Associations of Prenatal Exposure to Cadmium With Child Growth, Obesity, and Cardiometabolic Traits

Abstract

Prenatal cadmium exposure has been associated with impaired fetal growth; much less is known about the impact during later childhood on growth and cardiometabolic traits. To elucidate the associations of prenatal cadmium exposure with child growth, adiposity, and cardiometabolic traits in 515 mother-child pairs in the Rhea Mother-Child Study cohort (Heraklion, Greece, 2007–2012), we measured urinary cadmium concentrations during early pregnancy and assessed their associations with repeated weight and height measurements (taken from birth through childhood), waist circumference, skinfold thickness, blood pressure, and serum lipid, leptin, and C-reactive protein levels at age 4 years. Adjusted linear, Poisson, and mixed-effects regression models were used, with interaction terms for child sex and maternal smoking added. Elevated prenatal cadmium levels (third tertile of urinary cadmium concentration (0.571–2.658 μg/L) vs. first (0.058–0.314 μg/L) and second (0.315–0.570 μg/L) tertiles combined) were significantly associated with a slower weight trajectory (per standard deviation score) in all children (β = −0.17, 95% confidence interval (CI): −0.32, −0.02) and a slower height trajectory in girls (β = −0.30, 95% CI: −0.52,−0.09; P for interaction = 0.025) and in children born to mothers who smoked during pregnancy (β = −0.48, 95% CI: −0.83, −1.13; P for interaction = 0.027). We concluded that prenatal cadmium exposure was associated with delayed growth in early childhood. Further research is needed to understand cadmium-related sex differences and the role of coexposure to maternal smoking during early pregnancy.

Cadmium is a toxic pollutant to which exposure occurs through tobacco smoke, household dust, and foods grown in contaminated soil. Diet is the main source of environmental exposure to cadmium among non–occupationally exposed and nonsmoking populations (1). Over the past few decades, concern has risen about the potential health effect of low-level exposure to toxic metallic compounds, including cadmium, among environmentally exposed populations, especially in children and pregnant women (2).

The “developmental origins of health and disease” hypothesis posits that environmental exposures during early life may produce permanent changes in tissues’ structure, physiology, and function, leading to metabolic disease pathogenesis (3). Animal studies have shown embryotoxicity even at low levels of cadmium exposure (4–7). In human birth cohort studies, prenatal cadmium exposure has been associated with impaired fetal growth, reduced birth weight, and small-for-gestational-age birth (8–11). Low birth weight followed by rapid weight gain during early postnatal life has been associated with increased long-term risk of central obesity and obesity-related diseases (12, 13). Although there is evidence of adverse birth outcomes from prenatal cadmium exposure, much less is known about the impact of this exposure in later childhood on growth, obesity, and metabolic traits. Only a few studies have examined the association of maternal cadmium exposure with childhood growth trajectories (14–16), obesity (15), and blood pressure (17), and results have not been consistent. The majority of studies have suggested that the adverse outcomes of prenatal cadmium exposure are more prominent in girls than in boys (11, 14, 15, 17), although the underlying mechanisms of sex-specific associations with cadmium are not yet fully understood.

We examined whether maternal cadmium exposure during early pregnancy was associated with child growth trajectories, child adiposity, and metabolic traits at preschool age, and whether the associations were modified by child sex and maternal smoking during pregnancy, in a sample of mother-child pairs participating in a pregnancy cohort study in Heraklion, Greece, during 2007–2012 (18).

METHODS

Study population

The present analysis was conducted within the Rhea Mother-Child Cohort Study, a prospective longitudinal cohort study in Heraklion, Greece (19). The mothers were recruited into the cohort in early pregnancy, in connection with the first ultrasound examination (before 15 weeks of gestation) and were contacted again for child follow-up at 9 months, 18 months, and 4 years of age (2007–2012). The inclusion criteria were confirmed pregnancy, residency within the study area, being aged 16 years or older, and having a good understanding of the Greek language. In total, 879 singleton children participated in the age-4-years follow-up of the study, during which anthropometry and cardiometabolic factors were assessed in 779 children. Of those children, complete data on outcomes and cadmium exposure during pregnancy were available for 557 mother-child pairs, of which 42 pairs (7.5%) were excluded from the present analysis due to missing data on covariates. Because of the low percentage of participants with missing covariate data and the fact that no significant differences in exposure or outcomes were observed between included and excluded subjects (P values ranged from 0.17 to 0.97, based on the χ² test, Student’s t test, or the Mann-Whitney U test), we did not impute missing data for the covariates.

