Paternal and maternal preconception and maternal pregnancy urinary concentrations of parabens in relation to child behavior

Abstract

Background

Epidemiologic studies of the effects of parental preconception paraben exposures on child behavior are limited despite emerging evidence suggesting that such exposures may affect offspring neurodevelopment.

Objective

We investigated whether maternal and paternal preconception and maternal pregnancy urinary concentrations of parabens were associated with child behavior.

Methods

We analyzed data from the Preconception Environmental exposure And Childhood health Effects Study, an ongoing prospective cohort of children aged 6–13 years and their parents. We estimated covariate‐adjusted associations of log_e‐transformed urinary methyl, propyl, and butyl paraben concentrations (individually using linear regression models and as a mixture using quantile g‐computation) collected prior to conception and during pregnancy with Behavioral Assessment System for Children‐3 and Behavior Rating Inventory of Executive Function T‐scores (higher scores indicate more problem behaviors).

Results

This analysis included 140 mothers, 81 fathers, and 171 children (25 sets of twins); parents were predominantly non‐Hispanic white (88% for both mothers and fathers). In single paraben models, higher paternal preconception urinary propyl and methyl paraben concentrations were associated with higher Internalizing Problem T‐scores (propyl paraben $\beta$= 1.7; 95% confidence interval: 0.6, 2.8, methyl paraben $\beta$= 2.2; 95% confidence interval: 0.5, 3.9) and higher Behavioral Symptom Index T‐scores (propyl paraben $\beta$= 1.4; 95% confidence interval: 0.3, 2.5, methyl paraben $\beta$= 1.6; 95% confidence interval: ‐0.1, 3.3). Each quantile increase in the paternal mixture of three parabens was associated with a 3.4 (95% confidence interval: 0.67, 6.1) and 2.5 (95% confidence interval: 0.01, 5.0) increased internalizing problem and Behavioral Symptom Index T‐scores respectively. Higher paternal preconception ($\beta$= 1.0; 95% confidence interval: 0.04, 1.9) and maternal preconception ($\beta$= 1.1 95% confidence interval: ‐0.1, 2.2) concentrations of propyl paraben were associated with higher Behavior Rating Inventory of Executive Function Metacognition Index T‐scores in children, but the paraben mixtures was not.

Conclusion

In this cohort, paternal preconception urinary concentrations of propyl and methyl paraben were associated with worse parent‐reported child behaviors.

1. BACKGROUND

Prior studies have shown that exposure to endocrine‐disrupting chemicals (EDCs) during pregnancy may be associated with an increased risk of neurodevelopmental disorders in children. ¹ , ² Neuroendocrine disruption is hypothesized to be one possible mechanism for the neurotoxicity of EDCs as they not only mimic endogenous hormones but affect their metabolism and/or transport. ³ , ⁴ There is extensive research on prenatal and early childhood exposure to some EDCs, such as phthalates and bisphenol A, in relation to child neurodevelopment. ⁵ , ⁶ , ⁷ However, there is limited research examining child neurodevelopment and other EDCs, especially parabens. Parabens are a group of synthetic EDCs that are widely used as preservatives in foods, cosmetics, and pharmaceutical products. ⁸ A small number of published studies report inconsistent associations of pregnancy concentrations of parabens with various neurobehavioral outcomes. Some have found an increased risk of non‐typical development, increased physician diagnosis of childhood attention deficit hyperactive disorder (ADHD), and lower mental development scores among girls, ⁹ , ¹⁰ , ¹¹ while others found no association. ¹² , ¹³

Emerging evidence suggests that preconception chemical exposures in mothers and fathers have the potential to impact offspring neurodevelopment. ¹⁴ For instance, some studies suggest that parental smoking and exposure to phthalates in the preconception period affect child health outcomes. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ However, paternal preconception exposures, particularly in regard to chemical exposures, have rarely been considered together with maternal preconception exposures.

To understand the potential neurotoxic effects of exposure to parabens and address the gaps in our knowledge related to the potential effects of preconception exposures, we investigated whether maternal and paternal preconception and maternal pregnancy urinary concentrations of parabens were associated with parent‐reported problem behaviors among mother‐father‐child triads participating in the Preconception Environmental exposure And Childhood health Effects (PEACE) study.

