As cystatin C was measured exclusively during the 1999–2004 waves of the National Health and Nutrition Examination Survey (NHANES), our study analyzed data from 31,126 participants within this period. Since the Predicting Risk of cardiovascular diseases EVENTs (PREVENT) base model, which is used to define CKM, is applicable to U.S. individuals aged 30–79 years (16), our study only included subjects in this age group and excluded those with a classification of “Other Race—Including Multi-Racial.” In addition, we excluded individuals with missing variables required to define CKM (Supplementary Tables S1, S2), those without serum cystatin C and serum creatinine data, and pregnant women. The specific inclusion and exclusion criteria are presented in Supplementary Figure S1. Finally, 3,622 participants were included in our study. The NHANES program was approved by the National Center for Health Statistics (NCHS) Ethics Review Board, and all participants provided written informed consent (17). Since NHANES data are de-identified and anonymized during analysis, secondary analyses did not require additional ethical approval or informed consent.

Cystatin C levels were measured using a cystatin C immunoassay on an automated multi-channel analyzer (Siemens Dimension Vista 1,500, Siemens Healthcare Diagnostics) (18). For creatinine levels, an improved version of the Jaffé reaction, modified by Popper et al. and Seeling and Wuest, was used. Values below 0.6 mg/dL were considered insignificant and treated as missing data (19). Notably, there were discrepancies between the creatinine levels measured during the NHANES 1999–2000 and the gold standard reference method (i.e., the Roche coupled enzymatic assay performed on a Roche P Module instrument). Therefore, following the NHANES recommendations, Deming regression was applied to adjust the creatinine results for the NHANES 1999–2000 study (19).

eGFRcystatin was calculated using the CKD-EPI cystatin C (2012) equation, while eGFRcreatinine was calculated using the CKD-EPI creatinine (2009) equation (20, 21). The best eGFR (eGFRbest) is the average of eGFRcreatinine and eGFRcystatin (22). eGFRdiff is the difference between eGFRcystatin and eGFRcreatinine. eGFRdiff was analyzed as a continuous variable (per 10 mL/min/1.73 m2). Furthermore, participants were categorized into three groups according to their eGFRdiff: negative (<−15 mL/min/1.73 m2), midrange (−15 to 15 mL/min/ 1.73 m2), and positive (≥15 mL/min/1.73 m2) (13).

CKM Stage 0 comprised individuals who were not overweight (body mass index (BMI) <25 kg/m²) and had no metabolic risk factors (including hypertriglyceridemia [<135 mg/dL], hypertension, prediabetes, diabetes, and metabolic syndrome) or chronic kidney disease (CKD). CKM Stage 1 included individuals with a BMI ≥25 kg/m², a waist circumference ≥ 88/102 cm in women/men, and/or prediabetes, without other metabolic risk factors or CKD. CKM Stage 2 included individuals with metabolic risk factors (including hypertriglyceridemia [≥135 mg/dL], hypertension, diabetes, and metabolic syndrome) or CKD. CKM Stage 3 included individuals with CKM and a 10-year cardiovascular disease (CVD) risk ≥20%, as predicted by the PREVENT base model (where the values for age and other risk factors outside the validated PREVENT ranges were imputed as the upper and lower limits of these ranges) (16). CKM Stage 4 included individuals with clinical CVD (coronary heart disease, congestive heart failure, and stroke) (4, 6). Hypertension was defined as a history of the disease or elevated blood pressure measurements. Diabetes and prediabetes were diagnosed based on disease history, measured blood glucose levels, and glycohemoglobin levels. Metabolic syndrome (MetS) was identified based on the harmonized criteria set by the International Diabetes Federation. CKD was defined using eGFRbest and disease history. A more detailed definition of the variables used to define the CKM stages is presented in Supplementary Tables S1, S2. We defined CKM syndrome as CKM stage ≥2.

