The NHANES is an extensive research initiative conducted by the Centers for Disease Control and Prevention (CDC) in the United States to evaluate the health condition and associated health determinants among the U. S. population. Furthermore, the programs of NHANES have been sanctioned by the ethical review board of the National Center for Health Statistics (NCHS) (14).

In this study, we included participants from the NHANES between 2003 and 2018. The exclusion criteria were: under 20 years old, over 79 years old, pregnant, missing dietary intake data, with insufficient data for assessing CKM syndrome and with extreme energy intake data. Extreme energy intake was total energy intake < 500 kcal/day or > 5,000 kcal/day for females or <500 kcal/day or > 8,000 kcal/day males (15). Ultimately, our analysis included a final cohort of 15,233 participants (Figure 1).

The AHA has released diagnostic criteria for CKM syndrome, but the indicators involved in this criterion cannot be fully found in the NHANES database. Therefore, we referred to the diagnostic criteria applicable to the NHANES database mentioned in previous literature, which are very similar to those proposed by the AHA (4). Overall, CKM syndrome consists of 5 stages (stages 0 to 4), divided according to each individual’s obesity, cardiovascular, renal, and metabolic status. Please refer to Supplementary Table 1 for detailed diagnostic criteria.

Complex calculated indicators included the estimated glomerular filtration rate (eGFR) and 10-year cardiovascular disease risk. The eGFR was calculated using the 2021 race-and ethnicity-free Chronic Kidney Disease Epidemiology Collaboration creatinine equation which included serum creatinine and sex (16). And the 10-year CVD risk was assessed using prediction equations for absolute risk of total CVD, which included factors such as sex, age, blood lipids, blood pressure, diabetes, smoking status, eGFR, and the use of lipid-lowering and antihypertensive medications (17). High CVD risk was operationally defined as a ≥ 20% 10-year CVD risk. Please refer to Supplementary Table 2 for detailed calculation process. CKD stage was determined by eGFR and (Urine Albumin-to-Creatinine Ratio) UACR (18). We defined the CKM stage as a binary variable, with 0, 1, or 2 being nonadvanced CKM stages and stages 3 or 4 being advanced CKM stages (19).

Dietary intakes were from a 24-h dietary review survey. Trained investigators conduct interviews with respondents, with the first dietary survey conducted in person and the second conducted through telephone interviews 3 to 10 days after the first survey. Considering the accuracy and authenticity of face-to-face interview data, we only included data from the first dietary survey (20). In this study, we included 9 types of dietary metal intake, including potassium (K), calcium (Ca), magnesium (Mg), phosphorus (P), iron (Fe), copper (Cu), zinc (Zn), and selenium (Se).

Covariates encompassed a spectrum of demographic, lifestyle factors and energy intake. Demographic factors comprised gender (male, female), age (20–39, 40–59, and 60 years and above), ethnicity (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, and other race including multi-racial), educational attainment (below high school, high school graduate/GED or equivalent, and above high school), household poverty-to-income ratio (PIR) (≤1.3, 1.3–3.5, and above 3.5), and marital status (married/living with partner, widowed/divorced/separated, and unmarried). Lifestyle factors included smoking status, drinking status, and physical exercise. Participants were also classified into nonsmokers, former smokers, and current smokers based on their smoking habits. The status of alcohol intake was ascertained by the quantity of alcohol consumed last year, including nondrinkers (no alcohol consume), moderate drinkers (≤ 2 drinks/day for male, ≤ 1 drinks/day for female), heavy drinkers (> 2 drinks/day for male, > 1 drinks/day for female) (21). Low physical activity was defined as no moderate or vigorous physical activity in a typical week from self-reporting (1). The daily energy intake came from the dietary survey on the first day.

All statistical models were completed using R (version 4.4.1). Two-tailed p values < 0.05 were considered statistically significant. Given the multi-stage sampling methodology employed in NHANES, we utilized the dietary survey weight “WTDRD1” to stand for the U. S. population (22). In the descriptive statistical analysis, continuous variables, encompassing energy intake and 9 dietary metal intakes, were depicted through weighted medians (25th, 75th), while categorical variables were exhibited as counts (weighted percentages). Participants were grouped into two categories: non-advanced stages and advanced stages. We contrasted the distribution differences of the variables between these two groups, employing weighted Kruskal-Wallis tests for continuous variables and weighted chi-square tests for categorical variables (23). Spearman rank correlation analysis was implemented to appraise the correlation of dietary metal intake (24). To address the missing values of covariates, aiming to optimize the sample size to the fullest extent, we applied multiple imputation techniques in the “MICE” package (25).

To investigate the relationship between individual dietary metal intake and advanced CKM stages, we applied weighted logistic regression. Each dietary metal intake was stratified into four groups based on its quartiles (Q1-Q4), with the Q1 group serving as a reference point (26). Three models were constructed in this segment: Model 1, the unadjusted model; Model 2, adjusted for demographic variables; Model 3, adjusted for all covariates. Restricted cubic splines (RCS) regression model according to the quantity of knots that meet the minimal Akaike information criterion (AIC) each dietary metal intake was applied to evaluate the potential nonlinear relationship between dietary metal intake and advanced CKM stages (27).

