Estimation of the Cadmium Nephrotoxicity Threshold from Loss of Glomerular Filtration Rate and Albuminuria

Cadmium (Cd) is a pervasive, toxic environmental pollutant that preferentially accumulates in the tubular epithelium of the kidney. Current evidence suggests that the cumulative burden of Cd here leads to the progressive loss of the glomerular filtration rate (GFR). In this study, we have quantified changes in estimated GFR (eGFR) and albumin excretion (Ealb) according to the levels of blood Cd ([Cd]b) and excretion of Cd (ECd) after adjustment for confounders. ECd and Ealb were normalized to creatinine clearance (Ccr) as ECd/Ccr and Ealb/Ccr. Among 482 residents of Cd-polluted and non-polluted regions of Thailand, 8.1% had low eGFR and 16.9% had albuminuria (Ealb/Ccr) × 100 ≥ 20 mg/L filtrate. In the low Cd burden group, (ECd/Ccr) × 100 < 1.44 µg/L filtrate, eGFR did not correlate with ECd/Ccr (β = 0.007) while an inverse association with ECd/Ccr was found in the medium (β = −0.230) and high burden groups (β = −0.349). Prevalence odds ratios (POR) for low eGFR were increased in the medium (POR 8.26) and high Cd burden groups (POR 3.64). Also, eGFR explained a significant proportion of Ealb/Ccr variation among those with middle (η2 0.093) and high [Cd]b tertiles (η2 0.132) but did not with low tertiles (η2 0.001). With an adjustment of eGFR, age and BMI, the POR values for albuminuria were increased in the middle (POR 2.36) and high [Cd]b tertiles (POR 2.74) and those with diabetes (POR 6.02) and hypertension (2.05). These data indicate that (ECd/Ccr) × 100 of 1.44 µg/L filtrate (0.01–0.02 µg/g creatinine) may serve as a Cd threshold level based on which protective exposure guidelines should be formulated.


Introduction
Chronic kidney disease (CKD) is a progressive disease with high morbidity and mortality, affecting 8% to 16% of the world's population [1][2][3].CKD is diagnosed when the estimated glomerular filtration rate (eGFR) falls below 60 mL/min/1.73m 2 , termed low eGFR, or albuminuria, defined as albumin to creatinine ratios (ACR) ≥ 30 mg/g creatinine in women and ≥ 20 mg/g creatinine in men that persists for at least 3 months [1,2].Environmental exposure to the toxic metal pollutant cadmium (Cd) is linked to an increased risk of both low eGFR and albuminuria.Typically, studies of CKD associated with environmental Cd exposure employed levels of blood Cd ([Cd] b ) and excretion of Cd (E Cd ), normalized to creatinine excretion (E Cd /E cr ) as indicators of exposure [3,4].Notably, however, these Cd exposure indicators have been found to be also associated with an increased risks of diabetes [5][6][7][8][9][10] and hypertension [11][12][13][14][15][16]; both of which are major risk factors for CKD.Hypertension is known to be a cause and result of CKD [17,18].Furthermore, in a Dutch prospective cohort study, smoking, which is a source of Cd exposure, promoted kidney failure, evident from a reduction in eGFR to below 15 mL/min/1.73m 2 [19].It is imperative that these covariates are adequately adjusted for the estimation of risks of GFR loss and albuminuria due to renal Cd accumulation.
The dose-response relationship between Cd exposure and adverse kidney outcomes measured by a reduction in eGFR and albuminuria is not well defined.In comparison, numerous dose-effect studies employed tubular damage and tubulopathy reflected by tubular proteinuria as signs of the nephrotoxicity due to Cd accumulation.However, current evidence suggests that a sustained decrease in eGFR after Cd exposure is more suitable than the tubulopathy endpoint, especially for the purpose of determining protective exposure guidelines.
Also, current evidence suggests that excreted Cd (E Cd ) is related to tubular cell injury and death as a result of the cumulative burden of Cd [20].The excreted Cd emanates from tubular epithelial cells complexed with metallothionein (MT) as CdMT [21], the storage form of Cd.Consequently, it is logical to normalize E Cd to creatinine clearance (C cr ) rather than creatinine excretion (E cr ) because C cr is a surrogate for GFR, which is the measurable analogue of nephron number [3,20,22].
The present study aims to explore if there is a potential toxicity threshold level of Cd accumulation in kidneys.To achieve this aim, changes in eGFR and albumin excretion (E alb ) together with the risks of low eGFR and albuminuria were quantified in relation to both E Cd and [Cd] b .The confounding impacts of smoking, diabetes and hypertension were also evaluated.We collected data from women and men who resided in Cd-polluted and non-polluted regions of Thailand to obtain a wide range of exposure levels pivotal to the a dose-response analysis.
To depict an amount of Cd and albumin excreted per volume of filtrate, known also as primary urine, E Cd and E alb were normalized to C cr as E Cd /C cr and E alb /C cr , respectively [23].This C cr -normalization corrects for differences in urine flow rate (dilution) and the number of functioning nephrons among cohort participants; they are unaffected by creatinine excretion (E cr ) [23].Therefore, (E alb /C cr ) × 100 ≥ 20 mg/L filtrate identifies albuminuria in both men and women.In comparison, when E alb is adjusted to creatinine excretion (E cr ) as albumin to creatinine ratios (ACR), cut-off values for albuminuria in women and men differ because of universally lower E cr values in women than in men.

