Cadmium Exposure and Neurodevelopment in US Children
Cadmium Exposure and Neurodevelopment in US Children
We observed that children in the highest quartile of urinary cadmium had significantly higher odds of both LD and special education when compared with those in the lowest quartile. A few prior studies have linked cadmium exposure with LD, and these studies relied on hair and blood samples to assess cadmium exposure. Two case–control studies demonstrated higher hair cadmium concentrations in children with LD (Capel et al. 1981; Pihl and Parkes 1977). A third study also found higher hair cadmium concentrations in children with LD, but the difference was statistically significant only for males (Ely et al. 1981). In that study, the authors considered the sexes separately but did not present the type of statistical evaluation for interaction that we report here. Interestingly, in our study the sex–cadmium interaction was not significant, but the effect estimate for urinary cadmium was larger among males. A fourth study reported no association between LD and blood cadmium in the 1999–2000 NHANES data (Lee et al. 2007). Cadmium accumulates in the kidney, and urinary cadmium concentration is considered to be a marker of cumulative exposure/body burden, whereas blood cadmium is thought to be a better indicator of recent exposure (Lauwerys et al. 1994). The different exposure metric used in the Lee et al. (2007) study may help explain the discrepancy in our findings.
In addition to higher odds of LD, we found that children in the highest quartile of urinary cadmium also had higher odds of special education placement. Special education is a "catch-all" outcome that likely involves a variety of neurocognitive and behavioral dysfunctions, including LDs such as reading difficulties, dyslexia, ADHD, and language/communication disorders, as well as behavior problems, psychiatric conditions, and perhaps some physical dysfunctions. The broad heterogeneous nature of both special education and LD as outcome measures prevents inferences about specific learning or cognitive domains. We are not aware of prior epidemiologic studies directly relating cadmium exposure and special education. However, any of the previously mentioned animal or human studies that link cadmium exposure to adverse neurobehavioral/neurocognitive or general health outcomes may be relevant here, because these outcomes could lead to special education placement.
There are data supporting the biological plausibility of cadmium exposure as a risk factor for LD and special education placement. For example, cadmium can inhibit the calcium flux required for neurotransmitter release (Hirning et al. 1988; Nation et al. 1989) and might thereby disrupt the neural communication required for synaptic network formation during development. Cadmium has also been shown to influence the proliferation and differentiation of neuroblasts in culture (Gulisano et al. 2009), and there is evidence that cadmium could indirectly affect the developing brain by disrupting thyroid hormone function (Iijima et al. 2007).
Our findings for ADHD did not reach statistical significance, but the direction of the association suggests a possible decreased risk of ADHD diagnosis in children with urinary cadmium levels above the 25th percentile. There is only limited information from prior epidemiologic studies on cadmium exposure and ADHD/executive function. Lee et al. (2007) reported a nonsignificant trend of increasing odds of ADHD with increasing blood cadmium levels that was not present after adjustment for persistent organic pollutants based on an analysis of 1999–2000 NHANES data. As mentioned above, the different exposure metrics may partly explain the difference in our findings, because blood cadmium is a marker of recent exposure, whereas urine cadmium is a marker of chronic exposure (Lauwerys et al. 1994). Cao et al. (2009) reported a nonsignificant trend of increasing problem behavior scores at 7 years of age with increasing blood cadmium levels measured at 2 years of age, but there were no obvious trends with increasing blood cadmium in the attention/executive subdomains, the hyperactivity subdomain, or the ADHD index. The prospective approach is a strength of their study, but their unique population consisted solely of lead-poisoned children (blood lead levels of 20–44 μg/dL at enrollment), and the results may not be generalizable to non-lead-poisoned children.
