Associations of six insulin resistance-related indices with the risk and progression of cardio-renal-metabolic multimorbidity: evidence from the UK biobank

Abstract

Background

Insulin resistance (IR)-related indices are increasingly recognized as important contributors to cardio-renal-metabolic (CRM) diseases. However, most prior studies have focused on isolated outcomes or cross-sectional disease status, lacking dynamic insight into disease progression. This study aimed to evaluate the predictive value of six IR-related indices for CRM onset and multistage progression using multistate modeling, and to explore potential biological mechanisms through mediation analysis.

Methods

We included 327,692 CRM-free participants from the UK Biobank in this prospective cohort study. Six IR-related indices including triglyceride-glucose (TyG) index, TyG-body mass index (TyG-BMI), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), triglyceride/high-density lipoprotein cholesterol (TG/HDL-C) ratio, and metabolic score for insulin resistance (METS-IR), were calculated using established formulas. Cox proportional hazards and multistate models were used to assess associations with CRM incidence and progression. Predictive performance was evaluated using area under the curve (AUC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Mediation analysis was conducted for inflammatory, hepatic, and renal biomarkers.

Results

Over a mean follow-up of 13.6 years, 17.3% (n = 58,563) of participants developed at least one CRM disease. All six IR-related indices were significantly associated with CRM multimorbidity, both in terms of incidence and progression. In predicting the incidence of CRM diseases, TyG-WC, TyG-WHtR, and METS-IR demonstrated superior performance. For each 1-standard deviation (SD) increase in TyG-WC, the risks of developing first, double, and triple CRM diseases increased by 51.4%, 88.6%, and 128.7%, respectively (all P < 0.001), with similar gradients observed for TyG-WHtR and METS-IR. Multistate Cox models further confirmed consistent associations between IR-related indices and CRM progression, particularly for TyG-WHtR and TyG. Specifically, a 1-SD increase in TyG-WHtR was associated with a 65.3% increased risk of transitioning from healthy to first CRM disease, 34.6% from first to double, and 26.7% from double to triple CRM diseases (all P < 0.001). In predictive performance evaluation, TyG-WC, TyG-WHtR, and METS-IR achieved the highest AUC, NRI, and IDI values. Mediation analyses indicated that systemic inflammation, organ function, and especially kidney function partially mediated the observed associations.

Conclusion

In summary, IR-related indices, particularly TyG-WC, TyG-WHtR, and METS-IR, were observed to be associated with the presence and progression of CRM diseases. Their potential incorporation into risk assessment and prevention strategies, together with consideration of inflammatory and organ function pathways, might help reduce the burden of CRM multimorbidity; however, further prospective studies are needed to confirm these findings and clarify their clinical relevance.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-025-02928-w.

Introduction

Under the circumstances of increasing age-related population shifts and a surge in chronic disease prevalence, multimorbidity, characterized by the co-occurrence of multiple chronic conditions, has arisen as a major worldwide public health problem [1–3]. Among various disease clusters, cardio-renal-metabolic (CRM) multimorbidity, including cardiovascular disease (CVD), type 2 diabetes mellitus (T2DM), and chronic kidney disease (CKD), represents one of the most common and devastating combinations, contributing substantially to disability, health care burden, and premature mortality. These conditions often co-occur and interact through shared pathophysiological pathways, creating a feedback loop that exacerbates disease progression and complicates clinical management [4–7]. Evidence suggests that CRM multimorbidity does not occur at random but follows a sequential and cumulative trajectory, typically from a disease-free state to the first chronic condition, followed by the accumulation of additional diseases [8, 9]. However, most prior research had aimed at isolated endpoints, such as the incidence of CVD or T2DM, without accounting for this dynamic progression. Understanding how metabolic abnormalities influence transitions across different stages of CRM accumulation is essential for timely prevention and intervention.

Insulin resistance (IR) is a core metabolic disturbance implicated in the development of CVD, T2DM, and CKD [4, 10, 11]. It contributes to systemic inflammation, endothelial dysfunction, oxidative stress, and lipid abnormalities, all of which accelerate CRM disease onset [12]. Although the hyperinsulin-euglycemic clamp is the gold standard for IR assessment, its complexity limits its utility in large-scale population studies. In contrast, surrogate IR-related indices such as the triglyceride-glucose (TyG) index, TyG-body mass index (TyG-BMI), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), triglyceride/high-density lipoprotein cholesterol (TG/HDL-C) ratio, and metabolic score for insulin resistance (METS-IR) offer simple, cost-effective alternatives that have been associated with various chronic disease risks in prior studies [13–17]. Nevertheless, the comparative performance of these indices in predicting CRM multimorbidity across disease progression stages remains poorly understood. Furthermore, the potential biological pathways linking IR and CRM accumulation, particularly the role of systemic inflammation, liver dysfunction, and renal impairment, have not been fully elucidated. We selected these six IR-related indices based on their wide validation in previous literature, practicality for large-scale epidemiological studies, and the fact that they are more readily obtainable than many direct measures. These indices, which leverage routinely available metabolic parameters, have demonstrated correlations with gold-standard measures such as the hyperinsulinemic-euglycemic clamp or established biomarkers like homeostatic model assessment of insulin resistance (HOMA-IR), and to a certain extent, reflect population-level insulin resistance effectively.

Although Tian et al. [18] have employed multistate models to investigate the correlations between IR-related indices and cardiometabolic multimorbidity (CMM), incorporating multiple IR markers, mediation analyses, and discrimination metrics, their investigation primarily focused on the coexistence of CVD and T2DM. Building upon this foundation, our study expands the multimorbidity framework by including CKD, thereby defining a more comprehensive and clinically relevant construct: CRM multimorbidity. This extended disease cluster reflects a more complex and burdensome multimorbidity pattern commonly seen in real-world clinical settings, with greater implications for adverse outcomes and care challenges. Moreover, we incorporated and systematically evaluated six widely used IR-related indices (TyG, TyG-BMI, TyG-WC, TyG-WHtR, TG/HDL-C ratio, and METS-IR), enabling a more comprehensive comparison of their predictive performance. By deepening the scope of disease combinations and expanding the range of IR-related indices assessed, our study contributes additional value beyond methodological replication, enhancing the understanding of IR in the context of complex multimorbidity and informing more integrated prevention and management strategies.

Therefore, leveraging the rich longitudinal records from the UK Biobank, we focused on: (1) examine the correlations between six IR-related indices and the incidence of CRM diseases; (2) evaluate their predictive utility using multiple discrimination metrics; (3) investigate the role of these indices in disease progression using multistate modeling; and (4) explore the mediating effects of inflammation, liver function, and renal biomarkers. Our findings aim to enhance CRM disease prediction and shed light on mechanistic insights that may inform prevention strategies.

Methods

Study design and participants

The present study utilized data from the UK Biobank, a large-scale prospective cohort comprising more than 500,000 individuals recruited across England, Scotland, and Wales during the period from 2006 to 2010. Data collection included touchscreen questionnaires, physical measurements, and the collection of biological samples from participants. Further details about the cohort have been reported previously [19]. Ethical approval was granted by the North West Multicenter Research Ethics Committee, and all participants provided written informed consent before being enrolled in the study. The present analysis was performed based on data obtained from the UK Biobank under approved Application Number 177024.

Of the 501,961 participants in the UK Biobank study, we excluded those with a prior diagnosis of CVD (including ischemic heart disease (IHD) and stroke), T2DM, and CKD (n = 48,101), those without complete data on IR-related indices (n = 67,482), those with incomplete covariate data and potential mediator biomarker data (n = 58,686). Ultimately, 327,692 participants constituted the study population for the association analyses (Fig. 1A). To assess potential selection bias, baseline characteristics of included and excluded participants were compared, with results presented in Table S1.

Fig. 1.

Study design and criteria. A Enrollment flowchart. B Transition patterns. CRM: cardio-renal-metabolic, IR: insulin resistance, CRMD: cardio-renal-metabolic disease, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease

Evaluation of IR-related indices and blood biomarkers

At the initial evaluation, peripheral blood samples were randomly obtained from participants in the UK Biobank to measure biochemical markers such as glucose, TG, and HDL-C. Based on previously validated studies [20–22], six IR-related indices were obtained by means of the following established formulas:

In addition, blood biomarkers selected based on well-established pathophysiological mechanisms and validated through rigorous external quality control, were incorporated to assess their possible mediating effects in the correlations between six IR-related indices and CRM diseases [23–27]. The selected biomarkers encompassed three major categories: (1) systemic inflammatory markers, including C-reactive protein (CRP), white blood cell count (WBC), neutrophil count, monocyte count, lymphocyte count, and platelet count; (2) indicators of liver function, such as alanine aminotransferase (ALT), alkaline phosphatase (ALP), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), total bilirubin, total protein, and albumin; and (3) renal function markers, including cystatin C, creatinine, urate, and urea [18].

