metabolism [1]. In current clinical practice, cardiovascular risk assessment continues to rely predominantly on single biomarkers, such as low-density lipoprotein cholesterol (LDL-C) and fasting blood glucose (FBG). However, these parameters capture only isolated aspects of metabolic dysfunction and are insufficient for comprehensively characterizing an individual’s integrated cardiometabolic risk profile [2, 3]. Recent years have witnessed growing interest in composite indices that incorporate multiple pathophysiological dimensions. Among them, the triglyceride-glucose (TyG) index, a simple surrogate marker of insulin resistance, has demonstrated superior predictive performance for cardiovascular risk compared to traditional single biomarkers [4, 5]. A key limitation of the TyG index, however, is that its lipid component is based solely on the quantitative measure of triglyceride levels, which are susceptible to short-term metabolic fluctuations. More critically, this index does not account for qualitative, highly atherogenic lipoprotein characteristics, particularly small dense low-density lipoprotein cholesterol (sdLDL-C) [6, 7]. Small dense LDL particles exhibit a greater ability to penetrate the arterial intima, possess reduced antioxidant capacity, and have a prolonged circulation time, all of which contribute to their heightened atherogenic potential. Their levels are significantly elevated in insulin-resistant states and share common pathophysiological pathways with dysglycemia [8, 9]. Concurrently, obesity, especially visceral fat accumulation, is a central driver of insulin resistance and disturbances in glucose and lipid homeostasis [10, 11]. While several obesity-related indices, including body mass index (BMI), body roundness index (BRI), a body shape index (ABSI), and the Chinese visceral adiposity index (cVAI), have been associated with CVD risk [12–15], the optimal method for integrating these measures with biomarkers of lipoprotein quality and glycemic status to improve risk prediction remains unclear. To date, there is no comprehensive tool that successfully integrates atherogenic lipoprotein subtypes, glucose metabolism, and obesity-related phenotypes to enable the earlier and more accurate identification of high-risk individuals. Therefore, this study aimed to develop and evaluate a novel small dense low-density lipoprotein cholesterol-glucose (sdLG) index, integrated with obesity indicators, to enhance cardiometabolic risk stratification and assess its association with and predictive value for incident CVD. Methods Study design and population This study was a prospective cohort study based on the China Health and Retirement Longitudinal Study (CHARLS). CHARLS is a nationally representative, population-based longitudinal survey of Chinese adults aged 45 years and older. It adopts a multistage, stratified probability sampling strategy with probability proportional to size (PPS), covering 150 counties (districts) across 28 provinces in China, and has been shown to have good national representativeness [16]. The baseline survey was conducted in 2011, with follow-up surveys completed in 2013, 2015, 2018, and 2020. The participant selection process is shown in (Fig. 1). Individuals enrolled between 2011 and 2020 were considered for inclusion. Participants were excluded if they met any of the following criteria: (1) incomplete follow-up information or loss to follow-up (n = 1,155); (2) missing data on key variables (n = 11,757); (3) presence of extreme values in key variables (n = 121); (4) pre-existing cardiovascular disease at baseline, including heart disease or stroke (n = 621); or (5) age < 45 years (n = 82). After exclusions, a total of 3,969 participants were included in the final analysis from the initial 17,705 individuals. Flowchart of participant selection. CHARLS, China health and retirement longitudinal study; CVD, cardiovascular disease; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; FBG, fasting blood glucose; WC, waist circumference; DM, diabetes mellitus This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and was approved by the Institutional Review Board of Peking University (IRB No. IRB00001052-11015). Written informed consent was obtained from all participants prior to study participation. Calculation of indices Small dense low-density lipoprotein cholesterol (sdLDL-C) was estimated using a previously validated equation [17], as follows: sdLDL-C = LDL-C (mg/dL) − [1.43 × LDL-C (mg/dL) − (0.14 × ln(TG (mg/dL)) × LDL-C (mg/dL)) − 8.99]. The triglyceride–glucose (TyG) index was calculated according to the classical definition: TyG index = ln [TG (mg/dL) × FBG (mg/dL) / 2] [4]. Based on this, the sdLG index was constructed as follows: sdLG index = ln [sdLDL-C (mg/dL) × FBG (mg/dL) / 2]. To further evaluate the combined effects of sdLG index and obesity-related indicators, multiple sdLG-derived indices were constructed as follows [18–22]: The calculation methods for obesity-related indicators were based on previously validated formulas reported in the literature. Assessment of endpoint events The primary outcome of this study was incident CVD, including heart disease and stroke [23]. CVD events were identified based on standardized questionnaire items in CHARLS that have been widely used in previous studies. Specifically, heart disease was assessed by the question: “Has a doctor ever told you that you have had a heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems?”. Stroke was assessed by the question: “Has a doctor ever told you that you have had a stroke?”