Association Between Gestational Urinary Bisphenol A Concentrations and Adiposity in Young Children: The MIREC Study

Abstract

Background:

Bisphenol A (BPA) is a high production volume chemical and because of its use in many consumer products, exposure is ubiquitous. Gestational BPA exposure has been associated with excess adiposity in rodent studies, but not consistently in human studies. We investigated the relation between gestational BPA exposure and early childhood adiposity in a prospective cohort study of 719 mother-child pairs.

Methods:

We used data from the MIREC Study, a prospective Pan-Canadian pregnancy and birth cohort study. We measured BPA in urine samples collected at an average of 12.1 weeks (range: 6.3–15 weeks) gestation and measured children’s weight, height, waist/hip circumference, and subscapular/triceps skinfold thickness at an average age of 3.5 years (range: 1.9–6.2). We estimated covariate-adjusted associations of log₂-transformed BPA concentrations with child adiposity measures and examined whether these associations differed in boys and girls.

Results:

Median BPA concentrations were 0.8 ng/mL (IQR: 0.5–1.4). Among both boys and girls, each 2-fold increase in BPA concentrations was associated with higher waist-to-hip ratio (β: 0.003; 95% CI: 0.001, 0.005). The association of BPA with waist circumference and subscapular skinfold thickness was modified by sex (sex x BPA interaction p-values<0.2). In girls, each 2-fold increase in BPA concentrations was associated with a 0.2 cm (95% CI: 0.0, 0.5) and 0.15 mm (95% CI: 0.01, 0.30) increase in waist circumference and subscapular skinfolds, respectively. Associations were generally null or slightly inverse in boys.

Conclusions:

In this cohort, gestational urinary BPA concentrations were associated with subtle increases in girl’s central adiposity during early childhood.

Introduction

Environmental chemical exposures during sensitive periods of development may increase the risk of childhood obesity and cardiometabolic disorders. Experimental studies in animals show that gestational exposure to chemical obesogens can affect adipogenesis or reprogram neuroendocrine circuits involved in energy metabolism and eating behaviors, which in turn could affect offspring weight gain and adiposity.(1, 2) Some observational studies in humans show that gestational chemical exposures increase childhood obesity risk and may do so by affecting energy metabolism.(3–5) It is critical to determine if early life environmental chemical exposures increase the risk of childhood obesity given the ubiquity of these exposures,(6) the high global prevalence of childhood overweight or obesity (41 million children),(7) and difficulty in effectively reducing excess weight once it is acquired.(8)

Bisphenol A (BPA) is a suspected chemical obesogen used to produce polycarbonate plastics and resins that are found in some medical/dental equipment, consumer products, and food packaging. Exposure to BPA is ubiquitous among pregnant women in many countries and potential sources of exposure include handling thermal receipts and consumption of packaged, processed, or canned food and beverages.(9–16) A meta-analysis of 61 experimental studies in rodents found that prenatal or lactational exposure to BPA caused decreased body weight and increased fat pad weight.(17) Epidemiological studies (n’s=173 to 520) have reported inconsistent associations between gestational urinary BPA concentrations and childhood obesity; three found that BPA was associated with lower BMI,(18–20) two reported that BPA was associated with increased adiposity or risk of being overweight or obese,(21, 22) and three others reported no association.(23–25) In some rodent and human studies, the relation between BPA exposure and adiposity differed across the sexes. (17) (18–20, 22)

Given that prior epidemiological studies reported inconsistent and sex-specific associations between gestational BPA exposure and childhood obesity, there is a need to examine this association with larger studies that can precisely estimate potential sex-specific effects. Thus, we investigated the association of maternal urinary BPA concentrations during pregnancy with early childhood adiposity in 719 mother-child pairs from Canada.

Methods

Study Participants

We used data from the Maternal-Infant Research on Environmental Chemicals (MIREC) study, a prospective Pan-Canadian cohort that recruited women in the first trimester of pregnancy from obstetric and prenatal clinics in ten cities (11 study sites) from 2008–2011 (see Arbuckle et al., 2013 for details about eligibility, recruitment, and follow-up).(26) Briefly, eligibility criteria included the ability to consent and communicate in English or French, <14 weeks gestation at enrollment, >18 years of age, plans to deliver at a local hospital, and agree to participate in the cord blood collection component of the MIREC study. We excluded women who were carrying a fetus with a known malformation or abnormality and those with a history of major chronic disease, threatened abortion, or illicit drug use. Of the 8,716 women we approached to participate in MIREC, 5,108 (58.6%) were eligible, 2,001 (39.2%) consented to participate, and 1,861 had singleton live births.

