Tubular and Glomerular Kidney Effects in Swedish Women with Low Environmental Cadmium Exposure

1Identification of risk factors for chronic renal failure is essential in order to prevent reduction of life quality and life expectancy and to minimize the high costs of treatment. Cadmium is a widespread environmental pollutant known to cause renal damage. Apart from smoking, the major sources of cadmium exposure in the general population are cereals, vegetables, and shellfish. There is increasing evidence that toxic effects may occur at much lower exposure levels than those observed in occupational settings or in severely polluted environments. Still, the attempts to estimate the level of critical exposure for kidney effects have so far displayed large variations. Furthermore, possible effects in populations residing in areas with no particular industrial cadmium emission are undetermined.

2Cadmium accumulates in the renal cortex and induces tubular toxicity, which is first detected as increased urinary excretion of low-molecular-weight proteins and tubular enzymes. Glomerular dysfunction may also emerge, as demonstrated in heavily exposed subjects. It is not known, however, whether the glomerulus is affected by long-term low-level environmental exposure. Diabetes, an increasing health problem in many areas and one of the leading causes of incident end-stage renal disease, has been suggested to augment the risk of cadmium-induced kidney damage. Also, hypertension and intake of nephrotoxic nonsteroid anti-inflammatory drugs (NSAIDs) might interact with cadmium. However, these possible interactions need to be confirmed.

3The aim of the present investigation was to assess the association between cadmium concentrations in blood and urine and a series of markers of tubular and glomerular function. To minimize dilution of the effects, we focused on women at the age when the accumulation of cadmium in the kidney is at its maximum. Women have increased cadmium accumulation compared with men, due to a higher dietary cadmium absorption at low body iron stores. In addition, we assessed whether diabetes, hypertension, and the use of NSAIDs increased the risk. The study was conducted in an area without particular industrial cadmium emission.

4The Women’s Health in the Lund Area (WHILA) study, a population-based study of all women 50–59 years of age in the community of Lund, southern Sweden (n = 10,766), started in December 1995 and was extended in June 1999 to include health aspects of cadmium. This cohort was considered optimal for elucidation of remaining questions about dose–response relationships at low-dose cadmium exposure. The participation rate was 71% (n = 820). The exclusion criteria were renal cancer (n = 1) and lithium treatment (n = 3). Data were collected on various comorbidities, including diabetes and hypertension. Women were classed as having diabetes if they had a positive history, or if they had a non-fasting glucose > 8 mmol/L followed by a positive result in the oral glucose-tolerance test. Women were classed as hypertensive if they had received antihypertensive treatment or had a measured systolic and/or diastolic blood pressure ≥160 and ≥95, respectively (mean of two measurements after 15 min rest in seated position). Lists of medications and data on smoking were obtained, and weight and height were measured. Participants were asked to submit morning first-voided urine and blood samples. We obtained morning spot urine from 813 women and blood samples from 742. All samples were collected during 8 months from June 1999 through January 2000. The ethics committee at Lund University approved the WHILA study, and oral informed consent was obtained.

5Measurements included cadmium in blood as a measure mainly of ongoing exposure (expected to be fairly constant over time) and urine as a measure of body burden (cadmium in urine correlates well with cadmium in the kidney cortex;; ). To control for possible confounding/effect modification, we also determined lead in blood. We used the following effect markers: cystatin C in serum for calculation of glomerular filtration rate (GFR), creatinine clearance as markers of glomerular function, and human complex-forming protein (protein HC, α1-microglobulin), N-acetyl-β-d-glucosaminidase (NAG), and calcium in urine as markers of tubular damage.

6We measured cadmium, lead, and calcium using inductively coupled plasma mass spectrometry. Cystatin C was determined by immunonephelometry (Dade Behring, Marburg, Germany). GFR = 77.24 × (cystatin C)−1.2623 and creatinine clearance = [(140 – age) × body weight (kg)]/[0.85 × serum creatinine (μM)]. Creatinine was measured using a modified kinetic Jaffé method (Roche Diagnostics, Mannheim, Germany). We determined urinary protein HC by Mancini technique and polyclonal antibodies (DAKO A/S, Glostrup, Denmark), and urinary NAG with a colorimetric method (Roche, Shionogi & Co. Ltd., Osaka, Japan).

