Cardiovascular-liver-metabolic multimorbidity: preliminary insights into the associations with the triglyceride glucose index (TyG), its obesity-derived indices and their effects on long-term survival Authors: Shiming He, Yanfeng Liu, Guotai Sheng, Guobo Xie, Wei Wang, Hengli Lai, Yang Zou, Chao Wang Published: 2026 ABSTRACT AbstractBackgroundThe triglyceride-glucose (TyG) and its obesity‑derived indices are important composite parameters of insulin resistance coupled with obesity, and have been associated with cardiometabolic diseases. However, their associations with cardiovascular‑liver‑metabolic multimorbidity (CLMM) remain unclear. This study aims to systematically compare TyG and its obesity-derived indices to identify the optimal surrogate marker for CLMM, and to explore the potential impact of CLMM status on mortality.MethodsBased on a multi-stage sampling design of a national cohort, weighted logistic and Cox regression models were applied to evaluate the associations of each index with CLMM and long-term mortality. Discriminatory performance was compared using the area under the receiver operating characteristic curve (AUROC), net reclassification improvement, and integrated discrimination improvement. Additionally, weighted quantile sum regression, and mediation analysis were employed to explore the underlying disease mechanisms.ResultsThe study included 20,235 adults. Weighted logistic regression analyses for the individual indices revealed that the TyG and its obesity‑derived indices were all positively associated with CLMM, with TyG‑WHtR, TyG‑WWI, and TyG‑CVAI showing the strongest associations. Further longitudinal weighted interaction analyses revealed that CLMM status significantly amplified the mortality risk associated with the TyG and its obesity-derived indices. For CLMM identification, all indices showed favorable diagnostic performance (AUROC > 0.76), with TyG‑WWI and TyG‑CVAI achieving the best performance (AUROC > 0.83) and exhibiting the largest net reclassification improvement and integrated discrimination improvement. Joint XGBoost-SHAP analyses further indicated that TyG-CVAI and TyG-WWI were the major contributing variables for discriminating CLMM status. Exploratory mechanistic analyses suggested glucose metabolism as a core driver underlying CLMM development, with inflammatory factors partially mediating these associations.ConclusionThe TyG and its obesity-derived indices, particularly TyG-WWI and TyG-CVAI, exhibited significant incremental value for identifying CLMM, with CLMM status also conferring elevated subsequent mortality risk. Incorporating these indices into primary prevention monitoring systems for chronic diseases could facilitate the identification of individuals at high risk for CLMM and mortality.Graphical abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12933-026-03271-4.Keywords: Cardiovascular-liver-metabolic multimorbidity, CLMM, TyG and its obesity-derived indices, TyG-WWI, TyG-CVAI Research insights What is currently known about this topic? Cardiovascular-liver-metabolic multimorbidity (CLMM), defined as the co-occurrence of diabetes, cardiovascular disease, metabolic dysfunction-associated steatotic liver disease, and related conditions, is a growing global health burden. Cardiovascular-liver-metabolic multimorbidity (CLMM), defined as the co-occurrence of diabetes, cardiovascular disease, metabolic dysfunction-associated steatotic liver disease, and related conditions, is a growing global health burden. The triglyceride-glucose (TyG) and its obesity‑derived indices are important composite parameters of insulin resistance coupled with obesity, and have been associated with cardiometabolic diseases; however, their associations with CLMM remain unclear. The triglyceride-glucose (TyG) and its obesity‑derived indices are important composite parameters of insulin resistance coupled with obesity, and have been associated with cardiometabolic diseases; however, their associations with CLMM remain unclear. What is the key research question? Can the TyG and its obesity-derived indices (TyG-ABSI/TyG-CVAI/TyG-WWI/TyG-BRI/TyG-BMI/TyG-WHtR/TyG-WC) identify CLMM status? What is their value? Furthermore, does CLMM status influence long-term prognosis? Can the TyG and its obesity-derived indices (TyG-ABSI/TyG-CVAI/TyG-WWI/TyG-BRI/TyG-BMI/TyG-WHtR/TyG-WC) identify CLMM status? What is their value? Furthermore, does CLMM status influence long-term prognosis? What is new? Using a large population sample, this study represent the first investigation to establish the associations of the TyG and its obesity-derived indices with CLMM, and confirm that CLMM status is a significant risk factor to mortality. Using a large population sample, this study represent the first investigation to establish the associations of the TyG and its obesity-derived indices with CLMM, and confirm that CLMM status is a significant risk factor to mortality. Among all assessed TyG-derived indices, TyG-WWI and TyG-CVAI, which reflect insulin resistance and central adiposity, displayed the strongest associations with CLMM and demonstrated the optimal discriminatory performance. Among all assessed TyG-derived indices, TyG-WWI and TyG-CVAI, which reflect insulin resistance and central adiposity, displayed the strongest associations with CLMM and demonstrated the optimal discriminatory performance. How might this study influence clinical practice? Incorporating these indices into primary prevention monitoring systems for chronic diseases could facilitate the identification of individuals at high risk for CLMM and mortality. Incorporating these indices into primary prevention monitoring systems for chronic diseases could facilitate the identification of individuals at high risk for CLMM and mortality. Background Against the backdrop of accelerating global population aging and a sustained increase in the burden of chronic diseases, multimorbidity has emerged as a pivotal challenge in global public health [1], with multimorbidity patterns centered on cardiometabolic diseases (CMD) being widely endorsed by the research community [2]. In recent years, the global prevalence of metabolic dysfunction-associated steatotic liver disease (MASLD) has exhibited a rapid upward trend. Accumulating evidence indicates that hepatic steatosis is not an isolated pathological lesion, but rather engages in intricate pathological crosstalk with CMD. This understanding has gradually shifted the research perspective from individual CMD phenotypes toward multisystem disorders involving the interplay of cardiovascular, hepatic, and metabolic abnormalities, further promoting the proposal of the cardiovascular‑liver‑metabolic (CLM) concept [3, 4]. Of note, CMD and MASLD not only display prominent co‑occurrence but also share highly similar mechanistic pathways [3]. This highlights the non-independence of these diseases and underscores the importance of adopting a holistic perspective on health. In line with current concepts, CLM multimorbidity (CLMM) effectively integrates this theoretical framework: CLMM is not simply the sum of MASLD, cardiovascular disease (CVD), and diabetes, but rather reflects a state of multi-organ metabolic dysregulation involving concurrent impairments in the cardiovascular system, hepatic metabolic, and glucose metabolic systems [3, 4]. Compared with evaluating each disease separately, CLMM may better capture the cumulative