Association of dietary quality, biological aging, progression and mortality of cardiovascular-kidney-metabolic syndrome: insights from mediation and machine learning approaches

Abstract

Background

To investigate the association between the Dietary Inflammatory Index (DII), biological aging, and the staging and mortality of cardiovascular-kidney-metabolic (CKM) syndrome.

Methods

Data of 7,918 participants were derived from the National Health and Nutrition Examination Survey 2005–2018. Cross-sectional analyses using multivariable logistic regression were conducted to evaluate the relationship between DII and CKM staging. Cox proportional hazards models were employed to assess the impact of DII on mortality in CKM patients. Mediation analyses were performed to determine whether biological aging mediated DII-staging and DII-mortality association. Machine learning models were developed to classify CKM stages 3/4 and predict all-cause mortality, with SHapley Additive exPlanations (SHAP) used to interpret the contribution of DII components.

Results

Over a median follow-up of 9.3 years, 819 deaths were recorded. Higher DII were associated with an increased risk of advanced CKM stages [OR (95% CI): tertile 2, 1.39 (1.17, 1.65); tertile 3, 1.85 (1.56, 2.20)], and all-cause mortality [(HR (95% CI): tertile 2, 1.20 (1.01–1.43); tertile 3: 1.45 (1.21–1.73)]. The optimal risk stratification threshold for DII to predict all-cause mortality was 1.93. Mediation analyses revealed that biological aging accounted for 23% (95% CI: 18-28%) of the effect of DII on advanced CKM stages and 13% (95% CI: 8-22%) of the effect of DII on all-cause mortality. Furthermore, the Light Gradient Boosting Machine model showed strong performance in predicting advanced CKM staging (AUC: 0.896, 95% CI: 0.882–0.911), while Logistic regression performed better in predicting all-cause mortality (AUC: 0.857, 95% CI: 0.831–0.884). SHAP analysis revealed that intake of magnesium and n-3 fatty acid were associated with reduced risk of both advanced CKM stages and all-cause mortality.

Conclusion

DII, a marker of pro-inflammatory dietary patterns, was significantly linked to CKM syndrome progression and mortality, partly by influencing biological aging. This underscores the importance of diet quality in managing CKM staging and mortality risk.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12937-025-01175-9.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12937-025-01175-9.

Introduction

Cardiovascular-kidney-metabolic (CKM) syndrome, an age-associated disease, is characterized by a complex interplay of cardiovascular, renal, and metabolic dysfunctions [1]. Despite some research on traditional risk factors like systemic inflammation and oxidative stress [1, 2]the role of modifiable factors in CKM syndrome progression remains underexplored. The American Heart Association emphasizes the importance of dietary interventions but lacks specific recommendations for the different stages of CKM syndrome [1, 3]. CKM syndrome progresses through dynamic stages with distinct pathophysiology and management priorities [1, 4, 5]indicating that dietary impact may vary significantly across stages. However, the stage-specific effects of dietary patterns on CKM progression remain understudied. Investigating these links help inform personalized dietary strategies to improve health outcomes throughout the CKM spectrum.

Inflammation and oxidative imbalance are key mechanisms in cardiovascular-kidney-metabolic health [1, 6–8]. Previous studies reveal that various nutrients—both anti-inflammatory and pro-inflammatory—are linked to chronic systemic inflammation and oxidative homeostasis [9]. The Dietary Inflammatory Index (DII) reflects dietary inflammation based on anti-inflammatory and pro-inflammatory nutrient intake [10]. High DII levels have been reported to correlate with increased inflammation and chronic disease [11]. However, its relationship with the CKM syndrome remains unknown, highlighting the need to explore the effects of DII on cardiorenal metabolic health [12].

Although dietary patterns are a modifiable risk factor for cardiometabolic health, the underlying pathways are not fully understood [13]. Biological aging, marked by progressive cellular and molecular dysfunction, is increasingly recognized as a mediator linking diet and lifestyle to chronic disease pathophysiology [14, 15]. Mechanistic studies show that diet can affect aging processes, including mitochondrial dysfunction, epigenetic changes, and inflammation [16, 17]. These factors can disrupt metabolic balance and impair multi-organ crosstalk [18]. Clarifying these mechanisms could unveil novel targets for mitigating CKM risk through precision nutrition strategies tailored to attenuate aging-related pathways.

