Paternal mixtures of urinary concentrations of phthalate metabolites, bisphenol A and parabens in relation to pregnancy outcomes among couples attending a fertility center

Study participants included male partners of couples enrolled in the Environment and Reproductive Health (EARTH) Study, an ongoing prospective cohort of couples seeking infertility treatment at the Massachusetts General Hospital (MGH) Fertility Center aimed at evaluating environmental and dietary determinants of fertility ( Minguez-Alarcon et al. 2016b ). Men between the ages of 18–56 years and without a history of vasectomy were eligible to participate, and approximately 40% of those contacted by the research nurses were enrolled. Women between 18 and 45 years old were eligible to participate and approximately 60% of those contacted by the research staff enrolled. This prospective analysis included 210 couples who underwent at least one in vitro fertilization (IVF) cycle (N=300) between 2004 and 2017, and had a urine sample available that was collected during the treatment cycle to measure biomarker concentrations from the male partner. We excluded couples with no paternal urinary biomarker measurements obtained for the specific IVF cycle, as well as cycles using egg or sperm donors or cryo-thaw, between 2004 and 2017 (N=21 men, 35 IVF cycles). After the study procedures were explained, participants signed an informed consent form. The participant’s date of birth was collected at entry, and weight and height were measured by trained study staff. Body mass index (BMI) was calculated as weight (in kilograms) divided by height (in meters) squared. The participants completed a staff-administered questionnaire that contained additional questions on lifestyle factors, reproductive health, and medical history. The study was approved by the Human Subject Committees of the Harvard T.H. Chan School of Public Health, Partners HealthCare, and the Centers for Disease Control and Prevention (CDC). Men whose female partner was undergoing the IVF cycle provided one spot urine sample at the time of oocyte retrieval during each treatment cycle. Urine was collected in a sterile polypropylene specimen cup. Specific gravity (SG), was measured at room temperature at MGH using a handheld refractometer (National Instrument Company, Inc., Baltimore, MD, USA) calibrated with deionized water before each measurement. The urine was divided into aliquots, frozen, and stored at −80 °C. Samples were shipped overnight on dry ice to the CDC where they were stored at or below −40 °C until analysis. The CDC laboratory is certified by the Health Care Financing Administration to comply with the requirements set forth in the Clinical Laboratory Improvement Amendments of 1988 (CLIA ‘88). All analytical measurements follow strict quality control/quality assurance CLIA guidelines. For example, along with study samples, each analytical run includes a set of calibrators, reagent blanks, and high- and low-concentration quality control (QC) materials. Concentrations of the QCs are evaluated using standard statistical probability rules ( Caudill et al. 2008 ). As previously described ( Silva et al. 2007 ; Ye et al. 2005 ; Ye et al. 2006 ), online solid-phase extraction coupled with isotope dilution-high-performance liquid chromatography-tandem mass spectrometry was used to quantify the urinary concentrations of BPA, methylparaben, propylparaben, and of eight phthalate metabolites [monoethyl phthalate (MEP); mono-n-butyl phthalate (MBP); mono-isobutyl phthalate (MiBP); monobenzyl phthalate (MBzP); mono-2-ethylhexyl phthalate (MEHP); mono-2-ethyl-5-hydroxyhexyl phthalate (MEHHP); mono-2-ethyl-5-oxohexyl phthalate (MEOHP); mono-2-ethyl-5-carboxypentyl phthalate (MECPP)]. The limits of detection (LOD) ranged from 0.1 to 1.2 μg/L. Samples with a biomarker concentration below the LOD were assigned a value equal to the LOD divided by the square root of 2 prior to adjustment by SG as described previously ( Minguez-Alarcon et al. 2016a ). These CDC methods are public ( CDC 2018b , 2019b ) and have been used since the early 2000s for the analyses of tens of thousands of biological specimens, including those collected as part of CDC’s ongoing national survey, the National Health and Nutrition Examination Survey (NHANES). NHANES data have provided the most comprehensive assessment of Americans’ exposure to phthalates and phenols for decades ( CDC 2019a ). To account for urinary dilution, SG was included in the statistical models as a covariate ( Barr et al. 2005 ; Schisterman et al. 2005 ).

