Association between triglyceride–glucose muscle-loss index (TyG–MLI) and incident cardiovascular disease: a prospective cohort study in the UK biobank Authors: Yanwen Fang, Sizhuang Huang, Hao Zhang, Mingyang Li, Haihua Zhang, Weixian Yang, Mengyue Yu Published: 2025 ABSTRACT BackgroundInsulin resistance and muscle loss are interrelated processes linked to cardiovascular disease (CVD), but they are rarely assessed together. We investigated the association between a novel integrated biomarker—triglyceride–glucose muscle-loss index (TyG–MLI)—and incident CVD in a large prospective cohort.MethodsA total of 277,418 UK Biobank participants free of CVD at baseline were included. TyG–MLI was calculated as ln[(triglycerides × fasting glucose)/2] × (cystatin C/creatinine). Incident CVD was identified through hospital admissions and death registries. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs).ResultsOver a median follow-up of 15.1 years, 40,398 participants developed CVD. Crude CVD incidence increased steadily across TyG–MLI quartiles, and the cumulative incidence curves demonstrated clear separation on Kaplan–Meier analysis (log-rank p < 0.001). Higher TyG–MLI was independently associated with greater CVD risk when modeled either continuously (per 1-SD: HR 1.17; 95% CI 1.16–1.18) or categorically (Q4 vs. Q1: HR 1.48; 95% CI 1.44–1.53). Restricted cubic spline analyses demonstrated a non-linear association across the TyG–MLI distribution. Sensitivity analyses yielded consistent results. In body mass index-stratified analyses, TyG–MLI showed stronger associations with CVD than TyG index, particularly among normal-weight and underweight individuals.ConclusionsTyG–MLI, a composite marker reflecting insulin resistance and muscle loss, was independently associated with incident CVD in this large population-based cohort. These findings suggest that TyG–MLI captures cardiometabolic risk beyond traditional measures and may be particularly informative in individuals without obesity.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12933-025-03031-w. Research Insights What is currently known about this topic? Insulin resistance and muscle loss each increase CVD risk but are rarely evaluated together. No existing biomarker integrates both processes to characterize long-term CVD risk Insulin resistance and muscle loss each increase CVD risk but are rarely evaluated together. No existing biomarker integrates both processes to characterize long-term CVD risk Insulin resistance and muscle loss each increase CVD risk but are rarely evaluated together. No existing biomarker integrates both processes to characterize long-term CVD risk What is the key research question? Is TyG–MLI, which integrates markers of insulin resistance and muscle loss, independently associated with incident CVD in the general population? Is TyG–MLI, which integrates markers of insulin resistance and muscle loss, independently associated with incident CVD in the general population? Is TyG–MLI, which integrates markers of insulin resistance and muscle loss, independently associated with incident CVD in the general population? What is new? Higher TyG–MLI was independently associated with incident CVD in a large cohort, with stronger associations than TyG alone, particularly in normal- and low-BMI groups Higher TyG–MLI was independently associated with incident CVD in a large cohort, with stronger associations than TyG alone, particularly in normal- and low-BMI groups Higher TyG–MLI was independently associated with incident CVD in a large cohort, with stronger associations than TyG alone, particularly in normal- and low-BMI groups How might this study influence clinical practice? TyG–MLI may help highlight cardiometabolic vulnerability, especially among individuals without obesity whose metabolic or muscle-related risks may otherwise be underestimated TyG–MLI may help highlight cardiometabolic vulnerability, especially among individuals without obesity whose metabolic or muscle-related risks may otherwise be underestimated TyG–MLI may help highlight cardiometabolic vulnerability, especially among individuals without obesity whose metabolic or muscle-related risks may otherwise be underestimated Introduction Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality worldwide, representing a major challenge to global health [1]. Obesity and metabolic disturbances are currently regarded as major contributors of CVD development. Building on this recognition, the American Heart Association (AHA) recently introduced the cardiovascular–kidney–metabolic (CKM) framework, which underscores the continuous progression from obesity and metabolic dysfunction to subclinical and overt CVD [2]. However, metabolic disturbances are not restricted to obese individuals. Even those with normal or low body weight index (BMI) may harbor adverse metabolic profiles that heighten cardiovascular risk, a vulnerability often linked to latent insulin resistance (IR) and loss of skeletal muscle mass [3]. This highlights a limitation of CKM-based risk assessment and underscores the need for biomarkers that capture cardiometabolic vulnerability beyond adiposity, especially in non-obese populations. IR plays a central role in CVD pathogenesis, linking impaired glucose utilization and hyperglycemia with vascular injury and cardiometabolic dysfunction. To facilitate practical assessment, triglyceride–glucose (TyG) index, derived from triglycerides (TG) and fasting plasma glucose (FPG), has been widely validated as a surrogate of IR and consistently associated with multiple CVD [4]. Within CKM framework, TyG index and other surrogate markers of IR have demonstrated incremental prognostic value for risk stratification and outcome prediction [5, 6]. Given the close link between IR and adiposity, several composite indices combining TyG index with anthropometric measures such as BMI or waist circumference have been developed to capture the integrated risk profile [7, 8]. However, these indices primarily reflect obesity-related pathways and do not account for muscle status, an equally critical determinant of insulin sensitivity and cardiovascular health. This limitation is especially pronounced in non-obese individuals, in whom adverse outcomes are often driven by hidden IR and muscle loss rather than overt adiposity. Skeletal muscle, the principal site of glucose disposal, plays a pivotal role in maintaining systemic metabolic homeostasis. Accumulating evidence indicates that sarcopenia is closely associated with CVD [9]. The creatinine-to-cystatin C ratio (Cr/CysC), or sarcopenia index (SI), is a readily measurablebiomarker reflecting relative muscle mass. Lower SI has been linked to a higher risk of incident CVD and cardiovascular