**3. Results**

#### *3.1. Descriptive Characteristics of Study Population*

The demographic data, including the blood and urinary biochemistry and other clinical features for the study population of 392 Thai subjects, are shown in Table 1. The overall mean age was 34.9 years. Men were on average 4.1 years younger than women (*p* < 0.001). The mean urinary concentrations of Cd and Pb were 0.25 μg/<sup>L</sup> (0.45 μg/g of creatinine) and 0.89 μg/<sup>L</sup> (1.52 μg/g of creatinine), respectively. The mean eGFR (range) was 105 (70−139) mL/min/1.73 m2. The percentage of the eGFR <90mL/min/1.73 m<sup>2</sup> was similar in men and women (12.3% vs. 13.7%, respectively). The percentage of woman with low iron stores (ferritin levels ≤30 μg/L) were six times that of men (22.3% vs 3.6%, *p* < 0.001). Half of the men (49.7%) smoked (8.9 cigarettes per day), with an average duration of 10 years. There was no record of smoking in any of the women.



a Low body iron stores were defined as plasma ferritin ≤30 μg/L. b eGFR = estimated glomerular filtration rate, determined with Chronic Kidney Disease Epidemiology Collaboration (CKD−EPI) equations [19,21]. Data for age and the eGFR are arithmetic mean values ± standard deviation (SD). Data for all other continuous variables are geometric mean ± SD values. Numbers in parentheses are range. \* *p* ≤ 0.05 indicate mean or % differences between men and women based on the Mann–Whitney U-test or the Pearson chi-squared test, respectively.

The mean BUN, serum creatinine, and urinary creatinine were higher in men than women with *p* values of less than 0.001. The mean plasma protein concentration in men and women was similar (*p* = 0.385), as was the mean urinary protein concentration (*p* = 0.162). ECd/Ecr and EPb/Ecr showed marked differences between men and women. The mean ECd/Ecr was 1.3-fold lower in men than women (0.39 vs. 0.51 μg/g of creatinine, *p* < 0.001), while the mean EPb/Ecr was 1.9-fold lower in men than women (1.10 vs. 2.10 μg/g of creatinine, *p* < 0.001). Notably, ECd/Ccr and EPb/Ccr showed little gender differences. The mean for ECd/Ccr × 100 in men (0.36 μg/L) was nearly identical to that of women (0.34 μg/L), while the mean EPb/Ccr × 100 in men of showed a tendency to be lower than in women (1.02 vs. 1.48 μg/L, *p* = 0.062).

#### *3.2. Predictors of eGFR*

In the multivariable regression analysis for the eGFR with metal excretion rates normalized to Ccr (Table 2), the independent variables (urinary Cd, urinary Pb, age, BUN, serum ferritin, gender and smoking) accounted for 27.2%, 33.4%, 25.9%, 25.4% and 40.2% of the eGFR variability in the entire group, men, women, non-smokers, and smokers, respectively, with the *p* value being <0.001 for the entire group and all subgroups. In the entire group, the eGFR was not associated with urinary Pb (*p* = 0.115), but it showed an inverse association with age (β = −0.436, *p* < 0.001), BUN (β = −0.157, *p* = 0.001) and urinary Cd (β = −0.126, *p* = 0.006).

**Table 2.** Predictors of the estimated glomerular filtration rate (eGFR).


The eGFR is a continuous dependent variable. Independent variables are listed in the first column, including urine Cd as log [(ECd/Ccr) × 105], μg/<sup>L</sup> and urine Pb as log [(EPb/Ccr) × 105]. A standardized regression coefficient β indicates the strength of an association between the eGFR and an independent variable. \* *p* ≤ 0.05 identifies statistically significant associations. An adjusted *R<sup>2</sup>* value indicates the fraction of eGFR variation explained by independent variables. † *p* ≤ 0.05 indicates that the model explained a significant variability of eGFR levels.

In a subgroup analysis, the eGFR was inversely associated with urinary Cd (β = −0.132, *p* = 0.043) and urinary Pb (β = −0.130, *p* = 0.044), only in women. In contrast, the eGFR was not associated with urinary Cd (*p* = 0.219) or with urinary Pb (*p* = 0.333) in men, but it showed a positive association with plasma ferritin (β = 0.147, *p* = 0.017). Inverse associations of the eGFR with age and BUN were evident in all subgroups. The strength of an association between the eGFR and BUN was relatively stronger in male smokers (β −0.207, *p* = 0.016), compared with other subgroups, with β values being −0.125 in men (*p* = 0.044), −0.170 in women (*p* = 0.008), and −0.135 in non-smokers (*p* = 0.012).

In an equivalent multivariable regression analysis of the eGFR with metal excretion rates that were normalized to the excretion of creatinine (Table S1), an association between the eGFR and urinary Cd was not evident in the entire group or in any subgroups, as was the association of the eGFR and urinary Pb.

#### *3.3. Quantitation of E*ff*ects of Cadmium and Lead on the Decline of eGFR*

Figure 1 provides the results of a quantitative analysis of changes in the eGFR that was done by using metal excretion rates that were normalized to creatinine clearance. In the scatterplot of the eGFR against ECd/Ccr, a moderate inverse association was evident (β −0.249, *p* < 0.001) (Figure 1A). Six point two % of the eGFR reduction (*R<sup>2</sup>* = 0.062) could be attributed to Cd. An inverse association was evident also from the scatterplot of the eGFR against EPb/Ccr (Figure 1B). However, the strength of eGFR-EPb/Ccr association was insignificant (*p* = 0.314), and as little as 0.3% of eGFR variation could be attributed to Pb.

