Parental Preconception Exposure to Phenol and Phthalate Mixtures and the Risk of Preterm Birth

Abstract

Background:

Parental preconception exposure to select phenols and phthalates was previously associated with increased risk of preterm birth in single chemical analyses. However, the joint effect of phenol and phthalate mixtures on preterm birth is unknown.

Methods:

We included 384 female and 211 male (203 couples) participants seeking infertility treatment in the Environment and Reproductive Health (EARTH) Study who gave birth to 384 singleton infants between 2005 and 2018. Mean preconception urinary concentrations of bisphenol A (BPA), parabens, and eleven phthalate biomarkers, including di(2-ethylhexyl) phthalate (DEHP) metabolites, were examined. We used principal component analysis (PCA) with log-Poisson regression and Probit Bayesian Kernel Machine Regression (BKMR) with hierarchical variable selection to examine maternal and paternal phenol and phthalate mixtures in relation to preterm birth. Couple-based BKMR model was fit to assess couples’ joint mixtures in relation to preterm birth.

Results:

PCA identified the same four factors for maternal and paternal preconception mixtures. Each unit increase in PCA scores of maternal (adjusted Risk Ratio (aRR): 1.36, 95%CI: 1.00, 1.84) and paternal (aRR: 1.47, 95%CI: 0.90, 2.42) preconception DEHP-BPA factor was positively associated with preterm birth. Maternal and paternal BKMR models consistently presented the DEHP-BPA factor with the highest group Posterior Inclusion Probability (PIP). BKMR models further showed that maternal preconception BPA and mono(2-ethyl-5-hydroxyhexyl) phthalate, and paternal preconception mono(2-ethylhexyl) phthalate were positively associated with preterm birth when the remaining mixture components were held at their median concentrations. Couple-based BKMR models showed a similar relative contribution of paternal (PIP: 61%) and maternal (PIP: 77%) preconception mixtures on preterm birth. We found a positive joint effect on preterm birth across increasing quantiles of couples’ total mixture concentrations.

Conclusion:

In this prospective cohort of subfertile couples, maternal BPA and DEHP, and paternal DEHP exposure before conception were positively associated with preterm birth. Both parental windows jointly contributed to the outcome. These results suggest that preterm birth may be a couple-based pregnancy outcome.

Graphical Abstract

1. Introduction

Preterm birth, defined as birth before 37 completed weeks of gestation, is an important predictor of child health and development [1–4]. Although some risk factors have been identified, such as maternal age, race, socio-economic status, intrauterine infection, and multiple gestations, the underlying causes remain largely unknown [5–7]. Despite efforts to reduce known risk factors, a sustained reduction in the rate of preterm birth in the United States has yet to be achieved [8]. Growing evidence has linked environmental exposures, including non-persistent chemicals, to preterm birth [9–16].

Endocrine disrupting chemicals (EDCs) are exogenous chemicals that can interfere with any aspect of hormone action [17]. Phthalates and phenols are among the most studied EDCs due to concerns of their reproductive and developmental toxicity [18–29]. Humans are widely exposed to phthalates and phenols from use of plastics, food packaging materials, personal care products, and numerous other everyday consumer products [18, 30–36].

Exposures to some phthalates and phenols in the prenatal period are associated with poor pregnancy outcomes, including preterm birth [9–15]. Our previous work identified preconception as a susceptible period for the potential impact of phenols and phthalates on adverse birth outcomes [37–41]. Environmentally-driven epigenetic mechanisms in gametes during the pre- and peri-conception period likely contribute to the etiopathology of adverse pregnancy outcomes, and are considered a plausible mode of action for environmental chemicals [42–46].

One of the current challenges in the field of toxicology and environmental epidemiology is the study of chemical mixtures in human populations [47]. Thus, an increasing number of studies is exploring the effects of chemical mixtures on diverse health outcomes [48–51], highlighting the need to implement multipollutant models in environmental epidemiology [52–54]. Given that mixtures of environmental chemicals may interact with each other leading to additive, synergistic and/or antagonistic effects, the traditional one-chemical-at-a-time strategy may not represent real-world exposure scenarios in which humans are exposed to many chemical families [55–59]. Select phthalates and phenols share common exposure sources including diet (e.g., packaged food), cosmetics and other consumer products, potentially resulting in substantial co-exposure to both families of compounds in human populations [60–63], and possibly leading to mixture interactions as supported by toxicological data [64–66].

Using single chemical analyses, we previously reported that maternal and paternal preconception exposure to some phthalates and phenols were associated with an elevated risk of preterm birth [39, 40]. Given their simultaneous co-occurrence in real-life scenarios [67], the present study aimed to examine whether maternal and paternal preconception urinary concentrations of complex mixtures of phenols and phthalate metabolites interact to influence the risk of singleton preterm birth among couples attending a fertility clinic.