We obtained funding to conduct follow-up in 6 of the 10 cities with the largest number of participants to ensure that we could recruit our target sample size of approximately 800 singleton children (1,574 singletons eligible at Vancouver, Toronto, Hamilton, Montreal, Kingston, and Halifax study sites). A total of 783 singleton children born to women at these MIREC study sites completed an in-person follow-up when they were on average 3.5 years old (range: 1.9–6.2) (defined as early childhood hereafter). This in-person visit typically occurred in the participant’s home where trained research personnel collected urine and blood from the child, administered questionnaires, and measured child anthropometry. A total of 719 mother-child pairs were included in our primary analysis after excluding women missing urinary BPA concentrations (n=17) and covariates (n=45).

This research was approved by ethics review boards from Health Canada and CHU Sainte-Justine, and ethics committees at the participating hospitals. Potential participants were provided with information about the objectives and design of the study and asked to sign the informed consent forms for both the prenatal and early childhood follow-up portions of the study.

Gestational BPA Exposure Assessment

After collecting a single urine sample from women at an average of 12.1 weeks gestation (range: 6.3–15 weeks), we aliquoted the sample and froze the aliquots at −20 °C within 2 hours of collection before shipping them on dry ice to the MIREC coordinating center in Montreal where they were stored at −30 °C. Urine samples were shipped in batches to the Centre de Toxicologie du Québec, Institut National de Santé Publique du Québec for analysis. Total (conjugated + free) concentrations of BPA were quantified using previously described analytic chemistry methods.(11) The limit of detection (LOD) for this method was 0.2 ng/mL and we assigned values below the LOD a value of 0.1 ng/mL.

We accounted for individual variation in urine dilution by measuring urine specific gravity (SG) with a refractometer and standardizing urinary BPA concentrations using a formula adapted from Duty et al. (27):

$$P_S=P_i\left(\frac{S{G}_m-1}{S{G}_i-1}\right)$$

where P_s is the SG-standardized BPA concentration, P_i is the observed BPA concentration for the i-th woman, SG_m is the median SG (1.013), and SG_i is the observed SG for the i-th woman. We used SG-standardized concentrations in all of our analyses. Additional urine samples were collected in the 2^nd and 3^rd trimesters, but BPA was not quantified in these due to resource constraints.

Child Anthropometry Assessments

We measured children’s weight without shoes to the nearest 50 g using a Seca 874 scale. Using a Seca 216 stadiometer, we measured children’s height without shoes or head coverings to the nearest 0.1 cm. We measured children’s waist and hip circumference to the nearest 0.1 cm with a Seca 201 measuring tape at the mid-point between the upper iliac crest-lowest rib and maximum protuberance of the buttocks, respectively. Finally, we measured children’s subscapular and triceps skinfold thickness to the nearest 0.1 mm using a Lange skinfold caliper. All measures were taken in duplicate by a single research assistant, unless they differed by more than a pre-specified value, in which case a 3^rd measurement was taken. We took the average of two or three anthropometry measurements and used these for our analyses.

Using World Health Organization (WHO) references, we calculated children’s age- and sex-specific BMI, subscapular skinfold thickness, and triceps skinfold thickness z-scores.(28–30) In addition, we calculated waist-to-hip circumference ratio, subscapular-to-triceps skinfold thickness ratio, and sum of subscapular and triceps skinfold thickness.

Covariates

We considered adjusting for numerous factors that might confound the association between gestational urinary BPA concentrations and childhood adiposity. Trained interviewers administered standardized questionnaires at baseline, when urine samples were collected, to assess maternal age at delivery, race, education, marital status, employment, and pre-pregnancy BMI, as well as household income, self-reported smoking and alcohol consumption during pregnancy, and parity. During the 2^nd trimester, women completed standardized questionnaires assessing multivitamin and supplement use. We abstracted medical chart information to assess diagnosis of gestational diabetes or impaired glucose tolerance. During a study visit in the 3^rd trimester, trained staff measured women’s weight. We calculated gestational weight gain by taking the difference between weight at last visit prior to delivery and self-reported pre-pregnancy weight, and converting it to weight gain for gestational age z-scores according to pre-pregnancy BMI categories.(31) At the time of the childhood assessment, mothers completed standardized questionnaires that measured the duration of breastfeeding. We also measured father’s weight and height. Finally, we adjusted for study site since urinary BPA concentrations and measurement of children’s adiposity might vary across the study sites.

