Association between the triglyceride-glucose-frailty index and all-cause and cardiovascular mortality in individuals with cardiovascular-kidney-metabolic syndrome: the mediating role of the poverty-income ratio

Abstract

Background

cardiovascular-kidney-metabolic (CKM) syndrome involves complex interplay among metabolic, renal, and cardiovascular dysfunction. The triglyceride-glucose-frailty index (TyGFI) integrates insulin resistance (TyG) and functional decline (frailty), yet its association with mortality in CKM patients remains unexplored. Given that economic resources may influence mortality, whether the poverty–income ratio (PIR) mediates this association warrants investigation.

Methods

This retrospective cohort study drew from the NHANES database for the period from 2005–2018 and included 15,044 participants diagnosed with CKM syndrome. The TyGFI was calculated as the mathematical product of the TyG index [ln (triglycerides × fasting glucose/2)] and the frailty index. Mortality outcomes were ascertained by linking records to the National Death Index and tracking participants until December 31, 2019. Kaplan‒Meier survival curves, Cox regression analysis, restricted cubic spline (RCS), mediation analysis, and subgroup analyses were applied to explore the association between the TyGFI and mortality.

Results

During a median observation period of 82 months (IQR: 46–123 months), 1,103 deaths from all causes and 353 cardiovascular deaths occurred. Kaplan‒Meier curves revealed substantially reduced death rates among participants in the lowest TyGFI tertile. Relative to the lowest TyGFI tertile, the highest tertile exhibited more than threefold increased risks for total mortality (HR = 3.14; 95% CI 2.57–3.83; P < 0.001) and cardiovascular mortality (HR = 3.14; 95% CI 2.18–4.52; P < 0.001) following comprehensive adjustment. When analysed continuously, every 1-unit increase in TyGFI corresponded to a 56% increase in the risk for all-cause mortality and a 52% increase in cardiovascular mortality risk. Mediation analysis revealed that PIR partially mediate the TyGFI-mortality association, accounting for 7.66% of all-cause mortality and 8.48% of cardiovascular mortality relationships. Subgroup analysis revealed consistent associations across demographic strata, with a significant interaction for marital status.

Conclusions

TyGFI levels in individuals with CKM syndrome are nonlinearly associated with increased all-cause and cardiovascular mortality risk. Additionally, the PIR may partially mediate the TyGFI-mortality relationship, highlighting the possible role of socioeconomic factors. These findings support the use of the TyGFI as a valuable risk assessment tool for mortality evaluation in individuals with CKM syndrome.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-026-03076-5.

Introduction

Cardiovascular-Kidney-Metabolic (CKM) Syndrome represents a significant global public health challenge, affecting more than 500 million adults and contributing to more than 50% of adult mortality [1]. This syndrome encompasses a continuum of pathological conditions, including insulin resistance, adiposity, diabetes mellitus, chronic renal disease, and coronary heart disease [2]. The cardiovascular mortality risk increases progressively with advancing CKM stages, with patients in Stage 3 facing substantially higher mortality rates than those in earlier stages [3, 4]. These findings highlight the urgent need to identify modifiable metabolic factors associated with elevated mortality risk in CKM patients.

The triglyceride‒glucose (TyG) index, a measure of insulin resistance, has been extensively investigated for its association with cardiovascular outcomes [5, 6]. Extensive prospective cohort studies have consistently established that higher TyG index values correlate with increased cardiovascular and all-cause mortality risk among patients with metabolic syndrome [7–9]. Among 6383 NHANES 2005–2018 participants, those in the highest TyG index quartile had a 65% increase in cardiovascular mortality risk compared with the lowest quartile (HR = 1.65, 95% CI 1.32–2.07) [10]. Similarly, a UK Biobank study of 282,920 participants confirmed this association, with stronger effects observed in the CKM subgroup (HR = 1.38) [11]. Mechanistically, elevated TyG levels contribute to oxidative stress (e.g., increased malondialdehyde levels and reduced superoxide dismutase activity) and systemic inflammation (e.g., activation of the TLR4/NF-κB pathway), promoting endothelial dysfunction and the progression of atherosclerosis [12].

While existing studies have focused primarily on metabolic parameters, functional status measures, such as frailty, have received less attention despite their relevance in CKM populations. Frailty, a multidimensional condition characterized by diminished physiological capacity and increased susceptibility to adverse events, is prevalent among individuals with chronic conditions [13]. The Frailty Index (FI), derived from the cumulative deficit model, quantifies frailty as the proportion of health deficits—encompassing chronic diseases, functional limitations, and cognitive impairments—relative to a standardized set of variables, typically yielding a score between 0 and 1 [14, 15]. In middle-aged and elderly populations, higher frailty index levels are significantly associated with increased cardiovascular disease and stroke risk, with those in the highest frailty quartile demonstrating a 15-fold higher risk of cardiovascular events (OR = 15.09, 95% CI 9.65–23.60) and over 20-fold higher stroke risk (OR = 21.12, 95% CI 6.44–69.23) than their least frail counterparts [16]. The interaction between frailty and metabolic dysregulation, such as insulin resistance reflected by the TyG index, may exacerbate cardiovascular risk through shared mechanisms, including systemic inflammation and oxidative damage [17]. However, the combined utility of the TyG index and FI, potentially integrated as a composite triglyceride‒glucose‒frailty index (TyGFI), remains underexplored in CKM populations. While both the TyG index and FI have been independently associated with adverse outcomes, a composite index incorporating both metabolic and functional dimensions could provide a more comprehensive assessment of systemic health burden by capturing the co-occurrence of metabolic dysfunction and functional decline. However, its association with mortality in CKM patients requires further investigation.