2. Materials and Methods

2.1. Study cohort

In this present analysis, we used data from the Environment and Reproductive Health (EARTH) Study. The EARTH Study is a prospective preconception cohort of couples who were recruited from the Massachusetts General Hospital (MGH) Fertility Center starting in 2005 up until 2019. The EARTH Study investigates male and female environmental and nutritional factors and their relationship with fertility, pregnancy, and birth outcomes [68]. Details of the study can be found elsewhere [68]. In brief, men (18 to 55 years) and women (18 to 46 years) were eligible to participate either independently or as a couple. As couples undergo fertility care, they are followed at each cycle of attempted conception while under treatment observation. During this time, the EARTH Study follows and collects data on health and lifestyle information. Participants also provide urine and blood samples during follow-up.

The present study included 384 mothers and 211 fathers (203 couples) who gave birth to a singleton between 2005 and 2018 for whom we quantified phenol and phthalate biomarkers in at least one urine sample collected before conception of the index pregnancy. Trained study staff explained the study details and answered any questions before obtaining participant’s signed informed consents. The study was approved by the Institutional Review Boards of MGH, Harvard T.H. Chan School of Public Health, and the Centers for Disease Control and Prevention (CDC).

2.2. Exposure Assessment

Exposure assessment was based on spot urine samples provided by women first at study entry and then from up to two additional spot urine samples obtained during each fertility treatment cycle, corresponding with the early follicular phase. Together, these samples represented the preconception period and included all urine samples collected at baseline entry and at each cycle of attempted pregnancy, up to and including, the cycle of conception of the index birth. Correspondingly, men provided a spot urine sample at study entry and then an additional spot urine sample at every cycle at time of the fertility procedure.

Polypropylene specimen cups were used to collect urine samples. The urinary specific gravity (SG) was measured by a handheld refractometer (National Instrument Company, Inc., Baltimore, MD, USA). Samples were then divided into aliquots, and frozen for long-term storage at −80 °C. Samples were shipped to the CDC (Atlanta, GA, USA) on dry ice overnight for quantification of urinary phenol and phthalate metabolite concentrations using solid phase extraction coupled online with high performance liquid chromatography-isotope dilution tandem mass spectrometry [69]. Urinary concentrations were measured for four phenols, including bisphenol A (BPA), methylparaben, propylparaben, butylparaben, and eleven phthalate metabolites, including monoethyl phthalate (MEP); mono-n-butyl phthalate (MBP); mono-isobutyl phthalate (MiBP); monobenzyl phthalate (MBzP); mono(3-carboxypropyl) phthalate (MCPP); monocarboxyisooctyl phthalate (MCOP); monocarboxyisononyl phthalate (MCNP); and four di(2-ethylhexyl) phthalate (DEHP) metabolites [mono(2-ethylhexyl) phthalate (MEHP), mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethyl-5-oxohexyl) phthalate (MEOHP) and mono(2-ethyl-5-carboxypentyl) phthalate (MECPP)]. The limits of detection (LOD) ranged from 0.1 to 1.2 ng/ml. Concentrations below the LOD were assigned the LOD divided by the square root of two [70].

2.3. Outcome Assessment

We abstracted gestational age (days) from delivery records. In order to validate the delivery records estimates, we used guidelines for dating births following medically assisted reproduction from the American College of Obstetricians and Gynecologists (ACOG) [71]. For pregnancies conceived using in-vitro fertilization (IVF), gestational age was estimated as (outcome date – embryo transfer date +14 days + cycle day of transfer) [71]. For pregnancies using other forms of medically assisted reproduction or that were naturally conceived, gestational age was estimated as (birth date – cycle start date or last menstrual period date). All live births ending before 37 completed weeks’ gestation (<259 days) were considered preterm.

2.4. Covariates

Paternal and maternal age, education, race, and smoking status were self-reported via questionnaire at study entry. Height and weight of participants were measured at baseline by study staff. Body Mass Index (BMI) was calculated as kg/m². The treating infertility physician diagnosed the underlying cause of infertility using the Society for Assisted Reproductive Technology (SART) definitions [72, 73]. Data from electronic medical records were abstracted for mode of conception of the index birth and was dichotomized as any assisted reproductive technology (ART) procedures (i.e., all IVF protocols) versus non-ART protocols (i.e., all intrauterine inseminations, ovarian stimulation, or natural/non-medically assisted conceptions).

2.5. Statistical Analysis

Each biomarker concentration was multiplied by [(SGp-1)/(SGi-1)] in order to account for urinary dilution, where SGi is the SG of the participant’s sample and SGp is the mean SG for all male (mean=1.016) or all female (mean=1.015) participants included in the study [74]. Because the SG-adjusted biomarker concentrations were skewed, we natural log-transformed values to standardize the distribution and reduce the influence of extreme values. Given the short half-lives of phenols and phthalates, we averaged the natural log-transformed biomarker concentrations of multiple urine samples collected from study entry up to and including the sample from the cycle of conception of the index pregnancy per participant to estimate mean exposure in the preconception window. This strategy allows an improved exposure characterization, lowering the probability of attenuation bias, that is, the tendency towards null findings [75].