Data on age, gender, ethnicity, income level, and education level were collected using standardized demographic questionnaires (23). Data on smoking status were gathered through a smoking questionnaire, and physical activity was assessed through an activity questionnaire (23). The Healthy Eating Index (HEI) was calculated according to the HEI-2015 (24). The urine albumin-creatinine ratio (uACR, mg/g) is the ratio of urine albumin to creatinine. A more detailed definition of covariables is presented in Supplementary Table S3.

Data on all-cause and cardiovascular mortality status were obtained from the National Death Index (NDI). Cardiovascular mortality was defined using the International Classification of Diseases-10 (ICD-10) codes I00-I09, I11, I13, I20-I51, and I60-I69. Time was measured in months from the physical examination to either the date of death or the end of the follow-up period (31 December 2019).

Before analysis, we utilized the random forest imputation method to estimate the missing values, as the urine albumin-creatinine ratio had the highest proportion of missing values (13.94%, Supplementary Figure S2). According to the “highest proportion of missing values” principle, we performed complex sample analysis using fasting sample weights (25). Continuous variables were presented as weighted means and standard deviations (SDs), while categorical variables were expressed as weighted frequencies and proportions. The baseline characteristics were compared using eGFRdiff. The weighted Wilcoxon rank-sum test and the chi-squared test with Rao–Scott second-order corrections were used to compare group differences for continuous and categorical variables, respectively.

We first estimated the age-adjusted prevalence of each CKM stage by standardizing to the 2000 U.S. population census, which includes three age groups (30–39, 40–59, and 60–79 years) (26, 27). Then, we used ordinal logistic regression to explore the relationship between eGFRdiff and CKM stages. Model 1 was unadjusted for covariables. Model 2 included adjustments for age, gender, ethnicity, income level, educational level, smoking status, average physical activity level, HEI, and uACR. Model 3 was further adjusted for eGFRbest. To explore potential non-linear associations, we used restricted cubic splines (RCSs) in Model 3. The reference value was set at 20 mL/min/1.73m², with the 10th, 50th, and 90th percentiles of eGFRdiff serving as the three knots for the spline. Subgroup analyses were further performed based on age (cutoff value of 65 years), gender, ethnicity, uACR, smoking status, and physical activity.

We further used the Cox regression model to explore the association between eGFRdiff and both all-cause mortality and cardiovascular mortality in the participants with and without CKM syndrome. Model 1 was unadjusted for covariables. Model 2 included adjustments for age, gender, ethnicity, income level, educational level, smoking status, average physical activity level, HEI, and uACR. Model 3 was further adjusted for eGFRbest. To explore potential non-linear associations, we used RCSs in Model 3. The reference value was set at 20 mL/min/1.73m², with the 10th, 50th, and 90th percentiles of eGFRdiff serving as the three knots for the spline. Since the criteria for defining CKM stages include hypertension, blood glucose status, systolic blood pressure, triglycerides, and cholesterol, we excluded these variables from our models (28, 29).

Furthermore, we employed complete case analysis to avoid imputation bias for the missing data (n = 778). A two-sided p-value of <0.05 was considered statistically significant. Data analyses were performed using R 4.3.2.

Table 1 presents the baseline characteristics of the 3,622 participants in our study. The weighted mean age was 49.62 years (SD, 12.92 years); 50.74% of the participants were women, and 78.75% were non-Hispanic white. The unweighted proportions of the participants in CKM Stages 0–4 were as follows: 13.20% for Stage 0, 16.46% for Stage 1, 52.62% for Stage 2, 7.62% for Stage 3, and 10.10% for Stage 3.

Compared to the participants with the midrange eGFRdiff, those with the negative eGFRdiff were more likely to be current smokers and had higher levels of body mass index (BMI), waist circumference, systolic blood pressure (SBP), and triglycerides, in addition to higher rates of hypertension, diabetes, cardiovascular disease (CVD) and chronic kidney disease (CKD) stage (all p < 0.001). Conversely, participants with positive eGFRdiff showed better health indicators, including BMI, waist circumference, SBP, triglycerides, high-density lipoprotein, hypertension, diabetes, CVD, and CKD stage (all p < 0.001).

