SCIENTIFIC OPINION. EFSA Panel on food contact materials, enzymes, flavourings and processing aids (CEF) 2, 3

SCIENTIFIC OPINION Scientific Opinion on Bisphenol A: evaluation of a study investigating its neurodevelopmental toxicity, review of recent scientific literature on its toxicity and advice on the Danish risk assessment of Bisphenol A 1 ABSTRACT EFSA Panel on food contact materials, enzymes, flavourings and processing aids (CEF) 2, 3 European Food Safety Authority (EFSA), Parma, Italy Bisphenol A (BPA) is used in the manufacture of plastics, to produce reusable drinking bottles, infant feeding bottles and other food storage containers. EFSA was asked to evaluate a dietary developmental neurotoxicity study in rats (Stump, 2009) and recent scientific literature (2007-2010) in terms of relevance for the risk assessment of BPA. The impact of these studies on the current Tolerable Daily Intake (TDI) of 0.05 mg BPA/kg body weight (b.w.)/day as set by EFSA in 2006 was assessed. Advice on the Danish risk assessment underlying the Danish ban of BPA in food contact materials for infants aged 0-3 years is included. Overall, based on this comprehensive evaluation of recent toxicity data, the Panel on food contact materials, enzymes, flavourings and processing aids (CEF) concluded that no new study could be identified, which would call for a revision of the current TDI. This TDI is based on the No-Observed-Adverse-Effect-Level (NOAEL) of 5 mg/kg b.w./day from a multi-generation reproductive toxicity study in rats, and the application of an uncertainty factor of 100. This factor is regarded as conservative based on all information on BPA toxicokinetics. The Panel noted that some studies conducted on developing animals have suggested other BPA-related effects of possible toxicological relevance, in particular biochemical changes in brain, immune-modulatory effects and enhanced susceptibility to breast tumours. These studies had several shortcomings. At present the relevance of these findings for human health cannot be assessed. Should any new relevant data become available in the future, the Panel will reconsider this opinion. A minority opinion is expressed by a Panel member and presented in an Annex to this opinion. European Food Safety Authority, 2010 KEY WORDS Bisphenol A, BPA, CAS No. 00080-05-7, developmental toxicity, neurobehaviour, low dose effects. 1 On request from the European Commission, Questions No EFSA-Q-2009-00864, EFSA-Q-2010-01023 and EFSA-Q- 2010-00709, adopted on 23 rd September 2010. 2 Panel members: Arturo Anadón, Mona-Lise Binderup, Wilfried Bursch, Laurence Castle, Riccardo Crebelli, Karl-Heinz Engel, Roland Franz, Nathalie Gontard, Thomas Haertlé, Trine Husøy, Klaus-Dieter Jany, Catherine Leclercq, Jean-Claude Lhuguenot, Wim Mennes, Maria Rosaria Milana, Karla Pfaff, Kettil Svensson, Fidel Toldrá, Rosemary Waring, Detlef Wölfle. Part of this Opinion is not shared by a member of the Panel, Catherine Leclercq, who expressed a Minority Opinion related to parts II and IV (Overall conclusions) of this opinion. Correspondence: CEF-unit@efsa.europa.eu 3 Acknowledgement: The Panel wishes to thank the members of the Working Groups on Bisphenol A: David Bell (resigned on 15 January 2010), Wilfried Bursch, Berit Granum, Ulla Hass (her participation was limited to the evaluation of the Stump study until 26 March 2010), Edel Holene, Trine Husøy, Alberto Mantovani, Wim Mennes, Unni Cecilie Nygaard, Terry Parker, Ulrike Reuter, Emanuela Testai, Rosemary Waring, Detlef Wölfle for the preparatory work on this scientific opinion and EFSA staff: Anna F. Castoldi, Saghir Bashir, Laura Ciccolallo, Cristina Croera, Tomas Oberg, Roberta Pinalli Anne Theobald and Didier Verloo for the support provided to this scientific opinion. Suggested citation: EFSA Panel on food contact materials, enzymes, flavourings and processing aids (CEF). Scientific Opinion on Bisphenol A: evaluation of a study investigating its neurodevelopmental toxicity, review of recent scientific literature on its toxicity and advice on the Danish risk assessment of Bisphenol A. EFSA Journal 2010;8(9):1829. [110 pp.] doi:10.2903/j.efsa.2010.1829. Available online: www.efsa.europa.eu/efsajournal.htm European Food Safety Authority, 2010 1

SUMMARY Bisphenol A [2,2-bis(4-hydroxphenyl)propane, CAS Number 80-05-7] (BPA) is used as a monomer in the manufacture of polycarbonates and epoxy resins, as an antioxidant in PVC plastics and as an inhibitor of end polymerisation in PVC. Polycarbonates are used in food contact plastics such as reusable beverage bottles, infant feeding bottles, tableware (plates and mugs) and storage containers, whereas epoxy resins are used in protective linings for food and beverage cans and vats. Small amounts of BPA can potentially leach out from food containers into foodstuffs and beverages and therefore be ingested. BPA is permitted for use in food contact plastics in the European Union with a specific migration limit of 0.6 mg/kg food 4. In 2006, EFSA set the TDI for BPA at 0.05 mg BPA/kg body weight (b.w.)/day. This is based on the No-Observed-Adverse-Effect-Level (NOAEL) of 5 mg/kg b.w./day that has been identified in two multi-generation reproductive toxicity studies in rodents, where the critical effects were changes in body and organ weights in adult and offspring rats and liver effects in adult mice, respectively (EFSA, 2006). In 2008, EFSA reaffirmed this TDI, concluding that age-dependent toxicokinetics differences of BPA in animals and humans would have no implication for the default uncertainty factor (UF) of 100 and in turn for the TDI. The present opinion follows the requests of the European Commission (EC) to the Panel on food contact materials, enzymes, flavourings and processing aids (CEF) (I) to evaluate a dietary developmental neurotoxicity study of BPA in rats (Stump, 2009); (II) recent scientific literature (2007-2010) in terms of relevance for the risk assessment of BPA and impact on the current tolerable daily intake (TDI) of 0.05 mg BPA/kg body weight (b.w.)