Environment International 150 (2021) 106433

Contents lists available at ScienceDirect

Environment International

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

Prenatal exposure to bisphenols and cognitive function in children at 7 years of age in the Swedish SELMA study

Carl-Gustaf Bornehag a,b,*, Elin Engdahl c, Maria Unenge Hallerback¨ a, Sverre Wikstrom¨ d, Christian Lindh e, Jo¨elle Rüegg c,a, Eva Tanner b, Chris Gennings b a Karlstad University, Karlstad, Sweden b Icahn School of at Mount Sinai, New York City, USA c Uppsala University, Uppsala, Sweden d ¨ ¨ Orebro University, Orebro, Sweden e Lund University, Lund, Sweden

ARTICLE INFO ABSTRACT

Handling editor: Martí Nadal Background: Experimental evidence demonstrates that exposure to bisphenol A (BPA), and the recently intro­ duced alternatives bisphenol S (BPS) and bisphenol F (BPF) alter normal neurodevelopment. More research is Keywords: needed to evaluate the associations between exposure to individual BPA alternatives and neurodevelopmental Bisphenols outcomes in humans. BPA Objective: The present study aimed at examining the individual associations between prenatal BPA, BPS and BPF BPS exposure and cognitive outcomes in children at age 7 years. BPF Cognitive function Method: Women were enrolled in the Swedish Environmental Longitudinal Mother and Child, Asthma and Al­ Prenatal exposure lergy (SELMA) study, at gestational median week 10.0, and their children were examined for cognitive function at 7 years of age (N = 803). Maternal urinary BPA, BPS, and BPF concentrations were measured at enrollment and childreńs cognitive function at the age of 7 years was measured using the Wechsler Intelligence Scale for Children IV (WISC-IV). Results: All three bisphenols were detected in over 90% of the women, where BPA had the highest geometric mean concentrations (1.55 ng/mL), followed by BPF (0.16 ng/mL) and BPS (0.07 ng/mL). Prenatal BPF exposure was associated with decreased full scale IQ (β = 1.96, 95%CI; 3.12; 0.80), as well as with a decrease in all four sub scales covering verbal comprehension, perceptual reasoning, working memory and processing speed. This association corresponded to a 1.6-point lower IQ score for an inter-quartile-range (IQR) change in prenatal BPF exposure (IQR = 0.054–0.350 ng/mL). In sex-stratified analyses, significant associations with full scale IQ were found for boys (β = 2.86, 95%CI; 4.54; 1.18), while the associations for girls did not reach significance (β = 1.38, 95%CI; 2.97; 0.22). No significant associations between BPA nor BPS and cognition were found. Discussion: Prenatal exposure to BPF was significantly associated with childreńs cognitive function at 7 years. Since BPF is replacing BPA in numerous consumer products globally, this findingurgently call for further studies.

1. Introduction pregnant women globally (Woodruff et al., 2011; Braun et al., 2011). BPA is a recognized endocrine disrupting chemical (EDC), since it can Bisphenol A (BPA) was first synthesized in 1891 and its estrogenic interfere with hormonal signaling (WHO/UNEP, 2013). As hormones properties discovered by Edward Charles Dodds in the 1930s while are important for brain development, it may not be surprising that developing estrogenic pharmaceuticals. However, BPA was never prenatal exposure to BPA has been associated with impaired neuro­ commercialized for its estrogenic properties but instead for the usage in development in children, primarily with behavioural changes like anx­ epoxy resins and polycarbonate plastics where it has been extensively iety, depression, aggression and hyperactivity (Ejaredar et al., 2017; used during the past 60 years (Vogel, 2009). The ubiquitous use of this Mustieles et al., 2015; Rochester et al., 2018). The associations observed chemical is reflected in its presence in urine samples of >95% of between BPA exposure and neurodevelopmental outcomes show clear