Cohort Participants
To establish a clear dose-response relationship, a population exposed to a wide range of Cd doses is required.Therefore, we selected subjects from two population-based cross-sectional studies undertaken in a Cd-contaminated area of the Mae Sot District, Tak Province [24] and a non-contaminated location in the Nakhon-Si-Thammarat Province [25] of Thailand.More females (n = 354) were recruited to the present study than males (n = 118), given that an increase in mortality from kidney disease was found especially in women in a prospective cohort study of residents of a Cd-polluted area of Japan [26].
The data from a nationwide survey of Cd levels in soils and food crops indicated that environmental exposure to Cd in Nakhon Si Thammarat was low [27].In comparison, the Cd concentration of the paddy soil samples from the Mae Sot district exceeded the standard of 0.15 mg/kg and the rice samples collected from household storage contained four times the amount of the permissible Cd level of 0.1 mg/kg [28].An independent health survey revealed the prevalence of low eGFR among those resided in the Cd-contaminated area of the Mae Sot District to be 16.1% [29].
The study protocol for the Mae Sot group was approved by the Institutional Ethical Committees of Chiang Mai University and the Mae Sot Hospital (Approval No. 142/2544, 5 October 2001) [24].The Office of the Human Research Ethics Committee, Walailak University of Thailand, approved the study protocol for the Nakhon Si Thammarat group (Approval number WUEC-20-132-01, 28 May 2020) [25].
Prior to participation, all participants provided informed consent.They had lived at their current addresses for at least 30 years.Exclusion criteria were pregnancy, breastfeeding, a history of metal work and a hospital record or physician's diagnosis of an advanced chronic disease.Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg [30], a physician's diagnosis or prescription of antihypertensive medications.Diabetes was diagnosed when fasting plasma glucose levels ≥ 126 mg/dL (https://www.cdc.gov/diabetes/basics/getting-tested.html)(accessed on 12 August 2023) or a physician's prescription of anti-diabetic medications.

Assessment of Cadmium Exposure and Adverse Effects
Assessment of Cd exposure was based on a one-time measurement of blood Cd concentration ([Cd] b ) and the urinary excretion of Cd (E Cd ).Kidney functional assessment was based on E alb and eGFR.For these measurements, samples of urine and whole blood were collected after overnight fast.Blood samples were collected within 3 h of urine collection.Aliquots of blood and urine samples were stored at −80 • C for later analysis.
Levels of Cd in urine and blood ([Cd] u and [Cd] b ) were quantified by graphite furnace atomic absorption spectrometry with the Zeeman effect background correction system.Multielement standards (Merck KGaA, Darmstadt, Germany) were used for instrument calibration.The quality control and quality assurance of Cd quantitation were accomplished by simultaneous analysis of blood control samples (ClinChek, Munich, Germany) and the reference urine metal control levels 1, 2 and 3 (Lyphocheck, Bio-Rad, Hercules, CA, USA).
The limit of detection (LOD) for Cd in blood or urine, defined as 3 times the standard deviation of at least 10 blank sample measurements, was 0.3 µg/L for [Cd] b and 0.1 µg/L for [Cd] u .The sample blanks as reference standards for urine and blood were included in the assay together with samples of urine or blood from subjects.Deionized water was used to zero an instrument.The coefficient of variation in Cd in the reference urine and blood were within acceptable clinical chemistry standards.When a sample contained Cd below its LOD, the Cd concentration assigned was the LOD value divided by the square root of 2 [31].
Urinary and plasma creatinine concentrations ([cr] u and [cr] p ) were measured by the colorimetric method, based on the alkaline-picrate Jaffe's reaction [32].Urinary albumin concentration was determined by an immunoturbidimetric method [33,34].

Estimated Glomerular Filtration Rate (eGFR)
In theory, GFR is a measure of nephron function which is the product of the functioning nephron number and mean single nephron GFR [35][36][37].In practice, GFR is estimated from established chronic kidney disease epidemiology collaboration (CKD-EPI) equations and is reported as estimated GFR (eGFR) [34].These CKD-EPI equations have been validated with inulin clearance [38].
Male [cr] u = urine creatinine concentration (mg/dL).E x /C cr was expressed as the excretion of x per volume of filtrate [23].
Results obtained with C cr -normalized data are shown herein.Results of analogous analyses with E cr -normalized data are provided in Supplementary Material (SM).