The study was conducted according to the principles of the Helsinki Declaration, and it was approved by the ethical committee of the University Hospital in Heraklion, Greece, as well as the Regional Ethical Review Board in Stockholm, Sweden. Informed consent was obtained from all participants included in the present analysis.

Sample collection and trace element analyses

Concentrations of cadmium were measured in spot urine samples collected from the women during pregnancy (at a median of 13 (interquartile range, 12–15) weeks of gestation). Urinary cadmium level is a well-established biomarker of long-term cadmium exposure (20). The measurements were conducted at the Karolinska Institutet in Stockholm, Sweden, using inductively coupled plasma mass spectrometry (Agilent 7,700×; Agilent Technologies, Tokyo, Japan) (21). The urine samples were diluted 1:10 in 1% nitric acid (prepared from 65% nitric acid (Suprapur; Merck Sharp & Dohme, Darmstadt, Germany)). To minimize potential interference, cadmium (isotope 111) was measured in helium mode. The limit of detection (3 × standard deviation (SD) of the blank value) for urinary cadmium was less than 0.010 μg/L. No samples contained concentrations below the calculated limit of detection. Quality control was assessed by analyses of 2 commercial control materials (Seronorm Trace Elements Urine Blank (reference 201305, lot OK4636) and Seronorm Trace Elements Urine (reference 201205, lot NO2525); SERO AS, Billingstad, Norway) within each analytical run. The obtained mean urinary cadmium concentrations in the 2 control materials were 0.20 (SD, 0.01) μg/L (recommended value, 0.31 (SD, 0.05) μg/L; n = 21) and 5.0 (SD, 0.16) μg/L (recommended value, 5.06 (SD, 0.22) μg/L; n = 22), respectively. In general, the obtained mean urinary cadmium concentrations showed good agreement with the recommended reference values. Finally, cadmium urinary concentrations were adjusted for specific gravity, measured with a digital refractometer (EUROMEX RD712 Clinical Refractometer; Euromex Microscopen BV, Arhnem, Holland) in order to compensate for the variation in urine dilution (urinary concentration × [mean specific gravity (1.020) − 1/individual specific gravity − 1]) (22).

Child anthropometry

At the follow-up visits held at ages 9 months and 4 years, trained research assistants measured children’s weight and length/height, using validated scales (Seca 354 baby scale and Seca Bellisima 841 scale; Seca GmbH & Co. KG, Hamburg, Germany) and stadiometers (Seca 210 measuring mat and Seca 213 stadiometer; Seca GmbH & Co. KG) according to standard (operating) procedures. We abstracted data from the children’s medical records on weight and length/height from 3 months to 4 years of age (with a median of 19 (range, 2–42) measurements per child). We used both abstracted data from medical records and clinical measures at follow up visits (at ages 9 months and 4 years) to construct the child’s growth trajectories up to the age of 4 years. We calculated body mass index (BMI; weight (kg)/height (m)²) and converted raw values into sex- and age-specific SD scores (z scores) by using internally generated growth reference curves. The internally generated growth curves were estimated by fitting multilevel (mixed) models, with fractional polynomials of age and random effects for the child (23, 24). These internally generated SD scores were preferred for this study over external growth reference curves (e.g., from the World Health Organization (25)) since they provided the best fit to our population.

Rapid growth from birth to age 6 months was defined as a z score weight gain greater than 0.67 × SD (26). Children with a z score weight gain equal to or below 0.67 × SD were characterized as slow/average growers and constituted the reference group. We categorized weight status at age 4 years according to the BMI cutoff points for sex and age proposed by the International Obesity Task Force definitions (27).

Waist circumference was measured in duplicate to the nearest 0.1 cm in the standing position, at the high point of the iliac crest at the end of a gentle expiration, using a flexible tape measure (Seca 201; Seca GmbH & Co. KG). Triceps, subscapular, suprailiac, and quadriceps skinfold thicknesses were measured according to standardized techniques to the nearest 0.1 mm using a calibrated Harpenden caliper (Harpenden HSK-BI CE₀₂₁₀; Baty International Ltd., Burgess Hill, United Kingdom). Three complete sets of measurements were taken consecutively, and the mean value was used as the representative value for each site. The sum of the 4 aforementioned skinfold measurements was calculated as an indicator of subcutaneous fat.

