Exposure to bisphenols and asthma morbidity among low-income urban children with asthma

Abstract

Background:

Bisphenol A (BPA) has been linked with pediatric asthma development and allergic airway inflammation in animal models. Whether exposure to BPA or its structural analogues (BPS, BPF) is associated with asthma morbidity remains unknown.

Objective:

We examined associations between bisphenols and pediatric asthma morbidity.

Methods:

We quantified concentrations of BPA, BPS, and BPF in 660 urine samples from 148 predominantly low-income, African American children (5-17 years) with established asthma. We used biobanked biospecimens and data on symptoms, healthcare utilization, and pulmonary function and inflammation collected every 3 months over a year. We used generalized estimating equations to examine associations between concentrations or detection of urinary bisphenols and morbidity outcomes, and assessed heterogeneity of associations by sex.

Results:

We observed consistent positive associations between BPA exposure and asthma morbidity measures. For example, we observed increased odds of general symptom days (adjusted Odds Ratio, aOR:1.40;95% Confidence Interval,95%CI:1.02, 1.92), maximal symptom days (aOR:1.36;95%CI:1.00, 1.83), and emergency department visits (aOR:2.12;95%CI:1.28, 3.51) per 10-fold increase in BPA concentrations. We also observed evidence of sexually dimorphic effects; BPA concentrations were associated with increased odds of symptom days and healthcare utilization only among boys. Findings with BPS and BPF did not consistently point to associations with asthma symptoms or healthcare utilization.

Conclusion:

We found evidence to suggest that BPA exposure in a predominantly low-income, minority pediatric cohort is associated with asthma morbidity and that associations may differ by sex. Our findings support additional studies given the high pediatric asthma burden and widespread exposure to BPA in the U.S.

Graphical Abstract

Capsule Summary.

Exposure to bisphenol A, a xenoestrogen widely found in consumer products, was associated with increased asthma morbidity in a cohort of predominantly low-income, African American children with established asthma, especially among boys.

Introduction

Bisphenol A (BPA), a synthetic chemical used to manufacture polycarbonate plastics and epoxy resins, can be present in canned food and beverage linings, dental sealants, and other consumer goods (e.g., plastic dinnerware, toys, building materials).¹ Its pervasive use has led to widespread exposure in the U.S. general population. Diet is the main exposure route, although other routes, including inhalation, are possible.² BPA exposure is typically higher among U.S. children than adults, and among low-income and African American communities compared to other socioeconomic or ethnic groups.^3–5 While disparities in BPA exposure are reported among minority and low-income populations, the reasons for these disparities are not clear. It is thought that reduced access to fresh foods, low food security, and consequential consumption of processed foods (e.g., canned goods, heating and eating foods from plastic containers with BPA) could play a role.^{4, 6, 7} BPA is a xenoestrogen and concerns exist that environmental contaminants with endocrine disrupting properties could play a role in the development or exacerbation of allergic diseases like asthma.⁸

In experimental studies, BPA exposure augments the allergic immune response. BPA exposure in rodents increases production of proallergic cytokine IL-4 and antigen-specific IgE, and reduces levels of regulatory T cells, IFN-γ, and IL-10.^9–13 Low-dose BPA exposure in juvenile rodents has been linked with allergic airway inflammation. It is postulated that BPA may influence asthma risk through upregulation of T_H2 pathways by promoting eosinophilic airway inflammation.⁹ BPA has also been linked to oxidative stress,^{14, 15} which is implicated in asthma pathogenesis,¹⁶ and reported to contribute to poor asthma control and frequent, acute exacerbations by inducing airway inflammation and corticosteroid insensitivity.¹⁷ In vitro studies also show that BPA acts on both estrogen and androgen receptors.^{18, 19} Studies indicate that activation of estrogen receptors by BPA imparts effects on gene expression, activation of nuclear estrogen receptor-dependent signaling pathways, and selective receptor modulation.²⁰ The estrogenic properties of BPA could promote dendritic cell maturation²¹ or enhance antigen presentation and induce granulation of mast cells.²² It has also been suggested that BPA-estrogen receptor interactions could explain some of the associations between asthma and estrogen activity reported in epidemiologic investigations.^{22, 23}

While limited studies suggest that postnatal BPA exposure may increase risk of incident asthma and wheeze in childhood,^24–26 BPA’s role on asthma morbidity among children with established asthma remains unknown. This is a significant data gap, particularly among pediatric subpopulations with a high asthma burden^{27, 28} and disproportionate BPA exposures.^3–5 Moreover, little is known about structurally-similar BPA analogues- Bisphenol S (BPS) and Bisphenol F (BPF)- which may be used as BPA replacements in consumer products, including those labeled as “BPA-free”. BPS is used as a developer in thermal receipt paper and used in several industrial applications.²⁹ BPF is used to make epoxy resins and coatings for building materials and for plastics, dental sealants, and food packaging.³⁰ BPS and BPF have also been detected in personal care products (e.g., hair and skin care products, toothpaste),³¹ paper products (e.g., currency, mailing envelopes, airplane boarding passes),³² and in food commodities (e.g., dairy, meat and meat products, canned goods, cereals).³³ In vitro and in vivo studies indicate that these analogues, similar to BPA, may induce oxidative stress and have endocrine disrupting properties.^{14, 34, 35}