Clinical information on IVF outcomes were abstracted from the female partner’s electronic medical record by research staff. Infertility diagnosis was coded according to previously described definitions of the Society for Assisted Reproductive Technology (SART) ( SART 2015 ). Women underwent one of three controlled ovarian stimulation IVF treatment protocols after completing a cycle of oral contraceptives: (1) luteal phase gonadotropin-releasing hormone (GnRH)-agonist protocol, (2) follicular phase GnRH-agonist/Flare protocol, or (3) GnRH-antagonist protocol. Protocols were chosen by the treating physician after taking into consideration several factors as described elsewhere ( Minguez-Alarcon et al. 2016b ). Implantation was defined as a serum β-human chorionic gonadotropin (hCG) level >6mIU/mL, typically measured 17days after oocyte retrieval. An elevation in β-hCG with the confirmation of an intrauterine pregnancy on an ultrasound at 6 weeks was considered a clinical pregnancy. A live birth was defined as the birth of a neonate on or after 24 weeks of gestation. We present distribution of urinary biomarker concentrations using percentiles and geometric means ± standard deviations (SDs). We used Spearman correlation coefficients to assess correlations between urinary biomarker concentrations. To account for the potential correlation and better represent real-life exposures, we evaluated biomarkers of phthalates and phenols as a chemical mixture, using principal component analysis (PCA) in our primary analysis, and later confirmed results with Bayesian Kernel Machine Regression (BKMR) ( Bobb et al. 2015 ; Valeri et al. 2017 ). Due to right skewness, we log e -transformed urinary biomarker concentrations, then we standardized to create z-scores, and variables were included in a PCA with varimax rotation. We selected PCA components with eigenvalues>1 in our analysis ( O’Rourke and Hatcher 2013 ), and derived factor scores as a weighted linear combination of biomarkers using the loading factors. We presented participants’ demographics and couples’ reproductive characteristics using median ± interquartile ranges (IQRs) or counts (%). Multivariable Cox proportional hazards models for discrete survival time were used to estimate the associations between the PCA-derived factor scores and failure to achieve a live birth ( Maity et al. 2014 ; Missmer et al. 2011 ), using a robust sandwich covariance estimate to account for the multiple cycles per couple and adjusting for confounders. Results are presented as hazard ratios (HRs) with the corresponding 95% confidence intervals (CIs). The factor scores were evaluated in the main models continuously in interquartile range (IQR: 25th and 75th percentiles) increments. The HRs estimate the probability of failing a cycle at any point, conditional on not failing at an earlier point during the same cycle. Women were considered to be among those who could fail within a cycle for the duration of their initiated cycle, up until their specific point of failure. Interactions were also included in the models to examine strength of associations over three vulnerable periods: failing between embryo transfer and implantation, between implantation and clinical pregnancy, and between clinical pregnancy and live birth. We next used BKMR, a non-parametric approach for chemical mixtures that incorporates a variable selection component, and flexibly models the joint effect of chemicals with a kernel function ( Bobb et al. 2015 ; Valeri et al. 2017 ). In BKMR analyses we evaluated DEHP metabolites as a single summary measure (ΣDEHP), and obtained the relative importance of each component in the mixture-outcome association by estimating posterior inclusion probabilities (PIP). Each measurement of log e -transformed phthalate metabolite concentrations was treated as a continuous predictor and the other components of the mixture were set at their median while adjusting for confounders. Confounding was assessed using prior knowledge on biological relevance and descriptive statistics from our study participants. The variables considered as potential confounders included factors previously related to pregnancy outcomes ( Rooney and Domar 2014 ; Sharma et al. 2013 ), and factors associated with urinary biomarker concentrations and pregnancy outcomes in this study. Final models were adjusted for paternal and maternal ages and BMI, urine dilution (specific gravity) and year of collection, infertility diagnosis, and other PCA factor scores (not included in the BKMR models). Statistical analyses were performed with SAS (version 9.4; SAS Institute Inc., Cary, NC, USA), with exception of the BKMR models, which were conducted using the R package BKMR ( Bobb et al. 2018 ). To assess the robustness of the findings, we conducted a sensitivity analyses further adjusting for PCA-derived loading scores of maternal urinary BPA, paraben and phthalate metabolite concentrations for this specific subset of IVF couples with available biomarker data ( Minguez-Alarcon et al. 2019 ). Statistical tests were two-tailed and all p-values<0.05 were conventionally regarded as statistically significant, although emphasis was placed on consistency of findings across analyses.