* Corresponding author at: Sciences, Karlstad University, 651 88, Karlstad, Sweden. E-mail address: [email protected] (C.-G. Bornehag). https://doi.org/10.1016/j.envint.2021.106433 Received 9 October 2020; Received in revised form 29 January 2021; Accepted 29 January 2021 Available online 23 February 2021 0160-4120/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). C.-G. Bornehag et al. Environment International 150 (2021) 106433 differences depending on sex of the exposed foetus (Braun et al., 2011; the Regional Ethical Review Board (Uppsala, Sweden). Harley et al., 2013; Evans et al., 2014; Perera et al., 2016; Braun et al., We aimed at assessing cognitive function in 1,000 children at 7 years 2017b; Stacy et al., 2017; Jensen et al., 2019; Freire et al., 2020). of age. Families were invited consecutively based on the child’s age and Due to its health risks in sensitive populations, the European Union 924 were ultimately assessed. The current analysis includes 803 children and Canada now prohibit BPA in consumer products designed for chil­ with complete data on exposure, outcome, and potential confounding dren, such as baby bottles, food contact materials for baby food and toys variables (Table 1). (European Commission., 2011, Canada, 2010). However, as the use of BPA is decreasing, usage of BPA analogues is on the rise. Two common 2.2. Prenatal bisphenol exposure BPA analogues are Bisphenol S (BPS) and Bisphenol F (BPF). As with BPA, these analogues are present in canned food, thermal receipt papers, First-morning void urine samples were collected from mothers dur­ personal care products, and water pipes (Chen et al., 2016). The shift ing their first prenatal visit (median 10 weeks of pregnancy). Samples ◦ from BPA to BPF and BPS containing products is also observed in human were stored at –20 C and analysed at the Laboratory of Occupational exposure studies. For example, while BPA concentrations in urine and Environmental Medicine at Lund University, Lund, Sweden. decreased between 2011 and 2016 among pregnant women in Puerto The urine samples were analyzed by liquid chromatography tandem Rico, as well as between 2000 and 2014 in US adults, BPS concentrations mass spectrometry (LC-MS/MS) described by Gyllenhammar et al increased (Ashrap et al., 2018; Ye et al., 2015). Similarly, while BPA (Gyllenhammar et al., 2017) to quantify BPS, BPF and BPA. Briefly, concentration in Swedish mothers slightly declined between 2009 and samples were deconjugated using β-glucoronidase, and incubated for 30 2014, there was a clear increase in urinary BPF levels (Gyllenhammar min. Labeled internal standards for each compound were used. The et al., 2017). The presence of BPA in women of reproductive age is of samples were analyzed in a randomized order and in duplicate. interest since this chemical can cross the placenta and reach the foetus The limits of detection (LOD) were 0.03, 0.03, 0.2 ng/mL for BPS, (Liu et al., 2017). Furthermore, bisphenols likely reach the fetal brain BPF and BPA respectively. For concentrations below the LOD, reported since they have been detected in human adult brain, and thus are able to (measured) concentrations were used in the statistical calculations – cross the blood brain barrier. For example, one study detected BPA and (Perkins and Schisterman, 2006). The laboratory is a reference labora­ BPF in postmortem brain tissues in 23 out of 24 Dutch adults (Char­ tory for analyses of urinary BPA in the Erlangen inter-comparison pro­ isiadis et al., 2018). gram (Professor Dr. med. Hans Drexler, Institute and Out-Patient Clinic BPA, BPS, and BPF are structurally similar, and show similar endo­ for Occupational, Social and Environmental Medicine, University of crine disrupting properties (Rochester and Bolden, 2015; Rosenfeld, Erlangen-Nuremberg, Germany) with results within the tolerance limits. 2017). The estrogenic responses of all three bisphenols are well estab­ To adjust the bisphenol levels for urinary dilution, creatinine was lished (Molina-Molina et al., 2013; Rosenmai et al., 2014; Kitamura analyzed at the Department of , Lund University et al., 2005; Cano-Nicolau et al., 2016; Pelch et al., 2019) but these Hospital. The laboratory is accredited for creatinine analysis and is using compounds can, to varying extent, also bind to other hormonal receptors an enzymatic colorimetric method (Mazzachi et al., 2000). The unit for as well as interfere with steroid hormone synthesis (Rosenmai et al., adjusted bisphenol levels is nmol/mmol. 2014). For example, BPA and BPF (but not BPS) have been shown to ˇ bind the thyroid hormone transport protein transthyretin (Sauer et al., 2021), and all three bisphenols (although BPS to a lesser extent) display 2.3. Cognitive function in children at age 7 years ˇ clear anti-androgenic activity (Rosenmai et al., 2014; Sauer et al., 2021). There is experimental evidence that BPA, BPS and BPF affect neu­ Nine trained psychologists evaluated cognitive functioning using the rodevelopmental processes and behaviour (Yin et al., 2019; Gu et al., 2019; Catanese and Vandenberg, 2016; Kinch et al., 2015; Rosenfeld, Table 1 2017). Furthermore, there is increasing consistent epidemiological evi­ Description of the study participants. dence that prenatal exposure to BPA affects child neurobehavior (Mus­ Characteristics All Boys Girls p-value tieles et al., 2015). However, human studies specific to children * neurodevelopmental outcomes and exposure to the BPA alternatives are N = 803 N = 400 N = 403 scarce. One study has reported that higher concentrations of BPS during N (%) pregnancy is associated with lower psychomotor development in the Mothers education (university 543 272 271 0.819 level or higher) (67.6) (68.0) (67.2) children at two years of age (Jiang et al., 2020). We previously showed Mothers smoking during 27 (3.4) 15 (3.8) 12 (3.0) 0.544 that prenatal exposure to a mixture of 26 potential endocrine disrupting pregnancy chemicals was associated with lower IQ among 7-year old children in the Mean (SD) Swedish Environmental Longitudinal Mother and Child, Asthma and Mothers age (year) 31.4 (4.7) 31.4 31.5 0.916 (SELMA) study (Tanner et al., 2020). The whole mixture effect (4.6) (4.8) Mothers weight (kg) 69.4 (14) 69.0 69.9 0.350 was largely weighted by bisphenol compounds, warranting further (13) (14) investigation. Therefore, to inform the single chemical risk assessment Mothers IQ score** 114.4 114.6 114.2 0.714 paradigm, the present study aimed at examine the individual associa­ (16) (16) (15) ́ *** tions between prenatal BPA, BPS and BPF exposure and cognitive out­ Childs cognition scores Full Scale IQ 99.9 (13) 97.9 101.9 <0.001 comes in children at the age of 7 years. (13) (12) Verbal Comprehension 100.8 99.6 102.0 0.003 2. Methods (11) (12) (11) Perceptual Reasoning 106.2 105.8 106.8 0.218 2.1. Study population (12) (13) (12) Working Memory 89.8 (13) 88.3 91.3 0.002 (13) (14) The SELMA study is a population-based pregnancy cohort that Processing Speed 98.9 (15) 95.6 102.2 <0.001 recruited more than 2,300 women in the first trimester from prenatal (15) (15) ¨ clinics in Varmland county, Sweden, from November 2007 to March * Differences between boys and girls were analyzed with χ2 test for categorical 2010. Detailed recruitment and sample collection procedures have been variables and with t-test for continuous variables. described previously (Bornehag et al., 2015, Bornehag et al., 2012). ** Shortened Ravens Standard Progressive Matrices. Participants provided written consent and the study was approved by *** WISC-IV.

2 C.-G. Bornehag et al. Environment International 150 (2021) 106433

Wechsler Intelligence Scale for Children (WISC-IV), 4th edition verbal comprehension (102 vs. 100, p = 0.003), working memory (91 vs. (Wechsler, 2003). Four sub scales were used including the Verbal 88, p = 0.002), and processing speed (102 vs. 96, p < 0.001). There was Comprehension Index measuring verbal reasoning and comprehension, no difference in perceptual reasoning scores by sex. the Perceptual Reasoning Index measuring problem solving and visual Descriptive statistics for the three bisphenols are reported in Table 2 processing, the Working Memory Index measuring mental control, and and in Supplemental Table 2. BPA was found above the level of detection the Processing Speed Index measuring the speed and accuracy of visual (LOD) in all samples, while 97.4% and 91.4% of the samples contained information processing (Prifiteraet al., 2005). A full-scale IQ was used, BPS and BPF above the LOD, respectively. BPA had the highest GM reflecting overall cognitive abilities based on the sub-scales, which is concentrations (1.55 ng/mL), followed by BPF (0.16 ng/mL) and BPS indexed to have a mean and standard deviation of 100, and 15, (0.07 ng/mL). There were no significant differences in urine concen­ respectively. trations by sex of the foetus. Urinary levels of the bisphenols were weakly correlated (BPA vs. BPS, r = 0.18, p < 0.01; BPA vs. BPF, r = 2.4. Co-factors 0.14, p < 0.01; BPS vs. BPF r = 0.10, p < 0.01).

