3.4.1. Ecr-Normalized Dataset

As data in Figures 1 and 2 indicate, the differences between BMDL and BMDU figures of ECd/Ecr were small for both continuous and quantal endpoints. The BMDL-BMDU figures of ECd/Ecr calculated from Cd-dose and eGFR response models were higher in females than males.

**Figure 1.** BMD estimates of ECd/Ecr from eGFR reduction endpoint with BMR at 5%. ECd/Ecr and eGFR 264 data were fitted to an inverse exponential model (**a**), a natural logarithmic model (**b**), an exponential 265 model (**c**), and Hill model (**d**). Bootstrap curves were based on model averaging of ECd/Ecr BMD 266 estimates for all subjects (**e**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ecr 267 associated with a 5 % reduction in eGFR (**f**). 268 **Figure 1.** BMD estimates of ECd/Ecr from eGFR reduction endpoint with BMR at 5%. ECd/Ecr and eGFR data were fitted to an inverse exponential model (**a**), a natural logarithmic model (**b**), an exponential model (**c**), and Hill model (**d**). Bootstrap curves were based on model averaging of ECd/Ecr BMD estimates for all subjects (**e**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ecr associated with a 5 % reduction in eGFR (**f**).

a BMD value of ECd/Ecr in µg/g creatinine.

CI, confidence interval; U/L, BMDU/BMDL.

**Figure 2.** BMD estimates of ECd/Ecr from quantal eGFR endpoint with BMR at 10%. Bootstrap curves 269 were based on model averaging 95% confidence intervals of BMD of ECd/Ecr in males and females 270 (**a**) and in all subjects (**b**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ecr 271 associated with a 10% increase in prevalence of CKD (**c**). 272 **Figure 2.** BMD estimates of ECd/Ecr from quantal eGFR endpoint with BMR at 10%. Bootstrap curves were based on model averaging 95% confidence intervals of BMD of ECd/Ecr in males and females (**a**) and in all subjects (**b**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ecr associated with a 10% increase in prevalence of CKD (**c**).

3.4.2. Ccr-Normalized Dataset 273 For all subjects, the BMDL/BMDU of ECd /Ecr for continuous and quantal endpoints were 0.828/1.71 and 2.77/5.06 µg/g creatinine, respectively.

#### As data in Figures 3 and 4 indicate, the differences between BMDL and BMDU fig- 274 ures of ECd/Ccr were small for both continuous and quantal endpoints. The BMDL-BMDU 275 3.4.2. Ccr-Normalized Dataset

figures of ECd/Ecr calculated by Cd-dose and eGFR response models in males and females 276 were nearly identical. 277 For all subjects, the BMDL/BMDU of ECd /Ccr for continuous and quantal endpoints 278 were 10.4/24 and 56.1/83.1 ng/L filtrate, respectively. 279 As data in Figures 3 and 4 indicate, the differences between BMDL and BMDU figures of ECd/Ccr were small for both continuous and quantal endpoints. The BMDL-BMDU figures of ECd/Ecr calculated by Cd-dose and eGFR response models in males and females were nearly identical.

For all subjects, the BMDL/BMDU of ECd /Ccr for continuous and quantal endpoints were 10.4/24 and 56.1/83.1 ng/L filtrate, respectively.

**Figure 3.** BMD estimates of ECd/Ccr from eGFR reduction endpoint with BMR at 5%. ECd/Ccr and 280 eGFR data were fitted to an inverse exponential model (**a**), a natural logarithmic model (**b**), an ex- 281 ponential model (**c**), and Hill model (**d**). Bootstrap curves were based on model averaging of ECd/Ccr 282 BMD estimates for all subjects (**e**). Outputs of all fitted models as BMDL and BMDU estimates of 283 ECd/Ccr associated with a 5% reduction in eGFR (**f**). 284 **Figure 3.** BMD estimates of ECd/Ccr from eGFR reduction endpoint with BMR at 5%. ECd/Ccr and eGFR data were fitted to an inverse exponential model (**a**), a natural logarithmic model (**b**), an exponential model (**c**), and Hill model (**d**). Bootstrap curves were based on model averaging of ECd/Ccr BMD estimates for all subjects (**e**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ccr associated with a 5% reduction in eGFR (**f**).

