Associations of insulin resistance-related indices with the risk and progression of cardiometabolic multimorbidity in individuals with metabolic dysfunction-associated steatotic liver disease: a prospective cohort study

Abstract

Background

Insulin resistance (IR) is thought to be a major metabolic driver of both metabolic dysfunction-associated steatotic liver disease (MASLD) and cardiometabolic diseases (CMD). Although several IR-related indices have been linked to individual CMD, their associations with cardiometabolic multimorbidity (CMM) and stage-specific disease progression in individuals with MASLD remain unclear.

Methods

A total of 109,604 UK Biobank participants with MASLD who were free of CMD at baseline were included in this study. The analysis covered nine IR-related metrics, including the triglyceride-glucose (TyG) index, TyG-body mass index (TyG-BMI), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), TyG-body roundness index (TyG-BRI), TyG-a body shape index (TyG-ABSI), TyG-visceral adiposity index (TyG-VAI), TyG-weight-adjusted waist index (TyG-WWI), and the triglyceride-to-high-density lipoprotein cholesterol ratio (TG/HDL-C). Associations with incident CMM were estimated using Cox models. Multi-state models were applied to evaluate stage-specific transitions. Incremental predictive performance was evaluated using time-dependent ROC analyses, C-index, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Exploratory mediation analyses were further performed to explore possible biological pathways.

Results

Over a median follow-up of 15.9 years, 4944 participants developed CMM. Higher levels of all IR-related indices were linked to a greater risk of CMM. In the fully adjusted model, the strongest associations were observed for TyG-WHtR, TyG-BMI, and TyG-WC, with HRs (95% CIs) of 2.70 (2.45–2.97), 2.36 (2.16–2.58), and 2.33 (2.12–2.56), respectively, for the highest versus lowest quartile. Multistate analyses indicated that these indices showed stage-specific associations across the CMM trajectory. For transitions from a CMD-free state to single CMDs, the strongest associations were observed for T2D. Transitions from CHD or stroke were more likely to progress to CMM when exposed to higher IR-related indices. All indices modestly improved CMM prediction beyond conventional risk factors (all P < 0.001), with TyG-WHtR showing the best performance. Exploratory mediation analyses suggested that inflammatory, hepatic, and renal biomarkers jointly accounted for 11–23% of the associations.

Conclusions

Among individuals with MASLD, IR-related indices were significantly associated with the incidence and progression of CMM. Indices incorporating central adiposity, especially TyG-WHtR, provided the most informative risk estimates and modest incremental predictive value.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12933-026-03264-3.

Research Insights

What is currently known about this topic?

• λ IR is central to MASLD pathophysiology.

• λ Prior studies mainly examined single diseases rather than multimorbidity.

• λ Evidence on CMM progression in MASLD remains limited.

What is the key research question?

• λ Are IR-related indices associated with CMM onset and progression in MASLD?

• λ Which biomarker pathways may partly explain these associations?

What is new?

• λ All nine IR-related indices were associated with incident CMM and stage-specific progression in MASLD.

• λ TyG-WHtR and TyG-WC showed the strongest associations and risk-stratification performance.

• λ Inflammatory, hepatic, and renal biomarkers partly explained these associations.

How might this study influence clinical practice?

• λ IR-related indices, particularly TyG-WHtR and TyG-WC, may support early CMM risk stratification in MASLD.

Introduction

Metabolic dysfunction–associated steatotic liver disease (MASLD) has emerged as a common chronic metabolic disorder with substantial public health implications [1]. Rather than being confined to the liver, MASLD is now viewed as a systemic condition characterized by profound metabolic disturbances, including insulin resistance (IR), atherogenic dyslipidemia, and chronic low-grade inflammation [2, 3]. These abnormalities not only contribute to hepatic steatosis and disease progression, but also promote the development of cardiometabolic diseases (CMDs), particularly type 2 diabetes (T2D), coronary heart disease (CHD), and stroke [4, 5]. As a result, individuals with MASLD may be particularly susceptible to the accumulation of CMDs and the development of cardiometabolic multimorbidity (CMM), a state associated with substantial clinical burden and poor prognosis [6]. However, the trajectory of this progression has not been well characterized, highlighting the need for practical and reproducible markers to identify high-risk individuals and track disease progression.

Among the metabolic abnormalities underlying MASLD, IR is considered a central pathophysiological feature and a key mechanistic link between hepatic steatosis and CMD development [7, 8]. By promoting glucose dysregulation, lipid abnormalities, vascular injury, and systemic inflammation, IR may contribute not only to the onset of individual CMDs but also to the accumulation of multiple cardiometabolic conditions over time [5, 9]. The hyperinsulinemic–euglycemic clamp is considered the gold standard for quantifying IR, yet its invasive procedures, technical demands, and high cost preclude its use in routine clinical settings and large population-based studies. Alternative approaches, such as the homeostatic model assessment of IR (HOMA-IR), have been proposed, but their applicability is constrained by the need for insulin measurements and reduced accuracy in certain populations [10]. Consequently, increasing attention has been directed toward IR-related indices derived from routinely available clinical data. Among these, the triglyceride–glucose (TyG) index and its derivatives, including TyG-body mass index (TyG-BMI), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), TyG-body roundness index (TyG-BRI), TyG-a body shape index (TyG-ABSI), TyG-visceral adiposity index (TyG-VAI), and TyG-weight-adjusted waist index (TyG-WWI), as well as the triglyceride/high-density lipoprotein cholesterol (TG/HDL-C) ratio, have emerged as practical and cost-effective proxies for IR [11–13]. By integrating metabolic and anthropometric parameters, these composite indices capture a more comprehensive representation of IR and body fat distribution than single biomarkers, and have been validated in large-scale epidemiological studies as predictors of diverse cardiometabolic outcomes [14–18].

Previous studies have provided initial evidence linking IR-related indices to individual CMDs, mortality, and other adverse cardiometabolic outcomes [14–18]. However, evidence on their associations with incident CMM remains limited, and most available studies have focused on general populations rather than in individuals with MASLD. Importantly, CMM should be conceptualized not merely as the coexistence of multiple CMDs, but as a dynamic and progressive continuum. Individuals may progress from a cardiometabolic disease-free state to the occurrence of a first CMD, and subsequently to CMM, with each transition reflecting increasing metabolic complexity and clinical vulnerability [19–21]. Recent evidence has further emphasized the value of integrating metabolic dysfunction with broader indicators of physiological vulnerability for CMM risk stratification. For example, a national prospective cohort study showed that a combined cholesterol, high-density lipoprotein, glucose, and frailty index was associated with incident heart disease, stroke, diabetes, and CMM, and provided incremental predictive value for CMM risk assessment [22]. These findings support the concept that CMM development reflects the cumulative burden of metabolic dysregulation and systemic vulnerability rather than a single isolated event. Understanding this dynamic trajectory and its associated risk factors is therefore essential for both primary and secondary prevention, as well as for optimizing stage-specific clinical management [18, 23, 24]. In individuals with MASLD, IR may be particularly relevant across this continuum because it contributes to glucose dysregulation, atherogenic lipid profiles, systemic inflammation, vascular injury, and hepatic metabolic dysfunction [9, 25, 26]. However, whether routinely available IR-related indices are associated not only with incident CMM but also with stage-specific progression across the cardiometabolic disease trajectory remains unclear. In addition, although chronic inflammation and multi-organ dysfunction have been proposed as potential mechanisms linking IR-related indices to cardiometabolic outcomes [18, 23, 24], it is uncertain whether these biological pathways contribute to the development of CMM in individuals with MASLD.