We calculated descriptive statistics for biomarker concentrations and the percentage of values above the LOD. We also calculated Spearman correlation coefficients between biomarkers for maternal and paternal preconception windows respectively, as well as for biomarkers within couples. We examined the clinical and demographic characteristics of study participants and their singletons, and reported them as mean (± SD) or number (%).

We first used principal component analysis (PCA) with varimax rotation to classify phenol and phthalate biomarkers into uncorrelated components based on their correlations. This method reduces the number of components while retaining information from the original variables. Factors with eigenvalues greater than one were identified as principal components [76]. We then fit modified Poisson regression models [77] for the principal component scores and the binary preterm birth outcome to estimate risk ratio (RR) and 95% confidence interval (95% CI) per unit increase in the factor scores adjusting for the other factor scores. To make the results more intuitive and allow for nonlinearity, we fit Poisson regression across quartiles of PCA-derived factor scores. We conducted PCA and fit Poisson regression models separately for both maternal and paternal preconception biomarkers.

We then employed Bayesian Kernel Machine Regression (BKMR) with a probit kernel exposure–response function to separately examine maternal and paternal preconception phenol and phthalate mixtures in relation to preterm birth. We used BKMR to flexibly model the joint mixture effect, accounting for correlations and interactions between mixture components [78]. BKMR further allows the visualization of the exposure-response association for each biomarker within the mixture as well as the cumulative effects of the total mixture. We used BKMR with a hierarchical variable selection based on components obtained from PCA results (i.e., we grouped biomarkers according to the components identified by PCA). We calculated the group posterior inclusion probability (groupPIP) and conditional posterior inclusion probability (condPIP), with the former representing the probability of including a particular biomarker group within the model and the latter representing the probability that a biomarker is included within its principal component.

Results from BKMR are presented as 1) univariate exposure-response association of each individual biomarker concentration in relation to preterm birth estimate when holding the remaining biomarkers in the mixture at their median concentrations; 2) potential interactions within mixtures by estimating the change in preterm birth estimate comparing each individual biomarker concentration at 25^th to 75^th percentiles, when setting the remaining biomarkers at their 25^th, 50^th or 75th percentile levels; 3) the cumulative effect of the total mixture on preterm birth adjusting for covariates. The cumulative effect of the total mixture was displayed as the change in preterm birth estimate comparing all biomarkers at their median concentrations (reference) to the concentrations at each 5^th percentile, from the 25^th up to the 75^th percentile.

We selected a priori covariates as potential confounders based on substantive knowledge using a directed acyclic graph (DAG). Maternal age and BMI (continuous), education (<college, college, graduate degree), smoking status (never smoked versus ever (former or current)), race (Caucasian, Black/African American, Asian, Other) and ART versus non-ART mode of conception, were included in all models. We included the corresponding paternal covariates in addition to maternal ones in models examining the paternal exposure windows [81]. Statistical analyses were conducted with SAS (version 9.4; SAS Institute Inc., Cary, USA) and R package bkmr [82]. The interpretation of our results considered patterns of associations in light of previous epidemiologic findings rather than relying solely on p-values and their significance [83].

2.6. Sensitivity Analysis

To examine the association of couples’ exposure to mixtures of phenols and phthalates in relation to the risk of preterm birth, BKMR models were also fit to the 203 couples in the cohort. We separated mixtures into maternal and paternal preconception groups and used hierarchical variable selection to compare the parent-of-origin contribution to the study outcome [84]. We also restricted the maternal preconception analyses to 203 mothers who participated in the study with their partner in order to obtain more comparable results with the couple-based model.

3. Results

3.1. Study cohort

The study cohort comprised 384 mothers and 211 fathers (203 couples) with a mean age of 34.6 and 35.8 years and a mean BMI of 24.0 and 27.9 kg/m², respectively, at study entry (Table 1). Among the 384 singleton infants, mean (SD) gestational age was 39.3 (1.7) weeks and 8% (n=30) of infants were born preterm (Table 2). Background characteristics of mothers who enrolled as a couple (n=203) compared with those who enrolled without a partner (n=181) are presented in Table S1. Mothers who enrolled with their male partner were more likely to have a male factor infertility diagnosis at baseline (28% vs. 18%). They also had slightly higher geometric mean BPA and the sum of DEHP metabolite concentrations compared with those who enrolled without their male partner (BPA: 1.12 ng/mL vs. 1.06 ng/mL; sum of DEHP metabolites: 40.83 ng/mL vs. 37.78 ng/mL).

