Bone Resorption and Environmental Exposure to Cadmium in Women: A Population Study

Background Environmental exposure to cadmium decreases bone density indirectly through hypercalciuria resulting from renal tubular dysfunction. Objective We sought evidence for a direct osteotoxic effect of cadmium in women. Methods We randomly recruited 294 women (mean age, 49.2 years) from a Flemish population with environmental cadmium exposure. We measured 24-hr urinary cadmium and blood cadmium as indexes of lifetime and recent exposure, respectively. We assessed the multivariate-adjusted association of exposure with specific markers of bone resorption, urinary hydroxylysylpyridinoline (HP) and lysylpyridinoline (LP), as well as with calcium excretion, various calciotropic hormones, and forearm bone density. Results In all women, the effect sizes associated with a doubling of lifetime exposure were 8.4% (p = 0.009) for HP, 6.9% (p = 0.10) for LP, 0.77 mmol/day (p = 0.003) for urinary calcium, –0.009 g/cm2 (p = 0.055) for proximal forearm bone density, and –16.8% (p = 0.065) for serum parathyroid hormone. In 144 postmenopausal women, the corresponding effect sizes were –0.01223 g/cm2 (p = 0.008) for distal forearm bone density, 4.7% (p = 0.064) for serum calcitonin, and 10.2% for bone-specific alkaline phosphatase. In all women, the effect sizes associated with a doubling of recent exposure were 7.2% (p = 0.001) for urinary HP, 7.2% (p = 0.021) for urinary LP, –9.0% (p = 0.097) for serum parathyroid hormone, and 5.5% (p = 0.008) for serum calcitonin. Only one woman had renal tubular dysfunction (urinary retinol-binding protein > 338 μg/day). Conclusions In the absence of renal tubular dysfunction, environmental exposure to cadmium increases bone resorption in women, suggesting a direct osteotoxic effect with increased calciuria and reactive changes in calciotropic hormones.

Cadmium is a persistent environmental toxicant (Hogervorst et al. 2007;Lauwerys and Hoet 2001). Sources of cadmium pollution are past and present emissions from nonferrous industries, waste incineration, use of cadmium-containing phosphate fertilizers and sewage sludge, and the burning of fossil fuels (Lauwerys and Hoet 2001). Human exposure to cadmium occurs through consumption of contaminated food or water (Hogervorst et al. 2007;Watanabe et al. 2004) or by inhalation of tobacco smoke or polluted air (Hogervorst et al. 2007). Cadmium accumulates in the human body, in particular in the liver and kidneys, and has an elimination half-life of 10-30 years (Järup et al. 1998). The urinary excretion of cadmium over 24 hr is a biomarker of lifetime exposure (Järup et al. 1998). Cadmium causes glomerular and tubular renal dysfunction (Staessen et al. 1994) and increases calciuria (Staessen et al. 1991).
Current estimates suggest that > 200 million people worldwide have osteoporosis and that the prevalence of this disease is escalating (Reginster and Burlet 2006). Staessen et al. (1999) showed that low-level environmental cadmium exposure promotes osteoporosis and leads to a higher risk of fractures, especially in postmenopausal women. Women are at greater risk of developing cadmium toxicity than are men (Choudhury et al. 2001). Animal (Wang et al. 1994) and in vitro (Regunathan et al. 2003) studies suggest that cadmium might have direct toxic effects on bone, but convincing evidence for such an effect in humans does not exist. The development of biochemical assays that measure pyridinium crosslinks of collagen, which are specific markers of bone resorption (McLaren et al. 1992), greatly facilitates the exploration of cadmium's osteotoxicity. In view of the epidemic of osteoporosis (Reginster and Burlet 2006) and the ubiquitous distribution of cadmium pollution (Lauwerys and Hoet 2001), we used urinary crosslinks as a marker to investigate the possible direct osteotoxicity of cadmium (over and beyond its indirect effects on bone via increased calciuria) (Staessen et al. 1991) in Flemish women living in districts with low to moderate environmental cadmium pollution.