disease burden and cross-organ metabolic damage associated with multimorbidity in real-world clinical settings. Given that the current understanding of CLMM remains in its infancy, it is essential to employ simple tools for the early characterization of its pathogenesis and subsequent impact. It is well established that insulin resistance (IR) represents the major common pathogenic pathway underlying both CMD and MASLD [3], rendering analyses of IR highly meaningful for understanding CLMM. Currently, the hyperinsulinemic-euglycemic clamp remains the gold standard for assessing IR [5]. However, its invasive nature, high cost, and procedural complexity have limited its widespread adoption in clinical practice. To address the challenges in IR assessment, investigators have proposed several simple and efficient approaches in recent years, among which the triglyceride-glucose (TyG) index has gained the most widespread application [6, 7]. Emerging evidence suggests that the TyG index serves as a valuable surrogate marker of IR for predicting metabolic diseases and their adverse prognoses, supporting its broad clinical utility [8–16]. Furthermore, recent studies have demonstrated that novel composite indicators combining the TyG index with obesity-related indices can further improve the identification of IR [17, 18] and enable efficient early detection of individuals at high risk for fatty liver disease [19]. Further investigations into TyG-obesity-derived indices [TyG-a body shape index (TyG-ABSI); TyG-Chinese visceral adiposity index (TyG-CVAI); TyG-weight adjusted waist index (TyG-WWI); TyG-body roundness index (TyG-BRI); TyG-body mass index (TyG-BMI); TyG-waist-to-height ratio (TyG-WHtR); TyG-waist circumference (TyG-WC)] have supported the notion that this combined strategy reflects comprehensive metabolic risk at the population level more sensitively than the TyG index alone [20–34]. Nevertheless, it should be noted that the utility of the TyG and its obesity-derived indices as IR surrogates in the context of CLMM remains unknown: Are the TyG and its obesity-derived indices (TyG-ABSI/TyG-CVAI/TyG-WWI/TyG-BRI/TyG-BMI/TyG-WHtR/TyG-WC) applicable for identifying CLMM status? What is their value? Furthermore, does CLMM status further influence long-term prognosis? Based on this, the present study aimed to leverage data from the National Health and Nutrition Examination Survey (NHANES) to systematically compare TyG and its obesity-derived indices to identify the optimal surrogate marker for CLMM, and to explore the potential impact of CLMM status on mortality. Materials and methods Study population and data sources This study was based on the NHANES database, and data from 10 consecutive cycles between 1999 and 2018 were extracted to form the initial cohort. Methodologically, NHANES is a nationwide cross-sectional and longitudinal survey conducted in the United States. It employs a multistage, stratified probability sampling design and collects a comprehensive array of data, including demographic characteristics, physical examination measures, laboratory test results, and dietary interview information. The study protocol was approved by the Ethics Review Board of the National Center for Health Statistics, and all participants provided written informed consent upon enrollment. Of the initial 101,316 enrolled participants, the following exclusion criteria were applied sequentially: (1) age < 18 years (n = 42,112); (2) missing complete variables required to calculate the TyG and its obesity-derived indices (n = 34,915); (3) presence of factors interfering with the diagnosis of MASLD, including excessive alcohol consumption (n = 600), pregnancy (n = 159), and viral hepatitis (n = 25); (4) unclear diagnostic status of CLMM (that is, the diagnoses of the CLMM components including CVD, diabetes and MASLD could not be definitively confirmed; n = 1,928); and (5) samples with a statistical weight of zero (n = 1,342). Ultimately, 20,235 nationally representative participants were included in the comprehensive analysis (Fig. S1). Calculation of the TyG and its obesity-derived indices The TyG index was used as a surrogate marker for assessing IR and calculated as follows: . To further quantify metabolic status and fat distribution characteristics, we mathematically combined the TyG index with seven obesity indicators to construct a series of TyG-obesity-derived indices: TyG-BMI, TyG-WC, TyG-WHtR, TyG-WWI, TyG-BRI, TyG-CVAI, and TyG-ABSI (detailed calculation formulas are provided in Table S1) [31, 35]. To eliminate statistical bias arising from differences in units and dimensions, all exposure indices were standardized using Z‑scores prior to entry into the regression models, ensuring relative comparability of risk estimates across different indices. Definition of CLMM In the present study, CLMM was defined as the coexistence of two or more conditions among CVD, MASLD, and diabetes [3, 4]. For individuals with confirmed positivity for two diseases but missing data on the third, we retained them in the corresponding observed CLMM subgroups. Table S2 presents the characteristics of the CLMM patterns. The diagnostic criteria for each disease were as follows: CVD: Based on self-report from a standardized medical questionnaire, i.e., the participant reported a history of congestive heart failure, coronary heart disease, angina, myocardial infarction, or stroke as diagnosed by a physician or other health professional. MASLD: A two-step diagnostic approach was adopted. First, the presence of hepatic steatosis was assessed using the United States Fatty Liver Index (US-FLI) (≥ 30) [36] (detailed calculation formula is provided in Table S3). Additionally, participants were required to meet at least one of the following cardiometabolic risk criteria: (1) BMI ≥ 25 kg/m2; (2) central obesity (WC ≥ 94 cm in males or ≥ 80 cm in females); (3) fasting plasma glucose (FPG) ≥ 5.6 mmol/L, hemoglobin A1c (HbA1c) ≥ 39 mmol/mol (5.7%), or receiving glucose‑lowering therapy; (4) blood pressure ≥ 130/85 mmHg or receiving antihypertensive therapy; (5) plasma triglycerides (TG) ≥ 1.70 mmol/L or receiving specific lipid‑lowering therapy; (6) abnormal high‑density lipoprotein cholesterol (HDL‑C) (< 1.0 mmol/L in males or < 1.3 mmol/L in females) or receiving cholesterol‑lowering therapy [37]. Diabetes: Defined according to the 2021 American Diabetes Association criteria: FPG ≥ 7.0 mmol/L, HbA1c ≥ 6.5%, a physician-diagnosed history, or current use of glucose-lowering medications [38]. Ascertainment of long-term survival endpoints The endpoints were all-cause mortality and cardiovascular mortality, ascertained by extracting mortality data from the U.S. National Death Index with follow-up through December 31, 2019. Cardiovascular-specific deaths were classified according to the International Classification of Diseases, Tenth Revision (ICD-10) codes (I00-I09, I11, I13, I20-I51, and I60-I69). Covariates To account for potential confounding, the following baseline covariates were included in this study: (1) sociodemographic factors: age, sex, race, marital status, educational attainment, and household income level; (2) lifestyle factors: self-reported smoking status (never, former, or current) and drinking status (non-heavy or heavy); (3) anthropometric and physiological measures: height, weight, WC, and resting blood pressure (mean of consecutive measurements) measured by trained personnel