Table 1.

Parental characteristics from 384 mothers and 211 fathers participating in the Environment and Reproductive Health (EARTH) Study, 2005 – 2018.

Parental Characteristic	Maternal N = 384	Paternal N = 211
Age (years)
Mean (SD)	34.6 (4.0)	35.8 (4.6)
Age>35, n (%)	157 (41)	114 (54)
Race, n (%)
White	323 (84)	186 (88)
Black	10 (3)	4 (2)
Asian	35 (9)	14 (7)
Other	16 (4)	7 (3)
Body Mass Index (BMI, Kg/m2)
Mean (SD)	24.0 (4.1)	27.9 (6.3)
BMI >25, n (%)	120 (31)	144 (68)
Education, n (%)
< College	50 (13)	71 (34)
College Graduate	121 (32)	58 (28)
Graduate Degree	213 (55)	78 (38)
Smoking Status, n (%)
Never	289 (75)	145 (69)
Ever (former or current)	95 (25)	66 (31)
Infertility Diagnosis, n (%)
Male Factor	90 (23)	64 (30)
Female Factor	122 (32)	61 (29)
Unexplained	172 (45)	85 (41)
Primiparous, n (%)	320 (83)

Table 2.

Birth characteristics of 384 singletons from the Environment and Reproductive Health (EARTH) Study, 2005 – 2018.

Infant Characteristics	Births 2005 – 2018
	N = 384 a	N=203 b
Male, n (%)	195 (51)	100 (49)
Birth weight (grams)
Mean (SD)	3357 (536)	3353 (497)
min-max	1310 – 4790	1850 – 4790
Low birth weight
<2500 g, n (%)	18 (5)	6 (3)
Gestational age (weeks)
Mean (SD)	39.3 (1.7)	39.3 (1.5)
min-max	29–42	33–42
Preterm birth
<37 weeks, n (%)	30 (8)	14 (7)
Mode of conception, n (%)
ART c	223 (58)	126 (62)
Non-ART d	161 (42)	77 (38)

^aSingletons from total study population.

^bSingletons whose parents joined as a couple in the study.

^cAssisted Reproductive Technology (ART): fresh or frozen in-vitro fertilization protocols, including intracytoplasmic sperm injection.

^dNon-ART: intrauterine insemination with or without ovulation induction/stimulation; ovulation induction/stimulation with timed intercourse, or non-medically assisted/naturally conceived.

3.2. Urinary biomarker concentrations

We included a total of 1600 maternal and 557 paternal urine samples collected from study entry to the conception of the index pregnancy. Women provided on average 4.2 (median: 3; IQR: 2–5) and men on average of 2.6 (median: 2; IQR: 2–3) urine samples. The median time lag between urine sample collection and the conception date of the index pregnancy was 88 days for both maternal and paternal samples. The distribution of biomarkers and detection frequencies are presented in Table S2. Butylparaben had the lowest urinary detection frequency (62.8% for maternal preconception and 32.1% for paternal preconception). The remaining biomarkers had detection frequencies greater than or equal to 70%. Spearman correlation coefficient matrices are shown in Figure S1, S2 and S3 for maternal and paternal preconception, and couples’ biomarker concentrations, respectively. Spearman correlation coefficients ranged from −0.15 (MEHP and MCOP) to 0.98 (MEOHP and MEHHP) for the maternal preconception window (Figure S1), −0.06 (MEHP and MCOP) to 0.98 (MEOHP and MEHHP) for paternal preconception window (Figure S2), and −0.32 (maternal MEHHP and paternal MCOP) to 0.60 (maternal MECPP and paternal MECPP) for couples’ biomarker concentrations (Figure S3).

3.3. Maternal preconception mixtures

The maternal preconception PCA identified four principal factors (Table S3). Factor 1 was termed the DEHP-BPA factor, accounting for 33.2% of the total mixture variance, and showed high loading scores for MEOHP, MEHHP, MECPP, MEHP (metabolites of DEHP) and BPA. Factor 2 was termed the paraben factor, accounted for 16.1% of the total variance, and showed high loading scores for methylparaben, propylparaben, butylparaben (the three paraben biomarkers) and MEP. Factor 3 was termed the high molecular weight phthalate (HMWP) factor, accounted for 13.4% of the total variance, and showed high loading scores for MCOP, MCPP and MCNP. Factor 4 accounted for 10.2% of the total variance, and showed high loading scores for MiBP, MBP and MBzP. For simplicity, we termed factor 4 as low molecular weight phthalate (LMWP) factor although MBzP is a high molecular weight phthalate metabolite.