Statistical Analysis
Data were analysed with IBM SPSS Statistics 21 (IBM Inc., New York, NY, USA).The Mann-Whitney U test was used to compare two groups and Pearson's chi-squared test was used to assess differences in percentages.Distribution of the variables was examined for skewness and those showing right skewing were subjected to logarithmic transformation before analysis, where required.Departure from normal distribution of variables was assessed by one sample Kolmogorov-Smirnov test.
A multiple linear regression model analysis was used to identify predictors of eGFR and E alb /C cr .The multivariable logistic regression analysis was used to determine the prevalence odds ratio (POR) for low eGFR and albuminuria.Univariate analysis of covariance via Bonferroni correction in multiple comparisons was used to obtain covariate adjusted mean E alb /C cr and eta square (η 2 ) values.For all tests, p-values ≤ 0.05 were considered to indicate statistical significance.

Cohort Participants
Table 1 provides descriptive characteristics of the Thai cohort participants, who were recruited from a low-exposure locality and a high-exposure region due to environmental Cd pollution.
A total of 482 persons (354 women and 118 men), mean age of 51.8 years, were included in this investigation.The respective percentages of smoking and diabetes were 29.7% and 18.3%, while about half (48.5%) had hypertension.
The overall % low eGFR was 8.1% and % abnormal ACR was 15%.The overall % abnormal E alb /C cr was 16.9%.Of note, the % abnormal E alb was higher in the low eGFR group than normal eGFR group for both ACR and E alb /C cr criteria in both genders.Similarly, the % hypertension and diabetes were higher in the low eGFR than the normal eGFR group.

Predictors of eGFR
Table 2 provides the results of the multiple linear regression analysis to define the independent variables that contributed to differences in eGFR.
In an inclusive model, all eight independent variables explained 27.8% of the total variation in eGFR.Age, E Cd /C cr and diabetes were the three main influential independent variables.All other five independent variables did not show a significant association with eGFR.In a subgroup analysis, female eGFR was inversely associated with age (β = −0.511),E Cd /C cr (β = −0.126)and diabetes (β = −0.119).In comparison, male eGFR was associated with age only (β = −0.472).
In an analogous regression of eGFR with E Cd normalized to creatinine excretion (E Cd /E cr ) (Table S1), eGFR was inversely associated with age only (β = −0.442).No significant association was seen between eGFR and all other six variables, including E Cd /E cr .
To investigate the dose-response relationship of Cd exposure and changes in eGFR, we designated the low, middle and high E Cd /C cr tertiles to represent low, medium and high Cd burdens, respectively.In women, the cut-off values of (E Cd /C cr ) × 100 for the low, medium and high Cd burdens were ≤1.44, 1.45-3.26,>3.26 µg/L filtrate, respectively.Corresponding cut-off values of (E Cd /C cr ) × 100 in men were ≤1.25, 1.26-3.25,>3.25 µg/L filtrate.
Figure 1 provides scatterplots relating eGFR to Cd excretion rate in women and men and their subgroups according to Cd burden levels, described above.To investigate the dose-response relationship of Cd exposure and changes in eGFR, we designated the low, middle and high ECd/Ccr tertiles to represent low, medium and high Cd burdens, respectively.In women, the cut-off values of (ECd/Ccr) × 100 for the low, medium and high Cd burdens were ≤1.44, 1.45-3.26,>3.26 µg/L filtrate, respectively.Corresponding cut-off values of (ECd/Ccr) × 100 in men were ≤1.25, 1.26-3.25,>3.25 µg/L filtrate.
Figure 1 provides scatterplots relating eGFR to Cd excretion rate in women and men and their subgroups according to Cd burden levels, described above.In scatterplots including all women (Figure 1a) and all men (Figure 1b), a direct relationship between eGFR and E Cd /C cr was suggested.In subgroup analysis, a direct relationship of eGFR and E C d/C cr was seen in women and men who had low Cd burden (Figure 1c,d).In contrast, eGFR was inversely related to E Cd /C cr in women and men of the medium plus high Cd burden groups (Figure 1e,f).These data indicate that an E Cd /C cr of 1.44 µg/L filtrate may be a threshold level of Cd accumulation, above which eGFR began to drop.
To confirm or dispute the contrasting eGFR responses below or above E Cd /C cr of 1.44 µg/L filtrate, multiple regression analyses of eGFR across Cd burden groups were undertaken.Results are provided in Table 3.In distinction from a bivariate analysis (Figure 1c,d), the association between eGFR and E Cd /C cr was statistically insignificant in the low Cd burden group after adjustment for confounding variables (Table 3).In this low Cd burden group, eGFR was inversely associated with age only (β = −0.455),while its association with E Cd /C cr became statistically insignificant (Table 3).
Of interest, there was an inverse relationship between eGFR and E Cd /C cr in the medium and high Cd burden groups following a similar adjustment for confounders.All eight independent variables explained, respectively, 32.1% and 25.5% of total eGFR variation in the medium and high Cd burden groups.In the high Cd burden group, eGFR was more closely associated with E Cd /C cr (β = −0.349),compared with the medium Cd burden group (β = −0.284).Other predictors of eGFR in the medium Cd burden group were age (β = −0.505)and systolic blood pressure (β = −0.230).Age was the other predictor of eGFR (β = −0.410) in the high Cd burden group.