Therefore, using data from the UK Biobank, the present study aimed to: (1) evaluate the associations of nine IR-related indices with incident CMM in individuals with MASLD; (2) compare their predictive performance for CMM risk; (3) conduct exploratory mediation analyses of baseline inflammatory, hepatic, and renal biomarkers that may statistically account for part of the observed associations; and (4) characterize their associations with stage-specific progression across the CMM trajectory.

Methods

Study design and participants

This study was based on the UK Biobank, a large prospective cohort in the United Kingdom that enrolled more than 500,000 participants between 2006 and 2010. Baseline information was systematically procured through comprehensive touchscreen-based assessments, physical examinations, and the acquisition of diverse biological specimens, with the detailed methodology of the cohort established in prior literature [27, 28]. The UK Biobank received approval from the North West Multicenter Research Ethics Committee, and written informed consent was obtained from all participants.

After excluding individuals without MASLD at baseline (N = 354,485), 147,451 participants with MASLD were identified from the overall UK Biobank cohort. We further excluded those with missing data on IR-related indices (N = 11,684), prevalent CHD, stroke, or T2D at baseline (N = 19,190), and missing covariate data (N = 6973), leaving 109,604 participants for the primary analyses. For mediation analyses, 4,066 participants with incomplete data on candidate mediators were additionally excluded, resulting in an analytic sample of 105,538. The flowchart of the analytic sample selection is shown in Fig. 1.

Fig. 1.

Flowchart of participant selection for the main and mediation analyses

Assessment of MASLD

MASLD was defined according to the 2023 multisociety Delphi consensus as hepatic steatosis accompanied by at least one cardiometabolic abnormality [29]. Because liver imaging and histological data were not available for the full UK Biobank cohort, the hepatic steatosis component of MASLD was assessed using the fatty liver index (FLI), a validated non-invasive steatosis index derived from body mass index, waist circumference, triglycerides, and gamma-glutamyl transferase [30–32]. An FLI ≥ 60 was used to indicate hepatic steatosis, consistent with the original FLI study and subsequent validation studies, with this threshold showing good diagnostic performance, including a reported sensitivity of 87% and specificity of 86% [30, 33]. Cardiometabolic abnormalities were determined based on established criteria and included the following conditions [34, 35]: (1) BMI ≥ 25 kg/m² or WC ≥ 90 cm in men and ≥ 80 cm in women; (2) blood pressure ≥ 130/85 mmHg or current antihypertensive treatment; (3) triglycerides ≥ 1.70 mmol/L or use of lipid-lowering agents; and (4) high-density lipoprotein cholesterol (HDL-C) ≤ 1.0 mmol/L in men and ≤ 1.3 mmol/L in women, or use of lipid-lowering therapy. (5) HbA1c ≥ 5.7% (39 mmol/mol), fasting glucose ≥ 5.6 mmol/L, 2-h post-load glucose ≥ 7.8 mmol/L. Given that participants with preexisting diabetes had been excluded at baseline, diabetes-related criteria for MASLD were not considered, including a diagnosis of T2D or the use of glucose-lowering medications [35].

Assessment of outcomes

The primary endpoint was incident CMM, defined as the occurrence of at least two cardiometabolic diseases. CMDs included T2D, stroke, and CHD, which were identified according to the International Classification of Diseases, 10th Revision (ICD-10). T2D was defined using code E11, stroke using codes I60-I64 and I69, and CHD using codes I20-I25 [24]. Incident CMDs were identified using the UK Biobank “First occurrence”, which integrates information from self-reports, primary care, hospital admissions, and death registrations. The onset date of CMM was defined as the first recorded date on which a second CMD was documented during follow-up. Follow-up time was accrued from baseline until incident CMM, death, or the end of follow-up (August 31, 2025), whichever occurred earliest.

Assessment of IR-related indices

Baseline blood samples were used to measure biochemical markers, including glucose, triglycerides, and HDL-C. Drawing on previously published and validated methodologies [18, 24, 36], nine IR-related indices were subsequently calculated using established formulas:

1. TyG index = ln [TG (mg/dL) × glucose (mg/dL) / 2];

2. TyG-BMI = ;

3. TyG-WC = ;

4. TyG-WHtR = ;

5. TyG-ABSI = ;

6. TyG-WWI = ;

7. TyG-BRI = [];

8. TyG-VAI in male = TyG × WC (cm) / (39.68 + (1.88 × BMI [kg/m²])) × (TG [mmol/L]/1.03) × (1.31/HDL-C [mmol/L]) or TyG-VAI in female = TyG × WC (cm) / (36.58 + (1.89 × BMI [kg/m²])) × (TG [mmol/L]/0.81) × (1.52/HDL-C [mmol/L]);

9. TG/HDL-C ratio = TG (mg/dL) / HDL-C (mg/dL).

Assessment of biomarkers

Blood biomarker assays in the UK Biobank were performed under rigorous quality-control procedures [37]. Based on prior evidence, we selected candidate mediating biomarkers that may be involved in potential pathways linking IR-related indices to CMM [24, 31, 38–40]. These biomarkers were grouped into inflammatory, hepatic, and renal domains. The inflammatory panel included CRP, WBC count, neutrophil count, monocyte count, lymphocyte count, and platelet count; liver-related markers included ALT, ALP, AST, GGT, total bilirubin, total protein, and albumin; renal markers included cystatin C, creatinine, urate, and urea. Further information is presented in Table S1.

Covariates

Covariates were selected based on prior literature and clinical relevance [24, 31, 41] and included the following: (1) sociodemographic characteristics, including age, sex, ethnicity, educational levels, and Townsend deprivation index; (2) lifestyle behaviors, including smoking status, frequency of alcohol intake, and physical activity [42]; (3) medical history, including self-reported hypertension and cancer; (4) medication use, including lipid-lowering, and antihypertensive; and (5) family history, defined as a reported history of CMD among first-degree relatives. The covariates used in the analysis are described in Table S2, and missing data for covariates are detailed in Table S3.

Statistical analysis

Baseline characteristics were described according to CMM status. Continuous variables are presented as mean ± SD, whereas categorical variables are shown as counts and percentages.