Previous animal-based studies have demonstrated a variety of seemingly inconsistent effects of cadmium exposure on neurophysiology and activity levels. These findings include, for example, reduced exploratory activity and decreased time spent moving, but they also include hyperactivity, and evidence of changes in central nervous system dopamine and serotonin metabolism (Ali et al. 1986; Desi et al. 1998; Nation et al. 1989, 1990). These two neurotransmitter systems have been implicated in the etiology of ADHD (Faraone et al. 2005). The varied direction of effects on activity level seen in the animal literature might be related to differences in the timing of cadmium exposure during neurodevelopment, the presence of other uncontrolled variables, and/or the specific phenotype measured in each study. We lack detailed information on exposure timing, which makes it difficult to interpret the direction of the ORs for ADHD in the context of the animal literature (Andersen and Navalta 2004). If elevated cadmium exposure decreased activity levels, this may have made ADHD diagnosis less likely. It is also possible that cadmium may cause other neurocognitive dysfunctions that serve as competing risks to ADHD diagnosis (perhaps children with other diagnostic labels were less likely to receive an ADHD label), and the potential influence of chance should not be overlooked. Because these are cross-sectional data, the temporal relationship between exposure and outcome is not discernable, and it is possible that urinary cadmium concentration tended to decrease after ADHD diagnosis (e.g., children with ADHD may be more likely to have behaviors that decrease cadmium exposure, absorption, or excretion). In the lead-stratified analysis, we found evidence that the OR for ADHD is < 1 only among those with blood lead levels above the median. The interaction was not significant, but there is toxicologic evidence that cadmium exposure may attenuate lead-mediated increases in activity (Nation et al. 1990). Further research is needed to clarify these issues.
Previous cadmium risk assessments have considered renal effects to be the most sensitive end point of cadmium toxicity, and they identified urinary cadmium threshold levels that should protect against renal damage [EFSA 2009; World Health Organization/Food and Agriculture Organization of the United Nations (WHO/FAO) 2011]. Recent risk assessments by the European Food Safety Authority (EFSA 2009) and the WHO (WHO/FAO 2011) yielded urinary cadmium reference levels of 1 and 5.24 μg cadmium/g creatinine, respectively. When we excluded from the analyses the four study participants with urinary cadmium levels above the EFSA reference level, the associations between urinary cadmium and LD/special education were still evident (comparing the highest and lowest urinary cadmium quartiles: LD, OR = 3.25; 95% CI: 1.45, 7.28; special education, OR = 3.03; 95% CI: 1.14, 8.08). Thus, our work demonstrates associations with LD/special education at urinary cadmium levels below both the WHO and EFSA reference levels.
If these associations are replicated in other populations, then neurodevelopmental toxicity may be a sensitive end point to consider in future cadmium risk assessments. The EFSA and WHO risk assessments used toxicokinetic models to link creatinine-standardized urinary cadmium levels to dietary cadmium intake in order to estimate dietary intake standards (Amzal et al. 2009; EFSA 2009; WHO/FAO 2011). These toxicokinetic models were validated in adults (Amzal et al. 2009), but they are probably not appropriate to use among children, because exposure routes may differ (Weidenhamer et al. 2011) and because urinary creatinine concentration varies markedly with small increases in age among children (Barr et al. 2005). Recent work by Weidenhamer et al. (2011) suggests that mouthing or accidentally swallowing objects such as inexpensive jewelry may also contribute significantly to cadmium exposure in childhood. The extent to which these sources contributed to cadmium exposure in our population is unknown, but future risk assessments should not ignore the potential impact of non-food-based exposures or that current toxicokinetic models are not child specific.
Exposure. The use of urinary cadmium as an exposure metric is a strength of this study. Urinary cadmium integrates exposure over many years (Lauwerys et al. 1994); thus, if brain development is sensitive to cadmium exposure in any of the time periods represented by this exposure metric, then this effect could be detected in our analysis. However, it is not possible to determine etiologically relevant time windows of exposure or to confirm the temporal sequence of exposure and outcomes based on the available data. A shorter time-course exposure metric such as blood cadmium (Lauwerys et al. 1994), in the context of a longitudinal prospective study that measures exposure and outcome at several time points, may be able to determine if the associations are driven by exposure that occurs in specific developmental windows.