In this study, we examined associations between BPA, BPS, and BPF urinary concentrations and asthma morbidity among children with established asthma residing in a low-income, urban setting. Based on animal and limited epidemiologic studies, we hypothesized that exposure to bisphenols would be associated with increased morbidity.

Methods

Study Participants

For the present study, we used data and biobanked biospecimens from children who had participated in the Mouse Allergen and Asthma Cohort Study (MAACS), a 12-month prospective study of 148 Baltimore City children with asthma, ages 5 to 17 years. MAACs aimed to evaluate the impact of indoor allergens and air pollutants on clinical markers of asthma morbidity. Children were recruited between April 2007 and June 2009 from the Johns Hopkins Hospital Emergency Department (ED), health fairs, and word of mouth. Eligible children had received an asthma physician diagnosis ≥1 year prior to study enrollment, had a controller medication prescription or met National Asthma Education and Prevention Program criteria for persistent asthma,³⁶ and experienced ≥1 asthma exacerbation in the prior year. An exacerbation was defined as asthma symptoms requiring an emergency medical visit or an oral corticosteroid burst. Enrollment was limited to non-smokers as confirmed with rapid urine cotinine screening and non-pregnant females. The Johns Hopkins School of Medicine Institutional Review Board Study approved study protocols and informed consent was obtained from children’s parents/guardians. The involvement of the Centers for Disease Control and Prevention (CDC) in laboratory sample analyses did not constitute engagement in human subjects’ research.

Study procedures and clinical assessments

Clinic visits were conducted at baseline and every three months thereafter, for a total of up to 5 visits. At baseline, trained staff administered an allergy skin prick test with the MultiTest II device (Lincoln Diagnostics, Decatur, IL) as described previously.^{37, 38} Atopy was defined as ≥1 positive skin test response, with a wheal size of ≥2 mm greater than the negative control.^{39, 40} Blood was drawn to measure total IgE using the ImmunoCAP system (Pharmacia Diagnostics, Uppsala, Sweden).

At each visit, staff administered questionnaires to capture information on demographic characteristics (baseline only), pulmonary and allergic history, number of days with symptoms and rescue medication use in the prior 2 weeks, and asthma-related healthcare utilization in the prior 3 months. Asthma-related symptoms included: general symptoms (chest tightness, wheeze, cough), nocturnal wakening with symptoms, exercise-related symptoms (coughing or chest tightness when running/going upstairs), coughing without a cold, and slowed activity due to asthma. A “maximal symptom days” variable was generated by taking the maximum number of days contributed by general symptoms, nocturnal wakening with symptoms, or slowed activity due to asthma.^{40, 41} Healthcare utilization included acute care visits, ED visits, unscheduled physician visits, and hospitalization for asthma-related symptoms. Acute care visits included any unscheduled visit for asthma-related symptoms.

Spirometry was performed at each clinic visit with a KoKo spirometer (nSpire Health, Longmont, Colorado); percent predicted values of lung function measures were determined using Global Lung Initiative age-specific reference equations.⁴² Bronchodilator reversibility was defined as a ≥12% increase in forced expiratory volume in the first second (FEV₁) 15 minutes after administration of two puffs of short-acting β-agonist. Lung inflammation was assessed by measuring the fraction of exhaled nitric oxide (FENO) levels with a NIOX Mino (Aerocrine, Solana, Sweden) following American Thoracic Society/European Respiratory Society guidelines.⁴³

Exposure assessment

In vivo and in vitro studies suggest that BPF metabolism is similar to that of BPA, while studies on BPS are scarce.^44–46 Still, BPA, BPS, and BPF are thought to be rapidly metabolized to respective glucuronides and rapidly eliminated in urine.^{47, 48} Total urinary bisphenol levels, usually determined after deconjugation with β-glucuronidase/sulfatase, are considered to be a robust biomarker of exposure to these compounds in epidemiologic studies.^49–52 Thus, concentrations of total urinary species (free + conjugated) of bisphenols were quantified using on-line solid phase extraction coupled with high-performance liquid chromatography-isotope dilution tandem mass spectrometry.⁵³ Spot urine samples (n=660) were collected at each visit in polypropylene urine cups, aliquoted, and stored at −80°C until shipment to the CDC (Atlanta, GA) for analysis. Limits of detection (LOD) were 0.2 ng/mL (BPA, BPF) and 0.1 ng/mL (BPS). For concentrations <LOD, we used instrumental-reading values when a signal was detected, or randomly imputed values based on a log-normal probability distribution using maximum likelihood estimation when no signal was detected.⁵⁴ Concentrations of bisphenols were specific gravity-corrected to normalize for urine dilution.⁵⁵ Specific gravity was measured using a digital refractometer (ATAGO™3741, Tokyo, Japan).

