Integration of triglyceride-glucose index and biological aging to predict cardiovascular disease risk in cardiovascular–kidney–metabolic syndrome stages 0–3: a multiple prospective cohort study Authors: Runnan Shen, Pinrong Dong, Shuxiang Fang, Zirui Jia, Ayiguli Abudukeremu, Jiexin Li, Kai Huang, Xiaoqin Zuo, Yixuan Ma, Yilin Liu, Kai Huang, Yangxin Chen Published: 2026 ABSTRACT AbstractBackgroundThe triglyceride–glucose (TyG) index serves as a surrogate marker of insulin resistance and reflects underlying metabolic dysfunction, while biological aging (KDM-BA) independently contributes to cardiovascular vulnerability. However, no study has attempted to integrate metabolic dysfunction and biological aging to improve cardiovascular disease (CVD) risk prediction in individuals with cardiovascular–kidney–metabolic (CKM) syndrome stages 0–3.MethodsMultiple prospective cohorts were used, including 286,864 participants from the UK Biobank (UKB), 7237 from the China Health and Retirement Longitudinal Study (CHARLS), and 517 from Sun Yat-sen Memorial Hospital, Sun Yat-sen University, which were further included for external validation of the primary outcome. TyG-KDM-BA was calculated by multiplying the TyG index by KDM-BA. Cox proportional hazards models were used to evaluate the association. Time-dependent receiver operating characteristic (ROC) curves, as well as analyses of incremental predictive value, sensitivity analyses, and external validation were conducted to evaluate predictive performance.ResultsDuring a median follow-up of 15.27 years in UKB, 57,896 incident CVD events occurred (cumulative incidence: 20.18%). During a median follow-up of 7.84 years in CHARLS cohort, 1,610 incident CVD events were observed (cumulative incidence: 22.25%). CVD risk increased progressively across TyG-KDM-BA quartiles in both cohorts. In the fully adjusted model, each 1-SD increase in TyG-KDM-BA was associated with higher CVD risk in both cohorts (UKB: HR = 1.65, 95% CI 1.64–1.67; CHARLS: HR = 1.25, 95% CI 1.18–1.32). Participants in the highest quartile had significantly higher CVD risk than those in the lowest (UKB: HR = 3.97, 95% CI 3.86–4.09; CHARLS: HR = 1.92, 95% CI 1.63–2.25). TyG-KDM-BA was positively associated with CHD and stroke risk in both cohorts. Compared with conventional TyG-derived indices, TyG-KDM-BA achieved higher discrimination for CVD risk prediction, with an area under the curve of 0.657–0.659 over 5–15 years in UKB and moderate, consistent performance in CHARLS. Incremental analyses demonstrated that TyG-KDM-BA provided additional predictive value beyond conventional models and its individual components. External validation analyses supported the robustness of results.ConclusionTyG-KDM-BA integrates metabolic and aging dimensions, demonstrating superior predictive ability for CVD risk among individuals with CKM stages 0–3 compared to traditional TyG-derived indicators, and may serve as a more effective tool for the early screening and intervention.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12933-026-03192-2.Keywords: Cardiovascular disease, Cardiovascular–kidney–metabolic syndrome, Insulin resistance, Klemera–Doubal biological age Research insights What is currently known about this topic? Cardiovascular–kidney–metabolic (CKM) syndrome is associated with an increased risk of cardiovascular disease (CVD). Cardiovascular–kidney–metabolic (CKM) syndrome is associated with an increased risk of cardiovascular disease (CVD). Triglyceride-glucose (TyG) index-related indices serve as surrogate markers of insulin resistance and reflect underlying metabolic dysfunction, while biological aging independently contributes to cardiovascular vulnerability beyond chronological age. Triglyceride-glucose (TyG) index-related indices serve as surrogate markers of insulin resistance and reflect underlying metabolic dysfunction, while biological aging independently contributes to cardiovascular vulnerability beyond chronological age. No study has attempted to integrate metabolic dysfunction and biological aging to improve CVD risk prediction in individuals with CKM stage 0–3. No study has attempted to integrate metabolic dysfunction and biological aging to improve CVD risk prediction in individuals with CKM stage 0–3. What is the key research question? Can TyG-KDM-BA improve early CVD risk prediction among individuals with CKM syndrome? Can TyG-KDM-BA improve early CVD risk prediction among individuals with CKM syndrome? What is new? We developed a novel composite indicator (TyG-KDM-BA) integrating metabolic dysfunction and biological aging. We developed a novel composite indicator (TyG-KDM-BA) integrating metabolic dysfunction and biological aging. TyG-KDM-BA demonstrated superior predictive performance for CVD in CKM syndrome stages 0–3 compared with conventional TyG-related indices. TyG-KDM-BA demonstrated superior predictive performance for CVD in CKM syndrome stages 0–3 compared with conventional TyG-related indices. Findings were validated in two independent cohorts (UK Biobank and CHARLS) and in external real-world data. Findings were validated in two independent cohorts (UK Biobank and CHARLS) and in external real-world data. How might this study influence clinical practice? TyG-KDM-BA may enhance risk stratification and facilitate earlier identification of high-risk individuals for CVD in CKM populations. TyG-KDM-BA may enhance risk stratification and facilitate earlier identification of high-risk individuals for CVD in CKM populations. Background Cardiovascular disease (CVD) remains one of the leading causes of global mortality, with its onset and progression strongly associated with multiple traditional risk factors and metabolic factors [1]. In 2023, the American Heart Association (AHA) formally proposed the concept of Cardiovascular-Kidney-Metabolic (CKM) syndrome, emphasizing the complex pathophysiological interactions among the cardiovascular system, chronic kidney disease, and metabolic risk factors in driving cardiovascular risk [2]. Compared with isolated disease states, CKM syndrome reflects the cumulative impact of multiple mechanisms, including adipose tissue dysfunction, chronic inflammation, and metabolic syndrome (MetS) [3], resulting in an additive cardiovascular risk burden. According to AHA recommendations, CKM syndrome is classified into stages 0–4, with stages 0–3 representing the preclinical phase and encompassing more than 90% of the population [4]. Although most individuals at these stages are often asymptomatic, subclinical cardiovascular impairments may already be present. From a public health and preventive medicine perspective, early intervention during CKM stages 0–3 offers substantial health benefits and cost-effectiveness [5]. Therefore, the development of reliable risk prediction indicators for this population is essential. Nevertheless, the progression of CKM syndrome from stage 0 to stage 3 constitutes a continuous pathological trajectory. Early stages are primarily driven by insulin resistance (IR) and metabolic dysregulation, whereas the advanced stage 3 features marked organ functional decline and physiological reserve exhaustion [3]. Given this multidimensional nature, single-dimensional markers may not fully capture the overall disease burden. Therefore, integrating multidimensional biomarkers is essential for risk stratification across stages 0–3. IR is a central pathophysiological mechanism underlying CKM syndrome [6, 7], accelerating cardiovascular and renal damage through multiple interconnected pathways, thereby promoting the initiation and progression of CKM syndrome [8]. However, its accurate quantification remains challenging in large-scale clinical settings. The gold standard hyperinsulinemic-euglycemic clamp (HEC) is limited by poor feasibility in large-scale studies [9], whereas surrogate indicators like the homeostasis model assessment of IR (HOMA-IR) are subject to measurement bias and accuracy [10]. Therefore, there is a critical challenge to find an IR assessment tool. The Triglyceride-Glucose Index (TyG) is a simple metric calculated from fasting triglycerides (TG) and fasting plasma glucose (FPG) [11]. It has been widely validated as a reliable surrogate marker for IR and showed significant predictive value for CVD risk [4, 6, 8]. Recent studies have combined the TyG index with obesity-related measures to better capture metabolic dysfunction, such as TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), TyG-body mass index (TyG-BMI) [4, 12], TyG-Weight-Adjusted Waist Index (TyG-WWI) [13], TyG-Chinese Visceral Adipose Index (TyG-CVAI) [14], TyG-Body Roundness Index (TyG-BRI) [15], and TyG-A Body Shape Index (TyG-ABSI) [16, 17], which demonstrated superior predictive performance compared to the TyG index alone [8]. However, current TyG-related indicators primarily capture metabolic and adiposity abnormalities but not cumulative functional decline and physiological damage. Aging represents a fundamental risk factor for CVD. It promotes vascular injury and increases the risk of cardiovascular events through the synergistic effects of multiple pathways including chronic inflammation, oxidative stress, mitochondrial dysfunction, and endothelial injury [18–21]. However, chronological age alone fails to adequately reflect individual differences in metabolic function levels [22–25]. In contrast, biological age indicators constructed from multiple biomarkers more accurately characterize the aging process and have shown strong associations with cardiovascular risk and outcomes [19, 26, 27]. Among these, biological age estimated using the Klemera–Doubal method (KDM Biological Age) has been widely applied and linked to both CVD occurrence and metabolic dysfunction [27, 28]. In summary, during CKM stages 0–3, the TyG index effectively captures early metabolic abnormalities, while KDM-BA reflects cumulative systemic physiological damage. Given their complementary roles, this study constructed a composite indicator, TyG-KDM-BA, to integrate these two dimensions and improve cardiovascular risk assessment in individuals with CKM stages 0–3, and evaluate