We used Cox proportional hazards regression to estimate the associations between each IR-related index and incident CMM in participants with MASLD. Effect estimates were expressed as hazard ratios (HRs) and 95% confidence intervals (CIs) per standard deviation (SD) increment and across quartiles, with the lowest quartile serving as the reference. The proportional hazards assumption was checked with Schoenfeld residuals and showed no violation. A directed acyclic graph (DAG) was constructed to clarify the assumed relationships among IR-related indices, baseline covariates, candidate biomarker pathways, and incident CMM (Figure S1). Three adjusted models were constructed: Model 1 adjusted for sociodemographic characteristics; Model 2 further adjusted for lifestyle behaviors covariates, including smoking status, drinking frequency, and physical activity; and Model 3 additionally adjusted for medical history, medication use and family history of CMD. Candidate biomarkers evaluated in the mediation analysis were not included as covariates in these primary exposure–outcome models. Kaplan–Meier curves were generated to depict cumulative hazard across quartiles of each index, and group differences were assessed with the log-rank test. Furthermore, potential nonlinear dose–response relationships were evaluated using restricted cubic splines with four knots placed at the 5th, 35th, 65th, and 95th percentiles under Model 3 [43]. Discriminative performance was assessed using receiver operating characteristic (ROC) analysis, with the area under the curve (AUC) used to summarize overall classification ability. To account for censored observations and varying follow-up times in this prospective cohort, time-dependent ROC analyses were performed at 3, 5, and 10 years, and time-dependent AUCs were used to summarize discrimination at each prediction horizon. Conventional ROC curves and DeLong tests were also presented as supplementary comparisons among indices. To further examine whether IR-related indices provided incremental risk information beyond conventional risk factors, a conventional risk-factor model with the same covariates as Model 3 but without IR-related indices was compared with extended models additionally incorporating each index. Incremental model performance was evaluated using the C-index, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) [44, 45]. These analyses were used to assess changes in discrimination and reclassification after adding each IR-related index, rather than to evaluate the standalone clinical predictive utility of individual indices.

A multi-state model was further applied to examine the associations of nine IR-related indices with CMM progression [21]. This approach extends conventional Cox regression by allowing the estimation of risk across multiple stages of disease evolution simultaneously. Consistent with prior studies [41], a series of transition states were defined to characterize disease progression (Fig. 2): (A) from baseline (free of CMD) to incident T2D; (B) from baseline to stroke; (C) from baseline to CHD; (D) from T2D to CMM; (E) from stroke to CMM; and (F) from CHD to CMM. This framework enabled the evaluation of the role of IR-related indices in both the initial onset of CMD and their subsequent progression to multimorbidity. For participants who entered more than one disease stage on the same date, the event date of the preceding stage was assigned as 0.5 days earlier than that of the subsequent stage, consistent with previous reports [24].

Fig. 2.

Longitudinal transition patterns of cardiometabolic multimorbidity. Abbreviations: CMM, cardiometabolic multimorbidity; CHD, coronary heart disease; T2D, type 2 diabetes; MASLD, metabolic dysfunction-associated steatotic liver disease

Exploratory mediation analyses were conducted to investigate potential biomarkers mediating the associations between IR-related indices and incident CMM. All exploratory mediation analyses were adjusted for the same predefined baseline covariates as Model 3. In accordance with established approaches, a stepwise strategy was adopted to identify and quantify mediation effects. Firstly, two models were applied to preliminary screening the potential mediators [46]: (1) linear regression models were used to evaluate the relationships of IR-related indices and candidate biomarkers; (2) Cox proportional hazards models were fitted to assess the relationships between these biomarkers and the risk of CMM. Biomarkers that demonstrated significant associations in both models and consistent directions of effect were considered potential mediators. Secondly, exploratory mediation analyses were performed for each potential biomarker individually using the CMAverse package in R [47]. Finally, biomarkers demonstrating significant mediation were then jointly included in a multiple-mediator framework to estimate their combined mediation effect. The proportion mediated and its 95% CI were estimated using 1000 bootstrap resamples.

We conducted several sensitivity analyses. First, we excluded participants who developed CMM within the first five years of follow-up to minimize the potential influence of early subclinical disease. Second, we applied multiple imputation to address missing covariate data. Third, we fitted cause-specific Cox regression models to account for the competing risk of death before the occurrence of CMM. Fourth, to address the potential influence of excluding participants with baseline diabetes, we conducted an additional sensitivity analysis among MASLD participants with baseline diabetes but without baseline CHD or stroke. In this analysis, baseline diabetes was considered an initial single cardiometabolic disease state, and incident CMM was defined as the subsequent occurrence of CHD or stroke during follow-up. Potential effect modification by age (< 60 vs. ≥ 60 years), sex (female vs. male), and baseline lipid-lowering medication use (no vs. yes) was examined in stratified analyses. Interaction analyses were conducted by adding product terms between each IR-related index and the corresponding stratification variable to the fully adjusted model.

All tests were two-sided, and statistical significance was defined as P < 0.05. All analyses were performed using R software (version 4.5.2).

Results

Baseline characteristics

The analysis included 109,604 participants (mean age, 56.76 years; SD, 7.87), among whom 40.40% were female. Compared with those who did not develop CMM, participants who developed CMM were older, more likely to be male, less educated, and more socioeconomically deprived (all P < 0.001). They also had less favorable lifestyle profiles, including a higher prevalence of inadequate physical activity, smoking, and less frequent alcohol consumption (all P < 0.001). Moreover, they had a greater burden of comorbidities, higher medication use, and a higher proportion with a family history of CMD (all P < 0.001). All IR-related indices, including TyG, TyG-BMI, TyG-WC, TyG-WHtR, TyG-BRI, TyG-ABSI, TyG-VAI, TyG-WWI, and the TG/HDL-C ratio, were significantly higher in participants who subsequently developed CMM, indicating a more adverse metabolic profile at baseline (Table 1). Baseline characteristics by quartiles of the TyG are shown in Table S4.

Table 1.

Baseline characteristics of study population

Characteristics	Total population	CMM occurrence		P value
		No	Yes
n	109,604	104,660	4944
Age, year	56.76 ± 7.87	56.58 ± 7.88	60.57 ± 6.73	< 0.001
Sex				< 0.001
Female	44,281 (40.40)	42,584 (40.69)	1697 (34.32)
Male	65,323 (59.60)	62,076 (59.31)	3247 (65.68)
Ethnicity				< 0.001
Whites	103,847 (94.75)	99,272 (94.85)	4575 (92.54)
Non-Whites	5,757 (5.25)	5,388 (5.15)	369 (7.46)
Educational levels				< 0.001
University or college	31,943 (29.14)	30,917 (29.54)	1026 (20.75)
Others	77,661 (70.86)	73,743 (70.46)	3918 (79.25)
Townsend deprivation index	− 1.25 ± 3.10	− 1.28 ± 3.09	− 0.66 ± 3.35	< 0.001
Physical activity				< 0.001
Inadequate	35,046 (31.98)	33,289 (31.81)	1757 (35.54)
Adequate	74,558 (68.02)	71,371 (68.19)	3187 (64.46)
Smoking status				< 0.001
Never smoking	58,753 (53.60)	56,661 (54.14)	2092 (42.31)
Ever smoking	40,230 (36.70)	38,063 (36.37)	2167 (43.83)
Current smoking	10,621 (9.69)	9936 (9.49)	685 (13.86)
Frequency of drinking				< 0.001
Daily or almost daily	13,345 (12.18)	12,858 (12.29)	487 (9.85)
3–4 times/week	23,947 (21.85)	23,099 (22.07)	848 (17.15)
1–2 times/week	32,849 (29.97)	31,474 (30.07)	1375 (27.81)
1–3 times/month	14,834 (13.53)	14,153 (13.52)	681 (13.77)
Special occasions only	15,113 (13.79)	14,256 (13.62)	857 (17.33)
Never	9,516 (8.68)	8,820 (8.43)	696 (14.08)
Self-reported history of hypertension	38,572 (35.19)	35,763 (34.17)	2809 (56.82)	< 0.001
Self-reported history of cancer	8,121 (7.41)	7677 (7.34)	444 (8.98)	< 0.001
Use of antihypertensive medication	28,206 (25.73)	25,838 (24.69)	2368 (47.90)	< 0.001
Use of lipid-lowering medication	19,921 (18.18)	17,968 (17.17)	1953 (39.50)	< 0.001
Family history of cardiometabolic disease	86,600 (79.01)	82,499 (78.83)	4101 (82.95)	< 0.001
TyG	9.08 ± 0.49	9.07 ± 0.49	9.25 ± 0.58	< 0.001
TyG-BMI	285.09 ± 40.71	284.35 ± 40.17	300.72 ± 48.08	< 0.001
TyG-WC	924.03 ± 98.03	921.45 ± 96.18	978.64 ± 118.79	< 0.001
TyG-WHtR	5.44 ± 0.58	5.42 ± 0.57	5.77 ± 0.70	< 0.001
TyG-BRI	49.64 ± 12.76	49.33 ± 12.53	56.22 ± 15.50	< 0.001
TyG-ABSI	0.71 ± 0.06	0.71 ± 0.06	0.74 ± 0.07	< 0.001
TyG-VAI	29.97 ± 20.68	29.75 ± 20.42	34.63 ± 25.04	< 0.001
TyG-WWI	97.15 ± 7.75	96.95 ± 7.64	101.45 ± 8.74	< 0.001
TG/HDL-C ratio	4.78 ± 3.07	4.75 ± 3.04	5.42 ± 3.56	< 0.001