Child cardiometabolic traits

At the age-4-years examination, trained research assistants measured systolic and diastolic blood pressures on the child’s right arm after 5 minutes’ rest in the seated position, using an automatic oscillometric device (DINAMAP ProCare 400; GE Medical Systems, Information Technologies, Inc., Milwaukee, Wisconsin) with a cuff of the appropriate size for arm circumference. Five measurements were made at 1-minute intervals, and the average of all measurements was used for analysis (28). Nonfasting blood samples were collected from the children at the end of the visit in 10-mL gel separator Vacutainers (Becton Dickinson Medical Pharmaceutical Systems, Oxford, United Kingdom) with the use of standard procedures, and the samples were immediately spun, separated, and frozen at −80°C. Analyses of serum lipid concentrations (total cholesterol and high-density lipoprotein cholesterol) were performed via standard enzymatic methods (Medicon Hellas S.A., Athens, Greece) on an automatic analyzer (AU5400 high-volume chemistry analyzer; Olympus America, Inc., Melville, New York). Leptin level was measured via enzyme-linked immunosorbent assay (DLP00; R&D Systems, Inc., Minneapolis, Minnesota) on an automatic analyzer (MAGO Plus; Diamedix Corporation, Miami Lakes, Florida). C-reactive protein levels were measured with a high-sensitivity homogenous immunoassay (ORS 6199; Beckman Coulter Inc., Fullerton, California) on an automatic analyzer (AU5400 high-volume chemistry analyzer; Olympus America, Inc., Melville, New York). The limit of detection for C-reactive protein was equal to 0.002 mg/dL, and samples below the limit of detection (n = 20; 3.8%) were assigned the value of 0.5 × the limit of detection. All inter- and intraassay coefficients of variation were less than 5.5%.

Potential covariates

Personal interviews, together with self-administered questionnaires and medical records, were used to obtain information on possible confounding variables related to offspring outcomes. Maternal characteristics included: age at birth of the child, duration of education at recruitment (≤6 years, >6–≤12 years, or >12 years), national origin (Greek or non-Greek), parity (nulliparous or multiparous) before pregnancy, BMI before pregnancy, gestational diabetes, tobacco smoking during pregnancy (at 12th week of gestation; yes/no), and exposure to other pollutants during pregnancy (air pollution and organochlorine compounds). The children’s characteristics included: child sex (male/female), gestational age at birth (weeks), weight (g) at birth, and duration of breastfeeding (months).

Statistical analyses

Descriptive analyses of the study population’s characteristics, cadmium exposure, and outcomes were conducted. We fitted generalized additive models and visually assessed plotted splines to determine the linearity of exposure-outcome associations. We found evidence of nonlinear associations of prenatal cadmium exposure with outcomes (P-gain (defined as the difference in normalized deviance between the generalized additive models model and the linear model for the same exposure-outcome pair) greater than 0.1); therefore, cadmium exposure was analyzed categorically on the basis of tertiles (of cadmium concentration on the natural scale), where the third tertile was compared with the first and second tertiles combined. Because of right-skewed distributions, we log-transformed serum concentrations of leptin and C-reactive protein to satisfy model assumptions. For easier interpretation of association estimates for log-transformed outcomes, we exponentiated regression coefficients and report results as percent change (% change = [exp(β) − 1] × 100).

Multivariable-adjusted Poisson regression models with robust standard errors were used to estimate relative risks and 95% confidence intervals for the associations with binary outcomes (rapid growth and obesity), while linear regression models were applied to evaluate β coefficients and 95% confidence intervals for the associations with continuous outcome measures (BMI, waist circumference, sum of skinfolds, lipids, blood pressure, C-reactive protein, and leptin). Mixed-effects linear regression, fitted with a random child intercept and a random slope for child’s age, was used to compare growth trajectories between exposure groups (β coefficients and 95% confidence intervals) based on child BMI, weight, and height SD scores. The multivariable-adjusted models initially included the variables marginally related to outcomes in the bivariate models (P < 0.1), and thereafter variables were retained only if they had a P value less than 0.2 or modified the coefficient for cadmium by at least 10%. The final models adjusted for maternal age at delivery, maternal education, smoking during early pregnancy, prepregnancy BMI, parity, and the child’s sex and age at assessment. Statistical significance was set at P < 0.05 for all estimates.