. Participants were followed up in 2013, 2015, 2018, and 2020. Once a participant reported a first CVD event at any follow-up wave, subsequent follow-up was discontinued. Participants without CVD events were followed until 2020. The time to CVD onset was defined as the interval between the baseline year (2011) and the follow-up year in which CVD was first reported. Covariates Covariates included demographic characteristics, lifestyle factors, and comorbid conditions. Demographic variables included age, sex, and ethnicity. Lifestyle factors included current smoking status and current alcohol consumption. Socioeconomic variables included marital status, place of residence, and educational level. Comorbidities were determined based on self-reported physician diagnoses and included hypertension, type 2 diabetes mellitus, and dyslipidemia. All covariates were treated as potential confounders and were adjusted for in the statistical models. Statistical analysis Continuous variables were assessed for normality using the Shapiro–Wilk test. Normally distributed continuous variables are presented as mean ± standard deviation (SD), whereas non-normally distributed variables are presented as median (interquartile range, IQR). Categorical variables are presented as counts (percentages). Between-group comparisons were performed using analysis of variance (ANOVA) for continuous variables or the chi-square test for categorical variables, as appropriate. Given the very low proportion of missing data for variables other than the primary exposures (< 1%), a complete case analysis approach was adopted [24]. Participants with missing values in any variables were excluded to ensure consistency of model estimation. Extreme values were identified using the interquartile range (IQR) method, with values below Q1 − 1.5 × IQR or above Q3 + 1.5 × IQR defined as outliers and excluded from the analysis. To systematically compare the predictive performance of different indices while minimizing bias due to multiple comparisons, a stepwise analytical strategy was applied. First, the predictive performance of the sdLG index and the TyG index for incident CVD was compared using time-dependent area under the receiver operating characteristic curve (AUC), Concordance index (C-index), and net reclassification improvement/integrated discrimination improvement (NRI/IDI). Second, the same methods were used to compare all sdLG-derived indices, and the index with the optimal predictive performance was selected. Comparisons of AUCs were conducted using the DeLong test, and assessment of C-index values was performed using bootstrap resampling. An NRI > 0 and IDI > 0 were considered indicative of improved model performance. The cumulative incidence of CVD was estimated using Kaplan–Meier survival curves, and differences between groups were assessed using the log-rank test. In fully adjusted models, restricted cubic spline (RCS) analyses were applied to explore dose–response relationships between baseline and cumulative indices and CVD risk. If significant nonlinearity was observed, participants were stratified according to the identified inflection point and analyzed separately. Cox proportional hazards regression models were used to evaluate the associations between the selected index and the risk of incident CVD. The index was analyzed in three forms: baseline value, cumulative value, and trajectory pattern. Baseline and cumulative indices were modeled both as continuous variables and as categorical variables according to quartiles. The cumulative index was calculated as [(index value in 2011 + index value in 2015) / 2] × the time interval in years. Trajectories of the index were identified using k-means clustering. Four progressively adjusted models were constructed: Model 1 was unadjusted; Model 2 was adjusted for age, sex, and ethnicity; Model 3 was further adjusted for smoking status, alcohol consumption, educational level, marital status, and place of residence; and Model 4 was additionally adjusted for hypertension, diabetes, and dyslipidemia. Variance inflation factors (VIFs) were calculated to assess multicollinearity among covariates. In addition, time-dependent AUC, C-index, and NRI/IDI analyses were performed for the baseline, cumulative, and trajectory forms of the selected index under fully adjusted models to evaluate predictive performance. Subgroup analyses were conducted according to age, sex, ethnicity, smoking status, alcohol consumption, educational level, marital status, place of residence, hypertension, diabetes, and dyslipidemia. To assess the robustness of the findings, two sensitivity analyses were performed. First, logistic regression models were used to re-evaluate the associations to address potential inaccuracies in event timing. Second, to reduce potential reverse causality, participants who developed CVD in 2013 were excluded and the analyses were repeated. All statistical analyses were conducted using R software (version 4.5.1). A two-sided P value < 0.05 was considered statistically significant. The analyses were performed using the charlsR, cluster, survival, and nricens R packages. Results Baseline characteristics of participants Among the 3,969 participants included in the final analysis (1,753 men and 2,216 women), the median age was 58 years (IQR: 52–64), and the median follow-up duration was 9 years. During follow-up, a total of 1,029 incident CVD events were recorded. Baseline characteristics of participants stratified by incident CVD status are presented in (Table 1). Table 1Baseline characteristics of participants according to the occurrence of CVDVariableTotal(n = 3969)Non-CVD(n = 2940)CVD(n = 1029)P valueAge, years58 (52, 64)57 (51, 64)59 (53, 65)< 0.001Sex, n (%)< 0.010 Female2216 (55.8)1597 (54.3)619 (60.2) Male1753 (44.2)1343 (45.7)410 (39.8)Nationality, n (%)0.210 Han3736 (94.1)2776 (94.4)960 (93.3) Others233 (5.9)164 (5.6)69 (6.7)Education, n (%)< 0.010 High school and above372 (9.4)272 (9.2)100 (9.7) Middle school819 (20.6)643 (21.9)176 (17.1) Primary school and below2778 (70.0)2025 (68.9)753 (73.2)Residence, n (%)0.340 Rural2666 (67.2)1962 (66.7)704 (68.4) Urban1303 (32.8)978 (33.3)325 (31.6)Marriage, n (%)0.750 Married3929 (99.0)2909 (98.9)1020 (99.1) Unmarried40 (1.0)31 (1.1)9 (0.9)Drink, n (%)< 0.010 Current1337 (33.7)1027 (34.9)310 (30.1) Never2632 (66.3)1913 (65.1)719 (69.9)Smoke, n (%)< 0.010 Current1159 (29.2)895 (30.4)264 (25.7) Former302 (7.6)205 (7.0)97 (9.4) Never2508 (63.2)1840 (62.6)668 (64.9)Hypertension, n (%)< 0.001 No2495 (62.9)1943 (66.1)552 (53.6) Yes1474 (37.1)997 (33.9)477 (46.4)DM, n (%)< 0.001 No3408 (85.9)2558 (87.0)850 (82.6) Yes561 (14.1)382 (13.0)179 (17.4)Hyperlipidemia, n (%)< 0.001 No3019 (76.1)2276 (77.4)743 (72.2) Yes950 (23.9)664 (22.6)286 (27.8)BMI, kg/m²23.2 (21.0, 25.7)23.0 (20.9, 25.4)23.9 (21.5, 26.6)< 0.001Weight, kg57.9 (51.3, 65.5)57.3 (50.9, 64.9)59.4 (52.7, 67.5)< 0.001TG, mg/dL102.7 (73.5, 148.7)100.9 (71.7, 146.0)107.1 (79.7, 155.8)< 0.001FBG, mg/dL102.1 (94.5, 111.4)101.7 (94.3, 110.5)103.3 (95.4, 113.9)< 0.001WC, cm84.6 (78.0, 92.0)84.0 (77.4, 91.0)87.0 (80.0, 94.2)< 0.001Height, cm157.5 (151.9, 163.7)157.6 (152.0, 163.9)157.0 (151.6, 163.5)0.260LDL-C, mg/dL114.8 (93.9, 138.4)113.7 (93.2, 136.5)118.7 (96.3, 142.7)< 0.001HDL-C, mg/dL49.5 (41.0, 60.3)50.1 (41.0, 60.7)48.3 (40.2, 58.0)< 0.010TyG8.57 (8.21, 9.00)8.55 (8.19, 8.98)8.63 (8.29, 9.09)< 0.001sdLG7.47 (7.20, 7.74)7.45 (7.18, 7.72)7.54 (7.26, 7.81)< 0.001sdLG-BMI173.51 (153.48, 196.13)171.27 (151.62, 192.71)181.09 (159.65, 204.48)< 0.001sdLG-WC631.83 (569.87, 699.32)622.54 (562.99, 691.18)657.58 (592.63, 721.01)< 0.001sdLG-WHtR4.02 (3.60, 4.44)3.96 (3.56, 4.39)4.17 (3.74, 4.59)< 0.001sdLG-BRI30.28 (23.64, 38.54)29.39 (23.04, 37.48)33.14 (25.56, 41.74)< 0.001sdLG-cVAI692.86 (494.04, 917.31)663.06 (471.96, 885.52)766.76 (570.60, 1010.91)< 0.001sdLG-ABSI0.62 (0.58, 0.66)0.61 (0.58, 0.65)0.63 (0.59, 0.66)< 0.001sdLG-WWI83.16 (77.35, 88.97)82.46 (76.79, 88.33)85.15 (78.99, 90.81)< 0.001Abbreviations: CVD Cardiovascular Disease, DM Diabetes Mellitus, BMI body mass index, TG triglycerides, FBG fasting blood glucose, WC waist circumference, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, TyG triglyceride–glucose index, sdLG Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index, sdLG-BMI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body mass index, sdLG-WC Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist circumference, sdLG-WHtR Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist-to-height ratio, sdLG-BRI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body roundness index, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, sdLG-ABSI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-a body shape index, sdLG-WWI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-weight-adjusted waist index Baseline characteristics of participants according to the occurrence of CVD Abbreviations: CVD Cardiovascular Disease, DM Diabetes Mellitus, BMI body mass index, TG triglycerides, FBG fasting blood glucose, WC waist circumference, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, TyG triglyceride–glucose index, sdLG Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index, sdLG-BMI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body mass index, sdLG-WC Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist circumference, sdLG-WHtR Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist-to-height ratio, sdLG-BRI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body roundness index, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, sdLG-ABSI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-a body shape index, sdLG-WWI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-weight-adjusted waist index Compared with participants without incident CVD, those who developed CVD exhibited significantly more unfavorable metabolic and anthropometric profiles at baseline. Specifically, participants in the CVD group had higher age, triglycerides (TG), fasting blood glucose (FBG), body mass index (BMI), low-density lipoprotein cholesterol (LDL-C), waist circumference (WC), body weight, as well as higher levels of the TyG index, sdLG index, and all sdLG-derived indices, including sdLG-BMI, sdLG-WC, sdLG-WHtR, sdLG-BRI, sdLG-cVAI, sdLG-ABSI, and sdLG-WWI (all P < 0.05). Regarding demographic and clinical characteristics, the proportion of women was higher in the CVD group, and the prevalence of hypertension, diabetes, and dyslipidemia was significantly greater (all P < 0.05). In addition, participants with incident CVD had lower educational attainment and lower proportions of current smoking and alcohol consumption (P < 0.05). No significant differences were observed between the two groups with respect to height, ethnicity, place of residence, or marital status (all P > 0.05). Discriminative and reclassification performance of different indices Comparison of predictive performance between sdLG index and the traditional insulin resistance index Time-dependent AUC analyses showed that both the TyG index and the sdLG index exhibited similar temporal trends during follow-up, with a slight decline in AUC values during the early follow-up period (2–4 years), followed by a gradual increase over time (Fig. 2A). In the early follow-up period, the discriminative performance of the two indices was comparable. However, during the later follow-up years (years 7 and 9), the AUC values of sdLG index were consistently higher than those of TyG index. At year 9, the AUC for sdLG index was 0.565, compared with 0.552 for TyG index (P = 0.044) (Fig. 2A). Comparison of the time-dependent predictive performance between the TyG index, sdLG index, and sdLG-enhanced indices. A AUC trajectories for TyG index versus sdLG index; (B) C-index trajectories for TyG index versus sdLG index; (C) AUC trajectories for sdLG index alone versus sdLG index combined with anthropometric and visceral adiposity indices; (D) Corresponding C-index trajectories for the sdLG-based