using standardized protocols; (4) laboratory biochemical indicators: complete blood count parameters (white blood cell count: WBC, red blood cell count, hemoglobin, platelet count: PLT), lipid metabolism parameters (total cholesterol: TC, TG, HDL-C, low-density lipoprotein cholesterol), liver enzyme parameters (alanine transaminase: ALT; aspartate transaminase: AST; gamma-glutamyl transferase: GGT), and glucose metabolism parameters (FPG, HbA1c). Baseline medical history (e.g., hypertension) was comprehensively determined based on self-report, medication use, and laboratory clinical thresholds. Statistical analysis All statistical analyses were performed using R software (version 4.5). All core analyses in this study incorporated the complex sampling design factors of NHANES, including sample weights, stratification variables, and primary sampling units [39]. Specifically, given the multiple biomarkers involved in our research, we derived analytic weights based on the biochemical examination weights released by NHANES. We used “sdmvstra.y” as the stratification variable and “sdmvpsu.y” as the primary sampling unit. Prior to model construction, we thoroughly assessed data quality and covariates: highly correlated variables were excluded based on the variance inflation factor (Table S4) and the correlation matrix of covariates (Fig. S2). Given the high completeness of the dataset (over 90% fully complete) and the assessment indicating that missing values were randomly scattered (Table S5, Figs. S3–S4), complete-case analysis was used for the primary analyses. Baseline characteristics were presented stratified by CLMM status and tertiles of the TyG index. Subsequently, weighted multivariable logistic regression models (progressing from crude models to fully adjusted models sequentially incorporating sociodemographic factors, lifestyle behaviors, and biochemical indicators) were constructed to evaluate the independent associations of the TyG and its obesity‑derived indices with CLMM. Based on the final model, restricted cubic splines were used to flexibly fit the dose–response relationships. Multivariable weighted Cox regression was applied to assess the impact of CLMM status on the associations of the TyG and its obesity‑derived indices with all‑cause and cardiovascular‑specific mortality. Furthermore, within the non‑CLMM and CLMM populations, we conducted stratum-specific analyses to compare the mortality risks between high and low levels of each TyG‑related index. Scaled Schoenfeld residual plots were used to test the proportional hazards assumption for all Cox regression models. We evaluated the proportional hazards assumption using scaled Schoenfeld residual plots for all sixteen Cox models, including eight TyG-related indices for all-cause mortality and eight for cardiovascular mortality. For all-cause mortality, the smoothed residual curves closely overlapped the zero reference line without any time-dependent trends, confirming satisfactory proportional hazards. For cardiovascular mortality, only a slight upward deviation was observed during early follow-up, with no substantial violation of the proportional hazards assumption (Fig. S5). To overcome the limitations of linear models and quantify variable contributions, we employed the XGBoost algorithm to construct a model for identifying CLMM status, with eight TyG-related indices included as independent predictive variables. Study participants were stratified and randomly split into a training set and a test set at a 7:3 ratio according to CLMM status, with five‑fold cross‑validation applied for hyperparameter tuning in the training set. Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, Brier score, and decision curve analysis. After model validation, the SHAP framework was further applied to quantify the relative contribution of each TyG-related index to model outputs [40]. For the internal components of the composite indices, weighted quantile sum (WQS) regression was applied to evaluate their independent contributions to CLMM [41]. Additionally, the incremental identification performance of the TyG‑obesity‑derived indices beyond the TyG index alone was systematically quantified using the AUROC, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) [42, 43]. Mediation analysis was then conducted to explore the potential biological mediating effects of the inflammatory markers WBC and C-reactive protein (CRP) [44]. Finally, subgroup analyses and interaction tests were performed based on key sociodemographic and lifestyle variables (e.g., age, sex, race). To validate the robustness of the main findings, the following sensitivity analyses were performed: modified Poisson regression was used to calculate the prevalence ratio to correct for risk overestimation; models were refitted based on multiply imputed datasets; regression analyses were repeated after excluding all participants with excessive alcohol consumption; and the Fine‑Gray competing risk model was introduced in survival analyses to account for interference by non‑cardiovascular death. Considering that MASLD and diabetes, as components of CLMM, have partial conceptual overlap with the independent variables, we further constructed alternative outcomes by removing diabetes or MASLD separately (i.e., CVD+MASLD comorbidity patterns and diabetes + CVD comorbidity patterns) to cautiously interpret the current findings. Additionally, we adjusted for HbA1c and FPG in the models to test the robustness of the results. To evaluate the impact of differences in follow‑up duration across NHANES survey cycles, we also performed analyses with fixed 3‑year and 5‑year follow‑up windows and constructed Cox models stratified by NHANES survey cycle. Furthermore, considering the limitations in the operational definition of MASLD and the assessment of liver fibrosis, we analyzed the association between US-FLI‑defined MASLD and Fibrosis-4 index (FIB-4)-based fibrosis risk stratification. FIB-4 was categorized into low-risk (< 1.30), intermediate-risk (1.30–2.67), and high-risk (≥ 2.67) according to commonly used thresholds. We further compared the distribution of FIB-4 values and the proportions of fibrosis risk strata across different MASLD statuses. All statistical tests were two‑sided, and a P value < 0.05 was considered statistically significant. Results Baseline characteristics of the study population A total of 20,235 participants were included in this study, with an overall CLMM prevalence of 15.1% (n = 3056) (Table 1). In terms of sociodemographic and lifestyle characteristics, compared with non-CLMM individuals, those with CLMM were older, a higher proportion were male and non-Hispanic White individuals, were more likely to be unmarried, and more often had lower socioeconomic status (low income and low educational attainment), as well as lower rates of smoking and non-heavy alcohol use. Regarding clinical comorbidities, the prevalence of hypertension, diabetes, and CVD was significantly higher in this population. In terms of biochemical and anthropometric characteristics, the CLMM group exhibited higher BMI and WC, as well as a worse metabolic profile (higher levels of WBC, TG, AST, ALT, FPG, and HbA1c, and lower levels of TC, HDL‑C, LDL‑C, and PLT). Of note, the TyG index and all its obesity‑derived indices showed significantly higher exposure levels in the CLMM group. Table 1Summary of baseline characteristics of the study