Methods
Fieldwork. The Cadmium in Belgium study (CadmiBel, 1985(CadmiBel, -1989  ) included 1,107 Flemish participants randomly recruited from 10 districts in northeastern Belgium (Staessen et al. 1994). The participation rate was 78% (Staessen et al. 1994). The geometric mean cadmium concentration in the soil sampled from 85 kitchen gardens was 5.3 mg/kg (5th-95th percentile interval, 1.4-18.9) in six districts, which bordered on three zinc/cadmium smelters, and 0.9 mg/kg (0.4-1.6) in four districts, which were > 10 km away from the smelters (Hogervorst et al. 2007). The participants of the 10 districts had similar characteristics apart from exposure to cadmium (Staessen et al. 1994(Staessen et al. , 1999. We complied with all applicable requirements of U.S. and international regulations, in particular the Helsinki declaration for investigation of human subjects. The Ethics Review Board of the Medical Faculty of the University of Leuven approved the study. Participants gave informed consent at recruitment. From 1991 to 1996, in the framework of the Public Health and Environmental Exposure to Cadmium study (PheeCad), we invited 823 former CadmiBel participants, who had renewed their informed consent, for a measurement of bone density, of whom 614 (74.6%) responded. This cohort included 307 women, whose exposure to cadmium was exclusively environmental. Because of missing information, we excluded 13 women. Thus, the study population for the present analysis consisted of 294 women. Clinical measurements. As described elsewhere (Staessen et al. 1999), from 1991 to 1996 we measured bone density at the forearm just above the wrist by single photon absorptiometry (ND1100 bone density scanner; Nuclear Data Inc, Schaumburg, IL, USA). Distal scans of adult forearms traverse a mean of 35% trabecular bone, whereas in proximal scans this proportion declines to nearly 5%. Trained nurses measured the anthropometric characteristics of the women. They administered a questionnaire to collect information about the participants' lifestyle and medication intake. Socioeconomic status was coded and condensed into a scale with scores ranging from 1 to 3. Using published tables, we computed the energy spent in physical activity from body weight, time devoted to work and sports, and type of physical activity. Premenopause was defined as an active menstrual cycle throughout follow-up. Menopause was defined as the absence or cessation of periods during follow-up, confirmed by measurement of the serum concentration of follicle-stimulating hormone (FSH).
Biochemical measurements. At baseline (1985At baseline ( -1989 and follow-up (1991-1996), the participants collected a 24-hr urine sample in a wide-neck polyethylene container for the measurement of cadmium, calcium, retinolbinding protein, and creatinine. For these measurements, we applied the same analytical methods throughout the study. We measured serum and urinary creatinine (Bartels and Böhmer 1971) using an automated enzymatic technique (Technicon Autoanalyzer, Technicon Instruments, Tarrytown, NY, USA). At follow-up (1991At follow-up ( -1996, the nurses obtained a venous blood sample from the participants within 2 weeks of the bone density measurement. We measured the serum concentration of FSH (AR FSH Reagent Kit, Abbott 6C24-20; Louvain-la-Neuve, Belgium), calcitonin (CT-U.S.-IRMA Kit, BioSource Europe, Nivelles, Belgium ), and γ-glutamyltransferase, an index of alcohol intake, using commercially available kits. We determined bone-specific serum alkaline phosphatase activity on a COBAS-BIO centrifugal analyzer (Roche Diagnostics, Vilvoorde, Belgium), serum parathyroid hormone (PTH) by a two-site immunometric assay (Bouillon et al. 1990), serum and urinary calcium by compleximetry, and urinary retinol-binding protein, a biomarker of renal tubular dysfunction, by an automated nonisotopic immunoassay based on latex particle agglutination ).
To measure blood and urinary cadmium, we applied electrothermal atomic absorption spectrometry with a stabilized-temperature platform furnace and Zeeman background correction ). The external quality-control program did not show any time trend in the accuracy of the cadmium measurements. In the context of this article, NA 7 (4.9) -NA, not applicable. Values are arithmetic mean ± SD, geometric mean (5th-95th percentile interval), or number of women (%). Unless indicated otherwise, characteristics were measured at the time of the urine collection for crosslinks. To convert nanomoles of cadmium to micrograms, multiply by 0.11241; to convert millimoles of calcium to milligrams, multiply by 40.08; to convert picomoles of parathyroid hormone to nanograms, multiply by 9.428; to convert nanomoles calcitonin to micrograms, multiply by 3.4176.

A B
we used the average of the 24-hr urinary cadmium excretion at baseline (1985)(1986)(1987)(1988)(1989) and follow-up (1991-1996) as a measure of lifetime exposure and the blood cadmium concentration at the time of the urine collection for crosslinks (1991)(1992)(1993)(1994)(1995)(1996) as a biomarker of recent exposure. On the day of the bone density measurement (1991)(1992)(1993)(1994)(1995)(1996), the participants collected an exactly timed 4-hr urine sample for the measurement of the collagen pyridinium crosslinks, hydroxylysylpyridinoline (HP) and lysylpyridinoline (LP), and creatinine. We measured HP and LP by high-performance liquid chromatography, using a slight modification of the methods described by Black et al. (1998) and Uebelhart et al. (1990). The intra-and interassay precision was 5.9% and 7.0% for HP, and 9.5% and 10.2% for LP. The sensitivity of the assay was 1 pmol.