/day; (III) to provide advice on the Danish risk assessment underlying the Danish ban of BPA in food contact materials for infants aged 0-3 years. In order to provide a global view on the risk assessment of Bisphenol A, the CEF Panel decided to address the three mandates as given above in a single opinion and postponed the adoption of this comprehensive document to 23 rd September 2010. The three different questions raised by the Commission are dealt with in three different parts (PARTS I to III) of the opinion. PART IV presents an overview of the conclusions from PARTS I to III, together with an overall conclusion. PART I The GLP compliant study by Stump (2009) was performed according to OECD guideline 426 to address any uncertainty regarding potential neurodevelopmental effects of BPA. BPA was administered daily in the diet at concentrations of 0, 0.15, 1.5, 75, 750, and 2250 mg/kg feed to female Sprague-Dawley rats from gestational day (GD) 0 to postnatal day (PND) 21. The relative estimated intakes (in mg/kg b.w./day) were 0, 0.01, 0.12, 5.85, 56.4 and 164 during gestation and 0, 0.03, 0.25, 13.1, 129 and 410 during lactation. The CEF Panel considers this treatment schedule as relevant to human exposure in utero and via either breastfeeding or infant bottle feeding (in this study the estimated exposure of rat pups to BPA is ca. 30 times higher than that of bottle-fed infants). Dams were evaluated for general signs of toxicity and offspring were evaluated for general toxicity including developmental landmarks and for neurological effects, including behaviour and brain histopathology. For systemic toxicity (i.e. decreased body weight and/or body weight gain), a NOAEL of 5.85 mg/kg b.w./day during gestation was identified for both mothers and offspring. Male and female pups were tested for developmental landmarks and for neurological effects, including behaviour and brain histopathology. 4 Commission Directive 2002/72/EC of 6 August 2002 relating to plastic materials and articles intending to come into contact with foodstuffs 2

On PND 11 2 and 4 pups from the 750 and 2250 mg/kg feed groups, respectively, underwent popcorn seizures and convulsions. This effect was not reproducible in any other published study/report including a follow-up study by Stump (2009). The Panel concluded that the current NOAEL for BPA (5 mg/kg b.w./day) would be sufficiently low to exclude any concern for this effect. The study by Stump covers motor activity, learning and memory (spatial behaviour), auditory startle response, brain histopathology and morphology. The study does not cover some specific aspects of learning and memory (i.e. avoidance learning, schedule-controlled behaviour, and impulsiveness), anxiety-related behaviour or sexual dimorphic behaviour, but this does not invalidate the study. No treatment-related changes were observed in motor activity tests, auditory startle response or brain histopathology and morphology. According to the statistical analysis by the study authors, in the Biel maze swimming test for learning and memory on PND 62 the male offspring from the 0.15 mg/kg feed exposure group showed an increase in the number of errors in Path A trials 1-4, which could be interpreted as a delay in learning. The effect reached statistical significance when compared to concurrent controls. No such an effect was observed in pups studied on PND 22 or in any other exposure group on PND 62. When the swimming trials were conducted in the reverse path (i.e. the path B) no delay in learning was observed and also in the repeat Path A trials 11 and 12, no effect was observed in any exposure group, including the 0.15 mg/kg feed group. Therefore, the authors concluded that there were no changes in learning and memory. The Panel noted that in the Biel Maze test only the error counts were reported as a measure of learning and memory. The animals should have also been evaluated for effects on time-to-escape and checked for long term memory effects. EFSA s Assessment and Methodology Unit (AMU) applied a more appropriate statistical evaluation to the Biel Maze data. It was realised that the data suffer from censoring 5. Based on the re-analysis the Panel considered that no conclusion can be drawn from this study on the effect of BPA on learning and memory behaviour due to large variability in the data. Based on the body weight effects on dams and offspring and also taking into account the occurrence of seizures and convulsions in the two highest dose groups, which were not observed at the lower dose levels, the study supports the NOAEL which was derived from multigeneration studies in the past (5 mg/kg b.w./day), leading to a TDI of 0.05 mg/kg b.w./day. However, the Panel considered that this test on learning and memory was inconclusive and is of limited value in the risk assessment of BPA. Overall, based on the body weight effects on dams and offspring and taking into account the occurrence of seizures and convulsions only at the two highest doses, the study supports the previously identified NOAEL of 5 mg/kg b.w./day. However, the Panel considered that this study was inconclusive and is of limited value in the ongoing risk assessment of BPA. PART II The CEF Panel has reviewed toxicological data published between 2007 and July 2010 mainly focusing on toxicokinetic, human and animal toxicity studies. For risk assessment purposes only studies complying with these inclusion criteria were considered: full research papers in peer-reviewed journals available in the public domains and reporting original data; all human studies (except for purely biomonitoring studies). For the in vivo animal toxicity studies the focus was on low dose oral studies employing several test doses including at least one <5 mg/kg b.w./day and involving developmental exposure. 5 In statistical terminology censoring occurs when the value of an observation is only partially known. In the Biel water maze test the number of errors made by the rats that did not complete the maze within the 3 minute-limit that was used, was recorded and the time to escape was recorded as 180 seconds. These experiments are right censored, meaning that the rats could have escaped and made more errors, had they been given sufficient time. Therefore the time to escape and the number of errors recorded will have been underestimated. 3