Outcomes. We believe the diversity of outcomes evaluated is a strength of this study, because the combination of these three outcomes constitutes a screen for common neurodevelopmental dysfunctions. One limitation of these outcome measures is that they were derived from parent or proxy-respondent reports rather than neuropsychological evaluations. The use of ADHD treatments as an outcome would have likely identified only a subset of ADHD cases, resulting in low case numbers and analyses of limited power (Froehlich et al. 2007). Neuropsychiatric screening measures, such as the National Institute of Mental Health Diagnostic Interview Scale for Children that includes assessment of criteria for ADHD based on the 4th edition of the Diagnostic and Statistical Manual of Mental Disorders (CDC-NCHS 2006; Froehlich et al. 2009), would have been more objective, but these outcomes were not present in the publicly available NHANES data. Although outcomes were classified based on proxy-respondent reports, it is unlikely that outcome misclassification would be differential with respect to exposure unless the accuracy of reporting was related to unmeasured factors associated with cadmium exposure. Typically, nondifferential misclassification would be expected to bias associations toward the null (Rothman et al. 2008).
Confounding. The extensive covariate data available in NHANES, combined with the large sample size and high number of cases, allowed us to both evaluate and account for many potentially confounding variables. We sought to evaluate the association of cadmium exposure with the outcomes after accounting for other known correlates of the outcomes. We evaluated three sets of models: a) models adjusted only for creatinine, b) core models adjusted for primary potential confounders, and c) full models also adjusted for additional predictors of a priori interest. The conclusions from all three of these approaches were consistent. We further note that adjusting for iron deficiency using low hemoglobin had little effect on the results and does not alter the conclusions of this study (data not shown). As in any observational epidemiology study, we cannot rule out the possibility that confounding may have meaningfully affected our results. Potential sources of confounding in these analyses might include a lack of detailed information on the home environment (Bradley 1993) and parental psychopathology (Bellinger 2001).
Study design. The cross-sectional design of NHANES is a limitation of this study, because the temporal relationships between variables are not discernable. It is possible that higher cadmium exposure puts children at greater risk of LD/special education, but it is also possible that children with LD/special education have behaviors or prefer foods that increase their cadmium exposure. However, we are unaware of evidence supporting this reverse causation explanation.
The NHANES study design does offer strengths related to power and generalizability. To our knowledge, this is the largest study to evaluate associations between urinary cadmium and childhood learning/behavioral phenotypes. Because NHANES was designed to represent the noninstitutionalized U.S. population (CDC-NCHS 2010b), our findings should be generalizable to U.S. children 6–15 years of age.
Discussion
LD and Special Education
We observed that children in the highest quartile of urinary cadmium had significantly higher odds of both LD and special education when compared with those in the lowest quartile. A few prior studies have linked cadmium exposure with LD, and these studies relied on hair and blood samples to assess cadmium exposure. Two case–control studies demonstrated higher hair cadmium concentrations in children with LD (Capel et al. 1981; Pihl and Parkes 1977). A third study also found higher hair cadmium concentrations in children with LD, but the difference was statistically significant only for males (Ely et al. 1981). In that study, the authors considered the sexes separately but did not present the type of statistical evaluation for interaction that we report here. Interestingly, in our study the sex–cadmium interaction was not significant, but the effect estimate for urinary cadmium was larger among males. A fourth study reported no association between LD and blood cadmium in the 1999–2000 NHANES data (Lee et al. 2007). Cadmium accumulates in the kidney, and urinary cadmium concentration is considered to be a marker of cumulative exposure/body burden, whereas blood cadmium is thought to be a better indicator of recent exposure (Lauwerys et al. 1994). The different exposure metric used in the Lee et al. (2007) study may help explain the discrepancy in our findings.
In addition to higher odds of LD, we found that children in the highest quartile of urinary cadmium also had higher odds of special education placement. Special education is a "catch-all" outcome that likely involves a variety of neurocognitive and behavioral dysfunctions, including LDs such as reading difficulties, dyslexia, ADHD, and language/communication disorders, as well as behavior problems, psychiatric conditions, and perhaps some physical dysfunctions. The broad heterogeneous nature of both special education and LD as outcome measures prevents inferences about specific learning or cognitive domains. We are not aware of prior epidemiologic studies directly relating cadmium exposure and special education. However, any of the previously mentioned animal or human studies that link cadmium exposure to adverse neurobehavioral/neurocognitive or general health outcomes may be relevant here, because these outcomes could lead to special education placement.
