Maternal Cadmium Exposure during Pregnancy and Size at Birth: A Prospective Cohort Study

1Human exposure to the toxic metal cadmium (Cd) occurs mainly through food, such as cereals, seafood, and offal, and inhalation of tobacco smoke. Once absorbed, Cd has a long half-life in the body, especially in the kidneys. Chronic Cd exposure has been shown to adversely affect kidney and bone and to increase the risk of cancer and overall mortality. Cd also functions as an endocrine disruptor and may thus affect reproduction and child development. In general, women are more susceptible to Cd toxicity, mainly because of increased intestinal uptake of Cd given low iron stores, which are more prevalent in women than in men.

2There is, however, little information on the effects of Cd exposure in early life. Cd has been shown to be both embryotoxic and teratogenic in a variety of animal species, but this has not yet been confirmed in humans. Cd accumulates in human placenta, but the placenta is not a complete barrier, and Cd concentrations in cord blood increase with maternal exposure. There is increasing evidence of associations between maternal Cd exposure and adverse pregnancy outcomes, such as reduced size at birth and preterm delivery. However, studies that have reported associations included very few women (44–78 women), and associations with the same outcomes varied markedly among the studies. Residual confounding cannot be excluded because very few covariates were considered. In contrast, two larger studies (106–262 women) found no association between maternal Cd exposure and size at birth.

3In the present study, we took advantage of our large, population-based, longitudinal mother–child cohort in Bangladesh to assess the effects of maternal Cd exposure on size at birth. We previously reported that arsenic (As) exposure is associated with reduced size at birth. Additionally, pregnant women in this rural area had elevated concentrations of Cd in their placentas, and placental Cd was inversely associated with zinc (Zn) in cord blood, suggesting a possible effect of Cd on the transfer of Zn to the fetus. Thus, it is possible that Cd has adverse effects on pregnancy outcomes as well.

4Study area and design. The present study was nested in a population-based food and micronutrient supplementation trial in pregnancy in women in rural Bangladesh [Maternal and Infant Nutrition Interventions, Matlab (MINIMat)], which enrolled approximately 4,500 pregnant women from November 2001 through October 2003, as described previously. During monthly household visits, community health research workers offered a pregnancy urine test to women who had missed their menstrual period. In the case of a positive pregnancy test, the women were asked to donate a urine sample for future analysis and thereafter were invited to the nearest health center for further examination. The present study included the women who were identified as pregnant during one calendar year, from 1 February 2002 through 31 January 2003, who were enrolled in the MINIMat trial and had a singleton birth with measurements of size at birth. In total, 1,697 women fulfilled these criteria.

5The study has been approved by the research and ethical review committee at the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), as well as the regional ethics committee at Karolinska Institutet, Sweden. Oral and written consents were obtained from the women.

6Exposure assessment. Of the 1,697 women, 1,616 donated a urine sample, on average at gestational week 8, and this urine sample was used for assessment of Cd exposure. Urinary Cd is a well-documented biomarker of chronic exposure. These 1,616 women did not significantly differ in age, body mass index (BMI), socioeconomic status (SES), or parity, or with regard to characteristics of their newborns (birth weight, length, head and chest circumference) from those who did not provide a urine sample (n = 81). However, those who donated a sample had slightly longer gestation (median, 39.4 weeks) than those who did not (median, 39.0 weeks; p = 0.011).

7Cd in urine was measured with inductively coupled plasma mass spectrometry (ICPMS; model 7500ce; Agilent Technologies, Waldbronn, Germany) at Karolinska Institutet (Stockholm, Sweden). Samples were stored at –70°C in acid-washed plastic vials. The ICPMS was operated in helium mode and 111Cd was monitored. The sample preparation and analytical methods have been described elsewhere. No samples were below the limit of detection (LOD; < 0.02 μg/L). Quality control was assessed by analyses of two commercial control materials (SeronormTM Trace Elements Urine Blank, ref. 201305, lot OK4636, and SeronormTM Trace Elements Urine, ref. 201205, lot NO2525; SERO AS, Billingstad, Norway). The obtained mean (± SD) urinary Cd concentrations in the two control materials were 0.26 ± 0.05 μg/L (recommended value 0.31 ± 0.05) and 4.8 ± 0.14 μg/L (5.06 ± 0.22), respectively. We have previously measured the sum of inorganic As and its methylated metabolites, referred to as urinary As, in these women because many are highly exposed to As via drinking water. These measurements were performed with hydride generation atomic absorption spectroscopy. To compensate for variation in urine dilution, the final Cd and As concentrations were adjusted to the mean specific gravity (1.012 g/mL) of the urine samples.