Data are presented as mean ± standard deviation for continuous variables, and number (%) for categorical variables

ABSI, a body shape index; BMI, body mass index; BRI, body roundness index; CMM, cardiometabolic multimorbidity; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio; TyG, triglyceride-glucose index; VAI, visceral adiposity index; WC, waist circumference; WHtR, waist-to-height ratio; WWI, weight-adjusted waist index

Associations of nine IR-related indices with CMM among participants with MASLD

Over 15.9 years of median follow-up (interquartile range [IQR]: 14.0–16.9 years), 4,944 participants developed CMM. All IR-related indices showed positive associations with incident CMM in Model 3 (Table 2). Each SD increment in IR-related indices was linked to a significantly higher risk of CMM, with HRs (95% CIs) ranging from 1.15 (1.12–1.18) for TG/HDL-C to 1.37 (1.34–1.41) for TyG-WHtR in the fully adjusted model. Consistent findings were observed in quartile analyses. Compared to participants in Q1, participants in the highest Q4 had markedly elevated risks of CMM across all indices. The strongest associations were observed for TyG-WHtR (HR: 2.70, 95% CI 2.45–2.97), TyG-BMI (HR: 2.36, 95% CI 2.16–2.58), TyG-BRI (HR: 2.36, 95% CI 2.15–2.60), and TyG-WC (HR: 2.33, 95% CI 2.12–2.56), while relatively weaker but still significant associations were observed for TyG-VAI and TG/HDL-C (Table 2). Kaplan–Meier curves showed a stepwise increase in cumulative hazard across quartiles (all log-rank P < 0.001; Fig. 3), and spline analyses supported nonlinear associations between these indices and CMM risk (all P for overall and nonlinearity < 0.001) (Fig. 4).

Table 2.

Associations between insulin resistance-related indices and risk of cardiometabolic multimorbidity in participants with MASLD

	No. of cases/ total population	Incidence rate	HR (95% CI)
			Model 1	Model 2	Model 3
TyG
Per 1-SD increase			1.40 (1.37–1.44)	1.38 (1.35–1.42)	1.25 (1.22–1.28)
Quartile 1	931/27,401	2.17	Ref	Ref	Ref
Quartile 2	999/27,401	2.33	1.04 (0.95–1.14)	1.03 (0.94–1.13)	1.03 (0.94–1.13)
Quartile 3	1167/27,401	2.72	1.23 (1.12–1.34)	1.20 (1.10–1.31)	1.18 (1.08–1.28)
Quartile 4	1847/27,401	4.37	1.98 (1.83–2.15)	1.91 (1.77–2.07)	1.62 (1.49–1.76)
P-trend			< 0.001	< 0.001	< 0.001
TyG-BMI
Per 1-SD increase			1.53 (1.49–1.56)	1.52 (1.48–1.55)	1.32 (1.29–1.35)
Quartile 1	771/27,401	1.79	Ref	Ref	Ref
Quartile 2	1002/27,401	2.33	1.41 (1.29–1.55)	1.42 (1.29–1.56)	1.29 (1.17–1.41)
Quartile 3	1244/27,401	2.91	1.95 (1.78–2.13)	1.95 (1.78–2.14)	1.63 (1.49–1.79)
Quartile 4	1927/27,401	4.57	3.43 (3.15–3.74)	3.39 (3.10–3.69)	2.36 (2.16–2.58)
P-trend			< 0.001	< 0.001	< 0.001
TyG-WC
Per 1-SD increase			1.59 (1.55–1.63)	1.57 (1.54–1.61)	1.35 (1.32–1.38)
Quartile 1	637/27,401	1.46	Ref	Ref	Ref
Quartile 2	915/27,401	2.12	1.35 (1.22–1.50)	1.34 (1.21–1.48)	1.26 (1.14–1.40)
Quartile 3	1174/27,401	2.75	1.71 (1.55–1.89)	1.68 (1.52–1.85)	1.48 (1.34–1.63)
Quartile 4	2218/27,401	5.34	3.36 (3.07–3.68)	3.25 (2.96–3.55)	2.33 (2.12–2.56)
P-trend			< 0.001	< 0.001	< 0.001
TyG-WHtR
Per 1-SD increase			1.62 (1.58–1.65)	1.60 (1.56–1.64)	1.37 (1.34–1.41)
Quartile 1	562/27,401	1.29	Ref	Ref	Ref
Quartile 2	908/27,401	2.11	1.49 (1.34–1.66)	1.47 (1.33–1.64)	1.40 (1.26–1.55)
Quartile 3	1190/27,401	2.79	1.94 (1.75–2.14)	1.90 (1.71–2.10)	1.66 (1.50–1.83)
Quartile 4	2284/27,401	5.49	3.92 (3.57–4.30)	3.78 (3.44–4.15)	2.70 (2.45–2.97)
P-trend			< 0.001	< 0.001	< 0.001
TyG-BRI
Per 1-SD increase			1.50 (1.47–1.53)	1.49 (1.46–1.52)	1.31 (1.28–1.34)
Quartile 1	613/27,401	1.41	Ref	Ref	Ref
Quartile 2	935/27,401	2.17	1.40 (1.26–1.55)	1.39 (1.26–1.54)	1.30 (1.18–1.44)
Quartile 3	1238/27,401	2.90	1.84 (1.67–2.03)	1.81 (1.64–2.00)	1.54 (1.40–1.70)
Quartile 4	2158/27,401	5.18	3.42 (3.12–3.75)	3.33 (3.03–3.65)	2.36 (2.15–2.60)
P-trend			< 0.001	< 0.001	< 0.001
TyG-ABSI
Per 1-SD increase			1.45 (1.40–1.49)	1.41 (1.37–1.46)	1.27 (1.23–1.31)
Quartile 1	717/27,401	1.65	Ref	Ref	Ref
Quartile 2	1043/27,401	2.42	1.33 (1.21–1.47)	1.32 (1.19–1.45)	1.29 (1.17–1.42)
Quartile 3	1181/27,401	2.77	1.43 (1.30–1.58)	1.40 (1.27–1.54)	1.33 (1.21–1.46)
Quartile 4	2003/27,401	4.80	2.32 (2.12–2.54)	2.20 (2.01–2.41)	1.82 (1.66–2.00)
P-trend			< 0.001	< 0.001	< 0.001
TyG-VAI
Per 1-SD increase			1.23 (1.20–1.25)	1.21 (1.18–1.23)	1.16 (1.13–1.18)
Quartile 1	991/27,401	2.32	Ref	Ref	Ref
Quartile 2	1036/27,401	2.41	1.07 (0.98–1.17)	1.05 (0.96–1.15)	1.02 (0.93–1.11)
Quartile 3	1285/27,401	3.00	1.36 (1.25–1.47)	1.31 (1.21–1.43)	1.24 (1.14–1.34)
Quartile 4	1632/27,401	3.84	1.81 (1.67–1.96)	1.70 (1.57–1.84)	1.54 (1.42–1.67)
P-trend			< 0.001	< 0.001	< 0.001
TyG-WWI
Per 1-SD increase			1.55 (1.51–1.59)	1.52 (1.48–1.56)	1.33 (1.30–1.37)
Quartile 1	617/27,401	1.41	Ref	Ref	Ref
Quartile 2	915/27,401	2.12	1.30 (1.17–1.44)	1.28 (1.16–1.42)	1.24 (1.12–1.37)
Quartile 3	1154/27,401	2.70	1.52 (1.38–1.68)	1.49 (1.35–1.64)	1.39 (1.26–1.54)
Quartile 4	2258/27,401	5.44	2.87 (2.62–3.14)	2.73 (2.49–2.99)	2.14 (1.95–2.34)
P-trend			< 0.001	< 0.001	< 0.001
TG/HDL-C ratio
Per 1-SD increase			1.21 (1.18–1.24)	1.19 (1.16–1.22)	1.15 (1.12–1.18)
Quartile 1	968/27,401	2.27	Ref	Ref	Ref
Quartile 2	1106/27,401	2.58	1.13 (1.04–1.23)	1.11 (1.02–1.21)	1.07 (0.98–1.17)
Quartile 3	1251/27,401	2.92	1.30 (1.19–1.41)	1.26 (1.15–1.37)	1.19 (1.09–1.30)
Quartile 4	1619/27,401	3.80	1.75 (1.61–1.90)	1.65 (1.52–1.79)	1.52 (1.40–1.65)
P-trend			< 0.001	< 0.001	< 0.001