Covariate-adjusted modified Poisson regression models showed that each 1-unit increase in the maternal preconception DEHP-BPA factor score was positively associated with risk of preterm birth (RR 1.36; 95%CI: 1.00, 1.84), while no association was found for the other three factors (Table 3). Quartile analyses showed an elevated risk of preterm birth comparing the fourth quartile to the first quartile of maternal DEHP-BPA factor score, adjusting for covariates (RR 2.82; 95% CI: 0.97, 8.24) (Table S4).

Table 3.

Adjusted risk ratio (95% CI) of preterm birth per 1-unit increase in PCA-derived factor scores from 384 mothers and 211 fathers in the Environment and Reproductive Health (EARTH) Study, 2005–2018.

PCA-derived factors	Maternal		Paternal
	Risk Ratio (95% CI) a	P values	Risk Ratio (95% CI) b	P values
DEHP-BPA factor	1.36 (1.00, 1.84)	0.05	1.47 (0.90, 2.42)	0.13
Paraben factor	0.93 (0.65, 1.32)	0.68	1.43 (0.86, 2.38)	0.17
High molecular weight phthalate factor	0.88 (0.61, 1.26)	0.49	0.67 (0.38, 1.17)	0.16
Low molecular weight phthalate factor	0.96 (0.65, 1.41)	0.82	0.89 (0.51, 1.52)	0.66

Note: PCA, Principal component analysis; di-(2-ethylhexyl) phthalate (DEHP). All four PCA-derived factors were mutually adjusted for in the maternal/paternal models.

^aAdjusted for age (continuous), BMI (continuous), ART (yes/no), smoking (ever/never), education (categorical), race (categorical).

^bAdjusted for maternal and paternal age (continuous), maternal and paternal BMI (continuous), maternal and paternal smoking (ever/never), maternal education (categorical), maternal race (categorical), ART (yes/no).

Probit BKMR models for the maternal preconception biomarkers showed that the DEHP-BPA factor had the highest group posterior inclusion probability (p=0.65). Within the DEHP-BPA factor group, BPA produced the highest inclusion probability (p=0.73) (Table S5). The univariate exposure-response analysis found maternal preconception BPA concentration was positively associated with preterm birth estimate, when the remaining biomarkers were held at their median concentrations (Figure 1). No other discernible patterns were found for the other biomarkers examined (Figure 1). In examining the figures for potential interactions, we found no apparent differences in the associations between individual biomarkers at the 25^th versus 75^th percentiles and preterm birth estimates, comparing the results when concentrations of the remaining maternal biomarkers were set at their 25^th, 50^th, and 75^th percentile concentrations (Figure S4). Furthermore, we did not identify a clear trend of the cumulative effect of the total maternal preconception mixture on preterm birth estimate (Figure S5).

Figure 1.

Univariate dose-response association (estimates and credible intervals) of maternal preconception phenol and phthalate metabolite concentrations on preterm birth estimate, holding all other biomarkers at their median concentrations among 384 mothers in the Environment and Reproductive Health (EARTH) Study, 2005–2018.

Note: Bisphenol A (BPA); methylparaben; propylparaben; butylparaben; 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); mono(3-carboxypropyl) phthalate (MCPP); monocarboxyisooctyl phthalate (MCOP); monocarboxyisononyl phthalate (MCNP); models were adjusted for maternal age (continuous), BMI (continuous), ART (yes/no), smoking (ever/never), education (categorical), race (categorical).

3.4. Paternal preconception mixtures

The paternal preconception PCA identified the same four factors as the maternal preconception analysis, but with notable differences in the percent of total mixture variance explained by each factor (Table S6). The DEHP-BPA factor still accounted for the highest of the total variance (33.5%) with the HMWP, LMWP and paraben factors accounting for 15.4%, 13.9%, and 10.0% of the total variance, respectively. Poisson regression models showed that each 1-unit increase in the paternal preconception DEHP-BPA factor score was positively associated with preterm birth in the unadjusted model (RR 1.51; 95%CI: 0.98, 2.34), with modest attenuation after covariate adjustment (RR 1.47; 95%CI: 0.90, 2.42) (Table 3). A quartile-based analysis was not possible for paternal preconception DEHP-BPA factor scores as there were no preterm births within the first quartile (Table S4). We furthermore found a suggestive positive association between the paternal paraben factor and preterm birth (RR 1.43; 95%CI: 0.86, 2.38), but no obvious pattern was found between quartiles of paternal paraben factor scores and preterm birth risk (Table S7). Paternal preconception HMWP factor showed a potential negative association with preterm birth although the confidence interval was relatively wide (RR 0.67; 95% CI: 0.38, 1.17) (Table 3). No association was observed for paternal preconception LMWP factor and preterm birth (Table 3).