8Outcomes and covariates. Birth weight was measured with electronic or beam scales with precision of 10 g (UNICEF Uniscale; SECA Gmbh & Co., Hamburg, Germany), length with a locally produced wooden length board (accurate to 1 mm), and head and chest circumference with a flexible nonstretchable measuring tape (accurate to 1 mm). Approximately 40% of the women delivered at the health clinics and had birth anthropometry measured in connection with the delivery. A birth notification system was established for women who delivered at home, and birth anthropometry was most often measured within 24 hr. If more than 72 hr had passed since delivery, the weight was adjusted as described in detail elsewhere. Gestational age was calculated by subtracting the date for the last menstrual period from the date of birth. The date of last menstrual period was obtained by interviewing the women immediately after identification of pregnancy. For women who had forgotten this date (n = 26), we used the last menstrual period estimated by ultrasound measurements.

9Covariate information was obtained during household visits by the health research workers and at routine visits at the health facilities, as well as from the Health and Demographic Surveillance System database maintained by ICDDR,B (Dhaka, Bangladesh). Maternal height and weight were recorded at the health facilities, approximately in gestational week 8. Education was defined as years of formal schooling and categorized as no education, < 5 years, ≥ 5 to < 10 years, and ≥ 10 years. SES was classified according to a wealth index based mainly on household assets, as described previously. Maternal tobacco smoking and betel/tobacco chewing during pregnancy were defined as never or ever. The season of birth was categorized as premonsoon (January–May), monsoon (June–September), or postmonsoon (October–December).

10Statistical analyses. The statistical analyses were performed with Statistica for Windows (version 9.1; StatSoft, Inc., Tulsa, OK, USA) and STATA (version 11; StataCorp., College Station, TX, USA). All tests were two sided, and a p < 0.05 was considered statistically significant. Bivariate analyses between exposures, sizes at birth, and covariates were evaluated by Spearman’s correlation coefficient or analysis of variance depending on the type of data analyzed. The p-value for the difference between two independent groups was determined with the Mann–Whitney U-test.

11We further assessed associations between maternal urinary Cd (nontransformed) and size at birth (birth weight, birth length, head and chest circumference) using multivariable-adjusted linear regression analyses. The residual analyses indicated no major deviation from the assumptions of linear regression. We adjusted for variables significantly associated with either size at birth or urinary Cd (p < 0.05) in the bivariate analyses or that had previously been associated with size at birth. Maternal age and parity, and maternal education and SES were highly correlated (rs > 0.50). Therefore, we adjusted for maternal age and SES only, because they were more strongly associated with outcomes than were parity and maternal education. We evaluated colinearity by tolerance and variance inflation factor in the linear regression analyses. The final multiple linear regression analyses were adjusted for maternal age, BMI, SES, hemoglobin in gestational week 14, urinary As at gestational week 8, betel use, season of birth, gestational age at birth, and sex. Additionally, we stratified by sex and by SES (median split) and repeated the above analysis. Also, we assessed whether adding the different food (early and usual) and/or micronutrient supplementations (three different types) into the models modified the association between Cd and size at birth. We used “seemingly unrelated estimation” in STATA to adjust for conducting multiple tests on a set of correlated outcomes dealing with size at birth and compared these model estimates with estimates from the analysis described above.

12To assess the effect of maternal urinary Cd on the whole distribution of sizes at birth, we performed multivariable-adjusted quantile regression analyses using STATA. Linear regression models the relationship between a set of predictor variables and the mean of the response variable. Quantile regression models the relationship between a set of predictor variables and specific percentiles (or quantiles) of the response variable. Thus, quantile regression allows us to estimate associations between an exposure and the outcome at different levels of the outcome distribution, instead of associations with the mean value of the outcome only, and thus provides a method to determine whether a predictor variable, such as Cd, may have a different association with birth size among children with low birth size compared with children with normal birth size. Additionally, quantile regression estimates are robust to outliers in the data set. We assessed the 25th percentile (lowest quantile), 50th percentile (median quantile), and 75th percentile (highest quantile) for the different birth outcomes. Regression models for all three quantiles were estimated jointly using simultaneous quantile regression in STATA. Estimates, confidence intervals (CIs), and p-values were based on 200 bootstrap samples. The Wald test was used to test for difference between the jointly estimated regression coefficients across different quantiles.