In this study, we aim to: (1) examine the associations of DII with the staging and mortality of CKM syndrome; (2) investigate biological aging as a potential mediator between DII and the staging of CKM and mortality outcome; (3) explore the associations of DII components with CKM staging and mortality outcomes using machine learning and SHapley Additive exPlanations (SHAP) analysis.

Methods

Study population

Data were derived from the US National Health and Nutrition Examination Survey (NHANES), a nationally representative study assessing the health and nutritional status of the U.S. population. The NHANES protocol was approved by the Institutional Review Board (IRB) of the National Center for Health Statistics (NCHS), and informed consent was obtained from all participants. This study was conducted in accordance with the ethical principles laid out in the Declaration of Helsinki.

In this study, participants from NHANES 2005–2018 were included, excluding those under 20 years old, pregnant individuals, and those with missing data on CKM components, DII calculations, or biological aging biomarkers (Fig.S1).

Assessment of DII

NHANES collects dietary data via 24-hour recall interviews at Mobile Examination Centers. The DII, developed by Shivappa, includes 45 parameters and can be calculated with fewer than 30 nutrients [10]. Due to NHANES limitations, 27 components were used in this study, which preserved the stability of the DII’s accuracy [19]. Table S1 provided details on the dietary components of DII.

Assessment of biological aging

Biological age and phenotypic age were calculated using the R package “BioAge,” based on eight biomarkers, while phenotypic age was derived from nine variables [20].

Assessment of CKM syndrome

CKM syndrome was defined as having subclinical or clinical cardiovascular disease (CVD), chronic kidney disease (CKD), and metabolic disorders. Subclinical CVD was determined as a 10-year CVD risk of ≥ 20% or high-risk CKD [5]. Clinical CVD included a self-reported history of heart failure, coronary heart disease, heart attack, or stroke. CKD risk was stratified based on estimated glomerular filtration rate (eGFR) and urinary albumin/creatinine ratio based on the Kidney Disease: Improving Global Outcomes classification [21]. Metabolic disorders include overweight/obesity, abdominal obesity, prediabetes, diabetes, hypertension, dyslipidemia, and metabolic syndrome.

The 10-year CVD risk was assessed using the simplified CKM risk algorithm of Predicting Risk of Cardiovascular Disease Events (PREVENT), which considers factors like age, total cholesterol (TC), high-density lipoprotein cholesterol (HDL), systolic blood pressure (SBP) and eGFR [22]. The eGFR was calculated based on the 2021 CKD-EPI creatinine equation [23]. The detailed criteria were in Table S2-3.

Participants were stratified into two distinct groups based on the severity of their CKM syndrome: those with non-advanced CKM stages, encompassing stages 0 through 2, and those with advanced CKM stages, including stages 3 and 4.

Outcome

The outcomes included all-cause mortality, CVD mortality, and non-CVD mortality, which were sourced from the Centers for Disease Control and Prevention website and updated until December 31, 2019. The causes of mortality were identified using the tenth revision of the International Statistical Classification of Diseases and Related Health Problems.

Covariates

Data were collected through standardized questionnaires, physical exams, and laboratory analyses of blood and urine. Covariates included age, gender, race (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, or other), housing status (own, rent, or other), marital status (married, unmarried, or divorced/separated/widowed), smoking status (current smoker or not), alcohol consumption (ever consumed or not), citizenship (yes or no), employment status (employed or unemployed), and health insurance (private, government, or none).

Statistical analysis

All analyses were conducted using the R software (version 4.3.2). Continuous variables were assessed for normality via Shapiro-Wilk tests; parametric (ANOVA with Tukey post hoc) or nonparametric (Kruskal-Wallis with Dunn’s test) methods were applied accordingly. Categorical variables were compared using Chi-square test.

First, logistic regression was employed to evaluate the associations of DII with CKM staging. Models were adjusted for age, sex, ethnicity, marital status, citizenship, employment status, health insurance, and smoking status. Alcohol intake was not included as a covariate since it was inherently integrated into the DII calculation. Clinical indicators related to metabolic dysfunction, kidney function, or comorbidities were excluded from adjustments to avoid overadjustment, as these variables constitute the diagnostic criteria for CKM syndrome. Second, Kaplan-Meier (KM) survival curves were constructed to visualize survival differences across DII groups in CKM patients, with statistical significance assessed using the log-rank test. multivariable Cox proportional hazards regression was employed to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for mortality outcomes, adjusting for the aforementioned covariates. The proportional hazards assumption was verified using Schoenfeld residuals. Fourth, restricted cubic spline (RCS) analyses with four knots were conducted to explore potential nonlinear relationships between DII, biological aging, and both CKM staging and mortality outcomes. Fifth, sensitivity analyses were performed to ensure the robustness of the findings. These analyses restricted participants to those with variables (age, TC, HDL, SBP and eGFR) within ranges specified by the PREVENT equations, ensuring comparability across study populations.