All 56.1 83.1 1.481

a BMD value of ECd/Ccr in ng/L filtrate.

CI, confidence interval; U/L, BMDU/BMDL.

**Figure 4.** BMD estimates of ECd/Ccr from quantal eGFR endpoint with BMR at 10%. Bootstrap curves 285 were based on model averaging 95% confidence intervals of BMD of ECd/Ccr in males and females 286 (**a**) and in all subjects (**b**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ecr 287 associated with a 10% increase in the prevalence of CKD (**c**). 288 **Figure 4.** BMD estimates of ECd/Ccr from quantal eGFR endpoint with BMR at 10%. Bootstrap curves were based on model averaging 95% confidence intervals of BMD of ECd/Ccr in males and females (**a**) and in all subjects (**b**). Outputs of all fitted models as BMDL and BMDU estimates of ECd/Ecr associated with a 10% increase in the prevalence of CKD (**c**).

#### **4. Discussion**

**4. Discussion** 289 In a dose–response analysis of a large dataset from apparently healthy participants 290 (mean age 48.3 years), older age and higher BMI were independently associated with 291 higher risks of CKD, based on the low eGFR criterion (Table 2). These findings are con- 292 sistent with the literature reports of age, overweight and obesity as common CKD risk 293 factors [27–30]. In addition to these two risk factors, we have found the measure of long- 294 term exposure to Cd (ECd/Ecr) to be another independent risk factor of CKD (Table 3). An 295 association between low environmental Cd exposure and a decrease in eGFR to levels 296 commensurate with CKD has been observed in population-based studies in the U.S. [31– 297 In a dose–response analysis of a large dataset from apparently healthy participants (mean age 48.3 years), older age and higher BMI were independently associated with higher risks of CKD, based on the low eGFR criterion (Table 2). These findings are consistent with the literature reports of age, overweight and obesity as common CKD risk factors [27–30]. In addition to these two risk factors, we have found the measure of long-term exposure to Cd (ECd/Ecr) to be another independent risk factor of CKD (Table 3). An association between low environmental Cd exposure and a decrease in eGFR to levels commensurate with CKD has been observed in population-based studies in the U.S. [31–34], Taiwan [35] and Korea [36–38].

34], Taiwan [35] and Korea [36–38]. 298 In this study, the risk of CKD was increased by 6.2- and 10.6-fold, when ECd/Ecr ≤ 0.37 299 µg/g creatinine rose to 0.38−2.49 and ≥ 2.5 µg/g creatinine, respectively. These Cd-dose 300 dependent increases in the risk of CKD were strengthened by the results obtained from 301 the Ccr-normalized dataset where the risk of CKD was increased by 4.4- and 20.8-fold, 302 comparing ECd/Ccr ≤ 9.9 ng/L filtrates with ECd/Ccr of 10−49.9 and ≥ 50 ng/L filtrate, respec- 303 tively. This confirmation is noteworthy because normalizing ECd to Ecr can cause a wide 304 dispersion of dataset due to the interindividual differences in Ecr such as muscle mass 305 In this study, the risk of CKD was increased by 6.2- and 10.6-fold, when ECd/Ecr ≤ 0.37 µg/g creatinine rose to 0.38–2.49 and ≥ 2.5 µg/g creatinine, respectively. These Cd-dose dependent increases in the risk of CKD were strengthened by the results obtained from the Ccr-normalized dataset where the risk of CKD was increased by 4.4- and 20.8-fold, comparing ECd/Ccr ≤ 9.9 ng/L filtrates with ECd/Ccr of 10–49.9 and ≥50 ng/L filtrate, respectively. This confirmation is noteworthy because normalizing ECd to Ecr can cause a wide dispersion of dataset due to the interindividual differences in Ecr such as muscle mass which is unrelated to neither Cd exposure nor nephron function [11,12].

which is unrelated to neither Cd exposure nor nephron function [11,12]. 306 Because of such increased variance in datasets introduced by Ecr-normalization, the 307 effect of chronic exposure to low-dose Cd on eGFR was not realized. For example, a sys- 308 tematic review and meta-analysis of pooled data from 28 studies reported that the risk of 309 proteinuria was increased by 1.35-fold when comparing the highest vs. lowest category of 310 Cd dose metrics, but an increase in the risk of low eGFR was statistically insignificant (*p* = 311 Because of such increased variance in datasets introduced by Ecr-normalization, the effect of chronic exposure to low-dose Cd on eGFR was not realized. For example, a systematic review and meta-analysis of pooled data from 28 studies reported that the risk of proteinuria was increased by 1.35-fold when comparing the highest vs. lowest category of Cd dose metrics, but an increase in the risk of low eGFR was statistically insignificant (*p* = 0.10) [39]. An erroneous conclusion that chronic Cd exposure was not associated with a progressive eGFR reduction was also made in another systematic review [40].