Incidence rates are presented per 1,000 person-years. Hazard ratios (HRs) and 95% confidence intervals (CIs) are reported per one standard deviation increment in each index. Model 1 was adjusted for age, sex, ethnicity, educational levels, and Townsend deprivation index. Model 2 was further adjusted for frequency of drinking, smoking status, and physical activity. Model 3 was additionally adjusted for self-reported history of hypertension, history of cancer, use of lipid-lowering medication, use of antihypertensive medication, and family history of cardiometabolic disease

ABSI, a body shape index; BMI, body mass index; BRI, body roundness index; CI, confidence interval; CMM, cardiometabolic multimorbidity; HR, hazard ratio; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio; TyG, triglyceride-glucose index; VAI, visceral adiposity index; WC, waist circumference; WHtR, waist-to-height ratio; WWI, weight-adjusted waist index

Fig. 3.

Kaplan–Meier curves for incident cardiometabolic multimorbidity according to quartiles of insulin resistance-related indices in participants with MASLD. Abbreviations: TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BRI, body roundness index; ABSI, a body shape index; VAI, visceral adiposity index; WWI, weight-adjusted waist index; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio

Fig. 4.

Dose–response associations of insulin resistance-related indices with incident cardiometabolic multimorbidity in participants with MASLD. Models were adjusted for age, sex, ethnicity, educational levels, Townsend deprivation index, frequency of drinking, smoking status, physical activity, self-reported history of hypertension, history of cancer, use of lipid-lowering medication, use of antihypertensive medication, and family history of cardiometabolic disease. Abbreviations: TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BRI, body roundness index; ABSI, a body shape index; VAI, visceral adiposity index; WWI, weight-adjusted waist index; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio

Associations of IR-related indices with stage-specific progression of CMM among participants with MASLD

Multi-state models were used to evaluate transitions from a CMD-free state at baseline to T2D, stroke, and CHD, and subsequently to CMM (Table 3). Specifically, the associations were strongest for the transition from a CMD-free state to T2D, with HRs per SD increment ranging from 1.22 (95% CI 1.20–1.23) for the TG/HDL-C ratio to 1.81 (95% CI 1.79–1.84) for TyG-WHtR. All IR-related indices were likewise positively associated with incident CHD, with corresponding HRs ranging from 1.05 (95% CI 1.03–1.08) for TyG-BMI to 1.20 (95% CI 1.18–1.23) for TyG-ABSI per SD increment. In contrast, associations with incident stroke were generally weaker and were not statistically significant for several indices, including TyG, TyG-BMI, TyG-VAI, and the TG/HDL-C ratio. For progression from first CMD to CMM, positive associations were observed across all three downstream transitions. The strongest associations were observed for progression from CHD to CMM, with HRs ranging from 1.12 (95% CI 1.08–1.16) for the TG/HDL-C ratio to 1.47 (95% CI 1.42–1.53) for TyG-WHtR. The second strongest associations were observed for the transition from stroke to CMM, with corresponding HRs ranging from 1.07 (95% CI 1.00–1.15) for the TG/HDL-C ratio to 1.31 (95% CI 1.23–1.39) for TyG-WHtR. Overall, indices combining adiposity measures, particularly TyG-WHtR, TyG-WC, and TyG-BMI showed the strongest associations across these progression pathways.

Table 3.