Additinally, optimal risk stratification thresholds were determined using the “surv_cutpoint” function, which applies maximally selected rank statistics to identify the optimal DII thresholds for mortality risk stratification. Moreover, mediation analysis was performed using the R package “mediation” to assess the proportion of the dietary effect on mortality outcomes mediated by biological aging markers. Bootstrapping with 1,000 resamples was employed to estimate 95% CIs for indirect effects. Furthermore, the dataset was randomly split into a training set (70%) and a testing set (30%). Machine learning models were developed using eight algorithms—Logistic Regression (LR), Light Gradient Boosting Machine (LightGBM), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), eXtreme Gradient Boosting (XGBoost), Naive Bayes (NB), Decision Tree (DT), and Neural Network (NNET)—to predict advanced CKM stages (3–4) and all-cause mortality outcome. These models integrated dietary components derived from the DII along with demographic features (age, sex, race) to enhance predictive performance (Table S13). Hyperparameter tuning was conducted using 10-fold cross-validation combined with a randomized grid search in the training set (Table S16). Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1 score, as calculated in the validation set. SHAP values were calculated to quantify feature importance, identifying key dietary contributors to the outcomes. To enhance robustness, we calculated ensemble SHAP values for the top 5 models using an AUC-weighted approach [24]. Individual SHAP values were computed via the “fastshap” package. Model weights were determined by their relative AUC contributions. The final ensemble SHAP values were derived from the weighted sum of these individual SHAP values, prioritizing high-performing models while preserving overall feature attribution. The results of SHAP analysis were visualized using the “shapviz” package.

Results

Baseline characteristics

A total of 7,918 participants were included, of whom 3,853 (48.7%) were male, and 1,839 (23.2%) were classified as having advanced CKM stages. The median age was 50.6 ± 17.4 years, the median phenotypic age was 47.8 ± 18.5 years, and the median biological age was 45.3 ± 19.9 years. Participants with higher DII scores exhibited greater phenotypic age (tertile 1: 47.6 ± 17.5 years; tertile 2: 47.2 ± 18.5 years; tertile 3: 48.7 ± 19.3 years) and biological age (tertile 1: 43.3 ± 18.8 years; tertile 2: 44.9 ± 19.4 years; tertile 3: 47.7 ± 21.0 years). They also had a higher prevalence of advanced CKM stages, with distributions of 528 (20.2%) in tertile 1, 612 (22.7%) in tertile 2, and 699 (26.8%) in tertile 3 (Table 1). Additionally, participants with advanced CKM stages or those who died exhibited higher DII scores and biological aging biomarkers (all P < 0.001) (Table S4-S5).

Table 1.