Among these 210 couples enrolled in the EARTH study, detection frequencies for paternal urinary concentrations of BPA, parabens and phthalate metabolites exceeded 89% with the exception of MEHP (73%) ( Table 1 ). Urinary di-2-ethylhexyl phthalate (DEHP) metabolite concentrations (MEHP, MEHHP, MEOHP and MECPP) were highly correlated with each other (r=0.84–0.99), as were MBP and MiBP concentrations (r=0.78) and urinary methyl- and propylparaben concentrations (r=0.78) ( Table S1 ). Urinary concentrations of other biomarkers (BPA, methyl- and propylparaben, MEP and MBzP) were weakly to moderately correlated (r=0.18–0.72) with all other chemicals. Urinary phthalate scorrelations are similar to those observed in other studies among subfertile men in Asia ( Wang et al. 2020 ) and Europe ( Herr et al. 2009 ). All urinary biomarkers concentrations were included in the PCA analysis and three components were identified (eigenvalue>1), which accounted 56%, 15% and 10% of the total variance in urinary phthalate metabolites, BPA, and paraben concentrations, respectively ( Figure S1 ). Factor 1, which we refer to as the DEHP factor, was characterized by high loading scores (>0.45) of urinary concentrations of the four DEHP metabolites measured ( Figure S1 ). Factor 2, referred to as the BPA/non-DEHP factor, had high loading scores (>0.45) of urinary BPA, MBP, MiBP, MEP and MBzP concentrations. Factor 3, referred to as the paraben factor, had high loading scores (>0.45) of urinary methyl- and propylparaben concentrations. Although urinary concentrations of BPA exhibited highest loading scores for Factor 2, it also had a high contribution to Factor 1 (the DEHP factor). Men and their female partners had a median (interquartile range [IQR]) age and BMI of 35.4 (32.6, 39.3) and 34.0 (32.0, 37.5) years, and 26.9 (24.3, 29.2) and 22.8 (20.5, 25.8) kg/m 2 , respectively ( Table 2 ). Couples were predominantly white (88% of the men and 87% of the women), and approximately 30% ever smoked. Clinical pregnancy was observed in 153 (51%) cycles, and 125 (42%) cycles resulted in a live birth. Compared to included couples, those excluded were older and heavier and had less IVF cycle success ( Table S2 ). An interquartile range (25th and 75th percentiles) increase in the DEHP-related factor score was associated with elevated probability of failing prior to live birth (HR=1.41, 95% CI: 1.08, 1.81) ( Figure 1 ) and the association was stronger between implantation and clinical pregnancy (HR=1.57, 95% CI: 1.16, 2.13) as well as between clinical pregnancy and live birth (HR=1.44 95% CI: 1.02, 2.02) compared to before implantation (HR=1.27, 95% CI: 0.93, 1.75) ( Figure 1 ). The overall HRs of failure for the BPA/non-DEHP-related and paraben-related factor scores were HR=1.24 (95% CI: 0.97, 1.59) and HR=0.99 (95% CI: 0.80, 1.24), respectively, and the magnitude of the association was similar across the three cycle-specific time windows. Associations remained but were overall slightly weaker after further adjustment for PCA-derived factor scores of female partner’s urinary BPA, paraben and phthalate concentrations ( Table S3 ), and there were moderate within couple correlations for the DEHP (r=0.43), BPA/non-DEHP (r=0.25) and paraben (r=0.20) factor scores. The variable selection procedure from the BKMR analyses ( Table S4 ) showed that all components of the mixture provided some contribution to the mixture-outcome associations (all PIPs > 0.3), with the ΣDEHP concentrations being the most important contributor to the mixture-outcome association (PIP=0.64). However, when examining individual biomarkers within the context of the mixture, no significant associations were observed (data not shown).

