Exploring the interaction between klotho and TyG index in cardiovascular risk stratification: A metabolic-inflammatory network analysis with mediation and machine learning insights

Abstract

Background

This study explores the relationship of metabolic-inflammatory network and cardiovascular disease (CVD), offering new insights into the roles of Klotho and the Triglyceride-Glucose (TyG) index in CVD pathogenesis.

Methods

Data from 5402 adults (mean age: 58.04 ± 10.83 years; 50.96 % female) from the NHANES in 2007–2016 database were analyzed. We proposed a prediction model for CVD risk incorporating Klotho protein, TyG index, and their interaction. The predictive value of these factors was evaluated using machine learning techniques, including random forest analysis and CHAID decision tree modeling.

Results

The study found no association between serum alpha-Klotho levels and CVD risk. However, the TyG index was demonstrated to be a significant predictor of CVD risk, particularly when lifestyle and socio-economic factors were not accounted for. TyG values were associated with an increased risk of metabolic syndrome and CVD (Model 1 OR: 1.234; Model 2 OR: 1.268). There was a significant interaction between Klotho-TyG was observed (coefficient − 2.608 × 10⁶). In addition, the random forest model achieved an accuracy of 66.63 % with high specificity and precision, and in the CHAID model an error of 27 %.

Conclusions

This study underscores the TyG index as a key biomarker for CVD risk, with the Klotho-TyG interaction improving risk stratification, and supporting early screening, treatment, and personalized interventions for more effective CVD management.

1. Introduction

Cardiovascular disease (CVD) continues to be one of the leading causes of morbidity and mortality globally [1,2], accounting for nearly 32 % of all deaths across the world. As per WHO report 17.9 million people die due to CVDs every year [2]. The global burden of diabetes-related disease is expected to increase significantly by 2030, especially in low- and middle-income nations, thus presenting considerable socioeconomic challenges for health systems across the world [3]. CVD has become more common with increased complexity of the disease process due to multifactorial interactions of various risk factors, including metabolic syndrome, dyslipidemia, chronic inflammation [4]. Elucidating the determinants of CVD pathogenesis—and the dynamic networks of interaction between them-will enable the design of effective preventive and therapeutic strategies.

Inflammation has an established role in the pathogenesis of atherosclerosis, myocardial infarction, and heart failure. Chronic low-grade inflammation plays a crucial role in cardiovascular aging, triggering endothelial dysfunction, wall stiffness, and the progression of atherosclerotic plaques. These dysregulated inflammatory responses can become exacerbated and contribute to pathological cascades, including plaque rupture and thrombosis [[5], [6], [7]]. This assumes an intricate network involving pro-inflammatory cytokines, leukocyte recruitment, and oxidative stress. In addition, inflammation is a key mediator of the interplay between major risk factors, namely, atherosclerosis, hypertension, and metabolic syndrome. Thus, chronic inflammation has emerged as a key contributor to cardiovascular pathology, acting as a critical link between metabolic alterations and vascular pathogenesis. Characterizing the interplay between these risk factors is critical for improving our understanding of metabolic dysregulation and mitigating the global burden of CVD.

This underpins the basis for novel therapeutic strategies. Recent studies have focused on the Klotho protein, which is mainly secreted by renal and neuronal tissues, as a potential novel therapeutic target. Klotho, a type of transmembrane protein known to reduce oxidative stress, inflammation, and endothelial dysfunction [[8], [9], [10]]. Klotho can also be released in a soluble form into the circulation [11]. Circulating Klotho exerts antioxidative, anti-inflammatory, and endothelial-protective effects, and has been associated with multiple aging-related conditions, underscoring its potential as both a biomarker and therapeutic target [[12], [13], [14]]. Klotho-targeted therapeutic agents have the potential to greatly improve cardiovascular outcomes [[15], [16], [17]]. Moreover, identifying Klotho as a prognostic marker would allow earlier and tailored approaches toward the aim of improving cardiovascular outcomes. Importantly, Klotho function correlates with mechanisms associated with aging, making its role of particular interest in the aging context. Given the high frequency of CVD in aging populations, Klotho may play a pertinent role in cardiovascular disease. Gene analyses have indicated that Klotho functions as a crucial regulator in cancers, suggesting that Klotho targeting may also directly or indirectly influence cardiovascular aging and other age-related diseases. This makes Klotho a promising therapeutic target to address the interconnected pathophysiology of these conditions [18].