Data analyses

We calculated descriptive statistics to summarize baseline participant characteristics and biomarker concentrations across visits. Concentrations of BPA and BPS were log-normally distributed and log10-transformed in subsequent models. BPF was not detected as frequently and modeled as a dichotomous variable (<LOD vs. ≥LOD) in subsequent analyses.

Wilcoxon-Mann-Whitney, Kruskall Wallis, and Chi-square tests were used to assess whether BPA or BPS concentrations or BPF detection frequency differed based on select baseline participant characteristics. We assessed correlations between BPA and BPS concentrations across visits using Spearman correlations and assessed within- and between-child variability of concentrations by calculating intraclass correlation coefficients (ICC) using mixed-effects models. ICC analyses were restricted to children who provided ≥2 urine samples. We also compared average participant biomarker concentrations to those measured among U.S. children of similar ages (6-17 years) in the National Health and Nutrition Examination Survey (NHANES) during the same period (2007-2010) and more recently (2015-2016) using data on three different subsets: (1) a representative sample of U.S. children in the same age range as MAACS participants, regardless of race; (2) a representative sample of African American U.S. children given our cohort was predominantly African American; and (3) a representative sample of U.S. children matched on MAACS demographic characteristics (i.e., age, race, sex, and income). These comparisons were limited to uncorrected urinary biomarker concentrations as data on specific gravity were not available in NHANES.

To evaluate associations between concentrations of bisphenols and morbidity outcomes, we used generalized estimating equations (GEE) to account for our repeated measures design. Specifically, crude and multivariable binomial regression models were fit to model relationships between repeated exposure and binary outcome measures (e.g., symptoms, medication use, healthcare utilization) and linear regressions for continuous measures (e.g., FEV1/FVC%, FENO). To minimize bias while maximizing effect estimate precision, we identified covariates for model inclusion using a directed acyclic graph, including confounders and predictors of pediatric asthma morbidity that were not on the causal pathway. Covariates included age (in years), sex, race/ethnicity (Black/African American vs. Other), caregiver’s education level (less than high school, high school graduate, and some college or more), presence of smokers in the home, and season. To account for co-exposures and assess the independent effects of each bisphenol, all three bisphenols were included in each model. Because the bisphenols assessed are known or suspected endocrine disrupting compounds (EDCs) and prior studies have reported sexually dimorphic effects,¹⁵ we assessed effect measure modification by sex by including an interaction term in separate models.

To ensure the robustness of our results, we conducted sensitivity analyses. Because BPA and BPF exposure has been linked to obesity,^56–58 and obesity plays a role in asthma control and exacerbation,⁵⁹ we re-ran main models including children’s age- and sex-standardized body mass index (BMI) percentiles⁶⁰ modeled as a continuous covariate. We also re-ran models controlling for degree of atopy, a recognized risk factor for asthma morbidity;⁶¹ we included the total number of positive skin prick tests as a continuous model covariate to reflect each child’s degree of allergic sensitization. Lastly, we ran our main models additionally including other recognized risk factors for asthma morbidity, including degree of atopy and exposure to particulate matter.^{40, 61} Specifically, we included the total number of positive skin prick tests as a continuous model covariate to reflect each child’s degree of allergic sensitization of each child. For particulate matter, we ran our main models additionally including log₁₀-transformed indoor concentrations of PM_2.5 (i.e., particulate matter with an aerodynamic diameter <2.5 microns), which were measured in the participants’ homes within ±2 weeks of each clinic visit.

Analyses were performed with Stata/IC 14.2 software (StataCorp, College Station, Texas); and statistical significance criteria was set at p<0.05 and p<0.10 for main effects and effect measure modification, respectively.

Results

The mean (SD) age of participants was 11.2 (4.0) years and approximately 57% were males (Table 1). Children were predominantly African American (91%), had an annual household income <$35,000 (69%), and were receiving government-based health insurance (86%). Over half of the caregivers (63%) reported not having college education. Most children (91%) had ≥1 positive skin prick test response, and reported having an acute health care visit for asthma in the previous year (96%) with 82% reporting an ED visit for asthma during this period (Table 2). Additional clinical characteristics are presented in Table 2.

Table 1.