Evaluation of outcomes

We defined the primary outcome of this study as the occurrence of newly developed CRM diseases among participants during follow-up. Disease outcomes were identified based on diagnostic codes from the 10th revision of the International Classification of Diseases (ICD-10). The definition of CRM diseases was adapted from established criteria reported in prior studies [7, 9, 28]. We identified the first CRM disease as the first occurrence of any of the following during follow-up: CVD (including IHD (I20-I25) and stroke (I60-I69)), T2DM (E11) or CKD (N18) [29–31]. The presence of two coexisting CRM diseases was classified as double CRM diseases, while the simultaneous occurrence of all three conditions was defined as triple CRM diseases. We followed all participants from the date of enrollment in the UK Biobank until the earliest of the following events: incident outcome, death, loss to follow-up, or the end of the follow-up period. The final follow-up date was defined as May 31, 2024, and outcome data were obtained through linkage with national death registries.

Covariates

Based on existing literature and clinical practice [18, 31–33], we involved the following covariates: age (continuous, in years), sex, race, Townsend deprivation index (TDI) (continuous, with larger values reflecting increased deprivation), education levels (high (college or university degree, nursing, teaching and others), intermediate (A (advanced)/AS levels or equivalent, O (ordinary) levels/General Certificate of Secondary Education or equivalent and Certificate of Secondary Education or equivalent), low (National Vocational Qualification or Higher National Diploma or equivalent) or other (none of the above)) [34], smoking status (never, ever, current), drinking status (never, ever, current), sleep duration (< 7 h/day, 7-8 h/day, > 8 h/day), systolic and diastolic blood pressure (SBP and DBP, continuous, in mmHg), glycated hemoglobin (HbA1c, continuous, in mmol/L), and low-density lipoprotein cholesterol (LDL-C, continuous, in mmol/L).

Statistical analysis

Baseline demographic and clinical features were summarized through descriptive statistics, with participants grouped by the development status of their first CRM disease. Continuous variables are expressed as means ± standard deviations (SD) or medians with interquartile ranges (IQR), while categorical variables are reported as counts and percentages. Group comparisons were conducted using t-tests for normally distributed continuous variables, Kruskal-Wallis tests for those not normally distributed, and Pearson’s chi-square tests for categorical variables.

We employed multivariable Cox regression to investigate how six IR-related indices, entered as continuous variables standardized by SD increments, relate to the risk of CRM diseases (including first, double, and triple CRM diseases) and their individual components, namely IHD, stroke, T2DM, and CKD. Model 1 adjusted for age, sex, and race; and Model 2 incorporated further covariates including TDI, education level, smoking status, alcohol status, sleep duration, HbA1c, and LDL-C. Furthermore, the Kaplan-Meier (KM) method was conducted to compare the cumulative incidence of CRM diseases across quartiles of six IR-related indices, with statistical differences evaluated by the log-rank test. Additionally, restricted cubic spline (RCS) functions with four knots were employed to graphically depict the dose-response associations between six IR-related indices and the incidence of CRM diseases. Similar to previous studies [22, 35], to assess the added value of IR-related indices, we constructed a standard model using Model 2 covariates and an extended model that also included IR-related indices. We further calculated the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) to assess whether adding IR-related indices to the traditional risk model improved CRM diseases prediction [36]. To assess the discriminative ability of IR-related indices for predicting CRM diseases, receiver operating characteristic (ROC) curve analysis was performed, and we further calculate the area under the curve (AUC) to quantify the predictive accuracy of each index. The NRI, IDI, and ROC metrics in this study were derived from Cox regression analyses.

The multi-state model builds upon the competing risk framework to enable a thorough assessment of how risk factors affect different phases of disease development, including progression and prognosis. Previous literature has detailed this methodology [29, 37]. In the present study, we conducted a multi-state model to explore the impact of six IR-related indices on the temporal progression of CRM diseases, tracing transitions from a disease-free state to the onset of first, double, and triple CRM diseases. As illustrated in Fig. 1B, the progression trajectory of CRM diseases was divided into three transition stages: (1) from baseline (disease-free) to the onset of the first CRM disease, (2) from the first to double CRM diseases, and (3) from double to triple CRM diseases. For participants who transitioned between stages on the same day, the entry date for the earlier stage was set as 0.5 days prior to that of the subsequent stage, consistent with approaches used in previous research [29, 38]. For instance, the onset date of the first CRM disease was assigned as 0.5 days earlier than the date marking the transition to double CRM diseases. The progression from single to double to triple CRM diseases illustrates the accumulation of multiple comorbidities over time. While the first disease is included in subsequent multiple CRM categories, this framework is intended to describe the overall burden and pattern of CRM accumulation rather than a strictly sequential causal pathway. In secondary analyses, we further investigated the varying correlations of the six IR-related indices with disease progression according to the subtype of the initial CRM disease (IHD, stroke, T2DM, and CKD), including transitions from baseline to first CRM disease subtype, from first to double CRM diseases, and from double to triple CRM diseases. For all transition pathways, death was incorporated as an absorbing state to appropriately account for participants leaving the disease progression process.

For mediation analysis, the IR-related indices were chosen as exposure variables. Following established methodologies [27, 39], two models were constructed to evaluate the possible mediating roles of selected biomarkers. We conducted linear regression analyses to explore the correlations between IR-related indices and the biomarkers, while multivariable-adjusted Cox proportional hazards models explored the correlations between these biomarkers and CRM diseases. Biomarkers demonstrating statistical significance and consistent directional effects across both models were considered potential mediators. We further performed mediation analyses utilizing the ‘mediation’ package in R. We estimated the proportion mediated (PM), with 95% confidence intervals (CIs) derived through nonparametric bootstrap resampling based on 1,000 iterations.

We performed subgroup analyses to investigate the consistency of associations across key demographic and lifestyle factors, including age (< 60 vs. ≥60 years), sex, overweight/obesity status (BMI < 25 vs. BMI ≥ 25 kg/m²), smoking status (never vs. ever), and alcohol status (never vs. ever). Subsequently, sensitivity analyses were conducted by excluding participants who developed CRM diseases within the first two years of follow-up to minimize potential reverse causation, and those with fasting time less than 8 hours to reduce measurement bias in IR-related indices. Additional analyses were performed including baseline use of antihypertensive, antidiabetic, or lipid-lowering medications as covariates, and Fine-Gray competing risk models were applied to account for all-cause death. Statistical analyses were carried out using R version 4.4.0 (R Foundation for Statistical Computing). Statistical significance in biomarker-related analyses was determined by an FDR-adjusted P value (Benjamini-Hochberg method) of less than 0.05. In all remaining analyses, a two-sided P value of < 0.05 indicated statistical significance.

Results

Baseline characteristics of study population

Among the 327,692 participants enrolled in the final analysis, 58,563 (17.3%) developed at least one CRM disease during a mean follow-up of 13.6 years. As shown in Table 1, baseline characteristics differed significantly between individuals with and without CRM disease. Compared with those without CRM disease, affected individuals were older (59.74 ± 7.13 vs. 55.27 ± 8.06 years), more prone to be male (55.4% vs. 42.2%), and had larger socioeconomic deprivation as indicated by a less negative TDI (− 1.11 vs. −1.49) (all P < 0.001). They also had slightly higher proportions of Asian ethnicity (2.5% vs. 1.5%, P < 0.001) and lower levels of educational attainment, with fewer individuals holding higher education degrees (P < 0.001). Unhealthy lifestyle factors were more prevalent in the CRM disease group, including higher rates of current and former smoking, and more individuals reporting never or former alcohol use (all P < 0.001). SBP and DBP levels were also significantly elevated (all P < 0.001). Metabolic markers, including TyG, TyG-BMI, TyG-WC, TyG-WHtR, TG/HDL-C ratio, METS-IR, and HbA1c, were consistently higher in the CRM disease group (all P < 0.001), reflecting greater metabolic disturbance. Similarly, inflammatory markers (CRP, WBC, neutrophils, monocytes, and lymphocytes) and liver enzymes (ALT, AST, ALP, GGT) were elevated, while albumin was slightly reduced (all P < 0.001). Indicators of kidney function including cystatin C, creatinine, urate, and urea, were also significantly increased in individuals with CRM disease (all P < 0.001). In summary, participants with CRM disease exhibited an overall less favorable clinical and biochemical profile, characterized by adverse demographic, socioeconomic, lifestyle, metabolic, inflammatory, hepatic, and renal features.

Table 1.

