**3. Results**

Table 1 presents data concerning age, gender, blood pressure, smoking status, and renal function of subjects with low, moderate, and high environmental exposure to Cd. There were significant differences in age and percentages of women and smokers across the three exposure subsets. Female gender was overrepresented in the moderate exposure group. More than half of subjects in the high exposure group were smokers. Blood pressures were recorded in the low and moderate exposure groups; systolic and mean pressures were significantly higher in the latter.


**Table 1.** Study subjects drawn from three localities.

SBP = systolic blood pressure; DBP = diastolic blood pressure; MBP = Mean arterial pressure; CKD = chronic kidney disease; eGFR = estimated glomerular filtration rate; NAG = N-acetyl-β-<sup>d</sup>-glucosaminidase. MBP = DBP + (pulse pressure)/3, where pulse pressure = SBP − DBP. Data for age and eGFR are arithmetic mean values ± standard deviation (SD). Data for blood pressure are geometric mean values ± SD. Data for all other continuous variables are the median (25th, 75th percentile) values. \* Significant % differences among three groups (*p* < 0.05, Pearson Chi-Square test). \*\* Significant % differences between two groups (*p* < 0.001, Pearson Chi-Square test). † Significant mean differences among three groups (*p* < 0.001, Kruskal–Wallis test). ¶ Significant difference from the low exposure group (*p* = 0.003, Mann–Whitney U-test). ¶¶ Significant difference from the low exposure group (*p* = 0.014, Mann–Whitney U-test).

Mean eGFR fell and the percentages of stages 2 and 3 CKD rose with intensity of exposure. The mean serum creatinine concentration was higher in the high-Cd locality. In general, urine concentrations of Cd, NAG, and β2MG rose with Cd exposure, but [NAG]u and [β2MG]u were higher in the moderate Cd group than in the other two. ECd/Ccr, ENAG/Ccr, and <sup>E</sup>β2MG/Ccr followed the same pattern.

Figure 1A–D present scatterplots of eGFR against log(ECd/Ccr) in each exposure subset and the entire sample. In each subset, significant linear and quadratic relationships were documented, and quadratic *R*<sup>2</sup> values were slightly higher. Quadratic *R*<sup>2</sup> was 0.228 in the low-Cd group, 0.083 in the moderate-Cd group, 0.154 in the high-Cd group, and 0.378 in the entire sample. In the linear model, standardized β was −0.467 in the low-Cd group, −0.259 in the moderate-Cd group, −0.361 in the high-Cd group, and −0.598 in the entire sample.

**Figure 1.** ECd/Ccr as a predictor of the estimated glomerular filtration rate (eGFR). Scatterplots compare eGFR to log[(ECd/Ccr) × 105] in subjects grouped by locality (**A**–**C**) and in all subjects (**D**). Quadratic and linear coefficients of determination (*R*2) are provided together with corresponding equations, standardized β coefficients, and *p*-values.

Figure 2A–D present scatterplots of eGFR against log(ENAG/Ccr). As in Figure 1, significant, inverse linear and quadratic relationships were documented in subsets and the entire sample, and quadratic *R*<sup>2</sup> values were slightly higher. Quadratic *R*<sup>2</sup> was 0.055 in the low-Cd group, 0.216 in the moderate-Cd group, 0.381 in the high-Cd group, and 0.139 in the entire sample. In the linear model, standardized β was −0.206 in the low-Cd group, −0.447 in the moderate-Cd group, −0.605 in the high-Cd group, and −0.361 in the entire sample.

The quadratic curves in Figures 1D and 2D indicated that slopes describing rates of GFR reduction varied over the ranges of log[(ECd/Ccr) × 105] and log[(ENAG/Ccr) × 103]. We assumed that log[(ECd/Ccr) × 105] of 3.0 and log[(ENAG/Ccr) × 103] of 1.5 represented the excretion rates of Cd and NAG at which the rates of GFR reduction increased. Table 2 confirms that the slopes changed significantly at these points on the *x*-axes.