13The background characteristics and biomarkers of the mothers are presented in Table 1. Of the 1,616 mothers, 75% had urinary Cd concentration < 1 μg/L. Maternal age ranged from 14 to 45 years, with an average of 26 years. About one-third of the women had a BMI < 18.5 kg/m2. None of the women reported smoking during pregnancy, but approximately 38% reported betel chewing. About one-third of the women were primiparous, and the multiparous women had on average two children (range, one to eight). Table 1 also shows the descriptive data of the newborns; 52% of the newborns were boys, the average gestational age was 39 weeks (range, 29–45 weeks), and around 8% were born preterm (< 37 weeks). Birth weight ranged from 1,059 g to 4,000 g, with an average of 2,672 g. After excluding those born preterm, 38% still had a low birth weight (< 2,500 g).

14In the bivariate analyses, maternal urinary Cd was not significantly correlated with birth weight, length, and head or chest circumference but was positively correlated with maternal age and urinary As concentrations and negatively correlated with maternal education, SES, and hemoglobin (Table 2). In addition, birth outcomes were significantly correlated with gestational age at birth and with maternal height, weight, BMI, age, parity, SES, and level of education (Table 2).

15Urinary Cd did not differ between mothers of infants born preterm (median, 0.61 μg/L) and those born at term (0.62 μg/L; p = 0.77). Multiparous women had significantly higher urinary Cd concentrations (median, 0.70 μg/L) than did primiparous women (median, 0.50 μg/L; p < 0.001), but they were also older than the primiparous women (median, 29 vs. 21 years of age; p < 0.001). Betel chewing was not significantly associated with maternal urinary Cd or any of the birth outcomes (data not shown).

16In girls, urinary Cd was negatively associated with chest circumference (rs = –0.071, p = 0.048), birth weight (rs = –0.055, p = 0.13), and birth length (rs = –0.041, p = 0.25). In boys, urinary Cd was positively associated with birth length (rs = 0.068, p = 0.050) and birth weight (rs = 0.027, p = 0.43) and weakly associated with reduced chest circumference (rs = –0.014, p = 0.68). Head circumference was not correlated with maternal urinary Cd in girls or boys (rs = –0.0072, p = 0.25; and rs = –0.0072, p = 0.83, respectively.)

17Multivariable-adjusted linear regression analyses, including maternal age, BMI, SES, hemoglobin, urinary As, and betel use and infant season of birth, gestational age, and sex, showed significant associations of maternal urinary Cd with mean birth weight (unstandardized regression coefficient B = –31.0; 95% CI: –59, –2.8; p = 0.031; Table 3) and mean head circumference (B = –0.15; 95% CI: –0.27, –0.026; p = 0.018). Moreover, maternal urinary Cd was close to negatively associated with mean chest circumference (B = –0.14; 95% CI: –0.30, 0.021; p = 0.088), whereas no association was found with birth length. Sex was significant in all models, confirming that, on average, girls were smaller than boys (p < 0.001 for all outcomes). When a multiplicative interaction term between urinary Cd and sex was included in the multivariable-adjusted regression analyses (Table 3), the interaction term was not significantly associated with any of the birth outcomes (p > 0.05). The stratified analyses showed that the inverse associations between maternal urinary Cd and mean birth outcomes were evident only in girls (Table 3), whereas in boys, maternal urinary Cd was not associated with any of the birth outcomes. Evaluation of effect size with linear regression analysis showed that a 1-μg/L increase in maternal urinary Cd was associated with a decrement of 45 g (95% CI: –82.5, –7.3) in mean birth weight, 0.26 cm (95% CI: –0.43, –0.088) in mean head circumference, and 0.24 cm (95% CI: –0.44, –0.030) in mean chest circumference in girls. Results were consistent with those shown when associations were estimated using seemingly unrelated postestimation to account for multiple testing on correlated outcomes (p-values shown in Table 3). We further stratified for high and low SES (median split) and found that the effect estimates of Cd did not change significantly in girls or boys (data not shown).

18We also performed multivariable-adjusted linear regression analyses with gestational age at birth as the dependent variable, adjusting for all potential confounders. There was no significant association with maternal urinary Cd in the cohort as a whole, or in girls or boys (data not shown). Adding the different food and micronutrient supplementations that were ingested during pregnancy into the models did not modify the associations between Cd and size at birth (data not shown), even when the analysis was restricted to the girls.