its predictive performance across multiple cohorts. Methods Study design and participants Data were obtained from three independent cohorts: the UK Biobank (UKB), the China Health and Retirement Longitudinal Study (CHARLS), and an external hospital-based cohort from Sun Yat-sen Memorial Hospital, Sun Yat-sen University. UKB is a large-scale prospective cohort that recruited more than 500,000 participants from England, Scotland, and Wales between 2006 and 2010. Comprehensive data were collected through touchscreen questionnaires, physical examinations, oral interviews, biological samples, and imaging assessments [29]. Ethical approval was obtained from the North West Multi-Centre Research Ethics Committee (MREC 21/NW/0157), and all participants provided informed consent before enrollment. This study utilized UKB application number 100739. Among 502,140 participants, we first excluded individuals lacking information on CVD diagnosis or diagnosis timing (n = 55,539), missing follow-up duration (n = 179), prevalent CVD with incomplete diagnostic records at baseline (n = 20,162), lacking TyG-related indicators (n = 62,084) or KDM biological age (n = 47), and those with missing or non–CKM stage 0–3 classification (n = 62,438). Finally, we excluded participants with missing covariates (n = 15,007). The final analysis sample comprised 286,864 participants with CKM syndrome stage 0–3. CHARLS is a nationwide survey of Chinese adults aged ≥ 45 years, employing a multistage stratified probability proportional-to-size sampling strategy to recruit from 150 counties or districts across 28 provinces in China. The study adheres to the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of Peking University (IRB00001052-11015), and written informed consent was obtained. Data were collected through face-to-face interviews conducted by trained fieldworkers using standardized questionnaires [30]. From the baseline cohort of 17,708 individuals (2011–2012), the following groups were excluded: those < 45 years old (n = 508) to ensure consistency with the target population, those with a history of heart disease and stroke medication use at baseline (n = 2,975), individuals with missing follow-up outcomes (n = 815), and those with missing key variables (TyG index, KDM-BA calculation indicators, CKM staging indicators) (n = 6,173). A total of 7,237 individuals were included in the final analysis. The external hospital-based cohort was obtained from the inpatient database of Sun Yat-sen Memorial Hospital, including 1,810 individuals admitted between January 1, 2015, and June 1, 2025. From the baseline cohort of 1810 individuals, the following groups were excluded: those < 45 years old (n = 280), those with pre-existing cardiovascular disease or with the history of heart disease and stroke medication use at baseline (n = 443), individuals with missing follow-up outcomes (n = 104), and those with missing key variables (TyG index, KDM-BA calculation indicators, CKM staging indicators) (n = 466). A total of 517 individuals were included in the final analysis. Figure 1 illustrates the participant selection flowchart for the three cohorts. Flowchart of the study population. Note A: UK Biobank; B: CHARLS; C: Hospital-based cohort. Abbreviations: CVD, cardiovascular disease; TyG, triglyceride–glucose index; KDM-BA, biological age derived from the Klemera–Doubal method; CKM, Cardiovascular-Kidney-Metabolic; CHARLS, China Health and Retirement Longitudinal Study Definition of KDM-BA The Klemera–Doubal (KDM) method is a validated approach for estimating biological age (BA) by integrating multiple age-related biomarkers into a weighted composite index based on their linear associations with chronological age. This method has demonstrated robust predictive performance for age-related health outcomes in multiple cohort studies involving both British and Chinese populations [31]. Refer to prior research, we trained the baseline KDM-BA for participants in UKB using nine selected biomarkers: albumin, alkaline phosphatase, blood urea nitrogen, cholesterol, C-reactive protein, glycated hemoglobin, creatinine, systolic blood pressure, and forced vital capacity. These indicators collectively reflect key physiological domains, including hepatic and renal function, systemic inflammation, metabolic levels, and cardiopulmonary function [32]. Due to missing data for certain indicators in CHARLS and the need to ensure consistency with subsequent TyG index-related analyses, we selected urea nitrogen, total cholesterol, C-reactive protein, glycated hemoglobin, creatinine, systolic blood pressure, and platelet count to calculate KDM-BA in the validation set baseline, following other studies [28]. Subsequently, the trained KDM-BA algorithm was extrapolated and validated against external cohort hospital data, which employed the same biomarker set as UKB. Chronological age and estimated BA were recorded, and model performance was evaluated using the Pearson correlation coefficient (r), mean absolute error (MAE), and root mean square error (RMSE). All KDM-BA calculations were performed using the “BioAge” R package. Calculation of TyG index-related metrics The TyG index was calculated according to the established formula: TyG = ln [TG (mg/dL) × FBG (mg/dL) / 2] [11]. In UKB, biochemical measurements were obtained from peripheral venous blood samples collected at baseline, processed using standardized protocols, and stored at − 80 °C. Procedures for sample collection, transportation, laboratory testing, and quality control adhered strictly to the standard operating procedures of UKB, as documented on the official platform and in prior publications [33]. In CHARLS, blood samples were similarly collected during the baseline survey and analyzed according to standardized survey protocols, with quality control procedures described previously. Biochemical parameters in the external validation cohort were extracted from routine clinical laboratory test records. All assays were conducted by the hospital’s certified clinical laboratory, and data underwent standardized cleaning and validation procedures aligned with established quality control protocols. We constructed a composite index, TyG-KDM-BA, as the product of the TyG index and KDM-derived biological age: TyG-KDM-BA = TyG × KDM-BA. Definition of covariates Covariates were selected based on prior research and data availability. To avoid over-adjustment and mathematical coupling, variables that were components of, or closely related to, the TyG index or KDM-BA were excluded. In UKB cohort [4], the following covariates were included: sex, body mass index (BMI), white blood cell count, red blood cell count, hemoglobin concentration, smoking status (never, former, or current smoker), drinking status (never, former, or current drinker), cancer, educational attainment (high school and above vs other), presence of insomnia or sleep difficulties, weekly engagement in moderate-to-vigorous physical activity (MVPA), and Townsend Deprivation Index. Detailed definitions of covariates are provided in Table S1. Given differences in data structure and variable collection across cohorts, covariate selection was adapted accordingly while maintaining consistency in core confounders. Covariates selected for CHARLS cohort included sex, BMI, red blood cell count, white blood cell count, hemoglobin concentration, smoking status (never, former, or current smoker based on smoking history), and drinking status (never, former, or current drinker based on drinking history), cancer, educational attainment (high school and above vs other), insomnia or sleep difficulties and MVPA. Missing covariate data in CHARLS cohort were handled using multiple imputation to reduce potential bias and preserve statistical power [34]. In the external validation cohort, selected covariates included sex, BMI, red blood cell count, white blood cell count, hemoglobin concentration, smoking status (never or current smoker), alcohol consumption status (never or current drinker), cancer, and educational attainment (high school and above vs other). Outcome assessment The primary outcome in UKB was new-onset overall cardiovascular disease (CVD) among participants with CKM stages 0–3. CVD was defined as a composite endpoint including coronary heart disease (CHD), heart failure, stroke, atrial fibrillation or atrial flutter, and peripheral vascular disease [35–39]. Secondary outcomes included incident CHD and incident stroke. Participants without CVD at baseline were followed from their first assessment date until the first occurrence of a CVD event, death, or end of follow-up (August 31, 2024), whichever occurred first. Incident CVD events were ascertained through data linkage with hospital records, primary care records, and the national death registry. Outcome events were identified using International Classification of Diseases, 10th Revision (ICD-10) codes, as detailed in Table S2 [38, 40–42]. In CHARLS, outcomes were determined based on self-reported physician diagnoses obtained through standardized questionnaires [43]: “Has a doctor diagnosed you with a heart attack, coronary artery disease, angina, heart failure, or other heart-related issues?” and “Has a doctor diagnosed you with a stroke?” Participants were classified as having CVD if they responded affirmatively to either of the questions. In the hospital-based cohort, outcomes were primarily identified according to physician diagnoses of heart disease and stroke documented in clinical records. Definition of CKM syndrome stages According to the AHA, CKM syndrome is classified into stages 0 to 4, with stages 0–3 representing the early clinical phase. The classification criteria are as follows [3, 44, 45]: Stage 0: Individuals with normal BMI (non-Asian BMI < 25 kg/m2, Asian population < 23 kg/m2), normal WC (non-Asian women < 88 cm, men < 102 cm; Asian women < 80 cm, men < 90 cm), normal blood glucose, normal blood pressure, normal lipid levels, and no evidence of chronic kidney disease or subclinical cardiovascular disease. Stage 1: Individuals with overweight/obesity or early metabolic abnormalities, defined by