Logistic Regression Analysis of Low eGFR
Table 4 provides the results of the logistic regression analysis that evaluated the effects of Cd body burden and other six independent variables on the prevalence odds ratios (POR) for low eGFR.POR is defined in Section 2. 5.
In all subjects, the POR values for low eGFR increased with age (POR 1.118) and in those with diabetes (POR 3.024), medium Cd burden (POR 8.265) and high Cd burden (POR 3.643).All other four independent variables did not significantly affect the POR for low eGFR.Among women, the POR values for low eGFR increased with age (POR 1.114) and in the medium Cd burden group (POR 7.204).An increase in the POR for low eGFR in the high Cd burden group did not reach a statistically significant level (POR 3.218, p = 0.064).
An effect of Cd exposure on the POR for low eGFR in men could not be evaluated due a small number of men with low eGFR (n = 4).However, the POR for low eGFR among men was associated with hypertension (POR 2.478).

Multiple Regression Analysis of Albumin Excretion Rate
We used multiple regression analysis to identify if E Cd /C cr and other independent variables predicted rising E alb /C cr .We included eGFR in regression models to evaluate if functioning nephrons, indicated by eGFR values, had an independent effect on E alb /C cr .Results are provided in Table 5.
In the low and medium Cd burden groups, neither E Cd /C cr nor eGFR showed a significant association with E alb /C cr .However, diabetes and hypertension predicted E alb /C cr in the low Cd burden group, while BMI was a sole predictor of E alb /C cr in the medium Cd burden group.
In the high Cd burden group, E Cd /C cr and eGFR both were associated with E alb /C cr .There was an inverse association between eGFR and E alb /C cr (β = −0.214),while E Cd /C cr showed a direct relationship with E alb /C cr (β = 0.173).All other six independent variables did not have a significant effect on E alb /C cr .
Of note, the relationship between albumin excretion and eGFR was obscure, when E alb was normalized to E cr (E alb /E cr ).As data in Table S2 indicate, E alb /E cr did not show a significant inverse association with eGFR (β = −0.092,p = 0.078).
We next conducted another multiple regression analysis to evaluate associations of E Cd /C cr .eGFR and other independent variables on E alb /C cr in participants with normal kidney function (eGFR > 60 mL/min/1.73m 2 ) and those with a sign of CKD (eGFR ≤ 60 mL/min/1.73m 2 ).Results are provided in Table 6.
E alb /C cr was inversely associated with eGFR (β = −0.188)when all subjects were included in an analysis.In a subgroup analysis, no association was observed between E alb /C cr and eGFR in those with normal kidney function; there was a very strong inverse association of these two variables in those with CKD (β = −0.637).
In a subsequent analysis, scatterplots were used together with a univariate analysis to quantify the independent effects of E Cd /C cr , eGFR, diabetes and hypertension on E alb /C cr .Covariate adjusted mean E alb /C cr values were obtained for various subgroups, as shown in Figure 2.  cluded in an analysis.In a subgroup analysis, no association was observed between Ealb/Ccr and eGFR in those with normal kidney function; there was a very strong inverse association of these two variables in those with CKD (β = −0.637).
In a subsequent analysis, scatterplots were used together with a univariate analysis to quantify the independent effects of ECd/Ccr, eGFR, diabetes and hypertension on Ealb/Ccr.Covariate adjusted mean Ealb/Ccr values were obtained for various subgroups, as shown in Figure 2. All means were obtained via univariate covariance analysis with adjustment for covariates (age and BMI) and (smoking × hypertension × diabetes) interactions.
In the (E Cd /C cr ) ×100 < 1 µg/L filtrate group (Figure 2a), a scatterplot indicated that the correlation of E alb /C cr and E Cd /C cr was statistically insignificant.However, in an analysis restricted to those with normal eGFR, differences in E alb /C cr were observed in smokers and those with diabetes after adjustment for covariates and interactions (Figure 2b).
In the (E Cd /C cr ) × 100 ≥ 1 µg/L filtrate group (Figure 2c), a scatterplot indicated a statistically significant relationship between E alb /C cr and E Cd /C cr .An increment of E alb /C cr was observed in smokers in an analysis restricted to those with normal eGFR after adjustment for covariates and interactions (Figure 2d).