At enrollment, data on mothers’ education level (university or 3.1. Association between prenatal bisphenol exposure and cognitive higher) and smoking (any smoking during pregnancy reported at function in children recruitment) were collected by questionnaires and data on the mothers’ pre-pregnancy weight was collected from the National birth register. Prenatal BPF exposure was significantly associated with lower Maternal IQ was assessed using the shortened Ravens Standard Pro­ cognitive function in both crude and adjusted analyses (Fig. 1). The gressive Matrices (Van der Elst et al., 2013; Raven et al., 1998). These largest magnitude was observed for full scale IQ (β = 1.96, 95%CI; factors together with sex of the child and the mothers age were used for 3.12; 0.80), p = 0.001), when adjusting for sex of the child, and basic adjustments (Model 1) including 803 children. Missing values for maternal age, education, IQ, weight and smoking (Model 1), that is a the variables in Model 1 is described in Supplemental Table 1. 1.96 point decrease in IQ score per each log-unit increase in BPF con­ In a sensitivity analysis, we assessed additional covariates (Model 2), centrations, or a 1.6-point lower IQ score for an inter-quartile-range based on prior literature, including birthweight, prematurity (<37 week (IQR) change in prenatal BPF exposure (IQR = 0.05 – 0.35 ng/mL, of pregnancy) and parity from the National birth register and breast Table 2). Significant adjusted associations were also observed between feeding from questionnaires when the child was 6 months. When adding BPF and all four sub-scales, with the largest association observed for these additional co-factors to the statistical model the sample size processing speed: verbal comprehension (β = 1.10 (95%CI 2.16; became smaller (N = 712). 0.03), p = 0.044), perceptual reasoning (β = 1.51 (95%CI 2.67; 0.35), p = 0.011), working memory (β = 1.74 (-3.00; 0.47), p = 2.5. Statistics 0.007), and processing speed (β = 1.83 (95%CI 3.28; 0.38), p = 0.014). For univariate analyses, we log10-transformed the creatinine- In a sensitivity analysis, we assessed additional covariates including adjusted bisphenols to approximate a normal distribution and calcu­ birth weight, prematurity, parity, and breast feeding at six months of age lated geometric means (GM) and 95% confidenceintervals (95% CI). In in the statistical model (Eriksen et al., 2013; Oken et al., 2008). How­ the descriptive statistics presented, the χ2 test and t-tests were used to ever, no major changes in associations between prenatal BPF exposure assess potential differences between boys and girls. We also evaluated and cognitive function were observed (Fig. 1, Model 2). In analyses of Pearson correlations between creatinine-adjusted BPA, BPS, and BPF, bisphenol tertiles, the highest BPA and BPS were not associated with and tested for trends in exposure levels across the sampling period. WISC-IV scores. Compared to the lowest tertile of BPF, the highest tertile Associations between log-transformed prenatal bisphenol exposure was associated with lower IQ, working memory and processing speed (ng/mL) and cognitive function at 7 years of age was estimated with (Supplemental Table 2) indicating dose response associations. crude and adjusted linear regression. For the adjusted models we chose We observed no statistical interaction between child sex and covariates based on prior literature and our own data, (see directed bisphenol exposure on cognitive outcomes (Fig. 2). However, sex- acyclic graph (DAG) in Supplemental Fig. 1) (Eriksen et al., 2013; Bor­ stratified analyses adjusted for Model 1 covariates revealed that asso­ nehag et al., 2018); the final linear model was adjusted for sex of the ciations between BPF and cognitive functioning reached significancein child, and maternal age, education, IQ, weight and smoking (Model 1). boys but not in girls, although the direction of the beta coefficientswere Since prior literature suggests boys and girls to have different neuro­ in the same directions in both sexes. The association for IQ corresponded developmental responses to BPA exposure (Mustieles and Fernandez, to a 2.3-point decline in IQ score for an IQR change in prenatal BPF 2020), we examined the interaction effect of sex and exposure and exposure among boys. We found no evidence for an association between conducted sex-stratified analyses. In a sensitivity analysis, we assessed prenatal exposure for BPA or BPS and cognitive function in total nor sex- additional covariates based on prior literature, including birth weight, stratified analyses. prematurity, parity, and breast feeding at six months of age (Model 2) (Eriksen et al., 2013; Oken et al., 2008). We also assessed potential 4. Discussion nonmonotonicity using bisphenol tertiles as the exposure variable adjusting for Model 1 co-variates. This study is the first to show that prenatal BPF exposure is associ­ Analyses were conducted in SPSS, version 25. ated with impaired cognitive development in children. Our finding imply that boys may be more susceptible to BPF exposure in utero with 3. Results regards to cognitive development. This is in line with reports showing an overall higher risk of neurodevelopmental problems, specifically The study participants consisted of 803 mother–child pairs (400 boys cognitive deficits, in boys compared to girls prenatally exposed to and 403 girls) for which we had information on all variables described in different drugs (Traccis et al., 2020). The notion that boys would be Table 1. Sixty eight percent of the mothers achieved a college education more susceptible to prenatal chemical exposure is also reflected in an or higher and 3.5% were active smokers during pregnancy. The mean IQ overall higher risk of neurodevelopmental disorders in males, compared of the mothers was 114. Among the children, 4.4% were born before to females, which may be due to differences in biological mechanisms, gestational week 37. The mean full scale IQ score was 100 with signif­ for example hormone signaling, between sexes (May et al., 2019). The icant higher mean score in girls compared to boys (102 vs. 98, p < difference in association strength observed between BPF and cognition 0.001). Compared to boys, girls had also significantly higher scores for in the sexes is in line with previous findingswhere BPA has been shown

3 C.-G. Bornehag et al. Environment International 150 (2021) 106433

Table 2 Distribution of urinary levels of three bisphenols in 803 pregnant women and comparisons between mothers with boys and girls foetus. A further description including percentiles is available in Supplemental Table 2. Phenols Urinary levels (ng/mL) p-value*

All women Boys Girls (N = 803) (N = 400) (N = 403)

25% Median 75% GM GM GM (95% CI) (95% CI) (95% CI)

BPA 0.84 1.44 2.66 1.55 1.54 1.53 0.901 (1.46–1.64) (1.41–1.68) (1.41–1.66) BPS 0.04 0.07 0.12 0.07 0.07 0.07 0.584 (0.06–0.07) (0.06–0.08) (0.06–0.08) BPF 0.05 0.13 0.35 0.16 0.16 0.15 0.632 (0.14–0.18) (0.14–0.19) (0.13–0.18)

* Differences between boys and girls analyzed with t-test.