7Urinary spot samples need to be adjusted for dilution. Creatinine adjustment is most common, but a comparison of density and creatinine adjusted urinary cadmium indicated that creatinine did not adjust for all dilution-related variation of cadmium in urine. Because creatinine excretion is dependent upon meat intake and muscle mass, we chose to correct all urinary markers by the mean urinary density (1.015 g/mL) according to [urinary cadmium × (1.015 × 1,000) –1,000]/[(urinary density × 1,000) – 1,000]. However, creatinine-adjusted values are given for comparison.

8All the equipment was tested, and possible contamination was below the limit of detection. For cadmium and lead in blood and cadmium and calcium in urine, the limits of detection were 0.12 μg/L, 0.26 μg/L, 0.31 μg/L, and 1.6 mg/L, respectively. For results below the limit of detection (mainly urinary cadmium), the concentration was set as the value factually obtained in the analysis. The imprecision of the method, calculated as the coefficient of variation for duplicate measurements, was 7.4 and 3.1% for cadmium and lead in blood and 8.5 and 6.4% for cadmium and calcium in urine. The analytical accuracy for blood (Seronorm, batch 404107; Nycomed, Oslo, Norway) was as follows (mean ± SD): 0.67 ± 0.08 for cadmium and 29 ± 1.1 for lead (n = 21; recommended, 0.67–0.70 and 31–39 μg/L, respectively). The data for certified blood samples from the U.K. National External Quality Assessment Service (n = 11) deviated on average by ± 7.9% for target values of 1.8–8.9 μg cadmium/L and ± 6.1% for 52–352 μg lead/L. The results for urine (Seronorm, batch 102021) were 140 ± 8.1 mg/L (n = 20; recommended 130 mg/L) for calcium and 0.45 ± 0.07 μg/L (n = 20; recommended, 0.35 μg/L) for cadmium. The result for the certified urine samples from Centre de Toxicologie du Quebec Interlaboratory Comparison Program for cadmium was 0.76 ± 0.09 and 3.6 ± 0.22 μg/L (n = 11; certified 0.79 and 3.6 μg/L), respectively.

9The imprecision was 2.7% for cystatin C (n = 6), 16% for protein HC (n = 10; limit of detection = 0.7 mg/L), and < 10% for urinary NAG (n = 68).

10We used Spearman’s rank correlation analysis to assess univariate associations. The cadmium-associated kidney effect markers were further evaluated in multiple linear regression models, where each kidney effect marker was evaluated in relation to cadmium and confounders/covariates. The dependent effect markers (continuous) were not dichotomized. No log-transformation was needed as indicated by residual and goodness-of-fit analyses. We evaluated possible effect modification (interactions) for cadmium and lead. Lowest observed effect levels were assessed for each effect marker for categorized urinary cadmium using Dunnett’s post hoc test, including significant confounders and covariates in the models. All tests were two sided, and statistics were performed using SPSS (version 12.01; SPSS Inc., Chicago, IL, USA).

11The study population characteristics, exposure variables, and kidney effect markers are presented in Table 1. The proportion of subjects with diabetes was slightly higher in the present study compared with those participating in the whole WHILA cohort (6.4%). The proportion of hypertensive subjects was, however, similar. Those who had ever smoked had 90% higher cadmium concentrations in blood and 40% higher in urine compared with never-smokers, who had 0.30 μg/L and 0.45 μg/L cadmium in blood and urine, respectively.

12The univariate associations between cadmium and kidney effect markers as well as those with possible confounders and effect modifiers are shown in Table 2. Cadmium in both blood and urine was associated with all kidney effect markers except serum creatinine and urinary calcium, which were not included in further analysis. Using cystatin C instead of estimated GFR, and creatinine-adjusted markers in urine instead of density adjusted ones had no major impact on the results.