Blood Cadmium and eGFR as Predictors of Albuminuria
Table 7 provides the results of the logistic regression analysis that evaluated if [Cd] b , like E Cd , predicted an increase in risk of abnormally high E alb, defined as (E alb /C cr ) × 100 ≥ 20 µg/L filtrate.In all subjects, eGFR, diabetes, hypertension and [Cd] b were independent variables showing associations with the POR for albuminuria.Comparing those with [Cd] b < 0.82 µg/L, POR values for albuminuria were increased by 2.4-fold and 2.7-fold in those with [Cd] b of 0.83-2.63 and ≥2.64 µg/L, respectively.For every 1 mL/min/1.73m 2 loss of eGFR, the POR for albuminuria rose by 4.3%.There was no effect of Cd on albuminuria in an analysis with Ecr-normalized data (Table S3).
Among women, there was a [Cd] b -related increment of the POR for albuminuria.The POR values for high albuminuria were increased 3.4-fold and 3.8-fold in those with middle and high [Cd] b tertiles, respectively.In addition, eGFR, diabetes and hypertension were independent variables showing associations with the POR for albumin in women.For every 1 mL/min/1.73m 2 loss of eGFR, the POR for albuminuria rose by 4.5%.
Among men, [Cd] b did not show an association with POR for albuminuria, while eGFR did.For every 1 mL/min/1.73m 2 loss of eGFR, the POR for albuminuria rose by 4.5%, similar to the effect size found in women.
In a subsequent analysis, covariate adjusted mean E alb /C cr values were obtained for groups of participants with different eGFR levels (≤60, 61-90 and >90 mL/min/1.73m 2 ) to reveal an effect of GFR loss, reflected by a reduction in eGFR.Results are shown in Figure 3.
In the low [Cd] b tertile group, the relationship between E alb /C cr and eGFR was statistically insignificant (Figure 3a).Similarly, the covariate adjusted mean E alb /C cr values did not show a significant variation across three eGFR subgroups.
In the middle and high [Cd] b tertile groups, an inverse relationship was observed between E alb /C cr and eGFR (Figure 3c,e); an effect of GFR loss on E alb /C cr was apparent (Figure 3d,f).In these two [Cd] b tertile groups, the mean E alb /C cr value was highest, middle and lowest in those with eGFR ≤ 60, 60-90 and > 90 mL/min/1.73m 2 , respectively.Among men, [Cd]b did not show an association with POR for albuminuria, while eGFR did.For every 1 mL/min/1.73m 2 loss of eGFR, the POR for albuminuria rose by 4.5%, similar to the effect size found in women.
In a subsequent analysis, covariate adjusted mean Ealb/Ccr values were obtained for groups of participants with different eGFR levels (≤60, 61-90 and >90 mL/min/1.73m 2 ) to reveal an effect of GFR loss, reflected by a reduction in eGFR.Results are shown in Figure 3.

Discussion
Herein, we used the eGFR decline as a primary endpoint in our attempt to define a toxic accumulation level of Cd in the kidney.We focused on this eGFR because attenuation of eGFR decline is employed in clinical trials to evaluate the effects of treatment of CKD [1,2].Albuminuria was concurrently examined because it has often been associated with Cd exposure, along with low eGFR [4].Previous studies show that the burden of Cd as µg/g kidney tissue weight increases with age [39][40][41].They show also that sufficient tubular damage from the Cd accumulation reduces the GFR, leading to nephron atrophy and a decreasing GFR as a result [20,22,42,43].We hypothesized that this eGFR effect of Cd occurs at a very low kidney burden, producing E Cd /E cr below 5.24 µg/g creatinine, which is the current World Health Organization health-based exposure limit, detailed in Section 4.3.It is noteworthy that E Cd itself is indicative of tubular cell injury and death [3,20] and that E Cd /E cr of 5.24 µg/g creatinine signifies extensive tubular cell damage and death.
We analysed data from 482 Thai nationals, of which 8.1% had low eGFR and 15% had albuminuria (ACR criterion).Nearly half (48.3%) had hypertension, while 29.7% reported to be smokers and 18.3% had diabetes.There was a 1250-fold difference in E Cd /C cr values, ranging from 0.0002 to 0.25 µg/L filtrate (mean 0.0032 µg/L filtrate).There was a 667-fold difference in [Cd] b , ranging between 0.03 and 20 µg/L (mean 2.60 µg/L).The wide ranges of these Cd burden and exposure matrices together with sufficiently high numbers of smokers and those with diabetes and hypertension provide an ideal scenario to define with certainty a toxic threshold level of Cd.
Like E Cd /C cr , E Cd /E cr showed a large variation (1033-fold) and the mean E Cd /E cr (range) was 4.05 (0.03-31) µg/g creatinine (Table 1).However, in distinction from E Cd /C cr , there was no significant association between eGFR and E Cd /E cr (Table S1).Consequently, a dose-response analysis was not possibly determined.Also, when E alb was adjusted to creatinine excretion (E cr ) as E alb /E cr or ACR, an inverse relationship of ACR and eGFR was obscure (Table S2) and there was no effect of Cd on POR for albuminuria (Table S3).