mortality, underscoring its potential utility as a serum-based marker of sarcopenia in cardiovascular risk prediction [10]. Sarcopenia and IR are tightly linked, with loss of muscle mass reducing glucose disposal and IR in turn promoting further muscle deterioration [11, 12]. Epidemiological data also reinforce this link, as recent studies report that elevated TyG index and its derivative composite indices are strongly associated with a higher prevalence of sarcopenia in the general population [13, 14]. While existing biomarkers such as TyG index and SI each capture one dimension of IR or muscle loss, no composite metric has yet integrated these two interrelated processes. Therefore, we propose triglyceride–glucose muscle-loss index (TyG–MLI) as a novel biomarker capturing the combined burden of IR and muscle depletion. In this study, we sought to validate this concept by examining the association between TyG–MLI and incident CVD in a large, population-based cohort. Methods Study population The UK Biobank is a large, population-based prospective cohort that recruited over 500,000 participants aged 37–73 years from 22 assessment centres across England, Scotland, and Wales between 2006 and 2010. At baseline, participants completed touchscreen and nurse-administered questionnaires on sociodemographic characteristics, lifestyle behaviours, and medical history, underwent standardised physical measurements, and provided blood samples for biochemical assays. Among the 502,128 participants initially enrolled, we excluded 2 without baseline time information, 74,566 lacking TyG-related indices or missing creatinine (Cr) or cystatin C (CysC) measurements, 179,963 lacking covariate data, and 28,576 with prevalent CVD at baseline. The final analytic cohort consisted of 277,418 participants (Fig. 1). The pattern of missingness across covariates and baseline characteristics comparing included participants with those excluded due to missing covariates are presented in Supplementary Tables S1 and S2.Fig. 1Flow diagram of participant selection. Abbreviations: TyG, triglyceride–glucose index; CVD, cardiovascular disease Flow diagram of participant selection. Abbreviations: TyG, triglyceride–glucose index; CVD, cardiovascular disease Exposure assessment The exposure of interest was the composite biomarker TyG–MLI. At baseline, serum TG, FPG, Cr, and CysC were measured in the UK Biobank central laboratory using standardized protocols on the Beckman Coulter AU5800 analyzer. TyG index, a validated surrogate marker of IR, was calculated as ln[(TG × FPG)/2] [4]. Muscle-loss index (MLI) was defined as the reciprocal of SI (Cr/CysC × 100) [10] and was calculated as CysC/Cr. The scaling factor of 100 was omitted to maintain comparable numerical magnitude, as it does not affect biological interpretation. By taking the reciprocal, higher MLI values reflect greater muscle depletion after accounting for renal function. TyG–MLI was defined as ln[(TG × FPG)/2] × (CysC/Cr), consistent with the multiplicative formulation used in other TyG-derived composite indices [7, 8] to represent the coexistence of IR and muscle loss. Unit conversion factors used in these calculations are provided in Supplementary Table S3. Ascertainment of outcomes The primary outcome was incident CVD, defined as a composite of coronary heart disease (CHD), stroke, heart failure (HF), atrial fibrillation (AF), and peripheral artery disease (PAD). Secondary outcomes included each CVD subtype. All outcomes were identified through linked hospitalization and mortality records. Hospitalization data were obtained from the Hospital Episode Statistics for England, the Scottish Morbidity Record for Scotland, and the Patient Episode Database for Wales. Mortality information for England and Wales was sourced from the National Health Service Digital, and deaths in Scotland were identified via the National Records of Scotland. Diagnoses were coded using the International Classification of Diseases, 10th Revision (Supplementary Table S4). Hospital admission and mortality records in the UK Biobank are updated several times per year. Follow-up for this study was censored on 8 July 2024. Covariates Baseline information on age, sex, ethnicity, and educational attainment was obtained through touchscreen questionnaires at the assessment center. Socioeconomic status was evaluated with the Townsend deprivation index, derived from census-linked residential postal codes, where higher scores reflect greater area-level deprivation [15]. Lifestyle factors included smoking status (never, former, or current), alcohol intake (never, < 3 times per week, or ≥ 3 times per week), a cumulative diet risk score [16], sleep duration (< 6, 6–9, or > 9 h per day), and physical activity (PA). PA was assessed using self-reported questionnaires at baseline and converted into metabolic equivalent task scores following standard procedures. Participants were classified according to the 2017 UK guideline as meeting or not meeting the recommendation of at least 150 min of moderate-intensity activity or 75 min of vigorous-intensity activity per week. Clinical risk factors included hypertension, defined as self-report, measured blood pressure ≥ 140/90 mmHg, or use of antihypertensive medication; diabetes, defined as self-report, FPG ≥ 7.0 mmol/L, glycated hemoglobin ≥ 6.5%, or use of glucose-lowering therapy; and low-density lipoprotein cholesterol (LDL-C). Statistical analysis Baseline characteristics were summarized as means with standard deviations (SD), medians with interquartile ranges (IQR), or numbers with percentages, as appropriate. Incidence rates were calculated as the number of events per 1,000 person-years. Cumulative incidence curves were estimated using Kaplan–Meier methods and compared using log-rank tests. Associations between TyG–MLI and incident CVD were examined using Cox proportional hazards models, with results reported as hazard ratios (HRs) and 95% confidence intervals (CIs). Three models were fitted: Model 1 was unadjusted; Model 2 adjusted for demographic and lifestyle factors (age, sex, ethnicity, education, Townsend deprivation index, smoking status, alcohol intake, dietary score, sleep duration, and PA); and Model 3 further adjusted for clinical risk factors (hypertension, diabetes, and LDL-C). TyG–MLI was evaluated both as a continuous variable per 1-SD increase and as a categorical variable defined by quartiles. Proportional hazards assumptions were assessed using Schoenfeld residuals (Supplementary Fig. S1). Dose–response relationships and potential nonlinearity were evaluated using restricted cubic spline models with three knots. Prespecified stratified analyses were conducted according to BMI categories (< 18.5, 18.5–24.0, and ≥ 24.0 