13In the multiple linear regression analysis, each of the cadmium-associated kidney effect markers was tested separately, as were cadmium levels in urine and blood (Table 3). We included in the models the covariates age, body mass index (BMI), and blood lead, as well as the possible kidney-effect modifiers diabetes, hypertension, and use of NSAIDs. Never-smokers were analyzed separately. Cadmium in urine was significantly associated with GFR, creatinine clearance, protein HC, and NAG, after controlling for confounders and adjusting for other covariables (Table 3). Similar results were obtained for blood cadmium. In never-smokers, cadmium remained associated with protein HC and NAG and became significantly associated with creatinine clearance (blood cadmium). Lead was significantly associated with GFR and creatinine clearance.

14Because there were no associations between cadmium and GFR or creatinine clearance (for urinary cadmium) in never-smokers, we tested whether there was confounding through a non-cadmium-dependent effect of smoking by including pack-years in the multiple regression models. Smoking (pack-years) was not significantly associated with GFR or creatinine clearance (p ≥0.1).

15We assessed possible interactions between cadmium and blood lead, diabetes (insulin-treated vs. the rest), hypertension, or use of NSAIDs, and between blood lead and diabetes, hypertension, or use of NSAIDs. For NAG, there was an interaction between urinary cadmium and diabetes (insulin-treated vs. other diabetics and nondiabetics; regression coefficients (β): diabetics = 2.3, nondiabetics = 0.8; R2 = 0.10; p = 0.042) (Table 3). For protein HC, there was a close to significant interaction between urinary cadmium and diabetes (β: diabetics = 4.0, nondiabetics = 1.3; R2 = 0.09; p = 0.07), which became significant in never-smokers (β: diabetics = 27, nondiabetics = 1.9; R2 = 0.18; p < 0.001). Similar interactions were observed between blood cadmium and diabetes. Hypertension, NSAID use, and blood lead showed no significant interactions with cadmium exposure. However, there was an interaction between blood lead and diabetes for GFR in never-smokers (p = 0.005).

16Protein HC, NAG (diabetics excluded) (Figure 1A), and creatinine clearance (Figure 1B), after adjustment for blood lead, differed significantly between the group with lowest exposure level (urinary cadmium < 0.5 μg/L; mean, 0.36 μg cadmium/L = 0.48 μg cadmium/g creatinine) and that with the next lowest exposure level (0.50–0.75 μg/L; mean, 0.61 μg cadmium/L = 0.79 μg cadmium/g creatinine). For GFR, the group with the next highest exposure level [urinary cadmium, 0.75–1 μg cadmium/L; mean, 0.86 μg cadmium/L = 1.0 μg cadmium/g creatinine; adjusted for age, BMI, and blood lead (each categorized into four groups) and for NSAID use (into 0 or 1); Figure 1C] differed from the lowest level. For blood cadmium, associations were present in the exposure category 0.5–1 μg/L (mean, 0.69 μg/L) for protein HC (p = 0.036) and NAG (p = 0.024). For GFR, an association was seen only at blood cadmium > 1 μg/L (mean, 1.8 μg/L; p < 0.001) after adjustment for significant covariates.

17This population-based study of upper-middle-age women, representative of the general population of southern Sweden, showed clear associations between cadmium and the renal tubular-effect markers protein HC and NAG, even at the low levels of cadmium found in never-smokers. Cadmium potentiated the diabetes-induced effects on the kidney. There was also a clear association between cadmium and GFR or creatinine clearance.

18This study has several methodologic advantages, including the large sample size and high participation rate, individual exposure assessment with high analytical accuracy, and inclusion of several different outcomes of renal effects. Despite the low cadmium concentrations, we had a high analytical accuracy. Any imprecision would have caused a bias toward the null.

19The study population differed somewhat from the total WHILA population and Sweden (4–7%). Hence, there was a slight overrepresentation of diabetics, which may cause an overestimate of cadmium effects. However, because we controlled for diabetes in the statistical models, this is not a problem. Overcontrol and collinearity may occur in a statistical analysis such as that performed in this study. Smoking is then an obvious problem, which we handled by separate analysis in never-smokers. Lead and BMI were included, which means a risk of some overcontrol.

20Another problem, common in the interpretation of data from cross-sectional studies, is that the exposure is measured at the same time as the effects, which may not be the etiologically relevant period. This may be problematic for blood cadmium, because it largely reflects recent exposure, but not for urinary cadmium, which is a good estimate of the integrated low-level exposure over decades. It is known that kidney deterioration, due to both aging and high cadmium exposure, increases the excretion of cadmium in urine, resulting in lower kidney cadmium and eventually lower urinary cadmium. However, the present participants were below the age when the kidney cadmium starts to decrease, and the exposure was relatively low.