Fig. 1. Association between prenatal bisphenol exposure and cognitive function in children at 7 years of age. * Linear regression model 1: Adjusted for sex of the child, and maternal age, education, IQ, weight and smoking. ** Linear regression model 2: Adjusted for sex of the child, and maternal age, education, IQ, weight and smoking, plus birth weight, prematurity, parity, and breast feeding at six months of age.

4 C.-G. Bornehag et al. Environment International 150 (2021) 106433

Fig. 2. Sex-stratifiedassociation between prenatal bisphenol exposure and cognitive function at 7 years of age. Results from linear regression model 1 (adjusted for sex of the child, and maternal age, education, IQ, weight and smoking) on cognitive test scores at age 7 years, stratifiedby sex (N = 400 boys and 403 girls). p-values obtained for the association between bisphenol exposure and cognitive test scores, in sex-stratified analyses, and for the interaction effect of sex and exposure (SexxBP) on cognitive test scores. to affect the fetus differently depending on its sex (Braun et al., 2011; et al., 2017a) and 9–11 year old boys (Rodriguez-Carrillo et al., 2019) Harley et al., 2013; Evans et al., 2014; Perera et al., 2016; Braun et al., and placenta BPA have been associated to lower verbal scores in 4–5 2017b; Jensen et al., 2019; Braun et al., 2017a). Furthermore, a recent year old boys (Freire et al., 2020). However, our data are in line with study investigating associations between urinary BPF levels and meta­ most other studies where BPA has not been associated with cognitive bolic outcomes also found stronger associations in boys than in girls (Liu development (Casas et al., 2015; Nakiwala et al., 2018; Rodriguez- et al., 2019). Since we could not findany association with BPA or BPS in Carrillo et al., 2019; Minatoya et al., 2017). One reason for lack of/ neither boys nor girls, the results may indicate a functional difference null association with BPA findingsmay be that BPA has more impact on between these three bisphenols. neurodevelopment in the behavioral domain, which has been shown and Our lack of/null associations regarding BPA and cognitive develop­ discussed earlier (Mustieles and Fernandez, 2020; Rodriguez-Carrillo ment, in one way contradict some studies where prenatal or cord-blood et al., 2019). In the current study, behavioral outcomes were not BPA has been associated with lower full scale IQ at 7 years of age (Lin evaluated. et al., 2017), lower verbal scores in boys at 21 months of age, (Jensen When comparing our BPA, BPF and BPS detection frequencies and et al., 2019), and poorer working memory in 3 year old boys (Braun levels (Table 2) to other studies analyzing these three chemicals, all