Baseline demographic characteristics for MAACS children ages 5-17 years (n=148).^a

Demographic Characteristic	n (%)
Ageb
5-7 years	41 (27.7)
8-10 years	35 (23.7)
11-13 years	26 (17.6)
14-17 years	46 (31.1)
Sex
Male	85 (57.4)
Female	63 (42.6)
Child’s race
Black/African American	135 (91.2)
Otherc	13 (8.8)
Child body mass index (BMI)
Underweight (<5th percentile)	6 (4.1)
Normal Weight (5th-<85th percentile)	77 (52.7)
Overweight (85th-<95th percentile)	22 (15.1)
Obese (≥ 95th percentile)	41 (28.1)
Caregiver’s Education Level
Less than high school	42 (28.4)
High school graduate	51 (34.5)
Some college or more	55 (37.2)
Annual household income
<$35,000	92 (68.7)
≥$35,000	42 (31.3)
Insurance
Private/self-pay	20 (13.7)
Government-based (public)	126 (86.3)
Smoker lives in child’s home
No	64 (43.2)
Yes	84 (56.8)

^a.Information reported is for 148 MAACs children with BPA, BPS, and BPF biomarker data. Information was missing for select demographic characteristics at baseline (BMI: n=2; Annual household income: n=14; Health insurance: n=2).

^b.The mean(SD) age of MAACS children was 11.2(4.0) years.

^c.“Other” race category includes White (n=6); American Indian or Alaska Native (n=5); Other Pacific Islander (n=1); and Unknown (n=1).

Table 2.

Baseline clinical characteristics for MAACS children ages 5-17 years (n=148).^a

Clinical Characteristic
Allergic sensitization characteristics
Atopic (≥1 Positive skin prick test response); No.(%)		134(91.2)
Skin test sensitivities; No.(%)
	Cat	96(65.3)
	Cockroach	91(61.9)
	Dust mite	85(57.8)
	Mouse	78(53.1)
	Dog	26(17.7)
Total IgE (kU/L); Median (p25,p75)		190 (55.7, 458)
Lung function
FEV1 (% predicted); Mean (SD)		91.3(15.6)
FVC (% predicted); Mean (SD)		100.2(13.5)
FEV1/FVC%; Mean(SD)		80.6(9.6)
Bronchodilator reversibilityb; No./total (%)		37/130(28.5)
FeNO, ppb; Median (p25,p75)		33(16.0, 62.0)
Asthma-related health care utilization in the prior 12 months and medication use
Acute health care visitc; No.(%)		142(96.0)
ED visit; No.(%)		121(81.8)
Hospitalization; No.(%)		29(19.6)
Unscheduled doctor’s office visit; No.(%)		61(41.2)
Controller medication used; No.(%)		106(71.6)
Days of short acting β-agonist (SABA) use, prior 2 weeks; Mean(SD)		4.2(5.0)

^a.Information was missing on some participants (atopic status and skin sensitivities: n=1; IgE: n=3; FVC % predicted: n=16; FEV1 % predicted: n=16; FEV1/FVC%: n=16; bronchodilator reversibility: n=18; FENO: n=17).

^b.Bronchodilator reversibility is defined as an increase in FEV1 of 12% or greater following treatment with a short-acting β-agonist or SABA.

^c.Acute health care visits consists of a composite measure of any unscheduled healthcare visit for asthma-related symptoms (e.g., ED visits, hospitalization, and/or unscheduled doctor’s visits for asthma).

^d.Inhaled corticosteroid or leukotriene inhibitor.

Abbreviations: FEV 1: Forced exhaled volume in the first second; FVC: Forced vital capacity; FeNO: fractional exhaled nitric oxide; ED visit: Emergency department visit for asthma.

Summary statistics for specific gravity-corrected concentrations of bisphenols are provided in Table 3 (Repository Table E1 for uncorrected concentrations). Detection frequencies were 100% (BPA), 97% (BPS), and 66% (BPF). Geometric mean (GSD) concentrations were 3.6(2.2) ng/mL (BPA), 0.4(3.2) ng/mL (BPS), and 0.8(6.2) ng/mL (BPF). Baseline BPA concentrations differed by select demographic characteristics (Repository Table E2), including BMI (p=0.04); caregiver’s education (p=0.03); and season (p=0.01). Geometric mean (GM) BPA concentrations were higher among children with normal body weight (versus overweight/obese), children with less educated caregivers, and in samples collected in summer months. BPS concentrations were higher among children living with smokers (p=0.03). BPF detection frequency did not differ based on demographic characteristics.

Table 3.