kg/m2). Subgroup analyses were performed for age, sex, hypertension, and diabetes, and multiplicative interaction was assessed using likelihood ratio tests. To account for multiple comparisons, both Bonferroni correction (α = 0.0125) and False Discovery Rate (FDR) adjustment were applied. We additionally compared the area under the receiver operating characteristic curve (AUC) for TyG index and TyG–MLI, both alone and in combination with the AHA PREVENT risk model [17] (Supplementary Table S5), using DeLong’s test [18]. To assess the robustness of the findings, several sensitivity analyses were conducted. First, Fine–Gray subdistribution hazard models were applied to account for the competing risk of non-cardiovascular death. Second, to minimize potential reverse causation, a 2-year landmark analysis was performed by excluding events occurring within the first two years of follow-up. Third, to evaluate the impact of missing covariates and potential selection bias, analyses were repeated using multiple imputation based on chained equations. Fourth, because MLI may be affected by kidney function and TyG by statin therapy, additional models further adjusting for estimated glomerular filtration rate (eGFR) and statin use were fitted. Height-adjusted TyG–MLI was also constructed to account for variation in body size. Finally, internal geographic validation was conducted by re-estimating the fully adjusted models within each assessment region (England, Scotland, and Wales) separately. All statistical analyses were performed using R software (version 4.5.0). Two-sided p values < 0.05 were considered statistically significant. Results Baseline characteristics of participants Figure 2 illustrates the distribution of TyG–MLI in the analytic cohort, which was approximately normal. Based on this distribution, participants were categorized into quartiles (Q1–Q4). Table 1 presents baseline characteristics across TyG–MLI quartiles. Higher TyG–MLI quartiles corresponded to older age, a higher proportion of women, and greater prevalence of hypertension, diabetes, and adverse lifestyle factors such as smoking and physical inactivity. Overall, individuals with higher TyG–MLI exhibited a less favorable cardiometabolic profile.Fig. 2Distribution of TyG–MLI in the cohort. Histogram showing the distribution of TyG–MLI among 277,418 participants. Participants were categorized into quartiles: Q1 ≤ 8.41; Q2 = 8.42–9.53; Q3 = 9.54–10.87; Q4 ≥ 10.88. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss indexTable 1Baseline characteristics of participants according to quartiles of TyG–MLIVariableTotal(n = 277 418)Q1(n = 69 355)Q2(n = 69 355)Q3(n = 69 354)Q4(n = 69 354)Age, years56.01 (8.08)53.15 (8.14)55.25 (8.09)56.87 (7.81)58.76 (7.18)SexFemale146 671 (53%)21 736 (31%)30 801 (44%)40 432 (58%)53 702 (77%)Male130 747 (47%)47 619 (69%)38 554 (56%)28 922 (42%)15 652 (23%)EthnicityWhite258 938 (93%)63 690 (92%)65 280 (94%)65 280 (94%)64 688 (93%)Non-White18 480 (6.7%)5 665 (8.2%)4 075 (5.9%)4 074 (5.9%)4 666 (6.7%)EducationHigh101 831 (37%)29 608 (43%)27 321 (39%)24 648 (36%)20 254 (29%)Intermediate140 320 (51%)34 207 (49%)34 817 (50%)35 354 (51%)35 942 (52%)Low35 267 (13%)5 540 (8.0%)7 217 (10%)9 352 (13%)13 158 (19%)Townsend deprivation index-2.30 [-3.72, 0.14]-2.44 [-3.81, -0.13]-2.42 [-3.80, -0.11]-2.31 [-3.72, 0.13]-2.00 [-3.54, 0.67]Smoking statusNever154 499 (56%)42 232 (61%)39 417 (57%)37 400 (54%)35 450 (51%)Previous95 906 (35%)23 112 (33%)24 517 (35%)24 699 (36%)23 578 (34%)Current27 013 (9.7%)4 011 (5.8%)5 421 (7.8%)7 255 (10%)10 326 (15%)Alcohol intake frequency < 3 times/week129 900 (47%)29 370 (42%)30 710 (44%)32 931 (47%)36 889 (53%) ≥ 3 times/week129 189 (47%)37 072 (53%)35 083 (51%)31 781 (46%)25 253 (36%)Never18 329 (6.6%)2 913 (4.2%)3 562 (5.1%)4 642 (6.7%)7 212 (10%)Diet risk score5.44 (1.52)5.56 (1.52)5.49 (1.52)5.42 (1.52)5.30 (1.53)Sleep duration < 6 h13 098 (4.7%)2 672 (3.9%)2 829 (4.1%)3 266 (4.7%)4 331 (6.2%) > 9 h4 124 (1.5%)641 (0.9%)867 (1.3%)1 110 (1.6%)1 506 (2.2%)6–9 h260 196 (94%)66 042 (95%)65 659 (95%)64 978 (94%)63 517 (92%)Meets PA recommendations152 566 (55%)41 709 (60%)38 954 (56%)37 081 (53%)34 822 (50%)Body mass index, kg/m226.54 [24.02, 29.59]25.70 [23.53, 28.14]26.16 [23.79, 28.91]26.74 [24.11, 29.83]27.98 [24.92, 31.84]Waist circumference, cm89.86 (13.23)88.24 (11.68)89.13 (12.88)89.86 (13.57)92.21 (14.34)Triglycerides, mmol/L1.47 [1.03, 2.13]1.14 [0.83, 1.62]1.37 [0.99, 1.96]1.54 [1.11, 2.21]1.89 [1.37, 2.66]FPG, mmol/L4.92 [4.60, 5.29]4.86 [4.53, 5.20]4.90 [4.57, 5.25]4.93 [4.61, 5.31]5.00 [4.67, 5.45]LDL-C, mmol/L3.59 (0.85)3.41 (0.79)3.54 (0.81)3.64 (0.84)3.77 (0.90)Creatinine, µmol/L70.0 [63.0, 82.0]83.7 [76.1, 91.9]74.0 [67.1, 81.5]66.9 [60.7, 74.0]59.5 [53.9, 65.9]Cystatin C, mg/L0.88 [0.80, 0.98]0.84 [0.77, 0.91]0.87 [0.79, 0.96]0.88 [0.80, 0.97]0.93 [0.84, 1.03]TyG index8.70 (0.56)8.68 (0.57)8.68 (0.57)8.68 (0.56)8.76 (0.55)MLI1.27 (0.23)1.00 (0.08)1.18 (0.04)1.43 (0.05)1.58 (0.17)TyG–MLI9.91 [8.26, 11.92]7.75 [7.22, 8.26]8.99 [8.56, 9.46]10.08 [9.59, 10.60]11.92 [11.15, 12.95]Hypertension150 375 (54%)31 528 (45%)35 443 (51%)39 022 (56%)44 382 (64%)Diabetes17 083 (6.2%)2 057 (3.0%)3 003 (4.3%)4 326 (6.2%)7 697 (11%)Values are presented as mean ± standard deviation or median [interquartile range] for continuous variables, and number (percentage) for categorical variablesAbbreviations: FPG, fasting plasma glucose; LDL-C, low-density lipoprotein cholesterol; MLI, muscle-loss index; PA, physical activity; TyG, triglyceride–glucose; TyG–MLI, triglyceride–glucose muscle-loss index Distribution of TyG–MLI in the cohort. Histogram showing the distribution of TyG–MLI among 277,418 participants. Participants were categorized into quartiles: Q1 ≤ 8.41; Q2 = 8.42–9.53; Q3 = 9.54–10.87; Q4 ≥ 10.88. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index Table 1Baseline characteristics of participants according to quartiles of TyG–MLIVariableTotal(n = 277 418)Q1(n = 69 355)Q2(n = 69 355)Q3(n = 69 354)Q4(n = 69 354)Age, years56.01 (8.08)53.15 (8.14)55.25 (8.09)56.87 (7.81)58.76 (7.18)SexFemale146 671 (53%)21 736 (31%)30 801 (44%)40 432 (58%)53 702 (77%)Male130 747 (47%)47 619 (69%)38 554 (56%)28 922 (42%)15 652 (23%)EthnicityWhite258 938 (93%)63 690 (92%)65 280 (94%)65 280 (94%)64 688 (93%)Non-White18 480 (6.7%)5 665 (8.2%)4 075 (5.9%)4 074 (5.9%)4 666 (6.7%)EducationHigh101 831 (37%)29 608 (43%)27 321 (39%)24 648 (36%)20 254 (29%)Intermediate140 320 (51%)34 207 (49%)34 817 (50%)35 354 (51%)35 942 (52%)Low35 267 (13%)5 540 (8.0%)7 217 (10%)9 352 (13%)13 158 (19%)Townsend deprivation index-2.30 [-3.72, 0.14]-2.44 [-3.81, -0.13]-2.42 [-3.80, -0.11]-2.31 [-3.72, 0.13]-2.00 [-3.54, 0.67]Smoking statusNever154 499 (56%)42 232 (61%)39 417 (57%)37 400 (54%)35 450 (51%)Previous95 906 (35%)23 112 (33%)24 517 (35%)24 699 (36%)23 578 (34%)Current27 013 (9.7%)4 011 (5.8%)5 421 (7.8%)7 255 (10%)10 326 (15%)Alcohol intake frequency < 3 times/week129 900 (47%)29 370 (42%)30 710 (44%)32 931 (47%)36 889 (53%) ≥ 3 times/week129 189 (47%)37 072 (53%)35 083 (51%)31 781 (46%)25 253 (36%)Never18 329 (6.6%)2 913 (4.2%)3 562 (5.1%)4 642 (6.7%)7 212 (10%)Diet risk score5.44 (1.52)5.56 (1.52)5.49 (1.52)5.42 (1.52)5.30 (1.53)Sleep duration < 6 h13 098 (4.7%)2 672 (3.9%)2 829 (4.1%)3 266 (4.7%)4 331 (6.2%) > 9 h4 124 (1.5%)641 (0.9%)867 (1.3%)1 110 (1.6%)1 506 (2.2%)6–9 h260 196 (94%)66 042 (95%)65 659 (95%)64 978 (94%)63 517 (92%)Meets PA recommendations152 566 (55%)41 709 (60%)38 954 (56%)37 081 (53%)34 822 (50%)Body mass index, kg/m226.54 [24.02, 29.59]25.70 [23.53, 28.14]26.16 [23.79, 28.91]26.74 [24.11, 29.83]27.98 [24.92, 31.84]Waist circumference, cm89.86 (13.23)88.24 (11.68)89.13 (12.88)89.86 (13.57)92.21 (14.34)Triglycerides, mmol/L1.47 [1.03, 2.13]1.14 [0.83, 1.62]1.37 [0.99, 1.96]1.54 [1.11, 2.21]1.89 [1.37, 2.66]FPG, mmol/L4.92 [4.60, 5.29]4.86 [4.53, 5.20]4.90 [4.57, 5.25]4.93 [4.61, 5.31]5.00 [4.67, 5.45]LDL-C, mmol/L3.59 (0.85)3.41 (0.79)3.54 (0.81)3.64 (0.84)3.77 (0.90)Creatinine, µmol/L70.0 [63.0, 82.0]83.7 [76.1, 91.9]74.0 [67.1, 81.5]66.9 [60.7, 74.0]59.5 [53.9, 65.9]Cystatin C, mg/L0.88 [0.80, 0.98]0.84 [0.77, 0.91]0.87 [0.79, 0.96]0.88 [0.80, 0.97]0.93 [0.84, 1.03]TyG index8.70 (0.56)8.68 (0.57)8.68 (0.57)8.68 (0.56)8.76 (0.55)MLI1.27 (0.23)1.00 (0.08)1.18 (0.04)1.43 (0.05)1.58 (0.17)TyG–MLI9.91 [8.26, 11.92]7.75 [7.22, 8.26]8.99 [8.56, 9.46]10.08 [9.59, 10.60]11.92 [11.15, 12.95]Hypertension150 375 (54%)31 528 (45%)35 443 (51%)39 022 (56%)44 382 (64%)Diabetes17 083 (6.2%)2 057 (3.0%)3 003 (4.3%)4 326 (6.2%)7 697 (11%)Values are presented as mean ± standard deviation or median [interquartile range] for continuous variables, and number (percentage) for categorical variablesAbbreviations: FPG, fasting plasma glucose; LDL-C, low-density lipoprotein cholesterol; MLI, muscle-loss index; PA, physical activity; TyG, triglyceride–glucose; TyG–MLI, triglyceride–glucose muscle-loss index Baseline characteristics of participants according to quartiles of TyG–MLI Values are presented as mean ± standard deviation or median [interquartile range] for continuous variables, and number (percentage) for categorical variables Abbreviations: FPG, fasting plasma glucose; LDL-C, low-density lipoprotein cholesterol; MLI, muscle-loss index; PA, physical activity; TyG, triglyceride–glucose; TyG–MLI, triglyceride–glucose muscle-loss index Follow-up and incidence rates of cardiovascular outcomes The median follow-up duration was 15.1 years (IQR: 14.2–15.8 years), during which 40,398 composite CVD events occurred among 277,418 participants. The crude incidence rates per 1,000 person-years were 10.46 for CVD, 6.06 for CHD, 1.68 for stroke, 1.91 for HF, 2.70 for AF, and 1.35 for PAD. When stratified by TyG–MLI quartiles, event rates rose stepwise across quartiles (Supplementary Table S6). For example, composite CVD increased from 7.90 per 1000 person-years in Q1 to 13.70 in Q4. Figure 3 shows the Kaplan–Meier cumulative incidence curves, with clear separation across quartiles (log-rank p < 0.001), indicating a sustained elevation in CVD risk at higher TyG–MLI levels.Fig. 3Kaplan–Meier curves for incident CVD by TyG–MLI quartiles. Cumulative incidence curves are shown for TyG–MLI quartiles. Higher TyG–MLI was consistently associated with increased risk, with log-rank p < 0.001 for all outcomes. A CVD; B CHD; C stroke; D HF; E PAD; F AF. Numbers at risk are shown below each panel. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index; CVD, cardiovascular disease; CHD, coronary heart disease; HF, heart failure; PAD, peripheral artery disease; AF, atrial fibrillation Kaplan–Meier curves for incident CVD by TyG–MLI quartiles. Cumulative incidence curves are shown for TyG–MLI quartiles. Higher TyG–MLI was consistently associated with increased risk, with log-rank p < 0.001 for all outcomes. A CVD; B CHD; C stroke; D HF; E PAD; F AF. Numbers at risk are shown below each panel. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index; CVD, cardiovascular disease; CHD, coronary heart disease; HF, heart failure; PAD, peripheral artery disease; AF, atrial fibrillation Associations of TyG–MLI with cardiovascular outcomes We then assessed the independent associations of TyG–MLI with cardiovascular outcomes using Cox proportional hazards models (Table 2). After full adjustment for demographic, lifestyle, and clinical factors, each 1-SD increase in TyG–MLI was associated with a 17% higher risk of composite CVD (HR: 1.17; 95% CI: 1.16–1.18). When TyG–MLI was categorized into quartiles, participants in Q4 had a 48% higher risk of CVD compared with those in Q1 (HR: 1.48; 95% CI: 1.44–1.53). Restricted cubic spline analyses revealed a nonlinear dose–response association, with risks rising steeply at higher TyG–MLI levels (p for nonlinearity < 0.001; Fig. 4A). Similar patterns were observed for individual cardiovascular outcomes. For CHD, each 1-SD increase in TyG–MLI conferred a 15% higher risk (HR: 1.15; 95% CI: 1.13–1.16), with a 43% higher risk in Q4 versus Q1 (HR: 1.43; 95% CI: 1.38–1.49; Fig. 4B). For stroke, the corresponding HR per 1-SD was 1.14 (95% CI: 1.11–1.17), and the highest quartile had a 37% elevated risk (HR: 1.37; 95% CI: 1.27–1.49; Fig. 4C). Notably, the magnitude of association was stronger for HF, where each 1-SD increase was associated with a 27% higher risk (HR: 1.27; 95% CI: 1.25–1.29), and those in Q4 had a two-fold higher risk compared with Q1 (HR: 2.00; 95% CI: 1.86–2.15; Fig. 4D). For AF, TyG–MLI was consistently associated with increased risk, with HR per 1-SD of 1.17 (95% CI: 1.15–1.19) and a 42% excess risk in Q4 versus Q1 (HR: 1.42; 95% CI: 1.34–1.51; Fig. 4E). Likewise, a strong graded association was observed for PAD, with HR per 1-SD of 1.22 (95% CI: 1.20–1.25) and nearly an 80% higher risk in Q4 versus Q1 (HR: 1.80; 95% CI: 1.65–1.97; Fig. 4F).Table 2Associations of TyG–MLI with incident cardiovascular outcomesModel 1Model 2Model 3HR (95%CI)p-valueHR (95%CI)p-valueHR (95%CI)p-valueComposite CVDTyG–MLI per 1 SD1.22 (1.21–1.23) < 0.0011.20 (1.19–1.21) < 0.0011.17 (1.16–1.18) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.22 (1.18–1.25) < 0.0011.12 (1.09–1.16) < 0.0011.10 (1.07–1.13) < 0.001TyG–MLI Q31.37 (1.33–1.41) < 0.0011.25 (1.21–1.28) < 0.0011.18 (1.15–1.22) < 0.001TyG–MLI Q41.74 (1.69–1.79) < 0.0011.63 (1.58–1.68) < 0.0011.48 (1.44–1.53) < 0.001CHDTyG–MLI per 1 SD1.20 (1.18–1.21) < 0.0011.20 (1.18–1.21) < 0.0011.15 (1.13–1.16) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.21 (1.17–1.26) < 