5 C.-G. Bornehag et al. Environment International 150 (2021) 106433 studies found BPA to have the highest median concentration, varying cognition observed with the whole mixture approach, but does not find between 0.7 and 2.0 µg/L (In our study: 100% detection, GM 1.55 µg/L). any evidence that BPA or BPS has the same effect. Since correlations However, BPF was more abundant than BPS in studies from Sweden and between bisphenols were low, it is unlikely that these differing results China (Gyllenhammar et al., 2017; Jiang et al., 2020), while BPS was regarding BPA and BPS were solely due to collinearity in the WQS more abundant in the United States and Puerto Rico (Ashrap et al., 2018; mixture analysis or co-pollutant confounding in the single chemical Lehmler et al., 2018). We detected BPF and BPS in 91% and 97% of the analysis. Also, BPF and BPS are replacing BPA in consumer products, women (GM 0.16 µg/L and GM 0.07 µg/L respectively), compared to rather than co-occurring. Although we controlled for known con­ BPF detections between 41 and 98% (0.3–0.6 µg/L) and BPS detections founders, residual confounding may explain these differences. Further between 68 and 90% (0.04 – 0.5 µg/L) observed by others (Ashrap et al., research on bisphenol exposure sources and patterns may help identify 2018; Gyllenhammar et al., 2017; Jiang et al., 2020; Lehmler et al., additional unknown confounding variables. For comparison with our 2018). We can only speculate on the reasons for these differences in results and to obtain a better understanding of how bisphenols impact measured values. They could be due to methodological differences be­ neurodevelopment, we recommend that future epidemiologic studies tween analytical laboratories or actual exposure differences between also measure BPA analogues and incorporate both mixture and single populations due to different exposure sources or different collection chemical statistical analyses. times. We observed that one IQR increase in BPF exposure was associated We found median BPF levels in the SELMA mothers to be about 10 with a 2-point drop in full scale IQ. When we collected the samples in times lower than the BPA levels (Table 2), potentially adding an addi­ 2007–2010 almost all individuals were found exposed to BPF, and since tional dimension to the question why the associations observed with BPF this exposure is likely to increase due to BPA substitutions, we findsuch are much stronger than with BPA. However, EDCs are known to evoke magnitude worrying. A 2-point drop, on an individual level, may not be non-monotonic dose responses and thus to have different effects noticed by the affected individual but is worrying since a 2-point shift in depending on concentration (Vandenberg et al., 2012), which may IQ level of a population will lead to less individuals with very high IQ partly explain that BPF shows the strongest associations despite being and increase the number of individuals with very low IQ (Bellinger, present in lower levels than BPA. However, we did not findevidence for 2007). A shift in population IQ can lead to increased intellectual a non-monotonic dose response relationship in the current data. Alter­ disability and productivity loss, which brings high costs for the society natively, these compounds might have different toxicokinetic proper­ (Bellanger et al., 2015; Gaylord et al., 2020). Considering the fact that ties, and thus the actual foetal exposure might not correlate linearly to almost all (91%) of the SELMA mothers that were included in this study the levels in mother’s urine. Notably, in sheep, BPF has a longer half-life had BPF levels above the LOD, and that this exposure is likely to increase in maternal circulation, but a faster clearance from foetal circulation, due to BPA substitution, we consider our findings as alarming from a compared to BPA and BPS (Gingrich et al., 2019). public health perspective. It is intriguing, yet mechanistically difficult to explain, that only an association between BPF and cognition was observed. All three bisphenols assessed in this study have the potential to interfere with the 4.1. Strengths and limitations endocrine systems that are important for neurodevelopment and could potentially be involved in effects on cognition. Yet, experimental studies The study had additional strengths and limitations. The study design, have shown that BPA, BPS and BPF vary in their specific effects and the sample size and control for potential confounders gave us good potencies on different hormonal systems (Rochester and Bolden, 2015), opportunities to find relevant associations. However, although we which might underlie the observed differences. controlled for many relevant confounders, there may be residual con­ Although several hormones are important for cognitive develop­ founding from variables we were unable to collect. Furthermore, urinary ment, the most well established are thyroid hormones (THs). Distur­ levels of bisphenols may fluctuate over time due to short half-lives and bances in TH levels during early foetal development are known to impair the number of products that act as sources. One single measurement may cognitive function (Haddow et al., 1999; Korevaar et al., 2016; Pop therefore be questioned as a representative measure for exposure. To et al., 1999; Willoughby et al., 2014) and cause IQ loss in the child resolve this problem, we used first morning void samples, which we (Haddow et al., 1999; Korevaar et al., 2016). In the SELMA mothers, we regarded as the baseline level for that day. have observed a positive association between BPF levels and free triio­ dothyronine (FT3), which was not observed for BPA or BPS (Derakhshan 5. Conclusion et al., 2020), indicating a difference of these bisphenols in their effect on the thyroid system. Moreover, all three bisphenols in this study have This study shows for the first time that prenatal BPF exposure is been shown to interfere with TH signaling in different experimental associated with decreased cognitive function in children at 7 years, with systems (Zhu et al., 2018; Lee et al., 2019; Zhang et al., 2018; Huang significantassociations in boys. These findingsraise concerns since BPF et al., 2016; Lee et al., 2017). However, BPF may differ from BPA and is increasingly used to replace BPA, and thus exposure of humans and BPS in some aspects of this thyroid disrupting effects (Lee et al., 2019; wildlife is rising globally. Since there are currently no regulatory Lee et al., 2017). Nevertheless, more experimental results are needed to guideline values for BPF, our results will inform future risk assessments. decipher the role of BPF, and its possible differences compared to BPA and BPS, on TH signaling before concluding if BPF-mediated disturbance Funding in TH signaling may be an underlying cause in the association with lower IQ observed in this study. Since relatively little is known about This work was supported by the European Union’s Horizon 2020 how BPF can affect neurodevelopment, it is likely that other mechanisms Research and Innovation Programme [project: EDC-MixRisk; grant and pathways can contribute to the associations described here. number 634880], the Swedish research council for sustainable devel­ In this study population, we recently used a whole mixture approach opment (Formas) [grant number 2017-00513], and the Powering to identify chemicals of concern in relation to lower IQ among 26 Research Through Innovative Methods for Mixtures in Epidemiology chemicals in prenatal urine and serum using weighted quantile sum (PRIME) Program [grant number R01ES028811-01]. (WQS) regression (Tanner et al., 2020). We found BPF to have the largest contribution (weight) to the mixture. We also found that BPA and BPS contributed to the mixture effect, particularly for boys and girls Declaration of Competing Interest respectively, although at much lower weights compared to BPF. The analyses of the present study confirmsthe association between BPF and None.