Summary statistics and variability measures for specific gravity-corrected BPA, BPS, and BPF concentrations in ng/mL based on the total number of urine samples collected on 148 MAACS participants (n=660 samples).^a

	Specific gravity-corrected concentrations, ng/mL					Measure of biomarker variabilityb
Biomarker	LOD	%>LOD	GMa(GSD)	Median (p25, p75)	Max	ICC n≥2 (95%CI)
BPA	0.2	100.0	3.6(2.2)	3.4 (2.2, 5.6)	127.5	0.10 (0.05, 0.20)
BPS	0.1	96.8	0.4(3.2)	0.3(0.2, 0.7)	135.0	0.14 (0.08, 0.23)
BPF	0.2	65.9	0.8(6.2)	0.3(<LOD, 1.2)	629.9	--

^a.GM values reported are based on detectable concentrations, including instrumental reading values, for each biomarker.

^b.ICC _n≥2 (95%CI): indicates ICC values and respective 95% confidence intervals based on data from children who provided 2 or more samples during the 12-month follow-up period.

Abbreviations: LOD- Limit of detection; %LOD- Percent of samples with concentrations above the limit of detection for each specified analyte; GM- geometric mean; GSD: geometric standard deviation; 95%CI: 95% confidence interval; ICC: Intraclass correlation coefficient.

At baseline, higher median BPA concentrations were observed among children who reported having acute care (p50: 3.5 vs. 3.0 ng/mL, p=0.04) and ED visits for asthma-related symptoms (p50: 3.7 vs. 3.1 ng/ml, p=0.02) in the prior year compared to children who did not report these outcomes. This pattern was generally consistent across visits and similar patterns were observed for other healthcare utilization outcomes (Repository Table E3). For BPS, baseline median concentrations were similar among children reporting most healthcare utilization outcomes; however, this pattern was not consistent across visits (Repository Table E3). BPA and BPS concentrations were weakly correlated at each visit (Spearman-ρ: 0.18-0.30, p<0.05); and concentrations varied greatly within children over the 12-month follow-up period (ICC_BPA:0.10; ICC_BPS:0.14). Median BPA and BPF concentrations were higher in MAACS participants than in NHANES children from either sampling period, regardless of which subset of NHANES children was used in our comparisons (Figure 1). BPA and BPF GM concentrations were 1.2 to 2.8 times higher in MAACS participants compared to NHANES children. MAACS and NHANES children had similar median BPS concentrations.

Figure 1. Distribution of uncorrected urinary concentrations (ng/mL) of BPA, BPS, and BPF among 148 MAACS participants ages 5-17 years and children ages 6-17 years from the U.S. general population (NHANES 2007-2010, 2015-2016) using data from three subsets of children from NHANES (all children in the same age range, regardless of race; African American children; and a subset of children matched with MAACS participants based on.

Biomarker concentrations for each MAACS participant was first averaged prior to generating distributions presented. Biomarker concentrations for MAACS participants that were <LOD were imputed to LOD/√2 to allow for proper comparisons with NHANES data utilizing the same imputation method. Abbreviations: DF=Detection frequency

Adjusted associations between concentrations of bisphenols and asthma-related outcomes are presented in Table 4 (Repository Table E4 for crude associations). We observed consistent positive relationships between BPA and days of asthma symptoms in the prior two weeks, and healthcare utilization in the prior three months in multivariable GEE models. For every 10-fold increase in BPA concentrations we observed a 40% and 36% increased odds of having days with general symptoms and maximal symptoms, respectively (adjusted odds ratio, aOR:1.40,95% Confidence Interval, 0:1.02,1.92 and aOR:1.36,95%CI:1.00,1.83, respectively). We also observed positive, albeit non-statistically significant, associations with other symptoms. For healthcare utilization, we observed an 84% and 112% increased odds of reporting an acute care or an ED visit, respectively, per 10-fold increase in BPA concentrations (aOR:1.84,95%CI: 1.11, 3.03 and aOR:2.12,95%CI:1.28, 3.51, respectively). We also observed inverse, albeit non-statistically significant, associations between BPA and lung function parameters (Table 4). Inclusion of other covariates in sensitivity analyses, including degree of atopy and exposure to PM 2.5, did not materially affect results (Repository Tables E5–E7).

Table 4.

Adjusted associations between repeated urinary concentrations of BPA, BPS, and BPF and asthma morbidity measures among children ages 5-17 years participating in MAACS.^a,b,c