These studies were further assessed with respect to quality criteria (sufficient sample size, adequacy of control procedures, inclusion of positive controls when applicable, assessment of correlation between morphological and functional changes, and consideration of litter or dam as the appropriate statistical unit) in order to assess the validity and/or applicability of the individual findings to human risk assessment. Studies on toxicokinetics of BPA have demonstrated a significantly lower internal exposure after oral intake as compared to parenteral exposure. This confirms that toxicity studies with oral administration have higher relevance for human risk assessment of BPA in food than studies with parenteral administration. In addition, new findings in non human primates (both adults and newborns) further strengthen the view that BPA is eliminated faster in humans than in rodents. This fast BPA elimination in primates results in substantially lower internal exposure to free BPA in humans as compared to rodents. Even human premature infants can metabolise and excrete BPA efficiently (via glucuronidation and sulfation), as supported by recent human data and data in young monkeys. The use of the standard uncertainty factor (UF) of 10 to take into account interspecies differences is therefore considered quite conservative. In relation to in utero exposure, studies on transplacental transport of BPA and BPA-glucuronide in humans and rodents indicated that although transfer may occur, foetal free BPA levels are highly limited by the efflux pump P-glycoprotein, and placental BPA glucuronidation may also take place. The Panel noted that exposure to total BPA (the major constituent being the glucuronide) through lactation is limited to a very low fraction. Therefore, in utero exposure and exposure through lactation appear to be limited. The Panel recognises that inter-individual differences occur in expression of the isoenzymes responsible for the detoxification of BPA. However, even in persons with low expression of these enzymes, the metabolising capacity is still sufficient to eliminate free BPA from blood at the low levels of BPA resulting from consumer exposure. Dietary exposures of adults and infants aged 3-6 months were estimated to be up to 1.5 and 13 µg/kg body weight per day, respectively (EFSA, 2006), based on conservative estimates of food consumption and migration from food contact materials. These exposures are not anticipated to surpass the metabolic capacity for BPA in adults or infants. Recent epidemiological studies have suggested some statistically significant associations of BPA exposure (urine concentrations) and health effects (coronary heart disease, reproductive disorders) in adults and behavioural changes in young girls. The Panel noted that cross sectional epidemiological studies such as these can demonstrate statistical associations between BPA exposure and the presence (e.g. coronary heart disease) or absence (e.g. cancer, asthma) of health outcomes, but the inherent design of cross sectional studies does not allow establishment of a causal relationship between BPA exposure and health effects (e.g. chronic diseases). In addition, the Panel has identified some limitations in these studies, which raise further questions as to the significance of the reported findings. Therefore, the Panel could not draw any relevant conclusion for risk assessment from these studies. The animal studies on developmental and reproductive toxicology reporting effects at doses lower than 5 mg/kg b.w./day have severe shortcomings and were considered to be invalid. The Panel considers that the valid studies do not raise concern regarding reproductive and developmental toxicity of BPA at doses lower than 5 mg/kg b.w./day. Potentially significant biochemical changes, e.g. altered receptor expression in different brain regions (see section 5.3), have been reported. However, in the absence of a correlation with a functional adverse effect, the relevance of these observations for human health cannot be assessed. The impact of BPA on development of sexually dimorphic behaviour was addressed in the study by Ryan et al. (2010a), who observed a male-like reduced saccharin preference and inhibition of lordosis behaviour in female rat offspring from oestrogen-treated but not from BPA-treated dams. In the study reported by Stump et al. (2009) (See Part I) the effects of BPA on learning and memory behaviour were 4

inconclusive due to large variability in the data. Other recent studies have methodological shortcomings. The Panel does not consider the currently available data as convincing evidence of neurobehavioural toxicity of BPA. The study by Jenkins et al. (2009) is the first oral study on a possible BPA-induced enhancement of sensitivity of the mammary gland to carcinogen-induced breast tumour formation in rat offspring following lactational BPA exposure of pups. Using the same model of dimethylbenzanthracene (DMBA)-induced mammary carcinogenesis but in utero BPA exposure, Betancourt et al. (2010b) also reported an enhancement of susceptibility for mammary gland carcinogenesis. In consideration of the shortcomings in the design of both studies, in particular the uncertainty regarding the lactational as well as in utero exposure of the offspring to BPA, and of the limitations in reporting the Panel concluded that these results cannot be taken into consideration for derivation of a TDI. However, the Panel noted that at the highest dose level studied there is a shift of the ratio between cell proliferation and apoptosis in favour of cell multiplication in the mammary gland. In view of the mechanistic data obtained upon in utero exposure in other studies (see section 5.3) and the implications of an increased cell proliferation/apoptosis ratio in carcinogenesis, the effects reported by Jenkins and Betancourt deserve further consideration. Modulation of immune system-related parameters is also an emerging field also in BPA research. Several studies have reported changes in cytokines, changes in T-cell populations and other aspects of immune modulation. However, the studies were all suffering from shortcomings in experimental design and reporting. Therefore, at the moment, these studies cannot be taken into consideration for derivation of a TDI. In vitro- and in vivo-studies (not compliant with the selection criteria in section 3) on receptors, hormones, immune system, cell proliferation, apoptosis, proteomic, genomic and epigenetic changes have been presented to compile recent data on potentially relevant endocrine mechanisms of action of BPA. High doses of BPA (>5 mg/kg b.w./day) may have biochemical and molecular effects consistent with those observed with other oestrogenic substances. Effects have