We identified confounders by using directed acyclic graphs (DAGs) to select variables that were associated with both urinary BPA concentrations and childhood adiposity, but not those that were causal intermediates or colliders (Supplemental Figure 1).(11) In our final model, we adjusted for maternal race, education, age, marital status, smoking during pregnancy, pre-pregnancy BMI, household income, and study center. We additionally adjusted for child sex and age at assessment in models of waist and hip circumference and skinfold thickness to reduce the variance in these measures related to child sex and age. We conducted a complete case analysis because <10% of participants were missing covariate information. Household income, pre-pregnancy BMI, and maternal education were missing among 25, 18, and 2 women, respectively.

Statistical Methods

We began our statistical analyses by describing the central tendency and distribution of maternal urinary BPA concentrations and children’s BMI z-scores by covariates. We examined the shape of the relationship between log₂-transformed urinary BPA concentrations and adiposity measures using 3-knot restricted cubic polynomial splines.(32) We considered nonlinearity to be present when the p-value for the cubic polynomial spline term was <0.2.

We used multivariable linear regression to estimate the covariate-adjusted association between log₂-transformed urinary BPA concentrations and adiposity measures. In addition, we examined quintiles of urinary BPA concentrations in relationship to adiposity measures to further characterize the dose-response relationship. We used product interaction terms between child sex and gestational urinary BPA concentrations to estimate sex-specific associations and test the difference in these associations between boys and girls. We considered modification by child sex to be present when the p-value for this interaction term was <0.2.(33) Finally, because some prior experimental studies in animals suggest that folic acid can mitigate the obesogenic effects of BPA, we examined whether maternal-reported folic acid supplement use during the 1^st trimester modified the association between BPA concentrations and adiposity measures.(34, 35) All analyses were conducted using SAS version 9.4.

Secondary and Sensitivity Analyses

Because urinary BPA concentrations exhibit substantial within-person variation, we used regression disattenuation to estimate the association between urinary BPA concentrations and early childhood adiposity measures after correcting for non-differential BPA exposure measurement error.(36) We used the following formula:

$${\beta }_{\text{corr}}=\frac{{\beta }_{\text{obs}}}{ICC}$$

where β_corr is the measurement error corrected association, β_obs is the observed and uncorrected association, and ICC is the intraclass correlation coefficient for BPA. We assumed that the ICC was 0.2 based on the range of values reported in previous studies of pregnant women from North America (0.1 to 0.3).(12, 37, 38)

We conducted a series of sensitivity analyses to assess the robustness of our results to various confounder adjustments and assumptions. First, we compared unadjusted and covariate-adjusted models to estimate the magnitude of confounding present in our results. Second, we adjusted for all covariates except study site. Third, we adjusted for breastfeeding duration, weight gain during pregnancy, gestational diabetes or impaired glucose tolerance diagnosis, paternal BMI, gestational age at the time of urine collection, and child daycare attendance. Fourth, we used age- and sex-standardized subscapular and triceps skinfold measures as our outcomes instead of adjusting for child sex and age when examining these outcomes. Fifth, we examined whether our sex-specific results differed when we stratified by child sex instead of using product interaction terms.(39) Finally, we examined whether the association between BPA and child adiposity differed by child age because of the age range of children included pre-school and school-aged children.

Results

Women whose children completed early childhood follow-up were relatively similar to those who initially enrolled at the six study sites with regard to baseline covariates, but slightly more likely to be >35 years of age, graduate- or university-educated, non-smokers during pregnancy, and give birth to female infants (Supplemental Table 1, χ² p-values<0.1).BPA was detected in the urine of 620 women (86.2%). SG-standardized maternal urinary BPA concentrations ranged from 0.1 to 63 ng/mL (median 0.8 ng/mL) (Supplemental Table 2) and were not substantially different between women whose children did and did not complete follow-up (medians [25^th, 75^th]: 0.8 [0.3, 1.6] vs. 0.8 [0.3, 1.7] ng/mL). Median maternal urinary BPA concentrations did not vary by >0.5 ng/mL across categories of maternal sociodemographic or perinatal factors (Table 1).