Baseline characteristics of participants stratified by DII level

Variables	Total (n = 7,918)	DII tertile 1 (n = 2,613)	DII tertile 2 (n = 2,692)	DII tertile 3 (n = 2,613)	P value
Demographics
Age, years	50.6 (17.4)	51.0 (16.7)	49.9 (17.6)	50.9 (18.0)	0.046
Male, n (%)	3853 (48.7%)	1576 (60.3%)	1325 (49.2%)	952 (36.4%)	< 0.001
Ethnicity, n (%)					< 0.001
Mexican American	1303 (16.5%)	445 (17.0%)	457 (17.0%)	401 (15.3%)
Other Hispanic	764 (9.65%)	238 (9.11%)	237 (8.80%)	289 (11.1%)
Non-Hispanic White	3580 (45.2%)	1282 (49.1%)	1185 (44.0%)	1113 (42.6%)
Non-Hispanic Black	1582 (20.0%)	369 (14.1%)	575 (21.4%)	638 (24.4%)
Others	689 (8.70%)	279 (10.7%)	238 (8.84%)	172 (6.58%)
Current Smoker, n (%)	1619 (20.4%)	391 (15.0%)	531 (19.7%)	697 (26.7%)	< 0.001
Drinkers, n (%)	2466 (31.1%)	749 (28.7%)	816 (30.3%)	901 (34.5%)	< 0.001
Height, cm	167 (10.2)	170 (10.1)	167 (10.1)	165 (9.71)	< 0.001
Weight, kg	82.4 (20.9)	82.4 (20.0)	82.9 (21.0)	81.8 (21.5)	0.16
BMI, kg/m2	29.3 (6.65)	28.5 (6.11)	29.4 (6.62)	29.9 (7.10)	< 0.001
SRP measures
Married, n (%)	4239 (53.5%)	1561 (59.7%)	1454 (54.0%)	1224 (46.8%)	< 0.001
Employed, n (%)	4547 (57.4%)	1666 (63.8%)	1585 (58.9%)	1296 (49.6%)	< 0.001
Own a home, n (%)	5090 (64.3%)	1791 (68.5%)	1713 (63.6%)	1586 (60.7%)	< 0.001
Private insurance, n (%)	6375 (80.5%)	2195 (84.0%)	2155 (80.1%)	2025 (77.5%)	< 0.001
Citizenship, n (%)	6889 (87.0%)	2229 (85.3%)	2338 (86.8%)	2322 (88.9%)	< 0.001
Medical history
Overweight/ obesity, n (%)	5828 (73.6%)	1849 (70.8%)	2008 (74.6%)	1971 (75.4%)	< 0.001
Abdominal obesity, n (%)	2182 (27.6%)	686 (26.3%)	721 (26.8%)	775 (29.7%)	0.01
Hypertension, n (%)	4537 (57.3%)	1487 (56.9%)	1511 (56.1%)	1539 (58.9%)	0.11
Prediabetes, n (%)	4247 (53.6%)	1404 (53.7%)	1392 (51.7%)	1451 (55.5%)	< 0.001
Diabete mellitus, n (%)	1437 (18.1%)	410 (15.7%)	489 (18.2%)	538 (20.6%)	< 0.001
Metabolic syndrome, n (%)	3797 (48.0%)	1194 (45.7%)	1272 (47.3%)	1331 (50.9%)	< 0.001
Advanced CKM, n (%)	1839 (23.2%)	528 (20.2%)	612 (22.7%)	699 (26.8%)	< 0.001
Kidney function
eGFR, mL/min/1.73 m²	94.2 (21.3)	95.0 (19.9)	94.9 (21.0)	92.9 (22.8)	< 0.001
Urine ACR, mg/g	35.1 (215)	30.5 (176)	32.2 (237)	42.7 (227)	0.08
Biological aging
Phenotypic age	47.8 (18.5)	47.6 (17.5)	47.2 (18.5)	48.7 (19.3)	0.01
Biological age	45.3 (19.9)	43.3 (18.8)	44.9 (19.4)	47.7 (21.0)	0.02

Continuous variables were presented as mean (standard deviation). Categorical variables were expressed as number (percentage). Abbreviations: DII, dietary inflammation index; eGFR: estimated glomerular filtration rate; ACR: albumin-to-creatinine ratio.

Relationship between DII and CKM stages

As shown in Fig. 1, the DII exhibited a positive correlation with advanced CKM stages. Higher DII scores were linked to an increased risk of advanced CKM stages, with odds ratios (ORs) and 95% CIs of 1.39 (1.17, 1.65) for tertile 2 and 1.85 (1.56, 2.20) for tertile 3. In sensitivity analyses, after limiting the analysis to participants whose age, values of TC, HDL, SBP and eGFR ranged complied with those specified in the PREVENT equations, the associations of DII with advanced CKM stage remained consistent (Table S6-10). Additionally, a linear dose-response relationship was observed between DII and advanced CKM stages. As DII scores increased, the likelihood of advanced CKM stages increased (P for non-linearity = 0.41, Fig.S2).

Fig. 1.

Associations of dietary quality with the staging and mortality of CKM syndrome. Abbreviations: DII, dietary inflammation index; CKM, cardiovascular-kidney-metabolic syndrome

Association between DII and mortality in CKM patients

The KM survival curves (Fig. 2) revealed that for all-cause mortality outcomes, the tertile 1 of DII group had the highest probability of survival, whereas the tertile 3 of DII group had the lowest probability of survival, and all were statistically significant by Log-rank test (P < 0.001). Cox hazard model showed that higher tertiles of DII were significantly associated with an increased risk of all-cause (tertile 2: HR 1.20, 95% CI 1.01–1.43, tertile 3: HR 1.45, 95% CI 1.21–1.73), cardiovascular (tertile 2: HR 1.06, 95% CI 0.73–1.54, tertile 3: HR 1.72, 95% CI 1.21–2.45) and non-cardiovascular mortality (tertile 2: HR 1.24, 95% CI 1.01–1.51, tertile 3: HR 1.37, 95% CI 1.11–1.68). Morever, a linear dose-response relationship was observed between DII and all-cause mortality outcome (P for non-linearity = 0.35, Fig.S3).

