Prenatal exposure to bisphenol A and hyperactivity in children a … ADHD... · 2019. 12. 12. ·...

Contents lists available at ScienceDirect

Environment International

journal homepage: www.elsevier.com/locate/envint

Prenatal exposure to bisphenol A and hyperactivity in children: a systematicreview and meta-analysis

Johanna R. Rochestera,⁎, Ashley L. Boldena, Carol F. Kwiatkowskia,b

a The Endocrine Disruption Exchange, TEDX, Paonia, CO, United StatesbDepartment of Integrative Physiology, University of Colorado, Boulder, CO, United States

A R T I C L E I N F O

Handling Editor: Paul Whaley

Keywords:Bisphenol AAttention-deficit hyperactivity disorderNeurodevelopmentEndocrine disruptionSystematic reviewMeta-analysis

A B S T R A C T

Background: Attention-deficit hyperactivity disorder (ADHD) has increased in prevalence in the past decade.Studies attempting to identify a specific genetic component have not been able to account for much of theheritability of ADHD, indicating there may be gene-environment interactions underlying the disorder, includingearly exposure to environmental chemicals. Based on several relevant studies, we chose to examine bisphenol A(BPA) as a possible contributor to ADHD in humans. BPA is a widespread environmental chemical that has beenshown to disrupt neurodevelopment in rodents and humans.Objectives: Using the Office of Health Assessment and Translation (OHAT) framework, a systematic review andmeta-analysis was designed to determine the relationship between early life exposure to BPA and hyperactivity,a key diagnostic criterion of ADHD.Data sources: Searches of PubMed, Web of Science, and Toxline were completed for all literature to January 1,2017.Study eligibility criteria: For inclusion, the studies had to publish original data, be in the English language, includea measure of BPA exposure, and assess if BPA exposure affected hyperactive behaviors in mice, rats or humans.Exposure to BPA had to occur at< 3months of age for humans, up to postnatal day 35 for rats and up topostnatal day 40 for mice. Exposure could occur either gestationally (via maternal exposure) or directly to theoffspring.Study appraisal and synthesis methods: Studies were evaluated using the OHAT risk of bias tool. The effects inhumans were assessed qualitatively. For rodents exposed to 20 μg/kg/day BPA, we evaluated the study findingsin a random effects meta-analytical model.Results: A review of the literature identified 29 rodent and 3 human studies. A random effects meta-analysisshowed significantly increased hyperactivity in male rodents. In humans, early BPA exposure was associatedwith hyperactivity in boys and girls.Limitations, conclusions, and implications of key findings: We concluded that early life BPA exposure is a presumedhuman hazard for the development of hyperactivity. Possible limitations of this systematic review include de-ficiencies in author reporting, exclusion of some literature based on language, and insufficient similarity betweenhuman studies. SRs that result in hazard-based conclusions are the first step in assessing and mitigating risks.Given the widespread exposure of BPA and increasing diagnoses of ADHD, we recommend immediate actions tocomplete such risk analyses and take next steps for the protection of human health. In the meantime, precau-tionary measures should be taken to reduce exposure in pregnant women, infants and children. The presentanalysis also discusses potential mechanisms by which BPA affects hyperactivity, and the most effective avenuesfor future research.Systematic review registration number: Not available.

1. Introduction

In the past decade, increased rates of certain neurodevelopmentaldisorders in children such as autism, learning disabilities, and attention-

deficit hyperactivity disorder (ADHD) have raised concerns over thepossible causes of these disorders, and whether they might be due toearly environmental influences (Banerjee et al., 2007; de Cock et al.,2012). Specifically, lead exposure (Daneshparvar et al., 2016), tobacco

https://doi.org/10.1016/j.envint.2017.12.028Received 7 July 2017; Received in revised form 14 December 2017; Accepted 17 December 2017

⁎ Corresponding author.E-mail address: [email protected] (J.R. Rochester).

Environment International 114 (2018) 343–356

Available online 07 March 20180160-4120/ © 2017 Published by Elsevier Ltd.

T

http://www.sciencedirect.com/science/journal/01604120https://www.elsevier.com/locate/envinthttps://doi.org/10.1016/j.envint.2017.12.028https://doi.org/10.1016/j.envint.2017.12.028mailto:[email protected]://doi.org/10.1016/j.envint.2017.12.028http://crossmark.crossref.org/dialog/?doi=10.1016/j.envint.2017.12.028&domain=pdf

use (DiFranza et al., 2004), and alcohol consumption (Riley and McGee,2005) during pregnancy have been associated with the development ofADHD. This is concerning, as rates of ADHD may be increasing in theUS (Arnold et al., 2012; Visser et al., 2014; Boyle et al., 2011).

ADHD usually has an onset in early school-aged children and canpossibly persist into adulthood (Jain et al., 2017; Kessler et al., 2006),although some evidence suggests that adult ADHD is not associatedwith diagnosis in childhood (Moffitt et al., 2015). The disorder canaffect many behavioral aspects, such as attention (e.g., alertness andvigilance), executive function (e.g., working memory, response inhibi-tion, cognitive flexibility, and planning), and reduced response habi-tuation (Aguiar et al., 2010). ADHD is a heterogeneous disorder, con-sisting of many symptoms and co-morbidities that are present tovarying degrees (Aguiar et al., 2010; Matthews et al., 2014; Baird et al.,2000; Willcutt et al., 2005; Nigg et al., 2005). Individuals with ADHDmay also have problematic behaviors/disorders such as aggression,oppositional defiant disorders, conduct disorders, anxiety, depression,substance abuse, sleep disorders, Tourette syndrome, antisocial beha-viors, and learning disabilities (Aguiar et al., 2010; Baird et al., 2000;Ramos-Quiroga et al., 2013; American Psychiatric Association, 2013).

Evidence suggests that ADHD is associated with alterations in theprefrontal cortex and dysfunctional monoaminergic signaling in thebrain; however, the mechanisms have not been definitively established(Aguiar et al., 2010; Eubig et al., 2010). ADHD is usually treated with acombination of behavioral therapies and stimulant medications thattarget these signaling pathways, such as methylphenidate and amphe-tamine (which increase synaptic dopamine release), or norepinephrinereuptake inhibitors, such as atomoxetine (Ramos-Quiroga et al., 2013).

Studies suggest that heritability plays a role in the development ofADHD. Twin studies estimate heritability to be 70–80% (Matthewset al., 2014; Franke et al., 2009; Smith et al., 2009) but studies ofcandidate genes involved in ADHD etiology, which include specificcatecholaminergic and serotonergic signaling proteins, have not beenable to account for> 3–4% of the total variance in ADHD phenotype(Neale et al., 2010). Thus, it is possible that gene-environment inter-actions are inflating the heritability estimates (Matthews et al., 2014).

Environmental factors also likely play a role in the development ofADHD. Fetal exposures to alcohol, cigarette smoke, and lead have beenshown to be associated with development of ADHD and ADHD-symp-toms in children and in animal models (Banerjee et al., 2007; Eubiget al., 2010; Abbott and Winzer-Serhan, 2012). More recently, exposureto bisphenol A (BPA) has been studied as a possible contributor toADHD. BPA is widely used in plastics, epoxy resins (used to line cansand as dental sealants), food packaging and thermal receipts, and is alsofound in recycled paper products such as toilet paper (Michalowicz,2014). BPA, a known endocrine disruptor, interacts with several steroidreceptors and has been shown to disrupt numerous physiological sys-tems in animals (Vandenberg et al., 2013). Further, it has been linked tomany adverse health effects in both adults and children (Rochester,2013). In children, BPA has been implicated in neurobehavioral dis-ruption in those exposed both prenatally and postnatally (Rochester,2013; Braun et al., 2009; Perera et al., 2012; Maserejian et al., 2012;Casas et al., 2015) and ADHD-symptoms such as hyperactivity havebeen shown to be associated with exposure during development (Braunet al., 2009; Braun et al., 2011; Casas et al., 2015; Braun and Hauser,2011; Harley et al., 2013). A preliminary search of the literature onenvironmental chemical exposures and ADHD symptoms in rodentsfound evidence for associations between BPA and ADHD-like symptoms(see Problem Formulation, below).

From a mechanistic standpoint, there is also ample evidence thatBPA can contribute to ADHD; BPA has been shown to disrupt the ca-techolaminergic and serotonergic signaling systems, in vitro and in vivo(Komada et al., 2014; Ishido et al., 2004,Ishido et al., 2005, Ishidoet al., 2007; Masuo et al., 2004a; Mizuo et al., 2004; Zhou et al., 2009;Tian et al., 2010; Honma et al., 2006; Matsuda et al., 2012; Miyatakeet al., 2006; Yanagihara et al., 2005; Yoneda et al., 2003; Toyohira

et al., 2003; Itoh et al., 2012; Nakamura et al., 2010; Castro et al., 2013;Marquis and Haynes, 2010); these signaling systems are implicated inthe manifestation of ADHD (Aguiar et al., 2010; Swanson et al., 2007;Wilens, 2008). Thus, the human and animal evidence, along with themechanistic data, point to BPA as a possible environmental contributorto the development of ADHD. However, this connection has not pre-viously been systematically reviewed.

The National Institute of Environmental Health Sciences, Office ofHealth Assessment and Translation (OHAT) developed a systematicreview framework (Rooney et al., 2014) for environmental health re-search questions. The OHAT framework was adapted from well-estab-lished systematic review methods in clinical health research. It stan-dardizes the review process by providing transparent procedures forcollecting evidence, evaluating the validity of the study designs andmethods, rating confidence in the body of evidence, and integratinghuman and animal evidence for a final health effect conclusion. OHAThas conducted a review utilizing this framework; a draft version isavailable online (National Toxicology Program, 2013). We developed asystematic review methodology, based on this framework.