**Figure 1.** Comparing effects of cadmium and lead on eGFR change. The scatterplots show the relationship between the eGFR and log [excretion of Cd (ECd)/creatinine clearance (Ccr)) × 105] and between the eGFR and log [excretion of Pb (EPb)/Ccr) × 105] in all subjects (**A**,**B**). The linear equations and coefficients of determination (*R*2) are provided together with standardized β and *p*-values. The bars represent the mean values for the eGFR across urinary Cd and urinary Pb quartiles (**C**,**D**) with adjustments for various covariates and potential interactions. The numbers of subjects are provided for all subgroups. The geometric mean (GM) values (standard deviation) for ECd/Ccr × 100 in urinary Cd quartiles 1, 2, 3 and 4 were 0.12 (0.05), 0.30 (0.05), 0.48 (0.07) and 0.88 (0.44) μg/L, respectively. The GM (SD) for EPb/Ccr × 100 in urinary Pb quartiles 1, 2, 3 and 4 are 0.41 (0.36), 1.26 (0.10), 1.63 (0.14) and 2.73 (2.86) μg/L, respectively.

A generalized linear model (GLM) was then used to estimate the mean eGFR values for subgroups stratified by the quartiles of ECd/Ccr (Figure 1C) and the quartiles of EPb/Ccr (Figure 1D). After adjustments for age, covariates and the interactions, a negative effect of Cd on the eGFR was evident (*p* = 0.015). The estimated mean eGFR (standard error of mean, SEM) for males and females with urinary Cd in the fourth quartile was, respectively, 4.65 (1.72) and 4.94 (1.70) mL/min/1.73 m<sup>2</sup> lower than those with urine Cd in the first quartile (*p* = 0.021) and with urine Cd in the second quartile (*p* = 0.011), respectively. Distinct from Cd, the relationship between Pb and the eGFR was negligible and insignificant (*p* = 0.151) (Figure 1D).

Figure S1 provides the results of an equivalent quantitative analysis of changes in the eGFR that was done by using metal excretion rates that were normalized to the excretion of creatinine. In the scatterplot of the eGFR against ECd/Ecr (Figure S1A), an inverse association between the eGFR and ECd/Ecr was evident (β = −0.104, *p* = 0.040), but this relationship was weakened and became insignificant (*p* = 0.763) after adjustments for age, covariates and interactions (Figure S1C). In contrast, the scatterplot of the eGFR against EPb/Ecr indicated a marginal but non-significant positive association between the eGFR and EPb/Ecr (β = 0.080, *p* = 0.115) (Figure S1B). After adjustments for age, covariates and interactions, there were significant increases in eGFR levels across EPb/Ecr quartiles (*p* = 0.010) (Figure S1D). The estimated mean eGFR (SEM) for subjects with EPb/Ecr in the fourth quartile was 5.89 (1.77) mL/min/1.73 m<sup>2</sup> higher than those with EPb/Ecr in the second quartile (*p* = 0.003).

#### *3.4. The Prevalence Odds of Reduced eGFR across the Quartiles of Urinary Cd and Urinary Pb*

Table 3 provides the results of a logistic regression analysis of the POR for the reduced eGFR, defined as the eGFR at the 25th percentile or below (≤96 mL/min/1.73 m2). The POR for the reduced eGFR showed an inverse association with age (β= −0.071, *p* < 0.001) and gender (β = −1.020, *p* = 0.003). In addition, the POR for the reduced eGFR appeared to rise with urinary Cd in a dose-dependent manner. The POR for the reduced eGFR was 2.87 (95% CI: 1.32, 6.24), 2.51 (95% CI: 1.22, 5.18) and 1.70 (95% CI: 0.875, 3.29) in the urinary Cd quartile 4 (*p* = 0.008), quartile 3 (*p* =0.013) and quartile 1 (*p* = 0.117), respectively. The POR for the reduced eGFR was not associated with the urinary Pb quartile 2 (*p* = 0.198) or quartile 3 (*p* = 0.744), but it rose to 2.23 (95% CI: 1.04, 4.78) in the urinary Pb quartile 4 (*p* = 0.039).


**Table 3.** Prevalence odds ratios for the reduced eGFR across the ECd/Ccr and EPb/Ccr quartiles.

a POR = prevalence odds ratios for eGFR levels ≤96 mL/min/1.73 m2. The eGFR 96 mL/min/1.73 m<sup>2</sup> corresponds to the 25th percentile eGFR. b Low iron store status was defined as serum ferritin levels ≤ 30μg/L. \* *p* ≤ 0.05 indicates a statistically significant increment of POR, compared with the reference. The GM (SD) for ECd/Ccr and EPb/Ccr together with number of subjects in all urinary Cd and urinary Pb quartiles are as in Figure 1.

Table S2 provides the results of an equivalent logistic regression analysis for the reduced eGFR that was done by using urinary Cd and urinary Pb that were normalized to the excretion of creatinine. In this analysis, the POR for the reduced eGFR only showed an inverse association with age (β = −0.080, *p* < 0.001). No associations were seen between the reduced eGFR and urinary Cd or Pb in any quartiles of urinary Cd or Pb.