Fig. 2.

Kaplan-Meier survival curve for mortality by DII tertiles. Abbreviations: DII, dietary inflammation index

Associations of biological age with DII, CKM progression and mortality

Higher DII levels were associated with increased biological age [β ± standard deviation (SD): tertile 2, 2.16 ± 0.12; tertile 3, 3.98 ± 0.41] and phenotypic age (β ± SD: tertile 2, 0.82 ± 0.12; tertile 3, 1.52 ± 0.12). Significant correlations were observed between accelerated aging and advanced CKM stage, as well as with all-cause mortality, as indicated by measurements of both biological and phenotypic age, with all P < 0.001 (Table S11–S12).

As shown in Fig. S2-S3, linear relationships were observed between phenotypic age and advanced CKM stages (P for non-linearity = 0.52), as well as between biological age and all-cause mortality (P for non-linearity = 0.58). In contrast, non-linear associations were found between biological age and advanced CKM stages (P for non-linearity < 0.001), as well as between phenotypic age and all-cause mortality (P for non-linearity = 0.002).

Mediated effects of biological age on dietary quality and mortality

Mediation analyses revealed that biological age significantly mediated 23% (95% CI: 18%, 28%) of the effect of DII on advanced CKM stages and 13% (95% CI: 8%, 22%) of the effect on all-cause mortality. Similarly, phenotypic age mediated the correlation between DII and advanced CKM stages, contributing 36% (95% CI: 27%, 46%), and accounted for 23% (95% CI: 18%, 34%) of the association between DII and all-cause mortality (Fig. 3).

Fig. 3.

Mediation effects of biological aging on the associations of DII with cardiovascular-kidney-metabolic syndrome. Abbreviations: DII, dietary inflammation index; CKM, cardiovascular-kidney-metabolic

Optimal risk stratification cut-off points for DII on mortality

For all-cause mortality outcomes, the optimal risk stratification cut-off point for DII was 1.93. The KM survival curves revealed that all-cause mortality risk was notably elevated in patients with DII ≥ 1.93 compared to those with DII < 1.93 (Fig.S4).

Machine learning and SHAP analysis

As shown in Tables S14-15, there were no significant differences between the training and validation sets, indicating their comparability. For advanced CKM staging, ROC curves (Fig. 4A) was evaluated for all ML models in the validation set. LightGBM was considered to be the best model as it had the highest AUC: 0.896 (95%CI: 0.882 to 0.911), followed by LR and XGBoost model, with AUC of 0.895 (0.880, 0.910) and 0.895 (0.880–0.910). For all-cause mortality in CKM (Fig. 4B), LR, LightGBM and XGBoost models showed superior performance compared to other models, with AUCs of 0.857 (95%CI: 0.831 to 0.884), 0.848 (0.821, 0.875) and 0.843 (0.816–0.869). Additionally, the LightGBM, LR, and XGBoost models demonstrated superior accuracy and F1 scores relative to other models (Table S17-18).

Fig. 4.

Receiver operating characteristic curve of ML models. Abbreviations: ML, machine learning; XGBoost, eXtreme Gradient Boosting; NNET, Neural Network; SVM, Support Vector Machine; KNN, K-Nearest Neighbors; NB, Naive Bayes; LR, Logistic Regression; DT, Decision Tree; LightGBM, Light Gradient Boosting Machine

We calculated and ranked the weighted ensemble SHAP values for each DII component within the validation set. For predicting advanced CKM stages (Fig. 5A), the SHAP values revealed that total fat (0.0258), saturated fatty acid (SFA) (0.0163), magnesium (0.0148), n-6 fatty acid (0.0147), monounsaturated fatty acid (MUFA) (0.0112), polyunsaturated fatty acid (PUFA) (0.0102), alcohol (0.0078), protein (0.0044), folic acid (0.0041), β-carotene (0.0040), zinc (0.0039), n-3 fatty acid (0.0039) and caffeine (0.0033) were the primary contributors. As depicted in Fig. S5A, total fat intake exhibited a positive association with the risk of CKM progression. Conversely, anti-inflammatory components such as magnesium, folic acid, β-carotene and n-3 fatty acid demonstrated a negative correlation with the likelihood of advanced CKM staging. Notably, components like SFA were linked to a reduced progression risk, while PUFA showed an increased risk. These relationships should be interpreted cautiously due to potential non-causal confounding.