In the present review, we explored the connection between earlyBPA exposure and the development of ADHD in humans, as is suggestedby the human literature. The first step in a systematic review is to de-fine the protocol through a Population-Exposure-Comparator-Outcome(PECO) statement (Table 1). Our review included literature in whichrodents or humans were exposed prenatally (or early postnatally) toBPA, with control groups or low exposure groups included for com-parison. Outcomes included hyperactivity symptoms, diagnoses or be-haviors. Studies were experimental rodent studies, or epidemiologicalhuman studies. We analyzed the rodent studies through meta-analysis.Using the OHAT framework, we synthesized the human and animalevidence streams to arrive at a conclusion about the hazard early-lifeBPA exposure poses for hyperactivity in humans.

2. Methods

Our methodology was designed based on the OHAT frameworkdetailed in Rooney et al. (Rooney et al., 2014) and the National Tox-icology Program's (NTP) Draft Protocol for Systematic Review toEvaluate the Evidence for an Association Between Bisphenol A (BPA)Exposure and Obesity, Appendix 2 (National Toxicology Program,2013). Our protocol is available as Supplementary Information (Docu-ment 1), however, we did not publish or register the protocol prior tocarrying out our review. We acknowledge this may be a limitation, aspublication of systematic review protocols is recommended to promotetransparency and reduce the potential for bias.

Briefly, the framework includes seven steps: 1) ProblemFormulation and Protocol Development, 2) Search and Selection ofStudies for Inclusion, 3) Data Extraction, 4) Assessment of Quality ofIndividual Studies, 5) Rating of Confidence in the Body of Evidence(including meta-analysis), 6) Translation of Confidence Rating toEvidence of Health Effects, and 7) Integration of Animal and HumanEvidence for Hazard Identification Conclusions. Additionally, our ana-lysis was completed using the Preferred Reporting Items for SystematicReviews and Meta-Analysis (PRISMA) statement checklist (seeSupplementary Information Document 2) (Whaley et al., 2016).

2.1. Problem formulation and protocol development

This systematic review arose from an interest in neurodevelop-mental disorders, and how they may be associated with chemicals inthe environment. We started with a preliminary scoping review, priorto the methodology development, to determine which chemicals wereassociated with the development of ADHD in humans and animals.Many postnatal-onset, non-infectious diseases/disorders have unknownetiology and may be due to exposure to environmental factors in thewomb (Dolinoy et al., 2007; Heindel and Vandenberg, 2015).

J.R. Rochester et al. Environment International 114 (2018) 343–356

344

Therefore, our focus was on early life exposure: gestational to threemonths old in humans, and the corresponding development in rodents(i.e., gestation to prepuberty). These exposure time-frames were de-termined based on Bayer et al., 1993, which uses neurological devel-opment to equate human and rodent developmental timelines (Bayeret al., 1993). For the animal studies, we focused on mice and rats, asthey are the most widely used animal models in clinical studies, andseveral diseases and disorders have established mouse or rat models,including ADHD (Russell, 2011). Although other species (e.g., fish, in-sects) could have been included, we chose to focus on rodents as themost widely-used clinical models. From the initial scoping review, weidentified many chemicals associated with ADHD in humans and ro-dents. We chose BPA for several reasons: the robust body of literature inhumans and rodents, relevance to human health (i.e., current scientificinterest in the health effects of BPA and the high potential for humanexposure), and to highlight the relatively new idea that BPA may becontributing to ADHD. This led to our research question “Does earlyexposure to BPA contribute to the development of ADHD in humans?”We then formulated the Population, Exposure, Comparators, and Out-come (PECO) statement (Table 1), which included hyperactivity,among other endpoints related to ADHD. After our initial screeningwith this PECO statement (after endpoint identification but beforedata/outcome extraction), we further narrowed our question to “Doesearly life exposure to BPA contribute to postnatal hyperactivity?” toreduce the scope and increase the feasibility of our review, and in-cluded only the hyperactivity studies.

2.2. Search and selection of studies for inclusion

A comprehensive literature search was performed in order toidentify studies describing the effects of early life exposure to BPA onADHD-related behaviors. Experts were consulted for search criteriainclusion (see Table 2 for search terms). The search included all articlespublished and indexed for all literature up to January 1, 2017 that werein the English language. Electronic searches were performed inPubMed, Web of Science, and Toxline. Two reviewers (AB, JR) screenedall titles and abstracts independently, using the software Distiller SR®(Evidence Partners, Ottawa, Ontario, Canada (Evidence Partners)). Forinclusion in the systematic review, the studies had to contain original

data, be in the English language, include some measure of BPA ex-posure, and assess if BPA exposure affected the development of ADHDor associated behaviors in mice, rats or humans (e.g., hyperactivity,impulsivity, and inattention). Exposure to BPA had to occur at< 3months for humans, up to postnatal day 35 for rats and up to postnatalday 40 for mice, either gestational (via maternal exposure) or directlyto the offspring. The PECO statement, governing which studies wereincluded/excluded, is presented in Table 1. The two reviewers re-conciled study inclusion conflicts or discrepancies through discussion.Fig. 1 shows reasons for full text exclusions (a detailed list of included/excluded studies and reasons for exclusion is available upon authorrequest). We did not hand search the reference lists of included papers.After determining relevant endpoints and selecting studies for full re-view, but before data and outcome extraction, the specific ADHDsymptom “hyperactivity” was chosen.

2.3. Data extraction

Included studies were data-extracted by one reviewer (JR) and thenquality checked by an independent reviewer (AB). The data includedwere: author information, year of publication, animal model and strain,dose level, age dosed, route of exposure, dosing regimen, age whenmeasurements of hyperactivity were assessed, specific type of activity,sample size, and measurement of activity (i.e., mean, standard devia-tion, and standard error). Measures of hyperactivity in rodents includedtotal activity, total horizontal activity, spontaneous motor activity,

Table 1Population-Exposure-Comparator-Outcome (PECO) statement.

Inclusion criteria Exclusion criteria

Participants/Population (observational or experimental studies)

• Humans• Rodents •All non-human primates and non-rodent animals and organisms, includingwildlife, aquatic species, and plants

Exposure

• Exposure to bisphenol A (BPA) during any time between conception to 3months old inhumans, and conception to puberty in rodents (postnatal day 35 in rats and 40 in mice)

• Exposure to controlled doses of BPA via an exposure method (e.g. diet, injection, drinkingwater, gavage)

• Exposure to environmental exposures measure via biological samples of mothers duringgestation and neonates after gestation.

• Exposures measured via environmental monitoring, occupational exposures (i.e. surveys)

• Exposures to chemical mixtures in animals• Exposures to mixtures in humans where specific BPA exposures were notidentified

Comparators

• Low exposure quartiles (observational studies)• Vehicle-only treatment controls (experimental studies) •No determination of outcomes with respect to exposure level in humans

• No controls in experimental studiesOutcomes

• Diagnosis of ADHD or any related symptoms (humans)• Determination of activity levels or other ADHD symptoms via any methodology (e.g.questionnaires for humans, observation for humans/animals)

• All other non-ADHD physiological effects

Publication parameters

• Peer-reviewed• Original data• Studies must be published in English• No original data, e.g., editorials, reviews• Non-peer reviewed: Conference presentations or other studies published in

abstract form only, grant awards/proposals, and theses/dissertations

• Retracted articles

Table 2ADHD and BPA search logic.

Search logica

Disorder andsymptomsearch

(bisphenol-A OR BPA) AND (ADHD OR attention-deficitOR inattention OR inattentiveness OR sustained-attentionOR disruptive-behavior OR habituation OR hyperactivityOR overactivity OR hyperactive OR overactive OR impulseOR impulsiveness OR impulse-control OR impulsivity ORdelay-of-reinforcement OR delayed-reinforcement ORmotor-control OR motor-activity OR disinhibition ORinhibition)

a The search logic was identical for PubMed, Web of Science, and Toxline.


345

distance travelled, and the number of line crosses. In humans, BehaviorAssessment System for Children, Second Edition (BASC-2) scores,Conner's ADHD/DSM-IV (CADS), ADHD Criteria of Diagnostic andStatistical Manual of Mental Disorders-4th Edition (ADHD-DSM-IV),and the Conner's Parent Rating (CPRS) scales specifically assessinghyperactivity were included. Data were extracted from figures orgraphs using the Universal Desktop Ruler® (AVPSoft.com) software asneeded, with measurements taken in triplicate by a single reviewer andquality checked for accuracy by a second reviewer. Authors were con-tacted by email if pertinent data was not reported, with one follow-uprequest if we received no response.

2.4. Assessment of the quality of individual studies

Briefly, risk of bias of experimental methodology was assessed byanswering the 14 animal questions and the 11 human questions in theOHAT Risk of Bias Tool (see Table 3 in the NTP BPA Draft Protocol(National Toxicology Program, 2013)). The risk of bias questions usedin this analysis covered biases in: selection of subjects, performance ofimplementing methodology, accounting for confounding variables, at-trition/exclusion of subjects, detection/outcome assessment (includingquality of exposure assessment), selective reporting of outcomes, andstatistical methodology (see Supplementary Information Document 3

for details). Questions were answered using the following ratings:“definitely low risk of bias,” “probably low risk of bias,” “probably highrisk of bias,” “definitely high risk of bias,” or “not reported” based onstandardized responses. Two reviewers (JR, AB) assessed each studyindependently and discrepancies were resolved through discussion.Contact with authors was attempted up to two times by email if risk ofbias information was not reported.

The ratings of individual answers were then combined and used totier the studies according to study confidence (i.e., high, moderate, orlow). This tiering system is used to remove individual studies with lowconfidence. These are studies that have many components of high riskof bias in their experimental designs, which may lead to bias in thestudy results. (see Supplementary Information Document 4 and OHAT2015 Handbook for a detailed explanation of the tiering process(National Toxicology Program, 2015)).