2. METHODS

2.1. Study participants

This analysis of the PEACE study cohort included children aged 6−13 years whose parent(s) previously enrolled in the Environment and Reproductive Health (EARTH) study. The EARTH study was a prospective cohort of over 900 women and 500 men between the ages of 18 and 45 years and their partners who attended the Massachusetts General Hospital (MGH) Fertility Center. ¹⁹ Of the mother‐father‐child triads eligible to enroll in the PEACE study, 140 mothers, 82 fathers, and 171 children enrolled and were included in this analysis (Figure 1). A majority of enrolled children were conceived using in vitro fertilization (IVF) or intrauterine insemination (IUI); however, a substantial proportion (25%) of children were conceived without medical assistance. Children were excluded from PEACE study enrollment if a surrogate or gestational carrier was used. Additionally, children were excluded from this analysis if a parent did not complete behavioral questionnaires if maternal or paternal urinary concentrations of parabens were unavailable, if the child was less than 6 years of age, and, for the analysis with the Behavior Assessment System for Children‐Third Edition (BASC‐3), if the child was 12 years old or older.

FIGURE 1.

Flow chart of Preconception Environmental exposure And Childhood health Effects (PEACE) study participation.

We conducted this follow‐up with study mothers, fathers, and children beginning in July 2018. In‐person assessments were offered until March 2020 and then virtually thereafter due to the restrictions that resulted from the coronavirus disease 2019 (COVID‐19) pandemic. The PEACE study was approved by the Harvard T.H. Chan School of Public Health institutional review board. The EARTH study was approved by the Human Subject Committees of the Harvard T.H. Chan School of Public Health, MGH, and the Centers for Disease Control and Prevention (CDC). Verbal (for children < 7 years of age) or written (children > 7 years of age) assent was obtained from all child participants and written informed consent from their guardians. Parental and child participants were compensated with gift cards for their participation.

2.2. Parabens exposure assessment

We collected urine samples from mothers and fathers in polypropylene specimen cups. Mothers provided up to two preconception urine samples (collected on different days) for each fertility treatment cycle or one urine at EARTH study entry for naturally conceived pregnancies. Additionally, mothers provided up to three pregnancy urine samples (one per trimester) regardless of the mode of conception. Fathers provided one preconception urine sample per fertility cycle on the day that their partner underwent the fertility procedure or at EARTH study enrollment for medically unassisted pregnancies. Preconception samples were used only if associated with the cycle that resulted in the pregnancy of the child enrolled in the PEACE Study.

Research staff at the Harvard T.H. Chan School of Public Health and MGH measured the specific gravity (SG) of each urine sample using a handheld refractometer (National Instrument Company Inc). Urine samples were divided into aliquots, kept frozen at −80°C for one to three years, and shipped on dry ice overnight to the CDC laboratory for analysis. CDC laboratory staff quantified urinary concentrations of methyl, propyl, and butyl parabens using solid‐phase extraction coupled with high‐performance liquid chromatography–isotope dilution tandem mass spectrometry as described in detail elsewhere. ²⁰ The limits of detection (LODs) were 0.2 μg/L (propyl and butyl paraben) and 1.0 μg/L (methyl paraben). ²¹ Urinary paraben concentrations were standardized for urine dilution by SG using the following formula: P_s  =  P_i[(SG_m − 1) / (SG_i − 1)], where P_s is the SG‐standardized chemical biomarker concentration (μg/L), P_i is the measured biomarker concentration (μg/L), SG_i is measured specific gravity, and SG_m is the mean SG concentration for the maternal (1.016) or paternal (1.018) preconception or maternal pregnancy (1.014) samples. ²² Biomarker concentrations below the LOD were assigned a value equal to the LOD divided by the square root of 2 prior to adjustment by SG. ²³

2.3. Child behavior assessment

We assessed child behavior using the BASC‐3 and the Behavior Rating Inventory of Executive Function (BRIEF). The BASC‐3 can be administered as either child‐reported or parent‐reported assessments; for this study, we used the parent‐reported assessment. The BASC‐3 valid and reliable instrument that assesses overall behavioral and emotional functioning as well as specific problem and adaptive behaviors in children aged 6−11 years. ²⁴ The BASC‐3 includes 175 items with Likert‐style responses that are used to calculate 29 subscales and five composite scale scores. The BRIEF is a valid and reliable parent‐reported questionnaire that measures the extent to which executive dysfunction impairs a child's daily life. Executive dysfunction is a behavioral symptom that disrupts a person's ability to plan ahead, display self‐control, avoid distractions, and meet goals. ²⁵ , ²⁶ The BRIEF includes 86 items with Likert‐style responses that are used to calculate eight clinical scales, two broader indexes, and an overall composite score.