Roles of insulin resistance-related indices in the trajectory of cardiometabolic multimorbidity progression among participants with MASLD

Exposures	HRs (95% CIs)
	Transition A (Healthy—> T2D)	Transition B (Healthy—> Stroke)	Transition C (Healthy—> CHD)	Transition D (T2D—> CMM)	Transition E (Stroke—> CMM)	Transition F (CHD—> CMM)
No. of cases/ total population	12,371/109,604	4,083/109,604	10,988/109,604	1,988/12,371	896/4,083	2,060/10,988
Proportions	11.3%	3.7%	10.0%	16.1%	21.9%	18.7%
TyG	1.60 (1.57–1.62)*	0.97 (0.94–1.00)	1.10 (1.08–1.12)*	1.04 (1.00–1.08)*	1.20 (1.12–1.28)*	1.33 (1.27–1.39)*
TyG-BMI	1.67 (1.65–1.70)*	1.01 (0.97–1.04)	1.05 (1.03–1.08)*	1.10 (1.06–1.14)*	1.27 (1.19–1.35)*	1.42 (1.37–1.47)*
TyG-WC	1.74 (1.72–1.77)*	1.05 (1.01–1.08)*	1.16 (1.14–1.18)*	1.11 (1.08–1.16)*	1.31 (1.23–1.39)*	1.41 (1.36–1.47)*
TyG-WHtR	1.81 (1.79–1.84)*	1.08 (1.04–1.11)*	1.10 (1.08–1.12)*	1.11 (1.07–1.15)*	1.31 (1.23–1.39)*	1.47 (1.42–1.53)*
TyG-BRI	1.62 (1.60–1.64)*	1.09 (1.06–1.13)*	1.07 (1.05–1.09)*	1.10 (1.06–1.14)*	1.26 (1.18–1.33)*	1.37 (1.32–1.42)*
TyG-ABSI	1.48 (1.45–1.51)*	1.11 (1.08–1.15)*	1.20 (1.18–1.23)*	1.09 (1.04–1.13)*	1.21 (1.13–1.30)*	1.26 (1.20–1.32)*
TyG-VAI	1.27 (1.25–1.28)*	0.98 (0.95–1.01)	1.09 (1.07–1.11)*	1.04 (1.00–1.07)*	1.09 (1.02–1.16)*	1.16 (1.12–1.20)*
TyG-WWI	1.71 (1.68–1.74)*	1.10 (1.07–1.14)*	1.13 (1.11–1.16)*	1.08 (1.04–1.12)*	1.25 (1.16–1.33)*	1.39 (1.33–1.45)*
TG/HDL-C ratio	1.22 (1.20–1.23)*	0.97 (0.94–1.00)	1.13 (1.11–1.15)*	1.05 (1.02–1.09)*	1.07 (1.00–1.15)*	1.12 (1.08–1.16)*

*P < 0.05. HRs (95% CIs) are reported per one standard deviation increment in each index. Models were adjusted for age, sex, ethnicity, educational levels, Townsend deprivation index, frequency of drinking, smoking status, physical activity, self-reported history of hypertension, self-reported history of cancer, use of lipid-lowering medication, use of antihypertensive medication, and family history of cardiometabolic disease. Baseline MASLD indicates participants with MASLD who were free of type 2 diabetes, stroke, and coronary heart disease at study entry. All transitions are illustrated in Fig. 2

HR, hazard ratio; CI, confidence interval; CMM, cardiometabolic multimorbidity; CHD, coronary heart disease; T2D, type 2 diabetes; MASLD, metabolic dysfunction-associated steatotic liver disease; TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BRI, body roundness index; ABSI, a body shape index; VAI, visceral adiposity index; WWI, weight-adjusted waist index; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio

Incremental predictive performance of IR-related indices

We then assessed whether adding each IR-related index to the conventional risk model improved prediction of CMM (Table 4). The conventional model showed reasonable discrimination, with a C-index of 0.755 (95% CI 0.749–0.762). The addition of each index modestly improved predictive performance (all P < 0.001). Among all indices, TyG-WHtR provided the relatively greater incremental risk information, yielding the highest C-index of 0.773 (95% CI 0.767–0.779), followed by TyG-WC (C-index: 0.771, 95% CI 0.765–0.778) and TyG-BMI (C-index: 0.769, 95% CI 0.762–0.775). Similar patterns were observed for reclassification and discrimination metrics. TyG-WHtR showed the largest NRI (0.169, 95% CI 0.115–0.181) and IDI (0.022, 95% CI 0.013–0.027), indicating the greatest added value beyond conventional risk factors. These results were consistent with the ROC curves and pairwise AUC comparisons shown in Figures S2-S4.

Table 4.

Incremental predictive performance of insulin resistance-related indices for cardiometabolic multimorbidity in participants with MASLD

Model	C-index			NRI			IDI
	Estimate	95% CI	P value	Estimate	95% CI	P value	Estimate	95% CI	P value
Conventional model (Reference)	0.755	0.749–0.762	–	Ref	–	–	Ref	–	–
Conventional model + TyG	0.764	0.757–0.770	< 0.001	0.098	0.066–0.128	< 0.001	0.009	0.005–0.013	< 0.001
Conventional model + TyG-BMI	0.769	0.762–0.775	< 0.001	0.151	0.124–0.174	< 0.001	0.017	0.011–0.022	< 0.001
Conventional model + TyG-WC	0.771	0.765–0.778	< 0.001	0.163	0.140–0.183	< 0.001	0.021	0.016–0.027	< 0.001
Conventional model + TyG-WHtR	0.773	0.767–0.779	< 0.001	0.169	0.115–0.181	< 0.001	0.022	0.013–0.027	< 0.001
Conventional model + TyG-BRI	0.769	0.762–0.775	< 0.001	0.143	0.114–0.179	< 0.001	0.016	0.011–0.022	< 0.001
Conventional model + TyG-ABSI	0.763	0.756–0.769	< 0.001	0.099	0.069–0.119	< 0.001	0.009	0.006–0.012	< 0.001
Conventional model + TyG-VAI	0.760	0.754–0.767	< 0.001	0.055	0.034–0.085	< 0.001	0.004	0.001–0.006	< 0.001
Conventional model + TyG-WWI	0.768	0.762–0.775	< 0.001	0.131	0.077–0.150	< 0.001	0.015	0.007–0.018	< 0.001
Conventional model + TG/HDL-C ratio	0.760	0.753–0.766	< 0.001	0.055	0.012–0.079	< 0.001	0.003	0.001–0.005	< 0.001

Conventional models were adjusted for age, sex, ethnicity, educational levels, Townsend deprivation index, frequency of drinking, smoking status, physical activity, self-reported history of hypertension, self-reported history of cancer, use of lipid-lowering medication, use of antihypertensive medication, and family history of cardiometabolic disease

CMM, cardiometabolic multimorbidity; NRI, net reclassification improvement; IDI, integrated discrimination improvement; CI, confidence interval; TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BRI, body roundness index; ABSI, a body shape index; VAI, visceral adiposity index; WWI, weight-adjusted waist index; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio

Exploratory mediation analyses

As presented in Table S5, higher IR-related indices were broadly linked to unfavorable profiles of inflammatory, hepatic, and renal biomarkers, although the magnitude and direction of these associations varied across indices. Overall, stronger and more consistent exposure–mediator associations were observed for biomarkers related to systemic inflammation, including CRP, WBC, and neutrophil count, as well as for markers of liver and renal function such as alanine aminotransferase, cystatin C, urate, and urea. As shown in Table S6, several candidate biomarkers were positively associated with incident CMM, including CRP, WBC count, neutrophil count, monocyte count, ALT, ALP, AST, GGT, cystatin C, urate, and urea, whereas albumin showed an inverse association. In contrast, total bilirubin, total protein, and creatinine were not significantly associated with incident CMM. Exploratory mediation analyses identified multiple significant mediators in the relationships between IR-related indices and CMM (Fig. 5 and Table S7). At the individual biomarker level, neutrophil count, CRP, ALT, urate, and cystatin C consistently showed relatively large mediation proportions across different indices. The multiple-mediator models revealed that the evaluated pathways partially explained the association, with the proportion mediated ranging from 11.0 to 23.1%. The largest estimated overall mediation proportions were observed for TyG-BRI (23.1%, 95% CI 20.5–28.0%) and TyG-BMI (22.0%, 95% CI 18.4–25.5%).