0.0011.15 (1.10–1.19) < 0.0011.11 (1.07–1.15) < 0.001TyG–MLI Q31.35 (1.30–1.40) < 0.0011.29 (1.24–1.34) < 0.0011.19 (1.15–1.24) < 0.001TyG–MLI Q41.64 (1.58–1.70) < 0.0011.65 (1.59–1.72) < 0.0011.43 (1.38–1.59) < 0.001StrokeTyG–MLI per 1 SD1.24 (1.22–1.27) < 0.0011.18 (1.15–1.20) < 0.0011.14 (1.11–1.17) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.22 (1.14–1.32) < 0.0011.09 (1.01–1.17)0.0231.07 (0.99–1.15)0.085TyG–MLI Q31.39 (1.30–1.50) < 0.0011.18 (1.09–1.27) < 0.0011.12 (1.04–1.21)0.004TyG–MLI Q41.87 (1.74–2.00) < 0.0011.52 (1.40–1.64) < 0.0011.37 (1.27–1.49) < 0.001HFTyG–MLI per 1 SD1.33 (1.32–1.35) < 0.0011.29 (1.27–1.30) < 0.0011.27 (1.25–1.29) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.29 (1.20–1.39) < 0.0011.14 (1.05–1.23) < 0.0011.12 (1.04–1.21)0.002TyG–MLI Q31.67 (1.55–1.79) < 0.0011.41 (1.31–1.52) < 0.0011.36 (1.26–1.46) < 0.001TyG–MLI Q42.56 (2.40–2.74) < 0.0012.19 (2.03–2.35) < 0.0012.00 (1.86–2.15) < 0.001AFTyG–MLI per 1 SD1.19 (1.17–1.21) < 0.0011.17 (1.15–1.19) < 0.0011.17 (1.15–1.19) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.20 (1.13–1.27) < 0.0011.07 (1.01–1.13)0.0211.07 (1.01–1.14)0.015TyG–MLI Q31.36 (1.28–1.44) < 0.0011.20 (1.13–1.27) < 0.0011.20 (1.13–1.27) < 0.001TyG–MLI Q41.57 (1.48–1.65) < 0.0011.44 (1.35–1.53) < 0.0011.42 (1.34–1.51) < 0.001PADTyG–MLI per 1 SD1.30 (1.28–1.32) < 0.0011.24 (1.22–1.27) < 0.0011.22 (1.20–1.25) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.32 (1.21–1.44) < 0.0011.16 (1.06–1.26) < 0.0011.14 (1.04–1.24)0.002TyG–MLI Q31.62 (1.49–1.77) < 0.0011.36 (1.25–1.48) < 0.0011.29 (1.19–1.41) < 0.001TyG–MLI Q42.33 (2.15–2.52) < 0.0011.99 (1.83–2.17) < 0.0011.80 (1.65–1.97) < 0.001Model 1 was unadjusted. Model 2 was adjusted for sociodemographic and lifestyle factors, including age, sex, ethnicity, education, Townsend deprivation index, smoking status, alcohol intake, dietary score, sleep duration, and physical activity. Model 3 was further adjusted for clinical risk factors, including hypertension, diabetes, and low-density lipoprotein cholesterolAbbreviations: AF, atrial fibrillation; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HF, heart failure; HR, hazard ratio; PAD, peripheral artery disease; SD, standard deviation; TyG–MLI, triglyceride–glucose muscle-loss indexFig. 4Dose–response associations of TyG–MLI with CVD outcomes. Restricted cubic spline curves from Cox proportional hazards models show associations of TyG–MLI with CVD outcomes. Solid lines represent hazard ratios and shaded areas indicate 95% CIs. P values for overall and nonlinearity are shown. Models adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking, alcohol, diet, sleep, physical activity, hypertension, diabetes, and LDL-C. A CVD; B CHD; C stroke; D HF; E AF; F PAD. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index; CI, confidence interval; CVD, cardiovascular disease; CHD, coronary heart disease; HF, heart failure; AF, atrial fibrillation; PAD, peripheral artery disease; LDL-C, low-density lipoprotein cholesterol Table 2Associations of TyG–MLI with incident cardiovascular outcomesModel 1Model 2Model 3HR (95%CI)p-valueHR (95%CI)p-valueHR (95%CI)p-valueComposite CVDTyG–MLI per 1 SD1.22 (1.21–1.23) < 0.0011.20 (1.19–1.21) < 0.0011.17 (1.16–1.18) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.22 (1.18–1.25) < 0.0011.12 (1.09–1.16) < 0.0011.10 (1.07–1.13) < 0.001TyG–MLI Q31.37 (1.33–1.41) < 0.0011.25 (1.21–1.28) < 0.0011.18 (1.15–1.22) < 0.001TyG–MLI Q41.74 (1.69–1.79) < 0.0011.63 (1.58–1.68) < 0.0011.48 (1.44–1.53) < 0.001CHDTyG–MLI per 1 SD1.20 (1.18–1.21) < 0.0011.20 (1.18–1.21) < 0.0011.15 (1.13–1.16) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.21 (1.17–1.26) < 0.0011.15 (1.10–1.19) < 0.0011.11 (1.07–1.15) < 0.001TyG–MLI Q31.35 (1.30–1.40) < 0.0011.29 (1.24–1.34) < 0.0011.19 (1.15–1.24) < 0.001TyG–MLI Q41.64 (1.58–1.70) < 0.0011.65 (1.59–1.72) < 0.0011.43 (1.38–1.59) < 0.001StrokeTyG–MLI per 1 SD1.24 (1.22–1.27) < 0.0011.18 (1.15–1.20) < 0.0011.14 (1.11–1.17) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.22 (1.14–1.32) < 0.0011.09 (1.01–1.17)0.0231.07 (0.99–1.15)0.085TyG–MLI Q31.39 (1.30–1.50) < 0.0011.18 (1.09–1.27) < 0.0011.12 (1.04–1.21)0.004TyG–MLI Q41.87 (1.74–2.00) < 0.0011.52 (1.40–1.64) < 0.0011.37 (1.27–1.49) < 0.001HFTyG–MLI per 1 SD1.33 (1.32–1.35) < 0.0011.29 (1.27–1.30) < 0.0011.27 (1.25–1.29) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.29 (1.20–1.39) < 0.0011.14 (1.05–1.23) < 0.0011.12 (1.04–1.21)0.002TyG–MLI Q31.67 (1.55–1.79) < 0.0011.41 (1.31–1.52) < 0.0011.36 (1.26–1.46) < 0.001TyG–MLI Q42.56 (2.40–2.74) < 0.0012.19 (2.03–2.35) < 0.0012.00 (1.86–2.15) < 0.001AFTyG–MLI per 1 SD1.19 (1.17–1.21) < 0.0011.17 (1.15–1.19) < 0.0011.17 (1.15–1.19) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.20 (1.13–1.27) < 0.0011.07 (1.01–1.13)0.0211.07 (1.01–1.14)0.015TyG–MLI Q31.36 (1.28–1.44) < 0.0011.20 (1.13–1.27) < 0.0011.20 (1.13–1.27) < 0.001TyG–MLI Q41.57 (1.48–1.65) < 0.0011.44 (1.35–1.53) < 0.0011.42 (1.34–1.51) < 0.001PADTyG–MLI per 1 SD1.30 (1.28–1.32) < 0.0011.24 (1.22–1.27) < 0.0011.22 (1.20–1.25) < 0.001TyG–MLI Q1RefRefRefRefRefRefTyG–MLI Q21.32 (1.21–1.44) < 0.0011.16 (1.06–1.26) < 0.0011.14 (1.04–1.24)0.002TyG–MLI Q31.62 (1.49–1.77) < 0.0011.36 (1.25–1.48) < 0.0011.29 (1.19–1.41) < 0.001TyG–MLI Q42.33 (2.15–2.52) < 0.0011.99 (1.83–2.17) < 0.0011.80 (1.65–1.97) < 0.001Model 1 was unadjusted. Model 2 was adjusted for sociodemographic and lifestyle factors, including age, sex, ethnicity, education, Townsend deprivation index, smoking status, alcohol intake, dietary score, sleep duration, and physical activity. Model 3 was further adjusted for clinical risk factors, including hypertension, diabetes, and low-density lipoprotein cholesterolAbbreviations: AF, atrial fibrillation; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HF, heart failure; HR, hazard ratio; PAD, peripheral artery disease; SD, standard deviation; TyG–MLI, triglyceride–glucose muscle-loss index Associations of TyG–MLI with incident cardiovascular outcomes Model 1 was unadjusted. Model 2 was adjusted for sociodemographic and lifestyle factors, including age, sex, ethnicity, education, Townsend deprivation index, smoking status, alcohol intake, dietary score, sleep duration, and physical activity. Model 3 was further adjusted for clinical risk factors, including hypertension, diabetes, and low-density lipoprotein cholesterol Abbreviations: AF, atrial fibrillation; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HF, heart failure; HR, hazard ratio; PAD, peripheral artery disease; SD, standard deviation; TyG–MLI, triglyceride–glucose muscle-loss index Dose–response associations of TyG–MLI with CVD outcomes. Restricted cubic spline curves from Cox proportional hazards models show associations of TyG–MLI with CVD outcomes. Solid lines represent hazard ratios and shaded areas indicate 95% CIs. P values for overall and nonlinearity are shown. Models adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking, alcohol, diet, sleep, physical activity, hypertension, diabetes, and LDL-C. A CVD; B CHD; C stroke; D HF; E AF; F PAD. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index; CI, confidence interval; CVD, cardiovascular disease; CHD, coronary heart disease; HF, heart failure; AF, atrial fibrillation; PAD, peripheral artery disease; LDL-C, low-density lipoprotein cholesterol Joint associations of TyG index and MLI with incident CVD To further illustrate the rationale for constructing TyG–MLI and to disentangle the individual and combined contributions of IR and muscle loss, we classified participants into four joint categories using median cutoffs for TyG index and MLI (Supplementary Table S7). In fully adjusted models, compared with individuals with both low TyG and low MLI, those with high TyG index alone (HR = 1.12; 95% CI 1.09–1.15) and those with high MLI alone (HR = 1.23; 95% CI 1.19–1.27) demonstrated modest yet significant increases in incident CVD risk. Participants with high levels of both TyG index and MLI exhibited the greatest risk (HR = 1.33; 95% CI 1.29–1.37), forming a clear stepwise gradient. This pattern supports the conceptual foundation of TyG–MLI, indicating that the coexistence of metabolic stress and muscle depletion confers a greater cardiovascular burden than either component considered separately, thereby justifying the construction of a composite index to capture their joint effect. BMI-stratified analyses of TyG–MLI and TyG index for CVD risk In prespecified BMI-stratified analyses, TyG–MLI showed stronger associations with CVD than TyG index across BMI categories (Fig. 5). Among underweight participants, each 1-SD increase in TyG–MLI was associated with higher CVD risk (HR: 1.11; 95% CI: 1.02–1.22), whereas TyG index showed no significant association (HR: 1.00; 95% CI: 0.83–1.19). Quartile analyses indicated that only TyG–MLI, particularly at Q4, predicted elevated risk. In normal-weight participants, TyG–MLI demonstrated a strong association per 1-SD increase (HR: 1.16; 95% CI: 1.14–1.19), while TyG index showed only a marginal effect (HR: 1.02; 95% CI: 1.00–1.05). Quartile analyses further confirmed a graded risk increase for TyG–MLI, absent for TyG index. In overweight participants, both indices were significantly associated with CVD, but TyG–MLI conferred a stronger effect size (HR per 1-SD: 1.16; 95% CI: 1.15–1.17) compared with TyG index (HR: 1.06; 95% CI: 1.05–1.07). Quartile analyses also revealed steeper gradients for TyG–MLI. We additionally compared the discriminatory performance of TyG–MLI and TyG index in the overall cohort and within BMI strata (Supplementary Table S8). Overall AUC values were broadly similar, although TyG–MLI showed lower discrimination than TyG index in the total population. Within BMI categories, however, discrimination patterns paralleled the association results—more favorable for TyG–MLI in underweight individuals, comparable in normal-weight participants, and slightly lower in those with overweight.Fig. 5Associations of TyG index and TyG–MLI with CVD stratified by BMI. HRs and 95% CIs for CVD are shown for the TyG index and TyG–MLI, stratified by BMI: underweight (< 18.5 kg/m2), normal (18.5–24.0 kg/m2), and overweight (≥ 24.0 kg/m2). Models adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking, alcohol, diet, sleep, physical activity, hypertension, diabetes, and LDL-C. Abbreviations: BMI, body mass index; HR, hazard ratio; CI, confidence interval; TyG, triglyceride–glucose; TyG–MLI, triglyceride–glucose muscle-loss index; CVD, cardiovascular disease; LDL-C, low-density lipoprotein cholesterol Associations of TyG index and TyG–MLI with CVD stratified by BMI. HRs and 95% CIs for CVD are shown for the TyG index and TyG–MLI, stratified by BMI: underweight (< 18.5 kg/m2), normal (18.5–24.0 kg/m2), and overweight (≥ 24.0 kg/m2). Models adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking, alcohol, diet, sleep, physical activity, hypertension, diabetes, and LDL-C. Abbreviations: BMI, body mass index; HR, hazard ratio; CI, confidence interval; TyG, triglyceride–glucose; TyG–MLI, triglyceride–glucose muscle-loss index; CVD, cardiovascular disease; LDL-C, low-density lipoprotein cholesterol Subgroup analyses Subgroup analyses showed that higher TyG–MLI was consistently associated with increased CVD risk across all strata defined by age, sex, hypertension, and diabetes (Fig. 6). To evaluate effect modification, multiplicative interaction was tested using likelihood ratio comparisons. Given four subgroup tests, a Bonferroni-corrected significance threshold of p < 0.0125 was applied, under which all interactions reached statistical significance. However, after controlling the FDR, only age and sex interactions remained significant (both FDR-adjusted p < 0.001). Notably, the strength of association was greater among adults < 60 years (HR for Q4 vs. Q1: 1.83; 95% CI: 1.73–1.92) than among those ≥ 60 years (HR: 1.47; 95% CI: 1.41–1.53), and slightly stronger in women (HR for Q4 vs. Q1: 1.53; 95% CI: 1.43–1.64) compared with men (HR: 1.48; 95% CI: 1.43–1.54).Fig. 6Subgroup analyses of TyG–MLI and incident CVD. HRs and 95% CIs for incident CVD across quartiles of TyG–MLI are shown for subgroups defined by age, sex, hypertension, and diabetes. Models adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking, alcohol, diet, sleep, physical activity, hypertension, diabetes, and LDL-C. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index; HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease; LDL-C, low-density lipoprotein cholesterol Subgroup analyses of TyG–MLI and incident CVD. HRs and 95% CIs for incident CVD across quartiles of TyG–MLI are shown for subgroups defined by age, sex, hypertension, and diabetes. Models adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking, alcohol, diet, sleep, physical activity, hypertension, diabetes, and LDL-C. Abbreviations: TyG–MLI, triglyceride–glucose muscle-loss index; HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease; LDL-C, low-density lipoprotein cholesterol Sensitivity analyses To evaluate the robustness of our findings, we conducted a series of sensitivity analyses. First, to account for the competing risk of non-cardiovascular death, we applied Fine–Gray subdistribution hazard models. The association between TyG–MLI and incident CVD remained strong, with each 1-SD increase corresponding to a 17% higher subdistribution hazard (sHR = 1.17, 95% CI 1.17–1.19), and participants in the