Baseline characteristics of study population

Characteristics	Total population	CRM occurrence		P value
		No	Yes
N	327,692	269,129	58,563
Age (years)	56.06 (8.08)	55.27 (8.06)	59.74 (7.13)	< 0.001
Sex (%)				< 0.001
Female	181,703 (55.4)	155,576 (57.8)	26,127 (44.6)
Male	145,989 (44.6)	113,553 (42.2)	32,436 (55.4)
Ethnicity (%)				< 0.001
Asian	5,466 (1.7)	3,998 (1.5)	1,468 (2.5)
Black	4,290 (1.3)	3,391 (1.3)	899 (1.5)
Mixed	1,944 (0.6)	1,628 (0.6)	316 (0.5)
Other	4,800 (1.5)	3,958 (1.5)	842 (1.4)
White European	311,192 (95.0)	256,154 (95.2)	55,038 (94.0)
TDI	-1.42 (3.01)	-1.49 (2.96)	-1.11 (3.17)	< 0.001
Education (%)				< 0.001
High	159,260 (48.6)	135,512 (50.4)	23,748 (40.6)
Intermediate	98,333 (30.0)	82,378 (30.6)	15,955 (27.2)
Low	18,955 (5.8)	14,673 (5.5)	4,322 (7.4)
Other	51,104 (15.6)	36,566 (13.6)	14,538 (24.8)
Sleep duration (%)				< 0.001
< 7 h/day	80,651 (24.6)	64,365 (23.9)	16,286 (27.8)
7–8 h/day	223,969 (68.3)	187,052 (69.5)	36,917 (63.0)
> 8 h/day	23,072 (7.0)	17,712 (6.6)	5,360 (9.2)
Smoking status (%)				< 0.001
Never	183,065 (55.9)	154,951 (57.6)	28,114 (48.0)
Previous	110,793 (33.8)	88,341 (32.8)	22,558 (38.5)
Current	33,834 (10.3)	25,943 (9.6)	7,891 (13.5)
Alcohol status (%)				< 0.001
Never	13,127 (4.0)	9,999 (3.7)	3,128 (5.3)
Previous	10,581 (3.2)	7,862 (2.9)	2,719 (4.6)
Current	303,984 (92.8)	251,268 (93.4)	52,716 (90.0)
SBP (mmHg)	137.65 (18.63)	136.30 (18.32)	143.87 (18.77)	< 0.001
DBP (mmHg)	82.42 (10.12)	81.97 (10.04)	84.47 (10.21)	< 0.001
TyG	8.69 (0.55)	8.65 (0.53)	8.91 (0.59)	< 0.001
TyG-BMI	237.00 (47.77)	232.37 (45.18)	258.32 (53.28)	< 0.001
TyG-WC	781.23 (142.83)	765.48 (136.21)	853.61 (150.04)	< 0.001
TyG-WHtR	4.63 (0.80)	4.54 (0.76)	5.05 (0.85)	< 0.001
TG/HDL-C ratio	3.07 (2.45)	2.90 (2.29)	3.87 (2.91)	< 0.001
METS-IR	39.61 (8.95)	38.71 (8.42)	43.75 (10.08)	< 0.001
HbA1c (mmol/L)	35.45 (5.48)	34.76 (4.14)	38.63 (8.77)	< 0.001
LDL-C (mmol/L)	3.63 (0.84)	3.64 (0.82)	3.61 (0.91)	< 0.001
CRP (mg/L)	2.51 (4.20)	2.33 (3.98)	3.36 (4.99)	< 0.001
WBC (109 cells/Litre)	6.83 (1.98)	6.74 (1.94)	7.23 (2.09)	< 0.001
Neutrophil count (109 cells/Litre)	4.19 (1.39)	4.13 (1.36)	4.45 (1.48)	< 0.001
Monocyte count (109 cells/Litre)	0.47 (0.21)	0.46 (0.20)	0.51 (0.27)	< 0.001
Lymphocyte count (109 cells/Litre)	1.96 (1.07)	1.94 (1.06)	2.04 (1.11)	< 0.001
Platelet count (109 cells/Litre)	253.77 (59.32)	254.18 (58.62)	251.91 (62.42)	< 0.001
ALT (U/L)	23.21 (13.88)	22.62 (13.32)	25.93 (15.91)	< 0.001
ALP (U/L)	82.97 (25.25)	81.92 (24.66)	87.79 (27.29)	< 0.001
AST (U/L)	25.99 (9.86)	25.66 (9.41)	27.50 (11.57)	< 0.001
GGT (U/L)	36.20 (39.46)	34.34 (36.42)	44.76 (50.27)	< 0.001
Total bilirubin (umol/L)	9.11 (4.37)	9.13 (4.40)	9.04 (4.23)	< 0.001
Total protein (g/L)	72.53 (4.07)	72.51 (4.04)	72.60 (4.21)	< 0.001
Albumin (g/L)	45.27 (2.59)	45.35 (2.58)	44.94 (2.65)	< 0.001
Cystatin C (mg/L)	0.89 (0.14)	0.88 (0.13)	0.97 (0.18)	< 0.001
Creatinine (umol/L)	71.52 (14.59)	70.68 (13.55)	75.40 (18.13)	< 0.001
Urate (umol/L)	306.05 (78.81)	300.19 (76.97)	333.00 (81.52)	< 0.001
Urea (mmol/L)	5.34 (1.29)	5.28 (1.22)	5.64 (1.51)	< 0.001

CRM: cardio-renal-metabolic, TDI: Townsend deprivation index, SBP: systolic blood pressure, DBP: diastolic blood pressure, TyG: triglyceride-glucose index, BMI: body mass index, WC: waist circumference, WHtR: weight-to-height ratio, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, METS-IR: metabolic score for insulin resistance, HbA1c: glycated hemoglobin, LDL-C: low-density lipoprotein cholesterol, CRP: C-reactive protein, WBC: white blood cell count, ALT: alanine aminotransferase, ALP: alkaline phosphatase, AST: aspartate aminotransferase, GGT: gamma-glutamyltransferase

Correlations of six IR-related indices with CRM diseases

In the Cox model analysis, all six IR-related indices were associated with the risk of CRM diseases, with TyG-WC, TyG-WHtR, and METS-IR showing comparatively better predictive performance for CRM diseases and CRM components (Table 2). Per 1-standard deviation (SD) increase in TyG-WC, the risk of developing a first CRM disease increased by 51.4%, double CRM diseases by 88.6%, and triple CRM diseases by 128.7% (all P < 0.001). For TyG-WHtR, the corresponding risk increases were 48.5%, 82.8%, and 119.9%, and for METS-IR, the increases were 44.7%, 72.6%, and 99.4%, respectively (all P < 0.001). Regarding individual conditions, all indices were relatively stronger associated with the incidence of T2DM, with hazard ratios (HRs) ranging from 1.333 to 2.333. Notably, TyG-WC, TyG-WHtR, and METS-IR demonstrated relatively stronger associations with T2DM risk. Per 1-SD increase in TyG-WC, TyG-WHtR, and METS-IR, the risk of developing T2DM increased by 133.3%, 123.2%, and 97.8%, respectively (all P < 0.001). In the case of stroke, most indices retained statistical significance after adjustment, except for the TyG index. Similar to T2DM, TyG-WC, TyG-WHtR, and METS-IR demonstrated comparatively better predictive performance for stroke risk. Per 1-SD increase in TyG-WC, TyG-WHtR, and METS-IR, the risk of developing stroke increased by 6.6%, 7.4%, and 4.7%, respectively (all P < 0.001). For CKD, adjusted HRs ranged from 1.176 to 1.422. TyG-WC, TyG-WHtR, and METS-IR also demonstrated relatively stronger associations with CKD risk. Per 1-SD increase in TyG-WC, TyG-WHtR, and METS-IR, the risk of developing CKD increased by 42.2%, 37.7%, and 38.5%, respectively (all P < 0.001). Overall, indices incorporating central adiposity measures (TyG-WC and TyG-WHtR) and METS-IR tended to be associated with comparatively higher risk estimates for CRM outcomes.

Table 2.