been claimed to occur at low levels of BPA exposure, which could be independent of the classical hormone receptors. BPA has only weak binding affinities to these receptors, but these effects may alternatively be induced by cell membrane-triggered signalling pathways via protein kinases. However, in the absence of clear dose response curves and due to the shortcomings in experimental design, a conclusion cannot be reached on the implications of the observed biochemical and molecular changes or establish whether they have any impact on human health. Because of the lack of a common clearly defined mode of action of BPA at low doses, the toxicological relevance of the BPA effects described cannot be evaluated and the results cannot be taken into consideration for derivation of a TDI. While low dose effects of BPA are reported for some biochemical changes the Panel is not aware of any clearly reproducible adverse effect expressed specifically at low BPA doses only. EFSA has established an internal task force to initiate the development of a common strategy towards endocrine active substances. The Panel is aware of EFSA s ongoing work to monitor trends and developments in the assessment of health risks of endocrine active substances. PART III This part deals with the EFSA advice on the Danish risk assessment of BPA. The conclusion of the DTU Food Institute is based upon three major lines of arguments: (i) a degree of uncertainty with regard to the effects on learning ability, since in the study by Stump et al (2009) impaired learning ability was found in male offspring with low dosage of BPA; (ii) doubts on the monotonic ( normal ) dose-response for BPA; (iii) some endpoints which have not been considered, namely certain aspects of learning and memory (avoidance learning, schedule-controlled behaviour, impulsiveness), anxietyrelated behaviour and gender-specific (i.e., sexually dimorphic) behaviour. 5

With respect to the learning and memory endpoint of the Stump study, as examined in the Biel Maze test, the Panel concluded that the influence of BPA on learning and memory behaviour cannot be evaluated. Regarding this endpoint, the study is inconclusive and cannot be used for the risk assessment of BPA. It has been argued that BPA may show a non-monotonic dose-response curve. Low dose effects of BPA have been reported, which might be independent of the effect on the classical hormone receptors (See Part II). However, most of these studies have several shortcomings such as lack of dose responses and limitations in experimental design. The Panel is not aware of any clearly reproducible adverse effect expressed specifically at low BPA doses only. Altogether, the Panel concluded that the study by Stump et al. (2009) cannot be used for the risk assessment of BPA, because of large variability in the data. Therefore, the study is inconclusive and cannot impact on the risk assessment of BPA. Methodological shortcomings also apply to a number of studies addressing other neurobehavioural endpoints (e.g. learning and memory behaviour, anxietyrelated behaviour and gender-specific behaviour), which were considered invalid or inadequate for risk assessment purposes. The Panel does not consider the currently available data sufficiently indicative of neurobehavioural toxicity as an endpoint of concern for BPA. Overall, based on the comprehensive evaluation of recent human and animal toxicity data, the Panel concluded that no new study could be identified, which would call for a revision of the current TDI of 0.05 mg/kg b.w./day. This TDI is based on the NOAEL of 5 mg/kg b.w./day from a multi-generation reproductive toxicity study in rats, and the application of an uncertainty factor of 100, which is regarded as conservative based on all information on BPA toxicokinetics. The Panel noted that some studies conducted on developing animals have suggested other BPA-related effects of possible toxicological relevance, in particular biochemical changes in brain, immunemodulatory effects and enhanced susceptibility to breast tumours. These studies had many shortcomings. At present the relevance of these findings for human health cannot be assessed, though should any new relevant data become available in the future, the Panel will reconsider the current opinion. A minority opinion expressed by a Panel member is presented in an Annex to this opinion. 6

TABLE OF CONTENTS Abstract... 1 Summary... 2 Table of contents... 7 Background as provided by the European Commission... 9 Terms of reference as provided by the European Commission... 9 Interpretation of the terms of reference by the EFSA... 10 Part I Evaluation of the dietary developmental neurotoxicity study of Bisphenol A in rats by Stump (2009)... 11 Part I - Assessment... 11 1. Introduction... 11 1.1. Previous risk assessments... 12 1.2. Discussion of five developmental neurotoxicity studies, highlighted in previous evaluations. 15 2. Assessment of a newly available developmental neurotoxicity study in rats... 17 2.1. Summary of the study as reported by Stump (2009)... 17 2.2. The Panel s comments to the study design and results... 19 2.2.1. F0 data (Maternal generation)... 19 2.2.1.1. Detailed Clinical Observations... 20 2.2.2. F1 data (Litter)... 20 2.2.2.1. Postnatal survival, growth and developmental landmarks... 20 2.2.2.2. Detailed Clinical Observations (Subset A)... 20 2.2.2.3. Motor Activity (Subset A)... 21 2.2.2.4. Auditory Startle Test (Subset A)... 21 2.2.2.5. Biel Maze Swimming Trials (Subsets A and B)... 21 2.2.2.6. Macroscopic Examinations... 25 2.2.2.7. Neuropathology (Subsets A and C)... 25 2.3. Discussion of the study outcome... 26 Part I - Conclusions... 27 Part I - Documentation provided to EFSA... 29 Part I - References... 30 Part I - Abbreviations... 33 Part I - Appendices... 34 Part II - Review of recent scientific literature on the toxicity of Bisphenol A... 38 Part II - Assessment... 38 3. Introduction... 38 4. Toxicokinetics... 40 4.1. New pharmacokinetic studies in rats and monkeys (Doerge et al. 2010a, 2010b)... 41 4.2. The enzymes involved in BPA biotransformation... 43 4.3. In utero exposure and kinetics... 44 4.4. Neonatal exposure and kinetics... 47 4.5. Human physiologically based pharmacokinetic (PBPK) modelling... 49 4.6. BPA repeated exposure... 51 4.7. Summary and conclusions on toxicokinetics... 52 5. Toxicity... 52 5.1. Human studies... 52 5.1.1. Braun et al. (2009)... 52 5.1.2. Melzer et al. (2010)... 54 5.1.3. Other human studies... 55 5.1.4. General comments on human studies... 59 5.1.4.1. Analytical aspects... 59 5.1.4.2. Comment on cross-sectional studies... 60 5.1.5. Conclusions on human studies... 61 5.2. Animal toxicity studies... 61 5.2.1. Reproductive and developmental toxicity... 62 7