To assess whether our studied associations were modified by child sex and maternal smoking during early pregnancy, we included appropriate multiplicative interaction terms in the regression models. We stratified the sample when significant effect modification was detected, indicating a significantly different estimate of the association of exposure with outcomes between groups (P for interaction < 0.05). Analyses were conducted using Stata software, version 13.0 (StataCorp LLC, College Station, Texas).

RESULTS

Participating mothers were predominantly of Greek origin (94%) and had a mean age of 29.8 (SD, 5.0) years at delivery (Table 1). Approximately 30% of the mothers had received more than 12 years of schooling, 59% were multiparous, and 16% were smokers during the first trimester of pregnancy. Fifty-two percent of the children included in the analysis were boys, and 51% were vaginally delivered. At birth, mean weight was 3,237 (SD, 446) g, and the average gestational age was 38.2 (SD, 1.5) weeks. At 4 years of age, approximately 15% of the children were overweight and 7% were obese, with an overall mean BMI of 16.4 (SD, 1.9) (Table 1). The median maternal urinary cadmium concentration in early pregnancy (median gestational week, 13) was 0.5 μg/L (interquartile range, 0.3–0.7).

Table 1.

Characteristics of Mothers and Their Children at Birth, in Infancy, and at 4 Years of Age, Rhea Mother-Child Cohort Study, Heraklion, Greece, 2007–2012

Population Characteristic	Child’s Prenatal Cadmium Exposure Status									P Value
	Unexposeda			Exposedb			Overall
	No.	%	Mean (SD)	No.	%	Mean (SD)	No.	%	Mean (SD)
Maternal characteristics
Age at child’s birth, years			29.0 (5.0)			31.5 (4.5)			29.8 (5.0)	<0.001
Prepregnancy BMIc			24.5 (4.4)			25.4 (5.3)			24.8 (4.7)	0.048
Educational level										0.011
Low (≤6 years)	49	14.3		38	22.1		87	16.9
Medium (>6–≤12 years)	176	51.3		94	54.7		270	52.4
High (>12 years)	118	34.4		40	23.3		158	30.7
Parity										0.004
Nulliparous	155	45.2		55	32.0		210	40.8
Multiparous	188	54.8		117	68.0		305	59.2
Smoking during pregnancy										0.003
No	298	86.9		132	76.7		430	83.5
Yes	45	13.1		40	23.3		85	16.5
Infant and child (age 4 years) characteristics
Sex										0.596
Male	180	52.5		86	50.0		266	51.7
Female	163	47.5		86	50.0		249	48.3
Rapid BMI growth at age 6 monthsd										0.603
No	155	68.6		75	65.8		230	67.6
Yes	71	31.4		39	34.2		110	32.4
Age at follow up, years			4.2 (0.2)			4.2 (0.2)			4.2 (0.2)	0.792
Waist circumference, cm			53.6 (4.9)			53.4 (5.1)			53.5 (5.0)	0.729
IOTF BMI category										0.808
Nonobese	321	93.6		160	93.0		481	93.4
Obese	22	6.4		12	7.0		34	6.6
Sum of skinfolds, mm			40.3 (13.5)			41.0 (15.0)			40.5 (14.0)	0.555
Total cholesterol level, mg/dL			156.6 (27.9)			155.0 (26.3)			156.1 (27.4)	0.543
HDL cholesterol level, mg/dL			49.3 (11.1)			49.4 (12.1)			49.4 (11.4)	0.933
C-reactive protein level, mg/dLe	0.1 (1.2–2.9)			0.1 (1.3–3.1)			0.1 (1.2–2.9)			0.908
Leptin level, ng/mLe	1.8 (0.0–0.1)			1.9 (0.0–0.1)			1.8 (0.0–0.1)			0.356
Systolic blood pressure, mm Hg			90.3 (8.3)			90.4 (7.3)			90.3 (8.0)	0.923
Diastolic blood pressure, mm Hg			53.2 (5.1)			53.5 (4.7)			53.3 (5.0)	0.513

Abbreviations: BMI, body mass index; HDL, high-density lipoprotein; IOTF, International Obesity Task Force; SD, standard deviation.

^a First and second tertiles combined (0.058–0.570 μg/L).

^b Third tertile (0.571–2.658 μg/L).

^c Weight (kg)/height (m)².

^d Defined as a gain in SD score for BMI > 0.67 SD.