A significant relationship was seen between ECd and a decrease in eGFR with adjustment for covariates (Table 3). We subsequently applied the BMD method to our Ecr- and Ccr-normalized datasets to identify ECd/Ecr and ECd/Ccr values below which an adverse effect of Cd on eGFR can be discerned. The BMDL/BMDU figures of ECd/Ecr, estimated from the eGFR reduction endpoint were 0.839/1.81, 0.849/1.74 and 0.828/1.71 µg/g creatinine in males, females and all subjects, respectively (Figure 1). The corresponding BMDL/BMDU figures of ECd/Ccr were 11.3/24.3, 11.3/24.1 and 10.4/24 ng/L filtrate in males, females and all subjects, respectively (Figure 3).

The BMD values of Cd exposure levels calculated from toxic tubular cell injury and reduced tubular reabsorption of the filtered protein β2M can be found in numerous studies [41,42]. In contrast, a report of BMDL/BMDU of Cd exposure levels associated with eGFR reduction could only be found in a study of 790 Swedish women, aged 53–64 years, where the reported BMDL values for the glomerular endpoint were 0.7–1.2 µg/g creatinine [43]. These BMD values were slightly lower than those calculated for females in the present study (0.849/1.74 µg/g creatinine). The differences may be attributable to lower Ecr in Thai women than in Swedish women. Nevertheless, all these BMD values were lower than ECd/Ecr of 5.24 µg/g creatinine, which suggested to be a threshold level for the nephrotoxicity of Cd when Eβ2M/Ecr > 300 was used as the POD [8].

In our quantal eGFR endpoint analysis (Figure 2), the BMDL/BMDU values of ECd associated with a 10% increase in CKD prevalence were 2.77/5.06 µg/g creatinine (56.1/83.1 ng/L filtrate). These data suggested that population CKD prevalence was likely to be smaller than 10% at ECd/Ecr < 2.77 µg/g creatinine (<56.1 ng/L filtrates). Thus, the population health risk associated with ECd/Ecr < 2.77 µg/g creatinine could not be discerned. The impact of Cd exposure on GFR has long been underestimated due to the common practice of normalizing ECd to Ecr. The comparability of guidelines between populations could be improved by the universal acceptance of a consistent normalization of ECd to Ccr that eliminates the effect of muscle mass on Ecr, thereby giving a more accurate assessment of the severity of Cd nephropathy [10].

A tolerable intake level of 0.28 µg/kg body weight per day was derived in a risk calculation using pooled data from Chinese population studies [44]. This consumption level, equivalent to 16.8 µg/day for a 60 kg person, was derived from an Eβ2M/Ecr endpoint where the BMDL value of ECd/Ecr for such an endpoint was 3.07 µg/g creatinine. This BMDL estimate was 3.7-fold higher than the BMDL of 0.828 µg/g creatinine derived in the present study. In another Chinese population study, dietary Cd intake estimates at 23.2, 29.6, and 36.9 µg/d were associated with 1.73-, 2.93- and 4.05-fold increments in the prevalence of CKD, compared with the 16.7 µg/d intake level [45]. A diet high in rice, pork, and vegetables was associated with a 4.56-fold increase in the prevalence of CKD [45].

The European Food Safety Authority (EFSA) also used the β2M endpoint. However, the EFSA included an uncertainty factor (safety margin), and an intake of 0.36 µg/kg body weight per day for 50 years as an acceptable Cd ingestion level or a reference dose (RfD) [46]. The EFSA designated ECd/Ecr of 1 µg/g creatinine as the toxicity threshold level for an adverse effect on kidneys. This Cd excretion of 1 µg/g creatinine is 17 % higher than our NOAEL equivalent of Cd excretion of 0.828 µg/g creatinine.