Table 1:

Central tendency and variation of maternal urinary BPA concentrations during pregnancy and child BMI z-score in early childhood according to covariates: MIREC Study

Variable		BPA Median		BMI Mean
	N	(25th, 75th)	N	(SD)
Overall	719	0.8 (0.5, 1.4)	713	0.51 (0.91)
Maternal Age
18–25 years	23	1.3 (0.6, 2.7)	22	0.51 (0.95)
>25–35 years	427	0.8 (0.5, 1.4)	424	0.52 (0.9)
>35 years	269	0.8 (0.4, 1.3)	267	0.48 (0.92)
Maternal Race/Ethnicity
White	606	0.8 (0.5, 1.4)	601	0.49 (0.87)
Asian/Pacific Islander	30	0.5 (0.3, 1.2)	30	0.5 (0.97)
Other	52	0.9 (0.5, 1.5)	51	0.59 (1.11)
Multi-Racial/Ethnic	31	0.8 (0.5, 1.8)	31	0.58 (1.18)
Maternal Education
Graduate Degree	191	0.7 (0.4, 1.2)	189	0.54 (0.85)
University Degree	300	0.9 (0.5, 1.4)	297	0.51 (0.88)
Some College, Trade School, or Diploma	187	0.9 (0.5, 1.8)	186	0.49 (1.02)
High School or Less	41	0.9 (0.6, 1.4)	41	0.42 (0.86)
Marital Status
Married or Living with Partner	697	0.8 (0.5, 1.4)	691	0.51 (0.91)
Not Married or Living Alone	22	1.1 (0.6, 2)	22	0.39 (0.66)
Household Income (CAD)
>$100K	301	0.8 (0.5, 1.3)	299	0.45 (0.86)
$80K–100K	224	0.9 (0.4, 1.4)	221	0.52 (0.86)
$40K–<80K	125	0.8 (0.5, 2)	124	0.57 (0.96)
<$40K	69	1 (0.6, 2)	69	0.58 (1.14)
Maternal Pre-Pregnancy BMI
Underweight (<18.5)	20	0.9 (0.6, 1.6)	20	0.17 (0.86)
Normal Weight (18.5 to <25)	440	0.7 (0.4, 1.3)	436	0.38 (0.84)
Overweight (25 to <30)	145	1 (0.5, 1.5)	145	0.65 (0.93)
Obese (>30)	114	0.9 (0.6, 2)	112	0.86 (1.02)
Maternal Smoking During Pregnancy
No	696	0.8 (0.5, 1.4)	690	0.49 (0.88)
Yes	23	0.8 (0.6, 1.1)	23	0.97 (1.37)
Prenatal Folic Acid Supplement Usea
No	69	0.8 (0.4, 1.4)	69	0.70 (1.02)
Yes	597	0.8 (0.4, 1.4)	591	0.49 (0.90)
Child Sex
Male	356	0.9 (0.5, 1.5)	353	0.56 (0.94)
Female	363	0.8 (0.4, 1.4)	360	0.46 (0.88)

BPA: Bisphenol A, BMI: Body Mass Index, SD: Standard Deviation

^a-Sample size is lower because folic acid supplement use was assessed after the 1^st trimester visit and some women did not complete the questionnaire.

We assessed childhood adiposity at an average of 3.5 years (Median: 3.9, Standard Deviation: 0.9, range: 1.9–6.2). Children’s BMI z-scores were slightly higher than the WHO standard population (mean: 0.5 Standard Deviation Scores, Supplemental Table 3). A total of 159 (22.3%), 32 (4.5%), and 5 (0.7%) children had BMI z-scores >1, >2, and >3 Standard Deviation Scores, respectively. Adiposity measures were moderately to strongly correlated, with waist and hip circumferences having the strongest pairwise correlation among individual adiposity measures (Pearson r=0.81) (Supplemental Table 4). On average, child BMI z-scores were >0.1 Standard Deviation Scores higher among children born to women who had higher education, lower income, higher pre-pregnancy BMI, and those who were married, other or multi-racial/ethnic, not taking prenatal folic acid supplements, and reported smoking during pregnancy (Table 1).