Socioeconomic factors, such as the poverty–income ratio (PIR), further complicate the relationship between metabolic and functional factors and health outcomes in CKM patients. The PIR, a measure of socioeconomic disadvantage, influences access to health care, nutrition, and lifestyle factors, contributing to the progression of metabolic and functional decline [11, 18, 19]. Given that economic resources may influence mortality through multiple pathways, examining whether PIR mediates the TyGFI-mortality association may provide insights into the role of socioeconomic factors in CKM outcomes. Although prior studies have adjusted for traditional confounders using multistage Cox models [10], the mediating function of the PIR in the relationship between metabolic–functional measures and mortality remains poorly elucidated.

Therefore, this NHANES-based investigation examined the associations between the TyGFI and mortality outcomes (all-cause and cardiovascular) among individuals with CKM syndrome while evaluating the PIR as a potential mediator. These findings may provide deeper insights into the complex relationship between metabolic–frailty indices and socioeconomic factors in determining health outcomes among this vulnerable population.

Methods

Study population

This retrospective cohort study utilized data from the National Health and Nutrition Examination Survey (NHANES) from 2005–2018. The NHANES employs a cross-sectional survey design in which participants undergo baseline examinations during survey cycles. All exposure variables, including triglyceride levels, fasting glucose levels, and frailty index components for TyGFI calculation, were measured at baseline. The study population comprised 15,044 participants with cardiovascular-kidney-metabolic (CKM) syndrome, following the criteria outlined in the American Heart Association Presidential Advisory Statement. The exclusion criteria were as follows: (1) a lack of CKM syndrome diagnostic information, (2) incomplete triglyceride or glucose data and variables necessary for FI construction, and (3) inadequate mortality follow-up information. Participants were followed from their baseline examination date until death or December 31, 2019, through linkage to the National Death Index, establishing a clear temporal sequence between exposure measurement and outcome ascertainment.

The National Center for Health Statistics Institutional Review Board provided ethical approval for NHANES data collection. Written informed consent was obtained from each subject before participation [20]. Complete survey methodologies and documentation are available at https://www.cdc.gov/nchs/nhanes/.

Variable definitions

The TyGFI constituted the primary exposure. Baseline triglyceride and fasting glucose levels were assessed. The TyG index was calculated using the following formula: ln[triglycerides (mg/dL) × fasting glucose (mg/dL)/2] [21]. The frailty index employed methods established by Rockwood et al. [15], utilizing 49 health-related deficits spanning multiple physiological domains. The components included cognitive assessment (1 item), functional limitations (15 activities of daily living measures), depression screening (7 PHQ-9 components), comorbid conditions (self-reported chronic diseases, including arthritis, cardiovascular conditions, diabetes, and others), health care utilization patterns and general health assessment, physical performance and anthropometric evaluations (handgrip strength and BMI), and laboratory values (glycohemoglobin, haematologic parameters, and additional biomarkers). Each deficit was scored from 0 to 1 on the basis of severity. The frailty index represented the proportion of present deficits relative to the total number of assessed items (49), with continuous values between 0 and 1. The complete details of all 49 items are provided in Supplementary Table S1. The TyGFI was calculated as the mathematical product of the TyG index multiplied by the frailty index (TyGFI = TyG × FI), following the methodology established by Zhao et al.[16].

Mortality outcomes

The primary endpoints included all-cause and cardiovascular mortality. Mortality data were obtained from NHANES Public-use Linked Mortality databases through December 31, 2019. Death records underwent probabilistic matching with the National Center for Health Statistics and National Death Index repositories. Mortality classification utilized ICD-10 coding standards. Follow-up extended from baseline assessment until death or study termination [22]. All-cause mortality included fatalities from any cause, such as cardiac disease, malignancies, accidents, stroke, diabetes, and other conditions. Cardiovascular mortality comprised deaths due to cardiac disease and cerebrovascular events [23].

CKM syndrome staging classification (0–4)

On the basis of the American Heart Association clinical framework [3], CKM syndrome is stratified into five progressive stages (0–4). Stage 0 represents individuals without identifiable CKM-related risk determinants; Stage 1 encompasses patients manifesting early-stage metabolic perturbations, characterized by excess adiposity, excessive visceral fat accumulation, and/or impaired glycaemic homeostasis; Stage 2 represents individuals with established metabolic pathologies, specifically type 2 diabetes, hypertensive disorders, dyslipidaemia with elevated triglycerides, or concurrent chronic kidney dysfunction; Stage 3 indicates the emergence of preclinical cardiovascular structural and functional abnormalities, such as asymptomatic vascular atherosclerotic burden or left ventricular remodelling secondary to metabolic or nephropathic processes; and Stage 4 represents manifest cardiovascular clinical conditions, encompassing coronary disease, heart failure, cerebrovascular accidents, peripheral arterial disease, or cardiac rhythm disorders occurring within established CKM pathophysiology [2].

Covariates of interest

Subject information was retrieved from NHANES datasets, incorporating diverse demographic, behavioural, and clinical parameters. Sociodemographic factors included age, sex, ethnicity, marital status, education level, and poverty–income ratio. Lifestyle and health conditions included tobacco use, alcohol intake, hypertension, diabetes, and cerebrovascular disease history. Physical and biochemical measurements included body mass index (BMI), blood urea nitrogen (BUN), triglyceride (TG), glycosylated haemoglobin type A1c (HbA1c), fasting blood glucose (FBG), serum creatinine (Scr), and estimated glomerular filtration rate (eGFR). Variable selection was based on documented relationships with exposure and outcome measures, as established in prior epidemiological research [16].