**Figure 2.** ENAG/Ccr as a predictor of eGFR. Scatterplots compare eGFR to log[(ENAG/Ccr) × 103] in subjects grouped by locality (**A**–**C**) and in all subjects (**D**). Quadratic and linear coefficients of determination (*R*2) are provided together with corresponding equations, standardized β coefficients, and *p*-values.



The standardized β coefficient indicates the strength of the association of eGFR with log[(ECd/Ccr) × 105] or log[(ENAG/Ccr) × 103]. *R*<sup>2</sup> values are coefficients of determination that indicate the fraction of eGFR variation explained by ECd/Ccr or ENAG/Ccr. *p* ≤ 0.05 identifies statistically significant eGFR reduction rates or associations of eGFR with urinary Cd or NAG excretion.

Figure 3A–D present scatterplots of log(ENAG/Ccr) against log(ECd/Ccr) in the exposure subsets and the entire sample. In all subsets, the two ratios varied directly, significant linear and quadratic relationships were documented, and quadratic *R*<sup>2</sup> values were slightly higher. Quadratic *R*<sup>2</sup> was 0.108 in the low-Cd group, 0.114 in the moderate-Cd group, 0.269 in the high-Cd group, and 0.229 in the entire sample. In the linear model, standardized β was 0.325 in the low-Cd group, 0.327 in the moderate-Cd group, 0.507 in the high-Cd group, and 0.471 in the entire sample.

**Figure 3.** ECd/Ccr as a predictor of ENAG/Ccr. Scatterplots compare log[(ENAG/Ccr)×<sup>10</sup>3] to log[(ECd/Ccr) × 105] in subjects grouped by locality (**A**–**C**) and in all subjects (**D**). Quadratic and linear coefficients of determination (*R*2) are provided together with corresponding equations, standardized β coefficients, and *p*-values.

Figure 4A–D present scatterplots of log(Eβ2MG/Ccr) against log(ECd/Ccr) in the exposure subsets and the entire sample. Figure 4A demonstrates the absence of a relationship at the lowest Cd exposure (quadratic *R*<sup>2</sup> = 0.028, *p* = 0.088). At moderate and high exposure, the two ratios were directly related, significant linear and quadratic relationships were documented, and quadratic *R*<sup>2</sup> values were slightly higher (Figure 4B,C). Quadratic *R*<sup>2</sup> was 0.126 in the moderate exposure group, 0.204 in the high exposure group, and 0.370 in the entire sample. Quadratic and linear relationships in the entire sample were virtually identical (Figure 4D). In the linear model, standardized β was 0.067 in the low-Cd group, 0.334 in the moderate-Cd group, 0.450 in the high-Cd group, and 0.608 in the entire sample.

**Figure 4.** ECd/Ccr as a predictor of <sup>E</sup>β2MG/Ccr. Scatterplots compare log[(Eβ2MG/Ccr) × 104] to log[(ECd/Ccr) × 105] in subjects grouped by locality (**A**–**C**) and in all subjects (**D**). Quadratic and linear coefficients of determination (*R*2) are provided together with corresponding equations, standardized β coefficients, and *p*-values.

Figure 5A–D present scatterplots of log(Eβ2MG/Ccr) against log(ENAG/Ccr). Figure 5A demonstrates the absence of a relationship at the lowest Cd exposure (linear *R*<sup>2</sup> = 0.009, *p* = 0.225). At moderate and high exposure, the two ratios were directly related, significant linear and quadratic regressions were documented, and quadratic R<sup>2</sup> values were slightly higher (Figure 5B,C). Quadratic *R*<sup>2</sup> was 0.152 in the moderate exposure group, 0.426 in the high exposure group, and 0.288 in the entire sample. In the linear model, standardized β was 0.093 in the low-Cd group, 0.360 in the moderate-Cd group, 0.647 in the high-Cd group, and 0.536 in the entire sample.

**Figure 5.** ENAG/Ccr as a predictor of <sup>E</sup>β2MG/Ccr. Scatterplots compare log[(ENAG/Ccr) × 103] to log[(Eβ2MG/Ccr) × 104] in subjects grouped by locality (**A**–**C**) and in all subjects (**D**). Quadratic and linear coefficients of determination (*R*2) are provided together with corresponding equations, standardized β coefficients, and *p*-values.