Fig. 5.

SHAP values of key predictors for advanced CKM staging and all-cause mortality. Abbreviations: CKM, cardiovascular-kidney-metabolic; SHAP, SHapley Additive exPlanations

For predicting all-cause mortality (Fig. 5B), the SHAP values identified PUFA (0.0495), n-6 fatty acid (0.0490), total fat (0.0442), MUFA (0.0297), SFA (0.0129), magnesium (0.0102), n-3 fatty acid (0.0082), vitamin B2 (0.0063), zinc (0.0058), thiamine (0.0055), selenium (0.0044), vitamin A (0.0043) and dietary fiber (0.0035) as the key contributors. As depicted in Fig. S5B, a positive link was observed between SFA intake and the risk of all-cause mortality. In contrast, anti-inflammatory nutrients like magnesium, selenium, vitamin A, n-3 fatty acid and dietary fiber were negatively associated with all-cause mortality risk. However, vitamin B2, thiamine, MUFA, and zinc, were positively related to higher mortality risk; while specific fat intakes (total fat, n-6 fatty acid) negatively associated with lower mortality risk. This indicated that their effects might be influenced by underlying health conditions.

Discussion

This study investigated the associations between the DII, CKM staging, and mortality outcomes among patients with CKM, while evaluating the potential mediating role of biological aging in the pathways linking DII to mortality outcomes. RCS analysis revealed a significant positive linear association between DII scores and mortality outcomes. Patients in lower DII tertiles exhibited more favorable baseline characteristics and lower risks of all-cause, cardiovascular, and non-cardiovascular mortality. Higher DII scores were significantly associated with an increased likelihood of progression to advanced CKM staging. The optimal DII cut-off value for mortality stratification was determined to be 1.93. Mediation analysis demonstrated that biological aging significantly mediated the relationships between DII scores and both CKM staging and mortality outcomes. Additionally, LightGBM, LR and XGBoost model exhibited excellent performance in predicting advanced CKM staging and all-cause mortality. SHAP analysis was used to interpret the models, identifying the most influential predictors. For advanced CKM staging, elevated total fat intake was positively correlated with advanced CKM stage risk, whereas higher consumption of anti-inflammatory nutrients like magnesium, folic acid, β-carotene and n-3 fatty acid were associated with a lower risk. For all-cause mortality, increased intake of saturated fatty acid was linked to higher all-cause mortality risk, while anti-inflammatory nutrients such as magnesium, selenium, vitamin A, n-3 fatty acid and dietary fiber were associated with a lower risk. These findings underscore the potential importance of maintaining an anti-inflammatory dietary pattern and lifestyle in mitigating CKM progression and mortality risk.

The association between DII and the CKM progression and mortality risk can be interpreted through multifaceted biological mechanisms [11, 25]. First, DII focuses on the inflammatory-oxidative axis, a central aspect of CKM’s pathophysiology. Systemic inflammation can accelerate multi-organ damage, making it crucial to understand how DII influence this process [1, 6, 10]. Second, there’s growing evidence that dietary interventions can break the cycle of worsening conditions across different systems [13, 26, 27]. For example, pro-inflammatory diets might make renal-mediated hypertension worse, which in turn increases the workload on the heart and impairs blood sugar control [13, 26, 27]. Third, pro-inflammatory diets also promote metabolic dysregulation via insulin resistance and adipose tissue inflammation [28, 29]. Pro-inflammatory diets drive endothelial dysfunction, oxidative damage, and chronic inflammation, worsening cardiac remodeling, renal hypoxia, and insulin resistance [28, 30, 31]. This interaction forms a pathophysiological network where dietary patterns affect cardiovascular, renal, and metabolic systems simultaneously.