Effects of Cadmium on eGFR
A direct relationship between eGFR and E Cd /C cr was seen in women and men with a low-Cd burden (E Cd /C cr ≤ 0.014 µg/L filtrate in women and ≤0.012 µg/L filtrate in men) (Figure 1c,d).However, such eGFR/E Cd /C cr relationship was weak and became statistically insignificant when covariates were adjusted.In comparison, eGFR showed an inverse association with E Cd /C cr when E Cd /C cr values rose to levels > 0.014 µg/L filtrate in women and >0.012 µg/L filtrate in men (Figure 1e,f and Table 3).In reflecting a dose-response relationship, eGFR was more closely associated with E Cd /C cr in the high burden group (β = −0.349)compared to the medium Cd burden group (β = −0.284).An association of eGFR and E Cd /C cr was statistically insignificant in the low Cd burden group.
Cd burden, diabetes and age were three of seven variables that affected the POR for low eGFR (Table 4).In those with medium, high Cd burden and diabetes, the POR for low eGFR rose by 8.3-fold, 3.6-fold and 3.0-fold, respectively.Per every one year rise in age, there was an 11.8% (95% CI, 6.2, 17.6) increase in the POR for low eGFR.
In summary, no effect of Cd on eGFR was found in those with E Cd /C cr values < 0.012 µg/L filtrate.This E Cd /C cr value of 0.012 µg/L filtrate is in ranges with the no-observed-adverseeffect level (NOAEL) equivalent of E Cd /C cr at 0.010-0.024µg/L filtrate, obtained by the benchmark dose method [44].

Effects of Cadmium on Prevalence of Albuminuria
Most albumin (~80%) in the glomerular ultrafiltrate is reabsorbed in the S1 subsegment of the proximal tubule, where the receptor-mediated endocytosis involving the megalin/cubillin system is located [45][46][47].Reabsorption of albumin also occurs in the distal tubule and collecting duct, where the process is mediated by the NGAL/lipocalin-2 receptor system [48][49][50].Experimental studies show that albuminuria resulted from Cd selectively disabled albumin endocytosis which is mediated by the cubilin/megalin recep-tor system [51] and that Cd may diminish expression of megalin and ClC5 channels [52].However, data in Table 1 indicate that albuminuria was more prevalent in the low eGFR compared to normal eGFR groups, thereby suggesting that low eGFR could be a risk factor for albuminuria.We examined this phenomenon further, as described below.
In multiple regressions (Table 5), both eGFR and E Cd /C cr appeared to have an effect on E alb /C cr .In logistic regression (Table 6), the POR for albuminuria rose by 4.3% (95% CI: 2.6, 6.1) for every 1 mL/min/1.73m 2 loss of eGFR and it was increased by 2.4-fold to 2.7-fold in those who had [Cd] b ≥ 0.83 µg/L.A similar effect size of Cd on eGFR was seen in women and men, although the causes of albuminuria differed (Table 6).Additional evidence that Cd may affect eGFR and albumin reuptake simultaneously comes from a covariance analysis, where 9.3% to 13.2% of E alb /C cr variation could be attributable to eGFR levels in those with [Cd] b > 0.83 µg/L (Figure 3d,f).eGFR levels did not explain the variability of E alb /C cr in the low [Cd] b tertile group (F = 0.674, η 2 0.010, p = 0.511) (Figure 3b).
The results described above could be interpreted to suggest that albuminuria could be a consequence of the Cd-induced destruction of nephrons, together with a direct effect on the tubular reabsorption of albumin, possibly through the cubilin/megalin receptor system of endocytosis.These Cd effects were not found in those who had [Cd] b < 0.83 µg/L.

A Independent Effect of Smoking on Albuminuria
Smoking forms a significant Cd exposure source as cigarette smoke contains Cd in volatile metallic and oxide (CdO) forms, which have transmission rates 5 to 10 times higher than those that enter through the gut [53][54][55].Adverse kidney outcomes associated with smoking have been noted in the general populations [56,57].In the present study, evidence for an independent effect of smoking has emerged from analysis of covariance of albumin excretion data from those with normal kidney function.In Figure 2d, the mean E alb /C cr was higher in smokers compared to non-smokers of the same age, BMI and overall Cd burden [(E Cd /C cr ) × 100 ≥ 1 µg/L filtrate].This finding suggested that a possible Cd independent effect of smoking on albuminuria.Other constitutes of cigarette smoke are likely to have a role.In a Dutch prospective study involving 231 diabetic patients, Cd and active smoking were both found to be associated with progressive eGFR decline [19].
Of further note, the opposite effect was observed in the low Cd burden group [(E Cd / C cr ) × 100 < 1 µg/L filtrate] (Figure 2b), where the covariate adjusted mean E alb /C cr was lower in smokers than nonsmokers of the same low Cd burden.Cd-induced albuminuria may be mitigated in smokers by substances that accompany Cd.Tin, for example, has blood pressure-lowering effects [58,59].However, in our study, there was a high degree of statistical uncertainty due to a small number of smokers (n = 14) who had a low Cd burden (Figure 2b); hence, a further study is required to fully explore this.
An effect of diabetes on albumin excretion was also demonstratable in the present study; this effect was evident only in the low Cd burden group (Figure 2b).It is possible that we have underestimated the amount of albumin in urine samples from subjects with diabetes and high Cd burden (Figure 2d).The immunoturbidimetric method used has been reported to be unable to detect albumin molecules with altered conformation, which may be produced in people with CKD and diabetes [34].