During in‐person study visits (58 children), parents (46 mothers and two fathers) completed paper copies of the BASC‐3, BRIEF, and other questionnaires. After March 2020, parents completed the BASC‐3 via the Internet using the Q‐global online platform (102 completed online) and the BRIEF via the Internet using the PARI Connect online platform (113 completed online). Studies have shown there is no significant difference in median values between online responses and paper responses. ²⁷ For this analysis, we focused on three composite scores from the BASC‐3 that assessed features of prevalent childhood behavior disorders: externalizing problems (e.g., ADHD‐like behaviors) and internalizing problems (e.g., anxiety and depression), as well as an overall composite score, the Behavioral Symptoms Index (BSI, e.g., total problem behaviors). For the BRIEF questionnaire, we focused on three composite scores: Metacognition Index (child's ability to initiate, plan, organize, self‐monitor, and sustain working memory), Behavioral Regulatory Index (child's ability to shift cognitive set and modulate emotions and behavior via appropriate inhibitory control), and the Global Executive Composite (summary score of the BRIEF clinical scales). ²⁸ Composite and index scores were used in analyses as they represent larger domains of behavior. BASC‐3 scores are expressed as T‐scores, with a population reference mean and standard deviation of 50 and 10, respectively. ²⁴ BRIEF scores are also expressed as T‐scores and standardized for age and child sex assigned at birth. For both instruments, higher scores represent more problematic behavior or executive dysfunction; scores $\ge$60 are considered “at risk”, and scores $\ge$70 are considered to indicate “clinically significant” behavioral problems.

2.4. Covariates

We selected covariates as potential confounders or predictors of behavioral scores based on substantive knowledge and a directed acyclic graph (DAG) (Figure S1). Child and parental sociodemographic information was collected electronically through RedCap by parent report at the time of the PEACE study enrollment. These included child age, child and parental race and ethnicity, parental education, child sex assigned at birth, and the presence of a sibling in the home. Additional parental covariate information was collected at enrollment in the EARTH study and throughout participation including parental age at conception, pre‐pregnancy height and weight, smoking status, and mode of conception, such as IVF, IUI or non‐medically assisted conception.

2.5. Statistical analysis

We averaged log_e‐transformed SG‐adjusted urinary parabens concentrations from up to two samples collected before conception from the fertility cycle that resulted in the pregnancy. Similarly, we averaged log_e‐transformed SG‐adjusted urinary paraben concentrations from up to three samples collected during the pregnancy of the enrolled child. If only one urine sample was available, the concentrations of parabens for the single sample were used for that time period. We calculated descriptive statistics for the biomarker concentrations as well as the percentage over the LOD for paternal preconception, maternal preconception, and pregnancy parabens. Using Spearman's correlation coefficients, we calculated the correlation between individual urinary parabens during each of the three periods, as well as among periods.

2.5.1. Single paraben analysis

We estimated associations of paternal and maternal preconception and maternal pregnancy SG‐adjusted urinary log_e‐ concentrations of parabens with BASC‐3 and BRIEF index and composite T‐scores using multivariable generalized estimating equation linear models, accounting for correlation among multiple fetal gestations (twins) with an exchangeable correlation structure. Due to the high percentage of mothers (30% preconception, 41% pregnancy) and fathers (68%) with butyl paraben concentrations below the LOD of 0.2, in addition to continuous butyl paraben, we examined dichotomized butyl paraben (< LOD and ≥ LOD). Based on a priori knowledge and the aforementioned DAG, all analyses were adjusted for child sex assigned at birth, child race (white vs. non‐white), if the child had a sibling, maternal/paternal age, maternal/paternal education (high school and/or college education vs. post‐graduate education), and use of medically assisted conception (Assisted Reproductive Technologies and IUI vs. natural conception). Few covariates had any missing values, and those that did (maximum 3% missing) were replaced by the average or most frequent covariate value for continuous and categorical variables, respectively.