Similar to the maternal group, paternal preconception probit BKMR models found that the DEHP-BPA factor also had the highest group inclusion probability (p=0.67). However, for the paternal window, MEHP showed the highest inclusion probability within the DEHP-BPA factor (p=0.48) (Table S8). We observed a non-linear positive trend between paternal preconception MEHP concentrations and preterm birth estimate, when all other paternal biomarkers were held at their median concentrations (Figure 2). Negative trends were observed between paternal preconception MiBP and MCPP concentrations and preterm birth estimates, when all other paternal biomarkers were held at their median concentrations (Figure 2). No apparent interactions were found for any paternal individual biomarker with the remaining paternal biomarkers on preterm birth estimates (Figure S6). Lastly, there was no detectable trend in the cumulative effect of the total paternal preconception mixtures on preterm birth estimate (Figure 4).

Figure 2.

Univariate dose-response association (estimates and credible intervals) of paternal preconception phenol and phthalate metabolite concentrations on preterm birth estimate, holding all other biomarkers at their median concentrations among 211 fathers in the Environment and Reproductive Health (EARTH) Study, 2005–2018.

Note: Bisphenol A (BPA); methylparaben; propylparaben; butylparaben; 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); mono(3-carboxypropyl) phthalate (MCPP); monocarboxyisooctyl phthalate (MCOP); monocarboxyisononyl phthalate (MCNP); models were adjusted for maternal and paternal age (continuous), maternal and paternal BMI (continuous), maternal and paternal smoking (ever/never), maternal education (categorical), maternal race (categorical), ART (yes/no).

Figure 4.

Cumulative effect (estimates and credible intervals) across per 5^th quantile above and below medians of maternal, paternal and couples’ total preconception mixture on preterm birth estimate in the Environment and Reproductive Health (EARTH) Study, 2005–2018.

Note: Maternal model was restricted to women who participated as couples (n=203), paternal model included all fathers (n=211, i.e. 203 who participated as couples and 8 who participated independently), couples model included the 203 couples; reference group is the median concentration of the respective mixture; models for maternal and couples’ mixtures were adjusted for maternal age (continuous), BMI (continuous), ART (yes/no), smoking (ever/never), education (categorical), race (categorical); models for paternal mixtures were adjusted for maternal and paternal age (continuous), maternal and paternal BMI (continuous), maternal and paternal smoking (ever/never), maternal education (categorical), maternal race (categorical), ART (yes/no).

3.5. Couples’ preconception mixtures

In the probit BKMR model of couples (n=203), the inclusion probability of maternal (p=0.77) and paternal preconception mixtures (p=0.61) were similar, suggesting a comparable contribution of exposure from each parent to the outcome (Table S9). Within the paternal preconception mixture, MEHP (p=0.21) and MCPP (p=0.20) had the highest inclusion probabilities, while MEHHP (p=0.36) had the highest inclusion probability within the maternal preconception mixture (Table S9). We found a positive trend between maternal preconception MEHHP and preterm birth estimates when holding the remaining maternal and paternal biomarkers at their median concentrations (Figure 3). We observed a non-linear positive trend between paternal preconception MEHP concentrations and preterm birth estimate, when all the remaining paternal and maternal biomarkers were held at their median concentrations (Figure 3). No interactions were visible between couples’ individual biomarker and the other biomarker concentrations on preterm birth estimate (Figure S7). We did, however, find a positive trend of the cumulative effect of couples’ total phenol and phthalate mixture on preterm birth estimate (Figure 4).

Figure 3.

Univariate dose-response association (estimates and credible intervals) of couples’ preconception phenols and phthalates metabolite concentrations on preterm birth estimate, holding all other biomarkers at their median concentrations among 203 couples in the Environment and Reproductive Health (EARTH) Study, 2005–2018.

Note: Bisphenol A (BPA); methylparaben; propylparaben; butylparaben; 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); mono(3-carboxypropyl) phthalate (MCPP); monocarboxyisooctyl phthalate (MCOP); monocarboxyisononyl phthalate (MCNP);models were adjusted for maternal age (continuous), BMI (continuous), ART (yes/no), smoking (ever/never), education (categorical), race (categorical).

In maternal models restricted to the 203 mothers who enrolled with their male partners, we found an increased risk of preterm birth in relation to the DEHP-BPA factor (RR 1.80; 95% CI: 1.12, 2.88, per 1-unit increase in factor score) after adjusting for covariates (Table S10). In BKMR models, the DEHP-BPA factor had the highest group posterior inclusion probability (p=0.86), but MEHHP and MEOHP – and not BPA – accounted for the highest conditional posterior inclusion probability within the DEHP-BPA group (p=0.37) (Table S11). BKMR showed a positive trend between MEHHP and preterm birth estimate when the remaining maternal preconception biomarkers were held at their median concentrations among this restricted sample of women (Figure S8). No trend was observed between maternal preconception BPA and preterm birth estimate in the restricted model (Figure S8). Additionally, we found a significant positive cumulative effect of maternal preconception total mixtures assessed through quantiles on preterm birth estimate (Figure 4).