Associations between IR-related indices and risk of CRM diseases

	Model 1		Model 2
	HR (95% CI)	P value	HR (95% CI)	P value
TyG
First CRM disease	1.442 (1.430, 1.454)	< 0.001	1.325 (1.314, 1.336)	< 0.001
Double CRM diseases	1.753 (1.721, 1.787)	< 0.001	1.549 (1.520, 1.577)	< 0.001
Triple CRM diseases	2.130 (2.022, 2.244)	< 0.001	1.763 (1.678, 1.853)	< 0.001
IHD	1.242 (1.228, 1.258)	< 0.001	1.139 (1.124, 1.154)	< 0.001
Stroke	1.067 (1.048, 1.087)	< 0.001	1.011 (0.991, 1.030)	0.290
T2DM	2.205 (2.177, 2.234)	< 0.001	1.891 (1.868, 1.914)	< 0.001
CKD	1.296 (1.274, 1.319)	< 0.001	1.240 (1.218, 1.262)	< 0.001
TyG-BMI
First CRM disease	1.550 (1.539, 1.562)	< 0.001	1.428 (1.417, 1.439)	< 0.001
Double CRM diseases	1.901 (1.872, 1.930)	< 0.001	1.702 (1.675, 1.729)	< 0.001
Triple CRM diseases	2.275 (2.185, 2.369)	< 0.001	1.974 (1.893, 2.058)	< 0.001
IHD	1.312 (1.297, 1.327)	< 0.001	1.225 (1.210, 1.240)	< 0.001
Stroke	1.107 (1.088, 1.127)	< 0.001	1.034 (1.015, 1.054)	< 0.001
T2DM	2.187 (2.165, 2.208)	< 0.001	1.971 (1.951, 1.990)	< 0.001
CKD	1.460 (1.438, 1.482)	< 0.001	1.366 (1.345, 1.388)	< 0.001
TyG-WC
First CRM disease	1.674 (1.660, 1.688)	< 0.001	1.514 (1.501, 1.528)	< 0.001
Double CRM diseases	2.177 (2.137, 2.217)	< 0.001	1.886 (1.851, 1.921)	< 0.001
Triple CRM diseases	2.777 (2.643, 2.919)	< 0.001	2.287 (2.175, 2.406)	< 0.001
IHD	1.363 (1.346, 1.380)	< 0.001	1.256 (1.239, 1.274)	< 0.001
Stroke	1.147 (1.125, 1.169)	< 0.001	1.066 (1.045, 1.089)	< 0.001
T2DM	2.713 (2.680, 2.747)	< 0.001	2.333 (2.305, 2.362)	< 0.001
CKD	1.531 (1.505, 1.558)	< 0.001	1.422 (1.395, 1.448)	< 0.001
TyG-WHtR
First CRM disease	1.638 (1.625, 1.651)	< 0.001	1.485 (1.473, 1.497)	< 0.001
Double CRM diseases	2.106 (2.070, 2.142)	< 0.001	1.828 (1.797, 1.860)	< 0.001
Triple CRM diseases	2.660 (2.540, 2.786)	< 0.001	2.199 (2.098, 2.304)	< 0.001
IHD	1.361 (1.345, 1.377)	< 0.001	1.254 (1.237, 1.270)	< 0.001
Stroke	1.157 (1.137, 1.178)	< 0.001	1.074 (1.053, 1.095)	< 0.001
T2DM	2.577 (2.547, 2.606)	< 0.001	2.232 (2.206, 2.257)	< 0.001
CKD	1.485 (1.462, 1.510)	< 0.001	1.377 (1.354, 1.401)	< 0.001
TG/HDL-C ratio
First CRM disease	1.266 (1.259, 1.274)	< 0.001	1.206 (1.198, 1.213)	< 0.001
Double CRM diseases	1.361 (1.345, 1.377)	< 0.001	1.274 (1.258, 1.289)	< 0.001
Triple CRM diseases	1.359 (1.326, 1.393)	< 0.001	1.248 (1.216, 1.280)	< 0.001
IHD	1.187 (1.176, 1.199)	< 0.001	1.131 (1.119, 1.142)	< 0.001
Stroke	1.070 (1.052, 1.088)	< 0.001	1.025 (1.007, 1.043)	0.005
T2DM	1.417 (1.407, 1.427)	< 0.001	1.333 (1.324, 1.343)	< 0.001
CKD	1.226 (1.210, 1.242)	< 0.001	1.176 (1.160, 1.191)	< 0.001
METS-IR
First CRM disease	1.580 (1.568, 1.591)	< 0.001	1.447 (1.436, 1.458)	< 0.001
Double CRM diseases	1.950 (1.920, 1.979)	< 0.001	1.726 (1.700, 1.754)	< 0.001
Triple CRM diseases	2.339 (2.247, 2.435)	< 0.001	1.994 (1.913, 2.080)	< 0.001
IHD	1.332 (1.317, 1.347)	< 0.001	1.252 (1.237, 1.267)	< 0.001
Stroke	1.123 (1.103, 1.143)	< 0.001	1.047 (1.028, 1.067)	< 0.001
T2DM	2.236 (2.214, 2.258)	< 0.001	1.978 (1.959, 1.998)	< 0.001
CKD	1.493 (1.471, 1.516)	< 0.001	1.385 (1.363, 1.407)	< 0.001

IR: insulin resistance, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, TyG: triglyceride-glucose index, BMI: body mass index, WC: waist circumference, WHtR: weight-to-height ratio, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, METS-IR: metabolic score for insulin resistance, HbA1c: glycated hemoglobin, LDL-C: low-density lipoprotein cholesterol, HR: hazard ratio, CI: confidence interval

Model 1 was adjusted for age, sex, and ethnicity

Model 2 was adjusted for age, sex, ethnicity, Townsend deprivation index, educational levels, smoking status, alcohol status, sleep duration, HbA1c, and LDL-C

As shown in Fig. 2, Kaplan-Meier survival curves stratified by quartiles of the six IR-related indices demonstrated a clear stepwise increase in the cumulative incidence of first CRM disease. Individuals in the highest quartile (Q4) consistently exhibited the highest event rates, whereas those in the lowest quartile (Q1) had the lowest, indicating a robust dose-response trend across all indices. This gradient was evident not only for the initial occurrence of CRM disease but also for its progression to multiple comorbidities, supporting the potential clinical relevance of elevated IR levels in multimorbidity development (Fig. S1). RCS analyses further revealed significant non-linear associations between six indices and CRM disease risk (all P for non-linearity < 0.001). Across all indices, risk increased progressively with higher values, with steeper risk elevations observed in the upper ranges. Notably, TyG-WC and TyG-WHtR displayed more accentuated upward trajectories, suggesting comparatively stronger associations with CRM disease risk at elevated index levels (Fig. 3 and Fig. S2).

Fig. 2.

Kaplan-Meier curves of incident first CRM disease stratified by the quartiles of IR-related indices. IR: insulin resistance, CRM: cardio-renal-metabolic, TyG: triglyceride-glucose index, BMI: body mass index, WC: waist circumference, WHtR: weight-to-height ratio, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, METS-IR: metabolic score for insulin resistance

Fig. 3.

Dose-response relationship of IR-related indices with the risk of first CRM disease. IR: insulin resistance, CRM: cardio-renal-metabolic, TyG: triglyceride-glucose index, BMI: body mass index, WC: waist circumference, WHtR: weight-to-height ratio, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, METS-IR: metabolic score for insulin resistance

ROC curve analyses were conducted to evaluate the predictive performance of the six IR-related indices (Fig. S3). For the prediction of the first CRM disease, AUC values ranged from 0.743 to 0.755, with METS-IR demonstrating the highest AUC (0.755; 95% CI: 0.753-0.757), followed by TyG-WC and TyG-WHtR. For double CRM diseases, AUCs ranged from 0.805 to 0.827, with TyG-WHtR achieving the highest AUC (0.827; 95% CI: 0.823-0.831), followed by TyG-WC and METS-IR. Prediction of triple CRM diseases showed further improvement, with AUC values ranging from 0.860 to 0.892. TyG-WHtR again exhibited the greatest discriminative ability (AUC = 0.892; 95% CI: 0.883-0.900), followed by METS-IR and TyG-WC. In contrast, the TG/HDL-C ratio consistently showed relatively lower predictive capacity across all stages of CRM multimorbidity. Overall, among the IR-related indices evaluated, TyG-WHtR, TyG-WC, and METS-IR demonstrated comparatively stronger and more consistent predictive performance across the continuum of CRM disease burden, underscoring their potential clinical utility in risk assessment and stratification. These indices may therefore contribute to the development of future predictive models for CRM multimorbidity, offering practical and accessible measures for risk stratification.

Role of six IR-related indices in the trajectory of CRM multimorbidity progression

Multistate Cox regression analyses indicated that all six IR-related indices assessed in this study were significantly associated with the progression of CRM multimorbidity (Tables 3, 4, 5, 6, 7 and 8). Notably, TyG and TyG-WHtR demonstrated comparatively stronger and more consistent associations across all stages of transition. Specifically, each 1-SD increase in TyG-WHtR was associated with a 65.3% increased risk of transitioning from healthy to first CRM disease, 34.6% from first to double CRM diseases, and 26.7% from double to triple CRM diseases (all P < 0.001). For TyG, the respective increases were 68.9%, 29.3%, and 14.5% (all P < 0.05). When transitions from baseline to individual first CRM disease were examined, the IR-related indices were most strongly associated with the onset of T2DM. Per 1-SD increase in TyG and TyG-WHtR, the risk of developing T2DM increased by 225.3% and 177.4%, respectively (both P < 0.001). Associations with CKD were also prominent, each 1-SD increase in TyG and TyG-WHtR, the risk of developing CKD increased by 41.4% and 42.7%, respectively (both P < 0.001). For IHD, per 1-SD increase in TyG and TyG-WHtR, the risk of developing IHD increased by 27.7% and 30.9%, respectively (both P < 0.001). In contrast, associations with stroke were relatively weaker. Most indices lost statistical significance after adjustment, suggesting that the role of IR in cerebrovascular disease may be more indirect or constrained by fewer events. These indices lost their significance in the transition from CKD to double CRM, suggesting that their predictive value may diminish once CKD has developed. Importantly, given the very large sample size, some hazard ratios for individual associations were modest (e.g., close to 1.01), which may raise concerns about inflated statistical significance. However, even small relative risks can translate into a substantial absolute number of events at the population level, particularly when the disease or risk factor is common. This implies that, while the effect size for an individual may be small, the overall public health impact can be meaningful, helping to identify patterns of risk accumulation across large cohorts. Moreover, modest HRs can provide useful information for population-level surveillance, early detection of high-risk subgroups, and prioritization of preventive strategies, especially when combined with other risk indicators. Therefore, these findings should be interpreted with a dual perspective: cautiously at the individual level, but acknowledging their potential implications for population health management and strategic intervention planning. In summary, all six IR-related indices appeared to be associated with the progression of CRM multimorbidity to varying degrees, with TyG and TyG-WHtR showing relatively stronger and more consistent associations across different transition stages. These findings suggest their potential relevance in identifying individuals at elevated risk for disease accumulation, although further validation is warranted.