2.5. Rating confidence in the body of evidence

In this step, we analyzed the pertinent studies in each data stream(i.e., animal and human), which yielded an animal body of evidenceand a human body of evidence. For the animal evidence, only thestudies used in the meta-analysis were included (as discussed below in“Meta-Analysis”). The overall confidence in each body of evidence was

Fig. 1. Flow diagram of study selection process.Selection of included studies occurred in four phases: 1) Identification: potential studies were gathered from three databases; 2) Screening: studies were selected for full-text assessmentusing titles and abstracts, based on relevance; 3) Eligibility: studies were selected for inclusion using a priori criteria; 4) Inclusion: studies were selected for qualitative and quantitativeanalysis.


346

http://AVPSoft.com

evaluated. First, initial confidence in each of the bodies of evidence wasdetermined by evaluating the following key study design features:controlled exposure, exposure prior to the assessment of the outcome,outcome data that was collected at the individual level, and use of

comparison groups. The confidence scale was rated from “high con-fidence” to “very low confidence,” where “high confidence” has all fourfeatures, “moderate confidence,” three features, “low confidence,” twofeatures, and “very low confidence,” one or less features. A well-con-ducted experimental study will have all four features and thus start outwith high initial confidence. Most human epidemiological cohort stu-dies will have three factors (all but controlled exposure) and start out atmoderate initial confidence.

After the initial confidence rating was determined, factors that in-creased or decreased confidence in the body of evidence were con-sidered using a method adapted from the Grades of Recommendation,Assessment, Development and Evaluation (GRADE) approach (Balshemet al., 2011). The body of evidence was evaluated for downgradingusing the following factors: severity of risk of bias, unexplained in-consistency, indirectness with respect to animal models/routes of ex-posure, imprecision of the effect size, and publication bias. Next, up-grading factors were assessed that potentially increase confidence in thebody of evidence: large magnitude of effect, results showing presence ofdose-response (e.g., linear or non-linear), residual confounding (e.g.,bias towards the null), consistency across different models and popu-lation types, and any other relevant experimental features (e.g., rareoutcomes). Decreasing and increasing factors each had the potential tochange the confidence in the body of evidence by up to two levels.Finally, the initial confidence, decreasing factors, and increasing factorswere summed, generating the confidence in the body of evidencerating, which is scaled from “high” to “very low.”

2.6. Meta-analysis

Meta-analysis was used to determine if BPA administration resultedin hyperactive behaviors. For this analysis, only animal studies wereassessed because the body of literature had conflicting results that couldbe evaluated using meta-analytical techniques. Meta-analysis of en-vironmental studies can be difficult, due to the wide range of doses (ascompared to clinical studies which often include only one or twodoses). Although there are several techniques to address this issue (e.g.,pooling doses, splitting the control Ns, choosing one dose, etc.) wedecided that choosing one dose would introduce the least amount ofbias and heterogeneity. We chose 20 μg/kg/day because it was the dosewith the most information (studies), below the EPA's established tol-erable daily intake (TDI) of 50 μg/kg/day. Thus, any information ac-quired from this dose would indicate effects below the TDI, and wouldbe useful for new studies and/or risk assessments. We expanded thedoses included to range 15–25 μg/kg/day to account for dosing marginof error, as per the advice of an expert in the field. Studies that did notencompass these doses were not included in the meta-analysis. Further,only the studies used in the meta-analysis were analyzed for the otherparameters rating the confidence in the body of evidence (such aspublication bias, imprecision, risk of bias, etc.). The outcomes werecontinuous for all experiments and were expressed as means, standarddeviations, and sample sizes. All data from the experiments were ana-lyzed using a random effects model, as it was assumed a priori that thestudies differed for reasons beyond sampling error, therefore our set ofstudies represented a distribution of true effects (Borenstein et al.,2011). All analyses were performed using Comprehensive Meta-Ana-lysis Version 3 ® software (Biostat, Englewood, New Jersey, USA) andsummary effect sizes were expressed using Hedges's g (g), the unbiasedestimate of Cohen's d (i.e., the standardized mean difference)(Borenstein et al., 2011; Vesterinen et al., 2014). Statistical significancewas accepted at p < 0.05. Because of the evidence that ADHD/hy-peractivity is present at different rates in boys and girls (Aguiar et al.,2010), we decided, a priori, to include a subgroup analysis with respectto sex. We also ran confirmatory analysis to see if there were differencesbetween 15, 20, and 25 μg/kg/day. Publication bias was evaluatedquantitatively using an Egger's test. A qualitative analysis of the re-maining animal studies (significance and direction of effects) is

Table 3Characteristics of animal and human studies assessed in this review.

Animal Human

Number (%) Number (%)

Studies total 29 Studies total 3Mean sample size⁎ 13.2, range

3–30Mean sample size 282.2, range

225–361

Publication date Publication date

2016–2010 17 (59) 2016–2010 3 (100)2009–2000 11 (38) 2009–2000 0 (0)1999–1990 1 (3) 1999–1990 0 (0)

Sex Sex

Both 18 (62) Both 3 (100)Females only 1 (3) Females only 0 (0)Males only 10 (35) Males only 0 (0)

Doses assessed (μg/kg/day)

Number (%) ofstudies withdoses withineach range

Exposure range

< 1 1 (3) Braun (Braunet al., 2011)Gestational

< LOD-20 μg/gBPA (creatinineadjusted)

1–10 6 (21) Harley (Harleyet al., 2013)Gestational

< LOD-27 μg/LBPA (specificgravity adjusted)

11–100 17 (59) Casas (Casas et al.,2015) Gestational

< LOD-4.2 μg/gBPA (creatinineadjusted)

101–1000 12 (41)1001–10,000 11 (38)> 10,000 2 (7)

Age of exposure Age of Exposure

Gestational only 3 (10) gestational only 1 (0)Gestational and

prepubertal17 (59) gestational and

childhood2 (100)

Prepubertal only 9 (31) childhood only 0 (0)

Age of outcomemeasurement

Age of outcomemeasurement

After puberty 19 (66) after puberty 0 (0)Before puberty 8 (28) before puberty 3 (100)Both 2 (7) both 0 (0)

Measure ofhyperactivity

Measure ofhyperactivity

Horizontal activitya 25 (86) BASC-2 2 (67)Total activityb 4 (14) CADS 1 (33)

ADHD-DSM-IV 1 (33)CPRS 1 (33)

Animal model

Mouse 11 (38)Rat 18 (62)

BASC-2: Behavior Assessment System for Children, Second Edition; CADS: Conner'sADHD/DSM-IV Scale; CPRS: Conner's Parent Rating Scale; a, horizontal activity: totalhorizontal activity, spontaneous motor activity, distance travelled, and the number of linecrosses; b, total activity: total movement (measured in change in grams, via movement ona scale).

⁎ Studies where N's were not reported were not included (see text).


347

presented in Supplementary Information Document 6. These studieswere not included in determining the confidence in the body of evi-dence.

2.7. Translation of confidence rating to evidence of health effects

Human and animal evidence streams were assessed separately.Confidence in the body of evidence was combined with evidence of aneffect (assessed either through the meta-analysis [animal] or qualita-tively [human]) to determine whether the body of evidence supportedan association between exposure to BPA and hyperactivity. If the bodyof evidence shows evidence of an effect, the confidence rating translatesdirectly to an evidence rating of “high,” “moderate,” or “low,” but “verylow” translates to “inadequate” level of evidence for the health effect. Inthe OHAT framework, if there are no effects seen in the body of evi-dence, a “high” confidence in the body of evidence translates to “evi-dence of no effect,” while “moderate,” “low,” or “very low” confidencein the body of evidence translates to “inadequate” evidence (NationalToxicology Program, 2015).

2.8. Integration of animal and human evidence for hazard identificationconclusions

In the final step, human and animal evidence ratings were in-tegrated to determine the final hazard identification conclusion. Thehazard identification can then be used in the context of risk assessmentto make further recommendations (National Toxicology Program,2015). The OHAT framework has five possible hazard identificationconclusions: 1) known to be a hazard to humans, 2) presumed to be ahazard to humans, 3) suspected to be a hazard to humans, 4) notclassifiable as a hazard to humans, and 5) not identified to be a hazardto humans. These categories are similar to hazard classification labels ofother organizations (National Toxicology Program, 2015). If there wereto be no literature for a particular data stream (e.g., animal or human),the evidence for that data stream is determined to be “inadequate,” andthe conclusion is based on the existing evidence.

3. Results

3.1. Literature search, screening, and data extraction

Study selection is summarized in a flow diagram (Fig.1). Literaturesearches in PubMed identified 463 potentially relevant studies; Web ofScience, 966; and Toxline, 118. After removing duplicate records, thescreening of 913 records using inclusion criteria identified a total of 29

rodent (Komada et al., 2014; Ishido et al., 2004,Ishido et al., 2005,Ishido et al., 2007, Ishido et al., 2011; Masuo et al., 2004a,b; Tian et al.,2010; Matsuda et al., 2012; Adriani et al., 2003; Anderson et al., 2013;Ferguson et al., 2012; Kiguchi et al., 2007,Kiguchi et al., 2008; Negishiet al., 2004; Wolstenholme et al., 2013; Zhou et al., 2011; Farabolliniet al., 1999; Stump et al., 2010; Xu et al., 2007; Nakamura et al., 2012;Nagao et al., 2014; Negishi et al., 2003a,b; van Esterik et al., 2014;Kundakovic et al., 2013; Hass et al., 2016; Heredia et al., 2016; Hickset al., 2016; Rebuli et al., 2015) and three human (Casas et al., 2015;Braun et al., 2011; Harley et al., 2013) studies as relevant. The 29 ro-dent studies encompassed doses from 0.25 to 50,000 μg BPA/kg/day.For the meta-analysis, we included only doses from 15 to 25 μg BPA/kg/day, which included 12 studies and 25 independent effects. Table 3shows the summarized characteristics of the studies. In the includedanimal studies, hyperactivity was tested in several different apperati,including open field boxes, experimental cages, holeboards, tilt cages,and scales. The methods used to assess hyperactivity included infraredphotobeam technology, video and computer-based tracking systems,video and manual recording, live manual recording, tilt-switches, andmovement on a scale (measured in changes in grams) (Komada et al.,2014; Ishido et al., 2004,Ishido et al., 2005, Ishido et al., 2007, Ishidoet al., 2011; Masuo et al., 2004a,b; Tian et al., 2010; Matsuda et al.,2012; Adriani et al., 2003; Anderson et al., 2013; Ferguson et al., 2012;Kiguchi et al., 2007,Kiguchi et al., 2008; Negishi et al., 2003a,b, Negishiet al., 2004; Wolstenholme et al., 2013; Zhou et al., 2011; Farabolliniet al., 1999; Stump et al., 2010; Xu et al., 2007; Nakamura et al., 2012;Nagao et al., 2014;van Esterik et al., 2014; Kundakovic et al., 2013;Hass et al., 2016; Heredia et al., 2016; Hicks et al., 2016; Rebuli et al.,2015). The human measures of hyperactivity were assessed using theBASC-2, CADS, ADHD-DSM-IV, and Conner's Parent Rating scales(Casas et al., 2015; Braun et al., 2011; Harley et al., 2013).