Triglyceride-Glucose TyG is also a new marker of metabolic health and cardiovascular disease CVD risk. Previous meta-analysis has demonstrated positive correlation of TyG index with carotid artery intima–media thickness, peripheral artery high stiffness and incidence of cardiovascular events [[19], [20], [21], [22]]. Nonetheless, the TyG index became a more accessible marker for evaluation of patient population susceptible to CVDs, especially in patients with metabolic syndrome or diabetes.

While both Klotho deficiency and the TyG index have been widely studied in relation to cardiovascular health, the majority of previous publications have primarily evaluated their independent association with CVD, as opposed to the putative mechanisms linking these two factors to potential disease. Thus, the interrelationship of these markers and their joint contributions to overall CVD risk remain poorly understood. Klotho deficiency and elevated TyG index are associated with increased CVD risk, but the degree to which the mechanistic interaction between them contributes to overall risk is not clear.

The purpose of this study was to explore the mediating mechanisms that link metabolic health with cardiovascular disease, specifically, the relationship between serum Klotho levels, the TyG index and cardiovascular health, in a population-based large-scale study. Furthermore, a well-elucidated mechanistic understanding of these processes may also provide important clinical implications, specifically through the identification of early clinical screening tools, by revealing high-risk individuals and allowing diagnostic refinement for precise treatment and targeted intervention programs.

2. Methods

2.1. Study population and design: NHANES analysis

This study utilized data of individuals from 2007 to 2016 of NHANES. NHANES is a cross-sectional, nationally representative survey that assesses the health and nutritional status of the adult and child population in the United States. We included the participants from the NHANES in which serum alpha-Klotho had been measured, for the current study.

Eligibility of participants was determined by CVD data availability, alpha-Klotho measurements, and other covariates, including age, gender, race/ethnicity, education, body mass index (BMI), smoking status, alcohol status, and physical activity (Fig. 1).

Fig. 1.

Flowchart of study participants selection.

2.2. Serum alpha-Klotho protein (variable)

Serum alpha-Klotho concentrations were determined from frozen NHANES samples collected and tested during the period 2007–2016. The analysis was conducted using a commercially available ELISA kit produced by IBL International, Japan. All analyses, except for those on four fresh-frozen (pristine) samples from the Centers for Disease Control and Prevention, were conducted by The Northwest Lipid Metabolism and Diabetes Research Laboratories, Division of Metabolism, Endocrinology, and Nutrition, University of Washington.

All sample analyses were conducted in duplicate following the manufacturer's protocol and results were verified against the laboratory's standardized criteria for acceptability before reporting.

2.3. Triglyceride-glucose index (TyG) (metabolic indicator)

Blood specimens were collected using standard procedures and properly stored until further analysis. The TyG index was calculated as Ln[fasting triglycerides (mg/dL) × fasting glucose (mg/dL)/2] [23], as described previously, and used as a continuous variable in all analyses.

2.4. Cardiovascular diseases (disease)

Cardiovascular disease was assessed by asking participants, “Have you ever had any pain or discomfort in the chest?”. Participants who reported chest pain or discomfort of heart disease-related symptom yes were classified into the ‘CVDs-positive’; while those without symptoms were classified as ‘CVDs-negative’.

2.5. Con-founders and effect modifier

The following potential confounders were considered: age, gender, race/ethnicity, education, family poverty income ratio, body mass index (BMI), smoking status, alcohol status, physical activity, and sleep disturbance et al. These were self-reported through interviews with structure questionnaires.

2.6. Statistical analysis

Normality test of all continuous variables confirmed a normal distribution; thus, variables were presented as mean (standard deviation), and group differences were assessed using the t-test. Categorical variables were presented as percentages, with group differences assessed using the chi-square test. The relationship between Serum alpha-Klotho, TyG index, the interaction, and CVDs was analyzed using binary logistic regression models. In this study, the term “Serum alpha-Klotho levels × TyG” represents a multiplicative interaction term, constructed by mathematically multiplying serum alpha-Klotho concentrations with the TyG index to evaluate their joint effect. We used the bootstrap method for mediation analysis. Moreover, the role of metabolic indicators and gene interactions in risk prediction of CVDs was explored by introducing TyG index (triglyceride-glucose index) and Klotho-TyG interaction term. The machine learning study of random forest and cardinality automated interaction detection (CHAID) decision tree model was systematically used to investigate the performance and application value of metabolic index and gene interaction in risk stratification prediction of CVD [24].