2.5.2. Multi‐paraben analysis

Because parabens are often used as a mixture in products and we previously observed moderate to the high correlation amongst parabens, we estimated the effect of the paraben mixture using quantile g‐computation using the qgcomp R package. ²⁹ , ³⁰ , ³¹ Briefly, quantile g‐computation estimates the parameters of a joint marginal structural model to quantify the effect of modifying all components of the mixture simultaneously. Quantile g‐computation expresses the effects of increasing all paraben concentrations by one quartile simultaneously, conditional on confounders (child sex assigned at birth, child race, maternal/paternal age, and maternal/paternal education). While assuming linearity and additivity of associations, this approach also allows for accommodation of non‐linearity and non‐additivity of both the individual exposure effects, as well as the effect of the mixture. Models were run separately for each BASC‐3 and BRIEF scale of the three periods: paternal preconception, maternal preconception, and maternal pregnancy. This approach also provides weights that indicate the relative contribution of the individual components of the mixture to the positive and negative effects. We also implemented quantile g‐computation with 200 bootstrapped samples.

2.5.3. Sensitivity analysis

For parabens showing at least marginal associations (p‐value < 0.10) with a BASC‐3 or BRIEF scale in the single paraben models, we conducted additional analyses adjusting for urinary paraben concentrations in the other periods (i.e., maternal preconception and pregnancy, maternal and paternal preconception, or pregnancy and paternal preconception) to reduce the concern of residual confounding by exposures in other windows.

3. RESULTS

This analysis included 140 mothers, 82 fathers, and 171 children (including 25 sets of twins). The analysis for BASC‐3 was performed on a subset of the overall cohort (127 mothers, 77 fathers, and 156 children) because the version of the BASC‐3 used was valid in children ages 5−12 years. Parents were predominantly non‐Hispanic white (88% for both mothers and fathers), with a mean age at the conception of 34 years (mothers) and 35 years (fathers), and were generally highly educated (70% of mothers had a postgraduate degree; 46% of fathers had a post‐graduate degree) (Table 1). Similarly, children were predominately non‐Hispanic white (85%), with an average age of 9.3 years at enrollment, and 54% were assigned male at birth. Because this cohort was recruited from a fertility clinic, 62% of children were conceived through IVF, 13% through IUI, and 25% without medical assistance (Table 1).

TABLE 1.

Demographic Characteristics of Child and Parental Participants in the Preconception Environmental exposure And Childhood health Effects (PEACE) study ^* .

Characteristic	Child N = 171	Maternal N = 140	Paternal N = 81
Mean Age (years) [SD] At visit for child, at conception for parents	9.3 [1.7]	34 [3.6]	35 [4.4]
Race
White	145 (85%)	120 (88%)	69 (88%)
Black	2 (1%)	2 (1%)	2 (3%)
Asian	2 (1%)	10 (7%)	3 (4%)
Multi‐Racial	17 (10%)	0 (0%)	0 (0%)
Other	5 (3%)	4 (3%)	4 (5%)
Education	N/A	41 (30%)	43 (54%)
High school and/or college education	N/A	96 (70%)	37 (46%)
Post‐graduate education	N/A	41 (30%)	43 (54%)
Sex Assigned at Birth
Male	92 (54%)	N/A	N/A
Female	79 (46%)	N/A	N/A
Sibling Status
No Siblings	26 (15%)	N/A	N/A
One or more sibling	145 (85%)	N/A	N/A
Twins (N)	50 (25 sets)	N/A	N/A
Mode of conception
Unassisted Conception	42 (25%)	N/A	N/A
Intrauterine Insemination	22 (13%)	N/A	N/A
In Vitro Fertilization (with or without intracytoplasmic sperm injection)	107 (62%)	N/A	N/A

Note: all entries are N (%) except age.

^*Paternal education missing 1, paternal race missing 3, maternal race missing 4, maternal education missing 2.

In the preconception period, the majority of mothers (87%) provided two urine samples. During pregnancy, mothers provided one (5%), two (25%), or three (69%) samples. All fathers provided 1 preconception urine sample. Urinary paraben concentrations of parabens were similar to those observed in previous analyses of the larger EARTH cohort (Table 2). ³² , ³³ , ³⁴ Methyl and propyl paraben in each of the three time periods were highly correlated with one another (Spearman r = 0.71–0.84) (Figure 2). Log_e‐transformed urinary biomarker concentrations were weakly correlated within couples in the preconception period (Spearman r = −0.08–0.23) (Figure 2) and weakly to moderately correlated between maternal preconception and pregnancy periods (Spearman r = 0.13–0.56) (Figure 2). Mean BASC‐3 and BRIEF scores ranged from 49.3 to 49.9. Note that, 11%–18% of children had T‐scores greater than 60 (considered “at risk or clinically significant”), depending on the domain (Table S1).