3.2. Individual study quality

The risk of bias results are presented in Fig. 2 and SupplementaryInformation Document 4. All the studies had either high (tier 1) ormoderate (tier 2) confidence, so none of them were removed fromanalysis (see Supplementary Information Document 4 (NationalToxicology Program, 2015)). Only the studies analyzed in the meta-analysis are presented in Fig. 2. Supplementary Information Document4 presents the risk of bias data for all studies.

For an overall picture of risk of bias, the individual risk of biasratings were pooled by type of bias: Selection, Confounding,Performance, Attrition/Exclusion, Detection/Outcome, Reporting, andStatistical (Fig. 2). Studies were that were rated “−” (indicated by

Fig. 2. Risk of bias ratings.Pooled ratings for each type of bias (indicated at top) for animal and human studies included in the meta-analysis. ++ (dark green), definitely low risk of bias; + (light green), probablylow risk of bias; − (yellow), probably high of bias or not-reported; −− (red), definitely high risk of bias. Percentages are reported in Supplementary Information Document 6. (Forinterpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


348

yellow) were “probably high” in their risk of bias ratings, OR hadparameters where the answers were not reported by the study authors.

In general, the animal studies performed well in the areas ofDetection/Outcome bias, and Reporting bias, indicating that they hadlow risk of bias in these areas. The Attrition/Exclusion bias was slightlyhigher, with more studies having a “probably high” risk of bias, and thePerformance bias was even higher, with a few studies having “definitelyhigh” risk of bias. Areas with the highest risks of bias includedSelection, Confounding, and Statistical. These included areas such asblinding, allocation concealment, accounting for confounding factors,and the data passing statistical assumptions. It should be noted that ifthe authors do not report a parameter, it was not known if theseparameters represent true risks of bias or failure of the authors to reportthe parameter.

The quality of the human studies was high: all of studies (Casaset al., 2015; Braun et al., 2011; Harley et al., 2013) identified andcontrolled for relevant covariates, applied well-established methods tomeasure the outcome, and utilized valid methods that allowed for thequantification of BPA in study subjects (according to our methodology,adapted from the OHAT BPA Draft Protocol (National ToxicologyProgram, 2013), see SI S2 Document). Two of the studies did not useacceptable methods to measure covariates, and were rated “probablyhigh risk” for that individual parameter (see Fig. 2 for details) (Casaset al., 2015; Harley et al., 2013).

Since all the studies received either high or moderate ratings, nostudies were removed from the review. The answers and ratings of theindividual questions, as well as the tiering scores, are presented inSupplementary Information Document 4.

3.3. Rating confidence in the body of evidence

3.3.1. Determination of effectWe determined the effect in the human studies qualitatively by

looking at the effects reported in the primary studies and comparingthem overall. All three studies showed effects, but this differed by sex,time of assessments, time of exposure (i.e., prenatal vs. postnatal ex-posure), and who did the assessments (i.e., teacher vs. parents). InBraun et al., prenatal BPA exposure was positively associated withhyperactivity scores in 3 year old girls (Braun et al., 2011). Casas et al.found prenatal BPA to be associated with increased teacher reportedhyperactivity in 4-year old boys, but not in girls. Interestingly, at7 years old, hyperactivity was decreased in girls with respect to BPAexposure, but this association was not significant (Casas et al., 2015).Lastly, Harley et al. did not find associations with prenatal BPA ex-posure. They did find that childhood BPA exposure was associated withincreased teacher-reported hyperactivity at age 7 in girls (Harley et al.,2013), although these data were not included in our analysis, as welimited our analysis to prenatal/early post-natal exposures. It should benoted that there were differences in the populations studied, such astime of assessment. Because of this, we determined that the studieswere too different from each other to be compared in using meta-ana-lysis. For the animal literature, we conducted a meta-analysis to rectifydifferences between studies with positive, negative, and no effects. Themeta-analysis included doses of 15, 20, and 25 μg/kg/day, which ac-counted for 12 of the 29 animal studies and included 25 independentmeasures of hyperactivity. The three doses did not differ significantly(p < 0.993) and thus were combined for all subsequent analyses. TheSupplementary Information Document 5 contains summary statisticsincluding the standard mean difference and confidence intervals of eachstudy.

The forest plots (Figs. 3 and 4) show the individual effects of eachstudy, and the overall meta-values. When males and females wereanalyzed together, BPA exposure did not significantly impact hyper-activity (g= 0.088, 95% CI −0.057–0.232, p=0.237; Fig. 4). How-ever, when males were analyzed separately, there was a significantincrease in hyperactive behaviors with BPA exposure among males

(g= 0.243, 95% CI 0.038–0.448, p=0.020; Fig. 4).Due to observational evidence in humans suggesting a higher pre-

valence of ADHD in males, we completed an evaluation assessing dif-ferences between sexes. There were 12 and 13 data points for femalesand males, respectively. This analysis showed that the males differedsignificantly from females (p < 0.035). Heterogeneity of the data wascalculated using I2 (the distribution of effect sizes). This analysis in-dicates percentage of total variation across studies that is due to het-erogeneity rather than chance (Higgins et al., 2003). Our analysis didnot show a significant difference between studies overall with a verylow I2 value (I2= 5.042, p < 0.391), meaning that the differences ineffect sizes due to variables besides BPA treatment were small. Becausethe I2 value was so low (~5% heterogeneity), we did not pursue sub-group I2 analysis or meta-regression, as this would not have added tothe data for the purposes of hazard identification.

3.3.2. Publication biasAn Egger's Test analysis using the standard mean differences (SMDs)

was non-significant (p=0.33997, 2-tailed), indicating publication biasis unlikely. It has been shown that using SMDs to determine publicationbias can over-estimate asymmetry, leading to false-positive results(Zwetsloot et al., 2017). Since we did not find significant positive ef-fects, there was no danger of having a false-positive, so we did notconduct further analysis (for example by plotting the SMD against asample size-based precision estimate).

3.3.3. Confidence in the body of evidenceThe meta-analysis animal studies (12) and all the human studies (3)

were included in the analysis of the confidence in the body of evidence.For the animal studies, the initial confidence in the body of evidencewas high (++++) due to the fact that all the studies had controlledexposures, exposures prior to outcomes, individual outcome data, andcontrol/comparison groups (Macleod et al., 2008). The risk of bias wasdetermined for all the studies identified. Although there were morehigh quality studies than moderate quality studies, indicating the risk ofbias should not be downgraded (Rooney et al., 2014), we judged thatthe deficiencies in some parameters, such as blinding (important forbehavioral experiments) and randomization (which has been associatedwith increased effect sizes (Macleod et al., 2008)), presented a sub-stantial risk of bias, and thus we downgraded the body of evidence onelevel. Other downgradable factors (indirectness, imprecision, un-explained inconsistency, and publication bias) were acceptable and didnot warrant a downgrade. Because these factors were determined bymeta-analysis, only the studies used in the meta-analysis were includedfor these assessments. Specifically, no imprecision was found, viamethods outlined in the OHAT Handbook (National ToxicologyProgram, 2015, page 59). No unexplained inconsistency was foundusing I2statistics (0–40%, judged as “might not be important” (NationalToxicology Program, 2015, page 54). Similarly, quantitative analysis ofthe publication bias determined no bias. The animal body of evidencewas upgraded once due to plausible dose-responses. This was because,across the entire animal body of evidence, all the male significant ef-fects showed a linear dose-response. The methodology and data for thisanalysis is presented in Supplementary Document 7. No other factors(large magnitude of effect, residual confounding, consistency, or otherfactors) warranted an upgrade. Thus, the final rating for the animalbody of evidence was “high” (++++, Fig. 5).

For the human studies, we qualitatively assessed the body of evi-dence. The body of evidence was initially rated as moderate (+++)because of the following positive characteristics: outcomes prior toexposure, individual outcome data, and comparison groups (i.e., lowestquartile groups). Epidemiological studies inherently do not have con-trolled exposures, thus the body of evidence could not receive thehighest rating. The human body of evidence was downgraded once forhaving unexplained inconsistency (due to the fact that there were dif-ferent outcomes, based on age, sex, and outcome assessment). The body


349

of evidence was upgraded once for dose-response curves across allstudies that supported the plausible biological effects of BPA. BecauseBPA has been shown to illicit both non-monotonic and monotonic (i.e.,non-linear and linear) dose-response curves in a wide variety of or-ganisms (Vandenberg, 2014), we accepted both types of curves as abasis for upgrading. To determine if a dose response was present, all theeffects were standardized, pooled, and dose-response was examinedacross the entire human body of literature (rather than in each in-dividual study) for two age groups (3–4 year olds and 7 year olds), andfor each sex. The methodology and data are presented in SupplementalDocument 8. For the 3–4 year olds, there was a “U” shaped (or possiblypositive linear) response in girls, and for the boys there appeared to bean inverted “U” shaped response. While there was no significant effectin girls at 7 years old with prenatal exposures, there was a trend ofdecreasing hyperactivity (Harley et al., 2013). Interestingly, we foundthat there was a decreased linear dose-response across all studies(which assessed them) in girls at 7 years old. Thus, the final confidencerating for the human body of evidence was “moderate” (+++, Fig. 6).