Table 3.

 Association between TyG and the trajectory of CRM multimorbidity by multi-state model

Transition	HR (95% CI)	P value
Model 1: Baseline→First CRM disease→Double CRM diseases→Triple CRM diseases
Baseline→First CRM disease	1.6889 (1.6631, 1.7151)	< 0.001
First CRM disease→Double CRM diseases	1.2925 (1.2476, 1.3391)	< 0.001
Double CRM diseases→Triple CRM diseases	1.1453 (1.0291, 1.2747)	0.013
Model 2: Baseline→First CRM disease subtypes (IHD, Stroke, T2DM, and CKD)→Double CRM diseases→Triple CRM diseases
Baseline→IHD	1.2773 (1.2433, 1.3123)	< 0.001
Baseline→Stroke	0.9667 (0.9267, 1.0084)	0.116
Baseline→T2DM	3.2527 (3.1754, 3.3318)	< 0.001
Baseline→CKD	1.4136 (1.3558, 1.4738)	< 0.001
IHD→Double CRM diseases	2.0002 (1.8839, 2.1236)	< 0.001
Stroke→Double CRM diseases	0.7977 (0.7371, 0.8633)	< 0.001
T2DM→Double CRM diseases	1.6903 (1.5120, 1.8896)	< 0.001
CKD→Double CRM diseases	0.9701 (0.8935, 1.0533)	0.470
Double CRM diseases→Triple CRM diseases	1.1453 (1.0291, 1.2747)	0.013

TyG: triglyceride-glucose index, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, HR: hazard ratio, CI: confidence interval

Table 4.

Association between TyG-BMI and the trajectory of CRM multimorbidity by multi-state model

Transition	HR (95% CI)	P value
Model 1: Baseline→First CRM disease→Double CRM diseases→Triple CRM diseases
Baseline→First CRM disease	1.0076 (1.0075, 1.0078)	< 0.001
First CRM disease→Double CRM diseases	1.0044 (1.0040, 1.0048)	< 0.001
Double CRM diseases→Triple CRM diseases	1.0032 (1.0021, 1.0043)	< 0.001
Model 2: Baseline→First CRM disease subtypes (IHD, Stroke, T2DM, and CKD)→Double CRM diseases→Triple CRM diseases
Baseline→IHD	1.0041 (1.0038, 1.0044)	< 0.001
Baseline→Stroke	1.0000 (0.9995, 1.0005)	0.906
Baseline→T2DM	1.0146 (1.0144, 1.0148)	< 0.001
Baseline→CKD	1.0061 (1.0057, 1.0065)	< 0.001
IHD→Double CRM diseases	1.0085 (1.0078, 1.0091)	< 0.001
Stroke→Double CRM diseases	0.9989 (0.9980, 0.9999)	0.025
T2DM→Double CRM diseases	1.0089 (1.0077, 1.0101)	< 0.001
CKD→Double CRM diseases	0.9993 (0.9984, 1.0003)	0.192
Double CRM diseases→Triple CRM diseases	1.0032 (1.0021, 1.0043)	< 0.001

TyG: triglyceride-glucose index, BMI: body mass index, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, HR: hazard ratio, CI: confidence interval

Table 5.

Association between TyG-WC and the trajectory of CRM multimorbidity by multi-state model

Transition	HR (95% CI)	P value
Model 1: Baseline → First CRM disease → Double CRM diseases → Triple CRM diseases
Baseline → First CRM disease	1.0030 (1.0029, 1.0030)	< 0.001
First CRM disease → Double CRM diseases	1.0017 (1.0016, 1.0019)	< 0.001
Double CRM diseases → Triple CRM diseases	1.0014 (1.0009, 1.0018)	< 0.001
Model 2: Baseline → First CRM disease subtypes (IHD, Stroke, T2DM, and CKD) → Double CRM diseases → Triple CRM diseases
Baseline → IHD	1.0015 (1.0014, 1.0016)	< 0.001
Baseline → Stroke	1.0002 (1.0000, 1.0003)	0.066
Baseline → T2DM	1.0061 (1.0060, 1.0062)	< 0.001
Baseline → CKD	1.0022 (1.0021, 1.0024)	< 0.001
IHD → Double CRM diseases	1.0035 (1.0032, 1.0037)	< 0.001
Stroke → Double CRM diseases	0.9999 (0.9996, 1.0002)	0.549
T2DM → Double CRM diseases	1.0034 (1.0030, 1.0039)	< 0.001
CKD → Double CRM diseases	1.0000 (0.9996, 1.0003)	0.777
Double CRM diseases → Triple CRM diseases	1.0014 (1.0009, 1.0018)	< 0.001

TyG: triglyceride-glucose index, WC: waist circumference, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, HR: hazard ratio, CI: confidence interval

Table 6.

 Association between TyG-WHtR and the trajectory of CRM multimorbidity by multi-state model

Transition	HR (95% CI)	P value
Model 1: Baseline → First CRM disease → Double CRM diseases → Triple CRM diseases
Baseline → First CRM disease	1.6533 (1.6364, 1.6704)	< 0.001
First CRM disease → Double CRM diseases	1.3463 (1.3144, 1.3790)	< 0.001
Double CRM diseases → Triple CRM diseases	1.2669 (1.1782, 1.3624)	< 0.001
Model 2: Baseline → First CRM disease subtypes (IHD, Stroke, T2DM, and CKD) → Double CRM diseases → Triple CRM diseases
Baseline → IHD	1.3091 (1.2845, 1.3340)	< 0.001
Baseline → Stroke	1.0432 (1.0129, 1.0743)	0.005
Baseline → T2DM	2.7743 (2.7317, 2.8175)	< 0.001
Baseline → CKD	1.4269 (1.3880, 1.4668)	< 0.001
IHD → Double CRM diseases	1.7911 (1.7182, 1.8671)	< 0.001
Stroke → Double CRM diseases	0.9897 (0.9357, 1.0468)	0.717
T2DM → Double CRM diseases	1.7714 (1.6403, 1.9130)	< 0.001
CKD → Double CRM diseases	0.9854 (0.9296, 1.0445)	0.621
Double CRM diseases → Triple CRM diseases	1.2669 (1.1782, 1.3624)	< 0.001

TyG: triglyceride-glucose index, WHtR: weight-to-height ratio, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, HR: hazard ratio, CI: confidence interval

Table 7.

Association between TG/HDL-C ratio and the trajectory of CRM multimorbidity by multi-state model

Transition	HR (95% CI)	P value
Model 1: Baseline → First CRM disease → Double CRM diseases → Triple CRM diseases
Baseline → First CRM disease	1.0814 (1.0787, 1.0841)	< 0.001
First CRM disease → Double CRM diseases	1.0401 (1.0340, 1.0463)	< 0.001
Double CRM diseases → Triple CRM diseases	1.0125 (0.9963, 1.0289)	0.131
Model 2: Baseline → First CRM disease subtypes (IHD, Stroke, T2DM, and CKD) → Double CRM diseases → Triple CRM diseases
Baseline → IHD	1.0568 (1.0519, 1.0617)	< 0.001
Baseline → Stroke	1.0052 (0.9964, 1.0141)	0.249
Baseline → T2DM	1.1261 (1.1224, 1.1297)	< 0.001
Baseline → CKD	1.0704 (1.0625, 1.0784)	< 0.001
IHD → Double CRM diseases	1.0776 (1.0681, 1.0873)	< 0.001
Stroke → Double CRM diseases	0.9565 (0.9408, 0.9726)	< 0.001
T2DM → Double CRM diseases	1.0885 (1.0691, 1.1083)	< 0.001
CKD → Double CRM diseases	0.9893 (0.9713, 1.0077)	0.251
Double CRM diseases → Triple CRM diseases	1.0125 (0.9963, 1.0289)	0.131

TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, HR: hazard ratio, CI: confidence interval

Table 8.