4. Discussion

Our findings showed that mixtures containing DEHP metabolites and BPA as the main components were associated with an increased risk of preterm birth in both maternal and paternal preconception windows based on PCA results. We furthermore found a suggestive but less robust association between the paternal paraben factor and preterm birth. BKMR models provided complementary information on these associations in the context of mixtures. In the maternal preconception mixture analysis, BPA and MEHHP were identified as the strongest contributors to the risk of preterm birth when the remaining biomarkers were held at their median concentrations. In the paternal mixture analysis, MEHP was identified as the strongest contributor to preterm birth risk, showing a non-linear positive trend when all other paternal biomarkers were held at their median concentrations. No interactions within the total mixture in either maternal or paternal preconception windows were found. In sensitivity analyses restricted to couples, hierarchical variable selection coupled to BKMR models showed a similar relative contribution of paternal and maternal preconception biomarkers on the risk of preterm birth. A positive dose-response cumulative effect was observed between couple’s preconception exposure to mixtures and preterm birth risk estimates.

In the absence of other preconception cohorts with which to compare our results, the epidemiologic interpretation of the present mixture findings is based on preterm birth results from this same cohort examining maternal and paternal preconception exposure to phenols and phthalates using conventional single-pollutant analyses [39, 40]. We previously observed that maternal preconception exposure to BPA was associated with preterm birth, while paternal paraben exposure also pointed to an increased risk [39]. Regarding phthalates, single chemical analyses showed that both maternal and paternal preconception concentrations of DEHP metabolites were associated with increased risk of preterm birth. However, maternal DEHP exposure appeared to account for the association found among fathers’ exposure [40]. In single-pollutant analysis, we observed imprecise negative associations for two paternal preconception high molecular weight phthalates (MCPP and MCOP) and preterm birth [40]. However, no biologically plausible evidence exists to support this apparent protective finding. Given the imprecise estimate and limited plausible biological evidence, unmeasured confounding by lifestyles might be responsible for the observed seemingly protective association between paternal preconception HMWP factor and preterm birth, which limited further investigation.

In the maternal preconception BKMR model including the total cohort (i.e. 384 mothers), we found that BPA was positively associated with preterm birth (when holding all other biomarkers at their median concentrations), while there was no clear association between the total maternal mixture and preterm birth. However, among the 203 mothers who enrolled with their male partner, BKMR models showed MEHHP as the strongest contributor to preterm birth. Here, we found a positive association between the total maternal preconception mixture and preterm birth risk estimates. It is worth noting that the PCA results were largely consistent in both analyses, pointing to the DEHP-BPA factor as the main component associated with preterm birth. Women who enrolled as couples had slightly higher DEHP metabolites and BPA concentrations than those who did not, and their male partners showed a higher proportion of male infertility diagnosis. Any inconsistencies between the two models might suggest different chemical interplays (i.e. different cumulative effects) when DEHP metabolites and BPA were present at different concentrations, and/or different interplays between the effect of chemicals on progenitors based on slight variations in the proportion of parent-of-origin infertility diagnosis. Nevertheless, both BPA and MEHHP were identified as the chemicals of greatest concern among the phenols and phthalates examined in the maternal preconception window.

Together with the results obtained in single-chemical analyses for phenol [39] and phthalate [40] biomarkers, the current PCA and BKMR results support the following: 1) maternal preconception exposure to both BPA and DEHP appears to be associated with higher preterm birth risk; 2) paternal preconception concentrations of MEHP (a DEHP metabolite), but not BPA, appears to be associated with increased preterm birth; 3) a complex biological interplay seems to exist between maternal and paternal preconception exposures, with both progenitors contributing to preterm birth; 4) a cumulative joint effect of the total chemical mixture within couples may exist in relation to preterm birth. Below we discuss the existing biological and toxicological support.

We hypothesize that couples’ preconception exposure to phenols and phthalates may affect the male and female germline, especially at the level of epigenetic regulation during gametogenesis [85], which may persist and operate during embryogenesis, decidualization and/or placentation, predisposing to adverse pregnancy and birth outcomes, including a shorter gestation [86–88]. Although mechanistic data on preconception exposures are scarce, our maternal findings are plausible considering the established effect of BPA and DEHP in the ovary with related epigenetic modifications in ovary and oocytes which can be transmitted across generations [89–93]. Regarding the paternal findings, previous epidemiological study found MEHP was associated with decreased sex hormones and increased DNA fragmentation index among males [94]. Rodent studies have also confirmed the ability of DEHP and BPA to exert heritable epigenetic modifications in sperm, with implications for reproductive health in subsequent generations [95, 96], including post implantation losses [97].