There are data supporting the biological plausibility of cadmium exposure as a risk factor for LD and special education placement. For example, cadmium can inhibit the calcium flux required for neurotransmitter release (Hirning et al. 1988; Nation et al. 1989) and might thereby disrupt the neural communication required for synaptic network formation during development. Cadmium has also been shown to influence the proliferation and differentiation of neuroblasts in culture (Gulisano et al. 2009), and there is evidence that cadmium could indirectly affect the developing brain by disrupting thyroid hormone function (Iijima et al. 2007).
ADHD
Our findings for ADHD did not reach statistical significance, but the direction of the association suggests a possible decreased risk of ADHD diagnosis in children with urinary cadmium levels above the 25th percentile. There is only limited information from prior epidemiologic studies on cadmium exposure and ADHD/executive function. Lee et al. (2007) reported a nonsignificant trend of increasing odds of ADHD with increasing blood cadmium levels that was not present after adjustment for persistent organic pollutants based on an analysis of 1999–2000 NHANES data. As mentioned above, the different exposure metrics may partly explain the difference in our findings, because blood cadmium is a marker of recent exposure, whereas urine cadmium is a marker of chronic exposure (Lauwerys et al. 1994). Cao et al. (2009) reported a nonsignificant trend of increasing problem behavior scores at 7 years of age with increasing blood cadmium levels measured at 2 years of age, but there were no obvious trends with increasing blood cadmium in the attention/executive subdomains, the hyperactivity subdomain, or the ADHD index. The prospective approach is a strength of their study, but their unique population consisted solely of lead-poisoned children (blood lead levels of 20–44 μg/dL at enrollment), and the results may not be generalizable to non-lead-poisoned children.
Previous animal-based studies have demonstrated a variety of seemingly inconsistent effects of cadmium exposure on neurophysiology and activity levels. These findings include, for example, reduced exploratory activity and decreased time spent moving, but they also include hyperactivity, and evidence of changes in central nervous system dopamine and serotonin metabolism (Ali et al. 1986; Desi et al. 1998; Nation et al. 1989, 1990). These two neurotransmitter systems have been implicated in the etiology of ADHD (Faraone et al. 2005). The varied direction of effects on activity level seen in the animal literature might be related to differences in the timing of cadmium exposure during neurodevelopment, the presence of other uncontrolled variables, and/or the specific phenotype measured in each study. We lack detailed information on exposure timing, which makes it difficult to interpret the direction of the ORs for ADHD in the context of the animal literature (Andersen and Navalta 2004). If elevated cadmium exposure decreased activity levels, this may have made ADHD diagnosis less likely. It is also possible that cadmium may cause other neurocognitive dysfunctions that serve as competing risks to ADHD diagnosis (perhaps children with other diagnostic labels were less likely to receive an ADHD label), and the potential influence of chance should not be overlooked. Because these are cross-sectional data, the temporal relationship between exposure and outcome is not discernable, and it is possible that urinary cadmium concentration tended to decrease after ADHD diagnosis (e.g., children with ADHD may be more likely to have behaviors that decrease cadmium exposure, absorption, or excretion). In the lead-stratified analysis, we found evidence that the OR for ADHD is < 1 only among those with blood lead levels above the median. The interaction was not significant, but there is toxicologic evidence that cadmium exposure may attenuate lead-mediated increases in activity (Nation et al. 1990). Further research is needed to clarify these issues.
Implications for Cadmium Risk Assessments
Previous cadmium risk assessments have considered renal effects to be the most sensitive end point of cadmium toxicity, and they identified urinary cadmium threshold levels that should protect against renal damage [EFSA 2009; World Health Organization/Food and Agriculture Organization of the United Nations (WHO/FAO) 2011]. Recent risk assessments by the European Food Safety Authority (EFSA 2009) and the WHO (WHO/FAO 2011) yielded urinary cadmium reference levels of 1 and 5.24 μg cadmium/g creatinine, respectively. When we excluded from the analyses the four study participants with urinary cadmium levels above the EFSA reference level, the associations between urinary cadmium and LD/special education were still evident (comparing the highest and lowest urinary cadmium quartiles: LD, OR = 3.25; 95% CI: 1.45, 7.28; special education, OR = 3.03; 95% CI: 1.14, 8.08). Thus, our work demonstrates associations with LD/special education at urinary cadmium levels below both the WHO and EFSA reference levels.