highest quartile showing a 48% higher hazard compared with those in the lowest quartile (sHR = 1.48, 95% CI 1.43–1.54) (Supplementary Table S9). Second, to reduce potential reverse causality, we performed a 2-year landmark analysis excluding events occurring within the first two years of follow-up, which yielded results consistent with the primary findings (Supplementary Table S10). Third, to address possible selection bias arising from covariate missingness, we repeated the analyses using multiple imputation, and the associations remained materially unchanged (Supplementary Table S11). Fourth, because MLI may be influenced by renal function and TyG index may be affected by statin use, we additionally adjusted for eGFR and statin therapy. We also constructed a height-adjusted TyG–MLI to account for body size differences. All models produced directionally consistent and statistically robust associations (Supplementary Tables S12–S14). Sensitivity analyses using BMI ≥ 25 kg/m2 yielded comparable results (Supplementary Tables S15). Finally, although external validation was not feasible, we performed internal geographic validation across England, Scotland, and Wales. Region-stratified models demonstrated uniformly significant associations (Supplementary Table S16). Discussion In this large prospective cohort of more than 270,000 adults from the UK Biobank, we examined a novel composite biomarker—TyG–MLI—that integrates information on IR (via TyG index) and muscle status (via CysC-to-Cr ratio). Higher TyG–MLI levels were independently associated with increased risks of incident CVD and all major CVD subtypes. The associations were graded across the distribution, showed evidence of non-linearity at higher levels, and remained robust across extensive sensitivity and subgroup analyses. Notably, the association of TyG–MLI with CVD was stronger than that of TyG index alone among underweight and normal-weight participants. TyG index has been extensively studied as a simple surrogate of IR, with robust evidence linking it to incident CVD and cardiovascular mortality in diverse populations [19–21]. Several TyG-derived composites, such as TyG–BMI and TyG–waist circumference, have also been shown to improve predictive performance by capturing obesity-related pathways [7, 8, 22]. However, these indices are fundamentally anchored in adiposity and therefore less effective in non-obese individuals, who may nonetheless harbor significant cardiometabolic vulnerability. Parallel evidence has underscored the prognostic value of muscle-related biomarkers. The Cr-to-CysC ratio, also known as SI, has been consistently associated with frailty [23], sarcopenia [24, 25], and CVD [10, 26]. These data collectively suggest that both metabolic impairment and skeletal-muscle depletion independently contribute to cardiovascular risk, yet have rarely been evaluated within a unified framework. Despite growing evidence that the two conditions are mechanistically and epidemiologically intertwined, prior biomarkers have evaluated these domains in isolation, thereby overlooking individuals in whom concurrent metabolic stress and muscle deficiency coexist. No prior biomarker has combined these domains into a single construct. By introducing the TyG–MLI, our study addresses this gap and shows that jointly assessing metabolic stress and muscle loss provides important insights into CVD risk, particularly in non-obese populations. Our findings therefore extend existing literature by demonstrating that integrating IR and muscle-loss signatures may help identify vulnerable phenotypes that traditional TyG-derived or SI-based indices alone may fail to capture. The conceptual rationale for constructing TyG–MLI is supported by biological, methodological, and empirical considerations. Biologically, TyG index reflects metabolic impairment and IR, whereas MLI (CysC/Cr) captures relative skeletal-muscle depletion while accounting for renal function, and corresponds to the reciprocal of SI, with omission of the scaling factor not altering its biological meaning. Integrating TyG and MLI therefore enables the identification of individuals who concurrently exhibit metabolic dysfunction and diminished muscle reserve—a physiologically vulnerable profile linked to adverse cardiovascular outcomes. Methodologically, the multiplicative structure aligns with established TyG-derived composite measures, including TyG–BMI and TyG–waist circumference, which quantify the synergistic burden arising from coexisting cardiometabolic abnormalities and acknowledge that metabolic dysregulation and body-composition deficits often interact rather than act independently. Empirically, our joint-classification analysis demonstrated a clear stepwise gradient in CVD risk, with the highest hazards observed among participants with both high TyG and high MLI, supporting the plausibility of a multiplicative formulation and reinforcing the notion that concurrent metabolic stress and muscle depletion confer a greater risk than either component alone. Taken together, these considerations indicate that TyG–MLI functions as an integrative biomarker capturing the dual burden of IR and muscle loss and may provide additional insight into cardiometabolic vulnerability. BMI-stratified analyses offer important clarification on the context in which TyG–MLI is most informative. Across BMI categories, TyG–MLI showed clear and graded associations with incident CVD, but the magnitude of association varied meaningfully by body size. Compared with TyG index, TyG–MLI was more strongly associated with CVD in underweight and normal-weight individuals, suggesting that reliance on TyG index alone may underestimate risk in non-obese populations. Incorporating the muscle-loss component appears to capture an additional dimension of metabolic vulnerability that becomes particularly relevant when adiposity is low. However, the underweight subgroup included relatively few participants, leading to wider CIs and reduced statistical precision. Accordingly, estimates in this subgroup should be interpreted with caution. Findings from AUC analyses—secondary and not intended as a prediction-focused evaluation—showed parallel patterns of heterogeneity. Together, these observations indicate that the contribution of TyG–MLI is not uniform across the population. Its relevance appears greatest in individuals without obesity, where muscle-related metabolic impairment may carry greater cardiovascular significance and where TyG index alone may underestimate risk. These findings also highlight the need for cautious interpretation in small subgroups and for more rigorous prognostic evaluation in future research. Sex‐stratified analyses further demonstrated that the association between