19The multivariable-adjusted quantile regression analysis confirmed that maternal urinary Cd was associated with birth size in girls but not in boys (Figure 1). Moreover, the results indicate that the associations between maternal urinary Cd and size at birth were very similar across the different percentiles of each outcome.

20The present study, which to our knowledge is the largest to estimate effects of Cd on fetal development, shows an unexpected difference in the association between maternal Cd exposure during pregnancy and infant size at birth by sex, with inverse associations in girls but little evidence of an effect in boys. The strongest associations were found for head circumference and birth weight. A 1-μg/L increase in Cd in maternal urine during pregnancy was associated with a decrease in head circumference of 0.26 cm and in birth weight of 45 g in girls. The magnitude of the estimated effects is not necessarily of clinical importance for individual children but may have considerable public health relevance given the high prevalence of elevated exposure to Cd, mainly through the diet.

21The main strengths of this study include the prospective study design and the large sample size with a wide range of Cd exposure. Also, we had extensive data on potential confounders, including maternal age, nutrition, season of birth, and SES. We also adjusted for As exposure and betel chewing, which have previously been associated with size at birth. However, we cannot rule out the possibility that our findings may still be biased by unmeasured confounders or by residual confounding. We assessed individual maternal Cd exposure by concentrations in urine, a well-recognized biomarker of chronic Cd exposure. Urinary Cd concentrations were adjusted for specific gravity, because creatinine-adjustment is highly dependent on body composition and type of diet, which may also be associated with birth weight. The analytical measurements were performed with an ICPMS method with high sensitivity and efficient removal of potential interferences (e.g., molybdenum). We selected all the women who were identified as pregnant during one calendar year, to cover potential temporal variation, especially in size at birth. Furthermore, none of the women smoked, which otherwise would have been a major confounder, because smoking is associated with several adverse outcomes of pregnancy, including low birth weight, as well as increased Cd exposure.

22We tested the robustness of our findings using both conventional multivariate linear regression models and quantile regression models. The quantile regression estimates for the median generally agreed with the linear regression estimates in terms of direction and magnitude, indicating that the linear regression estimates were not unduly influenced by outlying values. We did not find any evidence that Cd differentially affected children of smaller size.

23A limitation of the study is that we did not measure Cd in cord blood, which might have provided more precise information of the prenatal Cd exposure. Two previous studies, although limited in size (49 and 45 women), found inverse associations between cord blood Cd concentrations and birth weight. Another study reported a significant inverse association with head circumference, but not with birth weight. However, the effect on fetal growth may at least partly be related to the accumulation of Cd in placenta because it is associated with impaired transport of nutrients to the fetus. Indeed, placental Cd concentrations have been negatively associated with size at birth in a few small studies. Maternal urinary Cd is associated with both placenta and cord blood Cd and would be a useful exposure measure for evaluating combined effects of direct fetal Cd exposure and indirect effects due to placental accumulation of Cd. Maternal urinary Cd concentrations were measured on only one occasion in early pregnancy, but urinary Cd concentrations have previously been shown to be rather constant throughout pregnancy.

24There are several possible mechanisms by which maternal Cd exposure directly or indirectly may affect size at birth. We have previously shown that pregnant women in Matlab had elevated Cd concentrations in their placentas and that these Cd concentrations were inversely association with Zn in cord blood, suggesting that Cd may disturb Zn transfer to the fetus. Before that, an interaction between Cd and Zn in the placenta, leading to decreased Zn transfer to the fetus, had been shown only in vitro and in women with increased Cd exposure via smoking. Because insufficient Zn transfer to the fetus is likely to have consequences for pregnancy outcomes, especially in terms of intrauterine growth retardation and preterm delivery, this has been suggested as one of the main mechanisms for the Cd-related reduction in birth weight. However, because we found a sex difference in the association between maternal urinary Cd and size at birth, that mechanism may only partly explain the present findings. Another possible explanation for our findings may be the recently identified endocrine-disrupting properties of Cd.