3.4. Translation of confidence rating to evidence of health effects

The meta-analysis for the animal body of evidence showed a sig-nificant effect for males. Therefore, the “high” confidence rating de-rived in Step 5 for the animal body of evidence was translated to a“high” level of evidence for hyperactivity. The human body of evidenceshowed significant effects in all three studies, although these effects

differed by age and sex (Casas et al., 2015; Braun et al., 2011; Harleyet al., 2013). Therefore, the “moderate” confidence rating from thehuman body of evidence was translated to a “moderate” level of evi-dence for hyperactivity.

3.5. Hazard identification

The hazard identification was determined by integrating the humanand animal ratings for evidence of health effects from the previous step,as shown in Fig. 7. The final rating identified BPA as a “presumed”hazard to human health with regard to hyperactivity.

4. Discussion

Our systematic review found that early BPA exposure is associatedwith a presumed hazard of hyperactivity in humans. Our conclusion isbased on “moderate” levels of evidence for the human and “high” levelsof evidence for animal literature. The systematic review included all theavailable literature up to January 1, 2017; 29 animal studies, and threehuman studies. A meta-analysis of studies in animals showed increasedhyperactivity with early life exposure to BPA in males. In humans, allthree studies showed significant effects of BPA on increased hyper-activity, although there were differences in the sexes and windows ofexposure. These results, along with a qualitative and quantitativeanalysis of the entire human and rodent bodies of literature (e.g., risk ofbias, publication bias, dose-response, etc.) were quantitatively

Fig. 3. Summary of studies measuring hyperactivity in animals dosed at 20 μg/kg BW using a random effects model.Each square and whisker represents the effect size and 95% confidence interval for the study, and the size of the square indicates the relative weight of the study in the calculation of themeta-value. The diamond represents the meta-value and the 95% confidence intervals. F: female; M: male; A: adult; J: juvenile; B: both sexes.


350

Fig. 4. Male vs. female hyperactivity using a random effects model.Each circle and whisker represents the effect size and 95% confidence interval for the study, and the size of the circle indicates the relative weight of the study in the calculation of themeta-value. The diamond represents the meta-value and the 95% confidence intervals. F: female; M: male; A: adult; J: juvenile; B: both sexes.

Fig. 5. Confidence in the body of evidence evaluationfor animal studies.+ indicates an upgrade in the confidence in the bodyof evidence; − indicates a downgrade in the con-fidence in the body of evidence. Summation of the“+’s” and “−‘s” in each column equals the number inthe bottom row. Summation of the “+’s” in the bottomrow results in the final rating in the bottom right-handbox.


351

integrated to arrive at the final hazard identification conclusion.This review also identified several interesting outcomes and ideas

for further research. These, along with a discussion of the strength andlimitations of the review and possible modes of action of BPA on neu-rodevelopment, are examined below.

4.1. Sex

In our analysis, BPA showed a significant effect on male, but notfemale, rodent hyperactivity. However, it should be noted that we didnot account for any other confounders in our analysis, so it is possiblethese differences were due to other factors. Interestingly, the humanresults in our review showed mixed outcomes: some had a larger effectin girls, and some in boys. These results, however, did not change our

hazard identification conclusion, as the goal is to identify and protectthe most sensitive populations, and both males and females are exposedin the general population. Because of this, any effect, regardless of sexor other possible variants to the population, is considered in the“Evidence of Health Effect” step of this methodology.

In the human population, ADHD is more prevalent in boys than girls(Aguiar et al., 2010), with the caveat that referral bias appears to causethe under-diagnosis of ADHD in girls (Gershon, 2002). Further, subtypedistributions appear to change in adulthood; studies indicate womenare diagnosed with the combined subtype more frequently than menand suffer greater impairment on several measures of ADHD symptoms(Robison et al., 2008; Rucklidge, 2010).

It is common for effects to vary across sexes (Palanza and Gioiosa,2008), particularly with endocrine disrupting chemicals, although the

Fig. 6. Confidence in the body of evidence evaluationfor human studies.+ indicates an upgrade in the confidence in the bodyof evidence; − indicates a downgrade in the con-fidence in the body of evidence. Summation of the“+’s” and “−‘s” in each column equals the number inthe bottom row. Summation of the “+’s” in the bottomrow results in the final rating in the bottom right-handbox.

Fig. 7. Hazard identification conclusion for early life exposure to BPA and hyperactivity.This shows the integration of human and animal evidence streams. A “high” level of evidence for animals and a “moderate” level of evidence for humans translate to the conclusion thatearly BPA exposure is associated with a presumed hazard of hyperactivity in humans.


352

exact mechanisms are often unclear. However, these sex differences dosuggest an endocrine basis for the differences in effects seen in thesestudies. Interestingly, a recent study found that being born with hy-pospadias, a disorder of the penis caused by disrupted steroids beforebirth, was associated with increased likelihood of having ADHD andother neurodevelopmental disorders (Butwicka et al., 2015). In two ofthe human studies in the present analysis, early BPA exposure was as-sociated with hyperactivity in girls, but not boys (Braun et al., 2011;Harley et al., 2013). In one study, hyperactivity in girls at age 7 wascorrelated with BPA exposure at 5 years of age but not gestational ex-posure (Harley et al., 2013). In contrast, the other human study showedthat gestational exposure to BPA was correlated with hyperactivity ingirls, while childhood exposure (i.e., 1–3 years of age) was not (Braunet al., 2011). A third study showed increased hyperactivity in boys, butnot girls, which attenuated with age (Casas et al., 2015). Additionalstudies could be done to understand the role of sex and windows ofsusceptibility with respect to hyperactive behaviors. These include afocus on female rodent data, which is lacking, as well as comparisons inrodents between different time periods of exposure during gestationand postnatally (i.e., early gestation vs. late gestation vs. postnatalexposures). Another interesting research avenue that has not been well-studied is whether levels of hyperactivity change as rodents age (as theyappear to do in humans). Although studies such as these would help tofurther understand the gene-environment interaction of the develop-ment of this disorder, the current body of evidence points to BPA ex-posure contributing to hyperactivity, and research resources should becarefully allotted for maximum protection of human health.

4.2. Dose and dose response

We chose to conduct a meta-analysis of the 20 μg/kg/day dose,because that was the most frequently studied dose in the body of evi-dence below the established TDI of 50 μg/kg/day. We included15–25 μg/kg/day in order to account for dosing margin of error withinthe studies (an analysis of these doses (Willcutt et al., 2005; Frankeet al., 2009; Vandenberg et al., 2013) showed no differences betweendoses). Because we found effects at 20 μg/kg/day, regulators should re-consider the acceptable “safe” level for BPA exposure in humans, as hasbeen previously suggested (Vandenberg et al., 2012), and new risk as-sessments of sensitive endpoints (such as neurodevelopment) should beconsidered. In the rodent studies, significant meta-effects were seen inthe males, but not in the females. When examining the male studiesacross the body of evidence, all the studies that showed individualpositive significant effects also showed a liner dose-response, with in-creasing doses resulting in increased hyperactivity. In humans, therewas also a dose response when all the studies were examined together;however, it appeared to be an inverted “U-shaped” (non-monotonic)response in boys, and a “U-shaped” (or positive linear) response in girls.Non-linear dose-response curves are common in low-dose and endo-crine mediated toxicity (Vandenberg et al., 2013) and it would not besurprising if dose responses differed across sexes, especially if one sexwas more susceptible than the other. According to current US risk as-sessment protocols, a thorough dose-response analysis and exposureassessment, including an exploration of the sex differences, timing ofexposure and assessment, and other sources of heterogeneity, would bethe next steps toward full understanding of the neurodevelopmentalrisks posed by early life exposure to BPA.

5. Strengths and limitations of systematic review

We chose to base our methodology on the OHAT framework, in part,because of its usefulness in integrating several data streams (i.e., invitro, in vivo, and human) even if one or more of the data streams hasfew (or no) studies. Often, there is little human data available, parti-cularly in the environmental toxicology/health fields, due to the timeand expense of conducting epidemiological studies and the ethical

barriers for human controlled trials that involve human exposure topotentially hazardous chemicals. Thus, it is important to have methodsfor using animal data to inform human health hazard conclusions; in-deed, animal models are traditionally used to study human health (deVries et al., 2014).

Additionally, one of the strengths of this framework is that it canadjust for potential flaws (such as increased risk of bias in the meth-odologies), to allow conclusions to be drawn based on the availableliterature. Another strength is that these potential deficits and data gapscan be identified in order promote research that will strengthen thebody of evidence in the most effective manner.

A possible limitation of the current systematic review was the ex-clusion of some data and some literature. All non-English papers wereexcluded from the search a priori due to a lack of resources for trans-lating services. The meta-analysis of the animal studies was based ononly doses ranging from 15 to 25 μg/kg/day, in order to reduce het-erogeneity. Because of this limitation, we recommend additional riskassessment/dose analysis for BPA with regard to this endpoint.