combinations. AUC, area under the receiver operating characteristic curve; C-index, Concordance index; TyG, triglyceride-glucose index; sdLG, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index; sdLG-WC, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist circumference; sdLG-BMI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body mass index; sdLG-WHtR, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist-to-height ratio; sdLG-ABSI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-a body shape index; sdLG-BRI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body roundness index; sdLG-cVAI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index; sdLG-WWI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-weight-adjusted waist index The C-index analysis yielded results consistent with those of the AUC analysis, with sdLG index demonstrating persistently superior discriminative ability compared with TyG index in the later follow-up period (Fig. 2B), suggesting a relative advantage of sdLG index in long-term risk prediction. Reclassification analyses performed at the median follow-up time (year 9) showed that, compared with TyG index, sdLG index yielded an integrated discrimination improvement (IDI) of 0.004 (P = 0.099) and a continuous net reclassification improvement (NRI) of 0.055 (P = 0.139). Both point estimates were positive, indicating a potential improvement in risk discrimination and reclassification with sdLG index (Table 2). Table 2Comparison of the incremental predictive value of sdLG index and sdLG-derived indices using IDI and NRIComparisonIDI (95% CI)P valueNRI (95% CI)P valueTyG vs. sdLG0.004 (-0.001, 0.009)0.0990.055 (-0.014, 0.085)0.139sdLG vs. sdLG-cVAI0.012 (0.002, 0.022)0.0200.087 (0.026, 0.131)< 0.001sdLG-ABSI vs. sdLG-cVAI0.015 (0.006, 0.026)< 0.0010.109 (0.059, 0.155)< 0.001sdLG-BRI vs. sdLG-cVAI0.004 (-0.002, 0.010)0.1780.038 (-0.018, 0.099)0.139sdLG-WWI vs. sdLG-cVAI0.011 (0.002, 0.021)0.0200.083 (0.040, 0.138)< 0.001sdLG-WHtR vs. sdLG-cVAI0.005 (0.001, 0.009)0.0200.055 (0.000, 0.090)0.059sdLG-WC vs. sdLG-cVAI0.005 (0.001, 0.009)< 0.0010.098 (0.048, 0.134)< 0.001sdLG-BMI vs. sdLG-cVAI0.007 (0.002, 0.013)0.0200.075 (0.021, 0.130)< 0.001Abbreviations: IDI Integrated Discrimination Improvement, NRI Net Reclassification Improvement, CI confidence interval, TyG triglyceride–glucose index, sdLG Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index, sdLG-BMI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body mass index, sdLG-WC Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist circumference, sdLG-WHtR Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist-to-height ratio, sdLG-BRI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body roundness index, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, sdLG-ABSI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-a body shape index, sdLG-WWI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-weight-adjusted waist index Comparison of the incremental predictive value of sdLG index and sdLG-derived indices using IDI and NRI Abbreviations: IDI Integrated Discrimination Improvement, NRI Net Reclassification Improvement, CI confidence interval, TyG triglyceride–glucose index, sdLG Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index, sdLG-BMI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body mass index, sdLG-WC Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist circumference, sdLG-WHtR Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist-to-height ratio, sdLG-BRI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body roundness index, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, sdLG-ABSI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-a body shape index, sdLG-WWI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-weight-adjusted waist index Comparison of predictive performance between sdLG index and sdLG index combined with obesity-related derived indices Further comparisons of predictive performance among sdLG index and sdLG index combined with obesity-related derived indices showed that all sdLG-derived indices exhibited overall better predictive performance than sdLG index alone. Although some variability was observed during the early follow-up period (2–4 years), the AUCs for all indices generally increased with longer follow-up duration (Fig. 2C). Among these indices, sdLG-cVAI consistently demonstrated the highest AUC across most follow-up periods and showed significantly better performance than sdLG index at years 4, 7, and 9 (adjusted P < 0.05). Among the remaining derived indices, only sdLG-WC showed significantly better predictive performance than sdLG index at year 7 (adjusted P < 0.05). The C-index trends further supported these findings, with sdLG-cVAI exhibiting the highest C-index across most follow-up periods (Fig. 2D). NRI/IDI analyses indicated that, compared with baseline models including sdLG-ABSI, sdLG-BRI, sdLG-WWI, sdLG-WHtR, sdLG-WC, sdLG-BMI, and sdLG, sdLG-cVAI generally showed improvements in discrimination and risk reclassification. Specifically, sdLG-cVAI demonstrated consistent improvement trends compared with sdLG-ABSI, sdLG-WWI, sdLG-WC, sdLG-BMI, and sdLG index (IDI = 0.005–0.015; NRI = 0.075–0.109; corresponding P values all < 0.05). In contrast, no significant differences were observed when compared with sdLG-BRI (IDI = 0.004; NRI = 0.038; all P > 0.05). When compared with sdLG-WHtR, the IDI showed improvement (IDI = 0.005, P = 0.02), whereas the NRI showed borderline significance (NRI = 0.055, P = 0.059) (Table 2). Based on these findings, sdLG-cVAI was identified as the sdLG-derived index with the relatively optimal predictive performance. Associations between sdLG-cVAI and incident CVD Trajectories of sdLG-cVAI identified by k-means clustering As shown in the elbow plot, the within-cluster sum of squares (WSS) decreased