 Association between METS-IR and the trajectory of CRM multimorbidity by multi-state model

Transition	HR (95% CI)	P value
Model 1: Baseline → First CRM disease → Double CRM diseases → Triple CRM diseases
Baseline → First CRM disease	1.0430 (1.0422, 1.0439)	< 0.001
First CRM disease → Double CRM diseases	1.0247 (1.0227, 1.0267)	< 0.001
Double CRM diseases → Triple CRM diseases	1.0175 (1.0115, 1.0235)	< 0.001
Model 2: Baseline → First CRM disease subtypes (IHD, Stroke, T2DM, and CKD) → Double CRM diseases → Triple CRM diseases
Baseline → IHD	1.0249 (1.0233, 1.0265)	< 0.001
Baseline → Stroke	1.0017 (0.9991, 1.0043)	0.201
Baseline → T2DM	1.0808 (1.0795, 1.0821)	< 0.001
Baseline → CKD	1.0347 (1.0323, 1.0370)	< 0.001
IHD → Double CRM diseases	1.0458 (1.0424, 1.0492)	< 0.001
Stroke → Double CRM diseases	0.9932 (0.9884, 0.9981)	0.007
T2DM → Double CRM diseases	1.0495 (1.0430, 1.0561)	< 0.001
CKD → Double CRM diseases	0.9967 (0.9916, 1.0019)	0.214
Double CRM diseases → Triple CRM diseases	1.0175 (1.0115, 1.0235)	< 0.001

METS-IR: metabolic score for insulin resistance, CRM: cardio-renal-metabolic, IHD: ischemic heart disease, T2DM: type 2 diabetes mellitus, CKD: chronic kidney disease, HR: hazard ratio, CI: confidence interval

Incremental predictive values of IR-related indices for CRM diseases incidence

The incremental predictive value analyses (Table 9) demonstrated that all six IR-related indices significantly improved the performance of the conventional model for predicting CRM multimorbidity. Among them, METS-IR showed the highest improvement in reclassification accuracy, with a continuous NRI of 0.382 (95% CI: 0.331-0.392; P < 0.001), followed closely by TyG-WHtR (NRI = 0.372; 95% CI: 0.319-0.381; P < 0.001). TyG-WC and TyG-BMI also yielded substantial improvements (NRI = 0.365 and 0.365, respectively; both P < 0.001). In contrast, the original TyG index (NRI = 0.271) and TG/HDL-C ratio (NRI = 0.277) contributed relatively less to risk reclassification. In terms of discrimination improvement, TyG-WHtR and METS-IR again outperformed other indices, with IDI values of 0.025 (95% CI: 0.017-0.027; P < 0.001) and 0.025 (95% CI: 0.017-0.026; P < 0.001), respectively. TyG-WC and TyG-BMI both showed moderate IDI increases (0.024 and 0.023, respectively), while the TG/HDL-C ratio demonstrated the smallest improvement. Overall, although the improvements in AUC, NRI, and IDI may appear modest, they demonstrate that adding these IR-related indices, particularly METS-IR and TyG-WHtR, enhances the model’s ability to discriminate and correctly reclassify individuals according to their risk of CRM multimorbidity. Clinically, this means that incorporating these biomarkers into conventional risk models could help identify individuals at higher risk who might benefit from closer monitoring, lifestyle interventions, or early therapeutic measures. Compared with existing prediction tools, these biomarkers provide complementary information that improves the precision of risk stratification at the population level.

Table 9.

Incremental predictive values of IR-related indices for first CRM disease

	Continuous NRI			IDI
	Estimate	95% CI	P value	Estimate	95% CI	P value
Conventional model
Conventional model + TyG	0.271	0.209-0.281	< 0.001	0.013	0.006-0.014	< 0.001
Conventional model + TyG-BMI	0.365	0.314-0.374	< 0.001	0.023	0.016-0.024	< 0.001
Conventional model + TyG-WC	0.365	0.315-0.375	< 0.001	0.024	0.017-0.026	< 0.001
Conventional model + TyG-WHtR	0.372	0.319-0.381	< 0.001	0.025	0.017-0.027	< 0.001
Conventional model + TG-HDL-C ratio	0.277	0.250-0.287	< 0.001	0.007	0.006-0.008	< 0.001
Conventional model + METS-IR	0.382	0.331-0.392	< 0.001	0.025	0.017-0.026	< 0.001

IR: insulin resistance, CRM: cardio-renal-metabolic, TyG: triglyceride-glucose index, BMI: body mass index, WC: waist circumference, WHtR: weight-to-height ratio, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, METS-IR: metabolic score for insulin resistance, NRI: net reclassification improvement, IDI: integrated discrimination improvement, HR: hazard ratio, CI: confidence interval

Mediating effects of blood biomarkers

We investigated the associations between IR-related indices and a panel of 17 biomarkers, their individual relationships with CRM disease risk, and their potential mediating effects in the pathway linking IR to CRM diseases (Fig. 4 and Fig. S4). Table S2 systematically presents the associations between six IR-related indices and 17 selected mediators, encompassing systemic inflammation, liver function, and renal function. The results show that all IR-related indices were significantly positively associated with inflammatory markers such as CRP, WBC, and neutrophil count. Notably, TyG and TyG-WHtR exhibited the strongest associations with CRP, suggesting a close link between IR and chronic inflammation. Regarding liver function, IR-related indices were positively associated with enzymes like ALT, AST, and GGT, with TyG showing a particularly high beta value for GGT (11.213), indicating a potential role of IR in hepatic metabolic dysregulation. Most IR-related indices were negatively associated with albumin and total protein levels, which may reflect chronic inflammatory states or nutritional disturbances. In terms of renal function, all IR-related indices were positively associated with creatinine, urate, and urea levels, with TyG-WHtR and TyG demonstrating the strongest associations with urate (beta = 33.547 and 31.136, respectively), further supporting the potential contribution of IR to renal impairment. Overall, all IR-related indices showed relatively strong and extensive correlations across various mediators, suggesting that they may serve not only as markers of IR but also as integrated indicators of metabolic disturbances and multisystem dysfunction.

Fig. 4.

Mediated proportion of selected biomarkers in the associations of six IR-related indices with first CRM disease. IR: insulin resistance, CRM: cardio-renal-metabolic, TyG: triglyceride-glucose index, BMI: body mass index, WC: waist circumference, WHtR: weight-to-height ratio, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, METS-IR: metabolic score for insulin resistance

Table S3 shows that various systemic biomarkers are significantly associated with the risk of CRM multimorbidity progression. Specifically, inflammatory markers such as CRP, WBC, neutrophils, and monocytes consistently demonstrated positive associations across different stages of CRM diseases (from initial onset to double and triple comorbidities), suggesting a potential role of inflammation in disease initiation and progression. Regarding liver function, elevated levels of ALT, AST, ALP, and GGT were positively associated with CRM disease risk, whereas total bilirubin and albumin showed negative associations. Notably, among kidney function markers, cystatin C and urea showed the most prominent associations, with the HR for cystatin C increasing from 4.22 in the first stage to 5.99 in the triple comorbidity stage. Overall, these findings suggest that systemic inflammation and liver-kidney dysfunction may synergistically contribute to CRM multimorbidity progression. Moreover, biomarkers such as cystatin C, monocyte count, and albumin demonstrated stable predictive value across different comorbidity stages and may serve as useful indicators for early identification and risk stratification in clinical settings.

Table S4 shows the results of the mediation analyses evaluating the indirect effects of these 17 biomarkers on the association between IR-related indices and CRM disease risk. We found all six IR-related indices influenced CRM disease risk partially through biological mediators related to systematic inflammation, liver function, and renal function. In the mediation analysis of first CRM disease, five indices (excluding TG/HDL-C ratio) exhibited significant indirect effects primarily through systemic inflammatory markers such as CRP and neutrophil count. Notably, neutrophil count mediated 4.39% of the effect of TyG on first CRM disease risk, suggesting that chronic low-grade inflammation may be a key pathway linking IR to disease risk and reflecting a potential causal relationship between IR and inflammation. Regarding liver function, ALT mediated 7.09% of the effect of TG/HDL-C ratio on first CRM disease risk, indicating that hepatic enzyme abnormalities may also play a mediating role between IR and CRM diseases. In contrast, kidney function biomarkers showed stronger and more consistent mediating effects across all indices. Cystatin C mediated up to 14.16% of the effect of TyG-BMI, while urate mediated 13.56% of the effect of TG/HDL-C ratio. The mediation ranges for cystatin C and urate were 7.84%-14.16% and 7.70%-13.56%, respectively, highlighting kidney dysfunction as a major intermediary in the IR-CRM pathway. It is worth noting that other mediation pathways, although included in the models, had 95% confidence intervals that included zero. This indicates that their indirect effects are uncertain and should be interpreted with caution; accordingly, they are not described in detail in the main text. In analyses of double and triple CRM diseases, the mediating roles of systemic inflammation and liver function were attenuated. Only kidney function markers (cystatin C and urate) retained partial mediation effects. In double CRM diseases, mediation proportions for urate ranged from 8.58% to 15.70%. In triple CRM diseases, significant mediation was observed only for TyG-BMI and METS-IR via kidney function indicators, while other IR-related indices showed no significant indirect effects. In summary, systemic inflammation, liver function, and kidney function may partially mediate the associations between IR-related indices and CRM disease risk, with kidney function showing the most robust and consistent mediating role. These findings suggest that future strategies aimed at reducing IR-related disease risk should consider multi-organ interactions, particularly the role of renal dysfunction, to better inform preventive and therapeutic approaches.