TABLE 2.

Summary statistics of the distributions of specific‐gravity standardized urinary biomarker concentrations (μg/L) before and during pregnancy among Preconception Environmental exposure And Childhood health Effects (PEACE) study mothers and fathers (2005–2017).

	Paternal Preconception			Maternal Preconception			Maternal Pregnancy
Biomarker*	N	LOD μg/L (% > LOD)	Geometric Mean (25th Percentile, 75th Percentile)	N	LOD μg/L (% > LOD)	Geometric Mean (25th Percentile, 75th Percentile)	N	LOD μg/L (% > LOD)	Geometric Mean (25th Percentile, 75th Percentile)
Methyl paraben	81	1.0 (94%)	29 (10, 73)	132	1.0 (99%)	158 (69, 381)	140	1.0 (100%)	109 (40, 313)
Propyl paraben	81	0.2 (99%)	2.9 (0.7, 8.8)	132	0.2 (98%)	27 (9.8, 89)	140	0.2 (98%)	18 (6.0, 61)
Butyl paraben	81	0.2 (32%)	<LOD (< LOD, 0.6)	132	0.2 (70%)	1.6 (0.3, 7.4)	140	0.2 (59%)	0.8 (0.2, 2.2)

FIGURE 2.

Within person and between period urinary paternal preconception, maternal preconception, and maternal pregnancy paraben spearman correlations.

3.1. Single paraben analysis

3.1.1. Paternal preconception period

Higher paternal preconception urinary propyl or methyl paraben concentrations were associated with higher mean internalizing problem T‐scores in children (propyl paraben $\beta$= 1.7; 95% confidence interval [CI]: 0.6, 2.8, methyl paraben $\beta$= 2.2; 95% CI: 0.5, 3.9) and BSI T‐scores (propyl paraben $\beta$= 1.4; 95% CI: 0.3, 2.5, methyl paraben $\beta$= 1.6; 95% CI: −0.1, 3.3) per 1 log_e increase (Figure 3 and Table S2). When further examining the subscales contributing to internalizing problem scores, the previous association was driven by the anxiety and depression subscales (Table S3). Higher paternal preconception propyl paraben concentrations were associated with higher BRIEF Metacognition Index T‐scores ($\beta$= 1.0; 95% CI: 0.04, 1.9). Otherwise, paternal preconception paraben concentrations were not associated with other BASC‐3 or BRIEF T‐scores (Figure 3 and Table S2).

FIGURE 3.

Adjusted¹ average difference in Behavior Assessment System for Children‐Third Edition (BASC‐3) and Behavior Rating Inventory of Executive Function (BRIEF) T‐scores per log_e increase in paternal preconception urinary paraben concentrations*. ¹Adjusted for child sex assigned at birth, child race (white vs. non‐white), whether the child has any sibling, paternal age (continuous, years), paternal education (high school and/or college education vs. post‐graduate education), and mode of conception (medically assisted vs. unassisted conception). * Methyl and propyl paraben are continuous concentrations while butyl paraben concentrations are dichotomized as below the limit of detection and above the limit of detection.

3.1.2. Maternal preconception period

With one exception, maternal urinary paraben concentrations were not associated with BASC‐3 or BRIEF scores (Figure 4 and Table S4). Higher maternal methyl paraben concentrations were associated with higher mean Metacognition Index T‐scores in children ($\beta$= 1.1; 95% CI: −0.1, 2.2 per log_e increase).

FIGURE 4.

Adjusted ¹ Average Difference in Behavior Assessment System for Children‐Third Edition (BASC‐3) and Behavior Rating Inventory of Executive Function (BRIEF) T‐scores per log_e Increase in Maternal Preconception Urinary Paraben Concentrations*. ¹Adjusted for child sex assigned at birth, child race (white vs. non‐white), whether the child has any sibling, maternal age (continuous, years), maternal education (high school and/or college education vs. post‐graduate education), and mode of conception (medically assisted conception vs. unassisted conception). * = p‐value < 0.1, ** = p‐value < 0.05. *Methyl and propyl paraben are continuous concentrations while butyl paraben concentrations are dichotomized as below the limit of detection and above the limit of detection.