sharply from k = 1 to k = 3 and plateaued thereafter, indicating diminishing returns in clustering performance beyond three clusters. Accordingly, k = 3 was selected as the optimal number of clusters (Fig. 3A and B). Trajectories of sdLG-cVAI identified by k-means clustering. A Selection of the optimal model (three-class solution) via the elbow method; (B) Visualization of the three distinct trajectory clusters derived from sdLG-cVAI at baseline and year 4; (C) Longitudinal sdLG-cVAI profiles for each trajectory cluster. sdLG-cVAI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Three distinct trajectories of sdLG-cVAI were identified: Cluster 1 (good control), characterized by low baseline sdLG-cVAI levels with a slight increase over time; Cluster 2 (moderate control), characterized by intermediate baseline levels with a gradual increase during follow-up; and Cluster 3 (poor control), characterized by high baseline sdLG-cVAI levels that remained persistently elevated throughout follow-up (Fig. 3C). These trajectory patterns reflected substantial heterogeneity in long-term cardiometabolic status. Baseline characteristics of participants grouped by baseline sdLG-cVAI, cumulative sdLG-cVAI (cusdLG-cVAI), and sdLG-cVAI trajectories are presented in Supplementary Tables S1–S3. Associations between sdLG-cVAI and CVD based on different classifications Kaplan–Meier survival analyses showed significantly higher cumulative incidence of CVD among participants in the highest quartiles of baseline sdLG-cVAI and cumulative sdLG-cVAI, as well as among those in the poor-control trajectory group (Log-rank test, P < 0.001) (Supplementary Figs. 1 A–1 C). Restricted cubic spline analyses demonstrated a significant linear positive association between baseline sdLG-cVAI and CVD risk (P for overall < 0.001; P for nonlinearity = 0.105) (Fig. 4A), as well as between cumulative sdLG-cVAI and CVD risk (P for overall < 0.001; P for nonlinearity = 0.522) (Fig. 4B). Restricted cubic spline analyses of the associations between sdLG-cVAI, cusdLG-cVAI, and cardiovascular disease risk. A Dose-response relationship of sdLG-cVAI with CVD risk; (B) Dose-response relationship of cusdLG-cVAI with CVD risk. sdLG-cVAI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index; cusdLG-cVAI, cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index The associations between sdLG-cVAI and incident CVD were further evaluated using Cox proportional hazards regression models (Table 3). When baseline sdLG-cVAI was analyzed as a categorical variable, a clear dose–response relationship was observed across quartiles from the unadjusted model to the fully adjusted model (Model 4). In Model 4, compared with participants in the lowest quartile (Q1), the risks of CVD increased by 31% (HR = 1.31; 95% CI: 1.08–1.60; P = 0.01), 47% (HR = 1.47; 95% CI: 1.21–1.79; P < 0.001), and 77% (HR = 1.77; 95% CI: 1.44–2.16; P < 0.001) for participants in Q2, Q3, and Q4, respectively. Per-standard-deviation (per-SD) increase in baseline sdLG-cVAI was associated with a 24% higher risk of CVD (HR = 1.24; 95% CI: 1.16–1.33; P < 0.001) in the fully adjusted model (Table 3). Table 3Associations of sdLG-cVAI, cumulative sdLG-cVAI, and changes in sdLG-cVAI with the risk of CVDCategoriesModel 1Model 2Model 3Model 4HR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valuesdLG-cVAI Quartile Q1RefRefRefRef Q21.38 (1.14, 1.68)0.0011.35 (1.11, 1.64)0.0031.35 (1.11, 1.64)0.0031.31 (1.08, 1.60)0.010 Q31.65 (1.36, 1.99)< 0.0011.57 (1.30, 1.90)< 0.0011.57 (1.30, 1.91)< 0.0011.47 (1.21, 1.79)< 0.001 Q42.17 (1.81, 2.60)< 0.0012.03 (1.69, 2.44)< 0.0012.04 (1.69, 2.46)< 0.0011.77 (1.44, 2.16)< 0.001P for trend< 0.001< 0.001< 0.001< 0.001sdLG-cVAI Per-SD increment1.33 (1.26, 1.42)< 0.0011.30 (1.23, 1.39)< 0.0011.31 (1.23,1.39)< 0.0011.24 (1.16, 1.33)< 0.001cusdLG-cVAI Quartile Q1RefRefRefRef Q21.47 (1.21, 1.79)< 0.0011.45 (1.19, 1.77)< 0.0011.45 (1.19,1.77)< 0.0011.41 (1.15, 1.72)< 0.001 Q31.71 (1.41, 2.07)< 0.0011.63 (1.34, 1.97)< 0.0011.63 (1.34,1.98)< 0.0011.52 (1.25, 1.86)< 0.001 Q42.28 (1.89, 2.73)< 0.0012.13 (1.77, 2.57)< 0.0012.15 (1.78, 2.60)< 0.0011.87 (1.53, 2.29)< 0.001P for trend< 0.001< 0.001< 0.001< 0.001cusdLG-cVAI Per-SD increment1.34 (1.26, 1.43)< 0.0011.31 (1.23, 1.40)< 0.0011.32 (1.24, 1.40)< 0.0011.25 (1.17, 1.34)< 0.001Change in the sdLG-cVAI Cluster 1RefRefRefRef Cluster 21.51 (1.29, 1.77)< 0.0011.45 (1.24, 1.70)< 0.0011.45 (1.24, 1.71)< 0.0011.39 (1.18, 1.63)< 0.001 Cluster 32.13 (1.80, 2.51)< 0.0011.99 (1.69, 2.36)< 0.0012.00 (1.69, 2.37)< 0.0011.75 (1.46, 2.10)< 0.001P for trend< 0.001< 0.001< 0.001< 0.001Model 1: unadjusted Model 2: adjusted for age, sex, nationalityModel 3: adjusted for age, sex, nationality, education, residence, marriage, smoke, drinkModel 4: adjusted for age, sex, nationality, education, residence, marriage, smoke, drink, hypertensionAbbreviations: DM hyperlipidemia, HR hazard ratio, CI confidence interval, SD standard deviations, Ref Reference, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, cusdLG-cVAI cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Associations of sdLG-cVAI, cumulative sdLG-cVAI, and changes in sdLG-cVAI with the risk of CVD Model 1: unadjusted Model 2: adjusted for age, sex, nationality Model 3: adjusted for age, sex, nationality, education, residence, marriage, smoke, drink Model 4: adjusted for age, sex, nationality, education, residence, marriage, smoke, drink, hypertension Abbreviations: DM hyperlipidemia, HR hazard ratio, CI confidence interval, SD standard deviations, Ref Reference, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, cusdLG-cVAI cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Similarly, for cumulative sdLG-cVAI, consistent positive associations with CVD risk were observed across all models; in the fully adjusted model, compared with participants in the lowest quartile (Q1), the risks of CVD increased by 41% (HR = 1.41; 95% CI: 1.15–1.72; P < 0.001), 52% (HR = 1.52; 95% CI: 1.25–1.86; P < 0.001), and 87% (HR = 1.87; 95% CI: 1.53–2.29; P < 0.001) for participants in Q2, Q3, and Q4, respectively. Per-SD increase in cumulative sdLG-cVAI was associated with a 25% higher risk of CVD (HR = 1.25; 95% CI: 1.17–1.34; P < 0.001) in the fully adjusted model (Table 3). In trajectory analyses, compared with participants in the good-control group, the risks of CVD increased by 39% (HR = 1.39; 95% CI: 1.18–1.63; P < 0.001) and 75% (HR = 1.75; 95% CI: 1.46–2.10; P < 0.001) for participants in the moderate-control and poor-control groups, respectively, in the fully adjusted model (Table 3). Multicollinearity diagnostics showed that all variance inflation factors were below 5, indicating stable model estimates (Supplementary Table S4). Predictive performance of baseline, cumulative, and trajectory forms of sdLG-cVAI Under fully adjusted models, time-dependent AUC analyses showed similar overall predictive performance for baseline sdLG-cVAI, cumulative sdLG-cVAI, and sdLG-cVAI trajectories (P > 0.05) (Fig. 5A). Consistent results were observed in the C-index analyses (Fig. 5B). Predictive performance of three sdLG-cVAI-based models over time. A Time-dependent AUC; (B) Time-dependent C-index. AUC, area under the receiver operating characteristic curve; C-index, Concordance index; sdLG-cVAI, Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index; cusdLG-cVAI, cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Further NRI/IDI analyses indicated that, compared with the baseline sdLG-cVAI and trajectory models, cumulative sdLG-cVAI did not show significant improvement (IDI = 0.001 / 0.001, P = 0.436 / 0.535; NRI = 0.038 / 0.019, P = 0.198 / 0.812) (Table 4). Table 4Comparison of the incremental predictive value of cumulative sdLG-cVAI using IDI and NRIComparisonIDI (95% CI)P valueNRI (95% CI)P valuesdLG-cVAI vs. cusdLG-cVAI0.001 (-0.001, 0.003)0.4360.038 (-0.020, 0.098)0.198cluster vs. cusdLG-cVAI0.001 (-0.002, 0.005)0.5350.019 (-0.052, 0.067)0.812Abbreviations: IDI Integrated Discrimination Improvement, NRI Net Reclassification Improvement, CI confidence interval, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, cusdLG-cVAI cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Comparison of the incremental predictive value of cumulative sdLG-cVAI using IDI and NRI Abbreviations: IDI Integrated Discrimination Improvement, NRI Net Reclassification Improvement, CI confidence interval, sdLG-cVAI Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index, cusdLG-cVAI cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Subgroup analysis Overall, whether sdLG-cVAI was analyzed as baseline level, cumulative exposure, or long-term trajectory, its positive association with CVD risk remained consistent across most predefined subgroups, indicating good robustness. There are no significant interactions in most subgroups (P for interaction > 0.05) ( Supplementary Figs. 2–4). Notably, effect modification differed across the three forms of sdLG-cVAI. In analyses of baseline sdLG-cVAI, diabetes mellitus emerged as a significant effect modifier (P for interaction = 0.017), with a steeper risk gradient observed among participants with diabetes. In cumulative sdLG-cVAI analyses, sex was the only variable showing a significant interaction (P for interaction = 0.025). Notably, the association between sdLG-cVAI and CVD risk manifested at lower quantiles in males, whereas this association attained statistical significance primarily at higher quantiles in females. In trajectory analyses, hypertension was identified as a significant effect modifier (P for interaction = 0.048), indicating differential associations between sdLG-cVAI trajectory patterns and CVD risk according to hypertension status. Sensitivity analyses Sensitivity analyses were conducted to assess the robustness of the findings. First, logistic regression models were applied to re-evaluate the associations between sdLG-cVAI and CVD risk. The direction and magnitude of the associations were largely consistent with those observed in the main analyses. In the fully adjusted model, when analyzed as categorical variables, participants in the highest quartile (Q4) of baseline sdLG-cVAI had a 97% higher risk of CVD compared with Q1 (OR = 1.97; 95% CI: 1.56–2.48; P < 0.001). For cumulative sdLG-cVAI, the corresponding increase in risk was 110% (OR = 2.10; 95% CI: 1.67–2.65; P < 0.001). For trajectory analyses, participants in the poor-control cluster had a 95% higher risk of CVD compared with those in the good-control cluster (OR = 1.95; 95% CI: 1.58–2.41; P < 0.001). Per-SD increase in baseline and cumulative sdLG-cVAI was associated with 30% (OR = 1.30; 95% CI: 1.20–1.41; P < 0.001) and 31% (OR = 1.31; 95% CI: 1.21–1.42; P < 0.001) higher risks of CVD, respectively (Supplementary Table S5). Second, to address potential reverse causality, analyses were repeated after excluding participants who developed CVD in 2013. The associations between sdLG-cVAI and CVD risk remained statistically significant. In the fully adjusted model, compared with Q1, participants in Q4 of baseline sdLG-cVAI had a 79% higher risk of CVD (HR = 1.79; 95% CI: 1.41–2.27; P < 0.001), and those in Q4 of cumulative sdLG-cVAI had a 110% higher risk (HR = 2.10; 95% CI: 1.64–2.67; P < 0.001). Participants in the poor-control trajectory group had an 89% higher risk of CVD compared with those in the good-control group (HR = 1.89; 95% CI: 1.52–2.34; P < 0.001). Per-SD increase in baseline and cumulative sdLG-cVAI was associated with 25% (HR = 1.25; 95% CI: 1.15–1.36; P < 0.001) and 27% (HR = 1.27; 95% CI: 1.17–1.38; P < 0.001) higher risks of CVD, respectively (Supplementary Table S6). These sensitivity analyses further supported the robustness of the main findings. Discussion