^e Values are expressed as median (interquartile range).

Higher maternal urinary concentrations of cadmium in early pregnancy (third tertile (0.571–2.658 μg/L) vs. first and second tertiles combined (0.058–0.570 μg/L)) were significantly associated with a slower weight (z score) trajectory from age 3 months to age 4 years (adjusted β = −0.17, 95% confidence interval (CI): −0.32, −0.02; Table 2 and Figure 1). Similarly, higher urinary cadmium concentrations were associated with slower height and BMI growth trajectories, but none of the associations reached statistical significance (Table 2 and Figure 1). No associations were observed with any of the child adiposity measures or with cardiometabolic outcomes (lipids, blood pressure, C-reactive protein, and leptin; Table 2). Further analyses showed that child’s sex and maternal smoking during early pregnancy modified the association of maternal urinary cadmium exposure during pregnancy with child growth measures (Table 3 and Figure 2). The greatest adverse cadmium-related associations were found for height SD score trajectories in girls (β = 0.04 (95% CI: −0.16, 0.24) in boys vs. β = −0.30 (95% CI: −0.52, −0.09) in girls; P for interaction = 0.025) and in children born to mothers who smoked during pregnancy (β = −0.04 (95% CI: −0.20, 0.12) in nonsmokers vs. β = −0.48 (95% CI: −0.83, −1.13) in smokers; P for interaction = 0.027). Among children born to mothers who smoked during pregnancy, associations of maternal urinary cadmium with different growth measures were similar in boys and girls (P for interaction > 0.05).

Table 2.

Associations of Maternal Urinary Cadmium Concentrations During Pregnancy With the Offspring’s Growth, Adiposity, and Cardiometabolic Traits, Rhea Mother-Child Cohort Study, Heraklion, Greece, 2007–2012

Outcome	Association With Maternal Urinary Cadmium Concentrationa,b
	β	95% CI
Growth from age 3 months to age 4 years
Weight (SD score)c	−0.17	−0.32, −0.02
Height (SD score)c	−0.11	−0.26, 0.04
BMId (SD score)c	−0.12	−0.26, 0.01
Rapid BMI growth at age 6 monthse,f	1.01	0.80, 1.50
Adiposity at age 4 years
Waist circumference, cm	−0.42	−1.36, 0.52
Sum of skinfolds, mm	−0.02	−2.62, 2.58
Obesityf,g	0.91	0.44, 1.88
Cardiometabolic traits at age 4 years
Total cholesterol level, mg/dL	−3.07	−8.79, 2.64
HDL cholesterol level, mg/dL	−0.07	−2.43, 2.28
Systolic blood pressure, mm Hg (SD score)	0.01	−0.14, 0.16
Diastolic blood pressure, mm Hg (SD score)	0.04	−0.06, 0.13
C-reactive protein level, mg/dLh	−11.0	−33.0, 18.0
Leptin level, ng/mLh	0.6	−15.0, 10.0

Abbreviations: BMI, body mass index; CI, confidence interval; HDL, high-density lipoprotein; SD, standard deviation.

^a Third tertile of cadmium concentration (0.571–2.658 μg/L) vs. first and second tertiles (0.058–0.314 μg/L and 0.315–0.570 μg/L, respectively) combined.

^b Results were adjusted for child sex, age at assessment (years), maternal age (years), smoking during pregnancy (yes/no), prepregnancy BMI, maternal education (low (≤6 years), medium (>6–≤12 years)), or high (>12 years)), and parity (primiparous/multiparous).

^c Mixed model regression estimate.

^d Weight (kg)/height (m)².

^e Defined as a gain in SD score for BMI > 0.67 SD.

^f Value is expressed as relative risk (95% CI).

^g Defined using the BMI cutoff point for sex and age proposed by the International Obesity Task Force (26).

^h Values are expressed as percent change (95% CI).

Figure 1.

Associations of maternal urinary cadmium concentrations during pregnancy with the offspring’s growth trajectories, Rhea Mother-Child Cohort Study, Heraklion, Greece, 2007–2012. The graph shows height (A), weight (B), and body mass index (BMI; weight (kg)/height (m)²) (C) standard deviation (SD) score trajectories from 3 months to 4 years of age for low (first and second tertiles, ≤0.57 μg/L) and high (third tertile, >0.57 μg/L) urinary cadmium concentrations during early pregnancy. Results were adjusted for child sex, age at assessment (years), maternal age (years), smoking during pregnancy (yes/no), prepregnancy BMI, maternal education (low (≤6 years), medium (>6–≤12 years), or high (>12 years)), and parity (primiparous/multiparous).