population grouped according to CLMM statusCLMMP-valueNoYesNo. of subjects (n)17,1793056Age (years)46.79 (17.64)62.71 (13.11)< 0.001 Sex (n, %) Male8035 (46.8)1661 (54.4)< 0.001Female9144 (53.2)1395 (45.6) Race (n, %) Mexican American3024 (17.6)619 (20.3)< 0.001Other Hispanic1431 (8.3)296 (9.7)Non-Hispanic White7749 (45.1)1374 (45.0)Non-Hispanic Black3323 (19.3)528 (17.3)Other Race - Including Multi-Racial1652 (9.6)239 (7.8) Marital status (n, %) Married7880 (46.3)1234 (40.6)< 0.001Other9123 (53.7)1805 (59.4) Household. Income (dollars, n%) Less than 20,0003807 (22.8)828 (27.9)< 0.00120,000–49,9994975 (29.7)997 (33.6)50,000 or more7950 (47.5)1145 (38.6) Education (n, %) Less than 9th grade1773 (10.3)572 (18.8)< 0.0019th–11th grade including 12th grade no diploma2384 (13.9)531 (17.4)High school graduate GED or equivalent3925 (22.9)702 (23.1)Some college or associate degree4969 (29.0)790 (26.0)College graduate and above4110 (23.9)448 (14.7) Smoking Status (n, %) Never9729 (56.7)1416 (46.4)< 0.001Former3937 (22.9)1146 (37.6)Current3496 (20.4)487 (16.0) Drinking Status (n, %) Non-heavy12,641 (79.8)2670 (97.1)< 0.001Heavy3202 (20.2)80 (2.9) Hypertension (n, %) No11,279 (65.7)772 (25.3)< 0.001Yes5897 (34.3)2284 (74.7) Diabetes (n, %) No16,420 (95.6)482 (15.8)< 0.001Yes759 (4.4)2574 (84.2) CVD (n, %) No16,558 (96.4)1648 (54.1)< 0.001Yes621 (3.6)1396 (45.9)SBP (mmHg)119.00 [109.00, 131.00]129.00 [118.00, 143.00]< 0.001DBP (mmHg)69.92 (11.74)69.58 (13.02)0.148BMI (kg/m2)27.14 [23.80, 31.20]31.60 [27.90, 36.50]< 0.001Weight (kg)78.96 (19.68)91.40 (22.19)< 0.001Height (cm)167.41 (10.06)166.62 (10.25)< 0.001WC (cm)96.29 (14.97)111.10 (14.80)< 0.001WBC (×109/L)6.40 [5.40, 7.70]7.10 [5.90, 8.40]< 0.001RBC (×1012/L)4.71 (0.50)4.72 (0.54)0.335HB (g/L)14.21 (1.54)14.18 (1.63)0.368PLT (×109/L)251.20 (65.88)238.33 (70.51)< 0.001HDL-C (mmol/L)1.34 [1.11, 1.66]1.16 [0.98, 1.39]< 0.001LDL-C (mmol/L)3.04 (0.91)2.75 (0.98)< 0.001TC (mmol/L)5.09 (1.07)4.85 (1.20)< 0.001TG (mmol/L)1.14 [0.80, 1.68]1.60 [1.14, 2.30]< 0.001ALT (U/L)20.00 [16.00, 27.00]23.00 [18.00, 31.00]< 0.001AST (U/L)22.00 [19.00, 27.00]23.00 [19.00, 29.00]< 0.001GGT (U/L)19.00 [14.00, 28.00]27.00 [19.00, 43.00]< 0.001FPG (mmol/L)5.38 [5.05, 5.78]7.16 [6.16, 8.89]< 0.001HbA1c (%)5.40 [5.10, 5.70]6.40 [5.80, 7.50]< 0.001TyG index8.53 (0.60)9.22 (0.72)< 0.001TyG-BMI240.54 (59.05)302.65 (68.13)< 0.001TyG-WC824.52 (155.05)1024.84 (158.90)< 0.001TyG-WHtR4.93 (0.93)6.16 (0.95)< 0.001TyG-WWI93.14 (10.78)107.95 (10.63)< 0.001TyG-ABSI0.69 (0.07)0.78 (0.07)< 0.001TyG-BRI40.89 [30.08, 53.32]63.27 [51.04, 78.36]< 0.001TyG-CVAI1039.12 [582.34]1813.07 [548.83]< 0.001WBC: white blood cell count; RBC: red blood cell count; Hb: hemoglobin; PLT: platelet count; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol; TC: total cholesterol; TG: triglyceride; ALT: alanine transaminase; AST: aspartate transaminase; GGT: gamma-glutamyl transferase; HbA1c: haemoglobin A1c; FPG: fasting plasma glucose; SBP: systolic blood pressure; DBP: diastolic blood pressure; CLMM: cardiovascular‑liver‑metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Summary of baseline characteristics of the study population grouped according to CLMM status WBC: white blood cell count; RBC: red blood cell count; Hb: hemoglobin; PLT: platelet count; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol; TC: total cholesterol; TG: triglyceride; ALT: alanine transaminase; AST: aspartate transaminase; GGT: gamma-glutamyl transferase; HbA1c: haemoglobin A1c; FPG: fasting plasma glucose; SBP: systolic blood pressure; DBP: diastolic blood pressure; CLMM: cardiovascular‑liver‑metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Further stratification by tertiles of the TyG index (Table S6) showed that the highest TyG index group (tertile 3, T3) was characterized by older age, a higher proportion of males, lower socioeconomic status, and a higher smoking rate. The proportions of individuals with hypertension, diabetes, CVD, and CLMM were also significantly increased in this group. Regarding biochemical indicators, the T3 group exhibited marked metabolic abnormalities, characterized by elevated levels of TC, TG, LDL‑C, FPG, HbA1c, ALT, AST, GGT, WBC, red blood cell count, hemoglobin, and PLT, along with decreased HDL‑C. In addition, conventional anthropometric parameters (BMI, WC) and all TyG‑related obesity‑derived indices were also at relatively higher levels in this group compared with the overall population. Association of the TyG and its obesity-derived indices with CLMM Weighted logistic regression analyses (Table 2) demonstrated that both the TyG and its obesity‑derived indices were significantly positively associated with CLMM. In the fully adjusted model (Model 3), each one‑standard-deviation increment in these indices was associated with a significantly elevated risk of prevalent CLMM. By contrast, TyG-WWI, TyG-ABSI, and TyG-CVAI demonstrated stronger associations with CLMM, with ORs of 6.32 (5.71–7.01), 4.49 (4.10–4.92), and 4.39 (4.05–4.76), respectively. Following tertile-based grouping, all indices revealed a marked dose-response relationship. In the highest tertile, TyG-WHtR, TyG-WWI, and TyG-CVAI yielded the most prominent associations, with ORs of 60.81 (45.34–81.57), 53.14 (39.76–71.01), and 75.91 (51.80-111.25), respectively. Table 2Association study of TyG and its obesity-derived indices with CLMM(OR, 95% CI)Model 1Model 2Model 3TyG (Per SD increase)2.99 (2.79, 3.19)2.91 (2.70, 3.14)3.98 (3.62, 4.37)TyG-BMI (Per SD increase)2.53 (2.41, 2.66)2.60 (2.45, 2.76)2.74 (2.57, 2.92)TyG-WC (Per SD increase)3.56 (3.35, 3.77)3.58 (3.33, 3.83)3.92 (3.63, 4.23)TyG-WHtR (Per SD increase)3.87 (3.65, 4.10)3.99 (3.72, 4.28)4.37 (4.05, 4.72)TyG-WWI (Per SD increase)5.15 (4.74, 5.59)5.18 (4.70, 5.69)6.32 (5.71, 7.01)TyG-ABSI (Per SD increase)4.02 (3.72, 4.34)3.80 (3.49, 4.14)4.49 (4.10, 4.92)TyG-BRI (Per SD increase)2.91 (2.77, 3.07)3.07 (2.88, 3.26)3.16 (2.96, 3.37)TyG-CVAI (Per SD increase)3.84 (3.60, 4.09)4.19 (3.88, 4.54)4.39 (4.05, 4.76) TyG index tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T23.06 (2.52, 3.71)2.77 (2.27, 3.38)3.67 (2.98, 4.53)T313.03 (10.92, 15.56)11.22 (9.25, 13.62)15.67 (12.80, 19.19) TyG-BMI tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T24.59 (3.70, 5.69)4.50 (3.60, 5.61)5.50 (4.38, 6.91)T316.88 (13.94, 20.43)16.73 (13.67, 20.47)20.49 (16.67, 25.19) TyG-WC tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T25.72 (4.33, 7.55)5.47 (4.11, 7.29)6.69 (5.01, 8.96)T332.41 (25.17, 41.73)30.72 (23.60, 40.00)38.86 (30.06, 50.22) TyG-WHtR tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T29.74 (7.24, 13.09)9.53 (6.97, 13.04)11.43 (8.35, 15.66)T350.58 (38.37, 66.66)49.16 (36.61, 66.01)60.81 (45.34, 81.57) TyG-WWI tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T28.35 (6.30, 11.07)7.60 (5.68, 10.18)9.01 (6.68, 12.15)T350.09 (38.42, 65.32)42.64 (32.11, 56.61)53.14 (39.76, 71.01) TyG-ABSI tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T25.13 (4.07, 6.46)4.68 (3.64, 6.01)5.48 (4.25, 7.06)T325.92 (21.11, 31.82)21.10 (16.72, 26.62)25.34 (19.83, 32.38) TyG-BRI tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T27.21 (5.58, 9.32)6.90 (5.21, 9.14)7.81 (5.84, 10.44)T332.83 (26.34, 40.92)32.73 (25.57, 41.89)36.67 (28.45, 47.26) TyG-CVAI tertile groups T11.00 (Ref)1.00 (Ref)1.00 (Ref)T210.31 (7.37, 14.42)10.00 (6.87, 14.54)12.15 (8.23, 17.92)T360.75 (43.78, 84.31)61.83 (42.84, 89.24)75.91 (51.80, 