6 C.-G. Bornehag et al. Environment International 150 (2021) 106433

Appendix A. Supplementary material Gingrich, J., Pu, Y., Ehrhardt, R., Karthikraj, R., Kannan, K., Veiga-Lopez, A., 2019. Toxicokinetics of bisphenol A, bisphenol S, and bisphenol F in a pregnancy sheep model. Chemosphere 220, 185–194. Supplementary data to this article can be found online at https://doi. Gu, J., Zhang, J., Chen, Y., Wang, H., Guo, M., Wang, L., Wang, Z., Wu, S., Shi, L., Gu, A., org/10.1016/j.envint.2021.106433. 2019. Neurobehavioral effects of bisphenol S exposure in early life stages of zebrafish larvae (Danio rerio). Chemosphere 217, 629–635. Gyllenhammar, I., Glynn, A., Jonsson, B.A., Lindh, C.H., Darnerud, P.O., Svensson, K., References Lignell, S., 2017. Diverging temporal trends of human exposure to bisphenols and plastizisers, such as phthalates, caused by substitution of legacy EDCs? Environ. Res. Ashrap, P., Watkins, D.J., Calafat, A.M., Ye, X., Rosario, Z., Brown, P., Velez-Vega, C.M., 153, 48–54. Alshawabkeh, A., Cordero, J.F., Meeker, J.D., 2018. Elevated concentrations of Haddow, J.E., Palomaki, G.E., Allan, W.C., Williams, J.R., Knight, G.J., Gagnon, J., urinary triclocarban, phenol and paraben among pregnant women in Northern O’Heir, C.E., Mitchell, M.L., Hermos, R.J., Waisbren, S.E., Faix, J.D., Klein, R.Z., Puerto Rico: Predictors and trends. Environ. Int. 121, 990–1002. 1999. Maternal thyroid deficiency during pregnancy and subsequent Bellanger, M., Demeneix, B., Grandjean, P., Zoeller, R.T., Trasande, L., 2015. neuropsychological development of the child. N. Engl. J. Med. 341, 549–555. Neurobehavioral deficits, diseases, and associated costs of exposure to endocrine- Harley, K.G., Gunier, R.B., Kogut, K., Johnson, C., Bradman, A., Calafat, A.M., disrupting chemicals in the European Union. J. Clin. Endocrinol. Metab. 100, Eskenazi, B., 2013. Prenatal and early childhood bisphenol A concentrations and 1256–1266. behavior in school-aged children. Environ. Res. 126, 43–50. Bellinger, D.C., 2007. Children’s cognitive health: the influence of environmental Huang, G.M., Tian, X.F., Fang, X.D., Ji, F.J., 2016. Waterborne exposure to bisphenol F chemical exposures. Altern. Ther. Health Med. 13, S140–S144. causes thyroid endocrine disruption in zebrafishlarvae. Chemosphere 147, 188–194. Bornehag, C.G., Carlstedt, F., Jonsson, B.A., Lindh, C.H., Jensen, T.K., Bodin, A., Jensen, T.K., Mustieles, V., Bleses, D., Frederiksen, H., Trecca, F., Schoeters, G., Jonsson, C., Janson, S., Swan, S.H., 2015. Prenatal phthalate exposures and Andersen, H.R., Grandjean, P., Kyhl, H.B., Juul, A., Bilenberg, N., Andersson, A.M., anogenital distance in Swedish boys. Environ. Health Perspect. 123, 101–107. 2019. Prenatal bisphenol A exposure is associated with language development but Bornehag, C.G., Lindh, C., Reichenberg, A., Wikstrom, S., Unenge Hallerback, M., not with ADHD-related behavior in toddlers from the Odense Child Cohort. Environ. Evans, S.F., Sathyanarayana, S., Barrett, E.S., Nguyen, R.H.N., Bush, N.R., Swan, S. Res. 170, 398–405. H., 2018. Association of Prenatal Phthalate Exposure With Language Development in Jiang, Y., Li, J., Xu, S., Zhou, Y., Zhao, H., Li, Y., Xiong, C., Sun, X., Liu, H., Liu, W., Early Childhood. JAMA Pediatr. 172, 1169–1176. Peng, Y., Hu, C., Cai, Z., Xia, W., 2020. Prenatal exposure to bisphenol A and its Bornehag, C.G., Moniruzzaman, S., Larsson, M., Lindstrom, C.B., Hasselgren, M., alternatives and child neurodevelopment at 2 years. J. Hazard. Mater. 388, 121774. Bodin, A., von Kobyletzkic, L.B., Carlstedt, F., Lundin, F., Nanberg, E., Jonsson, B.A., Kinch, C.D., Ibhazehiebo, K., Jeong, J.-H., Habibi, H.R., Kurrasch, D.M., 2015. Low-dose Sigsgaard, T., Janson, S., 2012. The SELMA study: a birth cohort study in Sweden exposure to bisphenol A and replacement bisphenol S induces precocious following more than 2000 mother-child pairs. Paediatr. Perinat. Epidemiol. 26, hypothalamic neurogenesis in embryonic zebrafish. Proc. Natl. Acad. Sci. 112, 456–467. 1475–1480. Braun, J.M., Kalkbrenner, A.E., Calafat, A.M., Yolton, K., Ye, X., Dietrich, K.N., Kitamura, S., Suzuki, T., Sanoh, S., Kohta, R., Jinno, N., Sugihara, K., Yoshihara, S., Lanphear, B.P., 2011. Impact of early-life bisphenol A exposure on behavior and Fujimoto, N., Watanabe, H., Ohta, S., 2005. Comparative study of the endocrine- executive function in children. 128, 873–882. disrupting activity of bisphenol A and 19 related compounds. Toxicol. Sci. 84, Braun, J.M., Muckle, G., Arbuckle, T., Bouchard, M.F., Fraser, W.D., Ouellet, E., 249–259. Seguin, J.R., Oulhote, Y., Webster, G.M., Lanphear, B.P., 2017a. Associations of Korevaar, T.I., Muetzel, R., Medici, M., Chaker, L., Jaddoe, V.W., de Rijke, Y.B., Prenatal Urinary Bisphenol A Concentrations with Child Behaviors and Cognitive Steegers, E.A., Visser, T.J., White, T., Tiemeier, H., Peeters, R.P., 2016. Association Abilities. Environ. Health Perspect. 125, 067008. of maternal thyroid function during early pregnancy with offspring IQ and brain Braun, J.M., Yolton, K., Stacy, S.L., Erar, B., Papandonatos, G.D., Bellinger, D.C., morphology in childhood: a population-based prospective cohort study. Lancet Diab. Lanphear, B.P., Chen, A., 2017b. Prenatal environmental chemical exposures and Endocrinol. 4, 35–43. longitudinal patterns of child neurobehavior. Neurotoxicology 62, 192–199. Lee, S., Kim, C., Shin, H., Kho, Y., Choi, K., 2019. Comparison of thyroid hormone Canada., G.O., 2010. Order Amending Schedule I to the Hazardous Products Act disruption potentials by bisphenols A, S, F, and Z in embryo-larval zebrafish. (bisphenol A). Chemosphere 221, 115–123. Cano-Nicolau, J., Vaillant, C., Pellegrini, E., Charlier, T.D., Kah, O., Coumailleau, P., Lee, S., Kim, C., Youn, H., Choi, K., 2017. Thyroid hormone disrupting potentials of 2016. Estrogenic Effects of Several BPA Analogs in the Developing Zebrafish Brain. bisphenol A and its analogues - in vitro comparison study employing rat pituitary Front. Neurosci. 