The prevalence of higher CKM stages was significantly higher in participants with negative eGFRdiff and lower in those with positive eGFRdiff (Supplementary Table S4). The age-adjusted prevalence rates for CKM Stage 3 were 10.06% [95% confidence interval (CI), 3.31 to 16.81] in the negative eGFRdiff group and 3.23% [95% CI, 2.22 to 4.24] in the positive eGFRdiff group. For CKM Stage 4, the rates were 14.9% [95% CI, 6.35 to 23.44] in the negative eGFRdiff group and 6.43% [95% CI, 5.08 to 7.78] in the positive eGFRdiff group.

When eGFRdiff was treated as a continuous variable, the common odds ratio (cOR) per 10 mL/min/1.73m² increase in eGFRdiff in the fully adjusted model was 0.86 [95% confidence interval (CI), 0.81 to 0.91] (Table 2). Compared to the midrange eGFRdiff, the cOR values for the negative and positive eGFRdiff were 1.88 [95% CI, 1.23 to 2.88] and 0.69 [95% CI, 0.58 to 0.83], respectively (Table 2). RCS analysis suggested no significant non-linear association between eGFRdiff and CKM stages (Figure 1).

The association between eGFRdiff and CKM stages can be found in the majority of the subgroups (Supplementary Figure S3). The cOR values for the participants aged <65 years, female participants, male participants, non-Hispanic white participants, and participants who had never smoked were 0.82 [95%CI, 0.75 to 0.89], 0.83[95%CI, 0.75 to 0.92], 0.89[95%CI, 0.83 to 0.95], 0.86[95%CI, 0.8 to 0.92], and 0.78[95%CI, 0.72 to 0.84], respectively. Although not statistically significant (p = 0.746), the cOR for the participants aged 65 years and older was 1.02 [95% CI, 0.92 to 1.12].

During the median follow-up period of 201 months, 853 participants (853/2548 participants with CKM syndrome, 33.48%) experienced all-cause mortality, and 265 participants (265/2548 participants with CKM syndrome, 10.4%) experienced cardiovascular mortality. In the fully adjusted model, for participants with CKM syndrome, the hazard ratios (HRs) per 10 mL/min/1.73m² increase in eGFRdiff were 0.88 [95% CI, 0.83 to 0.93] for all-cause mortality and 0.90 [95%CI, 0.81 to 1.00] for cardiovascular mortality, with the latter being marginally significant (p = 0.052) (Table 3). Compared to participants with midrange eGFRdiff, those with negative eGFRdiff had a 48% increased risk of all-cause mortality, while those with positive eGFRdiff had a 30% reduced risk of all-cause mortality. There was no significant difference in cardiovascular mortality risk between participants with negative or positive eGFRdiff and those with midrange eGFRdiff (Table 3). The RCS analysis indicated no significant non-linear association between eGFRdiff and both all-cause mortality and cardiovascular mortality in participants with and without CKM syndrome (Figure 2).

After further excluding 778 participants with missing covariate data (Supplementary Figure S2), a total of 2,844 participants were included in the complete case analysis. The cOR per 10 mL/min/1.73m² increase in eGFRdiff in the fully adjusted model was 0.84 [95% CI, 0.79 to 0.89] (Supplementary Table S5). Compared to participants with midrange eGFRdiff, those with negative eGFRdiff had a cOR of 1.97 [95% CI, 1.1 to 3.52] and those with positive eGFRdiff had a cOR of 0.66 [95% CI, 0.54 to 0.79] (Supplementary Table S5). No significant non-linear association was detected (Supplementary Figure S4).

During a median follow-up of 204 months, 538 participants included in the complete-case analysis experienced all-cause mortality and 159 participants experienced cardiovascular mortality. In the fully adjusted model, the hazard ratios (HRs) per 10 mL/min/1.73m² increase in eGFRdiff were 0.87 [95% CI, 0.82 to 0.93] and 0.97 [95% CI, 0.89 to 1.07] for all-cause mortality and cardiovascular mortality (Supplementary Table S6), respectively. No significant non-linear association was detected (Supplementary Figure S5).