111.25)OR: odds ratio; CI: confidence interval; CLMM: cardiovascular‑liver‑metabolic multimorbidity; other abbreviations as in Table 1Model 1: unadjustedModel 2: adjusted for sex, race, marital status, household income, education, smoking status, drinking statusModel 3: model 2 + WBC, Hb, PLT, LDL-C, ALT Association study of TyG and its obesity-derived indices with CLMM OR: odds ratio; CI: confidence interval; CLMM: cardiovascular‑liver‑metabolic multimorbidity; other abbreviations as in Table 1 Model 1: unadjusted Model 2: adjusted for sex, race, marital status, household income, education, smoking status, drinking status Model 3: model 2 + WBC, Hb, PLT, LDL-C, ALT Restricted cubic spline curves were fitted to characterize the dose-response associations between each exposure index and CLMM (Fig. 1). The analysis revealed that, except for the TyG index and TyG-ABSI, which showed a linear positive association with CLMM (P for nonlinearity > 0.05), the TyG-obesity‑derived indices exhibited significant nonlinear positive associations (P for nonlinearity < 0.05). Of note, these nonlinear curves generally exhibited a classic J‑shaped dose-response pattern, suggesting a potential threshold effect: the risk of prevalent CLMM remained relatively flat and low at lower quantiles of each index (a low‑risk plateau), but increased steeply once the exposure surpassed a specific nonlinear inflection point. The dose–response associations between TyG and its obesity-derived indices and the risk of CLMM. Models were adjusted for sex, race, marital status, household income, education, smoking status, drinking status, WBC, Hb, PLT, LDL-C, ALT. Abbreviations: CLMM: cardiovascular-liver-metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Incremental identification value and internal weight decomposition Figure 2 illustrates the comprehensive identification performance of the TyG and its obesity‑derived indices for CLMM. Compared with the basic TyG index (AUROC = 0.771), the novel derived indices incorporating obesity parameters significantly improved model discriminative capacity. Specifically, TyG‑CVAI and TyG‑WWI showed optimal performance in terms of model discrimination, with AUROC values of 0.839 and 0.834, respectively (DeLong test P < 0.001). In terms of risk reclassification performance, TyG‑WWI and TyG‑CVAI again achieved the greatest NRI and IDI (Table 3). Of note, TyG‑BMI not only showed no improvement in discrimination (DeLong test P = 0.297) but also significantly worsened reclassification (NRI = − 0.2744, IDI = − 0.0435; P < 0.001). Similarly, negative gains in reclassification performance were observed for TyG‑BRI. The comprehensive predictive performance of the TyG and its obesity‑derived indices for CLMM. Abbreviations: CLMM: cardiovascular-liver-metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Table 3Incremental predictive value of TyG–obesity-derivative indices versus TyG index for CMMAUROC(TyG index)AUROCDeLong PNRIP-valueIDIP-valueTyG-BMI0.7710.7650.297− 0.2744< 0.001− 0.0435< 0.001TyG-WC0.7710.819< 0.0010.2575< 0.0010.0286< 0.001TyG-WHtR0.7710.823< 0.0010.3072< 0.0010.0363< 0.001TyG-WWI0.7710.834< 0.0010.6406< 0.0010.0691< 0.001TyG-ABSI0.7710.805< 0.0010.3685< 0.0010.0289< 0.001TyG-BRI0.7710.803< 0.001− 0.04830.025− 0.00760.127TyG-CVAI0.7710.839< 0.0010.3764< 0.0010.0471< 0.001AUROC: area under the receiver operating characteristic curve; NRI: net reclassification improvement; IDI: integrated discrimination improvement; other abbreviations as in Table 1 Incremental predictive value of TyG–obesity-derivative indices versus TyG index for CMM AUROC: area under the receiver operating characteristic curve; NRI: net reclassification improvement; IDI: integrated discrimination improvement; other abbreviations as in Table 1 WQS regression analysis revealed that FPG had a relatively high contribution weight (Fig. 3; relative weight range: 50.7–77.2%). Notably, in the derived indices incorporating anthropometric parameters, the weight contributions of each body fat distribution index (17.18–38.5%) generally surpassed those of TG (10.8–18.5%). Relative weights of obesity index, FPG, and TG contributing to CLMM risk via WQS regression. Models were adjusted for sex, race, marital status, household income, education, smoking status, drinking status, WBC, Hb, PLT, LDL-C, ALT. Abbreviations: CLMM: cardiovascular-liver-metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Modifying effect of CLMM on the associations of the TyG and its obesity-derived indices with all-cause and cardiovascular mortality Using weighted multivariable Cox regression models, we identified that the TyG and its obesity-derived indices were significantly positively associated with both all‑cause and cardiovascular mortality (Tables S7-S8). By comparison, TyG-WWI, TyG-ABSI, and TyG-CVAI showed the strongest associations with mortality outcomes per standard deviation increment. Further stratified analyses revealed that CLMM status was an important effect modifier for poor prognosis (Tables S9–S10). Specifically, compared with individuals without CLMM and with low levels of TyG-related indices, those with CLMM had significantly higher risks of all-cause and cardiovascular mortality, regardless of whether their TyG-related indices were at low or high levels, particularly for cardiovascular mortality (Fig. 4). Additional stratum-specific analyses indicated that within the CLMM population, high levels of TyG-related indices were not consistently associated with greater mortality risks. In contrast, among the non-CLMM population, most elevated TyG-related indices were linked to higher risks of all-cause and cardiovascular mortality (Table S11). Radar charts illustrating the associations of TyG and its obesity-derived indices with all-cause (A) and cardiovascular (B) mortality in populations with and without CLMM. Models were adjusted for sex, race, marital status, household income, education, smoking status, drinking status, WBC, Hb, PLT, LDL-C, ALT. Abbreviations: CLMM: cardiovascular-liver-metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Machine learning interpretability analysis and mediation analysis We constructed an unadjusted XGBoost binary classification model to identify prevalent CLMM status, incorporating the eight TyG-related indices as independent predictors. The study participants were stratified by CLMM status and randomly divided into a training set and a test set at a ratio of 7:3. The training set comprised 14,165 participants, including 2142 with CLMM; the test set comprised 6070 participants, including 914 with CLMM. The prevalence of CLMM in the training set and test set was 15.12% and 15.06%, respectively, indicating a relatively balanced distribution of the outcome after stratified splitting. Model hyperparameters were tuned in the training set using 5-fold cross-validation, with the cross-validated AUROC as the optimization criterion. The final optimal model parameters were as follows: max_depth = 3, eta = 0.03, subsample = 0.90, colsample_bytree = 0.90, min_child_weight = 1, and gamma = 1, with an optimal boosting round of 218. The cross-validated AUROC of this model was 0.8785 ± 0.0105. Subsequently, we evaluated model performance in both the training set and the independent test set. In the independent test set, the XGBoost model demonstrated good discriminative ability, with an AUROC of 0.8821. Using the optimal threshold of 0.115 determined