meeting any of the following criteria: (1) Elevated BMI (non-Asian BMI ≥ 25 kg/m2, Asian populations ≥ 23 kg/m2); (2) Elevated WC (non-Asian women ≥ 88 cm, men ≥ 102 cm; Asian women ≥ 80 cm, men ≥ 90 cm); (3) Early-stage diabetes (HbA1c levels between 5.7% and 6.5%, or fasting blood glucose between 100 mg/dL and 126 mg/dL). Stage 2: Includes individuals with established metabolic disease or moderate-to-high-risk chronic kidney disease (CKD), defined by any of the following: (1) Fasting triglyceride level ≥ 135 mg/dL; (2) Hypertension (systolic blood pressure, SBP ≥ 140 mmHg; diastolic blood pressure, DBP ≥ 90 mmHg, or use of antihypertensive medication); (3) Diabetes (fasting blood glucose ≥ 126 mg/dL, HbA1c ≥ 6.5%, or use of antidiabetic medication); (4) Metabolic syndrome (meeting at least three of the following: elevated waist circumference, low HDL cholesterol, high triglycerides, hypertension, and impaired glucose metabolism); (5) Moderate-to-high risk CKD (according to KDIGO staging criteria: Stage 1 = 45 ≤ GFR < 60, Stage 2 = 30 ≤ GFR < 45). Stage 3: Individuals with subclinical cardiovascular abnormalities or very high-risk CKD. Stage 3 is defined by meeting any of the following criteria, including: (1) very high-risk CKD (KDIGO stage 3); or (2) elevated 10-year cardiovascular risk estimated by the Framingham risk score (women ≥ 21.5%; men ≥ 21.6%). Stage 4: Individuals with clinically diagnosed cardiovascular disease, identified through self-reported conditions such as heart failure, coronary artery disease, or stroke. Statistical analysis Baseline characteristics were summarized using descriptive statistics. Normally distributed continuous variables were expressed as mean ± standard deviation (SD), while skewed variables were expressed as median (interquartile range, IQR). Categorical variables were expressed as frequencies and percentages. Between-group comparisons were performed using the independent samples t-test for normally distributed continuous variables, the Kruskal–Wallis rank sum test for non-normally distributed continuous variables, and Pearson's chi-square test for categorical variables. Participants were divided into four groups (Q1-Q4) based on TyG-KDM-BA quartiles. CVD incidence rates were calculated for each quartile. Trends across quartiles were assessed using the Cochran–Armitage trend test. Associations between TyG-KDM-BA and incident CVD among participants with CKM stages 0–3 were evaluated using Cox proportional hazards regression models, with hazard ratios (HRs) and 95% CIs reported per standard deviation (SD) increase and by quartiles (Q2–Q4 vs Q1). Cumulative incidence rates for each outcome were assessed using Kaplan–Meier (KM) survival curves, with intergroup comparisons performed using log-rank tests. We constructed four stepwise-adjusted models: Model 1 was the unadjusted baseline model; Model 2 adjusted for demographic factors; Model 3 further adjusted for general health-related indicators; Model 4 additionally incorporated socioeconomic and lifestyle factors and potential confounders such as prior medical history. The specific covariates included in each model were appropriately adjusted based on data availability across different cohorts, with core confounders consistently retained. The proportional hazards (PH) assumption was assessed using Schoenfeld residuals based on the fully adjusted Cox model (Model 4). When violations were detected, additional analyses were conducted using stratified Cox models across follow-up time intervals and time-segmented Cox regression, in which follow-up time was categorized into 0–5, 5–10, 10–15, and ≥ 15 years. Multicollinearity among covariates was assessed using variance inflation factors (VIF). For variables with more than one degree of freedom, the generalized variance inflation factor (GVIF) was calculated and adjusted as GVIF ^ (1 / (2 × Df)) to ensure comparability with VIF. All variables met predefined criteria for acceptable collinearity (VIF < 5). Restricted cubic spline (RCS) analyses were conducted within the fully adjusted Cox model to explore potential nonlinear associations between TyG-KDM-BA and CVD risk. Four knots were placed at the 5th, 35th, 65th, and 95th percentiles, with the median value as the reference. When significant nonlinearity was detected, a two-stage Cox regression model was employed for threshold effect analysis, using the median as the cutoff point. Time-dependent receiver operating characteristic (ROC) curves were constructed to evaluate the discriminative ability of the TyG-KDM-BA index for predicting cardiovascular outcomes at multiple follow-up time points. The area under the curve (AUC) was calculated at predefined time horizons to assess temporal changes in predictive performance. The change in C-index (ΔC-index), Net Reclassification Improvement (NRI), and Integrated Discrimination Improvement (IDI) were calculated to evaluate the incremental predictive value of the TyG-KDM-BA index. Comparisons were performed across multiple models, including clinical covariates, TyG, KDM-BA, TyG and KDM-BA as separate covariates, and TyG-KDM-BA. All models were based on the fully adjusted Cox regression. Subsequently, pre-specified subgroup analyses were performed in UKB and CHARLS cohorts stratified by sex, age, hypertension, diabetes, metabolic syndrome, BMI categories, smoking status, drinking status, and CKM stages 0–3. Interaction terms between the TyG-KDM-BA index and subgroup variables were added to the Cox models to test for effect modification. To further explore stage-specific predictive performance, we calculated the Pearson linear correlation coefficient and Spearman's rank correlation between TyG-KDM-BA and CVD within each CKM stage. Violin plots were used to display the distribution characteristics. A series of sensitivity analyses were conducted to assess the robustness of the findings: (1) additional adjustments were performed for antihypertensive, lipid-lowering, and antidiabetic medication use in both cohorts; (2) since missing covariates were deleted in UKB, multiple imputation methods was performed to handle missing data, and the Cox proportional hazards models were reanalyzed; (3) excluding events occurring within the first two years of follow-up to reduce reverse causation; (4) analyses were repeated retaining extreme values, whereas the primary analyses applied 1st–99th percentile to reduce the influence of extreme values. Further analyses were conducted to address key methodological considerations. (1) To assess the robustness of the KDM-BA algorithm across cohorts with different biomarker availability, a sensitivity analysis was conducted in UKB by recalculating KDM-BA using only the 7 biomarkers available in CHARLS, and the corresponding TyG-KDM-BA was derived. The distributions were compared with those from the original model, and Pearson correlation coefficients were calculated. (2) To account for the competing risk, Fine–Gray subdistribution hazard models were applied, treating non-CVD death as a competing event. (3) To improve comparability of outcome definitions between cohorts, as UKB included both fatal and non-fatal events, whereas CHARLS primarily captured non-fatal events, an additional analysis was conducted in UKB by restricting the outcome to non-fatal CVD events, using the same covariate adjustments as in the primary analysis. (4) Heterogeneity between UKB and CHARLS was evaluated using a random-effects meta-analysis, with heterogeneity quantified by the I2 statistic and Cochran’s Q test. Finally, we used Sun Yat-sen Memorial Hospital cohort data as an external validation set to replicate the Cox model and assess its stable performance. Results Baseline characteristics of participants Baseline characteristics of participants are shown in Table 1. A total of 286,864 individuals were included from UKB, of whom 57,896 developed incident CVD during follow-up. The external validation cohorts comprised 7,237 participants from CHARLS (1,610 CVD events) and 517 participants from the hospital cohort (131 CVD events). Participants were divided into two groups based on whether they developed CVD during follow-up. Table 1The baseline characteristics of participants in UKB, CHARLS, and hospital-based cohortsBaseline VariablesUK Biobank (N = 286,864)CHARLS (N = 7237)Hospital-based Cohort (N = 517)Non-CVD(N = 228,968)CVD (N = 57,896)P ValueNon-CVD(N = 5627)CVD (N = 1610)P ValueNon-CVD(N = 386)CVD (N = 131)P ValueAge, years56.88 (48.91, 64.85)61.25 (54.55, 67.95)< 0.00158.48 (49.22, 67.74)60.12 (51.58, 68.65)< 0.00162.39 (49.65, 75.14)65.11 (54.42, 75.79)0.018Sex< 0.001< 0.0010.003 Female119,510 (52.20%)22,189 (38.33%)2868 (50.97%)935 (58.07%)227 (58.81%)57 (43.51%) Male109,458 (47.80%)35,707 (61.67%)2759 (49.03%)675 (41.93%)159 (41.19%)74 (56.49%)Education< 0.0010.0140.036 High school and above129,959 (56.76%)28,074 (48.49%)1610 (28.61%)410 (25.47%)76 (19.69%)38 (29.01%) Less than high school99,009 (43.24%)29,822 (51.51%)4017 (71.39%)1200 (74.53%)310 (80.31%)93 (70.99%)Smoking status< 0.0010.0040.324 Never126,450 (55.23%)26,404 (45.61%)3373 (59.94%)1005 (62.42%)323 (83.68%)104 (79.39%) Former78,918 (34.47%)23,905 (41.29%)429 (7.62%)147 (9.13%)–– Current23,600 (10.31%)7587 (13.10%)1825 (32.43%)458 (28.45%)63 (16.32%)27 (20.61%)Drinking status< 0.001< 0.0010.040 Never9487 (4.14%)2716 (4.69%)3215 (57.14%)977 (60.68%)335 (86.79%)123 (93.89%) Former7741 (3.38%)2745 (4.74%)410 (7.29%)151 (9.38%)–– Current211,740 (92.48%)52,435 (90.57%)2002 (35.58%)482 (29.94%)51 (13.21%)8 (6.11%)Hypertension62,923 (27.48%)25,427 (43.92%)< 0.0011047 (18.61%)531 (32.98%)< 0.001161 (41.71%)71 (54.2%)0.017Diabetes10,403 (4.54%)5,819 (10.05%)< 0.001250 (4.44%)113 (7.02%)< 0.00158 (15.03%)25 (19.08%)0.339Metabolic syndrome87,347 (38.15%)28,884 (49.89%)< 0.0013150 (55.98%)1123 (69.75%)< 0.001209 (54.15%)73 (55.73%)0.832Cancer18,011 (7.87%)5,635 (9.73%)< 0.00143 (0.76%)9 (0.56%)0.48964 (16.58%)33 (25.19%)0.040Weight, kg79.67 (64.80, 94.54)83.53 (67.13, 99.93)< 0.00158.06 (46.98, 69.15)59.98 (48.11, 71.86)< 0.00163.97 (52.11, 75.83)63.77 (52.26, 75.27)0.862Height, m1.69 (1.59, 1.78)1.70 (1.61, 1.80)< 