Prenatal exposure to phthalates and alternative plasticizers and emotional and behavioral outcomes in early childhood in the Environmental influences on Child Health outcomes (ECHO) cohort

Abstract

Background:

Evidence suggests prenatal phthalate exposure adversely affects children’s behavior. However, epidemiological studies on alternative plasticizers remain scarce. This study investigated associations of gestational exposure to phthalates and alternative plasticizers with internalizing and externalizing behaviors in children aged 1.5–5 years.

Methods:

The study included 2617 mother–child dyads from 13 Environmental influences on Child Health Outcomes (ECHO) cohorts. Maternal urine samples, primarily collected mid- to late-pregnancy, were analyzed for 27 phthalate metabolites and 6 alternative plasticizer metabolites. Based on detection frequency, metabolite concentrations were modeled either continuously or categorically (Group 1: non-detectable, 2: lower detectable, 3: higher detectable). Covariate-adjusted associations between individual metabolite concentrations and internalizing and externalizing T-scores on the Child Behavior Checklist for Ages 1½–5 were estimated using linear mixed-effects models. Effect modification by child sex was also examined.

Results:

An interquartile range increase in mono-benzyl phthalate (MBzP) and membership in Group 3, versus Group 1, for mono-hexyl phthalate (MHxP) were associated with higher externalizing T-scores (β_ext for continuous MBzP = 0.53, 95% CI: 0.05, 1.00; β_ext for MHxP Group 3 = 1.23, 95% CI: 0.35, 2.12). We observed no robust associations between phthalate metabolites and internalizing T-scores, nor between cyclohexane-1,2-dicarboxylic acid mono carboxyisooctyl ester (DINCH) metabolites and any behavioral outcomes. Child sex modified associations between several metabolites and externalizing T-scores, although the direction of effect varied by metabolite.

Conclusion:

This large-scale study suggests that prenatal exposure to several phthalates, but not to the alternative plasticizer DINCH, may be associated with a small-to-modest increase in externalizing behaviors in young children.

1. Introduction

Phthalates are a group of high-production-volume chemicals primarily used as plasticizers in various industrial and consumer products (Heudorf et al., 2007). High-molecular-weight phthalates are commonly found in toys, childcare articles, food packaging materials, pharmaceuticals, medical devices, vinyl flooring, wall covering, and automotive components, while low-molecular-weight phthalates are found in adhesives, coatings, and personal care products such as cosmetics, perfumes, lotions, shampoos, and nail polish (Meeker et al., 2009; Wang and Qian, 2021; Wang et al., 2019). Due to the adverse health effects of phthalate exposure (Benjamin et al., 2017; Hauser and Calafat, 2005; Heudorf et al., 2007; Lyche et al., 2009; Wang and Qian, 2021), the European Union, the United States (U.S.), and several other countries have implemented regulations or recommendations to restrict the use of specific phthalates in children’s products and food contact materials (Tumu et al., 2023; Wang and Qian, 2021). Although exposure to several phthalates has decreased over the last two decades (Domínguez-Romero et al., 2023; Jiang et al., 2023; Kasper-Sonnenberg et al., 2019; Vogel et al., 2023), phthalates remain widely detected in the general population and among pregnant women (Wang et al., 2019), potentially posing risks to children’s health (Almeida-Toledano et al., 2024; Wang and Qian, 2021). Furthermore, exposure to alternative plasticizers, such as di-iso-nonyl cyclohexane-1,2-dicarboxylate (DINCH) and di (2-ethylhexyl) terephthalate (DEHTP), has increased (Domínguez-Romero et al., 2023; Jiang et al., 2023; Kasper-Sonnenberg et al., 2019). Despite limited research, the endocrine-disrupting potential and toxicities of these alternative plasticizers have been reported (Gerofke et al., 2024; Qadeer et al., 2024).

Phthalates can cross the placenta and impact the developing fetus during crucial periods of neurodevelopment, when the fetal brain is particularly susceptible to environmental toxicants (Lucaccioni et al., 2021). Rodent studies suggest that there are neurobehavioral effects of perinatal phthalate exposure at doses higher than environmentally relevant levels but well below toxic doses, often showing sex-specific effects (Engel et al., 2021; Hlisníková et al., 2021; Mustieles et al., 2023). For example, perinatal exposure to di-2-ethylhexyl phthalate (DEHP) increased anxiety- and/or depressive-like behaviors in male rats and both sexes of mice (Carbone et al., 2013; Xu et al., 2015). Female rats perinatally exposed to di-iso-nonyl phthalate (DiNP) exhibited masculinization of learning behavior (Boberg et al., 2011). A mixture of DEHP, DiNP, and dibutyl phthalate impaired social behaviors, particularly in females (Morová et al., 2020). Evidence on the neurotoxic effects of alternative plasticizers is limited, although several experimental studies have reported that DINCH has endocrine-disrupting potential (Campioli et al., 2015; Campioli et al., 2017; Engel et al., 2018; Zughaibi et al., 2022) and can induce behavioral changes in zebrafish larvae (Saad et al., 2021).

Accumulating epidemiological research largely supports the link between gestational phthalate exposure and adverse neurodevelopmental outcomes in children (Ejaredar et al., 2015; Engel et al., 2021; Minatoya and Kishi, 2021). Most studies focusing on early childhood behavior have observed adverse associations for several phthalate metabolites (Choi et al., 2021; Cohen-Eliraz et al., 2023; Day et al., 2021; Dewey et al., 2023; Engel et al., 2010; England-Mason et al., 2020; Kim et al., 2018; Ku et al., 2020; Li et al., 2020; Philippat et al., 2017; Singer et al., 2017; Tsai et al., 2023), with only a few studies reporting null or inverse associations (Andreasen et al., 2023; Gascon et al., 2015; Minatoya et al., 2018). However, there is a limited number of studies examining the neurodevelopmental impacts of alternative plasticizers, which may be replacing phthalates, calling for timely research on this topic (Colicino et al., 2021; Guilbert et al., 2021; Özel et al., 2023; Park et al., 2023).