Our analysis was also potentially limited by the inability to differ-entiate between actual bias and the failure of authors to report howthey handled the methodological parameters evaluated in the sys-tematic review (e.g., blinding, randomization, statistical assumptions).The risk of bias for the animal literature was generally low, except for incertain domains, including Selection, Confounding, and Statistical(Fig. 2). Most deficits were in the areas of randomization, controllingfor confounding variables, unintended endocrine-disruptive exposures(i.e., phytoestrogens from rodent chow or BPA from housing), lack ofblinding of subjective measures (i.e., behavior), and statistical deficits(i.e., not testing for ANOVA [analysis of variance] assumptions). Betterratings of any or all of these risk of bias parameters would have allowedfor a stronger confidence in the animal body of evidence (see Fig. 2 andSupplementary Information Document 4). That is, if a methodologicalstep or question was not clearly reported in the study, it was rated“probably high” risk of bias (e.g., “there is insufficient informationprovided about how subjects were allocated to study groups”). Forexample, many of the statistical problems were due to the fact that theresearchers did not report whether or not they tested for assumptionswhen carrying out their ANOVA analyses, and did not respond to in-quiry. In total, all the rodent studies (29) were lacking the reporting ofone or more risk of bias parameters. Authors were contacted up to twotimes, but responses were received for only 59% of the studies. Becauseof the difficulty of contacting authors, authors should follow thoroughreporting guidelines, such as the ARRIVE guidelines (Kilkenny et al.,2010), and should include reporting of all risk of bias parameters.Further, peer-reviewers should insist that these elements be reportedclearly in submitted manuscripts. For a list of risk of bias parametersthat should be addressed in publications, see the Risk of Bias Tool (SI S2Document), and for clarification of the questions, see the OHAT fra-mework (Rooney et al., 2014; National Toxicology Program, 2015) andthe PRISMA checklist (SI S1 Document).

6. Mechanisms

In the OHAT systematic review framework, in vitro (and in vivomechanistic) studies are utilized as support for the in vivo human andanimal data. This information can have important implications for thefinal hazard identification rating (Rooney et al., 2014). In the case ofthis systematic review, there are no strongly established mechanismsfor ADHD or hyperactivity (Aguiar et al., 2010), and it would be un-founded to use the current in vitro literature to modify our final hazardrating. However, although BPA has been traditionally thought of asdisrupting the estrogen signaling system (Rubin, 2011), there is a largebody of literature linking BPA exposure to changes in the catechola-minergic and serotonergic signaling systems, which are suspected to beassociated with ADHD etiology (Aguiar et al., 2010; Swanson et al.,2007; Wilens, 2008). While an extensive review of this topic is beyond


353

the scope of this paper, we have included a short overview of BPA'seffects on the development of catecholaminergic and serotonergic re-gions of the brain relevant to ADHD.

ADHD has been associated with neurological dysfunction in hu-mans, particularly in the prefrontal cortex, caudate nucleus, cere-bellum, and corpus callosum. Catecholaminergic and serotonergic sys-tems have also been implicated in the mechanism of ADHD (Aguiaret al., 2010; Wilens, 2008). Dopamine, norepinephrine, and theirtransporters are important signaling molecules in the prefrontal cortex,and in the case of dopamine and dopamine transporter, the striatum.Disruptions in both of these brain areas are associated with ADHD(Aguiar et al., 2010), as well as disruptions in dopamine signaling,disrupted dopaminergic circuitry, and increased dopamine transporter(which reduces dopamine in the synapses). Stimulant drugs used totreat ADHD symptoms work to block dopamine transporter, and thusincrease dopamine levels in the synapses (Swanson et al., 2007; Wilens,2008).

Serotonin (5-HT) signaling has also been implicated in the etiologyof ADHD, although its contribution to the disorder is not as wellcharacterized. Neuroimaging studies have found that 5-HT transporter(SERT) activity is altered in ADHD patients, and there may be alteredsynthesis of 5-HT in ADHD (Oades, 2008). Further, interactions of do-pamine and 5-HT appear to be important in regards to hyperactivity. Ina knock-out dopamine transporter mouse model, drugs that modulated5-HT and SERT activities were shown to reduce hyper-locomotion(Gainetdinov et al., 1999). Some amphetamine-type stimulants used totreat ADHD also modulate serotonin activity, although it is thought thatthe therapeutic action of these drugs is not due to changes in extra-cellular 5-HT (Solanto et al., 2001).

Many studies have shown that BPA causes disruptions to the do-paminergic system in animals, generally downregulating the system(Komada et al., 2014; Ishido et al., 2004; Ishido et al., 2005; Masuoet al., 2004a; Mizuo et al., 2004; Zhou et al., 2009; Ishido et al., 2007;Tian et al., 2010; Honma et al., 2006; Matsuda et al., 2012). In vitroevidence also supports increased catecholaminergic activity with BPAtreatment (Miyatake et al., 2006; Yanagihara et al., 2005; Yoneda et al.,2003), and BPA may inhibit norepinephrine transporter (Toyohiraet al., 2003). BPA exposure has also been shown to alter the ser-otonergic system in vivo, generally appearing to upregulate activity.BPA exposure in rodents increased serotonin and its metabolite 5-hy-droxyindoleacetic acid (5-HIAA) (Honma et al., 2006; Itoh et al., 2012;Nakamura et al., 2010). Adult rats exposed to BPA also showed in-creased tryptophan hydroxylase 2 activity, the rate-limiting enzymeinvolved in serotonin synthesis (Castro et al., 2013). In vitro, however,BPA has been shown to inhibit the release of serotonin, by decreasingexocytotic function of cells (Marquis and Haynes, 2010). Clearly, earlyBPA exposure can alter these systems, although it is unclear what im-plications these data have for the development or etiology of hyper-activity/ADHD in rodents and humans.

7. Conclusions

We have found, using the OHAT systematic review framework, thatearly exposure to BPA is a presumed health hazard linked to hyperactivebehaviors in humans. This conclusion is based on the integration ofanimal and human literature in a rigorous analysis of the body of evi-dence that included evaluation of study quality, meta-analysis, and theconsideration of other factors such as publication bias and dose-re-sponse. This review also identified data gaps and the need for additionalrisk assessments, including assessments of timing of exposure and athorough dose-response analysis.

Although currently not widely used in the environmental scienceand regulatory arenas, systematic reviews such as this one are im-perative for understanding the relationship between environmentalexposures and health effects in humans. Likewise, systematic reviewsand meta-analyses are valuable tools that should be used in order to

translate science to policy for regulatory purposes, partly because theyallow for the assessment of endpoints that might not be evaluated intraditional health regulatory assessments (e.g., behavioral or non-cancer disease endpoints). Current efforts to devise appropriatemethods of assessing harmful chemicals should adopt systematic reviewas a valid means of hazard identification, due to its transparent, stan-dardized and comprehensive procedures. Further, dose-response andexposure analyses are warranted to characterize the risks of BPA ex-posure, so that they can be effectively managed. In the meantime, givenour conclusion that BPA is presumed to be a hazard to human health,steps should be taken to reduce exposure in the most vulnerable po-pulations.

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envint.2017.12.028.

Acknowledgments

The authors would like to thank Dr. Andrew Rooney for his criticalreview of an early draft of the manuscript and Christina Ribbens for herassistance with literature procurement. We would also like to thank Dr.Heather Patisaul and Dr. Laura Vandenberg for their expert consulta-tion.

Funding

The Arkansas Community Foundation, the Winslow Foundation, theCornell Douglas Foundation, and the Wallace Genetic Foundation pro-vided funding for this study.

Contributions

JRR and ALB conceived the research question, carried out the pro-tocol and meta-analysis, and wrote the manuscript. CFK participated inthe conception of the research question and writing and editing of themanuscript.

References

Abbott, L.C., Winzer-Serhan, U.H., 2012. Smoking during pregnancy: lessons learnedfrom epidemiological studies and experimental studies using animal models. Crit.Rev. Toxicol. 42 (4), 279–303.

Adriani, W., Della Seta, D., Dessì-Fulgheri, F., Farabollini, F., Laviola, G., 2003. Alteredprofiles of spontaneous novelty seeking, impulsive behavior, and response to D-am-phetamine in rats perinatally exposed to bisphenol A. Environ. Health Perspect. 111(4), 395–401.

Aguiar, A., Eubig, P.A., Schantz, S.L., 2010. Attention deficit/hyperactivity disorder: afocused overview for children's environmental health researchers. Environ. HealthPerspect. 118 (12), 1646–1653.

American Psychiatric Association, 2013. Diagnostic and statistical manual of mentaldisorders, (DSM-5®). American Psychiatric Pub.

Anderson, O.S., Peterson, K.E., Sanchez, B.N., Zhang, Z., Mancuso, P., Dolinoy, D.C.,2013. Perinatal bisphenol A exposure promotes hyperactivity, lean body composi-tion, and hormonal responses across the murine life course. FASEB J. 27 (4),1784–1792.

Arnold, L.E., Lofthouse, N., Hurt, E., 2012. Artificial food colors and attention-deficit/hyperactivity symptoms: conclusions to dye for. Neurotherapeutics 9 (3), 599–609.

Baird, J., Stevenson, J.C., Williams, D.C., 2000. The evolution of ADHD: a disorder ofcommunication? [Review]. Q. Rev. Biol. 75 (1), 17–35.

Balshem, H., Helfand, M., Schunemann, H.J., Oxman, A.D., Kunz, R., Brozek, J., et al.,2011. GRADE guidelines: 3. Rating the quality of evidence. J. Clin. Epidemiol. 64 (4),401–406. http://dx.doi.org/10.1016/j.jclinepi.2010.07.015. (PubMed PMID:21208779).

Banerjee, T.D., Middleton, F., Faraone, S.V., 2007. Environmental risk factors for atten-tion-deficit hyperactivity disorder. Acta Paediatr. 96 (9), 1269–1274.

Bayer, S.A., Altman, J., Russo, R.J., Zhang, X., 1993. Timetables of neurogenesis in thehuman brain based on experimentally determined patterns in the rat.Neurotoxicology 14 (1), 83–144 (PubMed PMID: 8361683).

Borenstein, M., Hedges, L.V., Higgins, J.P.T., Rothstein, H.R., 2011. Introduction to Meta-analysis. Wiley.

Boyle, C.A., Boulet, S., Schieve, L.A., Cohen, R.A., Blumberg, S.J., Yeargin-Allsopp, M.,et al., 2011. Trends in the prevalence of developmental disabilities in US children,1997–2008. Pediatrics 127 (6), 1034–1042. http://dx.doi.org/10.1542/peds.2010-2989. (PubMed PMID: 21606152).