Our findings further underscored the clinical relevance of the DII in mortality risk stratification for patients with CKM. Specifically, we identified an optimal cut-off value for DII, which could distinguish high-risk CKM populations and help design targeted intervention strategies. Previous researches have demonstrated that various dietary patterns can play a significant role in the management of chronic disease through mitigating oxidative stress and inflammation [31, 32]. For instance, Shivappa et al. found that individuals in the highest DII category showed a 36% increased risk of CVD incidence and mortality (relative risk = 1.36, 95% CI: 1.19, 1.57) [33]. Also, Chen et al. revealed that patients in the highest DII exposure category had a 28% higher risk of CKD than those in the lowest DII exposure category (relative risk = 1.28, 95%CI: 1.14, 1.44) [33].

Previous study have revealed that high DII diets promote aging-related inflammatory changes, leading to damage in blood vessels and kidney tissue [34–36]. Similarly, our results demonstrated that biological aging mediate the effects of DII on CKM progression and mortality. The interplay between biological aging and dietary inflammatory potential may elucidate the mechanistic pathways linking anti-inflammatory/pro-inflammatory diets to cardiorenal metabolic syndrome severity and mortality. Furthermore, whole body aging is primarily driven by cell senescence [37]and anti-senescence therapies have been reported to be promise for age-related diseases like idiopathic pulmonary fibrosis and diabetic kidney disease [38, 39]. These findings highlight anti-aging nutrition as a key strategy to break the cycle between DII and CKM syndrome.

This study represented the first integration of DII components with machine learning to predict CKM progression and mortality, demonstrating the utility of LightGBM, LR and XGBoost models in synthesizing dietary and clinical data for enhanced prognostic accuracy. Their strong performance can be explained by several factors [40]. Firstly, LR has a stable structure and prevents overfitting through regularization, ensuring good stability and generalization across different datasets. Secondly, LightGBM and XGBoost employ gradient boosting decision tree technology, which enhances generalization across diverse electronic medical datasets. Unlike KNN models, which are sensitive to noise and outliers and prone to overfitting in high-dimensional data, LightGBM and XGBoost effectively capture complex nonlinear relationships in electronic medical data, offering higher prediction accuracy. In contrast, NB models oversimplify assumptions of feature independence, limiting their predictive accuracy [40]. Due to these advantages, LightGBM and XGBoost have been extensively applied in various electronic health records (EHRs) for constructing disease prediction models [40, 41]. For instance, Qi et al. demonstrated that LightGBM outperformed other models in predicting CVD-cancer comorbidity, achieving an AUC of 0.951⁴⁰. Similarly, XGBoost has proved useful in predicting cardiac adverse events in patients receiving immune checkpoint inhibitors [42].

Our research indicated that anti-inflammatory nutrients, including magnesium, folic acid, β-carotene, selenium, vitamin A and dietary fiber were linked to a lower risk of advanced CKM staging or all-cause mortality. This is in line with previous studies on these nutrients’ associations to health results [43–46]. The health benefits of these nutrients likely come from their strong anti-inflammatory properties, which alleviate oxidative stress and neutralize reactive oxygen species (ROS) [36, 47]. Morever, magnesium is vital for genomic stability and cellular function. It acts as a key cofactor for enzymes that stabilize DNA structure, protecting against ROS-induced cellular damage, thus enhancing overall health [48]. Additionally, dietary fiber intake helps metabolic function, regulates immune balance, and repairs intestinal barrier damage lead exposure, thus improving overall health [46, 49].

This study also revealed distinct associations among various fatty acid components and advanced CKM staging and all-cause mortality. For instance, total fat was positively linked to advanced CKM staging yet negatively to all-cause mortality; SFA showed negative associations with advanced CKM staging but positive associations with all-cause mortality; conversely, polyunsaturated fatty acid were positively correlated with both outcomes. These results reflected the inconsistencies observed in existing researches [50–53]. A meta-analysis of 29 prospective large cohorts (n = 1,164,029) found inverse associations between total fat, MUFA, PUFA consumption and all-cause mortality [51]. However, another study found no significant trends in all-cause mortality across tertiles of SFA, MUFA, or PUFA intake, but highlighted the importance of the proportional composition of dietary fatty acid within total fat intake for long-term prognosis [52]. Given these inconsistent findings, further research is required to clarify the effects of different types of fat intake on health outcomes.