Statistical analyses. For database management and statistical analysis, we used SAS software (version 9.1; SAS Institute Inc., Cary, NC, USA). We logarithmically transformed variables with a non-Gaussian distribution. We represented the central tendency and spread of transformed variables by the geometric mean and the 5th to 95th percentile interval. We compared means and proportions by use of the large sample z-test and the chisquare statistic, respectively. We assessed longitudinal changes in proportions by McNemar's test. We compared parity across subgroups of women by Kruskall-Wallis test. All p-values refer to two-sided hypotheses.
We plotted mean values of the biomarkers of effect by quartiles of the exposure measures to ensure that there was no threshold phenomenon and that linear correlation techniques were appropriate. We investigated associations between biomarkers of effect and exposure using single and multiple linear regressions. We identified covariates by a stepwise regression procedure with the p-values for variables to enter and to stay in the model set at 0.15. Covariates considered for entry in the model were age, age squared, body mass index, smoking, γ-glutamyltransferase activity in serum, intake of diuretics, use of supplements of calcium and/or vitamin D, menopausal status, hormone replacement therapy and/or oral contraception, parity, energy spent in physical activity, socioeconomic status, and urinary retinolbinding protein.
The 5th-95th percentile interval of the blood cadmium concentration and the 24-hr urinary cadmium excretion spanned approximately a 10-fold increase. Because we had normalized the distributions of blood and urinary cadmium by a logarithmic transformation, we expressed changes in the biomarkers of effect as the effect sizes related to a doubling of blood or urinary cadmium. We estimated these responses and their 95% confidence interval (CI) by multiplying regression coefficients (± 1.96 × SE) by 0.3 (the logarithm of 2).

Characteristics of women.
The median interval between baseline and the follow-up examination was 6.6 years (range, 5.3-10.5 years). Based on the questionnaires administered at baseline and follow-up and the measurement of FSH in serum at follow-up, the study population included 150 premenopausal and 144 menopausal women. Table 1 lists the women's characteristics by menopausal status. Age at enrollment ranged from 20.4 to 73.6 years. As expected, bone density was lower in menopausal than in premenopausal women, whereas the opposite was true for the urinary excretion of crosslinks. The median time at which the participants started collecting urine for the measurement of crosslinks was 1000 hours (interquartile range, 0700-1200 hours).
Unadjusted analyses. In single regression analysis, we noticed positive correlations of the urinary excretion of HP (r = 0.23; p < 0.0001) and LP (r = 0.17; p = 0.003) with the 24-hr excretion of cadmium (Figure 1), which was mirrored by inverse associations of proximal (-0.32; p < 0.0001) and distal (-0.22; p = 0.0001) forearm bone density with the biomarker of lifetime exposure. In addition, there was a positive correlation between the 24-hr urinary excretion of calcium and cadmium (r = 0.19, p = 0.0009).
Adjusted analyses. In exploratory analyses, we studied the associations of biomarkers of effect with the 24-hr cadmium excretion across quartiles with adjustments applied for age, age squared, and menopausal status (Table 2). Urinary HP excretion increased significantly with higher 24-hr cadmium excretion (p for trend, 0.028). LP showed the same tendency (p for difference between the lowest and highest quartile, 0.095). The 24-hr calciuria increased significantly with higher 24-hr urinary cadmium (p for trend, 0.0002), with an opposite trend for PTH (p for difference between the lowest and highest quartile, 0.046) ( Table 2). In the minimally adjusted analyses across quartiles of 24-hr urinary cadmium excretion, the trends in the serum levels of total calcium, bone-specific alkaline phosphatase, and calcitonin did not reach significance.

Bone resorption and cadmium
Environmental Health Perspectives • VOLUME 116 | NUMBER 6 | June 2008 779 Values are arithmetic mean ± SD, geometric mean (5th-95th percentile interval), or number of women (%). a Calcium and/or vitamin D.
Our findings also remained consistent when we separately used the 24-hr urinary excretion of cadmium either at baseline or at follow-up as index of lifetime exposure and when we additionally adjusted for the time of day, at which participants collected urine for the measurement of HP and LP (data not shown).   Significance of the partial regression coefficients: NS, not significant; *0.1 < p < 0.05; **p ≤ 0.05; # p ≤ 0.01; ## p ≤ 0.001. Socioeconomic position and the 24-hr excretion of retinol-binding protein did not enter any model.