	Bisphenol A		Bisphenol S		Bisphenol F
Asthma Symptoms (n=148 children)	aOR (95% CI)	p-value	aOR (95% CI)	p-value	aOR (95% CI)	p-value
Coughing, wheezing, or chest tightness	1.40 (1.02, 1.92)	0.037	0.75 (0.57, 1.00)	0.046	1.14 (0.89, 1.47)	0.292
Nocturnal wakening with symptoms	1.46 (0.93, 2.29)	0.096	0.69 (0.50, 0.95)	0.023	0.99 (0.74, 1.32)	0.929
Exercise-related symptoms	1.46 (0.96, 2.20)	0.075	0.63 (0.45, 0.88)	0.006	1.03 (0.77, 1.38)	0.823
Cough without a cold	1.19 (0.69, 2.08)	0.530	0.72 (0.51, 1.03)	0.069	0.82 (0.57, 1.20)	0.310
Slowed activity due to asthma	1.23 (0.82, 1.83)	0.320	0.74 (0.52, 1.03)	0.076	1.04 (0.79, 1.39)	0.762
Maximal symptom days	1.36 (1.00, 1.83)	0.048	0.78 (0.59, 1.03)	0.076	1.09 (0.85, 1.38)	0.508
Short-acting Beta Agonist (SABA) use	1.30 (0.94, 1.78)	0.108	0.98 (0.76, 1.27)	0.880	0.97 (0.76, 1.25)	0.818
Healthcare Utilization (n=148 children)	aOR (95% CI)	p-value	aOR (95% CI)	p-value	aOR (95% CI)	p-value
Acute care visit	1.84 (1.11, 3.03)	0.018	0.84 (0.60, 1.18)	0.313	1.06 (0.76, 1.47)	0.752
Unscheduled doctor’s visit	1.23 (0.70, 2.15)	0.476	1.10 (0.69, 1.75)	0.680	1.01 (0.65, 1.56)	0.962
ED visit for asthma	2.12 (1.28, 3.51)	0.004	0.86 (0.58, 1.27)	0.451	0.94 (0.65, 1.36)	0.754
Hospitalization for asthma	2.24 (0.65, 7.71)	0.201	1.28 (0.61, 2.71)	0.514	1.80 (0.64, 5.08)	0.266
Lung Function (n=144 children)	β (95% CI)	p-value	β (95% CI)	p-value	β (95% CI)	p-value
FEV1 percent predicted	−0.90 (−3.52, 1.73)	0.503	0.79 (−0.62, 2.20)	0.273	−0.24 (−1.98, 1.50)	0.785
FVC percent predicted	−0.60 (−2.67, 1.47)	0.572	0.25 (−0.92, 1.42)	0.678	0.71 (−0.80, 2.23)	0.357
FEV1/FVC%	−0.09 (−1.55, 1.38)	0.907	0.50 (−0.32, 1.31)	0.233	−0.87 (−1.71, −0.03)	0.042

^a.Information on nocturnal awakening with asthma symptoms was missing for 2 children; information on lung function parameters was missing on 4 children.

^b.Models were adjusted for age, sex, race, caregiver education, season, presence of smokers in the home as a proxy for environmental tobacco smoke exposure, and co-exposure to other bisphenols (i.e., all bisphenols were included in the same model for each outcome).

^c.For dichotomous outcomes (asthma symptoms; healthcare utilization) we conducted binomial regression models using generalized estimating equations and log link to obtain PRs and for continuous outcomes (lung function) we ran linear regression models with generalized estimating equations to obtain beta coefficients. The analyses presented in this table represents data on up to 148 children with up to 660 urine samples collected over a 12-month period. BPA and BPS were modeled as continuous variables (log-10 transformed specific gravity-corrected concentrations) and BPF was modeled as a categorical variable (detected vs. not detected) based on detection frequencies. Abbreviations: aOR=Adjusted Odds Ratio; 95%CI: 95% Confidence Interval; β=Beta coefficient; ED=Emergency Department. Bolded effect estimates indicate p<0.05

We observed suggestive evidence of sexually dimorphic effects of BPA concentrations (p_int<0.10) for days of asthma symptoms in the prior two weeks (Figure 2). BPA concentrations were associated with increased odds of general symptoms and maximal symptom days only among boys (p<0.05). We observed this trend for other symptoms and healthcare utilization outcomes, though effect modification p-values were not significant. We did not observe sexually dimorphic effects with lung function or inflammation measures.

Figure 2. Adjusted associations between BPA and asthma-related symptoms by sex among MAACS children.^a.

Models are based on generalized estimating equations using the binomial log link function to account for repeated measures of exposure and outcomes and were adjusted for age, race, caregiver’s education, season, presence of smokers in the home, log₁₀-transformed concentrations of BPS and dichotomized (detected vs. not detected) exposure to BPF. b. One male and one female were missing information on nocturnal symptoms. Abbreviations: p_int: P-value for effect measure modification by sex

For BPS, we observed no consistent associations with asthma-related outcomes (Table 4). Effect estimates for healthcare utilization outcomes were not statistically significant and not in the same direction, while we observed a pattern of inverse associations with symptoms. Inclusion of other covariates did not appreciably affect the inferences for BPS and there was no evidence of sexually dimorphic effects for BPS (not shown). We observed an inverse association between BPF detection and FEV1/FVC%(β=−0.87; 95%CI:−1.71,−0.03); however, we observed no other associations with BPF. Inclusion of other covariates led to similar findings and there was no evidence of sexually dimorphic effects. Lastly, we observed no associations between concentrations of any bisphenols and FENO.