3.1.3. Maternal pregnancy period

Similar to the preconception period, maternal prenatal urinary paraben concentrations were generally not associated with BASC‐3 or BRIEF scores (Figure 5). When examining continuous pregnancy concentrations of butyl paraben higher concentrations were related to lower mean BSI T‐scores, however, the relationship did not persist when examining dichotomous butyl paraben as above and below the LOD (Table S5).

FIGURE 5.

Adjusted ¹ Average Difference in Behavior Assessment System for Children‐Third Edition (BASC‐3) and Behavior Rating Inventory of Executive Function (BRIEF) T‐scores per log_e increase in maternal pregnancy urinary paraben concentrations*. ¹Adjusted for child sex assigned at birth, child race (white vs. non‐white), whether the child has any sibling, maternal age (continuous, years), maternal education (high school and/or college education vs post‐graduate education), and mode of conception (medically assisted conception vs. unassisted conception). * = p‐value < 0.1, ** = p‐value < 0.05. *Methyl and propyl paraben are continuous concentrations while butyl paraben concentrations are dichotomized as below the limit of detection and above the limit of detection.

3.2. Paraben mixture analysis

Each one‐quartile increase in the paternal preconception paraben mixture was associated with higher internalizing problems T‐scores (ψ = 3.4; 95% CI: 1.0, 5.8) and BSI T‐scores (ψ = 2.5; 95% CI: −0.05, 5.1) (Table 3). Similar to the single paraben models, the association with internalizing problems was driven primarily by the anxiety subscale (Table S3). The association of the paternal preconception paraben mixture was driven primarily by propyl paraben (weights = 0.77 and 0.64 for internalizing problems T‐scores and BSI T‐scores, respectively) and to a lesser extent methyl paraben (weights = 0.23 and 0.36, respectively) (Figure S2).

TABLE 3.

Adjusted Difference in Behavior Assessment System for Children‐Third Edition (BASC‐3) and Behavior Rating Inventory of Executive Function (BRIEF) T‐scores with a one quartile increase in a mixture ^* of three parabens measured in the paternal preconception, maternal preconception and maternal pregnancy periods ¹ .

	Non‐bootstrapped model
	Sum of positive coefficients	Sum of negative coefficients	Ψ (95% CI)
Paternal preconception
Externalizing problems	2.73	−1.07	1.7 (−0.95, 4.3)
Internalizing problems	4.50	−1.09	3.4 (1.0, 5.8)
Behavioral symptoms index	3.36	− 0.86	2.5 (−0.05, 5.1)
Metacognition index	1.59	−1.54	0.04 (−2.3, 2.4)
Behavioral regulatory index	2.00	−2.51	−0.50 (−2.9, 1.8)
Global executive composite	1.73	−1.98	−0.24 (−2.5, 2.1)
Maternal preconception
Externalizing problems	1.27	−1.81	−0.54 (−2.3, 1.2)
Internalizing problems	0.44	−1.47	−1.0 (−2.9, 0.74)
Behavioral symptoms index	0.93	−1.68	−0.75 (−2.5, 1.0)
Metacognition index	1.23	−0.95	0.29 (−1.5, 2.1)
Behavioral regulatory index	0.85	−1.6	−0.75 (−2.5, 1.0)
Global executive composite	1.16	−1.27	−0.12 (−1.9, 1.7)
Maternal pregnancy
Externalizing problems	1.07	−2.02	−0.95 (−3.0, 1.1)
Internalizing problems	1.72	−2.93	−1.2 (−3.3, 0.97)
Behavioral symptoms index	1.43	−2.51	−1.1 (−3.1, 0.95)
Metacognition index	1.38	−1.29	0.09 (−1.8, 2.0)
Behavioral regulatory index	2.21	−2.20	0.01 (−1.9, 1.9)
Global executive composite	1.71	−1.69	0.01 (−1.9, 1.9)

^*Using quantile g‐computation which expresses the effects of increasing all paraben concentrations by one quartile simultaneously, conditional on confounders.

¹ Adjusted for child sex assigned at birth, child race (white vs. non‐white), paternal/maternal age (continuous, years), and paternal/maternal education (high school and/or college education vs. post‐graduate education).