5.2.1.1. Developmental studies by Howdeshell at al. (2008) and Ryan et al. (2010a)... 63 5.2.1.2. Other developmental toxicity studies... 64 5.2.2. Conclusions on animal toxicity studies... 70 5.2.2.1. Studies on developmental and reproductive toxicity... 70 5.2.2.2. Developmental neurotoxicity and neurodevelopmental studies... 70 5.2.2.3. Cell proliferation and apoptosis related to enhancement of tumourigenesis... 70 5.2.2.4. Studies on immunotoxicity... 70 5.3. Other information on the endocrine-mediated action of BPA... 70 5.3.1. Effects on receptors and hormone homeostasis... 71 5.3.2. Effects on signal transfer and gene expression... 73 5.3.3. Effects on the immune system... 74 5.3.4. Cytogenetic and epigenetic effects, cell proliferation and apoptosis... 74 5.3.5. Conclusions on the endocrine-mediated action of BPA... 76 Part II - Conclusions... 76 5.4. Conclusions on toxicokinetics... 77 5.5. Conclusions on human studies... 77 5.6. Conclusions on studies in animals... 78 5.6.1. Studies on developmental and reproductive toxicity... 78 5.6.2. Developmental neurotoxicity and neurodevelopmental studies... 78 5.6.3. Cell proliferation and apoptosis related to enhancement of tumourigenesis... 78 5.6.4. Studies on immunotoxicity... 78 5.7. Conclusions on endocrine-mediated action of BPA... 79 Part II - References... 80 Part II - Abbreviations... 94 Part III - Advice on the Danish risk assessment of Bisphenol A... 96 Part III - Assessment... 96 6. Introduction... 96 7. Conclusions of the Danish risk assessment... 96 8. Discussion of the Panel on the Danish risk assessment... 97 Part III - Conclusions... 98 Part III - References... 99 Part III - Abbreviations... 100 Part IV - Overall Panel conclusions... 101 9. Background... 101 10. Conclusions from PART I: Evaluation of the dietary developmental neurotoxicity study of Bisphenol A in rats... 102 11. Conclusions from PART II: Review of recent scientific literature on the toxicity of Bisphenol A... 103 11.1. Conclusions on toxicokinetics... 104 11.2. Conclusions on human studies... 104 11.3. Conclusions on studies in animals... 105 11.3.1. Studies on developmental and reproductive toxicity... 105 11.3.2. Developmental neurotoxicity and neurodevelopmental studies... 105 11.3.3. Cell proliferation and apoptosis related to enhancement of tumourigenesis... 105 11.3.4. Studies on immunotoxicity... 105 11.4. Conclusions on endocrine-mediated action of BPA... 105 12. Conclusions from PART III: Advice on the Danish risk assessment of Bisphenol A... 106 13. Overall conclusion... 107 Part IV - References... 108 Part IV - Abbreviations... 110 8

BACKGROUND AS PROVIDED BY THE EUROPEAN COMMISSION The substance 2,2-bis(4-hydroxyphenyl)propane, CAS Number 80-05-7, more commonly known as bisphenol A (BPA), is used as a monomer in the manufacture of polycarbonates and epoxy resins, as an antioxidant in PVC plastics and as an inhibitor of end polymerisation in PVC. Polycarbonates are used in food contact materials such as reusable beverage bottles, infant feeding bottles, tableware (plates and mugs) and storage containers. Epoxy resins are used in protective linings for food and beverage cans and vats. BPA was first evaluated in 1984 by the Scientific Committee on Food (SCF, 1986) for use in plastic materials and articles intended to come into contact with foodstuffs. At that time, the Committee allocated a Tolerable Daily Intake (TDI) of 0.05 mg/kg b.w.. This was based on 90-day and long-term oral studies in rats and mice, in which BPA was given via the diet. These showed an overall noobserved-adverse-effect level (NOAEL) of 25 mg/kg b.w./day for effects on body weight, taken from the 90-day rat study, to which an uncertainty factor of 500 was applied because of the incomplete database. The substance was re-evaluated in 2002 by the SCF in the light of new information published up to 2001. The Committee allocated a temporary TDI (t-tdi) of 0.01 mg/kg b.w.. This was based on 90- day and comprehensive three-generation studies in the rat, in which BPA was given via the diet. These showed an overall no-observed-adverse-effect level (NOAEL) of 5 mg/kg b.w./day for effects on body weight, taken from the comprehensive three-generation study in the rat, to which an uncertainty factor of 500 was applied because of the incomplete database. The SCF recommended that the t-tdi be reviewed once significant new data were available. The substance was re-evaluated in 2006 by EFSA in light of the new information published up to 2005. EFSA has, in its opinion of 26 November 2006, established for BPA (BPA) a tolerable daily intake (TDI) of 0.05 milligram per kilogram (mg/kg) body weight based on the no adverse effect level of 5 mg/kg body weight in rodent studies. A new opinion on the toxicokinetics of Bisphenol was adopted by EFSA on 9 July 2008. In this opinion, the AFC Panel considered that its previous risk assessment based on the overall NOAEL for effects in rats and using a default uncertainty factor of 100 can be regarded as conservative for humans. The Panel concluded that the differences in age-dependent toxicokinetics of BPA in animals and humans would have no implication for the assessment of BPA carried out by EFSA in 2006. Health Canada in its risk assessment issued on 18 April 2008 concluded on neurobehavioural toxicity that even though the weight of evidence supporting neurobehavioural effects in rodents at exposure below the established NOAEL for reproductive/developmental toxicity is limited, the outcome of the existing studies warrant the application of precaution for sensitive life stages such as pregnant women/foetus and infants. Based on this evaluation the government of Canada proposed legislation banning the use of polycarbonate in baby feeding bottles. In 2008 the Polycarbonate/BPA Global Group of the American Chemistry Council commissioned a neurobehavioural study to address the concerns raised by the Canadian Government. The results of this study have now been transmitted to the European Commission by Plastics Europe in the context of Article 11(5) of Regulation (EC) No 1935/2004. TERMS OF REFERENCE AS PROVIDED BY THE EUROPEAN COMMISSION In October 2009 the Commission asked the European Food Safety Authority (EFSA) to: - Assess the relevance of the dietary developmental neurotoxicity study of Bisphenol A in rats and its implication for hazard and risk assessment of Bisphenol A. 9