A significant mediation effect was noted when the 95 % confidence interval (CI) did not contain zero. R 4.3.2 software for Windows was used for statistical analyses and figures. P values below 0.05 were considered significant with a two-tailed p-value.

3. Results

3.1. The general characteristics from NHANES participants

A total of 5402 adults were included from 2007 to 2016 (Table 1). Participants who completed laboratory tests, examinations, and questionnaires were included in the analysis. The sample comprised 49.04 % males and 50.96 % females; aged from 40 to 79, with an average age was 58.04 ± 10.83 years. The sample included 14.96 % Mexican American, 10.77 % other Hispanic, and 74.27 % other participants. Additionally, 64.70 % of participants were married and living with partner. Apart from age, gender, and education status (P>0.05), significant differences were observed in the distribution of participants across ethnicity, marital status, family poverty income ratio, BMI, and other behavioral indices by serum alpha-Klotho levels from 2007 to 2016 (P<0.05). Therefore, these data were comparable.

Table 1.

Baseline characteristics based on the CVDs subjects.

Characteristics	CVDs				χ2/t	P
	Overall (N = 5402)	Non-CVDS (N = 3943)	CVDs (N = 1459)	%
Age					2.673	0.102
40–59	2857	2112	745	26.08
60–79	2545	1831	714	28.06
Gender					0.078	0.781
Male	2649	1929	720	27.18
Female	2753	2014	739	26.84
Ethnicity					8.449	0.015*
Mexican American	808	615	193	23.89
Other Hispanic	582	441	141	24.23
Others	4012	2887	1125	28.04
Marital status					16.920	<0.001**
Married and living with partner	3495	2615	880	25.18
Never married	444	310	134	30.18
Others	1462	1017	445	30.44
Education status					2.885	0.236
Less than high school	1347	963	384	28.51
High school or GED	1189	862	327	27.50
College and above	2866	2118	748	26.10
Family poverty income ratio					34.892	<0.001**
<1.0	263	168	95	36.12
1.0–2.0	2646	1867	779	29.44
≧3.0	2493	1908	585	23.47
BMI					17.744	<0.001**
Normal	1291	980	311	24.09
Overweight	1890	1408	482	25.50
Obesity	2221	1555	666	29.99
Vigorous physical activity					13.365	<0.001**
Yes	950	648	302	31.79
No	4452	3295	1157	25.99
Moderate physical activity					27.006	<0.001**
Yes	1946	1339	607	31.19
No	3456	2604	852	24.65
Sleep disturbance					237.412	<0.001**
Yes	1646	970	676	41.07
No	3756	2973	783	20.85
Alcohol status					1.560	0.212
Yes	3864	2802	1062	27.48
No	1538	1141	397	25.81
Smoke status					37.470	<0.001**
Yes	2707	1876	831	30.70
No	2695	2067	628	23.30
TyG	5402	8.72 ± 0.67	8.79 ± 0.68		11.847	<0.001**
Serum alpha-Klotho levels	5402	858.37 ± 306.34	859.49 ± 366.05		0.013	0.910
Serum alpha-Klotho levels×TyG	5402	98,360.24 ± 57,765.66	102,384.89 ± 68,119.63		4.676	0.031*

Note: 1. Percentage = CVDs / Overall ∗ 100 %; 2. *P < 0.05; **P < 0.01.

The overall self-reported incidence of CVDs was 27.01 %. There were no significant differences in age, gender, and educational status (P > 0.05).

However, sleep disturbances were significantly more frequent in CVDs group (41.07 %) compared to the Non-CVDs (P < 0.001). Both vigorous and moderate physical activities were more common among participants with CVDs (31.79 % vs, 31.19 %) (P < 0.001). Smoking and alcohol drinking were another significant factors, with a higher proportion of smokers (30.70 %) and drinkers (27.48 %) in the CVDs group (P < 0.001).