If these associations are replicated in other populations, then neurodevelopmental toxicity may be a sensitive end point to consider in future cadmium risk assessments. The EFSA and WHO risk assessments used toxicokinetic models to link creatinine-standardized urinary cadmium levels to dietary cadmium intake in order to estimate dietary intake standards (Amzal et al. 2009; EFSA 2009; WHO/FAO 2011). These toxicokinetic models were validated in adults (Amzal et al. 2009), but they are probably not appropriate to use among children, because exposure routes may differ (Weidenhamer et al. 2011) and because urinary creatinine concentration varies markedly with small increases in age among children (Barr et al. 2005). Recent work by Weidenhamer et al. (2011) suggests that mouthing or accidentally swallowing objects such as inexpensive jewelry may also contribute significantly to cadmium exposure in childhood. The extent to which these sources contributed to cadmium exposure in our population is unknown, but future risk assessments should not ignore the potential impact of non-food-based exposures or that current toxicokinetic models are not child specific.
Strengths and Limitations
Exposure. The use of urinary cadmium as an exposure metric is a strength of this study. Urinary cadmium integrates exposure over many years (Lauwerys et al. 1994); thus, if brain development is sensitive to cadmium exposure in any of the time periods represented by this exposure metric, then this effect could be detected in our analysis. However, it is not possible to determine etiologically relevant time windows of exposure or to confirm the temporal sequence of exposure and outcomes based on the available data. A shorter time-course exposure metric such as blood cadmium (Lauwerys et al. 1994), in the context of a longitudinal prospective study that measures exposure and outcome at several time points, may be able to determine if the associations are driven by exposure that occurs in specific developmental windows.
Outcomes. We believe the diversity of outcomes evaluated is a strength of this study, because the combination of these three outcomes constitutes a screen for common neurodevelopmental dysfunctions. One limitation of these outcome measures is that they were derived from parent or proxy-respondent reports rather than neuropsychological evaluations. The use of ADHD treatments as an outcome would have likely identified only a subset of ADHD cases, resulting in low case numbers and analyses of limited power (Froehlich et al. 2007). Neuropsychiatric screening measures, such as the National Institute of Mental Health Diagnostic Interview Scale for Children that includes assessment of criteria for ADHD based on the 4th edition of the Diagnostic and Statistical Manual of Mental Disorders (CDC-NCHS 2006; Froehlich et al. 2009), would have been more objective, but these outcomes were not present in the publicly available NHANES data. Although outcomes were classified based on proxy-respondent reports, it is unlikely that outcome misclassification would be differential with respect to exposure unless the accuracy of reporting was related to unmeasured factors associated with cadmium exposure. Typically, nondifferential misclassification would be expected to bias associations toward the null (Rothman et al. 2008).
Confounding. The extensive covariate data available in NHANES, combined with the large sample size and high number of cases, allowed us to both evaluate and account for many potentially confounding variables. We sought to evaluate the association of cadmium exposure with the outcomes after accounting for other known correlates of the outcomes. We evaluated three sets of models: a) models adjusted only for creatinine, b) core models adjusted for primary potential confounders, and c) full models also adjusted for additional predictors of a priori interest. The conclusions from all three of these approaches were consistent. We further note that adjusting for iron deficiency using low hemoglobin had little effect on the results and does not alter the conclusions of this study (data not shown). As in any observational epidemiology study, we cannot rule out the possibility that confounding may have meaningfully affected our results. Potential sources of confounding in these analyses might include a lack of detailed information on the home environment (Bradley 1993) and parental psychopathology (Bellinger 2001).
Study design. The cross-sectional design of NHANES is a limitation of this study, because the temporal relationships between variables are not discernable. It is possible that higher cadmium exposure puts children at greater risk of LD/special education, but it is also possible that children with LD/special education have behaviors or prefer foods that increase their cadmium exposure. However, we are unaware of evidence supporting this reverse causation explanation.
The NHANES study design does offer strengths related to power and generalizability. To our knowledge, this is the largest study to evaluate associations between urinary cadmium and childhood learning/behavioral phenotypes. Because NHANES was designed to represent the noninstitutionalized U.S. population (CDC-NCHS 2010b), our findings should be generalizable to U.S. children 6–15 years of age.
Source...