Missing data management

Supplementary Table S1 depicts the missing value patterns for the study variables. Although most covariates showed minimal missing data (< 5%), random forest-based multiple imputation was applied to address incomplete information within the CKM cohort (excluding exposure and outcome measures) to preserve eligible participants and minimize selection bias.

Statistical analysis

The subjects were stratified into tertiles (T1–T3) according to their TyGFI values. The Shapiro‒Wilk test was used to assess the normality of continuous data. Continuous variables are reported as the mean ± standard deviation, with intergroup comparisons conducted via one-way ANOVA. Categorical data are shown as counts (n) and proportions (%), with group differences evaluated through chi-square (χ²) analysis.

Mortality events were systematically documented during the follow-up period. Kaplan‒Meier survival analysis with log-rank tests was conducted to compare event-free survival rates across tertiles of the combined TyGFI. To evaluate the association between the combined TyGFI and all-cause mortality as well as cardiovascular mortality, univariate and multivariate Cox regression analyses were conducted, and hazard ratios (HRs) with 95% confidence intervals (CIs) were reported. Three sequential adjustment models were developed: Model 1 remained unadjusted; Model 2 incorporated demographic variables (age, sex, and race); and Model 3 included marital status, education level, tobacco use, alcohol intake, BMI, hypertension, cerebrovascular disease history, estimated glomerular filtration rate, and poverty–income ratio in addition to Model 2 variables. Variance inflation factors (VIFs) were used to assess collinearity, with all values under 5, confirming the absence of substantial multicollinearity issues. Using Model 3, restricted cubic spline (RCS) analysis with 4 knots was used to examine potential nonlinear relationships between the TyGFI and mortality outcomes. Hazard ratios in the RCS plots were displayed on a logarithmic scale (log scale) to facilitate visualization of the wide range of HR values and proportional relationships. The proportional hazards assumption for Cox regression models was evaluated using Schoenfeld residual tests, with detailed results provided in the supplementary materials (Table S15 and Fig. S3).

To examine the stability of the relationships between the TyGFI and mortality endpoints, several sensitivity analyses were performed to confirm the primary results. Initially, stratified analyses were used to investigate variations across sex, ethnicity (Mexican American, non-Hispanic Black, non-Hispanic White, other Hispanic, and additional racial groups), partnership status (married/cohabiting vs. others), and CKM syndrome progression (nonadvanced vs. advanced stages). Given incomplete data for certain components for computing the triglyceride–glucose index and frailty index product, multiple imputation was subsequently applied to missing values, followed by replication of the main analysis.

To assess whether the association between the exposure (TyGFI) and outcome variables (all-cause mortality and cardiovascular disease [CVD] mortality) was mediated by the poverty–income ratio (PIR), mediation analyses were conducted using the “mediation” package in R version 4.2.2 [24]. The mediation analysis quantified the total effect (overall association between the TyGFI and mortality outcomes), direct effect (association independent of PIR), and indirect effect (portion of the TyGFI-mortality association attributable to the PIR as a mediator) [25]. All the models were adjusted for potential confounders, including age, sex, race, body mass index (BMI), marital status, PIR, education, smoking status, drinking status, estimated glomerular filtration rate (eGFR), hypertension status, and stroke status. Statistical inference was based on 1,000 bootstrap resamples to calculate 95% confidence intervals for the mediation effects [26, 27].

All analyses were conducted using R software (version 4.2.2, http://www.Rproject.org) and Free Statistics software version 2.0, with statistical significance set at P < 0.05 (two-tailed).

Results

Participant baseline characteristics

The study flowchart is shown in Fig. 1. This analysis included 15,044 participants, with a mean age of 50.62 years, 50.89% females, and 42.14% non-Hispanic whites. Subject baseline features are detailed in Table 1. Table S2 presents the distribution of variables containing missing data. Across TyGFI tertiles, the participants in the highest tertile (T3) tended to be older and predominantly female. Relative to the T3 cohort, which had lower TyGFI categories, the T3 cohort had markedly elevated BMI and glucose parameters (FBG and HbA1c) and higher rates of comorbid conditions such as hypertension, diabetes, and cerebrovascular disease. In addition, kidney function was compromised in the T3 group, as evidenced by reduced eGFRs and increased serum creatinine levels, as well as diminished socioeconomic indicators. Advanced CKM syndrome phases (stages 3–4) were more frequent in T3 (36.27%) than in T1 (3.31%). The mortality outcomes revealed a pronounced gradient, with all-cause mortality increasing from 3.15 in T1 to 21.10% in T3, while cardiovascular mortality increased from 0.88 to 6.98%, all of which were statistically significant (P < 0.001). Tables S3 and S4 present the unadjusted associations between baseline characteristics and all-cause mortality and cardiovascular mortality, respectively.

Fig. 1.

Flowchart of participant selection from NHANES 2005–2018

Table 1.