We investigated associations of paternal mixtures of urinary BPA, parabens and phthalate metabolites concentrations with pregnancy outcomes applying PCA and BKMR methods. We identified three factors using PCA: the DEHP factor, characterized by high loading of urinary concentrations of four DEHP metabolites (MEHP, MEHHP, MEOHP and MECPP); the BPA/non-DEHP factor, which had high loading scores of urinary BPA, MBP, MiBP, MEP and MBzP concentrations; and the paraben factor, characterized by high loading scores of urinary methyl- and propylparaben concentrations. We found that paternal mixtures of urinary concentrations of DEHP metabolites were significantly related to higher probability of couples’ infertility treatment failure, even when adjusting for maternal PCA factor scores of urinary biomarker concentrations. Although non-significant, BKMR results were consistent with those observed in the PCA. Urinary concentrations of BPA, parabens and phthalate metabolites were generally higher than those reported in U.S. men from the general population (NHANES), especially for phthalate metabolites ( CDC 2019a ). Compared to men in other countries, urinary concentrations of some of these biomarkers among men in the EARTH Study were overall lower than those found in urines collected among subfertile men in Asia ( Wang et al. 2020 ) as well as Europe ( Herr et al. 2009 ) and Africa ( Omran et al. 2018 ). Although this is the first study investigating paternal mixtures of EDCs in relation to pregnancy outcomes in the EARTH Study, we have previously published evaluations of maternal chemical mixtures in relation to reproductive ( Minguez-Alarcon et al. 2019 ) and perinatal ( Chiu et al. 2018 ) health outcomes in the same study cohort, which also utilized the PCA approach. Interestingly, the PCA identified similar DEHP and non-DEHP factors using the same study cohort in two previous publications on maternal mixtures of BPA, parabens and phthalate metabolites in relation to pregnancy outcomes ( Minguez-Alarcon et al. 2019 ), as well as phthalate metabolites mixtures and birth weight ( Chiu et al. 2018 ), and also in other pregnancy cohorts ( Maresca et al. 2016 ). DEHP is commonly added to plastics to make them flexible and can be found in consumer products, flooring and wall coverings, food contact applications, and medical devices. Metabolites of DEHP include MEHP, MEHHP, MEOHP and MECPP. Other phthalate metabolites such as MBP, MiBP, and MBzP may have common sources and are mainly found in personal care products and other consumer products ( Braun et al. 2014 ; Hauser and Calafat 2005 ). Although BPA highly contributed to the PCA non-DEHP-derived factor, it also contributed to the DEHP factor. This may be explained by the fact that both DEHP and BPA are used in the manufacture of frequently used materials. For example, BPA is used in the manufacture of polycarbonate plastics, epoxy resin liners of canned foods, some dental sealants and composites, and thermal receipts ( Minguez-Alarcon et al. 2016c ). The paraben factor was characterized by relatively high urinary concentrations of both methyl- and propylparabens, as well as MEP. While diethyl phthalate, the parent compound of MEP, can be found in personal care products, parabens are used as preservatives in cosmetics, personal care products, pharmaceuticals, and food ( Andersen 2008 ). Thus, the identified principal components reflected exposure patterns, since PCA converts a set of observed variables into principal components based on the collinearity between the exposure variables rather than the underlying biological effects of a given mixture on the outcomes. Among women undergoing IVF from the same study cohort, we previously reported no significant associations of mixtures of female concentrations of BPA, parabens and phthalate metabolites with implantation, clinical pregnancy and live birth. However, we observed diminished pregnancy success with specific clusters of chemicals among women who underwent IVF cycles in earlier years of study when urinary concentrations of these EDCs were higher ( Minguez-Alarcon et al. 2019 ). While this is the first study considering mixtures when assessing paternal exposure to EDCs in relation to pregnancy outcomes in couples, we previously investigated individual paternal environmental exposures in relation to reproductive outcomes among couples from the same study cohort. Similar to our non-significant results on paternal mixtures of urinary parabens in relation to pregnancy outcomes, we previously found that paternal parabens assessed individually were unrelated to their female partners’ pregnancy outcomes ( Dodge et al. 2015b ). Also among men in the EARTH Study, we previously reported individual associations of increased paternal urinary concentrations of certain phthalate metabolites with decreased odds of implantation and live birth ( Dodge et al. 2015a ) as well as we observed reduced fertilization rate with increased paternal urinary concentrations of select flame retardants that were individually introduced in the statistical models ( Carignan et al. 2018 ). Buck-Louis and colleagues also emphasized the importance of the paternal role when assessing environmental exposures and couples’ reproductive outcomes ( Buck Louis et al. 2016 ). Specifically, they observed that men’s urinary concentrations of select phthalate metabolites were associated with reduced fecundability, measured as longer time to pregnancy, among 501 couples from the general population ( Buck Louis et al. 2014 ). Authors also found similar reduced couple’s fecundability with increased paternal exposure to lead, 2,2′,4,4′-tetrahydroxybenzophenone (a type of ultraviolet filter), and some polychlorinated biphenyls (PCB) congeners measured in serum samples, whereas they did not find any association with paternal exposure to other chemicals such as BPA and cadmium ( Buck Louis et al. 2016 ). The impact of the sperm epigenome, alterations in gene expression in the spermatozoa, by paternal environmental and lifestyle exposures have been proposed as a potential mechanism by which exposure to environmental chemicals in men may affect couple’s reproductive outcomes ( Chapin et al. 2004 ; Chen et al. 2016 ; Kumar et al. 2013 ; Rando 2012; Robinson et al. 2012 ; Siddeek et al. 2018 ). Sperm contributes more than paternal genetic material to the oocyte, sperm also transmits oocyte activation factors, centrosomes, messenger RNA, and microRNA ( Krawetz 2005 ; Kumar et al. 2012 ). The sperm epigenome is essential to embryogenesis, and modifications of sperm epigenetics (gene modifications in the spermatozoa) consequently affect fertilization and embryo development ( Jenkins and Carrell 2011 ). Perturbation of such epigenetic processes can result in reduced fertilization, poor quality embryos, and lower probabilities of live birth, with potential consequences to the offspring phenotype (Rando 2012; Soubry et al. 2014 ; Soubry 2015 ). A cross-sectional study among men undergoing IVF explored paternal urinary phthalate metabolites and sperm DNA methylation ( Wu et al. 2017 ). The authors found that ten phthalate metabolites were associated with 131 sperm differentially methylated regions that were enriched in genes related to growth and development, cell movement and cytoskeleton structure, suggesting that paternal adult environmental conditions may influence epigenetic reprogramming during spermatogenesis, and in turn, influence early-life development. Clearly, new research using mediation analyses needs to be conducted to explore whether the negative associations between paternal mixtures of EDCs and couple’s reproductive endpoints relate to changes in men’s sperm DNA methylation. The current study has several limitations. First, it may not be possible to generalize our findings to couples from the overall population of couples attempting conception. Nevertheless, approximately 15% to 25% of couples trying to achieve pregnancy are diagnosed with infertility ( Slama et al. 2012 ; Thoma et al. 2013 ), which make our study findings applicable to a significant proportion of couples seeking treatment. Second, exposure misclassification is possible given the relatively short biological half-lives of these non-persistent chemicals and the likely episodic nature of their exposures ( Braun et al. 2012 ). Third, residual confounding is still possible because other EDCs, which may be correlated with those included in the current analysis, were not measured or excluded from this analysis because of relatively low sample size or low detection frequencies. Fourth, we also used BKMR to further investigate which biomarkers, within the chemical groups detected from PCA analysis, were driving the observed associations, but we mostly observed null results. This may be due to the complex nature of the data (repeated clustered exposures) that is poorly captured by BKMR in this relatively small sample. It is warranted to replicate these analyses in a larger study. The biggest strength of this study is our ability to evaluate the exposure biomarkers in relation to reproductive endpoints as a chemical mixture, an approach that provides a better description of real-world multiple exposures, and accommodates confounding effects and possible correlations within the exposures of interest. Other strengths include the prospective design, complete follow-up of participants, which minimizes the risk of reverse causation; and comprehensive adjustment for other reproductive and lifestyle factors that could result in residual confounding. To conclude, mixtures of paternal urinary concentrations of DEHP metabolites were associated with higher couples’ infertility treatment failure highlighting the importance of exploring the paternal role when investigating environmental exposures on pregnancy endpoints. The association remained after additionally adjusting for maternal mixtures of urinary biomarker concentrations. Since this is the first study exploring paternal environmental mixtures on couples’ reproductive outcomes, other studies confirming these results and studying potential mechanisms that explain the associations are warranted.