by the Youden index based on the training set, the test set achieved a sensitivity of 0.8345, specificity of 0.7637, accuracy of 0.7719, positive predictive value of 0.3172, negative predictive value of 0.9723, F1-score of 0.4597, and Brier score of 0.0755. The AUROC was 0.8997 in the training set and 0.8821 in the test set, and their modest difference indicated no obvious evidence of overfitting (Table S12). Furthermore, decision curve analysis showed that the XGBoost model had a positive net benefit compared with the treat‑none strategy within a certain range of probability thresholds, and exhibited a relatively higher net benefit across most threshold ranges, suggesting its potential clinical utility (Fig. S6). After completing model construction, cross‑validation, performance evaluation in the independent test set, and DCA‑based clinical utility assessment, we further interpreted the final XGBoost model using the SHAP framework. The SHAP summary plot showed that TyG‑CVAI and TyG‑WWI contributed the most to the model output, followed by TyG‑WC (Fig. 5). SHAP summary plot showing the importance of TyG and its obesity-derived indices in predicting CLMM. Red points indicate higher feature values, while blue points indicate lower feature values. Abbreviations: CLMM: cardiovascular-liver-metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Exploratory mediation analyses further evaluated the potential mediating roles of inflammation-related markers, WBC and CRP, in the associations of TyG and its obesity-derived indices with CLMM. The results showed that WBC exhibited a partial mediating effect in the associations between all TyG-related indices and CLMM, with mediation proportions ranging from approximately 2 to 5%. Among these, the mediating effect of WBC was most pronounced for TyG and TyG‑ABSI, with relatively higher mediation proportions. After further including CRP as a more representative marker of chronic low‑grade inflammation, CRP also demonstrated a similar mediating effect; however, the overall mediation proportions were smaller, ranging from approximately 1 to 2%, with relatively wide 95% confidence intervals for some indices. Overall, the mediation analyses based on both WBC and CRP suggest that inflammation‑related pathways may partially mediate the associations between TyG‑related indices and CLMM, although the overall mediation proportions are limited (Fig. 6). Mediation proportion of selected biomarkers in the associations between TyG and its obesity-derived indices and the risk of CLMM. Models were adjusted for sex, race, marital status, household income, education, smoking status, drinking status, Hb, PLT, LDL-C, ALT. Abbreviations: CLMM: cardiovascular-liver-metabolic multimorbidity; TyG: triglyceride-glucose; TyG-ABSI: TyG-a body shape index; TyG-CVAI: TyG-Chinese visceral adiposity index; TyG-WWI: TyG-weight adjusted waist index; TyG-BRI: TyG-body roundness index; TyG-BMI: TyG-body mass index; TyG-WHtR: TyG-waist-to-height ratio; TyG-WC: TyG-waist circumference Subgroup analyses Subgroup analyses revealed significant effect modification by age and sex on the associations of the TyG and its obesity-derived indices with CLMM risk (Fig. S7). Specifically, age significantly modified the associations of the TyG index, TyG-WWI, and TyG-ABSI with CLMM (P-interaction < 0.001): the associations of these indices with CLMM were consistently stronger in participants aged < 60 years than in those aged ≥ 60 years (TyG: 4.58 vs. 3.45; TyG-WWI: 7.31 vs. 4.71; TyG-ABSI: 4.98 vs. 3.32). Furthermore, sex showed significant interaction effects on the associations of the TyG index, TyG‑WC, TyG-CVAI, and TyG‑ABSI with CLMM (Female vs. male: TyG index 4.74 vs. 3.42; TyG‑WC 4.22 vs. 3.67; TyG‑CVAI 6.31 vs. 3.54; TyG‑ABSI 3.98 vs. 5.35). Interactions were also observed for some indices in the race and educational attainment subgroups. However, marital status, household income level, and smoking/drinking status showed no significant modification effects on the associations of most TyG‑related obesity‑derived indices with CLMM (most P‑interaction > 0.05). Sensitivity analyses The high robustness of the main findings was supported by a series of sensitivity analyses. After applying modified Poisson regression to directly calculate prevalence ratios, excluding participants with excessive alcohol consumption, and performing multiple imputation for missing data, the overall association patterns of the TyG and its obesity-derived indices with CLMM did not change materially (Tables S13–S15). Furthermore, after accounting for the competing risk of non-cardiovascular death using the Fine-Gray model, the predictive associations of the TyG and its obesity-derived indices with long-term cardiovascular-specific mortality remained robust (Table S16). To address potential component overlap between CLMM and the independent variables in their definitions, we constructed two alternative comorbidity outcomes: CVD+MASLD comorbidity (with diabetes removed) and CVD+diabetes comorbidity (with MASLD removed), while additionally adjusting for HbA1c and FPG in the models. Although the effect estimates were attenuated in these sensitivity analyses, an overall positive association trend remained evident (Tables S17–S19). After adopting Cox models with fixed 3-year and 5-year follow-up windows as well as Cox models stratified by NHANES survey cycles, the modifying effect of CLMM status on the associations of TyG and its obesity derived indices with all-cause and cardiovascular mortality persisted (Table S20). Finally, compared with the non-MASLD population, the US-FLI-defined MASLD population had higher FIB-4 levels (Table S21), indicating that US-FLI-defined MASLD was associated with elevated FIB-4 levels and increased fibrosis risk. Discussion Based on a large population sample, this study represented the first investigation to establish the associations of the TyG and its obesity-derived indices with CLMM, and confirmed that CLMM status acts as a significant risk contributor to mortality. Among all TyG-derived indices assessed, TyG-WWI and TyG-CVAI, which characterize IR and central adiposity, displayed the strongest associations with CLMM and the optimal discriminatory performance. Mechanistic exploratory analyses further suggested that glucose metabolism may serve as a central driver of CLMM, with its pathological actions mediated in part by suppressing reverse cholesterol transport and enhancing inflammation. The disruption of glucose-lipid metabolic homeostasis driven by IR represents not only the pathogenesis of individual metabolic disorders but also the central pathological mechanism of CLMM [3, 4]. As a reliable surrogate for IR, the TyG index has been previously validated for risk assessment of diabetes, MASLD, and CVD [8–16]. However, analyses focusing on individual diseases cannot reflect the clustering pattern of these conditions within the same individual, nor can they fully capture the cumulative disease burden resulting from multi-organ metabolic injury. Additionally, the TyG index alone cannot quantify ectopic fat deposition and abnormal fat distribution—important pathological characteristics underlying the development and progression of CLMM. To address this limitation, previous studies have combined the TyG index with conventional anthropometric indices such as BMI, WC, and WHtR to improve the accuracy of risk assessment. In populations with