Additionally, economic status, measured by the family poverty income ratio, was significantly related to the occurrence of CVDs. Participants with a poverty income ratio below 1.0 were more common in the CVDs group (36.12 %), however, the ratio ≧3.0 were the lowest in the CVDs group (23.47 %) (P < 0.001). Marital status was also significantly associated with CVDs, where individuals who were never married or had no partners were more prevalent in CVDs (30.18 % and 30.44 %, respectively) (P < 0.001). Ethnicity showed a significant difference, with a higher proportion of “other” Ethnicity in CVDs (28.04 %) compared to the Non-CVDs (23.89 % for Mexican American and 24.23 %% for other Hispanic).

Moreover, BMI also differed significantly between normal, overweight and obesity groups (P < 0.001), with a highest prevalence of obesity in CVDs (29.99 %). As weight increased, so did to the risk of CVDs.

Additionally, the serum alpha-Klotho levels ranged from 152.500 pg/mL to 5038.300 pg/mL. The mean serum alpha-Klotho level was 858.67 ± 323.52 pg/mL. The TyG index ranged from5.65 to 12.33. The mean TyG level was 8.74 ± 0.67. The serum alpha-Klotho and TyG interaction levels ranged from 16,470.00 to 891,779.10. The mean serum alpha-Klotho and TyG interaction level was 99,447.24 ± 60,756.10 pg/mL. The TyG index, a metabolic indicator, was significantly elevated in the CVDs (8.79 ± 0.68) compared to the Non-CVDs (8.72 ± 0.67) (P < 0.001). Moreover, the interaction between serum alpha-Klotho levels and TyG was significant (P < 0.05).

3.2. The mediate impact of serum alpha-Klotho and TyG on CVDs

As Table 2 shown, the mediating effect (ACME) result showed the TyG index was significant mediating effect, and it has a negative mediating effect. The pathway through which the independent variable affects the dependent variable through the mediating variable was negative.

Table 2.

The mediate impact of serum alpha-Klotho, TyG, and CVDs.

Name	Estimate	95 % CI		P
		Lower	Upper
ACME	−2.608 × 10−6	−5.774 × 10−6	−5.053 × 10−7	0.014
ADE	4.717	−3.891 × 10−5	4.230 × 10−5	0.808
Total effect	2.108	−4.261 × 10−5	4.034 × 10−5	0.874
Prop mediate	−1.237	−2.356	1.436	0.872

However, neither the direct effect (ADE) nor the total effect was significant, which may imply that the effect of Serum alpha-Klotho on the dependent variable CVDs was mainly realized through the mediator variable TyG index and the effect was negative in this analysis.

TyG index was found that a mediate impact of the Serum alpha-Klotho on cardiovascular outcomes, strategies to enhance Klotho expression or supplementation could represent a novel approach to reducing cardiovascular risk in individuals with metabolic syndrome or insulin resistance.

As Fig. 2 showed the OR of serum alpha-Klotho levels for CVD was 1.238 (95 % CI 0.797–1.924). There was a significant interaction effect between alpha-Klotho and TyG index with the estimated coefficient being −2.608 × 10⁶ (range from −5.774 × 10⁶ to −5.503 × 10⁷). The TyG index also demonstrated an independent OR of 1.234 (95 % CI: 0.998–1.525) for the risk of CVD.

Fig. 2.

The mediate impact of serum alpha-Klotho level and TyG on CVDs.

These results indicated that the “TyG” variable had a significant relationship with the CVD. TyG index acted as a key mediator, with elevated TyG levels being dose-dependently associated with increased risk of CVD. In addition, the interaction term (alpha-Klotho*TyG) had the stronger association with CVD than TyG index, measured as its coefficient (−2.608 × 10⁶), while the different “serum alpha-klotho” categories did not.

3.3. The association between serum alpha-Klotho levels, TyG and CVDs

Comparison of the serum alpha-Klotho levels, the TyG index and their interaction to the risk of CVDs versus non-CVDs was shown in Table 3. As evidenced in the results, serum alpha-Klotho didn't significantly associate with CVD across all models (OR 1.174 to 1.246, P > 0.05). In contrast, the association between TyG index and CVD risk was significant in Model 1 (OR: 1.234, P = 0.05) as well as in Model 2 (OR: 1.268, P < 0.05), but a loss of significance was found in Models 3 and 4 after more adjustments. In models 3 (OR: 1.489, P > 0.05) and 4 (OR: 1.499, P > 0.05). The interaction term between serum alpha-Klotho and TyG index reached borderline significance, supporting a postulate of potential synergistic effect of serum alpha-Klotho and TyG index on the risk of CVD.