Subgroup and sensitivity analyses

To further explore the associations between IR-related indices and the risk of CRM diseases, we conducted a series of subgroup and sensitivity analyses. Subgroup analyses indicated that these associations remained generally consistent across various stratified populations, including age groups (< 60 vs. ≥60 years), sex (female vs. male), overweight/obesity status (BMI < 25 vs. ≥25 kg/m²), smoking status (never vs. ever smokers), and alcohol consumption (never vs. ever drinkers) (Fig. S5). Notably, relatively stronger associations tended to be observed when first or double CRM diseases were considered as outcomes, particularly among individuals aged < 60 years, females, those with overweight or obesity, and those with a history of smoking or drinking. These subgroups may therefore deserve greater attention in future public health interventions aimed at the early prevention and management of CRM multimorbidity. Sensitivity analyses further reinforced the robustness of our findings. After excluding individuals who developed CRM diseases within the first two years of follow-up (Fig. S6), the associations between IR-related indices and CRM disease risk remained statistically significant, with effect estimates largely comparable to those observed in the primary analyses. In this analysis, similar trends were observed whether IR-related indices were modeled as continuous or categorical variables. Additionally, after excluding participants with fasting time < 8 hours (Fig. S7), the results remained consistent in both magnitude and significance, lending further support to the stability of the observed associations. Similarly, sensitivity analyses further adjusting for the use of antihypertensive, antidiabetic, or lipid-lowering medications yielded results consistent with our main findings (Fig. S8). Moreover, when applying Fine-Gray competing risk models to account for the competing risk of all-cause death, the associations remained robust, with effect estimates comparable to those observed in the primary analyses (Fig. S9). Taken together, these subgroup and sensitivity analyses provide additional support for the potential utility of IR-related indices in predicting CRM disease risk across diverse demographic and clinical contexts, though further validation in independent populations is warranted.

Discussion

Utilizing data from 327,692 individuals in the UK Biobank, this large-scale prospective cohort study investigated the associations of six IR-related indices with the incidence and progression of CRM multimorbidity. All six IR-related indices were associated with the risk of CRM diseases, with TyG-WC, TyG-WHtR, and METS-IR showing comparatively better predictive performance for CRM diseases and CRM components. Multistate Cox regression analyses indicated that all six IR-related indices assessed in this study were significantly associated with the progression of CRM multimorbidity, especially TyG and TyG-WHtR demonstrated comparatively stronger and more consistent associations across all stages of transition. In addition, TyG-WHtR, TyG-WC, and METS-IR also achieved the highest AUC, NRI, and IDI. Mediation analysis indicated that systemic inflammation, liver function, and kidney function may partially mediate the associations between IR-related indices and CRM disease risk, with kidney function showing the most robust and consistent mediating role, offering insights into potential biological pathways that may link IR to CRM multimorbidity.

This study demonstrated that all six IR-related indices were significantly associated with the risk of CRM diseases and their individual components, with HRs showing a clear dose-response pattern as individuals transitioned from having a first disease to double and triple conditions. These findings suggest that IR may play a sustained and cumulative role in the development and progression of multimorbidity, acting as a central metabolic driver across chronic cardiometabolic and renal conditions. However, notable differences in predictive performance were observed among the six IR-related indices. Composite indices incorporating measures of central adiposity, particularly TyG-WC and TyG-WHtR, consistently showed relatively stronger associations with multimorbidity, often yielding higher HRs than the traditional TyG index. This may be attributed to the close relationship between abdominal obesity and insulin resistance, as visceral adipose tissue plays a critical role in metabolic dysregulation through pro-inflammatory cytokine release, insulin signaling disruption, and ectopic fat accumulation. Compared with BMI, waist-based indices may more accurately capture early metabolic dysfunction and systemic inflammation. Our findings align with previous studies. For example, in individuals with metabolic dysfunction-associated steatotic liver disease (MASLD), TyG-WHtR and TyG-WC demonstrated superior prognostic value for all-cause and cardiovascular mortality compared to other TyG-based indices [22, 40]. These indices have also been proposed as early screening tools for non-alcoholic fatty liver disease (NAFLD), metabolic-associated fatty liver disease (MAFLD), and liver fibrosis [41]. Similarly, a study by Dang et al. involving 11,937 U.S. adults found that TyG-WHtR and TyG-WC had better diagnostic accuracy for incident cardiovascular disease and cardiovascular mortality than the TyG index alone [42]. More recently, TyG-WHtR and TyG-WC have been emphasized as simple and cost-effective tools for identifying individuals at elevated risk of CMM in clinical settings [18]. Similarly, METS-IR also demonstrated favorable predictive performance for CRM disease risk, likely due to its integration of multiple metabolic parameters, including BMI, glucose, and triglycerides, which together provide a more comprehensive reflection of systemic insulin resistance and metabolic health. By contrast, the TG/HDL-C ratio showed comparatively weaker associations, which may be due to its focus on lipid metabolism while lacking glycemic or anthropometric components. With regard to outcome-specific heterogeneity, associations were generally strongest for T2DM and CKD, suggesting a more direct metabolic link between IR and these conditions. In contrast, the associations with stroke were the weakest, and in some cases (e.g., TyG), became non-significant after adjustment. This may indicate that IR plays a less dominant role in cerebrovascular pathology, or alternatively, that the limited number of stroke cases in this study reduced the statistical power to detect more robust associations. Importantly, all six indices showed progressively increasing HRs with the accumulation of CRM diseases, reinforcing the potential of IR as a key pathophysiological catalyst in the transition from single to multiple chronic conditions. From a clinical and public health perspective, these IR-related markers, especially TyG-WC, TyG-WHtR, and METS-IR, are non-invasive, cost-efficient, and readily accessible, making them promising tools for early risk stratification and targeted intervention. Future research should further explore the heterogeneity of these indices across specific populations (e.g., older adults, individuals with obesity or metabolic syndrome), and investigate their mechanistic links with chronic inflammation, oxidative stress, and hormonal dysregulation to better elucidate the pathways connecting IR and multimorbidity clustering.

Although numerous studies [10, 11, 15, 16] have explored the associations between IR-related indices and specific CRM conditions, such as T2DM, CKD, or CVD, or their combinations, the majority of these investigations have focused on cross-sectional snapshots of disease status. These studies typically assess whether individuals present with single or multiple conditions at a specific timepoint, without accounting for how IR-related indices influence the temporal progression of CRM diseases across distinct stages. Such a static approach limits our understanding of the continuous and evolving role of IR in disease development. Some recent studies have attempted to bridge this gap. For instance, a study by Tian et al. [18] employed multistate models, mediation analysis, and discrimination metrics to investigate the correlations between IR-related indices and CMM. However, their definition of multimorbidity was restricted to CVD and T2DM, omitting CKD which is a critical component of CRM disease risk. Another recent study [31] highlighted the role of the TyG-BMI index in predicting the transition from a disease-free state to single, double, and triple CRM diseases, indicating a persistent effect of IR throughout stages of disease clustering. However, no comparison of predictive performance of different IR-related indices at each transitional stage was conducted in this study. Building on these prior efforts, our study applied a multistate modeling framework to comprehensively evaluate six IR-related indices in relation to CRM disease progression from a dynamic, stage-specific perspective. Our findings provide important insights into the predictive capacity of various IR-related indices in the development of CRM multimorbidity. Notably, TyG and TyG-WHtR demonstrated relatively stronger and more consistent associations across all transitional stages, suggesting their potential utility as sensitive markers for cumulative metabolic risk. By integrating both glucose-lipid metabolism (TyG) and abdominal obesity (WHtR), TyG-WHtR may better capture the complex metabolic disturbances underlying multimorbidity. This is particularly relevant given the growing recognition of visceral adiposity as a central contributor to chronic inflammation, IR, and organ dysfunction [43, 44]. The strong association observed between IR-related indices and the onset of T2DM aligns with the well-established role of IR in the pathogenesis of diabetes. However, associations with CKD and IHD were also evident, whereas associations with stroke appeared relatively weaker. This heterogeneity might reflect differences in the pathophysiological pathways through which IR affects various organ systems. For instance, IR may contribute more directly to CKD and IHD via atherogenic dyslipidemia and glomerular hyperfiltration [45, 46], while its role in cerebrovascular disease may be more indirect, potentially mediated by hypertension or systemic inflammation, or influenced by a lower number of stroke events in the cohort. Furthermore, calculating transition probabilities for disease progression within each state of a multistate model may provide additional insights into the temporal dynamics of comorbidity trajectories, allowing for a more refined characterization of disease pathways, identification of high-risk subgroups, and potential implications for targeted prevention strategies. We suggest this as a promising direction for future studies.