Baseline characteristics of individuals

Characteristics	Overall	Tertiles of TyGFI			P value
	(N = 15,044)	T1 (N = 5015)	T2 (N = 5014)	T3 (N = 5015)
Age, mean (SD), years	50.62 (17.67)	43.15 (16.27)	50.35 (17.41)	58.35 (15.90)	< 0.001
Sex, (%)					< 0.001
Male	7388 (49.11%)	2889 (57.61%)	2417 (48.21%)	2082 (41.52%)
Female	7656 (50.89%)	2126 (42.39%)	2597 (51.79%)	2933 (58.48%)
Race, (%)					< 0.001
Mexican American	2370 (15.75%)	894 (17.83%)	767 (15.30%)	709 (14.14%)
Non-Hispanic White	6339 (42.14%)	2077 (41.42%)	2088 (41.64%)	2174 (43.35%)
Non-Hispanic Black	3026 (20.11%)	743 (14.82%)	1076 (21.46%)	1207 (24.07%)
Other Hispanic	1524 (10.13%)	538 (10.73%)	486 (9.69%)	500 (9.97%)
Other races	1785 (11.87%)	763 (15.21%)	597 (11.91%)	425 (8.47%)
Marital Status, (%)					< 0.001
Married/living with a partner	9148 (60.82%)	3215 (64.11%)	3175 (63.34%)	2758 (55.03%)
All others	5892 (39.18%)	1800 (35.89%)	1838 (36.66%)	2254 (44.97%)
Education level, (%)					< 0.001
< High school	3809 (25.35%)	1006 (20.07%)	1169 (23.33%)	1634 (32.64%)
≥ high school	11,219(74.65%)	4006 (79.93%)	3841 (76.67%)	3372 (67.36%)
Smoke status, (%)					< 0.001
Never	8279 (55.09%)	3070 (61.28%)	2852 (56.96%)	2357 (47.03%)
Former	3786 (25.19%)	1040 (20.76%)	1247 (24.91%)	1499 (29.91%)
Current	2964 (19.72%)	900 (17.96%)	908 (18.13%)	1156 (23.06%)
Drink, (%)	8391 (70.92%)	2983 (75.71%)	2900 (72.25%)	2508 (64.69%)	< 0.001
BMI, mean (SD), kg/m2	29.30 (7.09)	27.44 (6.09)	28.99 (6.57)	31.52 (7.91)	< 0.001
PIR	2.50 (1.61)	2.75 (1.65)	2.66 (1.62)	2.08 (1.49)	< 0.001
Hypertension, (%)	5867 (39.06%)	624 (12.46%)	1956 (39.10%)	3287 (65.62%)	< 0.001
Diabetes, (%)	3415 (23.01%)	273 (5.46%)	927 (18.82%)	2215 (45.04%)	< 0.001
Stroke, (%)	607 (4.04%)	18 (0.36%)	87 (1.74%)	502 (10.04%)	< 0.001
BUN, mean (SD), mg/dL	13.73 (6.09)	12.63 (3.97)	13.10 (4.81)	15.45 (8.25)	< 0.001
Scr, mean (SD), mg/dL	0.90 (0.47)	0.85 (0.19)	0.86 (0.27)	0.98 (0.74)	< 0.001
eGFR, mean (SD), ml/min/1.73m2	94.65 (22.74)	102.25 (18.16)	96.00 (20.98)	85.68 (25.33)	< 0.001
HbA1c, mean (SD), %	5.81 (1.13)	5.40 (0.51)	5.71 (0.95)	6.31 (1.49)	< 0.001
FBG, mean (SD), mg/dL	110.54 (36.25)	100.21 (15.62)	107.14 (29.37)	124.28 (50.29)	< 0.001
TG, mean (SD), mg/dL	129.65 (113.04)	112.90 (91.73)	126.43 (98.11)	149.63 (140.02)	< 0.001
TyG	8.65 (0.68)	8.45 (0.62)	8.62 (0.64)	8.89 (0.71)	< 0.001
FI	0.15 (0.10)	0.06 (0.02)	0.13 (0.02)	0.27 (0.09)	< 0.001
TyGFI	1.34 (0.91)	0.53 (0.19)	1.13 (0.19)	2.37 (0.80)	< 0.001
All-cause mortality, (%)	1610 (10.70%)	158 (3.15%)	394 (7.86%)	1058 (21.10%)	< 0.001
CVD mortality (%)	510 (3.39%)	44 (0.88%)	116 (2.31%)	350 (6.98%)	< 0.001
CKM stages, (%)					< 0.001
0	1278 (8.50%)	821 (16.37%)	370 (7.38%)	87 (1.73%)
1	2523 (16.77%)	1392 (27.76%)	857 (17.09%)	274 (5.46%)
2	8636 (57.40%)	2636 (52.56%)	3165 (63.12%)	2835 (56.53%)
3	853 (5.67%)	108 (2.15%)	306 (6.10%)	439 (8.75%)
4	1754 (11.66%)	58 (1.16%)	316 (6.30%)	1380 (27.52%)

Continuous variables are presented as means (SD) and categorical variables are presented as n (%). N, unweighted number of subjects; T, tertiles; SD, standard deviations; BMI, body mass index; PIR, poverty income ratio; BUN, blood urea nitrogen; Scr, serum creatinine; eGFR, estimated glomerular filtration rate; HbA1c, glycosylated hemoglobin type A1c; FBG, fasting blood glucose; TG, triglyceride; TyG, triglyceride–glucose index; FI, frailty index; CVD, cardiovascular disease; CKM, cardiovascular-kidney-metabolic

The association between the TyGFI and mortality in patients with CKM

During a median observation period of 82 months (IQR: 46–123 months), 1,103 total deaths were recorded, with 353 attributed to cardiovascular aetiology. Kaplan‒Meier survival curves revealed significant disparities in mortality outcomes for both endpoints across TyGFI tertile groups (Fig. 2, P < 0.0001 for each outcome). Patients in the lowest TyGFI tertile (T1) exhibited the most favourable survival, whereas those in the highest tertile (T3) showed markedly increased death rates.