In this nationwide prospective cohort study of middle-aged and older Chinese adults, we systematically evaluated the associations between the newly constructed sdLG index, its obesity-related derived indices, and the risk of incident CVD. The main findings were as follows: compared with the traditional TyG index, sdLG index demonstrated superior long-term risk discrimination; among all sdLG-derived obesity-related indices, sdLG-cVAI showed the relatively optimal discriminative ability and risk reclassification performance; moreover, baseline levels, cumulative exposure, and long-term trajectory patterns of sdLG-cVAI were all independently associated with incident CVD. These findings suggest that integrating lipoprotein quality, glycemic status, and visceral adiposity into a composite index may provide a more comprehensive assessment of cardiometabolic risk than traditional insulin resistance-related indicators. Compared with the classical TyG index, sdLG index exhibited stronger long-term predictive ability for CVD risk. This difference may be attributable to fundamental distinctions in the underlying pathophysiological information captured by the two indices. The lipid component of TyG index is mainly driven by plasma triglyceride levels, which reflect the overall burden of triglyceride-rich lipoproteins (TRLs) and their remnants rather than specific lipid metabolic pathways [6, 25]. Triglyceride levels are also highly sensitive to feeding status, circadian rhythm, and short-term metabolic fluctuations, resulting in substantial intra-individual biological variability that may limit the standardization and reproducibility of TG-based composite indices such as TyG index across different time points and populations [7, 26]. More importantly, TyG index does not directly capture a key mechanism of atherogenesis, namely qualitative remodeling of LDL subfractions. In contrast, sdLG index integrates glycemic burden with sdLDL-C, which more closely represents the atherogenic lipoprotein particles directly involved in atherosclerosis development. Small dense LDL particles are more prone to arterial wall penetration, oxidation, and prolonged circulation, making them particularly atherogenic [27–30]. Prospective studies have consistently demonstrated stable associations between sdLDL-related components and cardiovascular events, supporting their role as key contributors to the atherogenic lipid profile [31]. From a mechanistic perspective, therefore, sdLG index, by combining sdLDL-C with glucose, may more precisely capture the critical pathway linking metabolic dysregulation to vascular injury than TyG index. Our study further demonstrated that combining sdLG index with obesity-related indicators improved risk stratification, with sdLG-cVAI showing the most pronounced predictive advantage. This finding likely reflects the complementary nature of the information captured by these components. While sdLG index primarily reflects biochemical consequences of dysregulated glucose and lipid metabolism and atherogenic lipoprotein remodeling, visceral adiposity represents a key upstream driver of insulin resistance, chronic low-grade inflammation, and adverse lipid remodeling [32, 33]. Compared with conventional anthropometric measures such as BMI or waist circumference, the Chinese visceral adiposity index (cVAI) was specifically developed and validated to better reflect visceral fat dysfunction in the Chinese population [22, 34]. Thus, integrating sdLG index with cVAI may unify internal metabolic abnormalities and external adiposity-related phenotypes within a single quantitative framework, enabling more effective identification of individuals with both high exposure and high susceptibility to cardiovascular risk [22, 32, 33]. Importantly, this combination may represent a particularly atherogenic state, as small dense LDL particles more readily penetrate the arterial intima, exhibit reduced affinity for LDL receptors leading to prolonged circulation time, and are highly susceptible to oxidative modification, and these processes are further exacerbated within the insulin-resistant and pro-inflammatory milieu driven by visceral adiposity [8, 9, 32, 33]. Subgroup analyses in this study suggested potential effect heterogeneity in the association between sdLG-cVAI and CVD risk across different clinical contexts. Baseline sdLG-cVAI showed a significant interaction with diabetes status. This finding is biologically plausible, as diabetes is characterized by more pronounced atherogenic dyslipidemia and lipoprotein remodeling, and sdLDL-C particles in diabetic states are more susceptible to glycation and oxidative modification, thereby amplifying vascular injury and strengthening the association between metabolic exposure and cardiovascular events [35].In analyses of cumulative sdLG-cVAI, a significant interaction with sex was observed. This difference may reflect sex-specific variations in visceral fat accumulation, lipoprotein profiles, and hormonal milieu, particularly during the menopausal transition, when unfavorable changes in lipid metabolism become more pronounced [36, 37]. In trajectory analyses, hypertension was identified as a significant effect modifier. Previous studies have shown that elevated sdLDL-C levels can predict incident hypertension independent of traditional risk factors and lipid-lowering therapy, suggesting that sdLDL-C may be involved not only in atherosclerosis but also in metabolic and vascular processes related to blood pressure regulation [38]. Taken together, our findings provide population-based evidence supporting an integrated approach to cardiometabolic risk management. In middle-aged and older adults, focusing solely on LDL-C or glycemic indicators may be insufficient. Greater attention should be given to the combined effects of atherogenic lipoprotein subtypes, glycemic