Among all children and after adjustment for covariates, urinary BPA concentrations were positively associated with small increases in waist circumference and waist-to-hip circumference ratio (Table 2), but not with other measures of adiposity. Child sex suggestively modified the associations of maternal urinary BPA concentrations with child waist circumference, subscapular skinfold thickness, and sum of skinfolds (sex x BPA interaction p-values<0.2; Table 2). Each doubling of BPA concentrations was associated with a 0.2 cm (95% CI: 0, 0.5) and 0.15 mm (95% CI: 0.01, 0.30) increase in waist circumference and subscapular skinfold thickness among girls, respectively (Table 2). Among boys, the associations of BPA with waist circumference and subscapular skinfolds were null and slightly inverse, respectively (Table 2).

Table 2:

Unadjusted and adjusted difference in child adiposity measures in early childhood with 2-fold increase in specific gravity standardized maternal urinary BPA concentrations during pregnancy: MIREC Study^a

Anthropometry Measure	N	Unadjusted All Children (95% CI)	Adjusted All Children (95% CI)	Adjusted Boys (95% CI)b	Adjusted Girls (95% CI)b	BPA x Sex p-value
BMI Z-score	713	0.00 (−0.04, 0.05)	0.00 (−0.04, 0.05)	−0.03 (−0.10, 0.04)	0.02 (−0.03, 0.08)	0.21
Waist Circumference (cm)	687	0.28 (0.07, 0.49)	0.14 (−0.03, 0.31)	0.00 (−0.26, 0.26)	0.2 (0.0, 0.5)	0.16
Hip Circumference (cm)	684	0.18 (−0.06, 0.41)	−0.02 (−0.19, 0.15)	−0.09 (−0.36, 0.17)	0.0 (−0.2, 0.3)	0.49
Waist-to-Hip Ratio	683	0.002 (0.000, 0.005)	0.003 (0.001, 0.005)	0.002 (−0.001, 0.006)	0.004 (0.001, 0.007)	0.47
Subscapular Skinfold Thickness (mm)	673	−0.01 (−0.13, 0.11)	0.04 (−0.07, 0.15)	−0.11 (−0.28, 0.06)	0.15 (0.01, 0.30)	0.02
Tricep Skinfold Thickness (mm)	680	−0.14 (−0.33, 0.05)	−0.06 (−0.21, 0.09)	−0.07 (−0.30, 0.17)	−0.05 (−0.25, 0.15)	0.93
Skinfold Sum (mm)	672	−0.15 (−0.42, 0.13)	−0.01 (−0.24, 0.21)	−0.18 (−0.53, 0.16)	0.11 (−0.18, 0.41)	0.20
Subscapular: Triceps Skinfold Ratio	672	0.00 (−0.01, 0.02)	0.00 (−0.01, 0.02)	0.00 (−0.02, 0.01)	0.01 (0.00, 0.03)	0.21

^*-BMI: Body Mass Index.

^a-Adjusted for maternal race/ethnicity (White, Asian/Pacific Islander, Other, vs. Multi-Racial), education (Graduate Degree, University Degree, Some College/Trade School/Diploma, vs. High School or Less), age (continuous, years), marital status (Married or Living with Partner vs. Not Married or Living Alone), household income (>$100,000, >$80,000–100,00, >$40,000–80,000, vs <$40,000 CAD) smoking during pregnancy (any vs. none), maternal pre-pregnancy BMI (<18.5, 18.5 to <25, 25 to <30, vs. >30 kg/m²), and study center (categorical). Waist and hip circumference and skinfold measures were additionally adjusted for child sex (male vs. female and age (continuous, years) at assessment.

^b-Sex-specific associations are derived from a model including the above listed covariates, child sex, and an interaction term between log₂-transformed BPA and child sex.