Table 3.

Multiple logistic regression analysis on the association between serum alpha-Klotho, TyG, and CVDs.

		Category groups
		Non-CVDs	CVDs
Serum alpha-Klotho	Model 1	Reference	1.192 (0.854–1.664) 0.303
	Model 2	Reference	1.174 (0.840–1.641) 0.347
	Model 3	Reference	1.237 (0.881–1.737) 0.220
	Model 4	Reference	1.246 (0886–1.751) 0.206
TyG index	Model 1	Reference	1.234 (0.998–1.525) 0.052⁎⁎
	Model 2	Reference	1.268 (1.025–1.570) 0.029⁎⁎
	Model 3	Reference	1.184 (0.954–1.469) 0.126
	Model 4	Reference	1.120 (0.900–1.393) 0.309
Serum alpha-Klotho ∗ TyG	Model 1	Reference	1.475 (0.983–2.214) 0.061
	Model 2	Reference	1.390 (0.909–1.570) 0.129
	Model 3	Reference	1.489 (0.967–2.291) 0.071⁎
	Model 4	Reference	1.499 (0.974–2.309) 0.066⁎

Model 1: Unadjusted model.

Model 2: Adjusted ethnicity, marital status.

Model 3: Adjusted ethnicity, marital status, family poverty income ratio, vigorous work activity, moderate work activity, and smoke.

Model 4: Adjusted ethnicity, marital status, family poverty income ratio, vigorous work activity, moderate work activity, smoke, and BMI.

^⁎0.05 < P < 0.10.

^⁎⁎P < 0.05.

3.4. The machine learning study

By random forest evaluated, the accuracy of TyG index and Klotho-TyG interaction predicted CVD was 66.63 %. The specificity was 99.71 %. The precision was 83.33 %. MSE was 0.3338.The mean average precision was 0.8282. The AUC value was 0.673 (Fig. 3 shown).

Fig. 3.

The ROC curve of random forest analysis.

As Fig. 4 showed the CHAID decision tree model results showed that TyG index was the most important splitting variable, with higher values associated with higher risk of CVDs; the Klotho-TyG interaction term further significantly differentiated high-risk groups. The model error rate was 27 % with a standard error of 0.006, giving a robust prediction performance.

Fig. 4.

The CHAID decision tree model of machine learning study.

4. Discussion

Metabolic disorders and chronic inflammatory states have been well characterized as key factors in the initiation and advancement of CVD [25,26]. A web of interrelated physiological mechanisms closely associates metabolic disease and chronic inflammation. Obesity, for example, can lead to dyslipidemia, endothelial dysfunction, and increased oxidative stress, all of which promote CVD risk.

In this study, new mechanistic paradigms between metabolic dysfunction and CVD were described by linking the TyG index, a solid indicator of insulin resistance and metabolic alteration, to the serum alpha-Klotho signaling pathway. The findings demonstrated that serum alpha-Klotho acted as a mediator of TyG-induced cardiovascular risk in middle-aged and elderly populations, thereby advancing our understanding of the metabolic-inflammatory networks that uncovered their synergistic interactions as key regulatory nodes. These results not only refined current frameworks for cardiovascular risk stratification but also illuminated novel therapeutic targets for inflammation-driven cardiovascular disorders.

Notably, this study provides evidence for a strong association between the TyG index and CVD, suggesting that serum alpha-Klotho level may mediate cardiovascular outcomes. The risk effects of the TyG index for CVDs were mediated by serum alpha-Klotho. The most significant findings of the present study provide valuable insights into CVDs risk and the roles of serum alpha-Klotho and the TyG index on individual and interactive effects.