10, 112. (GH3) and thyroid follicular (FRTL-5) cells. Toxicol. In Vitro 40, 297–304. Casas, M., Forns, J., Martinez, D., Avella-Garcia, C., Valvi, D., Ballesteros-Gomez, A., Lehmler, H.J., Liu, B., Gadogbe, M., Bao, W., 2018. Exposure to Bisphenol A, Bisphenol F, Luque, N., Rubio, S., Julvez, J., Sunyer, J., Vrijheid, M., 2015. Exposure to bisphenol and Bisphenol S in U.S. Adults and Children: The National Health and Nutrition A during pregnancy and child neuropsychological development in the INMA- Examination Survey 2013–2014. ACS Omega 3, 6523–6532. Sabadell cohort. Environ. Res. 142, 671–679. Lin, C.C., Chien, C.J., Tsai, M.S., Hsieh, C.J., Hsieh, W.S., Chen, P.C., 2017. Prenatal Catanese, M.C., Vandenberg, L.N., 2016. Bisphenol S (BPS) alters maternal behavior and phenolic compounds exposure and neurobehavioral development at 2 and 7years of brain in mice exposed during pregnancy/lactation and their daughters. age. Sci. Total Environ. 605–606, 801–810. 158, 516–530. Liu, B., Lehmler, H.J., Sun, Y., Xu, G., Sun, Q., Snetselaar, L.G., Wallace, R.B., Bao, W., Charisiadis, P., Andrianou, X.D., van der Meer, T.P., den Dunnen, W.F.A., Swaab, D.F., 2019. Association of Bisphenol A and Its Substitutes, Bisphenol F and Bisphenol S, Wolffenbuttel, B.H.R., Makris, K.C., Van Vliet-Ostaptchouk, J.V., 2018. Possible with Obesity in United States Children and Adolescents. Diab. Metab. J. 43, 59–75. Obesogenic Effects of Bisphenols Accumulation in the Human Brain. Sci. Rep. 8, Liu, J., Li, J., Wu, Y., Zhao, Y., Luo, F., Li, S., Yang, L., Moez, E.K., Dinu, I., Martin, J.W., 8186. 2017. Bisphenol A Metabolites and Bisphenol S in Paired Maternal and Cord Serum. Chen, D., Kannan, K., Tan, H., Zheng, Z., Feng, Y.L., Wu, Y., Widelka, M., 2016. Environ. Sci. Technol. 51, 2456–2463. Bisphenol Analogues Other Than BPA: Environmental Occurrence, Human Exposure, May, T., Adesina, I., McGillivray, J., Rinehart, N.J., 2019. Sex differences in and Toxicity-A Review. Environ. Sci. Technol. 50, 5438–5453. neurodevelopmental disorders. Curr. Opin. Neurol. 32, 622–626. Derakhshan, A., Philips, E.M., Ghassabian, A., Santos, S., Asimakopoulos, A.G., Mazzachi, B.C., Peake, M.J., Ehrhardt, V., 2000. Reference range and method Kannan, K., Kortenkamp, A., Jaddoe, V.W.V., Trasande, L., Peeters, R.P., comparison studies for enzymatic and Jaffe creatinine assays in plasma and serum Korevaar, T.I.M., 2020. Association of urinary bisphenols during pregnancy with and early morning urine. Clin. Lab. 46, 53–55. maternal, cord blood and childhood thyroid function. Environ. Int. 146, 106160. Minatoya, M., Araki, A., Nakajima, S., Sasaki, S., Miyashita, C., Yamazaki, K., Ejaredar, M., Lee, Y., Roberts, D.J., Sauve, R., Dewey, D., 2017. Bisphenol A exposure Yamamoto, J., Matumura, T., Kishi, R., 2017. Cord blood BPA level and child and children’s behavior: A systematic review. J. Expo Sci. Environ. Epidemiol. 27, neurodevelopment and behavioral problems: The Hokkaido Study on Environment 175–183. and Children’s Health. Sci. Total Environ. 607–608, 351–356. Eriksen, H.L., Kesmodel, U.S., Underbjerg, M., Kilburn, T.R., Bertrand, J., Mortensen, E. Molina-Molina, J.M., Amaya, E., Grimaldi, M., Saenz, J.M., Real, M., Fernandez, M.F., L., 2013. Predictors of intelligence at the age of 5: family, pregnancy and birth Balaguer, P., Olea, N., 2013. In vitro study on the agonistic and antagonistic characteristics, postnatal influences, and postnatal growth. PLoS ONE 8, e79200. activities of bisphenol-S and other bisphenol-A congeners and derivatives via nuclear European Commission, 2011. Commission directive 2011/8/EU of 28 January 2011 receptors. Toxicol. Appl. Pharmacol. 272, 127–136. amending directive 2002/72/EC as regards the restriction of use of bisphenol A in Mustieles, V., Fernandez, M.F., 2020. Bisphenol A shapes children’s brain and behavior: plastic infant feeding bottles. Official Journal of the European Union. towards an integrated neurotoxicity assessment including human data. Environ. Evans, S.F., Kobrosly, R.W., Barrett, E.S., Thurston, S.W., Calafat, A.M., Weiss, B., Health 19, 66. Stahlhut, R., Yolton, K., Swan, S.H., 2014. Prenatal bisphenol A exposure and Mustieles, V., P´erez-Lobato, R., Olea, N., Fernandez,´ M.F., 2015. Bisphenol A: Human maternally reported behavior in boys and girls. Neurotoxicology 45C, 91–99. exposure and neurobehavior. Neurotoxicology 49, 174–184. Freire, C., Vela-Soria, F., Beneito, A., Lopez-Espinosa, M.J., Ibarluzea, J., Barreto, F.B., Nakiwala, D., Peyre, H., Heude, B., Bernard, J.Y., Beranger,´ R., Slama, R., Philippat, C., Casas, M., Vrijheid, M., Fernandez-Tardon, G., Riano-Galan, I., Fernandez, M.F., 2018. In-utero exposure to phenols and phthalates and the intelligence quotient of Project, I, 2020. Association of placental concentrations of phenolic endocrine boys at 5 years. Environ. Health 17, 17. disrupting chemicals with cognitive functioning in preschool children from the Oken, E., Osterdal, M.L., Gillman, M.W., Knudsen, V.K., Halldorsson, T.I., Strom, M., Environment and Childhood (INMA) Project. Int. J. Hyg. Environ. Health 230, Bellinger, D.C., Hadders-Algra, M., Michaelsen, K.F., Olsen, S.F., 2008. Associations 113597. of maternal fish intake during pregnancy and breastfeeding duration with Gaylord, A., Osborne, G., Ghassabian, A., Malits, J., Attina, T., Trasande, L., 2020. Trends attainment of developmental milestones in early childhood: a study from the Danish in neurodevelopmental disability burden due to early life chemical exposure in the National Birth Cohort. Am. J. Clin. Nutr. 88, 789–796. USA from 2001 to 2016: A population-based disease burden and cost analysis. Mol. Pelch, K.E., Li, Y., Perera, L., Thayer, K.A., Korach, K.S., 2019. Characterization of Cell Endocrinol. 502, 110666. Estrogenic and Androgenic Activities for Bisphenol A-like Chemicals (BPs). In Vitro