Braun, J.M., Hauser, R., 2011. Bisphenol A and children's health. Curr. Opin. Pediatr. 23


354

https://doi.org/10.1016/j.envint.2017.12.028https://doi.org/10.1016/j.envint.2017.12.028http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0005http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0005http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0005http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0010http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0010http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0010http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0010http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0015http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0015http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0015http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0020http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0020http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0025http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0025http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0025http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0025http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0030http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0030http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0035http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0035http://dx.doi.org/10.1016/j.jclinepi.2010.07.015http://dx.doi.org/10.1016/j.jclinepi.2010.07.015http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0045http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0045http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0050http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0050http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0050http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0055http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0055http://dx.doi.org/10.1542/peds.2010-2989http://dx.doi.org/10.1542/peds.2010-2989http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0065

(2), 233–239.Braun, J.M., Yolton, K., Dietrich, K.N., Hornung, R., Ye, X., Calafat, A.M., et al., 2009.

Prenatal bisphenol A exposure and early childhood behavior. Environ. HealthPerspect. 117 (12), 1945–1952.

Braun, J.M., Kalkbrenner, A.E., Calafat, A.M., Yolton, K., Ye, X., Dietrich, K.N., et al.,2011. Impact of early-life bisphenol A exposure on behavior and executive functionin children. Pediatrics 128 (5), 873–882.

Butwicka, A., Lichtenstein, P., Landen, M., Nordenvall, A.S., Nordenstrom, A.,Nordenskjold, A., et al., 2015. Hypospadias and increased risk for neurodevelop-mental disorders. J. Child Psychol. Psychiatry 56 (2), 155–156. http://dx.doi.org/10.1111/jcpp.12290. (PubMed PMID: 25048198).

Casas, M., Forns, J., Martinez, D., Avella-Garcia, C., Valvi, D., Ballesteros-Gomez, A.,et al., 2015. Exposure to bisphenol A during pregnancy and child neuropsychologicaldevelopment in the INMA-Sabadell cohort. Environ. Res. 142, 671–679. http://dx.doi.org/10.1016/j.envres.2015.07.024. (PubMed PMID: 26343751).

Castro, B., Sanchez, P., Torres, J.M., Ortega, E., 2013. Effects of adult exposure to bi-sphenol a on genes involved in the physiopathology of rat prefrontal cortex. PLoS One8 (9), e73584.

de Cock, M., Maas, Y.G., van de Bor, M., 2012. Does perinatal exposure to endocrinedisruptors induce autism spectrum and attention deficit hyperactivity disorders? Rev.Acta Paediatr. 101 (8), 811–818.

Daneshparvar, M., Mostafavi, S.A., Zare Jeddi, M., Yunesian, M., Mesdaghinia, A., Mahvi,A.H., et al., 2016. The role of lead exposure on attention-deficit/hyperactivity dis-order in children: a systematic review. Iran J. Psychiatry 11 (1), 1–14 (PubMed PMID:27252763; PubMed Central PMCID: PMCPMC4888135).

DiFranza, J.R., Aligne, C.A., Weitzman, M., 2004. Prenatal and postnatal environmentaltobacco smoke exposure and children's health. Pediatrics 113 (4 Suppl), 1007–1015(PubMed PMID: 15060193).

Dolinoy, D.C., Weidman, J.R., Jirtle, R.L., 2007. Epigenetic gene regulation: linking earlydevelopmental environment to adult disease. Reprod. Toxicol. 23 (3), 297–307.http://dx.doi.org/10.1016/j.reprotox.2006.08.012. (PubMed PMID: 17046196).

van Esterik, J.C., Dolle, M.E., Lamoree, M.H., van Leeuwen, S.P., Hamers, T., Legler, J.,et al., 2014. Programming of metabolic effects in C57BL/6JxFVB mice by exposure tobisphenol A during gestation and lactation. Toxicology 321, 40–52. http://dx.doi.org/10.1016/j.tox.2014.04.001. (PubMed PMID: 24726836, Epub 2014/04/15).

Eubig, P.A., Aguiar, A., Schantz, S.L., 2010. Lead and PCBs as risk factors for attentiondeficit/hyperactivity disorder. Environ. Health Perspect. 118 (12), 1654–1667.

Evidence Partners. DistillerSR. URL:http://distillercer.com/products/distillersr-systematic-review-software/ Ottawa, Ontario, Canada 2015 [updated 2015; cited2015 April 10]. Available from:http://distillercer.com/products/distillersr-systematic-review-software/.

Farabollini, F., Porrini, S., Dessi-Fulgheri, F., 1999. Perinatal exposure to the estrogenicpollutant bisphenol A affects behavior in male and female rats. Pharmacol. Biochem.Behav. 64 (4), 687–694.

Ferguson, S.A., Law, C.D., Abshire, J.S., 2012. Developmental treatment with bisphenol Acauses few alterations on measures of postweaning activity and learning.Neurotoxicol. Teratol. 34 (6), 598–606.

Franke, B., Neale, B.M., Faraone, S.V., 2009. Genome-wide association studies in ADHD.Hum. Genet. 126 (1), 13–50. http://dx.doi.org/10.1007/s00439-009-0663-4.(PubMed PMID: 19384554; PubMed Central PMCID: PMC3774416).

Gainetdinov, R.R., Wetsel, W.C., Jones, S.R., Levin, E.D., Jaber, M., Caron, M.G., 1999.Role of serotonin in the paradoxical calming effect of psychostimulants on hyper-activity. Science 283 (5400), 397–401 (PubMed PMID: 9888856).

Gershon, J., 2002. A meta-analytic review of gender differences in ADHD. J. Atten.Disord. 5 (3), 143–154 (PubMed PMID: 11911007).

Harley, K.G., Gunier, R.B., Kogut, K., Johnson, C., Bradman, A., Calafat, A.M., et al., 2013.Prenatal and early childhood bisphenol A concentrations and behavior in school-agedchildren. Environ. Res. 126, 43–50. http://dx.doi.org/10.1016/j.envres.2013.06.004. (PubMed PMID: 23870093; PubMed Central PMCID: PMC3805756).

Hass, U., Christiansen, S., Boberg, J., Rasmussen, M.G., Mandrup, K., Axelstad, M., 2016.Low-dose effect of developmental bisphenol A exposure on sperm count and beha-viour in rats. Andrology 4 (4), 594–607. http://dx.doi.org/10.1111/andr.12176.(PubMed PMID: 27089241).

Heindel, J.J., Vandenberg, L.N., 2015. Developmental origins of health and disease: aparadigm for understanding disease cause and prevention. Curr. Opin. Pediatr. 27(2), 248–253. http://dx.doi.org/10.1097/MOP.0000000000000191. (PubMed PMID:25635586).

Heredia, L., Belles, M., MI LL, Domingo JL, Linares V., 2016. Behavioral effects in mice ofpostnatal exposure to low-doses of 137-cesium and bisphenol A. Toxicology 340,10–16. http://dx.doi.org/10.1016/j.tox.2015.12.002. (PubMed PMID: 26719215).

Hicks, K.D., Sullivan, A.W., Cao, J., Sluzas, E., Rebuli, M., Patisaul, H.B., 2016. Interactionof bisphenol A (BPA) and soy phytoestrogens on sexually dimorphic sociosexualbehaviors in male and female rats. Horm. Behav. 84, 121–126. http://dx.doi.org/10.1016/j.yhbeh.2016.06.010. (PubMed PMID: 27373758; PubMed Central PMCID:PMCPMC4996731).

Higgins, J.P., Thompson, S.G., Deeks, J.J., Altman, D.G., 2003. Measuring inconsistencyin meta-analyses. BMJ 327 (7414), 557–560. http://dx.doi.org/10.1136/bmj.327.7414.557. (PubMed PMID: 12958120; PubMed Central PMCID: PMCPMC192859).

Honma, T., Miyagawa, M., Suda, M., Wang, R.S., Kobayashi, K., Sekiguchi, S., 2006.Effects of perinatal exposure to bisphenol A on brain neurotransmitters in female ratoffspring. Ind. Health 44 (3), 510–524.

Ishido, M., Masuo, Y., Kunimoto, M., Oka, S., Morita, M., 2004. Bisphenol A causes hy-peractivity in the rat concomitantly with impairment of tyrosine hydroxylase im-munoreactivity. J. Neurosci. Res. 76 (3), 423–433.

Ishido, M., Morita, M., Oka, S., Masuo, Y., 2005. Alteration of gene expression of Gprotein-coupled receptors in endocrine disruptors-caused hyperactive rats. Regul.

Pept. 126 (1–2), 145–153.Ishido, M., Yonemoto, J., Morita, M., 2007. Mesencephalic neurodegeneration in the

orally administered bisphenol A-caused hyperactive rats. Toxicol. Lett. 173 (1),66–72.

Ishido, M., Masuo, Y., Terasaki, M., Morita, M., 2011. Rat hyperactivity by bisphenol A,but not by its derivatives, 3-hydroxybisphenol A or bisphenol A 3,4-quinone. Toxicol.Lett. 206 (3), 300–305.

Itoh, K., Yaoi, T., Fushiki, S., 2012. Bisphenol A, an endocrine-disrupting chemical, andbrain development. Neuropathology 32 (4), 447–457.

Jain, R., Jain, S., Montano, C.B., 2017. Addressing diagnosis and treatment gaps in adultswith attention-deficit/hyperactivity disorder. Prim Care Companion CNS Disord. 19(5). http://dx.doi.org/10.4088/PCC.17nr02153. (PubMed PMID: 28906602).

Kessler, R.C., Adler, L., Barkley, R., Biederman, J., Conners, C.K., Demler, O., et al., 2006.The prevalence and correlates of adult ADHD in the United States: results from theNational Comorbidity Survey Replication. Am. J. Psychiatry 163 (4), 716–723.http://dx.doi.org/10.1176/ajp.2006.163.4.716. (PubMed PMID: 16585449; PubMedCentral PMCID: PMCPMC2859678).

Kiguchi, M., Fujita, S., Lee, J., Shimizu, N., Koshikawa, N., 2007. Behavioral responses tomethylphenidate and apomorphine in rats exposed neonatally to bisphenol-A. J. OralSci. 49 (4), 311–318.