- Update, if necessary, the currently applicable tolerable daily intake for Bisphenol A by May 2010. In relation to the above request, the European Commission clarified in a letter dated 3 February 2010 that EFSA should address any new scientific evidence that may affect the conclusions of the previously adopted opinions on BPA by May 2010. On 30 March 2010, EFSA received an additional request from the European Commission to advise by 15 April 2010 on the Danish risk assessment underlying the Danish ban, on 26 March 2010, of Bisphenol A in food contact materials for infants aged 0-3 years. INTERPRETATION OF THE TERMS OF REFERENCE BY THE EFSA In order to provide a global view on the risk assessment of Bisphenol A, the CEF Panel decided to address the three mandates received from the European Commission in a single opinion and postponed the adoption of this comprehensive document to September 2010. The present opinion consists of four distinct parts: PART I: Evaluation of the dietary developmental neurotoxicity study of Bisphenol A in rats PART II: Review of recent scientific literature on the toxicity of Bisphenol A PART III: Advice on the Danish risk assessment underlying the Danish ban of Bisphenol A in food contact materials for infants aged 0-3 years. PART IV: Overall conclusions 10

PART I EVALUATION OF THE DIETARY DEVELOPMENTAL NEUROTOXICITY STUDY OF BISPHENOL A IN RATS BY STUMP (2009) PART I - ASSESSMENT 1. Introduction Bisphenol A (BPA; 2,2-bis(4-hydroxyphenyl)propane, CAS Number 80-05-7) is primarily used as a monomer in the manufacture of polycarbonate (PC) and epoxy resins. Regarding food contact applications, polycarbonate is used to make food containers such as reusable beverage bottles, infant feeding bottles, tableware (plates and mugs), microwave ovenware and storage containers including reservoirs for water dispensers. Epoxy resins are used to make protective coatings and linings for food and beverage cans and vats. BPA is permitted for use in food contact materials in the European Union (EU) with a specific migration limit of 0.6 mg/kg food, under Commission Directive 2002/72/EC of 6 August 2002 relating to plastic materials and articles intended to come into contact with foodstuffs. It is also permitted for food contact use in other countries such as the USA and Japan. Another field of potential use of BPA is in water distribution networks for human consumption. BPA-based epoxyphenolic resins are widely used as a surface-coating on residential drinking water storage tanks (Bae et al., 2002). For use of BPA in materials in contact with drinking water no legislation at the EU level is available. Individual Member States may have their own national legislation in force. BPA is also used in a variety of non-food applications: epoxy resin based paints, PVC medical devices, wood fillers, adhesives, surface coatings, printing inks, carbonless and thermal paper, flame retardants, tyre and brake fluid manufacture, resin-based composites and sealants used in dentistry (EFSA, 2006). PC finds wide application in construction materials, light fittings and other non food consumer products. As a result of food contact uses, consumers may be exposed to BPA via their diet. Small amounts of BPA can potentially leach out from food containers into foodstuffs and beverages and can therefore be ingested (EFSA, 2006; EC, 2003; 2008). Based on calculations with a EU environmental exposure and risk assessment model (EUSES), for the general population the main route of human exposure to BPA is the oral route (EC, 2003; 2008). There is a concern that BPA has potential endocrine-disrupting properties, which may adversely impact physical, neurological and behavioural development. BPA is considered as a weak oestrogenlike agent (NTP-CERHR, 2008), and a multitude of biological effects has been postulated for BPA (see also Part II of this Opinion). Toxicokinetic data indicate that BPA is rapidly absorbed from the gastrointestinal tract, and that formation of BPA-glucuronide is the major pathway of BPA biotransformation in primates and in rats. Conjugation of BPA is considered a deactivation reaction since BPA-conjugates are devoid of endocrine activity. Major differences in disposition of BPAglucuronide have been reported in primates, including humans, and rodents due to different pathways of elimination from the liver. In humans, 80% of the ingested BPA is excreted in the urine within 5 hours. In rodents, enterohepatic circulation of BPA takes place and this causes a slower clearance of BPA from the organism than in humans (the half-life in rodents is 19-78 hrs; EFSA, 2006). The toxicology of BPA has been extensively investigated by industry, government and academic research groups in short and long term animal tests, including several reproductive toxicity studies, multi-generation exposure studies and a cancer bioassay. The acute toxicity of BPA is low through all exposure routes relevant to humans (EC, 2003). Repeated oral toxicity studies have indicated the intestine, liver, kidney and reproductive system as the possible target organs/systems for toxicity in laboratory animals at doses ranging from 25 to 500 mg/kg b.w./day (EFSA, 2006; NTP-CERHR, 2007). BPA was shown not to be genotoxic in bacteria and in mammalian cells, or tumourigenic in mice or rats (NTP, 1982) exposed for life to high doses of BPA by dietary administration (EC, 2003; EFSA, 2006). 11