Fig. 2.

Kaplan–Meier survival analysis stratified by TyGFI tertiles in adults with CKM syndrome. A All-cause mortality and B cardiovascular mortality outcomes. TyGFI, triglyceride glucose-frailty index; CKM, cardiovascular-kidney-metabolic syndrome

Table 2 presents the relationships between TyGFI levels and mortality endpoints in individuals with CKM. Following comprehensive confounder adjustment, each 1-unit increase in the TyGFI corresponded to a 56% increase in all-cause death risk (HR = 1.56; 95% CI 1.47–1.65; P < 0.001) and a 52% increase in cardiovascular death risk (HR = 1.52; 95% CI 1.37–1.68; P < 0.001). Similarly, each 1-SD increase in TyGFI was associated with a 49% increase in all-cause mortality (HR = 1.49; 95% CI 1.42–1.58; P < 0.001) and a 46% increase in cardiovascular mortality (HR = 1.46; 95% CI 1.33–1.60; P < 0.001). According to the results of the categorical analysis, compared with the reference tertile participants, middle tertile participants exhibited increased all-cause death risk (HR = 1.62; 95% CI 1.32–1.99; P < 0.001) and cardiovascular death risk (HR = 1.58; 95% CI 1.09–2.30; P = 0.017); the highest tertile subjects demonstrated an over threefold increase in both all-cause (HR = 3.14; 95% CI 2.57–3.83; P < 0.001) and cardiovascular mortality (HR = 3.14; 95% CI 2.18–4.52; P < 0.001). Trend P values were significant for both endpoints (P < 0.001), with trend HRs showing an 82% increase in all-cause risk (HR = 1.82, 95% CI 1.66–2.00) and an 84% increase in cardiovascular risk (HR = 1.84, 95% CI 1.56–2.17) per tertile increase. The findings remained consistent across the adjustment models, suggesting that the TyGFI may serve as a valuable tool for risk assessment in individuals with CKM syndrome.

Table 2.

Cox regression model for TyGFI and mortality in adults with CKM syndrome

Exposure	Model 1 HR (95% CI)		P value	Model 2 HR (95% CI)	P value	Model 3 HR (95% CI)	P value
All-cause mortality
TyGFI (per 1-unit)	1.99 (1.92–2.06)		< 0.001	1.65 (1.59–1.73)	< 0.001	1.56 (1.47–1.65)	< 0.001
TyGFI (per 1-SD)	1.87 (1.8–1.93)		< 0.001	1.58 (1.52–1.64)	< 0.001	1.49 (1.42–1.58)	< 0.001
TyGFI tertiles
T1 (0.08–0.82)	Reference			Reference		Reference
T2 (0.82–1.49)	2.53 (2.1–3.04)		< 0.001	1.56 (1.3–1.88)	< 0.001	1.62 (1.32–1.99)	< 0.001
T3 (1.49–5.98)	7.76 (6.56–9.17)		< 0.001	3.67 (3.09–4.35)	< 0.001	3.14 (2.57–3.83)	< 0.001
P for trend	2.87 (2.67–3.1)		< 0.001	2.05 (1.9–2.22)	< 0.001	1.82 (1.66–2.00)	< 0.001
CVD mortality
TyGFI (per 1-unit)	2.06 (1.93–2.2)		< 0.001	1.7 (1.58–1.83)	< 0.001	1.52 (1.37–1.68)	< 0.001
TyGFI (per 1-SD)	1.93 (1.82–2.05)		< 0.001	1.62 (1.51–1.73)	< 0.001	1.46 (1.33–1.60)	< 0.001
TyGFI tertiles
T1 (0.08–0.82)	Reference			Reference		Reference
T2 (0.82–1.49)	2.68 (1.89–3.79)		< 0.001	1.55 (1.1–2.2)	0.013	1.58 (1.09–2.30)	0.017
T3 (1.49–5.98)	9.25 (6.76–12.66)		< 0.001	3.97 (2.89–5.47)	< 0.001	3.14 (2.18–4.52)	< 0.001
P for trend	3.18 (2.77–3.65)		< 0.001	2.19 (1.89–2.52)	< 0.001	1.84 (1.56–2.17)	< 0.001

Model 1: no covariates were adjusted, Model 2: Adjusted for age, sex and race, Model 3: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoke, drink, eGFR, hypertension, stroke

CI, confidence interval; HR, hazard ratio; T, tertiles; CVD, cardiovascular disease; TyGFI, TyGFrailty index; BMI, body mass index; PIR, poverty income ratio; eGFR, estimated glomerular filtration rate; CVD, cardiovascular disease; CKM, cardiovascular-kidney-metabolic

Four-knot restricted cubic spline modelling was used to evaluate the dose‒response patterns between the TyGFI and death outcomes (Fig. 3). The TyGFI exhibited significant nonlinear associations with both mortality endpoints. Threshold effect analysis was performed using two-piecewise linear regression models to delineate potential inflection points of the TyGFI in relation to mortality outcomes. Significant threshold effects were identified for both all-cause mortality and cardiovascular mortality (P < 0.001 for both outcomes). Detailed results, including hazard ratios and confidence intervals for each segment defined by the inflection points, are provided in Supplementary Tables S13 and S14.