0.0011.58 (1.46, 1.70)1.57 (1.45, 1.69)0.01981.60 (1.52, 1.69)1.62 (1.54, 1.69)0.064BMI, kg/m227.95 (23.41, 32.48)28.84 (23.82, 33.85)< 0.00122.87 (18.24, 27.50)23.73 (18.45, 29.01)< 0.00124.91 (20.7, 29.13)24.31 (20.72, 27.89)0.112SBP, mm Hg140.61 (121.56, 159.67)145.89 (126.10, 165.68)< 0.001128.72 (107.84, 149.59)134.15 (111.86, 156.43)< 0.001136.40 (116.55, 156.24)133.79 (112.76, 154.83)0.216DBP, mm Hg83.14 (72.73, 93.54)83.76 (72.85, 94.66)< 0.00175.04 (63.14, 86.93)77.26 (64.69, 89.82)< 0.00181.76 (69.58, 93.94)78.65 (67.12, 90.18)0.009FBG, mg/dL89.01 (76.05, 101.97)90.20 (75.64, 104.76)< 0.001102.06 (83.88, 120.24)103.14 (83.52, 122.76)< 0.00191.80 (70.20, 113.40)91.80 (63.90, 119.70)0.997HbA1c, %5.38 (4.91, 5.85)5.47 (4.95, 5.99)< 0.0015.10 (4.60, 5.60)5.20 (4.60, 5.80)< 0.0015.80 (4.9, 6.7)5.80 (4.6, 7)0.191Scr, mg/dL0.80 (0.58, 1.02)0.84 (0.61, 1.07)< 0.0010.76 (0.52, 0.99)0.75 (0.52, 0.97)< 0.0010.79 (0.52, 1.05)0.90 (0.54, 1.27)< 0.001TyG8.78 (8.23, 9.32)8.87 (8.30, 9.43)< 0.0018.65 (7.99, 9.30)8.75 (8.08, 9.42)< 0.0018.66 (8.1, 9.23)8.73 (8.05, 9.41)0.308KDM-BA56.93 (48.90, 64.95)61.24 (54.45, 68.04)< 0.00158.48 (49.06, 67.89)60.12 (51.46, 68.77)< 0.00162.38 (49.59, 75.18)65.16 (54.56, 75.76)0.015KDM-BA advance0.05 (− 0.53, 0.62)− 0.01 (− 0.69, 0.68)< 0.001− 0.09 (− 0.96, 0.78)− 0.09 (− 0.95, 0.78)0.9968− 0.01 (− 0.64, 0.61)0.05 (− 0.71, 0.81)0.390TyG-KDM-BA499.87 (421.06, 578.68)542.74 (474.90, 610.58)< 0.001505.55 (416.46, 594.64)525.66 (442.04, 609.28)< 0.001540.62 (424.93, 656.3)567.94 (470.6, 665.29)0.009CKM stage< 0.001< 0.0010.330Stage 010,983 (4.80%)1940 (3.35%)768 (13.65%)125 (7.76%)23 (5.96%)3 (2.29%)Stage 141,462 (18.11%)5444 (9.40%)1090 (19.37%)216 (13.42%)54 (13.99%)21 (16.03%)Stage 2140,305 (61.28%)34,806 (60.12%)3062 (54.42%)1024 (63.60%)233 (60.36%)77 (58.78%)Stage 336,218 (15.82%)15,706 (27.13%)707 (12.56%)245 (15.22%)76 (19.69%)30 (22.9%)Data are presented as mean (SD) or median (interquartile range) for continuous variables and n (%) for categorical variablesSmoking and drinking status in the hospital cohort were e recorded as a binary history (yes/no), rather than detailed categories (never, former, current)CVD, cardiovascular disease; BMI, body mass index; WC, waist circumference; SBP, systolic blood pressure; DBP, diastolic blood pressure; FBG, fasting blood glucose; HbA1c, hemoglobin; Scr, serum creatinine; TyG, triglyceride–glucose index; KDM-BA, Klemera–Doubal method biological age; KDM-advance, KDM biological age advancement; CKM, Cardiovascular-Kidney-Metabolic; CHARLS, China Health and Retirement Longitudinal Study The baseline characteristics of participants in UKB, CHARLS, and hospital-based cohorts Data are presented as mean (SD) or median (interquartile range) for continuous variables and n (%) for categorical variables Smoking and drinking status in the hospital cohort were e recorded as a binary history (yes/no), rather than detailed categories (never, former, current) CVD, cardiovascular disease; BMI, body mass index; WC, waist circumference; SBP, systolic blood pressure; DBP, diastolic blood pressure; FBG, fasting blood glucose; HbA1c, hemoglobin; Scr, serum creatinine; TyG, triglyceride–glucose index; KDM-BA, Klemera–Doubal method biological age; KDM-advance, KDM biological age advancement; CKM, Cardiovascular-Kidney-Metabolic; CHARLS, China Health and Retirement Longitudinal Study In UKB cohort, participants who developed CVD were older and more likely to be male, had lower educational attainment, higher rates of smoking and alcohol consumption, higher body mass index, and a greater proportion of advanced CKM stage (stage 3) compared with those who remained CVD-free. They also exhibited a higher prevalence of comorbid hypertension, diabetes, metabolic syndrome, and cancer history (all P < 0.001). Additionally, individuals who developed CVD showed less favorable cardiometabolic and inflammatory profiles, including higher systolic blood pressure, glycated hemoglobin, triglyceride levels, and TyG-KDM-BA (all P < 0.001). In CHARLS cohort, compared with those who remained CVD-free, participants who developed CVD were older, more likely to have lower educational attainment, higher rates of smoking and alcohol consumption, higher body mass index, and a greater proportion of advanced CKM stage (stage 3), and also exhibited a higher prevalence of comorbid hypertension, diabetes, and metabolic syndrome. Notably, participants who developed CVD were more likely to be female (P < 0.001), whereas no significant difference in cancer history was observed, likely reflecting the limited number of cancer cases. In the hospital-based validation cohort, participants who developed CVD were older, more likely to be male, and had a higher prevalence of hypertension, elevated creatinine, and higher TyG-KDM-BA levels compared with those who remained CVD-free. Most other baseline characteristics generally showed similar directional trends but did not reach statistical significance. This may be attributed to the small sample size and differences in population structure of the hospital cohort. Incidence of CVD, CHD, and stroke The incidence rates of outcomes in both cohorts are presented in Table S3. The median follow-up duration was 15.27 years (IQR: 13.44, 17.10) in UKB. During this period, 57,896 participants developed incident CVD, corresponding to a cumulative incidence of 20.18% and an incidence density of 144.80 per 10,000 person-years (95% CI: 143.63–145.99). Kaplan–Meier curves stratified by quartiles of TyG-KDM-BA (Fig. 2) demonstrated progressively higher cumulative CVD incidence across increasing quartiles, with rates of 8.59%, 16.82%, 23.88%, and 31.44% in Q1 through Q4, respectively (log-rank P < 0.001). Similar graded patterns were observed for secondary outcomes. Specifically, 32,147 participants developed CHD (cumulative incidence: 11.21%), with rates of 4.82%, 9.28%, 13.04%, and 17.68% in Q1 through Q4, and 8,263 developed stroke (cumulative incidence: 2.88%), with rates of 1.17%, 2.16%, 3.45%, and 4.74% in Q1 through Q4. KM curves for cumulative incidence of CVD, CHD, and stroke by TyG-KDM-BA quartiles. Note A,D CVD; B, E CHD; C, F Stroke. Abbreviations: CVD, cardiovascular disease; CHD, coronary heart disease; TyG-KDM-BA, triglyceride-glucose index combined with biological age; CHARLS, China Health and Retirement Longitudinal Study In CHARLS, the median follow-up duration was 7.84 years (IQR: 5.70–9.98). A total of 1,610 participants developed CVD (cumulative incidence: 22.25%), including 1,290 cases of CHD (17.83%) and 470 cases of stroke (6.49%). Kaplan–Meier analyses similarly indicated increasing risks across TyG-KDM-BA quartiles (log-rank P < 0.001). Although fluctuations were observed in the stroke survival curves, likely due to the relatively small number of events, the overall risk stratification pattern was consistent with that observed in UKB. Association of TyG-KDM-BA with new-onset CVD risk Four Cox proportional hazards models were applied to assess the association between TyG-KDM-BA and the incidence of CVD, CHD, and stroke among participants with CKM stages 0–3. In UKB (Table 2), higher TyG-KDM-BA levels were consistently associated with increased risks of CVD, CHD, and stroke in unadjusted analyses, and these associations remained robust after sequential adjustment for demographic characteristics, lifestyle factors, and clinical variables. In the fully adjusted model, each one–standard deviation increase in TyG-KDM-BA was associated with a 65% higher risk of incident CVD (HR: 1.65; 95% CI: 1.64–1.67), a 65% higher risk of CHD (HR: 1.65; 95% CI: 1.63–1.67), and a 72% higher risk of stroke (HR: 1.72; 95% CI: 1.68–1.76). Quartile-based analyses further demonstrated a clear dose–response relationship, with participants in the highest TyG-KDM-BA quartile exhibiting markedly elevated risks compared with those in the lowest quartile (P for trend < 0.001): HRs were 3.97 (95% CI: 3.86–4.09) for CVD, 3.88(95% CI: 3.74–4.03) for CHD, and 4.17 (95% CI: 3.86–4.50) for stroke. Table 2Associations between TyG-KDM-BA and CVD, CHD, and stroke in participants of UKB with CKM syndrome stages 0–3Model 1Model 2Model 3Model 4HR(95%CI)P ValueHR(95%CI)P ValueHR(95%CI)P ValueHR(95%CI)P ValueCVD ~ TyG-KDM-BA (Per SD)1.71(1.69–1.72)< 0.0011.69(1.68–1.71)< 0.0011.66(1.65–1.68)< 0.0011.65(1.64–1.67)< 0.001CVD ~ TyG-KDM-BA (Q1-Q4) Q1ReferenceReferenceReferenceReference Q22.05 (1.99–2.11)< 0.0012.06 (1.99–2.12)< 0.0012.04 (1.98–2.10)< 0.0012.03 (1.97–2.09)< 0.001 Q33.05 (2.96–3.13)< 0.0013.06 (2.98–3.15)< 0.0013.03 (2.95–3.12)< 0.0013.01 (2.93–3.11)< 0.001 Q44.23 (4.11–4.35)< 0.0014.19 (4.07–4.31)< 0.0014.02 (3.91–4.14)< 0.0013.97 (3.86–4.09)< 0.001 P for trend< 0.001< 0.001< 0.001< 0.001CHD ~ TyG-KDM-BA (Per SD)1.71 (1.69–1.73)< 0.0011.69 (1.67–1.71)< 0.0011.66 (1.64–1.68)< 0.0011.65 (1.63–1.67)< 0.001CHD ~ TyG-KDM-BA (Q1-Q4) Q1ReferenceReferenceReferenceReference Q22.00 (1.92–2.08)< 0.0012.00 (1.92–2.08)< 0.0011.99 (1.91–2.07)< 0.0011.99 (1.91–2.07)< 0.001 Q32.91 (2.80–3.03)< 0.0012.93 (2.82–3.05)< 0.0012.91 (2.80–3.03)< 0.0012.90 (2.78–3.01)< 0.001 Q44.15 (4.00–4.31)< 0.0014.10 (3.95–4.26)< 0.0013.94 (3.79–4.09)< 0.0013.88 (3.74–4.03)< 0.001 P for trend< 0.001< 0.001< 0.001< 0.001Stroke ~ TyG-KDM-BA (Per SD)1.76 (1.72–1.81)< 0.0011.75 (1.71–1.80)< 0.0011.73 (1.69–1.77)< 0.0011.72 (1.68–1.76)< 0.001Stroke ~ TyG-KDM-BA (Q1-Q4) Q1ReferenceReferenceReferenceReference Q21.89 (1.74–2.05)< 0.0011.89 (1.74–2.05)< 0.0011.87 (1.72–2.04)< 0.0011.88 (1.73–2.04)< 0.001 Q33.09 (2.86–3.33)< 0.0013.09 (2.86–3.35)< 0.0013.05 (2.82–3.30)< 0.0013.05 (2.82–3.30)< 0.001 Q44.38 (4.06–4.73)< 0.0014.34 (4.03–4.68)< 0.0014.20 (3.90–4.53)< 0.0014.17 (3.86–4.50)< 0.001 P for trend< 0.001< 0.001< 0.001< 0.001Model Adjustments in UKB:Model 1: an unadjusted modelModel 2: adjusted sexModel 3: adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, and drinking statusModel 4: fully adjusted model, further adjusted for insomnia or sleep difficulties, cancer, moderate-to-vigorous physical activity (MVPA), educational level, and the Townsend deprivation index at recruitment, in addition to the covariates