Among all children, there was some evidence of a J-shaped relation of BPA with waist circumference and subscapular skinfold thickness (non-linearity p-values=0.17 and 0.08, respectively) (Figure 1). However, these J-shaped relations were only observed in girls, where mean waist circumference and subscapular skinfold thickness decreased slightly up to BPA concentrations of 1 ng/mL and increased more dramatically thereafter (non-linearity p-values=0.22 and 0.19) (Figure 1). Differences in covariate-adjusted mean waist circumference and subscapular skinfold thickness among girls were greatest between the 5^th and 2^nd (reference group) quintiles of urinary BPA concentrations (waist circumference: 1.1 cm; 95% confidence interval [CI]: −0.1, 2.2 and subscapular skinfolds: 0.7; 95% CI: 0.0, 1.4) (Supplemental Table 5). Maternal intake of folic acid did not modify the association between maternal BPA and childhood adiposity (interaction p-values>0.2) (results not shown).

Figure 1:

Adjusted and smoothed association of maternal urinary BPA concentrations during pregnancy with child adiposity measures in early childhood among MIREC Study participants: Stratified by child sex ^a,b

a- Smoothed was done using a 3-knot restricted cubic polynomial spline of urinary BPA concentrations.

b-Adjusted for maternal race/ethnicity (White, Asian/Pacific Islander, Other, vs. Multi-Racial), education (Graduate Degree, University Degree, Some College/Trade School/Diploma, vs. High School or Less), age (continuous, years), marital status (Married or Living with Partner vs. Not Married or Living Alone), household income (>$100,000, >$80,000–100,00, >$40,000–80,000, vs <$40,000 CAD) smoking during pregnancy (any vs. none), maternal pre-pregnancy BMI (<18.5, 18.5 to <25, 25 to <30, vs. >30 kg/m²), and study center (categorical). Waist and hip circumference and skinfold measures were additionally adjusted for child sex (male vs. female and age (continuous, years) at assessment.

c-Non-linearity p-values for BMI, subscapular skinfold thickness, triceps skinfold thickness, waist circumference, and hip circumference among all children/boys/girls were 0.13/0.34/0.28, 0.08/0.52/0.19, 0.55/0.19/0.82, 0.17/0.73/0.21, and 0.41/0.95/0.57, respectively.

Secondary and Sensitivity Analyses

The pattern of associations between urinary BPA concentrations and early childhood adiposity were similar after correcting for non-differential BPA exposure measurement error (Supplemental Table 6). Compared to non-corrected associations, corrected associations of BPA with girl’s waist circumference, waist-to-hip ratio, and subscapular skinfolds were larger, but less precise (Supplemental Figure 3). For example, each 2-fold increase in urinary BPA concentrations was associated with a 1.21 cm (95% CI: 0.11, 2.30) increase in waist circumference with correction and a 0.24 cm (95% CI: 0.02, 0.46) increase without correction.

The associations of BPA with waist circumference and subscapular skinfolds among girls were similar when we excluded women who had gestational diabetes or impaired glucose tolerance during pregnancy or adjusted for gestational weight gain, breastfeeding, paternal BMI, gestational age at the time of urine collection, or child daycare attendance (Supplemental Table 7). Our results were unchanged when we used WHO subscapular and triceps skinfold z-scores instead of age- and sex-adjusted measures (Supplemental Table 8).

The associations of BPA with BMI, waist and hip circumference, waist-to-hip ratio, and subscapular skinfolds thickness was similar in younger (1.9–3.9 years) and older (3.9–6.2 years) children (Supplemental Table 9, BPA x age interaction p-values>0.35). However, BPA was inversely associated with triceps skinfolds thickness in older children (β per doubling in BPA: −0.45; 95% CI: −0.81, −0.09), but not younger children (β per doubling of BPA: 0.10; 95% CI: −0.18, 0.37) (BPA x age interaction p-value=0.009). The association of BPA with waist circumference and subscapular skinfolds thickness was similar in both younger and older girls (BPA x age interaction p-values>0.75).

Discussion

In this prospective pregnancy and birth cohort, maternal urinary BPA concentrations were associated with small increases in waist-to-hip ratio, waist circumference, and subscapular skinfold thickness in girls, but not boys. These associations in girls were stronger at higher levels of urinary BPA concentrations; there were no associations between BPA and these adiposity measures at BPA concentrations below the median, but positive associations at BPA concentrations greater than the median. Associations between BPA and adiposity measures in boys were generally null or slightly inverse.