Both the maternal preconception and maternal pregnancy mixture of parabens were not associated with BASC‐3 or BRIEF scores. We observed similar associations when implementing quantile g‐computation with bootstrapping (data not shown). Propyl and methyl parabens also had positive weights in the maternal preconception and pregnancy periods, although paraben mixtures were not associated with any BASC‐3 or BRIEF outcomes (Table 3 and Figures S3 and S4).

3.3. Sensitivity analysis

When adjusting for either the other parent's or other period's concentrations of the same paraben, almost all paternal preconception relations remained unchanged (Table S6). However, when adjusting paternal preconception methyl paraben concentrations for maternal preconception or pregnancy methyl paraben concentrations, there was no longer a marginal relationship with BASC‐3 BSI T‐scores (Table S6). Additionally, when adjusting for maternal pregnancy methyl paraben concentrations, the relationship between maternal preconception methyl paraben concentrations and the BRIEF Metacognition Index T‐scores was attenuated (Table S7). When adjusting for maternal preconception butyl paraben concentrations, the relationship between maternal pregnancy butyl paraben concentrations and BASC‐3 BSI T‐scores was attenuated (Table S8).

4. DISCUSSION

In this preconception cohort, higher paternal preconception urinary methyl and propyl paraben concentrations, as well as their mixture, were associated with increased parent‐reported behavior problems in their children; these associations appeared to be primarily driven by associations with the anxiety and depression subscales. Results for paternal paraben concentration were generally robust when adjusting for maternal urinary paraben concentrations in preconception and pregnancy. While in single paraben models, higher paternal and maternal preconception urinary methyl and propyl paraben concentrations were associated with increased Metacognition Index T‐scores, this relationship did not remain in the mixture analysis.

There is limited research on the impact of parabens on neurodevelopment in animals or during the maternal and paternal preconception periods. However, a few human studies found that parabens may affect semen quality (i.e.; abnormal morphology, reduced motility, inhibiting mitochondrial function). ³⁵ , ³⁶ , ³⁷ These effects on spermatogenesis suggest that impacts on sperm have the potential to play a role in the biological mechanisms through which paternal exposures may impact child neurodevelopment. ³⁸ , ³⁹ However, it is important to note that these studies in germ cells and gametes did not explore paraben associations with epigenetic changes in sperm, which are the hypothesized mechanism through which paternal exposures may impact children's health. ¹⁴ , ⁴⁰ , ⁴¹ The impact of maternal pregnancy EDC exposure on the interactions between proteins (Ras‐activated Ca2+influx), receptors (NMDA glutamate receptor activity and 5‐hydroxytryptamine receptor activity), and signaling pathways (Grin2A interactions) are further potential mechanisms of action of EDCs in mediating brain function. ⁴²

With the exception of butyl paraben concentrations and BSI T‐scores, we found no associations between maternal pregnancy paraben concentrations and child behavior. Several other cohort studies also did not find any association between maternal pregnancy paraben concentrations and child neurobehavior. Specifically, in a French mother‐child cohort, urinary concentrations of a mixture of methyl, ethyl, butyl, and propyl parabens during pregnancy had close to null associations with internalizing behaviors (assessed by the Strengths and Difficulties Questionnaire and completed by mothers) in boys at age 3 years. ¹² Similarly, no associations were observed when authors evaluated each of the pregnancy paraben biomarkers separately. ¹³ However, studies from other cohorts did find associations between maternal pregnancy paraben concentrations and child neurodevelopment and behavior. In the Markers of Autism Risks in Babies Learning Early Signs Cohort Study in California, a mixture of urinary biomarkers of phenols and parabens (including methyl, ethyl, butyl, and propyl parabens) during pregnancy was associated with nontypical development at 3 years which was measured using the Autism Diagnostic Observation Scale and the Mullen Scales of Early Learning. ⁹ In a prenatal cohort in China of mother‐child pairs, authors observed that pregnancy urinary concentrations of methyl paraben and the sum of methyl, ethyl, butyl, and propyl parabens were associated with lower mental developmental index scores in girls, but not boys, at 2 years. ¹¹ In the GESTation and the Environment prospective observational pregnancy cohort in Canada, infant's methyl paraben concentrations in meconium were associated with reported ADHD by physician diagnosis at the age of 6−7 years. ¹⁰