Fig. 3.

Restricted cubic splines depicting the dose–response relationships between TyGFI and mortality outcomes across CKM stages. Adjustment factors included age, sex, race, BMI, marital status, PIR, education level, smoking status, alcohol consumption, eGFR, hypertension, and stroke history. A Association between TyGFI and all-cause mortality in CKM stages 0–4; B association between TyGFI and all-cause mortality in non-advanced CKM stages; C association between TyGFI and all-cause mortality in advanced CKM stages; D association between TyGFI and cardiovascular mortality in CKM stages 0–4; E association between TyGFI and cardiovascular mortality in non-advanced CKM stages; F association between TyGFI and cardiovascular mortality in advanced CKM stages. TyGFI, triglyceride glucose–frailty index; CKM, cardiovascular-kidney-metabolic; HR, hazard ratio; BMI, body mass index; PIR, poverty income ratio; eGFR, estimated glomerular filtration rate

To comprehensively evaluate potential effect modification, we conducted stratified analyses across multiple clinically relevant variables, including demographic factors, lifestyle behaviours, anthropometric measures, and clinical conditions. The primary TyGFI-mortality relationships across key demographic and clinical strata are shown in Fig. 4, with complete subgroup analyses presented in Supplementary Tables S9 and S10. The associations between the TyGFI and mortality outcomes persisted across all the subgroups, with hazard ratios consistently above unity, which demonstrated consistent associations. Significant interactions (P for interaction < 0.05) were observed for marital status (P = 0.044), age (P < 0.001), BMI (P < 0.001), the PIR (P = 0.001), education level (P = 0.008), and stroke history (P = 0.001) for all-cause mortality and for age, BMI, the PIR, diabetes mellitus, and FBG for cardiovascular mortality (detailed results in Supplementary Tables S9 and S10. The remaining variables showed no significant interactions (P > 0.05), indicating consistent TyGFI effects across these strata.

Fig. 4.

Subgroup analysis of the association between TyGFI and mortality in adults with CKM syndrome stratified by baseline characteristics. Adjusted for all covariates except for this subgroup of variables. A Association between TyGFI and all-cause mortality; B association between TyGFI and cardiovascular mortality. TyGFI, triglyceride glucose–frailty index; CKM, cardiovascular-kidney-metabolic; HR, hazard ratio; CI, confidence interval

Sensitivity analyses

To assess result stability, sensitivity analysis was conducted using multiple imputation techniques for missing value handling, and the primary analytical procedures were subsequently repeated on the complete imputed dataset (Table S5). The sensitivity analysis results corroborated the primary findings, thereby confirming the stability and validity of our conclusions.

PIR as a mediator in TyGFI-mortality associations among CKM populations

Mediation analysis was conducted to evaluate whether the PIR acts as an intermediary variable in the association between the TyGFI and mortality outcomes among individuals with CKM syndrome. As depicted in Fig. 5, the PIR demonstrated significant mediating effects on the TyGFI-mortality relationship within this population. The analysis revealed that the PIR mediated 7.66% of the total effect linking the TyGFI to all-cause mortality and 8.48% of the total effect linking the TyGFI to cardiovascular mortality.

Fig. 5.

Mediating role of PIR in the relationship between TyGFI and mortality in individuals with CKM syndrome. A PIR partially mediates the association between TyGFI and all-cause mortality; B PIR partially mediates the association between TyGFI and cardiovascular mortality

Discussion

Using a comprehensive cohort of CKM syndrome adults from the NHANES (2005–2018), this study revealed that the TyGFI is nonlinearly associated with increased all-cause and cardiovascular mortality risk. Furthermore, the poverty income ratio may function as a mediator in this relationship. These findings contribute valuable insights for clinical risk assessment and may facilitate enhanced risk stratification strategies in the management of individuals with CKM syndrome.

While this investigation represents the direct examination of TyGFI-mortality relationships, the two fundamental elements comprising the TyGFI—TyG index and frailty index—have individually demonstrated significant mortality associations in prior studies, establishing theoretical support for our combined measure. Accumulating evidence has shown that elevated TyG index values are significantly associated with a increased risk of cardiometabolic diseases and mortality across heterogeneous study populations [28, 29]. Zhao et al.’s [16] study examined the use of the TyGFI, which demonstrated remarkably strong associations with cardiovascular outcomes: the highest TyGFI quartile had ORs of 15.09 for CVD and 21.12 for stroke in CHARLS and ORs of 4.98 for CVD and 12.98 for stroke in NHANES. Building upon this foundational work, our study extends the application of the TyGFI to a clinically distinct population—individuals with CKM syndrome—and examines mortality outcomes rather than disease incidence. Additionally, while both studies calculate the TyGFI as TyG × FI, our use of Hakeem et al.’s [15] 49-item FI (including depressive symptoms and laboratory biomarkers) differs from Zhao et al.’s 40-item FI [14]. Our observed hazard ratios for the highest TyGFI tertile (all-cause mortality HR: 3.14; cardiovascular mortality HR: 3.14) in CKM individuals are consistent with the findings of Zhao et al., although the effect magnitude differs, which may reflect differences in outcome measures (mortality vs. incidence), population characteristics (CKM syndrome vs. general population), and FI construction methodology. Importantly, our study identifies the PIR as a partial mediator in the TyGFI-mortality association (mediating 7.66% and 8.48% of the effects on all-cause and cardiovascular mortality, respectively), revealing socioeconomic pathways not explored in their work. This mediating role of the PIR is further corroborated by Yogeswaran et al.’s [19] study of 2,464 cancer survivors from NHANES 2003–2014, which demonstrated significant associations between low PIR and mortality in U.S. cancer survivors, providing additional empirical support for the importance of socioeconomic factors in mortality risk stratification.