included in Model 3TyG-KDM-BA, triglyceride–glucose index combined with biological age; CVD, cardiovascular disease; CHD, coronary heart disease; CKM, Cardiovascular-Kidney-Metabolic; CI, confidence interval; SD, standard deviation Associations between TyG-KDM-BA and CVD, CHD, and stroke in participants of UKB with CKM syndrome stages 0–3 Model Adjustments in UKB: Model 1: an unadjusted model Model 2: adjusted sex Model 3: adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, and drinking status Model 4: fully adjusted model, further adjusted for insomnia or sleep difficulties, cancer, moderate-to-vigorous physical activity (MVPA), educational level, and the Townsend deprivation index at recruitment, in addition to the covariates included in Model 3 TyG-KDM-BA, triglyceride–glucose index combined with biological age; CVD, cardiovascular disease; CHD, coronary heart disease; CKM, Cardiovascular-Kidney-Metabolic; CI, confidence interval; SD, standard deviation In CHARLS (Table 3), TyG-KDM-BA showed consistent associations with cardiovascular outcomes. After full adjustment, each standard deviation increase in TyG-KDM-BA was associated with higher risks of CVD (HR: 1.25; 95% CI: 1.18–1.32), CHD (HR: 1.23; 95% CI: 1.16–1.31), and stroke (HR: 1.29; 95% CI: 1.17–1.43). Similarly, quartile analyses revealed a graded increase in risk across TyG-KDM-BA categories (P for trend < 0.001), with the highest quartile showing significantly elevated hazards for CVD (HR: 1.92; 95% CI: 1.63–2.25), CHD (HR: 1.92; 95% CI: 1.60–2.30), and stroke (HR: 1.84; 95% CI: 1.35–2.50), compared with the lowest quartile. Table 3Associations between TyG-KDM-BA and CVD, CHD, and stroke in participants of CHARLS with CKM syndrome stages 0–3Model 1Model 2Model 3Model 4HR (95%CI)P ValueHR (95%CI)P ValueHR(95%CI)P ValueHR (95%CI)P ValueCVD ~ TyG-KDM-BA (Per SD)1.34 (1.27–1.40)< 0.0011.34 (1.28–1.41)< 0.0011.33 (1.26–1.39)< 0.0011.25 (1.18–1.32)< 0.001CVD ~ TyG-KDM-BA (Q1–Q4) Q1ReferenceReferenceReferenceReference Q21.43 (1.22–1.67)< 0.0011.45 (1.24–1.70)< 0.0011.43 (1.22–1.67)< 0.0011.36 (1.16–1.59)< 0.001 Q31.89 (1.63–2.19)< 0.0011.93 (1.67–2.24)< 0.0011.88 (1.62–2.18)< 0.0011.72 (1.48–2.01)< 0.001 Q42.28 (1.97–2.65)< 0.0012.33 (2.01–2.71)< 0.0012.24 (1.93–2.60)< 0.0011.92 (1.63–2.25)< 0.001 P for trend< 0.001< 0.001< 0.001< 0.001CHD ~ TyG-KDM-BA (Per SD)1.29 (1.22–1.36)< 0.0011.29 (1.23–1.37)< 0.0011.28 (1.21–1.35)< 0.0011.23 (1.16–1.31)< 0.001CHD ~ TyG-KDM-BA (Q1–Q4) Q1ReferenceReferenceReferenceReference Q21.39 (1.17–1.65)< 0.0011.43 (1.21–1.70)< 0.0011.41 (1.19–1.68)< 0.0011.38 (1.16–1.65)< 0.001 Q31.73 (1.47–2.04)< 0.0011.80 (1.53–2.12)< 0.0011.76 (1.49–2.08)< 0.0011.69 (1.42–2.00)< 0.001 Q42.09 (1.76–2.46)< 0.0012.16 (1.83–2.54)< 0.0012.10 (1.78–2.47)< 0.0011.92 (1.60–2.30)< 0.001 P for trend< 0.001< 0.001< 0.001< 0.001Stroke ~ TyG-KDM-BA (Per SD)1.52 (1.39–1.66)< 0.0011.51 (1.38–1.66)< 0.0011.49 (1.36–1.63)< 0.0011.29 (1.17–1.43)< 0.001Stroke ~ TyG-KDM-BA (Q1–Q4) Q1ReferenceReferenceReferenceReference Q21.57 (1.15–2.14)0.0041.55 (1.14–2.12)0.0051.51 (1.11–2.06)0.0091.29 (0.94–1.76)0.113 Q32.47 (1.85–3.29)< 0.0012.43 (1.82–3.24)< 0.0012.33 (1.75–3.12)< 0.0011.79 (1.33–2.41)< 0.001 Q42.98 (2.23–3.97)< 0.0012.94 (2.20–3.93)< 0.0012.78 (2.08–3.72)< 0.0011.84 (1.35–2.50)< 0.001 P for trend< 0.001< 0.001< 0.001< 0.001Model Adjustments in CHARLS:Model 1: an unadjusted modelModel 2: adjusted sexModel 3: adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, and drinking statusModel 4: fully adjusted model, further adjusted for insomnia or sleep difficulties, cancer, moderate-to-vigorous physical activity (MVPA) and educational level, in addition to the covariates included in Model 3TyG-KDM-BA, triglyceride–glucose index combined with biological age; CVD, cardiovascular disease; CHD, coronary heart disease; CKM, Cardiovascular-Kidney-Metabolic; CI, confidence interval; SD, standard deviation; CHARLS, China Health and Retirement Longitudinal Study Associations between TyG-KDM-BA and CVD, CHD, and stroke in participants of CHARLS with CKM syndrome stages 0–3 Model Adjustments in CHARLS: Model 1: an unadjusted model Model 2: adjusted sex Model 3: adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, and drinking status Model 4: fully adjusted model, further adjusted for insomnia or sleep difficulties, cancer, moderate-to-vigorous physical activity (MVPA) and educational level, in addition to the covariates included in Model 3 TyG-KDM-BA, triglyceride–glucose index combined with biological age; CVD, cardiovascular disease; CHD, coronary heart disease; CKM, Cardiovascular-Kidney-Metabolic; CI, confidence interval; SD, standard deviation; CHARLS, China Health and Retirement Longitudinal Study The PH assumption was tested using Schoenfeld residuals, which revealed a violation for the TyG-KDM-BA index in UKB (P < 0.001), whereas no significant violations were observed in CHARLS and hospital cohorts (all P > 0.05). In the stratified Cox models across follow-up time intervals, TyG-KDM-BA remained significantly associated with CVD risk (HR: 1.19, 95% CI: 1.17–1.20, P < 0.001). In the time-segmented analysis (Table S4), TyG-KDM-BA was consistently associated with increased CVD risk across all follow-up periods, with HRs ranging from 1.59 to 1.71. These results indicate that, despite violation of the PH assumption, the association between TyG-KDM-BA and CVD risk remained stable over time. Multicollinearity testing of covariates across both cohorts revealed no significant multicollinearity issues in the model variables (Tables S5 and S6). Nonlinear association between TyG-KDM-BA and CVD risk We conducted RCS modeling in UKB and CHARLS to systematically evaluate the shape of the association between TyG-KDM-BA and cardiovascular outcomes. As shown in Fig. 3, TyG-KDM-BA demonstrated significant nonlinear dose–response relationships with the risks of incident CVD, CHD, and stroke in individuals with CKM stages 0–3 (P < 0.05). RSC and two-piece Cox models for TyG-KDM-BA and risk of CVD, CHD, and stroke. Note A, D, G, J. for overall CVD; B, E, H, K. for CHD; C, F, I, L. for stroke. A–F for UKB, model fully adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, drinking status, insomnia or sleep difficulties, cancer, MVPA, educational level, and the Townsend deprivation index at recruitment. I–L for CHARLS, model fully adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, drinking status, insomnia or sleep difficulties, cancer, MVPA, and educational level. Abbreviations: RCS, Restricted cubic spline; CVD, cardiovascular disease; CHD, coronary heart disease; TyG-KDM-BA, triglyceride-glucose index combined with biological age; BMI, body mass index; MVPA, moderate-to-vigorous physical activity; CHARLS, China Health and Retirement Longitudinal Study Based on risk inflection points identified via RCS curves, two-stage Cox regression models were subsequently applied to assess potential threshold effects. In UKB cohort, when TyG-KDM-BA values were below the identified threshold of 517.82, each unit increase was associated with a significant rise in CVD risk (HR: 1.009). Above this threshold, the association persisted but with an attenuated slope (HR: 1.004). A similar pattern was observed in CHARLS, with a threshold of 501.59: TyG-KDM-BA was strongly associated with CVD risk below the threshold (HR: 1.005, whereas the increase in risk plateaued at higher levels (HR: 1.001). Comparable nonlinear trends and threshold characteristics were also observed for CHD and stroke, suggesting a consistent pattern across cardiovascular outcomes. ROC curve evaluation Time-dependent ROC analyses were performed in UKB cohort (Fig. 4). The results demonstrated robust cardiovascular disease predictive capability of the TyG-KDM-BA index over 5, 10, and 15 years (AUC: 0.659, 0.657, 0.657), with comparable performance for CHD (AUC: 0.663, 0.651, 0.643) and stroke (AUC: 0.642, 0.642, 0.645). Notably, within the five-year prediction window, the TyG-KDM-BA index consistently outperformed other TyG-derived indicators (TyG: 0.546, TyG-BMI: 0.569, TyG-WC: 0.604, TyG-WHtR: 0.596). Furthermore, its AUC values remained stable as the prediction window extended to 15 years, indicating good temporal stability in long-term risk discrimination. ROC curve comparison of different insulin resistance indices for predicting 5, 10, and 15 year risks of CVD, CHD, and stroke in UKB. Note A–C: CVD; D–F: CHD; G–I: stroke. 5, 10, and 15 year risks are shown from left to right. Abbreviations: ROC, receiver operating characteristic; AUC, the area under the curve; CVD, cardiovascular disease; CHD, coronary heart disease; TyG, triglyceride–glucose index; TyG-KDM-BA, triglyceride-glucose index combined with biological age; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio In CAHRLS, the TyG-KDM-BA index demonstrated moderate but consistent predictive ability for CVD, CHD, and stroke across 3, 5, and 8 year follow-up windows (Fig. 5). TyG-KDM-BA index over 3, 5, and 8 years (AUC: 0.567, 0.588, 0.569), with comparable performance for CHD (AUC: 0.559, 0.581, 0.565) and stroke (AUC: 0.609, 0.620, 0.620). Predictive performance of the TyG-KDM-BA index was comparable to or superior to other indicators across different prediction windows. The index achieved its highest discriminative ability at the 5 year time point, exceeding that of conventional TyG-related measures (TyG: 0.539, TyG-BMI: 0.559, TyG-WC: 0.560, TyG-WHtR: 0.565). Although slightly lower performance was observed at the 3 year and 8 year time points, the overall trend suggested consistent and stable predictive capability across time horizons. ROC curve comparison of different insulin resistance indices for predicting 3, 5, and 8 year risks of CVD, CHD, and stroke in CHARLS. Note A–C: CVD; D–F: CHD; G–I: stroke. 