Discussion

In this study, we assessed associations between urinary concentrations of BPA and select analogues (BPS, BPF) and asthma morbidity, and lung function and inflammation in a cohort of predominantly low-income, urban, African American children with established asthma. Overall, we observed consistent positive associations between BPA concentrations and asthma morbidity measures, including with general and maximal symptom days, and with acute care and ED visits for asthma. We found evidence of sexually dimorphic effects, with only boys exhibiting higher odds of asthma morbidity with BPA. By contrast, we found no consistent relationships between BPS or BPF and asthma morbidity measures, and neither BPA nor its analogues were associated with lung inflammation as measured by FENO.

BPA concentrations in our cohort were substantially higher than those observed in Mexican, Chinese, or U.S. children from the general population,^{62, 63} and similar to those measured in minority children in New York City.^{24, 64} For example, specific gravity-corrected GM concentrations for BPA among a sample of 60 Chinese children were 0.40 ng/mL compared to 3.6 ng/mL in MAACS children, while median uncorrected BPA concentrations among 159 minority children in New York City were 3.4 ng/ml compared to 3.3 ng/mL in MAACS children. BPS and BPF concentrations were also higher in our cohort compared to 283 children from South China (e.g., uncorrected median concentrations- BPS: 0.30 vs. 0.03, respectively; and 0.30 vs. 0.19, respectively).⁶³ Differences in concentrations of bisphenols among populations could reflect different consumption dietary patterns (e.g., differences in the amount or frequency of consumption of canned goods or beverages in polycarbonate bottles, heating foods in plastic containers with BPA) and consumer product uses; with the exception of one study,⁶³ prior studies collected urine samples around the same time period as the present study.

Similar to low-income New York City children,⁶⁵ baseline BPA concentrations were highest during summer months. Seasonal differences could relate to dietary patterns, such as greater consumption of beverages in canned or plastic containers in the summer. BPA can migrate from reusable polycarbonate water bottles into water,⁶⁶ and use of polycarbonate water bottles has been positively associated with urinary BPA concentrations,⁶⁷ particularly during summer months.⁶⁸ Children of normal weight had higher GM BPA concentrations at baseline than overweight children, contrary to findings in prior pediatric studies.^{58, 69} GM BPA concentrations were also lower among participants with caregivers who were not high school graduates, a trend not reported in previous studies. Lastly, similar to children ages 1 to 8 years in Ohio,⁷⁰ and minority children in New York City ages 6-10 years,⁶⁴ BPA concentrations tended to vary greatly within individuals (ICCs<0.20). To our knowledge, no studies have assessed temporal variability of BPA analogues among children, although one study reported great intra-individual variability for BPA and BPF concentrations among adult men.¹⁴

While prior studies have evaluated associations between pre- and postnatal BPA exposures and asthma development and prevalence,^{24, 71–73} no studies have assessed the role of bisphenols on asthma morbidity among children with established asthma. Yet, one study conducted among low-income children in New York City with comparable GM BPA urinary concentrations reported positive associations with incident asthma at ages 5 to 12 years.²⁴ Postnatal BPA concentrations were also associated with increased risk of wheeze at age 7 years.²⁴ Two other pediatric studies also reported positive associations and comparable effect sizes as those observed in our study between exposure to BPA and incident asthma and wheeze,^{25, 26} although one other study reported null associations between BPA and wheeze.⁷¹

Although the exact mechanisms by which BPA could elicit respiratory effects remain unknown, several have been proposed, including oxidative stress and upregulation of Th2 pathways via promotion of eosinophilic airway inflammation.^{9, 15} A recent pediatric case-control study also reported that genetic polymorphisms of oxidative stress-related genes may modulate the effect of BPA on asthma.⁷⁴ In adults, BPA has been associated with oxidative stress and inflammation markers in postmenopausal women, suggesting that BPA could bind to estrogen receptors more actively when estrogen levels are low, triggering cellular responses associated with oxidative stress and inflammation. In one experimental study, ovalbumin + low-dose BPA exposure during the juvenile period of development in male mice altered serum levels of anti-inflammatory corticosterone and estrogen receptor 2 messenger RNA expression in the lungs and enhanced inflammatory cell infiltration and protein expression of Th2 cytokines and chemokines, among other immune-activated responses, suggesting that BPA could exacerbate allergic airway inflammation by enhancing Th2 responses through immune system disruption. Unfortunately, no biomarkers of oxidative stress (e.g., malondialdehyde (MDA), and 8-hydroxydeoxyguanosine (8-OHdG)), were measured in our cohort, preventing us from examining associations between BPA and oxidative stress.. We also did not observe associations with FENO, a non-invasive marker of eosinophilic airway inflammation. It is plausible that no associations were observed with FENO because most children had eosinophilic airway inflammation at baseline. Further studies are warranted to confirm our findings and identify the underlying pathways by which BPA exposure could increase asthma morbidity risk.