To further elucidate the potential biological mechanisms linking IR to increased CRM disease risk, we systematically assessed the correlations between six IR-related indices and 17 key biomarkers, and examined their mediating effects on the association between IR and CRM diseases. All IR-related indices showed significant positive correlations with markers of systemic inflammation, with the strongest association observed between TyG-WHtR and CRP. These findings suggest that systemic inflammation may play an important intermediary role in the IR-CRM pathway. Mechanistically, IR can activate inflammatory signaling pathways such as NF-κB, leading to increased secretion of pro-inflammatory cytokines (e.g., IL-6, TNF-α), which promote chronic low-grade inflammation and oxidative stress [47, 48]. These processes may contribute to endothelial dysfunction and atherosclerosis, ultimately resulting in multi-organ damage and increased multimorbidity risk. Previous studies [18, 21] have also reported partial mediation by inflammatory markers in the relationship between TyG-related indices and cardiovascular disease, supporting our observations. In addition, we found consistent positive associations between IR-related indices and liver function markers. Among them, ALT exhibited the strongest mediating effect in the relationship between IR-related indices and CRM diseases. Given the central role of the liver in insulin metabolism, hepatic insulin resistance may lead to increased lipogenesis, elevated inflammatory cytokine release, and higher liver enzyme levels, thereby exacerbating systemic metabolic dysfunction. Renal biomarkers, particularly cystatin C and urate, demonstrated the most prominent mediating effects across first, double, and triple CRM diseases. These results highlight renal dysfunction as a key component in the biological pathway linking IR and multimorbidity, similar to previous research [18]. Mechanistically, IR may contribute to glomerular hyperfiltration, endothelial injury, and tubulointerstitial fibrosis, which in turn impair renal function. Declining kidney function may further amplify systemic inflammation and metabolic dysregulation, forming a deleterious feedback loop. In summary, our findings suggest that systemic inflammation, liver dysfunction, and renal impairment may serve as important mediators linking elevated IR-related indices to CRM disease development. These results improve our understanding of the underlying biological pathways and highlight the potential value of jointly assessing IR-related indices and organ function biomarkers for improved risk stratification. In clinical settings, such combined evaluation could help identify individuals at elevated risk and inform targeted preventive strategies. Further studies are warranted to confirm the generalizability of these mechanisms across diverse populations and explore their potential as therapeutic targets.

From a clinical translation perspective, our findings suggest that IR-related indices, particularly TyG-WHtR, TyG-WC, and METS-IR, may serve as valuable adjuncts to existing risk stratification tools for CRM multimorbidity. TyG-WHtR and TyG-WC, which incorporate waist circumference, exhibited stronger dose-response relationships across CRM progression stages, suggesting a greater sensitivity to metabolic disturbances driven by abdominal obesity and insulin resistance. METS-IR, which integrates BMI, glucose, and triglycerides, demonstrated consistent and robust predictive power across different CRM stages, making it a potentially reliable tool for evaluating systemic metabolic risk. Taken together, incorporating these indices into clinical decision-making and public health screening strategies may facilitate more targeted, early-stage prevention and personalized intervention approaches.

This study has several notable strengths. First, the large sample size from the UK Biobank substantially enhanced the statistical power and reliability of our analyses. Second, comprehensive adjustment for a wide range of potential confounders, along with consistent findings across sensitivity and subgroup analyses, supports the robustness of the results. Third, the availability of detailed biomarker data allowed for in-depth exploration of potential biological mechanisms linking IR with CRM diseases. However, several limitations of this study should also be noted. Firstly, the relatively low response rate of the UK Biobank cohort may have introduced selection bias and affected the generalizability of the findings. To further assess this issue, we compared baseline characteristics between included and excluded participants. While P values were statistically significant for most comparisons due to the large sample size, the standardized mean differences (SMDs) were generally small (mostly < 0.1), with only age and ethnicity slightly above 0.1, suggesting that the risk of substantial selection bias is likely limited (Table S1). Secondly, as the cohort is predominantly composed of White individuals, the generalizability of our findings to other ethnic populations is limited. We acknowledge that the lack of external validation is a limitation of the current study. Future research is warranted to replicate and validate these findings in more diverse cohorts to enhance external validity. Thirdly, baseline undiagnosed subclinical conditions may have influenced insulin resistance status, potentially introducing bias into the observed associations. To address this, sensitivity analyses excluding participants who developed CRM outcomes within the first two years of follow-up were performed, and the results remained consistent with the main analyses. Notably, both IR-related indices and potential mediating biomarkers were measured only at baseline. While repeated measurements could better clarify the temporal relationships between exposures, mediators, and outcomes and strengthen causal inference, in the UK Biobank they were collected only in a relatively small subset of participants who attended the repeat assessment, thereby limiting sample size and potentially introducing selection bias. Therefore, our analyses relied primarily on baseline measurements to ensure sufficient sample size, representativeness, and analytical stability. Future studies with more complete longitudinal data are warranted to further investigate these relationships. Furthermore, the IR-related indices in this study were calculated based on non-fasting samples, which may have introduced some measurement error; however, we addressed this by conducting a sensitivity analysis excluding participants with fasting time less than 8 hours. Notably, this study included only six IR-related indices (TyG, TyG-BMI, TyG-WC, TyG-WHtR, TG/HDL-C ratio, and METS-IR), without assessing classic insulin sensitivity/resistance-derived indices (e.g., HOMA-IR) or gold-standard methods such as the euglycemic-hyperinsulinemic clamp. These indices were selected for their feasibility, reproducibility, and prior validation in large epidemiological cohorts, making them practical for studies involving tens of thousands of participants. Previous studies have shown that these surrogate indices can reasonably reflect insulin resistance, although they are not perfect substitutes. Nevertheless, reliance on surrogate measures may not fully capture the complex physiology of insulin resistance and could introduce measurement bias or underestimate effect sizes. Future studies employing gold-standard methods are warranted to validate and extend these findings. Furthermore, although our findings demonstrate associations between surrogate IR-related indices and CRM outcomes, additional research is needed to explore their clinical applicability. Early identification of high-risk individuals may inform preventive strategies, including tailored lifestyle interventions or pharmacotherapy, and integrating these indices into existing guidelines could potentially improve patient outcomes, although this remains to be confirmed in prospective studies.

Conclusion

In summary, IR-related indices, particularly TyG-WC, TyG-WHtR, and METS-IR, were found to be associated with the presence and progression of CRM diseases. Their potential incorporation into risk assessment and prevention strategies, alongside consideration of inflammatory and organ function pathways, might contribute to reducing the burden of CRM multimorbidity; however, further prospective studies are needed to confirm these findings and assess their clinical applicability.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ALP

Alkaline phosphatase

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

AUC

Area under the curve

BMI

Body mass index

CI

Confidence interval

CKD

Chronic kidney disease

CRM

Cardio-renal-metabolic

CRP

C-reactive protein

CVD

Cardiovascular disease

GGT

Gamma-glutamyltransferase

HbA1c

Glycated hemoglobin

HDL-C

High-density lipoprotein cholesterol

HR

Hazard ratio

IDI

Integrated discrimination improvement

IHD

Ischemic heart disease

IQR

Interquartile ranges

IR

Insulin resistance

KM

Kaplan-Meier

LDL-C

Low-density lipoprotein cholesterol

NRI

Net reclassification improvement

PM

Proportion mediation

RCS

Restricted cubic spline

ROC

Receiver operating characteristic

SD

Standard deviation

TDI

Townsend deprivaion index

T2DM

Type 2 diabetes mellitus

TG

Triglyceride

TyG

Triglyceride-glucose

WBC

White blood cell

WC

Waist circumference

WHtR

Weight-to-height ratio

Author contributions

LL contributed to the conception and design of the work. LL and HH performed data acquisition, processing, and interpretation. LL, GY, XJ, YW engaged in literature review and data visualization. LL composed the original manuscript, while all authors engaged in essential revisions. All authors reviewed the manuscript and authorized the ultimate draft. All authors assert that they had complete access to all study data and take responsibility for submitting it for publication.

Funding

This work was supported by Beijing Anzhen Hospital, Capital Medical University, and Beijing Institute of Heart Lung and Blood Vessel Diseases. This work was funded by the Natural Science Foundation of Beijing Municipality (7202037).

Data availability

Data supporting the findings of this study from the UK Biobank team (http://www.ukbiobank.ac.uk/). The data and methods that support the findings of this study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The UK Biobank was approved by the North West Multicenter Research Ethics Committee, with all participants providing written informed consent. Ethical approval and informed consent were waived as the UK Biobank data is publicly available and does not include identifiable information.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.