Eight prior studies using prospective cohorts with sample sizes ranging from 173 to 520 have observed inconsistent associations between gestational BPA exposure and excess childhood adiposity or risk of childhood overweight/obesity (Table 3).(18–25) Three studies reported no associations between gestational urinary BPA concentrations and child adiposity measures (n’s=173, 194, and 520).(23–25) Three other studies reported that higher gestational urinary BPA concentrations were associated with lower BMI and that these associations were stronger in girls (n’s=235, 297, and 311).(18–20) Finally, another two studies reported that gestational urinary BPA concentrations were associated with increased waist circumference, BMI, and risk of being overweight or obese (n’s=300 and 402).(21, 22) In one of these studies, the positive association between gestational urinary BPA concentrations and adiposity at age 7 years was stronger in girls than boys (n=300).(22)

Table 3:

Summary of literature examining the association of gestational urinary BPA concentrations and childhood adiposity

Study	Location	N	BPA Central Tendency (ng/mL)	Number of urines	Urine Timing	Adiposity Measure	Age at Adiposity Assessment	Results
Braun et al. 2014	Ohio, USA	297	Median: 2.1; IQR: 1.1, 3.9	2	2nd/3rd	BMI	2–5 years	Inverse in girls. Non-significant sex x BPA interaction p-value.
Harley et al. 2013	Californi a, USA	311	Median: 1.1 IQR: 0.5, 1.9	2	2nd/3rd	BIA, BMI, WC, SF	9 years	Inverse in girls. Sex x BPA interaction p-value <0.2
Vafeiadi et al., 2016	Greece	235	GM: 1.2 IQR: 0.7, 2.0	1	1st	BMI, SF, WC	4 years	Null. Some evidence that high (>80th percentile) BPA associated with higher BMI in boys and lower BMI in girls from age 1–4 years. Similar associations with WC/SF.
Valvi et al. 2013	Spain	402	GM: 2.6	2	2nd/3rd	BMI, WC	4 years	Positive association of BPA with BMI and WC.
Hoepner et al., 2013	New York City	300	Mean: 3.1	1	3rd	BIA, BMI, WC	5 and 7 years	Null for BMI. BPA associated with higher FMI, BF, and WC. Stronger association with FMI in girls. Some sex x BPA terms were significant.
Philippat et al. 2014	France	520	Median: 2.5 5th/95th: 0.7, 11	1	3rd	Weight	3 years	Null. Study was conducted only in boys.
Buckley et al. 2016	New York City	173	GM: 1.3 IQR: 0.6–2.3	1	3rd	BIA	7–9 years	Positive association, but weak. Association was weaker in girls, but sex x BPA was non-significant.
Yang et al. 2017	Mexico City	194	Median: 0.7 IQR: 0.5, 1.1	1	3rd	BMI, SF, WC	10 years	Inverse association with all parameters. Non-significant sex x BPA interaction.

^*-Abbreviations: BIA-Bioelectric Impedance Analysis, BMI-Body Mass Index, GM-Geometric Mean, SF-Skinfolds,

WC-Waist Circumference

This study and several prior ones suggest that gestational urinary BPA concentrations are associated with both increased and decreased adiposity in girls, but not among boys.(18–20, 22) However, the results do not appear to be related to sample size, study location, number or timing of urine sample collections, maternal urinary BPA concentrations during pregnancy, or type or age of child adiposity assessments (Table 3). Future systematic reviews and meta-analyses may be able to identify potential reasons for the observed heterogeneity.

We speculate that the inconsistent results of studies examining the association between gestational urinary BPA concentrations and child adiposity could be due to the high degree of within-person variability of urinary BPA concentrations and resulting exposure misclassification.(12, 38) Indeed, a previous simulation study demonstrated that the association between a health outcome and non-persistent biomarker with a high-degree of within-person variability (e.g., BPA) could produce a literature with both positive and negative effect estimates.(40) Our secondary analyses showed that the observed associations between urinary BPA concentrations and early childhood adiposity among girls became stronger, but less precise, once we accounted for BPA exposure misclassification. This study and prior ones show that measurement error correction methods, including the one employed here, can be used to estimate the range of potential effects of BPA on human health outcomes.(41, 42) In the absence of studies with multiple serial BPA measures during pregnancy these correction methods may provide more plausible effect estimates.(43)