burden, and visceral adiposity on vascular risk [32, 33]. In resource-limited or primary care settings, sdLG index and its composite indices may offer a practical means of initial risk stratification using routine laboratory tests and basic anthropometric measurements, thereby helping clinicians identify individuals who may benefit from intensified preventive strategies. Several limitations of this study should be acknowledged. First, the study population consisted of Chinese adults aged 45 years and older; therefore, the generalizability of the findings to younger populations or other ethnic groups remains uncertain. Second, CVD outcomes were based on self-reported physician diagnoses, which may introduce misclassification bias. Third, because index measurements were available only at wave 1 and wave 3, the calculation of cumulative exposure may not fully capture long-term dynamic changes. Future studies incorporating more frequent repeated measurements may further improve exposure assessment. Fourth, as an observational study, residual confounding cannot be completely excluded despite comprehensive adjustment for potential confounders. Fifth, sdLDL-C was estimated using a validated equation rather than direct measurement, which may introduce potential estimation bias. However, given that direct assays are currently inaccessible in most medical institutions, a formula-based approach offers greater clinical utility and feasibility for implementation in primary care settings. Sixth, the selection of the optimal index was primarily based on its predictive performance at baseline. This approach may introduce potential selection bias, as indices that perform modestly in cross-sectional baseline analyses could exhibit stronger predictive value when evaluated using longitudinal metrics, such as cumulative exposure or trajectory patterns. Therefore, the superiority of sdLG-cVAI should be interpreted with caution, and future studies incorporating fully longitudinal model-based comparisons are warranted. Finally, in the CHARLS database, information on the initiation or adjustment of lipid-lowering drugs and hypoglycemic drugs during follow-up is very limited. These factors may affect the longitudinal changes of sdLG-cVAI and CVD risk assessment values. This should be regarded as a potential source of residual confounding factors. Conclusion In conclusion, sdLG index combined with obesity-related derived indices, particularly sdLG-cVAI, was significantly associated with incident cardiovascular disease and demonstrated favorable predictive performance. These findings suggest that composite indices integrating lipoprotein quality, glycemic status, and visceral adiposity may provide valuable tools for cardiovascular risk stratification and primary prevention. Supplementary Information Supplementary Material 1. Acknowledgements We thank all participants and the CHARLS project team. Abbreviations Cardiovascular Disease China Health and Retirement Longitudinal Study Hazard Ratio Confidence Interval Receiver Operating Characteristic Area Under the Curve Concordance index Net Reclassification Improvement Integrated Discrimination Improvement Standard Deviation Interquartile Range Restricted Cubic Spline Odds Ratio Variance Inflation Factor Waist Circumference Fasting blood glucose Triglycerides Small dense low-density lipoprotein cholesterol Low-density lipoprotein cholesterol High-Density Lipoprotein Cholesterol Body Mass Index Diabetes Mellitus Triglyceride-glucose index Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index Waist-to-Height Ratio Body roundness index Chinese visceral adiposity index A body shape index Weight-adjusted waist index Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body mass index Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist circumference Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-waist-to-height ratio Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-body roundness index Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-a body shape index Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-weight-adjusted waist index Cumulative Small Dense Low-Density Lipoprotein Cholesterol-Glucose Index-Chinese visceral adiposity index Authors' contributions XD and XX designed the study and drafted the manuscript; TW and QL collected and analyzed the data; XL, PQ, SL, ZL, HL, and YS visualized the results. KZ and XX read and revised the manuscript. All authors read and approved the final manuscript. Funding This work was supported by the research fund of Sichuan Medical and Health Care Promotion institute (KY2022QN0309); Sichuan Provincial Medical Association Youth Innovation Project (Q23021); Health Commission of Sichuan Province Medical Science and Technology Program (24QNMP100); Chengdu Medical Research Project (2025092); Medical Research Projects and Youth Innovation Projects of Sichuan Medical Association (S20250018); Nursing Research Topics of Sichuan Nursing Association (H25022). Data availability The CHARLS datasets are accessible in the CHARLS home website (http://charls.pku.edu.cn/en). Declarations Ethical Approval The CHARLS study was approved by the Institutional Review Board (IRB) of Peking University (IRB00001052-11015). The study protocol complies with the ethical standards of the 1964 Declaration of Helsinki and its subsequent amendments, and has been approved by the National Health Statistics Research Ethics Review Committee. All participants provided written informed consent. 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. Xumao Deng, Tao Wu and Qiyuan Liang contributed equally to this work. Contributor Information Kun Zhang, Email: zhangkunyyo@cmc.edu.cn. Xiang Xiao, Email: xxiang1001@163.com. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The CHARLS datasets are accessible in the CHARLS home website (http://charls.pku.edu.cn/en).