Discussion
Population-based studies from Belgium (Staessen et al. 1999), Sweden (Åkesson et al. 2006;Alfvén et al. 2000;Järup and Alfvén 2004), Japan (Honda et al. 2003), and China (Zhu et al. 2004) showed an association between osteoporosis and low-level environmental cadmium exposure. The interpretation of these findings was that cadmium-induced renal tubular damage Staessen et al. 1994) attenuated the calcium reabsorption in the nephron, resulting in hypercalciuria (Staessen et al. 1994) and demineralization of bones (Järup and Alfvén 2004;Staessen et al. 1999), particularly in menopausal women (Staessen et al. 1999). In keeping with experimental studies, the present study supports the interpretation that, in women, cadmium decreases bone density through a direct osteotoxic effect. Indeed, we found consistent associations between biomarkers of bone resorption, the urinary pyridinium crosslinks HP and LP, and biomarkers of lifetime and recent exposure to cadmium. In addition, serum PTH levels decreased with higher cadmium exposure, as might be expected when a toxic substance induces release of calcium from bone tissue. If the hypercalciuria in cadmium-exposed subjects were attributed entirely to excessive calcium loss in the renal tubules, one would expect instead an increase in PTH with higher exposure to compensate for the urinary calcium loss. A possible adaptive response favoring bone formation might occur, especially in postmenopausal women, with serum levels of bone-specific alkaline phosphatase activity and calcitonin correlating positively with cadmium exposure, as observed in the present study. To our knowledge, only one previous population study has addressed the possible association between bone resorption and lowlevel cadmium exposure. In 820 Swedish women 53-64 years of age, Åkesson and colleagues (2006) measured forearm bone mineral density, calciotropic hormones, and the urinary LP concentration not standardized for creatinine. They assessed exposure to cadmium, not from the 24-hr urinary excretion, but from the concentration in fresh urine samples and blood. The median values were 4.6 nmol/L (0.52 µg/L) and 3.4 nmol/L (0.38 µg/L), respectively. In our current study, the corresponding concentrations in urine and blood were 5.2 nmol/L (0.58 µg/L) and 8.0 nmol/L (0.90 µg/L). In multivariateadjusted analyses, Åkesson et al. (2006) reported inverse associations (p < 0.05) of bone density and serum PTH with the urinary cadmium concentration and a positive relation between urinary LP and cadmium. These associations persisted in never-smokers, who had the lowest, mainly dietary, cadmium exposure. For LP, there was a significant interaction between menopause and urinary cadmium. The associations with blood cadmium were not significant in the Åkesson et al. study, except for PTH. Laboratory studies strongly support the epidemiologic evidence for a direct osteotoxic effect of cadmium. In experimental animals exposed to cadmium, bone demineralization begins early after the start of cadmium exposure, well before the onset of kidney damage (Wilson and Bhattacharyya 1997). In cultures of bone marrow cells, cadmium accelerated the differentiation of new osteoclasts from their progenitor cells and enhanced the activity of mature osteoclasts (Wilson et al. 1996). In female mice, bilateral ovariectomy enhanced the osteotoxicity of cadmium (Comelekoglu et al. 2007). However, at the molecular level, the effects of cadmium on bone tissue need further clarification. Cadmium stimulated bone resorption by the up-regulation of the production of prostaglandin E 2 in osteoblasts through enhanced expression of phospholipase A 2 and cyclooxygenase (Miyahara et al. 1992). Exposure of human osteoblast-like cells to cadmium also produced an increase in caspase-3 activity and nuclear changes characteristic of apoptosis, including marginalization and condensing of chromatin and DNA fragmentation (Coonse et al. 2007). Experiments in genetically engineered mice suggested that the effects of cadmium on bone tissue require c-Src (Regunathan et al. 2002) and might be mediated via a p38 mitogen-activated phosphokinase pathway (Regunathan et al. 2003), but are independent from c-Fos expression (Regunathan et al. 2002).