The TyG index is a well-established marker of insulin resistance and metabolic disorder, both of which are major risk determinant of CVD risk [27]. The significant association between the TyG index and CVDs observed in this study initial models (Model 1 and Model 2 without lifestyle and socioeconomic factors) aligned with existing research, which showed that higher TyG values correlated with an increased risk of metabolic syndrome, and CVDs (Model 1 OR: 1.234, 95 % CI 0.998–1.525; Model 2 OR: 1.268, 95 % CI 1.025–1.570). This implies that TyG may act as a metabolic driver of CVD pathogenesis [28]. Interestingly, after adjusting for additional variables such as physical activity, smoking status, and BMI in models 3 and 4, the significance of the TyG index's association with CVD risk weakened, suggesting that lifestyle factors might influence this relationship. This implies that the odds of the TyG index for CVD risk might be affected by lifestyle and metabolic markers, highlighting the significance of overall risk assessment in the clinical aspect of both the TyG index and CVD risk.

This study also found that serum alpha-Klotho was unable to independently predict CVD risk in any adjusted model (P > 0.05); however, it suggested that serum alpha-Klotho may play a mediating role in the TyG index's associated with CVD risk, indicating a more nuanced involvement. This finding was contrary to several previous studies postulated that the role of alpha-keto was protective against age-related diseases especially acted on the kidney function and vascular health by modulating oxidative stress and inflammation [29,30]. Nevertheless, the non-significance in this study might be related to the heterogeneity of the population sample, the complexity of CVD, or potential confounding factors including lifestyle and socioeconomic status that were controlled for in the models. In this study, therefore, it was confirmed that the disagreement may be due to the multiple environmental, lifestyle, genetic, and metabolic factors that can affect CVDs [[31], [32], [33]]. The other reason could also be the limitations of the serum alpha-Klotho measurement in covering the entire biological activity, which should be more accurate or may need functional assessments.

Our mediation analysis revealed that Klotho exerts protective effects by mitigating inflammation-driven metabolic disturbances, while TyG serves as a critical biomarker reflecting systemic metabolic health. The serum alpha-Klotho×TyG interaction had the greatest effect overall on CVD risk of the variables considered, possibly representing synergistic or antagonistic pathways connecting metabolic dysregulation and aging-related pathways (interaction effect: −2.608 × 10⁶, 95 % CI −5.774 × 10⁶–5.503 × 10⁷). This study highlights the relevance of considering individual and interactive influences of TyG and serum alpha-Klotho in the context of stratifying cardiovascular risk.

This study further analyzed the clinical value of the TyG index and alpha-Kloto*TyG interaction effects for predicting CVD using machine learning ensembles. According to the random forest analysis indicated the accuracy of TyG index and Klotho-TyG interaction predicted CVD was 66.63 %. The specificity was 99.71 %.The precision was 83.33 %. The AUC value was 0.673. These results implied the high specificity and precision means that in practice, the model is effective in reducing false positives, the recall is low, and the model performed well in negative class identification and can be used as a primary screening tool.

Moreover, the CHAID decision tree model error rate was 27 % with a standard error of 0.006, giving a robust prediction performance. This analysis further confirmed the TyG index can be used as a basic indicator for risk screening of CVDs, especially in high TyG populations, and the serum alpha-Klotho*TyG interaction effect further distinguished high-risk populations, suggesting that genes and metabolism-related variables should be integrated into risk stratification and precision medicine research in the future. Therefore, the CHAID model indicated that the TyG index combined with the Klotho-TyG interaction effect for comprehensive analysis has the advantage of easy interpretability and stratified analysis the risk of CVD, which may provide strong support for early prevention and personalized intervention for CVD.

Above all, theoretical implications of this study holds potential implications in terms of targeting serum alpha-Klotho and TyG, as they may play an essential role in interrupting metabolic-inflammation network [34], culminating in arrest or even reversal of CVDs. These biomarkers had a better predictive capacity for CVD risk assessment, confirmed by multiple machine learning models providing a strong model for early diagnosis and treatment. This doubled-target provided appropriate evidence to design new types of therapeutic drugs aimed at reversing both the metabolic component and the inflammatory component of cardiovascular pathology. These findings have translational potential, and future solid-phase studies are warranted and open therapeutic avenues to exploit the serum Klotho-TyG axis.