The Cd toxicity threshold level, RfD and an acceptable consumption level derived from the β2M excretion above ≥300 µg/ g creatinine do not appear to be without an appreciable health risk. In theory, health-risk assessment should be based on the most sensitive endpoint with consideration given to subpopulations with increased susceptibility to Cd toxicity such as children.

In the present study, the body burden of Cd, measured as ECd/Ecr, was increased by 3-fold in men and women who smoked cigarettes (Table 1). These results are expected, given that the tobacco plant accumulates high levels of Cd in its leaves, and the volatile metallic Cd and oxide (CdO) generated from cigarette burning are more bioavailable than Cd that enters the body through the gut [47,48].

The diet is the main Cd exposure source for non-smoking and non-occupationallyexposed populations. In a temporal trend analysis of environmental Cd exposure in the U.S., the mean urinary Cd fell by 29% in men (0.58 vs. 0.41 µg/g creatinine, *p* < 0.001) over 18 years (NHANES 1988–2006), but not in women (0.71 vs. 0.63 µg/g creatinine, *p* = 0.66) [49]. Such a reduction in Cd exposure among men was attributable to a decrease in smoking prevalence [50]. In contrast, total diet studies in Australia [51], France [52], Spain [53] and the Netherlands [2] reported that dietary Cd exposure levels among young children exceeded the current health guidance values. These data are concerning for the reasons below.

CKD is a progressive syndrome with high morbidity and mortality and affects 8% to 16% of the world's population [27–30]. An upward trend of its incidence continues, while an adverse effect of Cd on eGFR and the risk of CKD have increasingly been reported. Higher Cd excretion was associated with lower eGFR in studies from Guatemala [54] and Myanmar [55]. The effect of Cd exposure on eGFR observed in children is particularly concerning. In a prospective cohort study of Bangladeshi preschool children, an inverse relationship between urinary Cd excretion and kidney volume was seen in children at 5 years of age. This was in addition to a decrease in eGFR [56]. Urinary Cd levels were inversely associated with eGFR, especially in girls. In another prospective cohort study of Mexican children, the reported mean for Cd intake at the baseline was 4.4 µg/d, which rose to 8.1 µg/d after nine years, when such Cd intake levels showed a marginally inverse association with eGFR [57].

#### **5. Conclusions**

Environmental exposure to Cd, old age, and elevated BMI are independent risk factors for reduced eGFR. For the first time, the BMDL/BMDU figures of Cd excretion levels associated with a decrease in eGFR have been computed for men and women. The narrow BMDL-BMDU differences indicate the high degree of statistical certainty in these derived NOAEL equivalent figures. The BMDL/BMDU estimates of the Cd excretion associated with a decrease in eGFR in all subjects are 0.828/1.71 µg/g creatinine. The BMDL/BMDU estimates of Cd excretion associated with a 10% increase in the prevalence of CKD are 2.77/5.06 µg/g creatinine. These NOAEL equivalents indicate a decrease in eGFR due to Cd nephropathy occurs at the body burdens lower than those associated with Cd excretion of 5.24 µg/g creatinine and an increase in β2M excretion above 300 µg/g creatinine. The established nephrotoxicity threshold level for Cd is outdated and is not protective of human health. Human health risk assessment should be based on current scientific research data.

**Author Contributions:** Conceptualization, S.S.; methodology, S.S., S.Y. and A.B.Ð.; formal analysis, S.S. and A.B.Ð.; investigation, S.S. and S.Y.; resources, G.C.G. and D.A.V.; writing—original draft preparation, S.S.; writing—review and editing, G.C.G. and D.A.V.; project administration, S.S. and S.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** This study analyzed archived data taken from published reports [17,18]. Ethical review and approval were not applicable.

**Informed Consent Statement:** All participants took part in the study after giving informed consent.

**Data Availability Statement:** All data are contained within this article.

**Acknowledgments:** This work was supported with resources from the Kidney Disease Research Collaborative, Translational Research Institute and the Department of Nephrology, Princess Alexandra Hospital. It was also supported by the resources of the Department of Toxicology "Akademik Danilo Soldatovi´c", University of Belgrade-Faculty of Pharmacy, Serbia.

**Conflicts of Interest:** The authors have declared no potential conflict of interest.