Morever, our study revealed that protein and alcohol intake were negatively associated with advanced CKM staging risk. Similarly, prior studies have linked protein and alcohol consumption to reduced risks of accelerated aging [54]. This inverse relationship might stem from the confounding from moderate alcohol consumption, which is often associated with healthier lifestyles and higher socioeconomic status among moderate drinkers compared to abstainers or heavy drinkers [55]. Additionally, adequate protein intake may lower mortality risk by preserving muscle mass and function, preventing sarcopenia and enhancing metabolic health [56]. Furthermore, our study found positive associations between thiamine, vitamin B2, and zinc intake and the prognosis of CKM patients. Previous research has demonstrated a U-shaped relationship between dietary thiamine intake and new-onset hypertension, suggesting that thiamine intake above the inflection point may increase mortality risk [57]. Additionally, the observed increased risk may be due to either the unhealthy components that often accompany vitamin B2-rich foods or its potential to disrupt mitochondrial function and energy metabolism [58]. Regarding zinc, its association with health outcomes remains inconclusive. For instance, Lee et al. found higher zinc intake to be beneficial in reducing CVD mortality [59]whereas Liu et al. observed no significant correlation between zinc intake and thyroid dysfunction [60].

The integration of dietary components from the DII into machine learning models represents a novel approach, advancing clinical applicability by prioritizing modifiable factors for targeted interventions. Future research should focus on elucidating dose-response dynamics and synergies between nutrients to optimize personalized dietary strategies in CKM management, thereby bridging the gap between predictive modeling and actionable clinical recommendations. Additionally, exploring biological aging as a mediator offers novel insights into how diet influences CKM pathogenesis, providing a framework for future aging-related interventions. Our study also has several limitations. First, its cross-sectional nature precludes confirming causality between dietary quality, biological aging, and CKM syndrome. Second, plant antioxidants were excluded from DII calculations due to NHANES data constraints, potentially introducing bias. Third, DII assessments were based on 24-hour dietary recall interviews, which cannot adequately reflect long-term dietary habits. Moreover, single 24-hour recalls are subject to recall bias and cannot accurately capture an individual’s usual dietary intake or daily variations in consumption. Fourth, unaccounted confounding factors and measurement inaccuracies may bias results, necessitating further validation through additional cohort and experimental studies.

Conclusion

In conclusion, this study establishes a significant positive linear association between the DII and mortality outcomes in patients with CKM disease, with lower DII tertiles linked to reduced mortality risks. Higher DII scores are associated with advanced CKM progression, and biological aging significantly mediates the link between DII and outcomes. LightGBM, LR, and XGBoost models effectively predict advanced CKM staging and mortality. These findings highlight the importance of anti-inflammatory diets in mitigating CKM progression and mortality, while underscoring the potential of machine learning in personalized risk assessment. Future research should validate these findings and explore underlying mechanisms to optimize dietary interventions.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

CKM

Cardiovascular-kidney-metabolic

DII

Dietary inflammatory index

NHANES

National health and nutrition examination survey

SHAP

SHapley additive explanations

CVD

Cardiovascular disease

CKD

Chronic kidney disease

eGFR

Estimated glomerular filtration rate

PREVENT

Predicting risk of cardiovascular disease events

HR

Hazard ratio

OR

Odd ratio

CI

Confidence interval

RCS

Restricted cubic spline

TC

Total cholesterol

HDL

High-density lipoprotein cholesterol

SBP

Systolic blood pressure

LR

Logistic regression

SVM

Support vector machine

KNN

K-nearest neighbors

NB

Naive bayes

DT

Decision tree

NNET

Neural network

LightGBM

Light gradient boosting machine

XGBoost

eXtreme gradient boosting

AUC-ROC

Area under the receiver operating characteristic curve

Author contributions

Junfeng Ge contributed to the data analysis and drafted the manuscript. Lin Zhu contributed to the data analysis, critical revising the manuscript and the interpretation of data. Sijie Jiang, Wenyan Li and Rongzhan Lin provided technical, and material support. Jun Wu and Fengying Dong provided administrative support. Jing Deng and Yi Lu designed, supervised the study, and contributed to the critical revision of the manuscript. All authors read and approved the final manuscript.

Funding

The Natural Science Foundation of China (82100804), the Natural Science Foundation of Hunan Province (2022JJ30525 and 2023JJ40601), General Topic of the 2024 Annual Health Research Project Hunan Provincial Health Commission (W20243163).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study adhered to the principles outlined in the Declaration of Helsinki, with approval from the Institutional Review Board (IRB) of the National Center for Health Statistics (NCHS) for all procedures involving participants. Written informed consent was obtained from all individuals involved in the study.

Competing interests

The authors declare no competing interests.