25Glucocorticoid exposure in utero has been associated with impaired fetal growth. A key regulator in the fetoplacental barrier toward maternal glucocorticoids is 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2), which inactivates maternal cortisol to cortisone. In fact, it has been shown that placental 11β-HSD2 gene expression is reduced in pregnancies with intrauterine growth restriction, and small-for-gestational-age babies appear to have elevated cortisol concentrations. However, the underlying mechanism is unknown. Cd seems to be one factor affecting 11β-HSD2; 1 μM CdCl2 (cadmium chloride) markedly reduced 11β-HSD2 gene activity, mRNA, and protein levels in cultured human trophoblastic cells in vitro. On the other hand, very high Cd exposure of pregnant rats (50 mg/L in drinking water) did not alter placental 11β-HSD2, but corticosterone concentrations (corresponding to cortisol in humans) increased and the pup weights decreased. Notably, the effect of prenatal glucocorticoid exposure appears to be sex specific. In mothers receiving glucocorticoid treatment for asthma, 11β-HSD2 activity was significantly decreased in placentas of female but not male fetuses, and this was associated with increased umbilical cord blood cortisol levels and reduced birth weight. Thus, decreased 11β-HSD2 activity and increased concentrations of cortisol may be fundamental features in female fetal growth impairment.

26Cd may also interfere with the insulin-like growth factor (IGF) axis and thereby reduce fetal growth in a sex-specific manner. Rats given drinking water with 50 mg Cd/L had lower concentrations of IGF-1 and insulin-like growth factor-binding protein 3 (IGFBP-3) than did controls. In human pregnancies, both IGF-1 and IGFBP-3 are usually positively associated with birth weight. Furthermore, two epidemiologic studies have found that IGF-1 and IGFBP-3 levels in umbilical cord blood/plasma were higher in female than in male infants.

27We also hypothesize that Cd may affect the sex-specific programming, which begins very early in pregnancy and may contribute to the vulnerability of the developing fetus to maternal stressors. In an experimental study on chick embryos, Cd was found to down-regulate gene expression (levels) of DNA methyltransferases regulating the de novo DNA methylation, which is essential for normal embryogenesis. Interestingly, female embryonic stem cells were found to lose DNA methylation at a much faster rate than did male stem cells.

28Low birth weight is a significant public health problem worldwide, and it has been associated with neonatal mortality and morbidity, as well as increased risk for the development of diseases later in life. The association between Cd and reduced birth weight in girls (an estimated decrease of 45 g for a 1-μg/L increase in maternal urinary Cd) corresponds to about one-quarter of the estimated effect of smoking during pregnancy on birth weight (~ 200 g lower birth weight for mothers who smoke than for nonsmoking mothers), part of which may be mediated via the Cd inhaled with the tobacco smoke. To further put the effect size into perspective, an increase in birth weight of 37 g was achieved by supplementation with folic acid and iron in a double-blind randomized community trial in rural Nepal. In the present study, associations between Cd and birth size outcomes in girls were independent of SES and similar in small and larger infants; however, the relevance of our findings for birth outcomes in better-nourished populations needs to be evaluated.

29We believe that the main source of Cd in this population is the diet, especially plant-derived foods, because none of the women smoked. The median maternal urinary Cd concentration (median, 0.63 μg/L) was higher than that reported in young women in the United States (geometric mean, 0.27 μg/g creatinine) and Europe (median, 0.31 μg/L), probably due to the rice-based diet, because rice is known to take up Cd from the soil to a greater extent than many other plants. Rice samples collected from 63 families in the study area contained on average 69 μg Cd/kg dry weight (10th and 90th percentiles, 11–98 μg Cd/kg), which would correspond to a daily intake of 25–35 μg Cd, assuming a daily consumption of 400–500 g rice grains for an adult person. This daily Cd intake is similar to that in other Asian countries where rice is the main staple food and also to that in most individuals consuming vegetarian diets. It is essential to investigate other potential sources of Cd, particularly vegetables and wheat, as well as agricultural fertilizers that may be a major contributor to Cd in agricultural soil.

30We found significant inverse associations between maternal Cd exposure and birth anthropometry in girls, especially head circumference and birth weight, but no associations in boys. These findings, along with evidence from previous experimental and observational studies, suggest that there may be sex differences in both toxicokinetics (i.e., intestinal uptake) and toxicodynamics of Cd. It is essential to follow the children through childhood to clarify whether the apparent effect on growth remains and whether the early-life Cd exposure is associated with other health outcomes later in childhood. This study clearly emphasizes the need to consider early-life effects of Cd in health risk assessments and to reduce Cd pollution worldwide.