a specific or single disease, studies by Dang et al. [20] and Min et al. [24] indicated that TyG-WC and TyG-WtHR possess important independent predictive value for mortality risk among patients with CVD and MASLD. This observation has also been validated in recent studies on CMM. Multiple analyses based on the CHARLS cohort, published separately by Zhang et al., Ruan et al., and Wang et al., demonstrated that elevated TyG and its obesity-derived indices were significantly associated with an increased risk of incident CMM [13, 45, 46]. Meanwhile, prospective evidence from Tian et al. based on the UK Biobank also suggested that multiple IR-related indicators were closely associated with the progression trajectories of CMM [29]. Similarly, the application of TyG-related indices in cardiovascular–kidney–metabolic syndrome (CKM syndrome) has attracted growing attention. For instance, a study by Wu et al. using NHANES data reported that an elevated TyG index was associated with a higher risk of advanced CKM syndrome [47]. Moreover, several studies have specifically examined the positive correlations of TyG and its obesity‑derived indices with future cardiovascular events and adverse prognoses among individuals with CKM syndrome [22, 23, 34, 48–50]. Collectively, these studies suggest that TyG‑related indices hold potential value for identifying multisystem metabolic risk and for prognostic assessment. However, prior research has primarily focused on the interplay among cardiovascular, renal, and metabolic abnormalities, while insufficient attention has been paid to hepatic metabolic dysfunction, particularly the role of MASLD within the framework of multimorbidity. Given this research gap, the present study extended the cardiometabolic multimorbidity framework to incorporate the hepatic metabolic dimension, focusing on CLMM defined as the co-occurrence of CVD, MASLD, and diabetes, and systematically compared the associations of TyG and its obesity‑derived indices with CLMM. The findings revealed that all TyG-related obesity-derived indices were positively associated with CLMM, among which TyG-WHtR, TyG-WWI, and TyG-CVAI showed the highest correlation coefficients. In summary, the inclusion of MASLD means that CLMM no longer merely reflects the additive effects of general cardiometabolic abnormalities, but rather highlights the important roles of hepatic lipid deposition and visceral fat accumulation. Therefore, TyG-CVAI and TyG-WWI, which reflect visceral fat burden and abnormal body fat distribution, may be more suitable than the TyG index alone for identifying the complex metabolic abnormalities associated with CLMM. This notion is further supported by the superior discriminative performance of TyG-CVAI and TyG-WWI observed in our SHAP and ROC analyses. These findings are highly consistent with recent evidence [31, 33, 34]. In studies by Yue et al. and Wang et al., TyG-CVAI demonstrated superior predictive performance for stroke and CKM-related outcomes based on comparisons across multiple indices [33, 34]. Notably, in contrast to the study by Li et al., which found that TyG-BMI performed excellently in individuals with CKM stages 0–3 [22], the present study observed that TyG-BMI exhibited no incremental identification value for CLMM. This discrepancy may be primarily attributed to the inherent limitations of BMI. As a conventional obesity index, BMI cannot adequately assess visceral fat distribution or skeletal muscle mass, and thus fails to accurately capture metabolically abnormal leanness or muscle loss commonly observed in the CLMM population [51–53], thereby reducing its risk assessment efficacy in complex multimorbidity settings. With respect to long-term survival outcomes, substantial evidence has supported the predictive value of TyG-WHtR or TyG-WC for poor prognoses, including all-cause and cardiovascular mortality. For example, in individuals with CKM, Zhang et al. reported that TyG-WHtR exhibited the strongest association with cardiovascular mortality, while TyG-WC showed the strongest correlation with all-cause mortality [23]. Additional studies across diverse populations have further validated the reliability of obesity-derived indices such as TyG-WHtR and TyG-WC in predicting long-term mortality risk [20, 21, 24, 25, 28]. Against this background, after clarifying the associations of TyG and its obesity-derived indices with CLMM and mortality, the present study further focused on the modifying effect of CLMM status on the relationship between these indices and mortality risk. Results of stratified analyses showed that, among individuals without CLMM, high levels of TyG-obesity-derived indices were associated with only a moderate increase in mortality risk. However, in the presence of CLMM, both all-cause and cardiovascular mortality risks were markedly increased. Further joint stratified analyses revealed that among individuals with pre-existing CLMM, high levels of TyG and its obesity-derived indices were not consistently associated with further elevated mortality risk. This finding suggests that when an individual already has multimorbidity involving CVD, MASLD, and diabetes, the mortality risk may be more influenced by the multimorbidity burden represented by CLMM itself, disease severity, underlying changes in nutritional status, and residual confounding, rather than reflecting an additive or multiplicative amplification of mortality risk by TyG-related indices. Within this joint exposure assessment framework, we further compared the hazard ratios of different TyG-obesity-derived index combinations. The study revealed that although TyG-WC and TyG-WHtR retained certain predictive value for long-term mortality risk, TyG-CVAI exhibited substantially superior risk discrimination ability. Collectively, elevated TyG‑CVAI not only increases the risk of CLMM but also serves as a powerful predictor of mortality in patients with CLMM. From a pathophysiological standpoint, this advantage may stem from the more comprehensive nature of CVAI. Unlike BMI or WC alone, CVAI integrates age, BMI, WC, and key lipid parameters (TG and HDL‑C), allowing for more precise quantification of visceral fat accumulation [36]. Accordingly, it more sensitively captures chronic low‑grade inflammation and endothelial injury driven by visceral obesity. Such sustained metabolic stress may accelerate irreversible target organ damage in CLMM patients, ultimately contributing to a marked increase in long-term mortality risk. To further explore the core drivers and pathological pathways underlying CLMM, we performed WQS analysis to estimate the contribution weights. The results demonstrated that FPG played a dominant role in driving the risk of prevalent CLMM, suggesting that long-term glucotoxicity may be a critical trigger for the entire pathological cascade. Sustained hyperglycemic exposure not only directly impairs vascular endothelial function [54, 55] but also induces irreversible structural damage to the cardiovascular and hepatic microenvironments by accelerating the accumulation of advanced glycation end products [56–58]. Additionally, exploratory mediation analyses revealed that WBC and CRP partially mediated the associations of TyG and its obesity‑derived indices with CLMM, with mediated proportions ranging from 1–5%. This finding suggests that inflammatory status may be involved in or reflect the associations between TyG‑related indices and