Table 3.

Associations of Maternal Urinary Cadmium Concentrations During Pregnancy With the Offspring’s Growth, According to Offspring’s Sex and Maternal Smoking During Early Pregnancy, Rhea Mother-Child Cohort Study, Heraklion, Greece, 2007–2012

Growth Trajectory	Association With Maternal Urinary Cadmium Concentrationa,b
	Offspring’s Sex					Maternal Smoking Status
	Male		Female		P for Interaction	Nonsmoker		Smoker		P for Interaction
	β	95% CI	β	95% CI		β	95% CI	β	95% CI
Weight (SD score)	−0.05	−0.25, 0.16	−0.30	−0.52, −0.09	0.134	−0.13	−0.30, 0.03	−0.34	−0.70, 0.02	0.422
Height (SD score)	0.04	−0.16, 0.24	−0.30	−0.52, −0.09	0.025	−0.04	−0.20, 0.12	−0.48	−0.83, −1.13	0.027
BMIc (SD score)	−0.12	−0.30, 0.07	−0.10	−0.30, 0.09	0.741	−0.14	−0.29, 0.01	−0.04	−0.37, 0.30	0.431

Abbreviations: BMI, body mass index; CI, confidence interval; SD, standard deviation.

^a Third tertile of cadmium concentration (0.571–2.658 μg/L) vs. first and second tertiles (0.058–0.314 μg/L and 0.315–0.570 μg/L, respectively) combined.

^b Estimates were obtained using mixed-effects models, with results adjusted for child sex, age at assessment (years), maternal age (years), smoking during pregnancy (yes/no), prepregnancy BMI, maternal education (low (≤6 years), medium (>6–≤12 years)), or high (>12 years)), and parity (primiparous/multiparous).

^c Weight (kg)/height (m)².

Figure 2.

Associations of maternal urinary cadmium concentrations during pregnancy with the offspring’s height trajectories, by sex and maternal smoking during early pregnancy, Rhea Mother-Child Cohort Study, Heraklion, Greece, 2007–2012. The graph shows height standard deviation (SD) score trajectories from 3 months to 4 years of age for low (first and second tertiles, ≤0.57 μg/L) and high (third tertile, >0.57 μg/L) urinary cadmium concentrations during early pregnancy for boys (A), girls (B), children born to mothers who did not smoke during pregnancy (C), and children whose mothers smoked during pregnancy (D). Results were adjusted for child sex, age at assessment (years), maternal age (years), smoking during pregnancy (yes/no), prepregnancy body mass index (weight (kg)/height (m)²), maternal education (low (≤6 years), medium (>6–≤12 years), or high (>12 years)), and parity (primiparous/multiparous).

DISCUSSION

In this study, we found that elevated prenatal cadmium exposure was associated with delayed childhood growth patterns, especially height, and that the associations were stronger in girls and in children born to mothers who smoked during the first trimester of pregnancy. Prenatal cadmium exposure did not appear to be associated with child adiposity or cardiometabolic health.

Our results are consistent with the findings of a few previous studies that have evaluated the assocation between prenatal cadmium exposure and child growth (14–16), some of which also showed that the differences in attained weight and height in childhood were apparent only in girls (14, 15). In a longitudinal study of 1,505 5-year-old children in rural Bangladesh, girls’ lifelong cadmium exposure, including maternal urinary cadmium in pregnancy, was inversely associated with attained weight and height; girls in the 95th percentile of exposure were about 1 kg lighter and 2.2 cm shorter than those in the fifth percentile of exposure (14). Similarly, in a study of 114 Flemish children aged 7–9 years by Delvaux et al. (15), umbilical cord blood cadmium concentrations were inversely associated with girls’ weight, BMI, waist circumference, and sum of skinfolds. We did not find any significant association with waist circumference or sum of skinfolds, but this may be attributed to the fact that we assessed these outcomes much earlier in life.