To systematically evaluate the clinical utility of the composite TyGFI index, we assessed the independent associations of its components—TyG and FI—with all-cause and cardiovascular mortality (Supplementary Tables S6–S7; Figs. S1–S2), along with formal interaction analyses (Supplementary Tables S8–S9). The results demonstrated a submultiplicative joint effect, with the FI as the dominant contributor to mortality risk.

This prominent role of functional decline is corroborated by Burton et al.’s [30] systematic review and meta-analysis of 31 studies involving 18,868 acute stroke patients, which reported a pooled frailty prevalence of 24.6% (95% CI 16.2–33.1%), and in 7 studies suitable for meta-analysis, the association between frailty and mortality yielded an odds ratio of 3.71 (95% CI 2.41–5.70), demonstrating a comparable effect magnitude to the observed TyGFI associations. Despite the established importance of frailty alone, the continuous TyGFI measure offers several distinct advantages. First, it provides a single, parsimonious metric that integrates information from both metabolic and functional domains, thereby avoiding the need to evaluate multiple indices separately. Second, the multiplicative formulation (TyG × FI) generates a continuous product whose value increases with increasing combined metabolic–functional burden, assigning higher values to individuals with elevations in both domains. Third, although TyG alone did not remain independently associated after full adjustment, its inclusion in the composite index may help identify subgroups in which metabolic dysfunction and functional decline coexist—a phenotype that potentially reflects multidimensional deterioration in CKM syndrome. Fourth, the TyGFI demonstrated a robust monotonic dose–response relationship across its tertiles (Table 2 and Supplementary Table S5), supporting straightforward risk stratification. Taken together, our findings suggest that the TyGFI provides a comprehensive assessment of two pathophysiologically related yet independently contributing domains—metabolic dysfunction and functional decline—that together capture the systemic burden characteristics of CKM syndrome.

The strong association observed between the TyGFI and mortality highlights the value of integrating metabolic and functional assessments among individuals with CKM syndrome. Our analyses suggest several key explanations for these findings. First, the additional value of the TyGFI arises from capturing both functional deficits and the metabolic environment, as frailty status affects the impact of metabolic dysfunction. Notably, among individuals with low FI, elevated TyG was not significantly associated with mortality, whereas high FI was consistently linked to increased risk regardless of TyG status, underscoring the importance of functional decline in mortality risk (Supplementary Tables S8–S9). Furthermore, compared with AIP in a comparable CKM population (HR: 1.19–1.38) [31], the TyGFI demonstrated substantially stronger associations (HR: 3.14), representing a 2–threefold difference in effect magnitude. This difference highlights that the integration of functional vulnerability assessment by the TyGFI captures dimensions of mortality risk that are absent from purely metabolic indices. Threshold effect analysis revealed significant inflection points at TyGFI values of 2.29 (95% CI 2.23–2.35) for all-cause mortality and 2.19 (95% CI 2.12–2.25) for cardiovascular mortality (Supplementary Tables S13 and S14). Below these thresholds, the TyGFI was strongly associated with mortality. These results highlight the need for intensified monitoring and early intervention in CKM syndrome patients with elevated TyGFI values, offering cut-off values for risk stratification and personalized management. Third, bidirectional pathophysiological relationships support the integration of metabolic and functional assessments. Insulin resistance (reflected by the TyG index) promotes chronic inflammation, oxidative stress, and mitochondrial dysfunction through activation of the NF-κB and JAK/STAT pathways [32–35], accelerating sarcopenia and functional decline [36]. Conversely, frailty-associated sarcopenia, reduced physical activity, and malnutrition worsen insulin resistance [32, 37]. Vascular endothelial dysfunction constitutes a convergent mechanism linking both conditions [37, 38]. Collectively, these findings underscore that functional decline is a pivotal contributor to mortality risk in patients with CKM syndrome. Moreover, the incorporation of both metabolic and functional parameters, as reflected in the TyGFI, enhances the robustness and clinical applicability of risk assessment in this high-risk cohort.

The subgroup analyses revealed several clinically important effect modifications, with marital status being particularly noteworthy. The significant interaction effect of marital status on all-cause mortality (P = 0.044) revealed that the TyGFI-mortality association was stronger in married/cohabiting individuals (HR = 1.66) than in those without partners (HR = 1.46). Several mechanisms may explain these findings. First, while married individuals typically benefit from spousal support and better health monitoring [39, 40], severe metabolic–functional impairment (high TyGFI values) may represent a critical vulnerability threshold where the physiological burden overwhelms these protective factors[41]. Second, the combined emergence of frailty and insulin resistance in married individuals could indicate accelerated physiological decline that spousal connections do not entirely shield against [42]. Third, relationship dynamics, including caregiving burden and stress associated with a partner’s health decline, may amplify stress responses in individuals with elevated TyGFI values, potentially exacerbating mortality risk [43, 44].

Beyond marital status, our comprehensive subgroup analysis revealed additional significant interactions with age, BMI, the PIR, education level, and comorbidity profiles (Supplementary Tables S10–S11). These interaction patterns suggest that TyGFI-based risk stratification may benefit from personalization according to patient sociodemographic and clinical profiles. Clinicians should consider enhanced monitoring for socially isolated individuals and integrated assessments incorporating social support systems alongside traditional metabolic parameters.