There is evidence that prenatal and/or lactational exposure to high BPA doses (i.e., those causing maternal toxicity) can adversely affect the development of laboratory animals, reducing survival, birth weight, and growth of offspring early in life, and modify the onset of puberty in males and females (EFSA, 2006; NTP-CERHR, 2008). Other studies in rodents reported that also low BPA doses (i.e., those that are more relevant to human environmental exposure) administered during development can produce a number of effects including neural and behavioural alterations in rats and mice. However, the validity of these studies as well as the implications of these low dose behavioural for human health is heavily debated (EFSA, 2006; NTP-CERHR, 2008). 1.1. Previous risk assessments In recent years, several international agencies and bodies have comprehensively reviewed the available toxicological studies on BPA (AIST, 2005; SCF; 2002; EC, 2003; 2008; EFSA, 2006; 2008; Health Canada, 2008; NTP-CERHR, 2007; 2008, AFSSA, 2010a, U.S. FDA, 2010a).The current opinion was partly presented as a draft at a meeting in Parma with experts from Member States on 26 March 2010. It was also communicated to the U.S. Federal Drug Administration (FDA), the Food Standards Australia New Zealand (FSANZ), the New Zealand Food Safety Authority (NZFSA), the World Health Organisation (WHO) and Health Canada. In 1984, the Scientific Committee on Food (SCF, 1986) evaluated BPA for use in plastic materials and articles intended to come into contact with foodstuffs. The Committee set a Tolerable Daily Intake (TDI) of 0.05 mg/kg b.w.. This was based on 90-day and long-term dietary studies in rats and mice 6. An overall no-observed-adverse-effect level (NOAEL) of 25 mg/kg b.w./day was identified for effects on body weight (90-day rat study), to which an uncertainty factor (UF) of 500 was applied because of the incomplete database. In 2002, BPA was re-evaluated by the SCF. The SCF considered the overall oral NOAEL for BPA to be 5 mg/kg b.w./day, based on the results of a comprehensive three-generation study in the rat, showing significant reductions in adult body weight and pup body and organ weights (Tyl et al., 2002). A temporary TDI of 0.01 mg BPA/kg b.w./day was derived by applying an UF of 500 (100 for inter- and intra-species differences, and 5 for uncertainties in the database) to the NOAEL of 5 mg/kg b.w./day. In 2003, the European Union published a comprehensive Risk Assessment Report (EU-RAR) for BPA from all sources of exposure in the context of Council Regulation (EEC) No. 793/93 on the evaluation and control of existing substances. The key health effects of BPA through different exposure routes were considered to be eye and respiratory tract irritation, skin sensitisation, repeated dose toxicity to the respiratory tract, effects on the liver and reproductive toxicity (effects on fertility and on development). With respect to human health risks, a need for further research was identified, to resolve the uncertainties surrounding the potential for BPA to produce adverse effects on neurological and neurobehavioural development at low doses (EC, 2003). Several studies addressing these endpoints are discussed in section 1.2. In 2005, the Japanese National Institute of Advanced Industrial Science and Technology (AIST) published a comprehensive human health risk assessment report based on a thorough review of the toxicological profile of BPA combined with estimates of human exposure. The key toxicological endpoints for human health risks posed by BPA were reduction in body weight gain, effects on the liver and reproductive toxicity. For all three endpoints the NOAEL or the calculated Benchmark Dose Lower Limit (BMDL) were in the range of 5-50 mg/kg b.w./day. Comparison of the most realistic exposure estimates with the NOAELs and BMDL led to Margins of Exposure (MOEs) in the range of 85,000-1,800,000. The MOEs were also sufficiently large (>1,000) for adults and children when more 6 SCF (1986) refers to CIVO (currently TNO-voeding, Zeist, The Netherlands) report number R 6229, november, 1979. This report was not available to the CEF Panel. 12

conservative, worst-case exposure estimates were used for the comparison. AIST concluded that the current BPA exposure levels were unlikely to pose unacceptable risks to human health. In 2006, on a request from the European Commission, EFSA re-assessed BPA for use in food contact materials, focusing on its effects on reproduction and the endocrine system. The Scientific Panel on Food Additives, Flavourings, Processing Aids and Materials in Contact with Food (AFC) confirmed the validity of the overall NOAEL of 5 mg BPA/kg b.w./day from a comprehensive three generation study in the rat (Tyl et al., 2002), taken as the key study by the SCF in 2002. This NOAEL was further supported by the results of a newer two-generation reproductive toxicity study in mice, which also provided a NOAEL of 5 mg/kg b.w./day for liver effects (Tyl et al., 2007; same study as Tyl et al., 2006, cited in EFSA, 2006). Because the database concerning reproduction and development had been considerably strengthened since the 2002 SCF report, the Panel considered that the additional UF of 5 was no longer needed. Thus, the AFC Panel set a full TDI for BPA at 0.05 mg BPA/kg b.w., by applying a default UF of 100 to the overall NOAEL of 5 mg/kg b.w./day. This revised UF was considered to be conservative in view of the well-described species differences in toxicokinetics, showing a low systemic concentration of free BPA in humans compared with rats. The Panel found that intakes of BPA through food and drink were well below the TDI, even for infants and children (EFSA, 2006). In April 2008, the EU-RAR (EC, 2008) was updated after submission and evaluation of the twogeneration reproductive study in mice by Tyl et al. (2007) along with the new data on human exposure and effects of BPA that had become available since the original risk assessment report was completed (EC, 2003). The Rapporteur (UK) came to the conclusion that there was no need for further information and/or testing and for risk reduction measures beyond those which were already being applied. However, Denmark, Sweden and Norway considered that the results of four neurodevelopmental studies (Adriani et al., 2003; Carr et al., 2003; Negishi et al., 2004; Ryan and Vandenbergh, 2006) warranted further consideration (EC, 2008). These studies will be addressed in more detail in part I, section 1.2. In June 2008, the Norwegian Scientific Committee for Food Safety (VKM) issued an opinion regarding the assessment of the same four studies (Adriani et al., 2003; Carr et al., 2003; Negishi et al., 2004; Ryan and Vandenbergh, 2006) as highlighted in the EU-RAR (EC, 2008), on developmental neurotoxicity of BPA following low dose exposure. The VKM concluded that these studies did not provide sufficient evidence for setting a robust lower NOAEL for BPA than that previously