7 C.-G. Bornehag et al. Environment International 150 (2021) 106433

Estrogen and Androgen Receptors Transcriptional Activation, Gene Regulation, and Stacy, S.L., Papandonatos, G.D., Calafat, A.M., Chen, A., Yolton, K., Lanphear, B.P., Binding Profiles. Toxicol. Sci. 172, 23–37. Braun, J.M., 2017. Early life bisphenol A exposure and neurobehavior at 8 years of Perera, F., Nolte, E.L.R., Wang, Y., Margolis, A.E., Calafat, A.M., Wang, S., Garcia, W., age: identifying windows of heightened vulnerability. Environ. Int. 107, 258–265. Hoepner, L.A., Peterson, B.S., Rauh, V., Herbstman, J., 2016. Bisphenol A exposure Tanner, E.M., Hallerback,¨ M.U., Wikstrom,¨ S., Lindh, C., Kiviranta, H., Gennings, C., and symptoms of anxiety and depression among inner city children at 10–12 years of Bornehag, C.-G., 2020. Early prenatal exposure to suspected endocrine disruptor age. Environ. Res. 151, 195–202. mixtures is associated with lower IQ at age seven. Environ. Int. 134, 105185. Perkins, N.J., Schisterman, E.F., 2006. The inconsistency of “optimal” cutpoints obtained Traccis, F., Frau, R., Melis, M., 2020. Gender Differences in the Outcome of Offspring using two criteria based on the receiver operating characteristic curve. Am. J. Prenatally Exposed to Drugs of. Abuse. 14. Epidemiol. 163, 670–675. van der Elst, W., Ouwehand, C., van Rijn, P., Lee, N., van Boxtel, M., Jolles, J., 2013. The Pop, V.J., Kuijpens, J.L., van Baar, A.L., Verkerk, G., van Son, M.M., de Vijlder, J.J., shortened Raven Standard Progressive Matrices: item response theory-based Vulsma, T., Wiersinga, W.M., Drexhage, H.A., Vader, H.L., 1999. Low maternal free psychometric analyses and normative data. Assessment 20, 48–59. thyroxine concentrations during early pregnancy are associated with impaired Vandenberg, L.N., Colborn, T., Hayes, T.B., Heindel, J.J., Jacobs Jr., D.R., Lee, D.H., psychomotor development in infancy. Clin. Endocrinol. (Oxf.) 50, 149–155. Shioda, T., Soto, A.M., Vom Saal, F.S., Welshons, W.V., Zoeller, R.T., Myers, J.P., Prifitera, A., Saklofske, D.H., Weiss, L.G., 2005. WISC-IV clinical use and interpretation: 2012. Hormones and endocrine-disrupting chemicals: low-dose effects and Scientist-practitioner perspectives. Academic Press. nonmonotonic dose responses. Endocr. Rev. 33, 378–455. Raven, J., Raven, J.C., Court, J.H., 1998. Manual for Raven’s progressive matrices and Wechsler, D., 2003. Wechsler Intelligence Scale for Children, fourth ed. Pearson, San vocabulary scales., Oxford, UK, Oxford Psychologists Press; San Antonio, TX: The Antonio, TX. Psychological Corporation. WHO/UNEP 2013. State of the science of endocrine disrupting chemicals - 2012. Rochester, J.R., Bolden, A.L., 2015. Bisphenol S and F: A Systematic Review and Willoughby, K.A., McAndrews, M.P., Rovet, J.F., 2014. Effects of maternal Comparison of the Hormonal Activity of Bisphenol A Substitutes. Environ. Health hypothyroidism on offspring hippocampus and memory. Thyroid 24, 576–584. Perspect. 123, 643–650. Vogel, S.A., 2009. The politics of plastics: the making and unmaking of bisphenol a Rochester, J.R., Bolden, A.L., Kwiatkowski, C.F., 2018. Prenatal exposure to bisphenol A “safety”. Am. J. Public Health 99 (Suppl 3), S559–S566. and hyperactivity in children: a systematic review and meta-analysis. Environ. Int. Woodruff, T.J., Zota, A.R., Schwartz, J.M., 2011. Environmental chemicals in pregnant 114, 343–356. women in the United States: NHANES 2003–2004. Environ. Health Perspect. 119, Rodriguez-Carrillo, A., Mustieles, V., Perez-Lobato, R., Molina-Molina, J.M., Reina- 878–885. Perez, I., Vela-Soria, F., Rubio, S., Olea, N., Fernandez, M.F., 2019. Bisphenol A and Ye, X., Wong, L.-Y., Kramer, J., Zhou, X., Jia, T., Calafat, A.M., 2015. Urinary cognitive function in school-age boys: Is BPA predominantly related to behavior? Concentrations of Bisphenol A and Three Other Bisphenols in Convenience Samples Neurotoxicology 74, 162–171. of U.S. Adults during 2000–2014. Environ. Sci. Technol. 49, 11834–11839. Rosenfeld, C.S., 2017. Neuroendocrine disruption in animal models due to exposure to Yin, N., Liang, X., Liang, S., Liang, S., Yang, R., Hu, B., Cheng, Z., Liu, S., Dong, H., bisphenol A analogues. Front. Neuroendocrinol. 47, 123–133. Liu, S., 2019. Embryonic stem cell-and transcriptomics-based in vitro analyses reveal Rosenmai, A.K., Dybdahl, M., Pedersen, M., Alice Van Vugt-Lussenburg, B.M., that bisphenols A, F and S have similar and very complex potential developmental Wedebye, E.B., Taxvig, C., Vinggaard, A.M., 2014. Are structural analogues to toxicities. Ecotoxicol. Environ. Saf. 176, 330–338. bisphenol a safe alternatives? Toxicol. Sci. 139, 35–47. Zhang, Y.F., Ren, X.M., Li, Y.Y., Yao, X.F., Li, C.H., Qin, Z.F., Guo, L.H., 2018. Bisphenol ˇ ˇ ˇ Sauer, P., Svecova,´ H., Grabicova,´ K., Gonül¨ Aydın, F., Mackulak, T., Kodeˇs, V., Blytt, L. A alternatives bisphenol S and bisphenol F interfere with thyroid hormone signaling D., Henninge, L.B., Grabic, R., Kocour Kroupova,´ H., 2021. Bisphenols emerging in pathway in vitro and in vivo. Environ. Pollut. 237, 1072–1079. Norwegian and Czech aquatic environments show transthyretin binding potency and Zhu, M., Chen, X.Y., Li, Y.Y., Yin, N.Y., Faiola, F., Qin, Z.F., Wei, W.J., 2018. Bisphenol F other less-studied endocrine-disrupting activities. Sci. Total Environ. 751, 141801. Disrupts Thyroid Hormone Signaling and Postembryonic Development in Xenopus laevis. Environ. Sci. Technol. 52, 1602–1611.

8