Several key points regarding the clinical utility of the TyGFI merit discussion. Although the 49-item frailty index offers comprehensive assessments across multiple physiological domains, its complexity may hinder its routine clinical implementation, especially in time-constrained primary care settings. However, several factors support its potential value: (1) Many components (laboratory values, medication counts, and comorbidity data) are routinely available in electronic health records, enabling automated calculation; (2) the comprehensive nature of the TyGFI is particularly suitable for detailed risk assessment in specialized CKM clinics or geriatric units rather than as a rapid screening tool; and (3) in research and population health surveillance, the TyGFI provides valuable insights into metabolic-functional vulnerability and its association with disease outcomes. Additionally, user-friendly digital tools or smartphone applications could facilitate the calculation of the TyGFI in clinical settings. Despite these practical considerations, our findings underscore the importance of integrating functional assessment with metabolic parameters in CKM syndrome management, regardless of the specific tool used.

Several limitations should be acknowledged. First, although we conducted Fine-Gray competing risk analysis to account for noncardiovascular deaths (Supplementary Table S12), the attenuated effect sizes in the competing risk model compared with traditional Cox regression (T3 vs. T1: SHR = 2.18 vs. HR = 3.14) underscore the importance of considering competing events when interpreting cardiovascular mortality associations, particularly in populations with substantial comorbidity burden. Nevertheless, the preservation of statistical significance and the pronounced mortality gradient in the competing risk framework support the robustness of our findings regarding the association between the TyGFI and cardiovascular mortality. Second, our exclusion of participants with incomplete clinical data may limit generalizability to individuals with irregular health care follow-up or insufficient assessments. Third, this study utilized NHANES data representing the U.S. population, which may restrict extrapolation to other ethnic groups with different genetic backgrounds and health care systems. Fourth, the observational design precludes the establishment of causality, limiting our results to associational relationships between the TyGFI and mortality endpoints. Fifth, despite adjusting for measured confounders, residual confounding from unmeasured variables remains a concern. Sixth, regarding the mediation analysis, PIR showed a statistically significant but modest mediation proportion (7.66–8.48%). This modest proportion indicates that while PIR captures the socioeconomic component of the pathway, the majority of the TyGFI effect operates through other mechanisms beyond economic status. This study specifically examined the socioeconomic pathway represented by PIR. Comprehensive examination of the mediators simultaneously would require different analytical frameworks and substantially larger sample sizes. Future research is needed to further elucidate the relative contributions of different pathways linking metabolic-functional burden to mortality. Seventh, some frailty components were assessed by self-report, which may be subject to measurement errors due to recall bias or misclassification, potentially affecting the accuracy of our findings. Eighth, the complexity of the TyGFI may limit its clinical applicability. The 49-item frailty index requires comprehensive data collection from multiple sources, which may be challenging in time-constrained clinical settings. Future research should explore simplified versions or electronic health record integration strategies to enhance practical feasibility. Ninth, consistent with NHANES survey design, only baseline data were collected without longitudinal assessment of medication use or clinical management during follow-up. Secular trends in metabolic disease treatment may have influenced outcomes, potentially attenuating the observed associations toward null. Nevertheless, we still observed strong associations (HR = 3.14 for highest vs. lowest tertile).

Conclusion

In individuals with CKM syndrome, the TyGFI is nonlinearly associated with increased all-cause and cardiovascular mortality risk. Additionally, mediation analysis identified the PIR as a potential partial mediator in the TyGFI–mortality association, highlighting the possible role of socioeconomic factors in this relationship. This study supported the clinical value of the TyGFI as a novel mortality evaluation tool among individuals with CKM syndrome.

Abbreviations

CKM

cardiovascular-kidney-metabolic

TyG

Triglyceride–glucose index

FI

Frailty index

HR

Hazard ratio

PIR

Poverty income ratio

CI

Confidence interval

DM

Diabetes mellitus

RCS

Restricted cubic spline

ANOVA

One-way analysis of variance

IQR

Interquartile range

BUN

Blood urea nitrogen

Scr

Serum creatinine

eGFR

Estimated glomerular filtration rate

HbA1c

Glycosylated hemoglobin type A1c

FBG

Fasting blood glucose

CVD

Cardiovascular disease

SD

Standard deviation

TG

Triglycerides

TNF-α

Tumor necrosis factor-alpha

IL-6

Interleukin-6

CRP

C-reactive protein

VIFs

Variance inflation factors

Author contributions

H.F.Z., Y.W. and N.W. conceived and designed the study. Y.W. and M.L. performed data acquisition, analysis, and interpretation. Y.W., N.W., and M.L. drafted the manuscript. J.W.L. conducted the statistical analysis. H.F.Z. and H.B.L. provided administrative, technical, and material support. H.F.Z. supervised the study.

Funding

This work was supported by grant 82471294 (to Dr. Hong-Fei Zhang) from the National Natural Science Foundation of China and grant 2024A1515012256 (to Dr. Hong-Fei Zhang) from the Natural Science Foundation of Guangdong Province, China.

Data availability

Data supporting the findings of this study are publicly available from the National Health and Nutrition Examination Survey (NHANES) official website (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx).

Declarations

Ethics approval and consent to participate

The National Health and Nutrition Examination Survey protocols were approved by the National Center for Health Statistics Ethics Review Board. Written informed consent was obtained from all participants prior to participation.

Competing interests

The authors declare no competing interests.