We observed a consistent pattern of increased odds of morbidity among boys, but not girls, for several outcomes. Previous studies have reported similar patterns of increased risk of allergic disease and morbidity among boys in association with EDC exposures during pre-pubescence (8-14 years).^75–78 One potential explanation for these observed sex differences is that BPA exposure could enhance the allergic immune response, increasing susceptibility to adverse respiratory effects among males. It is plausible that BPA, being a xenoestrogen, exerts different endocrine disrupting effects in boys than girls due to their different hormonal milieu; current evidence suggests that sex hormones and EDCs may play a role in the development and/or function of the lungs and the immune system.^79–83 In males with severe asthma, greater androgen levels have been associated with better lung function⁸⁴ and it is possible that BPA exposure leads to an imbalance in circulating hormones; however, further investigations are needed to confirm findings and elucidate biological mechanisms by which sex differences could occur.

Study limitations include lack of data on pubertal stage, preventing us from assessing the effects of pubertal development on our exposure-outcome relationships. Pubertal development differentially influences asthma symptom progression among males and females,⁸⁵ and may also influence how BPA affects asthma morbidity during adolescence. Future studies should evaluate the effects of BPA exposure during different developmental stages to identify potential windows of susceptibility because disruption in the production and metabolism of sex hormones may dictate differences in susceptibility by sex and life stage. It is also plausible that the effects of BPA on asthma morbidity depend on the degree of atopy, making atopy a confounder, mediator, or effect modifier. However, we were unable to assess mediation or effect modification because 91% of participants were atopic. We also cannot rule out the possibility of unmeasured confounding and spurious findings. Additionally, our study was conducted on a predominantly low-income, urban minority pediatric cohort, limiting generalizability to other populations. While outcomes like healthcare utilization preceded exposures, collection of other outcomes obtained closer in time or concurrently to our exposures helped overcome this limitation. Although diet is the main BPA exposure route, we cannot rule out the possibility that other chemicals present in foods (e.g., phthalates) may be responsible for the associations observed. Because our analyses were conducted after all data collection concluded using biobanked biospecimens, we were also not able to prospectively collect information on potential sources of exposure to bisphenols in our study population. However, our study population is predominantly low-income and African American and exposure in this population is thought to arise predominantly from diet.^{4, 6, 7} Future studies should evaluate whether dietary exposures to mixtures of immunotoxic chemicals impact asthma morbidity. Identifying specific modifiable exposure sources of BPA in low-income, urban settings will also be important.

Despite acknowledged limitations, this study has several strengths. This is the first study to examine BPA, BPS, and BPF exposure in relation to measures of asthma morbidity in a pediatric population suffering a high asthma burden and high exposures to select bisphenols. The prospective study design and evaluation of repeated exposure and several outcome measures are major strengths that allowed for improved exposure and outcome characterization. Prospective data collection in a diseased cohort also minimized the potential for recall bias. This is also the first study to characterize temporal variability for BPS concentrations in children. We also controlled for important confounders (e.g., race, income, exposure to air pollution, BMI, degree of allergic sensitization) and bisphenol co-exposures, allowing us to assess the independent effects of each bisphenol on the outcomes. Noteworthy, while BPA is being phased out of some consumer products, recent national biomonitoring data still demonstrate widespread exposure in the USA,⁵ underscoring the continued relevance of our findings. Furthermore, our study contributes some of the first data on BPA alternatives (BPS and BPF) in relation to human health.

In summary, we found evidence to suggest that BPA exposure among a predominantly low-income, minority cohort of children with asthma may increase asthma morbidity, at least for boys. We found no consistent associations between BPS or BPF and asthma morbidity. Our findings warrant replication given widespread BPA exposure in the U.S. population and the potential clinical implications for asthma control guidelines. If the present findings are confirmed in future studies, then avoiding or limiting contact with BPA sources (e.g., consumption of canned goods, storing or heating food in plastic containers containing BPA) may be advisable among children with asthma.

Clinical Implications.

BPA exposure among predominantly low-income, African American children with asthma may be associated with asthma morbidity. Findings warrant replication given the high asthma burden and disparate BPA exposures reported among African Americans.

Abbreviations

aOR

Adjusted Odds Ratio

BPA

Bisphenol A

BPF

Bisphenol F

BPS

Bisphenol S

BMI

Body mass index

CI

Confidence Interval

cOR

Crude Odds Ratio

DF

Detection frequency

GEE

Generalized estimating equation

GM

Geometric mean

GSD

Geometric standard deviation

ICC

Intraclass correlation coefficient

NHANES

National Health and Nutrition Examination Survey