Weighted Quantile Sum (WQS) regression model is extensively utilized in environmental epidemiology to explore the impacts of pollutant mixtures on health outcomes. We substituted environmental pollutants with dietary metal intake and assessed the impact of a mixture of nine dietary metal intakes on advanced CKM stages. In essence, WQS regression categorized various dietary metal intakes into quantiles and computed a weighted index represents the combined effects of the mixture of 9 dietary metal intakes. Additionally, WQS regression established two models by hypothesizing a positive or negative correlation between the mixture and the outcome (28). Specifically, each dietary metal intake will possess a corresponding weight denoting its contribution to advanced CKM stages in each model. If the weight of a dietary metal intake surpasses 0.11 (1/9), it is deemed to play a predominant role in the mixture. The dataset was randomly divided into a training set (50%) and a validation set (50%), and was sampled 1,000 times to achieve more precise results. We also established a Quantile-based computation (gqcomp) model to substantiate the outcomes of WQS regression. This method integrates the straightforward inference of WQS regression with the adaptability of g-computation (29). Unlike WQS regression, the qgcomp model does not presuppose a positive or negative correlation between the mixture and the outcome separately. It also generates a weight for each component to articulate the significance of each component within the mixture. The WQS regression and qgcomp models are not yet suitable for weighted analysis of NHANES (30).

Ultimately, we established subgroup analyses to assess the potential interaction between the dietary metal mixture and subgroup variables including age, sex, smoking status, drinking status, physical activity and energy intake (Classified by median calorie count).

The basic characteristics of 15,233 participants were displayed in Table 1. 12,931 (approximately 88.2%) were in nonadvanced stages of CKM and 2,302 participants (approximately 11.8%) were in advanced stages of CKM. Moreover, about half of the participants (50.2%) were female. By comparing the differences in the distribution of different demography and lifestyle variables between the two groups, we found that there were significant differences between the two groups in all the variables. In terms of metal intake, participants in the advanced stage might have lower levels of all metal intakes (All p < 0.05).

Supplementary Figure 1 showed the correlation heatmap of 9 metal intakes. All the metal intakes are positively correlated. The highest correlation was observed between Mg and Cu (Correlation coefficient = 0.85).

Table 2 showed the associations between single metal intake and advanced CKM stages in weighted logistic regression models. The population was grouped into Q1-Q4 according to each metal intake. In the unadjusted model, K (Q4), Ca (Q3, Q4), Mg (Q2, Q3, Q4), Na (Q2, Q3, Q4), P (Q3, Q4), Fe (Q3, Q4), Cu (Q2, Q3, Q4), Zn (Q3, Q4) and Se (Q2, Q3, Q4) were negative with advanced CKM stages compared with Q1. After adjusting for demography variables, we found that K (Q4), Ca (Q4), Mg (Q2, Q4), Na (Q2, Q3, Q4), P (Q3, Q4), Fe (Q4), Cu (Q2, Q3, Q4), Zn (Q3, Q4) and Se (Q2, Q3, Q4) were still negative associations with advanced CKM stages compared with Q1. In the fully-adjusted model, only Q2 (OR = 0.74, 95% CI = 0.60, 0.92), Q3 (OR = 0.74, 95% CI = 0.58, 0.93) and Q4 (OR = 0.73, 95% CI = 0.55, 0.95) groups of Cu intake remain negative associations with advanced CKM stages compared with Q1 group (All p < 0.05).

The results of RCS regression models were displayed in Figure 2. After adjusting for all covariates, higher intake of Cu was linearly associated with lower risk of advanced CKM stages (p for overall < 0.05 and P for non-linear > 0.05). However, higher intake of Na was linearly associated with higher risk of advanced CKM stages (p for overall < 0.05 and P for non-linear > 0.05). No significant association was found between the intake of other metals and advanced CKM stages (All p for overall > 0.05).

Subsequently, the associations between the metal intake mixture and advanced CKM stages were examined using WQS regression and qgcomp models (Figure 3). Due to the single metal intake model indicating a significant negative correlation between copper intake and advanced CKM stages, we firstly hypothesized a negative relationship between metal intake mixture and advanced CKM stages. However, we did not find a significant effect of metal ingestion mixtures (OR: 1.10, 95% CI: 0.95, 1.26). Conversely, when we hypothesized a positive relationship between metal intake mixture and advanced CKM stages, no significant effect was observed either (OR: 1.13, 95% CI: 0.98, 1.29; Supplementary Figure 2). The qgcomp model yielded similar results to the WQS regression, corroborating an insignificant effect of metal ingestion mixtures (OR: 1.00, 95% CI: 0.94, 1.08; Figure 3B). Although the overall effect of multiple metal intake was not significant, copper intake seemed to be associated with a low risk of advanced CKM stages.

We further evaluated the potential interaction between the Cu intake and subgroup variables, including age, sex, smoking status, drinking status, physical activity and energy intake (Table 3). The association between Cu intake and advanced CKM stages was robust across subgroups. In some subgroups, we still found a correlation between groups of high copper intake and lower risk of advanced CKM stages compared with the lowest Cu intake group.