Our findings might have important implications for environmental policies, especially those designed to protect women's health. Roughly, 200 million people worldwide suffer from osteoporosis (Reginster and Burlet 2006). In the United States, there are an estimated 44 million osteoporosis patients, of whom 30 million are women (Reginster and Burlet 2006). After menopause, osteoporosis occurs at an accelerated rate. The studies in Belgium (Staessen et al. 1999) and China (Zhu et al. 2004) demonstrated loss of bone mineral density in relation to cadmium exposure, which was more severe in women (Staessen et al. 1999;Zhu et al. 2004), particularly after the onset of menopause (Staessen et al. 1999). Itai-Itai disease in Japan, an advanced stage of cadmium-induced osteomalacia and osteoporosis combined with kidney disease, occurs almost exclusively in older women (Vahter et al. 2007). Women have a higher body burden of cadmium than men (Vahter et al. 2007). Low iron stores that are common during pregnancy and before menopause lead to an upregulation of the duodenal metal transporter, which has a high affinity for cadmium (Tallkvist et al. 2000;Vahter et al. 2007). In twin studies, the Bone resorption and cadmium Environmental Health Perspectives • VOLUME 116 | NUMBER 6 | June 2008 781 Table 5. Independent associations of forearm bone density and effect biomarkers in urine with current exposure as reflected by blood cadmium.

Forearm bone density
Biomarkers in urine Proximal (g/cm 2 ) Distal (g/cm 2 ) HP (log nmol/mmol creatinine) LP (log nmol/mmol creatinine) Calcium (mmol/day) Significance of the partial regression coefficients: NS, not significant; **p < 0.05; ## p < 0.001. All models were adjusted for the same covariates as in Table 3. heritability of the blood cadmium concentration was 65% in nonsmoking women, but only 13% in nonsmoking men (Björkman et al. 2000). Finally, experimental studies showed stronger effects of cadmium on calciotropic hormones and on the metabolism of calcium and phosphate in female than in male rats (Brzóka and Moniuszko-Jakoniuk 2005). The present study has limitations and strengths. Although our results were consistent after multiple adjustments and in sensitivity analyses, we cannot exclude residual confounding. We replicated the findings of Åkesson et al. (2006), albeit at higher exposure levels. However, the Swedish investigators measured LP and cadmium in concentration units on the same spot urine sample. We measured the crosslinks standardized to creatinine and 24-hr cadmium excretion on different samples, and we therefore excluded the possibility of a spurious association due to varying degrees of the concentration of urine in individual samples. Moreover, we also found significant dose-effect associations with blood cadmium, which with the exception of PTH was not the case in the Swedish study (Åkesson et al. 2006). Some experts consider LP as a more specific marker of bone resorption than HP (Robins 1983). However, both LP and HP originate from mature collagen. In most circumstances, bone collagen degradation is the major contributor to both crosslink compounds in urine, due to the low turnover rate of other tissues (McLaren et al. 1992). The urinary excretion of pyridinium crosslinks has a diurnal variation, with levels peaking in the morning (Schlemmer et al. 1992). Our results were consistent when we accounted for the starting time of the urine collection.
Cadmium is a ubiquitous and persistent environmental contaminant. In our study, zinc smelters began to emit cadmium into the atmosphere in 1888. The last zinc smelter shut down in 2002. Even though annual emissions dropped from 125,000 kg in 1950 to 130 kg in 1989, the historical pollution of the soil remains a source of exposure via food contamination and the inhalation of house dust (Hogervorst et al. 2007). In the United States, ecologic studies demonstrated cadmium pollution, not only close to industrial (Gale et al. 2004) or mining (Peplow and Edmonds 2004) settlements, but in agricultural (Schmitt et al. 2006) and coastal (Karouna-Renier et al. 2007) areas as well. Japanese women remain currently more exposed to cadmium than other rice-dependent populations in Asia and other parts of the world (Watanabe et al. 2004). Satarug and Moore (2004) predicted that the continuing mobilization of cadmium from once nonbioavailable geologic matrices into biologically accessible materials could gradually increase over the next 10-20 years and amplify the upward trend in osteoporosis in aging populations worldwide. Globally, in 2000, there were an estimated 9.0 million osteoporotic fractures (Johnell and Kanis 2006). These fractures caused the loss of 5.8 million disabilityadjusted life-years, of which 51% occurred in Europe and the Americas (Johnell and Kanis 2006). Cadmium is also nephrotoxic (Staessen et al. 1994) and increases the risk of lung cancer (Nawrot et al. 2006). Regulators must realize that because of these health effects and the very long biological half-life of cadmium (Järup et al. 1998), exposure due to human activities is unacceptable.
In conclusion, cadmium is an osteotoxic pollutant that increases bone resorption. Even in the absence of cadmium-induced renal tubular dysfunction, low-level environmental exposure to cadmium increases calciuria with reactive changes in calciotropic hormones. The provisional tolerable daily intake of cadmium via food is currently 1 µg/kg per day (Nordberg 2004). The question arises whether, in the light of the present findings and the disability associated with osteoporotic fractures in aging populations (Johnell and Kanis 2006), regulators should not lower this threshold, particularly for women.