Threshold-Based Exposure Guidelines
For most people, exposure to Cd is inevitable because the metal is present in nearly all food types.In Japan, for example, rice and its products, green vegetables and cereals and seeds plus potatoes constituted, respectively, to 38%, 17% and 11% of total dietary Cd exposure [60].In response, health guidance such as a tolerable intake level of Cd and a reference dose (RfD) were determined [61][62][63][64].The Joint FAO/WHO Expert Committee on Food Additives and Contaminants (JECFA) suggested a tolerable monthly intake (TMI) of Cd to be 25 µg per kg body weight per month, equivalent to 0.83 µg per kg body weight per day (58 µg/day for a 70 kg person), and E Cd /E cr of 5.24 µg/g creatinine was adopted as a nephrotoxicity threshold value [63].Both figures were based on a risk assessment model based solely on a β 2 -microglobulin excretion rate (E β2M /E cr ) ≥ 300 µg/g creatinine, termed tubular proteinuria, as a toxic endpoint.
The European Food Safety Authority (EFSA) used also the β 2 M endpoint but with the inclusion of an uncertainty factor (safety margin).The EFSA suggested E Cd /E cr of 1 µg/g creatinine to be the toxicity threshold and 0.36 µg/kg body weight per day (25 µg/day for a 70 kg person) as the RfD [64].The higher JECFA guidelines are adopted by most countries.These health guideline values exceed the nephrotoxicity threshold identified in the present study.In theory, exposure guidelines that provide sufficient health protection should only be determined from the most sensitive endpoint with consideration given to subpopulations with increased susceptibility [54] and the most recent scientific knowledge should be considered.
Of concern, eGFR decline due to Cd nephropathy has increasingly been observed in both children and adult populations.Lower eGFR values were found to be associated with higher Cd excretion rates in studies from Guatemala [65] and Myanmar [66].In a prospective cohort study of Bangladeshi preschool children, an inverse relationship between E Cd and kidney volume was seen in children at 5 years of age.This was in addition to a decrease in eGFR [67].E Cd was inversely associated with eGFR, especially in girls.In another prospective cohort study, the reported mean for Cd intake among Mexican children was 4.4 µg/d at the baseline and rose to 8.1 µg/d after nine years, when such Cd intake levels showed a marginally inverse association with eGFR [68].

Conclusions
Based on a dose-response relationship analysis of data from 482 non-occupationally exposed persons with a 1250-fold difference in Cd burden and a 667-fold difference in levels of blood Cd, environmental exposure to Cd was confirmed to be closely associated with a declining GFR and albuminuria.An independent effect of smoking on albuminuria has also been observed in smokers who had normal kidney function.The current World Health Organization Cd exposure limit of 5.24 µg/g creatinine (E Cd /E cr ), which is solely based on an increment of excretion of β 2 -microglobulin above 300 µg/g creatinine, underestimates the level at which Cd induces kidney damage.Our results show that when a declining GFR is considered along with albuminuria, the no-observed-adverse-effect level (NOAEL) equivalent is 0.01-0.02µg/g creatinine.Now is the time to acknowledge there is no safe level of Cd exposure.

Figure 1 .
Figure 1.The relationship between eGFR and cadmium excretion rate.Scatterplots relate eGFR to log[(ECd/Ccr) × 10 5 ) in women (a) and men (b), women (c) and men (d) with low Cd burden and women (e) and men (f) with medium plus high Cd burden.Coefficients of determination (R 2 ) and

Figure 1 .
Figure 1.The relationship between eGFR and cadmium excretion rate.Scatterplots relate eGFR to log[(E Cd /C cr ) × 10 5 ) in women (a) and men (b), women (c) and men (d) with low Cd burden and women (e) and men (f) with medium plus high Cd burden.Coefficients of determination (R 2 ) and p-values are provided for all scatterplots together with numbers of participants in all subgroups.The cut-off values of (E Cd /C cr ) × 100 for low, medium and high Cd burden in women were ≤1.44, 1.45-3.26,>3.26 µg/L filtrate, respectively.Corresponding cut-off values of (E Cd /C cr ) × 100 in men were ≤1.25, 1.26-3.25,>3.25 µg/L filtrate.