21The present cadmium concentrations are comparable with, or slightly higher than, those in other recent studies from Sweden and the United States but lower than those in more contaminated areas of Europe and much lower than in certain areas in Japan. Despite the present low cadmium levels, there were clear effects on the kidney. The associations between cadmium and biomarkers of several different renal effects support causality. It is unlikely that they are merely a result of parallel phenomena, impaired tubular reabsorption (protein HC), or increased general turnover of tubular cells (NAG). The associations with blood cadmium also preclude such an interpretation. Because smoking is a major source of cadmium exposure, the possibility of confounding through a non-cadmium-dependent effect of smoking must be considered. However, because we found cadmium-associated effects on NAG and protein HC even in never-smokers, and there was no effect of smoking on creatinine clearance or GFR, this is unlikely.

22The lead levels were low. The association between blood lead and GFR and creatinine clearance may indicate either an effect on GFR at low lead exposure or reverse causality. In the case of cadmium, reverse causality seems highly unlikely. Even though data may imply that a decrease in GFR causes increased blood cadmium concentrations, the inverse associations between the glomerular effect markers and cadmium in urine rather indicate that reduced GFR does not reduce the clearance of cadmium. In addition, lead is bound to high-molecular-weight plasma albumin, and cadmium to metallothionein, a small polypeptide that is easily filtered through the glomerulus.

23The lowest observed effect level, defined as the mean urinary cadmium in the exposure category that displayed significantly different levels of effect markers compared with the lowest urinary cadmium category, was 0.6 μg cadmium/L (0.8 μg/g creatinine), corresponding to approximately 20 μg cadmium/g kidney cortex. The lowest observed effect level is lower than in previous studies that observed effects at low-level cadmium exposure. This is probably due to homogeneity of the population, absence of healthy worker effects, and good precision in analyses of the exposure and effect markers.

24More important, we found a similar lowest observed effect level for creatinine clearance as for the tubular markers. Although usually considered an index of glomerular function, the creatinine clearance may partly reflect a proximal tubular dysfunction, because creatinine is not only filtrated but also secreted in the tubuli. On the other hand, the cadmium-associated increase in GFR, occurring in the next highest cadmium stratum (0.86 μg/L urine = 1.0 μg/g creatinine), clearly indicates an effect on the glomerular function. An observation in this contex that supports an effect of cadmium on GFR is the reported ecologic association between end-stage renal disease and distance to cadmium-emitting industrial plants.

25Cadmium has been suggested to cause hypertension, but no such effect was seen here, in agreement with other studies. Also, we did not observe any synergism between cadmium and hypertension on the kidney effects. However, there might be a dilution of the group by cases with mild hypertension. Further, it has been reported, both from experimental and epidemiologic studies, that cadmium increases the risk of type II diabetes, which was not supported by the present study. As expected, diabetes affected the kidney function, although only in insulin-dependent women, of whom about half had type II diabetes. More important, we found an interaction between cadmium and diabetes, as suggested in previous studies. Hence, the lowest observed effect level is expected to be lower in diabetics but could not be evaluated because of too few cases. The incidence of diabetes is increasing, and because diabetes is the leading cause of end-stage renal disease, this has important public health implications. The incidence of renal replacement therapy in Sweden is 125 per million, with an estimated increased prevalence of 5% per year.

26The nephrotoxic effects in the present study appear small in a clinical context, and only a few percent of the variances were explained by cadmium. However, the increase in the effect markers indicates renal toxicity, which should be considered an early sign of severe health effects. Because it concerns a large segment of the population worldwide, the results are of public health concern. Although the cadmium-induced kidney effects in several studies have been associated with decreasing GFR, a positive aspect is that progression of the very early effect may not always occur when the exposure is substantially decreased. It should, however, be emphasized that in areas with exposure to cadmium mainly through diet, the long half-time of cadmium in the soil will hamper a decrease of the exposure. Thus, far-reaching mitigation will be needed in addition to actions against smoking.