CLMM. This may be related to chronic low‑grade inflammation accompanying visceral fat accumulation and metabolic disturbances [59, 60]. Subgroup analyses in the present study also revealed several differential findings, most notably in age-stratified groups: exposure indices (TyG index, TyG-WWI, and TyG-ABSI) exhibited stronger associations with CLMM in young and middle-aged individuals (< 60 years). We speculate that this phenomenon may be attributed to the strong competing risk effect in the elderly population. With advancing age, natural senescence, apoptosis, and other non-metabolic comorbidities gradually become the dominant drivers of target organ damage [61, 62], and this superimposed effect statistically dilutes the relative risk attributable solely to IR and obesity. Conversely, metabolic abnormalities in young and middle-aged adults represent a purer, more aggressive pathological phenotype with minimal influence from aging and non-metabolic factors, thereby allowing the risk predictive value of TyG-obesity-derived indices to be more fully expressed. Besides age, sex also exerted a significant modifying effect on the TyG index, TyG-WC, and TyG-CVAI, a difference that may be attributed to sex hormone‑mediated dimorphism in fat distribution. Women, particularly premenopausal women, are protected by estrogen and tend to accumulate fat predominantly in subcutaneous depots [63–65], a pattern that is less likely to induce metabolic disturbances. Moreover, risk heterogeneity identified in specific ethnic and educational subgroups highlights the critical moderating role of social determinants of health in the development and progression of CLMM. Such heterogeneity may be closely associated with population-specific differences in long-term epigenetic modifications, dietary patterns, and cumulative chronic environmental stress [66–69]. Strengths and limitations This study has distinct methodological and epidemiological advantages. Based on a large, nationally representative multi-ethnic cohort, our findings are generalizable to broad real-world populations. Moreover, we applied a spectrum of advanced statistical approaches, including SHAP, WQS analysis, mediation analysis, and competing risk models. The robustness of the core findings was further validated by multiple sensitivity analyses, including multiple imputation and modified Poisson regression. Nevertheless, several limitations should be acknowledged. First, the results of the logistic regression analyses in this study should be interpreted as associations between TyG-related indices and CLMM. ROC, improvement index, and SHAP analyses only reflect the cross-sectional ability of these indices to distinguish individuals with and without CLMM, rather than predictive performance for future CLMM risk. Future prospective cohort studies based on longitudinal follow-up data are warranted to further validate whether the TyG and its obesity-derived indices have value in predicting incident CLMM, and to clarify the temporal sequence and potential causal relationships between these indices and the development and progression of CLMM. Second, all metabolic composite indices were derived from a single baseline measurement, failing to account for the cumulative impact of longitudinal dynamic trajectories of exposure levels on clinical outcomes. Third, although the multivariable models were extensively adjusted for numerous clinical and socioeconomic variables, residual confounding from potential unmeasured factors such as specific genetic polymorphisms and detailed dietary patterns could not be completely ruled out. Fourth, in the NHANES project, CVD information is primarily derived from self-reported medical history obtained through standardized questionnaires, which may introduce risks of recall bias and disease misclassification. Fifth, because the calculation of the TyG and its obesity‑derived indices depends on FPG, TG, and multiple anthropometric measurements, many participants were excluded due to missing key variables. This may have led to systematic differences between the final analytical sample and the excluded population, thereby introducing potential selection bias. Additionally, some participants could not be classified for CLMM status because of incomplete data on CVD, diabetes, or MASLD, and were thus excluded. Although we used NHANES complex sampling weights to enhance the representativeness of our results, weighting adjustment cannot fully eliminate selection bias due to missing data and sample exclusion. Sixth, in this study, MASLD was operationally defined based on a US-FLI score ≥ 30 together with criteria for metabolic abnormalities, rather than confirmed by liver imaging or biopsy, which may introduce some degree of disease misclassification. Conclusion The TyG and its obesity-derived indices, particularly TyG-WWI and TyG-CVAI, exhibit significant incremental value for identifying CLMM, with CLMM status also conferring elevated subsequent mortality risk. Incorporating these indices into primary prevention monitoring systems for chronic diseases could facilitate the identification of individuals at high risk for CLMM and mortality. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Acknowledgements We are grateful to the NHANES project team for data support and to colleagues at Nanchang University for reviewing and revising the English language of the manuscript. Abbreviations Cardiometabolic diseases Metabolic dysfunction-associated steatotic liver disease Cardiovascular-liver-metabolic multimorbidity Insulin resistance Cardiovascular disease Triglyceride-glucose TyG-a body shape index TyG-Chinese visceral adiposity index TyG-weight adjusted waist index TyG-body roundness index TyG-body mass index TyG-waist-to-height ratio TyG-waist circumference White blood cell count Platelet count High-density lipoprotein cholesterol Triglyceride Haemoglobin A1c Fasting plasma glucose Odds ratios Weighted quantile sum Area under the receiver operating characteristic curve Net reclassification improvement Integrated discrimination improvement United States Fatty Liver Index Author contributions CW, HL-L and WW contributed to conceptualization, methodology, and writing of the original draft. SM-H contributed to visualization, investigation, supervision, data cleaning, data curation, and writing—review and editing. YF-L, GB-X, GT-S contributed to visualization, investigation, supervision, and writing—review and editing. YZ contributed to supervision, data curation, and funding acquisition. All authors read and approved the final manuscript. Funding This work was supported by the Natural Science Foundation of Jiangxi Province [20232BAB216004]. Data availability The datasets are available for download from the NHANES home website (https://www.cdc.gov/nchs/nhanes/index.htm). Declarations Ethics approval and consent to participate The study protocol was approved by the Ethics Review Board of the National Center for Health Statistics, and all participants provided written informed consent upon enrollment. 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. Contributor Information Wei Wang, Email: wwangcvri@163.com. Hengli Lai, Email: laihengli@163.com. Yang Zou, Email: jxyxyzy@163.com. Chao Wang, Email: ahmu_wangchao@outlook.com. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The datasets are available for download from the NHANES home website (https://www.cdc.gov/nchs/nhanes/index.htm).