reported by EFSA, i.e. 5 mg/kg b.w./day. In order to eliminate any uncertainty regarding potential developmental effects of BPA at low doses, the VKM recommended that a GLP compliant study should be carried out according to OECD guideline 426, using a broad concentration range from the very low doses up to those with known maternal effects. The new study evaluated in Part I of the current opinion is in compliance with this recommendation. In its opinion of July 2008, EFSA further assessed the possible age-dependent toxicokinetics of BPA in animals and humans and their implications for hazard and risk assessment of BPA in food, taking into account the most recent information and data available. The AFC Panel considered the exposure of a human foetus to free BPA as negligible, due to the increased maternal capacity for conjugation of BPA with glucuronide during pregnancy, whereas the foetal rat would be exposed to free BPA from the maternal circulation (due to decreased glucuronidation capacity). The AFC Panel also concluded that human neonates would have sufficient capacity to biotransform BPA to hormonally inactive conjugates at the BPA exposure levels reported in the EFSA opinion of 2006 and the EU-RAR (EC, 2003; 2008). Because of these metabolic differences, exposure to free BPA, and therefore toxicity, would be greater in rats of any age than in humans on an equivalent dose basis. The Panel concluded that the differences in age-dependent toxicokinetics of BPA in animals and humans would have no implication for its earlier risk assessment and reaffirmed the TDI value of 0.05 mg/kg b.w./day (EFSA, 2008) and the conservative nature of the uncertainty factor used (100). In August 2008, the Health Canada s Food Directorate reported in its assessment of BPA from food packaging applications that, based on the overall weight of evidence, the current dietary exposure to 13

BPA through food packaging uses was not expected to pose a health risk to the general population, including newborns and young children (Health Canada, 2008). Health Canada did not revise the provisional tolerable daily intake (ptdi) for BPA of 0.025 mg/kg b.w./day which had been set by them in 1996 based on the lowest NOEL for general toxicity parameters of 25 mg/kg b.w./day observed in a 90-day study in rats (NTP, 1982). However, Health Canada recommended that the general principle of ALARA (as low as reasonably achievable) should be applied to limit BPA exposure from food packaging applications of newborns and infants. This advice was based on the results of a number of neurodevelopmental and behavioural studies in experimental animals, which suggest a heightened sensitivity during stages of development in rodents. At the same time, these studies presented interpretational uncertainties related to methodological concerns, raising questions as to the actual significance of these findings in the assessment of the potential risk of BPA to human health. Health Canada concluded that although highly uncertain, these data sets suggest the need for more focused attention on products consumed by newborns and infants. In 2009, the Health Canada s Food Directorate issued a statement (Health Canada, 2009) endorsing the same conclusions on the safety of BPA first drawn by them in 2008. Nevertheless, due to uncertainties in the neurodevelopmental database the Government of Canada reiterated the need to limit BPA exposure of infants and young children from food packaging applications, i.e. from pre-packaged infant formula products. In September 2008, the U.S. NTP-CERHR (National Toxicology Program - Center for the Evaluation of Risks to Human Reproduction) published a monograph on the potential human reproductive and developmental effects of BPA. This included the NTP Brief and the Expert Panel Report that had been published the year before (NTP-CERHR, 2007). The NTP expressed some concern ( Some concern is the midpoint on a five-level scale, ranging from negligible to serious ) for effects on the brain, behaviour, and prostate gland in foetuses, infants, and children at current human exposures to BPA. The NTP had minimal concern for effects on the mammary gland and for an earlier age for puberty for females, in foetuses, infants, and children at current human exposure levels to BPA. NTP expressed negligible concern that exposure of pregnant women to BPA will result in fetal or neonatal mortality, birth defects, or reduced birth weight and growth in their offspring. NTP also expressed negligible concern that exposure to BPA will cause reproductive effects in non-occupationally exposed adults and minimal concern for workers exposed to higher levels in occupational settings. On 15 October 2009, EFSA received a request from the European Commission to assess the relevance of a new study on possible neurodevelopmental effects of BPA (Stump, 2009) and, if necessary, to update the existing TDI accordingly. The study in question was commissioned by the American Chemistry Council to address concerns raised by the Canadian government. The study was conducted according to OECD guideline 426 as recommended by the VKM in 2008. This study is discussed in depth in Part I, section 2. After the release of the previous EFSA opinions on BPA by the former Panel on additives, flavourings, processing aids and materials in contact with food (AFC Panel) and the Panel on food contact materials, enzymes, flavourings and processing aids (CEF Panel), several hundreds of new papers on BPA were published, which required consideration. An update review on these new studies is presented in Part II of this Opinion. During the preparation of this opinion, the U.S. FDA (2010b), the Danish DTU Food (2010), the French Food Safety Agency (AFSSA, 2010a, b) and the German Institute for Risk Assessment (BfR, 2010) published updates on BPA, including comments on the study by Stump (2009). The Danish risk assessment is discussed in Part III of this opinion, as requested in the terms of reference of this opinion. In January 2010, the U.S. FDA (2010b) stated that standard toxicity studies had thus far supported the safety of low level exposure of humans to BPA. However, FDA had some concern for human development in terms of potential neurobehavioural and reproductive effects that had been reported in some novel approach-based studies testing for subtle effects. At the same time, FDA emphasised 14