3-, 5-, and 8- risks are shown from left to right. Abbreviations: ROC, receiver operating characteristic; AUC, the area under the curve; CVD, cardiovascular disease; CHD, coronary heart disease; TyG, triglyceride–glucose index; TyG-KDM-BA, triglyceride-glucose index combined with biological age; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; CHARLS, China Health and Retirement Longitudinal Study Overall, TyG-KDM-BA demonstrated consistent and time-stable discriminatory performance across both cohorts and multiple cardiovascular outcomes. Incremental predictive value To further evaluate incremental predictive value, additional analyses using ΔC-index, NRI, and IDI were performed in UKB and CHARLS (Figure S1-S2). The addition of TyG-KDM-BA improved predictive performance compared with multiple alternative models. In UKB, TyG-KDM-BA demonstrated significant incremental predictive value compared with clinical covariates (ΔC-index = 0.0653, IDI = 0.0151, NRI = 0.1957, all P < 0.001), TyG alone (ΔC-index = 0.1305, IDI = 0.0224, NRI = 0.3699; all P < 0.001), KDM-BA alone (ΔC-index = 0.0270, IDI = 0.0075, NRI = 0.0622; all P < 0.001). It also remained superior to the model including TyG and KDM-BA as separate covariates (ΔC-index = 0.0236, IDI = 0.0067, NRI = 0.0542; all P < 0.001). Consistent results were observed in CHARLS. Subgroup analysis Stratified subgroup analyses were conducted in UKB and CHARLS according to age, gender, BMI, hypertension, diabetes, metabolic syndrome, smoking status, alcohol consumption, and CKM stage (Fig. 6). Across most subgroups, TyG-KDM-BA showed directionally consistent positive associations with the risks of CVD, CHD, and stroke. In general, each SD increase in TyG-KDM-BA was associated with an HR greater than 1 for CVD, supporting the robustness of its risk association across diverse population strata. Forest plot of subgroup analysis for the association between TyG-KDM-BA and CVD. Abbreviations: CVD, cardiovascular disease; TyG-KDM-BA, triglyceride–glucose index combined with biological age; BMI, body mass index; CKM, Cardiovascular-Kidney-Metabolic; CHARLS, China Health and Retirement Longitudinal Study In UKB, significant effect modification was observed for all subgroups for CVD (P for interaction < 0.001). Specifically, the association between TyG-KDM-BA and CVD risk was stronger among younger individuals (< 60 years), including < 45 years (HR: 2.18) and 45–60 years (HR: 1.59), than among those aged ≥ 60 years (HR: 1.34). Also, the associations are stronger in women (HR: 1.77) than in men (HR: 1.62), and more pronounced in participants without hypertension, diabetes, or metabolic syndrome. Further detailed analysis of interaction terms at different levels within each subgroup variable revealed (Tables S7). Significant interactions were observed for current smoking and current drinking(P for interaction < 0.05), whereas no significant interactions were found for former smoking or former drinking. For CHD and stroke outcomes (Figures S3 and S4), subgroup-specific association patterns were generally consistent with those observed for CVD. In CHARLS, the overall direction of the association between TyG-KDM-BA and CVD risk was consistent with that in UKB. However, statistical significance was not reached in several subgroups, particularly among older adults, participants with hypertension or diabetes, and those with CKM stage 3. Nevertheless, heterogeneous trends across age, sex, hypertension, and diabetes status remained generally comparable to those in UKB (Fig. 6, Table S8). Notably, in CKM syndrome stage–specific analyses within UKB, the association was strongest in CKM stage 1 (HR: 2.00; 95% CI: 1.94–2.07) and weakest in stage 3 (HR: 1.16; 95% CI: 1.13–1.19), with both interaction effects reaching statistical significance (P < 0.001). In CHARLS, a similar attenuation trend was observed in advanced stages, but the interaction effect was not statistically significant. Further examination of TyG-KDM-BA distributions across CKM stages revealed a marked reduction in inter-individual variability among participants with stage 3, accompanied by diminished linear and monotonic associations with CVD events (Table S9; Figure S5), potentially limiting its discriminatory capacity in this subgroup. Sensitivity analysis A series of sensitivity analyses was conducted to evaluate the robustness of the primary findings. Overall, the results remained materially unchanged across all analyses (Figure S6–S7), supporting the robustness of the observed associations. These included additional adjustments for medication use, multiple imputation for missing covariates, exclusion of events occurring within the first two years, and retention of extreme values, which yielded results consistent with the primary models. Recalculation of KDM-BA using the 7 biomarkers available in CHARLS yielded highly consistent results compared with the original 9-biomarker model. The distributions of KDM-BA (57.84 vs 57.80) and TyG-KDM-BA (509.04 vs 508.52) were nearly identical, with high correlations (r = 0.987 and r = 0.989, respectively). Consistently, the associations between TyG-KDM-BA and cardiovascular outcomes were largely unchanged, with only small variations in HRs across CVD (1.68 vs 1.65), CHD (1.68 vs 1.65), and stroke (1.78 vs 1.72) (Table S10). These findings indicate that differences between the two KDM-BA calculation approaches have a negligible impact on both the exposure distribution and its association with outcomes, supporting the robustness and cross-cohort comparability of the TyG-KDM-BA index. In competing risk analyses using Fine-Gray models, the associations between TyG-KDM-BA and CVD, CHD, and stroke remained largely consistent with the primary Cox models. The subdistribution hazard ratios (sHRs) were 1.66 (95% CI: 1.64–1.67) for CVD, 1.62 (95% CI: 1.60–1.64) for CHD, and 1.68 (95% CI: 1.64–1.72) for stroke, which were very similar to the corresponding Cox estimates. These results suggest that competing risk bias is unlikely to materially influence the observed associations (Table S11). When restricting the outcome to non-fatal CVD events in UKB, the association remained robust (HR: 1.63 vs 1.65 in the primary analysis) (Table S12). However, the effect estimate remained higher than that observed in CHARLS (HR: 1.25), and substantial heterogeneity was detected between cohorts (I2 = 98.8%, P < 0.001), although the direction of association was consistent. External validation External validation using an independent hospital-based cohort yielded results broadly concordant with the main Cox regression analyses. In the independent hospital-based cohort (Table 4), TyG-KDM-BA was significantly associated with increased risks of CVD (HR: 1.44; 95% CI: 1.20–1.74), CHD (HR: 1.40; 95% CI: 1.15–1.69), and stroke (HR: 1.85; 95% CI: 1.08–3.18) (all P < 0.05) in the fully adjusted Cox regression model. Quartile-based analyses further supported these findings. Participants in the highest quartile of TyG-KDM-BA exhibited a significantly increased risk compared with those in the lowest quartile, whereas the intermediate quartiles showed non-significant and relatively similar risk estimates. Importantly, a significant dose–response relationship was observed across quartiles in all models (all P for trend < 0.05). Overall, these findings from the external cohort corroborate the primary analyses and provide additional evidence for the robustness and generalizability of the association between TyG-KDM-BA and cardiovascular outcomes. Table 4Associations between TyG-KDM-BA and incident CVD, CHD, and stroke in the external validation cohortModel 1Model 2Model 3Model 4HR (95%CI)P ValueHR (95%CI)P ValueHR (95%CI)P ValueHR (95%CI)P ValueCVD ~ TyG-KDM-BA(Per SD)1.37 (1.15–1.64) < 0.0011.46 (1.21–1.75)< 0.0011.43 (1.19–1.71)< 0.0011.44 (1.20–1.74)< 0.001CVD ~ TyG-KDM-BA (Q1–Q4) Q1ReferenceReferenceReferenceReference Q21.41 (0.85–2.32)0.1811.56 (0.94–2.59)0.0831.64 (0.98–2.74)0.0611.62 (0.97–2.72)0.067 Q31.40 (0.84–2.33)0.1931.54 (0.93–2.58)0.0961.59 (0.95–2.67)0.0751.59 (0.95–2.67)0.079 Q41.89 (1.17–3.04)0.0092.16 (1.32–3.52)0.0022.03 (1.24–3.33)0.0052.03 (1.23–3.34)0.006 P for trend0.0130.0030.0080.008CHD ~ TyG-KDM-BA(Per SD)1.33 (1.11–1.61)0.0031.41 (1.17–1.71)< 0.0011.39 (1.15–1.67)< 0.0011.40 (1.15–1.69)< 0.001CHD ~ TyG-KDM-BA (Q1–Q4) Q1ReferenceReferenceReferenceReference Q21.46 (0.87–2.43)0.1481.62 (0.96–2.72)0.0681.70 (1.00–2.89)0.0491.67 (0.98–2.83)0.059 Q31.34 (0.79–2.27)0.2821.47 (0.86–2.52)0.1561.52 (0.89–2.60)0.1261.49 (0.87–2.56)0.146 Q41.76 (1.06–2.90)0.0282.00 (1.20–3.34)0.0081.89 (1.12–3.18)0.0171.86 (1.10–3.14)0.021 P for trend0.0450.0150.0290.036Stroke ~ TyG-KDM-BA(Per SD)1.78 (1.02–3.10)0.0421.86 (1.05–3.29)0.0331.83 (1.03–3.25)0.0411.85 (1.08–3.18)0.025Stroke ~ TyG-KDM-BA (Q1–Q4) Q1ReferenceReferenceReferenceReference Q20.90 (0.15–5.40)0.9080.95 (0.16–5.81)0.9600.91 (0.14–5.73)0.9181.11 (0.18–7.04)0.910 Q32.04 (0.46–9.14)0.3512.16 (0.47–9.82)0.3202.15 (0.47–9.86)0.3232.57 (0.55–12.09)0.233 Q43.51 (0.91–13.61)0.0693.80 (0.95–15.24)0.0603.50 (0.84–14.46)0.0844.28 (1.02–17.95)0.047 P for trend0.0380.0320.0430.024Model Adjustments in the external validation cohort:Model 1: an unadjusted modelModel 2: adjusted sexModel 3: adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, and drinking statusModel 4: fully adjusted model, further adjusted for cancer and educational level, in addition to the covariates included in Model 3TyG-KDM-BA, triglyceride–glucose index combined with biological age; CVD, cardiovascular disease; CHD, coronary heart disease; CKM, Cardiovascular-Kidney-Metabolic; CI, confidence interval; SD, standard deviation Associations between TyG-KDM-BA and incident CVD, CHD, and stroke in the external validation cohort CVD ~ TyG-KDM-BA (Per SD) CHD ~ TyG-KDM-BA (Per SD) Stroke ~ TyG-KDM-BA (Per SD) Model Adjustments in the external validation cohort: Model 1: an unadjusted model Model 2: adjusted sex Model 3: adjusted for sex, BMI, hemoglobin, red blood cell count, white blood cell count, smoking status, and drinking status Model 4: fully adjusted model, further adjusted for cancer and educational level, in addition to the covariates included in Model 3 TyG-KDM-BA, triglyceride–glucose index combined with biological age; CVD, cardiovascular disease; CHD, coronary heart disease; CKM, Cardiovascular-Kidney-Metabolic; CI, confidence interval; SD, standard deviation Discussion This study primarily utilized two large