Fig. 5.

Mediated proportions of selected biomarkers in the associations between insulin resistance-related indices and incident cardiometabolic multimorbidity in participants with MASLD. A. the mediation proportions of individual biomarkers for each insulin resistance-related index, and B. the corresponding overall mediation proportions with 95% confidence intervals. Models were adjusted for age, sex, ethnicity, educational levels, Townsend deprivation index, frequency of drinking, smoking status, physical activity, self-reported history of hypertension, history of cancer, use of lipid-lowering medication, use of antihypertensive medication, and family history of cardiometabolic disease. Abbreviations: TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BRI, body roundness index; ABSI, a body shape index; VAI, visceral adiposity index; WWI, weight-adjusted waist index; TG/HDL-C, triglyceride-to-high-density lipoprotein cholesterol ratio

Subgroup and sensitivity analyses

The main results were materially unchanged in a series of sensitivity analyses, including analyses excluding CMM events that occurred within 5 years, analyses based on multiple imputation for missing covariates, analyses accounting for the competing risk of death, and analyses among MASLD participants with baseline diabetes but without baseline CHD or stroke (Tables S8–S11). Across these analyses, the direction and magnitude of associations were largely unchanged. In stratified analyses, associations were generally stronger among younger participants for several indices (Table S12). Additional heterogeneity by sex was also observed for some indices, including TyG, TyG-BMI, TyG-BRI, and the TG/HDL-C ratio (Table S13). When stratified by baseline lipid-lowering medication use, positive associations persisted in both users and non-users, with broadly comparable effect estimates; statistically significant interactions were observed for most indices, except for the TG/HDL-C ratio (Table S14). Overall, these results support the robustness of the associations between IR-related indices and incident CMM.

Discussion

Principal findings

In this large prospective cohort of individuals with MASLD, higher levels of all nine IR-related indices were independently associated with an increased risk of incident CMM. Among these indices, TyG-WHtR, TyG-WC, and TyG-BMI showed the strongest associations and the best overall predictive performance. Multi-state modelling further suggested that these indices were related not only to disease onset but also to subsequent progression across different stages of CMM. Specifically, the associations were strongest for the transition from a CMD-free state to T2D, with HRs ranging from 1.22 to 1.81 per SD increment. For progression from first CMD to CMM, TyG-WHtR, TyG-WC, and TyG-BMI showed the strongest associations. In addition, exploratory mediation analyses suggested that biomarkers reflecting systemic inflammation, liver dysfunction, and renal impairment jointly explained 11.0% to 23.1% of these associations. Collectively, these findings suggest that IR-related indices, particularly those incorporating central adiposity, may help identify individuals with MASLD who are at elevated risk of developing CMM.

Comparison with previous studies on IR-related indices and CMM incidence

Our results are generally in line with previous studies showing that IR-related indices are linked to adverse cardiometabolic outcomes [41, 48, 49]. Recent evidence has further suggested that routinely available metabolic indices may be useful for CMM risk stratification. For example, a national prospective cohort study by Fan et al. showed that a combined cholesterol, high-density lipoprotein, glucose, and frailty index was associated with incident heart disease, stroke, diabetes, and CMM, and provided incremental predictive value for CMM risk assessment [22]. This study is particularly relevant because it highlights that CMM risk may reflect not only metabolic dysregulation but also the accumulation of broader physiological vulnerability. Similar evidence has also been observed in related multimorbidity frameworks. Another prospective cohort study reported that the TyG index and its related markers were significantly linked to CMM development and trajectory [24]. One UK Biobank analysis showed that several IR-related indices were associated with the risk and progression of cardio-renal-metabolic multimorbidity [18], and another prospective study observed positive associations between multiple IR-related indices and incident CMM among participants with hypertension [13]. In addition, a recent UK Biobank study showed that IR-related indices were associated with the occurrence of CVD and cardiovascular mortality in individuals with MASLD [31]. However, whether these findings can be directly generalized to individuals with MASLD remains uncertain. MASLD is characterized by hepatic steatosis, exacerbated IR, dyslipidemia, and chronic low-grade inflammation, which may amplify the cardiometabolic consequences of IR. Building on this evidence, our study extends the focus from single cardiovascular outcomes to CMM, providing a more integrated assessment of disease burden in individuals with MASLD.

Notably, our findings suggest that IR-related indices incorporating adiposity-related parameters exert a more pronounced influence on CMM risk. This observation indicates that the development of CMM in MASLD is closely linked not only to abnormalities in glucose and lipid metabolism but also to excessive fat accumulation, particularly central adiposity. From a biological perspective, visceral adipose tissue promotes the flux of free fatty acids to the liver via the portal circulation, thereby exacerbating hepatic steatosis and IR [26]. It also contributes to chronic inflammation, adipokine imbalance, and endothelial dysfunction. The nonlinear patterns observed in the RCS analyses indicate that the impact of IR on cardiometabolic risk may involve threshold-dependent biological processes. Rather than increasing uniformly, risk appears to accelerate once metabolic dysfunction exceeds a certain level, consistent with the concept of a metabolic tipping point. This transition may reflect the breakdown of compensatory mechanisms and the activation of pathological pathways such as oxidative stress, mitochondrial dysfunction, and systemic inflammation [50]. The temporal separation in cumulative incidence across exposure levels further supports the view that these indices may capture sustained metabolic burden over time and may therefore be useful for identifying individuals at long-term risk of CMM. We also observed significant interactions by baseline lipid-lowering medication use for most IR-related indices. The associations between IR-related indices and incident CMM persisted among both users and non-users of lipid-lowering medication, with broadly comparable effect estimates and slightly stronger associations among non-users for several adiposity-integrated TyG indices. Since TyG and related indices incorporate triglycerides and, in some indices, anthropometric parameters, lipid-lowering therapy may partially modify the lipid-related metabolic information captured by these measures. In contrast, among participants not using lipid-lowering medication, elevated IR-related indices may more directly reflect untreated atherogenic dyslipidemia, insulin resistance, and central adiposity-related metabolic disturbance.

Comparison with previous studies on IR-related indices and CMM progression

Multistate modelling, which conceptualizes CMM as a dynamic and progressive continuum rather than a single static endpoint, indicated that the associations of IR-related indices varied across different transition stages of CMM. Specifically, these indices were most strongly associated with the transition from a CMD-free state to incident T2D, whereas associations with incident stroke were weaker and were not statistically significant for several indices. For subsequent progression from first CMD to CMM, the strongest associations were observed for the transition from CHD to CMM, followed by the transition from stroke to CMM, while associations for the transition from T2D to CMM were comparatively weaker. Taken together, these stage-specific results indicate that the contribution of IR may differ across the cardiometabolic disease continuum. IR may exert a stronger influence during earlier stages of disease initiation, particularly for pathways leading to T2D, where impaired glucose metabolism is central. By contrast, stroke is a more heterogeneous outcome with multifactorial determinants beyond metabolic dysfunction. The weaker associations observed for the transition from T2D to CMM may indicate that, once diabetes is established, subsequent progression is influenced not only by baseline IR but also by factors such as disease duration, treatment, glycaemic control, and accumulated vascular damage. Overall, these findings support the view that cardiometabolic disease progression is dynamic, progressive, and stage specific.