Infertility, the inability to achieve a pregnancy after several months of attempted conception, is estimated to affect about 10–15% of all couples worldwide ( Inhorn and Patrizio 2015 ; Mascarenhas et al. 2012 ) and male factor, defined according to World Health Organization (WHO) reference values for semen quality, is the underlying or contributing cause of infertility in around 40% of the cases ( Skakkebaek et al. 2016 ). Despite the increasing use of medical treatment among couples with decreased fecundity ( CDC 2018a ), success rates of live birth among couples undergoing assisted reproductive technology (ART) treatments have remained stable ( CDC 2018a ). Thus, identifying modifiable factors, such as environmental exposures, that can predict human fertility has become a major clinical and public health concern. Bisphenol A (BPA), parabens and phthalates are chemicals with endocrine disrupting activities which can be found in numerous consumer and personal care products ( Diamanti-Kandarakis et al. 2009 ; WHO 2012 ) leading to widespread general population exposure ( CDC 2019a ). Urinary concentrations of some of these endocrine disrupting chemicals (EDCs) have been associated with poor semen quality in epidemiologic studies ( Jurewicz et al. 2017 ; Minguez-Alarcon et al. 2016c ; Radke et al. 2018 ). Interestingly, we previously showed that urinary concentrations of some phthalates attenuated the observed decline in semen quality in men attending the Massachusetts Hospital Fertility Center between 2000 and 2017 ( Minguez-Alarcon et al. 2018 ). Whether the potential effect of these chemicals on decreasing semen quality has any consequences on the female partner’s pregnancy outcomes is understudied. This is particularly important because a growing literature suggests that paternal environmental and lifestyle exposures may affect the health of offspring via the sperm epigenome ( Chapin et al. 2004 ; Chen et al. 2016 ; Kumar et al. 2013 ; Rando 2012; Robinson et al. 2012 ; Siddeek et al. 2018 ). The sperm epigenome assesses alterations in gene expression in the spermatozoa and is integral to embryogenesis, and sperm epigenetic modifications (gene modifications in the spermatozoa) play a critical role in fertilization potential and embryo development ( Jenkins and Carrell 2011 ). Alteration of such epigenetic processes may result in reduced embryonic development, fertilization, and pregnancy, consequently influencing offspring phenotype (Rando 2012; Soubry et al. 2014 ; Soubry 2015 ). Although our study was not designed to identify biological mechanisms through which paternal exposure to endocrine disruptors may alter couples’ pregnancy outcomes (e.g. via sperm epigenetics), it can help clarify the role of paternal environmental exposures on couples’ reproductive health. Thus, we investigated whether paternal urinary BPA, parabens, and phthalates biomarkers were associated with pregnancy outcomes among couples attending a fertility center. Urinary biomarker concentrations were evaluated as a mixture because people are likely to be exposed to multiple chemicals ( Birnbaum 2012 ; Bobb et al. 2015 ). Using the same study cohort, we have previously published on maternal ( Hauser et al. 2016 ; Minguez-Alarcon et al. 2015 ; Minguez-Alarcon et al. 2016a ) and paternal ( Dodge et al. 2015a ; Dodge et al. 2015b ) urinary concentrations of BPA, parabens and phthalate biomarkers, assessed individually, in relation to pregnancy outcomes.