The mechanisms behind the prenatal cadmium-related sex differences in relation to later childhood growth are not clear. Cadmium accumulates in the placenta (21), which may affect the sexually dimorphic epigenetic pattern in placenta and cord blood (29). In the placenta, sexual epigenetic dimorphism has been observed in genes encoding proteins involved in growth/transcription factor signaling, immune function, and membrane transport (30). In a small study of 24 mother-infant pairs, sex-specific associations between cadmium in placenta and DNA methylation in placenta were observed (31). In 319 women from the Newborn Epigenetic Study, elevated maternal blood cadmium concentration was associated with lower birth weight, and results suggested that an association between prenatal cadmium exposure and lower offspring DNA methylation at regulatory sequences of imprinted genes may be sex- and gene-specific (32). In another study of 127 mother-child pairs in Bangladesh, maternal blood cadmium concentrations during pregnancy showed sex-specific associations with DNA methylation in cord blood (33). In girls, cadmium appeared to mainly affect genes (mostly hypomethylated) related to organ development, morphology, and bone mineralization, while in boys, cadmium affected genes (mostly hypermethylated) encoding for cell death. The cadmium-affected sites were also associated with lower birth weight. In addition, some of the sites were also affected in blood among 4.5-year-old children, suggesting that these epigenetic alterations induced in utero persist in later childhood. Further large-scale longitudinal studies are needed, exploring the persistency of early-life epigenetic alterations and their potential relationship to childhood growth and development.

Our results suggested that the adverse association of prenatal cadmium exposure with child growth was greater among children whose mothers smoked. This was most likely due to the higher cadmium exposure in mothers who smoked during pregnancy (median urinary cadmium concentration, 0.55 μg/L; range, 0.12–2.66 μg/L) in comparison with the nonsmoking mothers (median, 0.40 μg/L; range, 0.06–2.37 μg/L), as tobacco smoke is a major source of cadmium (20). Obviously, the role of smoking in cadmium toxicity is complex, and we cannot totally exclude the possibility that cadmium may be a proxy for a combined effect of multiple toxicants in tobacco smoke on child growth. On the other hand, numerous studies and meta-analyses have shown that maternal smoking during pregnancy is rather more associated with obesity in the offspring than impaired growth (34–36). Moreover, recent studies have found no adverse association of maternal smoking during pregnancy with childhood bone mass; if anything, prenatal exposure was associated with increased bone mass, probably related to childhood adiposity, not the intrauterine mechanism of skeletal growth (37, 38). Cadmium exposure in 10-year-old children has been associated with increased bone resorption and demineralization, suggesting that cadmium may affect bone tissue from an early age (39). A negative influence on bone development may translate to impaired bone growth in children and thus explain the observed reduction in attained height among children of mothers who smoke. Unfortunately, our sample size did not allow us to evaluate the combined interaction of cadmium, child sex, and maternal smoking.

Strengths of our study include the population-based prospective design, the fairly large sample size, the longitudinally collected growth data from early infancy through childhood, and the detailed childhood body fat and cardiometabolic measurements. Unlike previous studies, we considered a set of different anthropometric parameters, not only height and weight but also waist circumference and skinfolds, which made it possible to distinguish markers of subcutaneous fat from markers of abdominal fat. In contrast to studies that have relied only on umbilical cord blood levels, we assessed individual maternal cadmium exposure in the first half of pregnancy, a critical time window for organogenesis, by means of cadmium concentrations in urine, a well-recognized biomarker of chronic cadmium exposure (20). Urinary cadmium concentrations were adjusted for specific gravity, because creatinine adjustment is highly dependent on body composition and type of diet (22). Use of a population that is fairly homogenous with regard to factors such as diet, country of origin, and socioeconomic status can help reduce uncontrolled confounding. The level of attrition in the Rhea cohort is similar to that found in other birth cohort studies (40, 41), and we found no evidence of differences in prenatal cadmium exposure between participants and nonparticipants.

The study also had some limitations. We studied a broad spectrum of outcomes, and chance findings are always of concern when multiple comparisons are performed; however, because the outcomes were not independent, Bonferroni correction or a similar type of correction for multiple testing would be overly conservative (42, 43). Finally, we acknowledge that although we incorporated extensive information on social and environmental factors, residual confounding due to other unmeasured confounders, such family income and characteristics of the home environment, may still have occurred.

In conclusion, this study showed that elevated maternal cadmium exposure during pregnancy was associated with delayed growth in early childhood. Further research is needed to understand cadmium-related sex differences and the role of coexposure to maternal smoking during pregnancy.

Abbreviations

BMI

body mass index

CI

confidence interval

SD

standard deviation.