A previous analysis done with 158 children with a mean age of 4.6 (min:2.1, max:8.9) years whose parents participated in the EARTH study, found no association between the sum of concentrations of parabens measured in preconception or pregnancy periods and BASC‐2 scores. ⁴³ While this finding does differ from our paternal and maternal preconception results, this result could relate in part to the different age of measurement and that child behavioral issues may be more apparent for parental reports at the older age of children for this analysis. Additionally, the previous analysis examined only the association with the sum of parabens while our analysis examined both the single paraben associations and implemented mixtures analysis methods to examine the mixture of three parabens. This previous analysis also used preconception maternal and paternal paraben concentrations that were measured from any urine sample obtained from study entry and at any fertility treatment cycle up until and including the index cycle of conception of the child under study.

While our study has many strengths, it also has limitations. This analysis was conducted on a modest sample size. Only 30% of eligible mothers and fathers have enrolled in PEACE. This is due in part to strict enrollment criteria, challenges associated with conducting follow‐up during the COVID‐19 pandemic, and the long interval since the last contact with EARTH participants. This modest sample size has the possibility to decrease the precision of our estimates and precluded us from exploring sex‐stratified associations. Our study sample included 25 sets of twins and while we adjusted for the correlation between twins in our linear regression model, given the study sample size we were unable to additionally adjust for the correlation between siblings. We are unaware of other studies that have been able to link detailed prospectively collected preconception environmental exposure data in both mothers and fathers, especially of parabens, with child health. Being able to leverage the rich covariate database from the EARTH study is also a strength of our study. The use of two valid and reliable parental report questionnaires, the BASC‐3 and BRIEF to assess child behavioral assessment is another strength. While studying a cohort that is relatively homogenous in terms of race and socioeconomic status may reduce the generalizability of the findings, it also reduces the potential for confounding from these factors. Exposure misclassification is possible as parabens have relatively short biological half‐lives and people are episodically exposed. Thus, single measures of urinary concentrations may not necessarily reflect a person's long‐term exposure to parabens, ³¹ although previous analysis in the EARTH cohort suggested fair reliability between repeated paraben concentrations before and during pregnancy. ⁴⁴ Furthermore, a strength of our study was that the exposure assessment relied on multiple urine samples that have been averaged for both maternal preconception and pregnancy periods for most women, resulting in a more precise measurement of biomarker concentrations during these periods.

5. CONCLUSIONS

In the PEACE study preconception prospective cohort, we found that higher paternal preconception urinary methyl and propyl paraben concentrations and their mixture were associated with increased parent‐reported internalizing (anxiety/depression) behaviors, as well as overall increased behavioral problems. Generally, we found no association between maternal pregnancy or preconception paraben concentrations with child behavior. Thus, our results add to the literature suggesting the potential importance of the preconception period as a window of vulnerability for child health outcomes, especially paternal preconception exposure. Additional research can help validate our findings and explore possible mechanisms underlying these associations, such as impacts on sperm and oocyte epigenetics.

AUTHOR CONTRIBUTIONS

Jordana Leader conducted study visits, data collection, cleaning and analysis, and writing and preparing the manuscript. Lidia Mínguez‐Alarcón and Paige L. Williams provided expertise and guidance with data analysis as well as editing and reviewing of the manuscript. Jennifer B. Ford conducted study recruitment and visits as well as provided expertise in regard to the EARTH study cohort. Olivia Chagnon conducted a study on recruitment and visits. Ramace Dadd provided expertise in regard to the EARTH study cohort. Emily Oken edited and reviewed the manuscript. Antonia M. Calafat supervised phthalate and BPA biomarkers quantification and edited and reviewed the manuscript. Russ Hauser and Joseph M. Braun conceptualized this research study, oversaw the research investigation, acquired funding for this research study, and edited and reviewed the manuscript.

CONFLICT OF INTEREST STATEMENT

Joseph Braun was compensated for serving as an expert witness for plaintiffs involved in litigation over PFAS‐contaminated drinking water. The other authors declare no conflict of interest.

DISCLAIMER

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC). Use of trade name is for identification only and does not imply endorsement by the CDC, the Public Health Service, or the U.S. Department of Health and Human Services.

Supporting information

Supporting Information

Leader J, Mínguez‐Alarcón L, Williams PL, et al. Paternal and maternal preconception and maternal pregnancy urinary concentrations of parabens in relation to child behavior. Andrology. 2025;13:22–33. 10.1111/andr.13576

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.