prospective cohorts, UKB and CHARLS, and incorporated a real-world hospital-based cohort for external validation. It evaluates a novel composite indicator, TyG-KDM-BA, which integrates insulin resistance and biological aging for the first time. TyG-KDM-BA was strongly associated with increased risks of CVD, CHD, and stroke in CKM stages 0–3 and remained robust after multivariable adjustment and across multiple sensitivity analyses. RCS analyses revealed a nonlinear association between TyG-KDM-BA and cardiovascular risk. Compared with the conventional TyG-related indices, TyG-KDM-BA demonstrated improved risk discrimination and provided significant incremental predictive value beyond both individual components and conventional model specifications. Subgroup analyses demonstrated heterogeneity by age and CKM stage and stronger associations among individuals without overt cardiometabolic comorbidities. These findings were further supported by sensitivity analyses and external validation, underscoring the robustness of the results. IR is widely recognized as a central upstream driver of cardiometabolic disease, particularly in the early stages of CKM syndrome [8]. The TyG index, a validated surrogate marker of insulin resistance, has demonstrated predictive value for cardiovascular outcomes in populations with early cardiometabolic dysfunction [43]. However, most TyG-derived indicators primarily capture metabolic burden and adiposity-related phenotypes, with limited consideration of physiological aging. Given that aging represents a fundamental biological substrate underlying cardiovascular disease, integrating metabolic dysfunction with biological aging may provide a more comprehensive assessment of cardiometabolic risk [46]. Therefore, we developed the TyG-KDM-BA index to better capture the interplay between metabolic dysregulation and systemic aging. Extensive previous studies have consistently demonstrated that TyG and its obesity-related derivatives are associated with adverse cardiovascular outcomes. For instance, Liu et al. reported that the highest quartile of TyG and obesity-composite indices, including TyG-BMI, TyG-WC, and TyG-WHtR, significantly increased the risk of CVD and death (HR ranges of 1.30–1.58) [4]. Similarly, Dang et al. reported heterogeneous predictive performance of TyG-related indicators across cardiovascular outcomes, with TyG-WHtR showing the strongest association with myocardial infarction and TyG-WC demonstrating stronger associations with angina pectoris, heart failure, and overall CVD [12]. However, these markers primarily reflect metabolic burden, which generally demonstrate modest risk estimates (HR approximately 1.2–1.6) [47, 48]. In contrast, TyG-KDM-BA exhibited a larger magnitude of association with CVD and secondary outcomes in our study. This benefit might result from TyG-KDM-BA better integration of biological aging data and glucose-lipid metabolism, which more fully reflects the pathophysiological mechanisms that connect metabolic diseases to aging. The consistency of results across sensitivity analyses and external validation further supports the robustness and potential clinical utility of TyG-KDM-BA as a cardiovascular risk predictor for individuals with early CKM syndrome. Notably, weaker effect estimates were observed in CHARLS. This attenuation should be interpreted with caution and is likely attributable, at least in part, to differences in outcome ascertainment between cohorts, as CVD events in UKB were identified through registry-based records, including both fatal and non-fatal events, whereas CHARLS relied on self-reported physician diagnoses capturing predominantly non-fatal events. In addition, differences in population characteristics, including older age structure, shorter follow-up duration, and a potentially higher burden of competing risks, may have further contributed to the observed attenuation. We observed a nonlinear association between TyG-KDM-BA and CVD risk in both UKB and CHARLS cohorts. Similar nonlinear patterns have been reported in previous studies. Miao et al. found a J-shaped association between TyG and both cardiovascular mortality and all-cause mortality in patients with heart failure [49], while the large-scale community cohort CHERRY study found an inverted L-shaped nonlinear association between the TyG index and CVD risk, with a pronounced increase beyond a threshold level [50]. In contrast, we found that CVD risk increased steeply at lower levels and gradually plateaued beyond the reference point. This pattern suggests that TyG-KDM-BA may capture a critical transition phase in cardiometabolic deterioration, providing stronger early warning signals within the low-to-moderate range. Early elevations may reflect endothelial dysfunction, chronic low-grade inflammation, and the accelerated accumulation of insulin resistance [51]. Although compensatory mechanisms may still partially buffer cardiovascular damage, incremental metabolic deterioration could exert a disproportionately greater impact on cardiovascular risk, resulting in the steep portion of the curve. Once a threshold is exceeded, cardiovascular injury may already be substantial, and individuals may have entered a state of advanced metabolic disturbance. Under such conditions, further increases in the index may contribute relatively less additional risk, leading to a plateau or flattening of the curve [52]. These findings highlight the potential value of identifying threshold ranges of TyG-KDM-BA to facilitate earlier detection of high-risk individuals. ROC analysis demonstrated that TyG-KDM-BA achieved superior discrimination performance compared with traditional TyG-related markers, particularly in UKB cohort. Although discrimination was attenuated in CHARLS, it consistently performed comparably to or better than other indices. Prior studies have reported only moderate discriminative capacity for traditional TyG-based markers (AUC approximately 0.55–0.60) [15, 53]. Importantly, TyG-KDM-BA provided significant incremental predictive value compared with clinical covariates alone, TyG or KDM-BA alone, and also models including TyG and KDM-BA as separate covariates, as evidenced by consistent improvements in ΔC-index, NRI, and IDI. These findings indicate that the composite index captures complementary information from metabolic dysfunction and biological aging that is not fully reflected when these components are considered independently. Thus, the finding further increases the potential use of TyG-KDM-BA as a thorough risk stratification tool and broadens the range of applications for TyG-based markers in cardiovascular risk prediction. Notably, chronological age was not included as a covariate in the primary Cox models because TyG-KDM-BA intrinsically incorporates biological age, which is highly correlated with chronological age and reflects age-related physiological decline. Multicollinearity diagnostics confirmed substantial collinearity between TyG-KDM-BA and chronological age, which could compromise model stability and interpretability. Therefore, rather than adjusting for age, we evaluated potential effect modification through stratified and interaction analyses. Age-stratified analyses revealed marked heterogeneity. In UKB cohort, stronger associations with CVD and CHD were observed among individuals < 60 years, whereas estimates were attenuated in those aged ≥ 60 years. Similar directional patterns were observed in CHARLS, although statistical significance was reduced, likely due to the older age structure and fewer events. This pattern is supported by prior studies. Wan et al. reported higher predictive performance of the TyG index for cardiovascular outcomes in younger populations [54], and a study conducted in Iranians also found a more pronounced association in younger individuals [55]. One possible explanation is that younger adults generally have a lower baseline absolute cardiovascular risk, making relative risk elevations associated with metabolic dysfunction more discernible. In contrast, among older individuals, metabolic abnormalities are more prevalent and age-related pathological processes are already widespread, which may reduce the discriminatory capacity of TyG-related markers. These findings suggest that TyG-KDM-BA may have greater prognostic value in younger and middle-aged adults. We further observed that the association between TyG-KDM-BA and cardiovascular risk was strongest in stages 0 and 1, and progressively weakened in stages 2 and 3 across both UKB and CHARLS. Results of the distribution and correlations showed that as CKM syndrome severity increased, TyG-KDM-BA values rose steadily, while inter-individual variability narrowed. Consistently, the correlation remained stable across stages 0–2 but markedly weakened in stage 3. Similar phenomena have been reported previously. For instance, Tan et al. observed that the inverse association between estimated glucose disposal rate (eGDR) and mortality was more pronounced in CKM stage 1 than in later stages in both NHANES and CHARLS cohorts [56]. Collectively, these findings suggest that TyG-KDM-BA may offer superior risk discrimination during early stages of CKM syndrome, facilitating the identification of high-risk individuals prior to irreversible metabolic dysfunction or organ damage. In advanced CKM stages, however, disease complexity and overlapping complications may attenuate its predictive utility, underscoring the need for complementary clinical or imaging-based markers. Beyond age and CKM stage, TyG-KDM-BA demonstrated heterogeneous predictive performance across demographic and clinical subgroups. Overall, stronger associations tended to appear among individuals with fewer overt cardiometabolic risk factors. Similar patterns have been reported previously. Hong et al. found attenuated associations among individuals with established hypertension [57]. Moreover, among acute coronary syndrome patients without standard modifiable cardiovascular risk factors, the TyG index remained predictive of adverse outcomes [58]. These findings suggest that TyG-KDM-BA may be particularly useful for