5. Limitations

The observations from this study must be interpreted with a few limitations. One such limitation is that the reliance on self-reported cardiovascular disease introduces a methodological limitation, as such assessments may be subject to recall bias and misclassification. And the heterogeneity of the study population may have influenced the findings, as variations in genetic, lifestyle, and environmental factors could impact the generalizability of the results. The sampling population did not fully reflect all demographics, limiting the overall usefulness of the data. Secondly, although the analysis took into account a wide range of potential confounding factors including physical activity, smoking and BMI, residual confounding cannot be ruled out. Not accounted for or inadequately measured exposures—like approaches to eating, hereditary tendencies or different metabolic diseases—might have influenced the noticed associations between the TyG index, alpha-Klotho and CVD risk. Thirdly, although serum alpha-Klotho is measured as a circulating marker, this indicator has its own intrinsic limitations; all current testing methods may not reflect their biological activity; it requires further functional and mechanistic studies to clarify their role in cardiovascular health. Last but not the least, it should be noted that, owing to the cross-sectional design of this study, causal relationships and directionality of associations between TyG, alpha-Klotho and CVDs could not be established. And although our analysis shows that the combined Klotho-TyG index exhibits superior predictive power for CVD compared with Klotho alone, an essential issue remains unresolved—namely, identifying the specific threshold at which this composite index can reliably stratify individuals at elevated risk of CVD. Longitudinal studies are needed to fill these gaps, to understand the temporal relationships involved in these findings, and to validate the findings using an additional independent database.

6. Conclusions

In this study, a solid methodological foundation is established that combines mediation analysis and machine learning to disentangle the complexities of metabolic- inflammatory interactions. Using machine learning allowed to move beyond traditional linear models and to uncover non-linear associations and interactions between TyG index, serum alpha-Klotho and CVD risk. As a result, this method improved predictive performance and also yielded high-risk sub-populations that may greatly value targeted interventions. These findings highlight the complex interplay between metabolic dysregulation and chronic inflammation as integral components of the CVD pathogenesis. A network analysis of comprehensive metabolic-inflammatory disorders identified serum alpha-Klotho protein and the TyG as the key nodes of this complex network. In summary, their interactive involvement between serum Klotho and TyG is not only related to metabolic and inflammatory pathways but also might be proposed as an early diagnosis and screening tool, and a possible therapeutic strategy for intervention and prevention of CVD.

Abbreviations

CVD

Cardiovascular disease

TyG

Triglyceride-glucose

CHAID

Cardinality automated interaction detection

OR

Odds ratio

ACME

Average causal mediation effect

ADE

Average direct effect

Clinical trial number

Not applicable.

CRediT authorship contribution statement

Jieqing Min: Data curation, Conceptualization. Yunjuan Yang: Writing – review & editing, Writing – original draft, Visualization, Software, Project administration, Methodology, Formal analysis, Data curation, Conceptualization.

Consent for publication

Not applicable.

Ethics approval and consent to participate

The Ethics Review Committee of the National Center for Health Statistics approved the NHANES study protocol, and all participants provided written informed consent before participation.

Funding

This work was supported in part by Yunnan Provincial Grant for Yunnan Provincial Grant for the Academic Leadership (grant No. 2024AC350014), 2024 Yunnan Provincial Expert Basic Research Workstation (grant NO. 2024-164), and by the 16th Batch of Kunming Grant for the Young Academic and Technical Leadership (grant NO. KMRCD-2018011), Research on the Association between Chronic Diseases and Common Diseases in Children and Health Behavior Intervention Strategies (YNAPM2025-006), the Grant for Kunming Healthcare Commission Technology Centre (2023-sw(Ji)-09), the Grant for Kunming Municipal Health Science and Technology Leading Talent Project(2024-sw-09), the Clinical study on artificial intelligence assisted screening of congenital heart disease in children based on electrocardiogram and ultrasound imaging data(2024YNLCYXZX0444), Xishan Major Grant for the Academic Leadership and a grant by Xishan District Bureau of Science and Technology (grant NO. 34 Xikezi), Yunnan International Joint Innovation Platform (Special Project for Establishing a Science and Technology Innovation Center for South and Southeast Asia, 202503AP140034).

Declaration of competing interest

The authors declare no competing interests.

Data availability

The NHANES data are available from https://www.cdc.gov/nchs/nhanes/.