Kiguchi, M., Fujita, S., Oki, H., Shimizu, N., Cools, A.R., Koshikawa, N., 2008.Behavioural characterisation of rats exposed neonatally to bisphenol-A: responses toa novel environment and to methylphenidate challenge in a putative model of at-tention-deficit hyperactivity disorder. J. Neural Transm. 115 (7), 1079–1085.

Kilkenny, C., Browne, W., Cuthill, I.C., Emerson, M., Altman, D.G., Group NCRRGW,2010. Animal research: reporting in vivo experiments: the ARRIVE guidelines. Br. J.Pharmacol. 160 (7), 1577–1579. http://dx.doi.org/10.1111/j.1476-5381.2010.00872.x. (PubMed PMID: 20649561; PubMed Central PMCID: PMCPMC2936830).

Komada, M., Itoh, S., Kawachi, K., Kagawa, N., Ikeda, Y., Nagao, T., 2014. Newborn miceexposed prenatally to bisphenol A show hyperactivity and defective neocortical de-velopment. Toxicology 323, 51–60. http://dx.doi.org/10.1016/j.tox.2014.06.009.(PubMed PMID: 24953688).

Kundakovic, M., Gudsnuk, K., Franks, B., Madrid, J., Miller, R.L., Perera, F.P., et al., 2013.Sex-specific epigenetic disruption and behavioral changes following low-dose inutero bisphenol A exposure. Proc. Natl. Acad. Sci. U. S. A. 110 (24), 9956–9961.http://dx.doi.org/10.1073/pnas.1214056110. (PubMed PMID: 23716699; PubMedCentral PMCID: PMC3683772; Epub 2013/05/30).

Macleod, M.R., van der Worp, H.B., Sena, E.S., Howells, D.W., Dirnagl, U., Donnan, G.A.,2008. Evidence for the efficacy of NXY-059 in experimental focal cerebral ischaemiais confounded by study quality. Stroke 39 (10), 2824–2829. http://dx.doi.org/10.1161/STROKEAHA.108.515957. (PubMed PMID: 18635842).

Marquis, B.J., Haynes, C.L., 2010. Evaluating the effects of immunotoxicants using carbonfiber microelectrode amperometry. Anal. Bioanal. Chem. 398 (7–8), 2979–2985.

Maserejian, N.N., Trachtenberg, F.L., Hauser, R., McKinlay, S., Shrader, P., Tavares, M.,et al., 2012. Dental composite restorations and psychosocial function in children.Pediatrics 130 (2), e328–38. http://dx.doi.org/10.1542/peds.2011-3374. (PubMedPMID: 22802599; PubMed Central PMCID: PMC3408688).

Masuo, Y., Ishido, M., Morita, M., Oka, S., 2004a. Effects of neonatal treatment with 6-hydroxydopamine and endocrine disruptors on motor activity and gene expression inrats. Neural Plast. 11 (1–2), 59–76.

Masuo, Y., Morita, M., Oka, S., Ishido, M., 2004b. Motor hyperactivity caused by a deficitin dopaminergic neurons and the effects of endocrine disruptors: a study inspired bythe physiological roles of PACAP in the brain. Regul. Pept. 123 (1–3), 225–234.

Matsuda, S., Matsuzawa, D., Ishii, D., Tomizawa, H., Sutoh, C., Nakazawa, K., et al., 2012.Effects of perinatal exposure to low dose of bisphenol A on anxiety like behavior anddopamine metabolites in brain. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 39(2), 273–279.

Matthews, M., Nigg, J.T., Fair, D.A., 2014. Attention deficit hyperactivity disorder. Curr.Top. Behav. Neurosci. 16, 235–266. http://dx.doi.org/10.1007/7854_2013_249.(PubMed PMID: 24214656; PubMed Central PMCID: PMC4079001).

Michalowicz, J., 2014. Bisphenol A–sources, toxicity and biotransformation. Environ.Toxicol. Pharmacol. 37 (2), 738–758. http://dx.doi.org/10.1016/j.etap.2014.02.003.(PubMed PMID: 24632011).

Miyatake, M., Miyagawa, K., Mizuo, K., Narita, M., Suzuki, T., 2006. Dynamic changes indopaminergic neurotransmission induced by a low concentration of bisphenol-A inneurones and astrocytes. J. Neuroendocrinol. 18 (6), 434–444.

Mizuo, K., Narita, M., Yoshida, T., Narita, M., Suzuki, T., 2004. Functional changes indopamine D3 receptors by prenatal and neonatal exposure to an endocrine disruptorbisphenol-A in mice. Addict. Biol. 9 (1), 19–25.

Moffitt, T.E., Houts, R., Asherson, P., Belsky, D.W., Corcoran, D.L., Hammerle, M., et al.,2015. Is adult ADHD a childhood-onset neurodevelopmental disorder? Evidence froma Four-Decade Longitudinal Cohort Study. Am. J. Psychiatry 172 (10), 967–977.http://dx.doi.org/10.1176/appi.ajp.2015.14101266. (PubMed PMID: 25998281;PubMed Central PMCID: PMCPMC4591104).

Nagao, T., Kawachi, K., Kagawa, N., Komada, M., 2014. Neurobehavioral evaluation ofmouse newborns exposed prenatally to low-dose bisphenol A. J. Toxicol. Sci. 39 (2),231–235.

Nakamura, K., Itoh, K., Yoshimoto, K., Sugimoto, T., Fushiki, S., 2010. Prenatal andlactational exposure to low-doses of bisphenol A alters brain monoamine con-centration in adult mice. Neurosci. Lett. 484 (1), 66–70.

Nakamura, K., Itoh, K., Dai, H., Han, L., Wang, X., Kato, S., et al., 2012. Prenatal andlactational exposure to low-doses of bisphenol A alters adult mice behavior. BrainDev. 34 (1), 57–63.

National Toxicology Program, 2013. Draft Protocol for Systematic Review to Evaluate theEvidence for an Association between Bisphenol A (BPA) and Obesity, Appendix 2:Guideance for Assessing Risk of Bias in the BPA-Obesity Systematic Review. URL


355

http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0065http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0070http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0070http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0070http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0075http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0075http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0075http://dx.doi.org/10.1111/jcpp.12290http://dx.doi.org/10.1111/jcpp.12290http://dx.doi.org/10.1016/j.envres.2015.07.024http://dx.doi.org/10.1016/j.envres.2015.07.024http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0090http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0090http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0090http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0095http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0095http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0095http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0100http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0100http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0100http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0100http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0105http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0105http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0105http://dx.doi.org/10.1016/j.reprotox.2006.08.012http://dx.doi.org/10.1016/j.tox.2014.04.001http://dx.doi.org/10.1016/j.tox.2014.04.001http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0120http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0120http://distillercer.com/products/distillersr-systematic-review-software/http://distillercer.com/products/distillersr-systematic-review-software/http://distillercer.com/products/distillersr-systematic-review-software/http://distillercer.com/products/distillersr-systematic-review-software/http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0125http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0125http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0125http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0130http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0130http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0130http://dx.doi.org/10.1007/s00439-009-0663-4http://dx.doi.org/10.1007/s00439-009-0663-4http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0140http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0140http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0140http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0145http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0145http://dx.doi.org/10.1016/j.envres.2013.06.004http://dx.doi.org/10.1016/j.envres.2013.06.004http://dx.doi.org/10.1111/andr.12176http://dx.doi.org/10.1111/andr.12176http://dx.doi.org/10.1097/MOP.0000000000000191http://dx.doi.org/10.1097/MOP.0000000000000191http://dx.doi.org/10.1016/j.tox.2015.12.002http://dx.doi.org/10.1016/j.yhbeh.2016.06.010http://dx.doi.org/10.1016/j.yhbeh.2016.06.010http://dx.doi.org/10.1016/j.yhbeh.2016.06.010http://dx.doi.org/10.1136/bmj.327.7414.557http://dx.doi.org/10.1136/bmj.327.7414.557http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0180http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0180http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0180http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0185http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0185http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0185http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0190http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0190http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0190http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0195http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0195http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0195http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0200http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0200http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0200http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0205http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0205http://dx.doi.org/10.4088/PCC.17nr02153http://dx.doi.org/10.1176/ajp.2006.163.4.716http://dx.doi.org/10.1176/ajp.2006.163.4.716http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0220http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0220http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0220http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0225http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0225http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0225http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0225http://dx.doi.org/10.1111/j.1476-5381.2010.00872.xhttp://dx.doi.org/10.1111/j.1476-5381.2010.00872.xhttp://dx.doi.org/10.1016/j.tox.2014.06.009http://dx.doi.org/10.1016/j.tox.2014.06.009http://dx.doi.org/10.1073/pnas.1214056110http://dx.doi.org/10.1073/pnas.1214056110http://dx.doi.org/10.1161/STROKEAHA.108.515957http://dx.doi.org/10.1161/STROKEAHA.108.515957http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0250http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0250http://dx.doi.org/10.1542/peds.2011-3374http://dx.doi.org/10.1542/peds.2011-3374http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0260http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0260http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0260http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0265http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0265http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0265http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0270http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0270http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0270http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0270http://dx.doi.org/10.1007/7854_2013_249http://dx.doi.org/10.1007/7854_2013_249http://dx.doi.org/10.1016/j.etap.2014.02.003http://dx.doi.org/10.1016/j.etap.2014.02.003http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0285http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0285http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0285http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0290http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0290http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0290http://dx.doi.org/10.1176/appi.ajp.2015.14101266http://dx.doi.org/10.1176/appi.ajp.2015.14101266http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0300http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0300http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0300http://refhub.elsevier.com/S0160-4120(17)31196-0/rf0305http://refhub.

Prenatal exposure to bisphenol A and hyperactivity in children a … ADHD... · 2019. 12. 12. ·...

Documents

Transcript of Prenatal exposure to bisphenol A and hyperactivity in children a … ADHD... · 2019. 12. 12. ·...