TyG–MLI and incident CVD was consistently stronger in women than in men, even after identical multivariable adjustment. This observation is biologically plausible: women generally have lower baseline skeletal-muscle mass, greater vulnerability to metabolic dysregulation, and distinct hormonal and inflammatory profiles, all of which may amplify the adverse consequences of reduced muscle reserve [27]. In this context, a higher TyG–MLI may represent a more sensitive marker of metabolic–muscular imbalance in women, thereby translating into greater relative cardiovascular risk. Similarly, age-stratified analyses revealed that younger participants (< 60 years) exhibited stronger TyG–MLI–CVD associations than their older counterparts. This pattern suggests that metabolic–muscular dysregulation may confer disproportionately greater risk earlier in adulthood, when competing risks are lower and pathogenic metabolic disturbances may exert more direct vascular effects. Together, these findings highlight important effect modification by sex and age, emphasizing the need to consider demographic context when interpreting TyG–MLI and supporting future work to elucidate potential age- and sex‐specific mechanisms linking muscle loss, IR, and cardiovascular risk. Several biological mechanisms may explain the observed associations between TyG–MLI and cardiovascular outcomes. First, the reciprocal relationship between IR and skeletal muscle loss creates a vicious cycle. IR reduces glucose uptake in muscle through impaired insulin receptor substrate signaling and decreased translocation of glucose transporter-4, thereby limiting energy supply for protein synthesis and accelerating proteolysis [12]. Conversely, declining muscle mass reduces the largest reservoir for glucose disposal, amplifying systemic hyperglycemia and perpetuating metabolic stress [28]. Second, both insulin resistance and sarcopenia converge on inflammatory and oxidative pathways. Chronic low-grade inflammation, characterized by elevated tumor necrosis factor-α, interleukin-6, and C-reactive protein, activates nuclear factor kappa B signaling and promotes endothelial dysfunction and vascular remodelling [29]. IR augments mitochondrial reactive oxygen species production, while sarcopenia is linked to impaired mitochondrial biogenesis and autophagy, further driving oxidative stress [30]. These processes accelerate atherosclerotic plaque development and destabilization. Third, muscle loss has systemic effects beyond classical metabolism. Sarcopenia reduces muscle-derived myokines such as irisin and myonectin, which normally exert cardioprotective, anti-inflammatory, and lipid-regulating effects. Loss of these signals may therefore exacerbate cardiometabolic dysfunction and vascular injury independent of glucose metabolism [9]. Fourth, sarcopenia contributes to reduced physiological reserve and frailty. Low muscle strength compromises cardiorespiratory fitness and vascular compliance, increasing susceptibility to hemodynamic stress and adverse outcomes independent of traditional risk factors [31, 32]. From a translational perspective, TyG–MLI shows potential as a clinically informative biomarker. It is derived from laboratory measurements hat are widely available in many healthcare systems and can be incorporated into routine assessment without major changes to existing workflows. TyG–MLI may be particularly useful for identifying individuals at elevated cardiovascular risk despite having normal or low BMI, a group in whom traditional adiposity-based approaches may underestimate metabolic vulnerability. For these individuals, integrating muscle-related information into risk assessment could offer meaningful clinical value. However, Cys C is not uniformly available across all primary-care or resource-limited settings, and its incorporation into large-scale screening programs would require careful consideration. Further work is needed to evaluate cost-effectiveness, feasibility, and potential implementation strategies in diverse healthcare environments. In addition, future studies should also aim to establish validated, population-specific cutoff values for TyG–MLI and assess their predictive utility in clinical practice. On a broader scale, TyG–MLI may also provide insights into the interplay between metabolic and musculoskeletal health, encouraging a more holistic approach to CVD prevention. Given global trends of aging, rising sarcopenia prevalence [33], and increasing cardiometabolic disease burden [34], integrative tools like TyG–MLI may help bridge clinical silos and guide early intervention strategies targeting both metabolic control and muscle preservation. This study has several limitations. First, the observational nature of the analysis precludes causal inference. Although extensive covariate adjustment was performed, residual or unmeasured confounding cannot be fully excluded, and the associations reported here should be interpreted as non-causal. Second, both TyG index and MLI were assessed only once at baseline. Single-timepoint biomarker measurement may not capture long-term metabolic or muscular fluctuations and could introduce nondifferential exposure misclassification, most commonly biasing associations toward the null. Third, MLI (CysC/Cr) is an indirect biochemical surrogate rather than a direct quantitative measure of muscle mass or sarcopenia. Accordingly, interpretations related to “muscle loss” should be made with caution. Fourth, although robustness was evaluated extensively through sensitivity analyses, subgroup analyses, and region-stratified models, the study lacked external validation in an independent cohort. Such internal consistency cannot substitute for external confirmation, and replication in diverse populations is needed before broader clinical application. Finally, more than 90% of UK Biobank participants are of White European ancestry. The limited ethnic diversity constrains the generalizability of our findings, and further validation in multi-ethnic cohorts is warranted. Conclusions In conclusion, TyG–MLI—a composite measure integrating IR and muscle loss—was independently associated with incident CVD in this large prospective cohort. Compared with TyG index, TyG–MLI appeared to offer additional insight into cardiometabolic risk, particularly among individuals without obesity. These findings suggest that incorporating muscle-related information may help capture aspects of cardiometabolic vulnerability not fully reflected by conventional metabolic markers. Further studies are warranted to evaluate the clinical utility, feasibility, and incremental prognostic value of TyG–MLI across diverse populations and clinical settings. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1