Figure 3 .
Figure 3. Albumin excretion rates in participants grouped by blood cadmium and eGFR levels.Scatterplots relate log[(Ealb/Ccr) × 10 4 ] to eGFR in participants with low (a), middle (c) and high [Cd]b Figure Albumin excretion rates in participants grouped by blood cadmium and eGFR levels.Scatterplots relate log[(E alb /C cr ) × 10 4 ] to eGFR in participants with low (a), middle (c) and high [Cd] b tertiles (e).Coefficients of determination (R 2 ), p-values and numbers of subjects are provided for all scatterplots.Bar graphs depict mean log[(E alb /C cr ) ×10 4 ] in participants of low (b), middle (d) and high [Cd] b tertiles (f) who had eGFR ≤ 60, 61-90 and > 90 mL/min/1.73m 2 .All means were obtained via univariate covariance analysis with adjustment for covariates and interactions.The cut-off values of [Cd] b for low, middle and high tertiles were < 0.82, 0.83-2.63 and ≥ 2.64 µg/L, respectively.Arithmetic means (SD) of [Cd] b in low, middle and high tertiles were 0.31 (0.29), 1.86 (1.98) and 5.58 (3.28), respectively.For all tests, p-values ≤ 0.05 indicate statistically significant levels.

Table 1 .
Characteristics of participants according to eGFR levels and gender.

Table 2 .
Predictors of eGFR in women and men.

Table 3 .
Comparing strength of association of eGFR with cadmium excretion rate across cadmium burden groups.
eGFR, estimated glomerular filtration rate; β, standardized regression coefficient; adjusted R 2 , coefficient of determination.Coding: male, 1, female, 2; non-smoker, 1, smoker, 2; non-diabetic, 1, diabetic, 2. β indicates strength of association of eGFR with eight independent variables (first column).Adjusted R 2 indicates the proportion of eGFR variance explained by all independent variables.The cut-off values of (E Cd /C cr ) × 100 for the low, medium and high Cd burdens in women were ≤1.44, 1.45-3.26,>3.26 µg/L filtrate, respectively.Corresponding cut-off values of (E Cd /C cr ) × 100 in men were ≤1.25, 1.26-3.25,>3.25 µg/L filtrate.The numbers of subjects in low, moderate and high Cd burdens were 147, 148 and 147, respectively.For all tests, p-values ≤ 0.05 indicate a statistically significant effect of individual independent variables on eGFR.

Table 4 .
44evalence odds ratios for low eGFR in relation to cadmium body burden and other independent variables.Effect of hypertension was assessed by Fisher's exact test using data from 118 men.The cut-off values of (E Cd /C cr ) × 100 for the low, medium and high Cd burdens in women were ≤1.44, 1.45-3.26,>3.26 µg/L filtrate, respectively.Corresponding cut-off values of (ECd/Ccr) × 100 in men were ≤1.25, 1.26-3.25,>3.25 µg/L filtrate.For all tests, p-values ≤ 0.05 indicate a statistically significant effect of individual independent variables to the POR for low eGFR.

Table 5 .
Comparing strength of association of albumin excretion rate with cadmium excretion rate across cadmium burden groups.
eGFR, estimated glomerular filtration rate; β, standardized regression coefficient; adjusted R 2 , coefficient of determination.Coding: male, 1, female, 2; non-smoker, 1, smoker, 2; non-diabetic, 1, diabetic, 2. β indicates strength of association of eGFR with eight independent variables (first column).Adjusted R 2 indicates the proportion of eGFR variance explained by all independent variables.The cut-off values of (E Cd /C cr ) × 100 for the low, medium and high Cd burdens in women were ≤1.44, 1.45-3.26,>3.26 µg/L filtrate, respectively.Corresponding cut-off values of (E Cd /C cr ) × 100 in men were ≤1.25, 1.26-3.25,>3.25 µg/L filtrate.The numbers of subjects in low, medium and high Cd burdens were 147, 148 and 149, respectively.For all tests, p-values ≤ 0.05 indicate a statistically significant effect of individual independent variables on eGFR.

Table 6 .
Comparing strength of association of albumin excretion rate with cadmium excretion rate and other variables in participants grouped by eGFR.× 10 4 ] with eight independent variables (first column).Adjusted R 2 indicates the proportion of log[(E alb /C cr ) × 10 4 ] variance explained by all independent variables.Normal and low eGFR groups were defined as eGFR > 60 and ≤60 mL/min/1.73m 2 , respectively.Respective numbers of subjects with normal and low eGFR were 416 and 31.For all tests, p-values ≤ 0.05 indicate a statistically significant effect of individual independent variables on albumin excretion rate.

Table 7 .
Prevalence odds ratios for albuminuria in relation to blood Cd levels and other independent variables., prevalence odds ratio; CI, confidence interval; BMI, body mass index; eGFR, estimated glomerular filtration rate; [Cd] b , blood Cd concentration.Coding, male 1, female 2; non-smoker 1, smoker 2; non-diabetic 1, diabetic 2; code 1 is referent.Data were generated from logistic regression relating POR for albuminuria to seven independent variables (first column).Arithmetic means (SD) of blood Cd in low, middle and high tertiles were 0.31 (0.29), 1.86 (1.98), 5.58(3.28),respectively.Corresponding numbers of subjects were 157, 164 and 161, respectively.For all tests, p-values ≤ 0.05 indicate a statistically significant effect of individual independent variables to the POR for albuminuria. POR