Gestational BPA exposure may affect a number of biological pathways that are related to childhood obesity risk, as well as the sex-specific distribution of adipose tissue. Both BPA and its glucuronide conjugate can increase the differentiation of human pre-adipocytes into adipocytes, possibly by activating glucocorticoid receptors or classical and non-classical estrogen receptor pathways.(44–47) Developmentally, estrogen and its associated receptors play a sex-specific role in programming body fat distribution and function, as well as changes in body fat distribution across the lifespan.(48) For example, in postmenopausal women, estrogen replacement therapy reduces total body weight and adiposity, including central adiposity.(49) In addition, BPA can affect androgen, estrogen, and glucocorticoid metabolism, which in turn may increase or decrease circulating levels of these hormones in the fetus.(50, 51) For instance, BPA is associated with androgen levels, which are generally higher in male fetuses,(52) and this may explain the sex-specific associations observed in this and some prior studies. A cross-sectional study in adolescents reported that urinary BPA concentrations were inversely associated with testosterone in boys, but positively associated with testosterone in girls.(53)

This study has some strengths and limitations worth noting. First, we used a sensitive and specific biomarker to assess BPA exposure between 6.3 and 15 weeks of gestation. However, BPA exposure is episodic in nature and BPA has a short biological half-life. Thus, a single urine sample may not adequately characterize exposure.(12, 38, 54) Assuming non-differential BPA exposure misclassification, there could be as much as 40% attenuation in effect estimates when using a single urine sample to assess exposure.(40) Moreover, we assessed BPA exposure during a period of ~9 weeks in early pregnancy and urinary BPA concentrations may be subject to pregnancy-induced changes in metabolism.(55) However, prior studies show little to no changes in BPA concentrations from the 1^st to 3^rd trimesters and our results did not change when we adjusted for the time of urine sample collection.(37, 38)

Second, we had a larger sample size than previous studies and this enhanced our ability to estimate potential sex-specific effects of BPA. However, median urinary BPA concentrations in the present study were considerably lower than those in some prior studies in the United States and Europe (0.8 vs. ~2 ng/mL); we would expect this to diminish our ability to detect potential associations since most of our participants had a narrow range of low-level BPA exposure. Indeed, our analyses using splines showed that the positive association between BPA and adiposity in girls was more apparent at higher urinary BPA concentrations. Third, while we adjusted for a rich set of covariates, there is the possibility for residual confounding by factors associated with both gestational BPA exposure and child adiposity. For instance, a woman’s diet, which is the predominant source of BPA exposure,(56) would be influenced by household dietary patterns, which in turn influence child diet and their adiposity (Supplemental Figure 2).(57) Moreover, we did not consider the potential for cumulative or interactive effects of multiple environmental chemicals on childhood adiposity, but a variety of methods are becoming available that could do this, while also examining non-linear dose-response function.(58, 59) Fourth, some prior studies have observed that early childhood urinary BPA concentrations were associated with childhood adiposity, but we were not able to examine this in the present study.(18) Finally, we measured several features of children’s central and peripheral adiposity using valid and reliable measurements that are strongly correlated with gold-standard assessments of fat mass (i.e., dual energy x-ray absorptiometry).(60) However, we measured these across a relatively wide age range, which might have induced some misclassification of adiposity. Reassuringly, our results were similar in both younger and older girls.

Conclusion

In this cohort, urinary BPA concentrations during pregnancy were associated with increased central adiposity among girls during early childhood. While the magnitude of these associations was subtle, they could have large effects at the population level to increase the risk of children being overweight or obese. Future studies examining the potential obesogenic effects of BPA should consider using measurement error correction methods or collect multiple urine samples to reduce BPA exposure misclassification and comprehensively assess childhood adiposity since BPA exposure may affect both the amount and distribution of fat mass. While these findings do not fully resolve the specific role of gestational BPA exposure in the etiology of childhood obesity, they contribute to a broader literature suggesting that exposure to some environmental chemicals during pregnancy could increase the risk of childhood obesity.

Highlights.

• Increasing gestational BPA exposure was associated with higher waist-to-hip ratio in early childhood.

• Gestational BPA exposure was associated with greater central adiposity in early childhood among girls, but not boys.

• Gestational BPA exposure was not associated with child BMI Z-score in either sex

• These findings suggest that gestational environmental chemical exposures may play a role in the development of early childhood obesity.