Incremental predictive value for CMM risk stratification

Beyond etiological associations, IR-related indices provided modest incremental value for risk stratification beyond conventional clinical factors. This observation is consistent with previous studies showing that IR-related indices can improve prediction of cardiovascular outcomes [51, 52]. The present study extends these findings to CMM and further highlights the added value of indices incorporating central adiposity. From a clinical perspective, these findings are particularly relevant in MASLD populations, where identifying individuals at high risk of disease accumulation remains challenging [29]. Because IR-related indices are calculated from routinely collected clinical and anthropometric variables, they offer a practical, low-cost, and scalable approach for early risk identification. The enhanced performance of indices incorporating waist-related measures may reflect their ability to capture visceral adiposity, which is a key determinant of metabolic dysfunction in MASLD [53]. By integrating both metabolic and body composition information, these indices may better represent the underlying pathophysiological complexity than single-dimensional markers. These findings support their potential role as accessible markers for CMM risk assessment in MASLD populations.

Potential biological mechanisms

Our exploratory mediation analyses also offer preliminary evidence regarding the biological mechanisms potentially linking IR-related indices to CMM. In our study, higher levels of IR-related indices were associated with a broad range of biomarkers, particularly those reflecting systemic inflammation, liver function, and renal function, and many of these biomarkers were in turn associated with incident CMM. Among them, CRP, neutrophil count, ALT, cystatin C, urate, and urea showed relatively prominent mediation across several indices. Jointly, these biomarkers accounted for 11.0% to 23.1% of the observed associations, suggesting that inflammatory activation, hepatic dysfunction, and renal impairment may partly explain the link between IR and CMM. These findings are biologically plausible, as IR is closely connected with chronic low-grade inflammation, hepatic steatosis and dysfunction, and early renal impairment [54]. These exploratory findings suggest that biomarkers related to inflammation, liver dysfunction, and renal impairment may explain part of the observed associations, particularly for indices incorporating both metabolic and anthropometric components. This may be because these composite measures capture not only IR but also body fat distribution, particularly central adiposity. Given that visceral fat is metabolically active and can increase free fatty acid delivery to the liver while sustaining chronic inflammation, these mechanisms may partly explain why adiposity-integrated indices showed stronger overall associations with CMM.

Strengths and limitations

To the best of our knowledge, this is among the first studies to comprehensively evaluate a broad panel of IR-related indices in relation to incident CMM and stage-specific disease transitions in individuals with MASLD. However, several limitations should be acknowledged. First, although the UK Biobank offers a large and well-characterized cohort, most participants were of European ancestry, which may restrict the generalizability of our findings to other populations. Additional studies in more ethnically diverse populations are therefore warranted to confirm the applicability of these results across different demographic groups. Second, as an observational study, our analysis is inherently limited in its ability to establish causal relationships. Although we adjusted for multiple potential confounders, there may still be residual confounding from unmeasured or incompletely captured factors, such as genetic factors, dietary habits, or changes in medication use over time. Third, the mediation analyses should be interpreted as exploratory. IR-related indices and candidate mediators were measured at baseline; therefore, temporal precedence between exposure and mediator could not be fully established. Future longitudinal studies incorporating repeated biomarker assessments are needed to further clarify these potential pathways and better characterize the dynamic nature of these relationships. Fourth, although time-dependent ROC analyses and C-index, NRI, and IDI were used to evaluate predictive performance, external validation is needed before these indices can be applied for routine clinical risk stratification. Finally, because imaging-based liver fat assessments and histological data were not available for the full UK Biobank cohort, hepatic steatosis was defined using FLI ≥ 60. Therefore, some misclassification of hepatic steatosis is possible.

Conclusion

Utilizing a large prospective cohort, this study found that higher levels of IR-related indices were independently associated with increased risks of both incident and progressive CMM among individuals with MASLD. These associations were nonlinear and varied across disease stages, with indices incorporating central adiposity, particularly TyG-WHtR and TyG-WC, showing the strongest effect estimates and incremental predictive utility. These findings suggest that IR-related indices derived from routine clinical measures may aid early risk stratification and targeted prevention of CMM in this high-risk population. In addition, exploratory mediation analyses identified biomarkers reflecting chronic inflammation and hepatic and renal dysfunction as potential mediators, providing further insight into biological pathways that may help explain the associations between IR-related indices and CMM development and progression.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 ^{(1.1MB, docx)}

Abbreviations

ABSI

A body shape index

ALP

Alkaline phosphatase

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

AUC

Area under the curve

BMI

Body mass index

BRI

Body roundness index

C-index

Concordance index

CHD

Coronary heart disease

CI

Confidence interval

CMD

Cardiometabolic disease

CMM

Cardiometabolic multimorbidity

CRP

C-reactive protein

CVD

Cardiovascular disease

DAG

Directed acyclic graph

FDR

False discovery rate

FLI

Fatty liver index

GGT

Gamma-glutamyl transferase

HbA1c

Glycated hemoglobin

HDL-C

High-density lipoprotein cholesterol

HOMA-IR

Homeostatic model assessment of insulin resistance

HR

Hazard ratio

ICD-10

International classification of diseases, 10th Revision

IDI

Integrated discrimination improvement

IQR

Interquartile range

IR

Insulin resistance

MASLD

Metabolic dysfunction-associated steatotic liver disease

NRI

Net reclassification improvement

PA

Physical activity

PM

Proportion mediated

RCS

Restricted cubic spline

ROC

Receiver operating characteristic

SD

Standard deviation

SLD

Steatotic liver disease

T2D

Type 2 diabetes

TDI

Townsend deprivation index

TG

Triglyceride

TG/HDL-C

Triglyceride-to-high-density lipoprotein cholesterol ratio

TyG

Triglyceride-glucose index

VAI

Visceral adiposity index

WBC

White blood cell count

WC

Waist circumference

WHtR

Waist-to-height ratio

WWI

Weight-adjusted waist index

Author contributions

QHe: Conceptualization, Methodology, Writing—original draft. MS: Visualization, Investigation, Data curation, Writing—review & editing. JY: Visualization, Investigation, Supervision, Writing—review & editing. YW: Data curation. YS: Supervision, Funding acquisition, Writing—review & editing. ZL: Supervision, Funding acquisition, Writing—review & editing. QHa: Conceptualization, Supervision, Writing—review & editing. All authors read and approved the final manuscript.

Funding

This research received no external funding.

Data availability

Data supporting the findings of this study from the UK Biobank (http://www.ukbiobank.ac.uk/). Data is available with permission of the UK Biobank.

Declarations

Ethics approval and consent to participate

The UK Biobank study was approved by the North West Multi-centre Research Ethics Committee (MREC reference: 21/NW/0157), and all participants provided written informed consent before enrolment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.