Although epigenetic marks in gametes are erased during fertilization and then during embryo preimplantation development, imprinted genes can escape this process [98]. Genomic imprinting, the monoallelic parent-of-origin-dependent expression of a subset of specific genes, is required for normal development, fetal growth, gestation length and metabolism among other functions [99, 100]. Interestingly, both BPA and DEHP have been shown to alter the epigenetic regulation of imprinted genes in gametes [31, 101, 102], potentially explaining the contribution of both parents to birth outcomes such as preterm birth.

In relation to the potential cumulative effect of the total chemical mixture on preterm birth, emerging evidence highlights the frequent exacerbation in response to the so-called “cocktail” effect [47]. For example, co-administration of DEHP in mice reduces the threshold dose at which BPA disrupts blastocyst implantation [63]. In rats, mixtures of BPA and DEHP led to transgenerational effects including reproductive endpoints [65], biologically supporting the potential existence of mixture effects mediated by epigenetic mechanisms.

The observed PCA-derived factors in this study, namely DEHP and BPA factor, paraben factor, HMWP factor and LMWP factor, were similar between preconception mothers and fathers, and in line with previously published analyses in both the EARTH Study [50, 54] and other cohorts [103]. The DEHP-BPA factor suggests that both chemicals share exposure sources, perhaps from packaged food and beverages [60]. Parabens and LMWP, including MEP, MBP and MiBP are mainly used in cosmetics, fragrance and consumer products. Thus, the paraben factor comprising parabens and MEP seems to indicate cosmetic sources of exposure [104].

The major strength of our study was the implementation of two complementary statistical methods to evaluate the contribution of chemical mixtures on the risk of preterm birth among prospective parents. PCA and BKMR methods complemented each other providing coherent answers to our initial research question [68]. PCA grouped the mixtures based on correlations [76] while BKMR used the information on the mixture groups derived by PCA, examining both the overall mixture effect and the effects of each component within the context of the overall joint exposure. BKMR also provided information on potential interactions and cumulative effects of the total mixture, although we did not observe any interactions. Furthermore, BKMR models coupled to hierarchical variable selection [84, 105] estimated the relative contribution of maternal and paternal preconception windows. We were furthermore able to account for the increasingly relevant role of both maternal and paternal factors in pregnancy outcomes [81]. Nevertheless, we remained uncertain whether or not our findings were generalizable to fertile couples since subfertile men and women may have increased susceptibility to environmental perturbations, and preterm birth may be associated with underlying treatment or infertility factors [106]. However, our results remained consistent with the existing literature on BPA and DEHP in relation to adverse outcomes in both subfertile [37, 38, 41, 107] and non-subfertile study populations [108]. Another major strength of the present work was our use of repeated urine samples across the preconception period to quantify exposure to phthalates and phenols. This approach reduced exposure misclassification and the probability of attenuation bias [75], though exposure misclassification was still possible given the short biological half-lives and the episodic nature of exposure to these non-persistent chemicals. Our prospective preconception cohort design was a major advantage relative to other designs reported in prior research on this topic [109, 110]. The modest number of preterm birth cases limited our ability to study the subtypes of preterm birth, including preterm premature rupture of membranes, placental abruption, and spontaneous preterm labor. Caution should be taken when interpreting the couple-based results due to the reduced sample size for couples and the more modest number of preterm cases. Another limitation is that, while human populations are normally exposed to many chemical families at low doses, our mixtures analyses focused on non-persistent chemicals used in plastics manufacturing (e.g., BPA, DEHP) and in personal care products (e.g., parabens). We additionally did not consider potential co-exposure confounding during the prenatal window, given the inherent complexity of the models investigated. However, we did not observe prenatal co-exposure confounding of couple’s DEHP and BPA preconception associations with preterm birth in our previous single-pollutant analyses within this same dataset [39, 40].

5. Conclusions

A complex interplay was observed between parental preconception exposure to mixtures of non-persistent chemicals, with both paternal and maternal windows of exposure jointly contributing to the risk of preterm birth. Specifically, maternal BPA and DEHP, and paternal DEHP exposure before conception, appeared to increase the risk of preterm birth in the context of the joint preconception mixture. Additionally, maternal and paternal preconception windows showed similar contributions supporting that preterm birth should be regarded as a couple-based pregnancy outcome. Our results highlight the need to consider couples’ preconception exposure to chemical mixtures, showing the suitability and advantages of PCA coupled to BKMR for achieving this purpose. Future studies should deepen the mechanistic understanding of these exposure-outcome associations and validate our findings in a fertile population of pregnancy planners.

Highlights.

• Mixtures of parental preconception phenols and phthalates were examined.

• Parental preconception DEHP-BPA mixture was positively associated with preterm birth.

• Couples’ preconception mixtures showed a cumulative effect on preterm birth.

• Maternal and paternal preconception mixtures contributed similarly to the outcome.

• Results suggest that preterm birth may constitute a couple-based outcome.