identifying high-risk individuals who appear clinically low-risk based on conventional factors. Mechanistically, the TyG-KDM-BA index shows a synergistic interaction between metabolic dysfunction and systemic physiological decline, two major factors contributing to cardiometabolic vulnerability. Among individuals with CKM syndrome, insulin resistance contributes to cardiovascular injury through several interrelated biological pathways. First, impairment of the PI3K/Akt signaling pathway disrupts insulin-mediated nitric oxide production while enhancing MAPK pathway activity, resulting in endothelial dysfunction and early arterial stiffening [59]. Second, insulin resistance induces dysregulated glucose metabolism and subsequent hyperglycemia, which promotes chronic low-grade inflammation, thereby accelerating atherosclerotic processes [60, 61]. Third, increased reactive oxygen species generation and downregulation of key mitochondrial regulators, including PGC-1α and SIRT1, further exacerbate oxidative stress and mitochondrial dysfunction, impair cellular energy homeostasis, and promote vascular aging [62]. Meanwhile, age-related declines in mitochondrial function, accumulation of oxidative damage, endothelial senescence, and chronic systemic inflammation collectively contribute to myocardial fibrosis, vascular remodeling, and arterial stiffness, thereby increasing vulnerability to cardiometabolic injury [25, 63]. The convergence of IR-related metabolic stress and aging-associated vulnerability may amplify cardiovascular risk beyond the effect of either process alone. This study first proposes the TyG-KDM-BA composite index, integrating insulin resistance and biological aging information, to assess CVD, CHD, and stroke risks across CKM stages 0–3. Major strengths include the use of multiple prospective cohorts with large sample sizes, diverse populations, long follow-up durations, and rigorous outcome adjudication. The robustness of our findings was further supported by extensive sensitivity analyses and external validation using real-world hospital data. Several limitations should be acknowledged. First, differences in KDM-BA construction due to variations in biomarker availability may affect cross-cohort comparability. Although sensitivity analyses using a harmonized 7-biomarker model showed highly consistent results, KDM-BA remains based on predefined clinical biomarkers and may not fully capture molecular aging processes. Second, differences in outcome ascertainment between cohorts may limit direct comparability. Cardiovascular events were identified through registry-based records, including both fatal and non-fatal events in UKB, whereas CHARLS relied on self-reported physician diagnoses, capturing predominantly non-fatal events and potentially introducing recall and misclassification bias. Although sensitivity analyses restricted to non-fatal CVD in UKB showed minimal attenuation, cross-cohort comparisons should still be interpreted with caution. Third, the use of CKM staging may introduce incorporation bias, as stages 2–3 are defined based on metabolic abnormalities that overlap with components of the TyG-KDM-BA index, potentially leading to overestimation of associations in later stages. Fourth, competing risk analyses could not be performed in CHARLS due to the lack of cause-specific mortality data, which may introduce bias related to competing risks. Although a violation of the PH assumption was observed in UKB, time-stratified analyses demonstrated stable associations, suggesting limited impact on the overall conclusions. Fifth, the external validation cohort was hospital-based with a relatively small sample size and higher baseline risk, which may introduce selection bias and limit generalizability; thus, it should be considered as a validation rather than a population-representative sample. In addition, despite the established reproducibility of UKB, a healthy volunteer bias cannot be excluded [64]. Finally, residual confounding and reverse causality may persist due to the observational design, and the TyG-KDM-BA was assessed only at baseline; future studies should explore longitudinal changes or cumulative exposure. In conclusion, TyG-KDM-BA is a novel composite indicator that integrates insulin resistance and biological aging and demonstrates robust predictive value for incident cardiovascular disease among individuals with CKM stages 0–3, particularly in the early disease phase. Compared with traditional TyG-related indices, it more effectively identifies individuals at elevated risk despite relatively favorable conventional risk profiles. The observed threshold and saturation patterns suggest that the predictive utility of TyG-KDM-BA may vary across different CKM stages, underscoring the importance of integrating disease staging in clinical applications. Future studies should validate this indicator in more diverse populations, explore its potential value in guiding personalized interventions, and elucidate its underlying biological mechanisms in CKM disease progression. Conclusion This study first constructs a novel composite indicator, TyG-KDM-BA, by integrating TyG with KDM-BA, and evaluates its association with incident cardiovascular disease among individuals with CKM stages 0–3. Across multiple prospective cohorts, higher TyG-KDM-BA levels were consistently associated with increased CVD risk and demonstrated improved predictive performance compared with conventional TyG-based indices. TyG-KDM-BA also provided significant incremental predictive value beyond both its individual components and conventional model specifications. The association was more pronounced in the low-to-moderate range of TyG-KDM-BA, supporting its potential utility for early risk identification. The robustness of these findings was supported by sensitivity analyses and external validation. Overall, TyG-KDM-BA integrates metabolic dysfunction and biological aging into a single quantitative measure and may serve as a useful tool for cardiovascular risk stratification in early CKM populations. Further studies are warranted to validate its clinical applicability in broader populations and to elucidate the biological mechanisms underlying its association with cardiovascular risk. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors thank the participants and research teams of the UK Biobank and CHARLS for their valuable contributions. We also acknowledge Sun Yat-sen Memorial Hospital for supporting the hospital-based cohort data collection. Abbreviations A Body Shape Index American Heart Association Biological age Body mass index Body roundness index China Health and Retirement Longitudinal Study Chronic kidney disease Cardiovascular–kidney–metabolic Confidence interval Chinese visceral adiposity index Cardiovascular disease Diastolic blood pressure Estimated glomerular filtration rate Fasting plasma glucose Generalized variance inflation factor High-density lipoprotein cholesterol Hyperinsulinemic–euglycemic clamp Hazard ratio Insulin resistance Klemera–Doubal method Klemera–Doubal method-derived biological age Low-density lipoprotein cholesterol Metabolic syndrome Restricted cubic spline Receiver operating characteristic Standard deviation Triglycerides Triglyceride-glucose index Triglyceride-glucose-A Body Shape Index Triglyceride-glucose-body mass index Triglyceride-glucose-body roundness index Triglyceride-glucose-Chinese visceral adiposity index Triglyceride-glucose-Klemera–Doubal method biological age Triglyceride-glucose-waist circumference Triglyceride-glucose-waist-to-height ratio Triglyceride-glucose-weight-adjusted waist index UK Biobank Waist circumference Waist-to-height ratio Weight-adjusted waist index Author contributions YC, KH, and RS designed the research. PD, RS, SF, and ZJ conducted data analyses and drafted the manuscript. AA, JL, KH, XZ, YM, and YL contributed to interpreting the results and revising the manuscript. All authors approved the final version of the manuscript. YC is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Funding This study was supported by the National Key R&D Program of China (2024YFF0507300), Science and Technology Program of Guangzhou City (2023B01J1011), The National Natural Science Foundation of China (U23A20397, 82271609, 81800420), Guangdong Basic and Applied Basic Research Foundation (2023B1515120088), and Sun Yat-sen Pilot Scientific Research Fund (YXQH202625). Data availability Data supporting this study's results are available from the official websites of the UK Biobank (https://www.ukbiobank.ac.uk/), China Health and Retirement Longitudinal Study (http://charls.pku.edu.cn). The hospital-based cohort data are not publicly available due to privacy and ethical restrictions. Access to these data is subject to approval by Sun Yat-sen Memorial Hospital. Declarations Ethics approval and consent to participate UKB was approved by the Northwest Multicenter Research Ethics Committee (Ref: 21/NW/0157), and all participants provided written informed consent. CHARLS was approved by the Biomedical Ethics Committee of Peking University (IRB00001052-11015), and written informed consent was obtained from all participants. The hospital-based cohort study was approved by the Ethics Committee of Sun Yat-sen Memorial Hospital. The requirement for informed consent was waived due to the retrospective nature of the study and the use of anonymized data. Consent for publication Not applicable. 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. Runnan Shen, Pinrong Dong, Shuxiang Fang and Zirui Jia contributed equally as co-first authors. Contributor Information Kai Huang, Email: huangk37@mail.sysu.edu.cn. Yangxin Chen, Email: chenyx39@mail.sysu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Data supporting this study's results are available from the official websites of the UK Biobank (https://www.ukbiobank.ac.uk/), China Health and Retirement Longitudinal Study (http://charls.pku.edu.cn). The hospital-based cohort data are not publicly available due to privacy and ethical restrictions. Access to these data is subject to approval by Sun Yat-sen Memorial Hospital.