The National Institutes of Health (NIH) Environmental influences on Child Health Outcomes (ECHO) consortium provides a collaborative framework that was initiated with 69 U.S. pregnancy and pediatric cohorts with over 107,000 participants to understand the impacts of environmental exposures on child health outcomes (Knapp et al., 2023). A recent large-scale study combining data from three ECHO cohorts found that a phthalate mixture and several individual metabolites in prenatal maternal urine were weakly associated with higher total problem scores at 4–6 years of age, with more pronounced effects observed among males (Barrett et al., 2024). However, that study did not measure alternative plasticizers and used total problem scores, which sum the scores of externalizing, internalizing, sleep problems, and other behavioral scales. Therefore, our study expands on this work by investigating an extensive set of phthalate and alternative plasticizer metabolites to determine their associations with internalizing and externalizing behaviors in young children, using a larger, geographically and sociodemographically diverse ECHO population from 13 cohorts. Secondarily, we examined two potential effect modifiers: (a) child sex, based on previous sex-dimorphic associations (Jankowska et al., 2021; Palanza et al., 2021), and (b) the child opportunity index (COI), a measure of neighborhood-level resources and conditions, given that disparities in neighborhood factors may amplify the effects of phthalate exposure through co-occurring stressors, potentially influencing neurodevelopment (Barrett and Padula, 2019; Dickerson et al., 2023; Galvez et al., 2018; Payne-Sturges et al., 2023).

2. Methods

2.1. Study population

This study initially considered ECHO cohorts that (a) contributed prenatal maternal urine samples for the quantification of a multi-class chemical panel measuring more than 100 contemporary and emerging chemicals, including phthalate and alternative plasticizer metabolites, and (b) assessed children’s behaviors at ages 1.5–5 years using the Child Behavior Checklist for Ages 1½–5 (CBCL/1½–5). Among the 14,426 biological mother–child dyads (hereafter referred to as mother–child dyads) from 16 cohorts meeting these criteria (Supplementary Fig. S1), not all samples collected by the cohorts could be analyzed due to budgetary constraints. Accordingly, each cohort selected either a single spot or first morning void sample per participant, with some larger cohorts limiting their selection to a subset of eligible pregnancies based on their own criteria. We excluded participants with samples not chosen for chemical analysis (n = 10,130); those with missing sample collection dates, specific gravity data, or phthalate biomarker concentrations (n = 11) (described below); and those for whom samples were collected on or after the delivery date (n = 41). We further excluded participants without complete CBCL/1½–5 data (n = 1590). Finally, after excluding three cohorts with small sample sizes (n < 30), the final sample consisted of 2617 mother–child dyads from 13 cohorts. This included all singleton births (n = 2596), and for families with twins or multiple siblings, one randomly selected twin or only the eldest child was included (n = 21). Information on the study characteristics of these 13 cohorts is provided in Supplementary Table S1. Over 90% of maternal urine samples were collected during the second and third trimesters of pregnancy. We analyzed data collected and harmonized by ECHO before October 2023.

The study protocol received approval from either the ECHO single Institutional Review Board (IRB) or the local IRB of each cohort. Each cohort was responsible for obtaining written informed consent from the pregnant participants and, if necessary, the children’s parents or guardians. The Johns Hopkins Bloomberg School of Public Health IRB reviewed and approved the work of the ECHO Data Analysis Center (DAC).

2.2. Quantification of urinary metabolites of phthalates and alternative plasticizers

Each cohort stored urine samples at −80 °C and shipped aliquots of urine samples from pregnant participants on dry ice to the Human Health Exposure Analysis Resource (HHEAR) laboratory for the quantification of phthalate and alternative plasticizer metabolites. Detailed methods on sample preparation and instrumental analysis are described elsewhere (Zhu et al., 2021). Briefly, 500 μL urine aliquots were spiked with an isotope-labeled internal standard mixture and β-glucuronidase/arylsulfatase. After incubation, the samples were extracted using ABS ELUT-Nexus cartridges (Varian, Walnut Creek, CA, USA), concentrated, and reconstituted. The target compounds were analyzed using an ExionLC system (SCIEX, Redwood City, CA, USA) coupled with an AB SCIEX QTRAP 5500+ triple quadrupole mass spectrometer (Applied Biosystems, Foster City, CA, USA) equipped with an electrospray ionization source in both positive and negative modes. A total of 33 metabolites were quantified, including 27 phthalate metabolites and six alternative plasticizer metabolites. Mono-n-butyl phthalate (MnBP) and mono-isobutyl phthalate (MiBP) co-eluted and were therefore quantified as a composite. The limits of detection (LODs) for the metabolites ranged from 0.01 ng/mL to 0.35 ng/mL, except for phthalic acid (5 ng/mL).

Along with procedural blanks, two reagent blanks, matrix blanks, and matrix spikes were analyzed with each batch. As trace levels of several metabolites were detected in reagent blanks, the blank concentrations were subtracted from the study sample concentrations. Additionally, three replicates of HHEAR quality control (QC) pools (A/B), one of each Standard Reference Material (SRM3672/SRM3673, NIST, Gaithersburg, MD, USA), and 136 pairs of blinded duplicate aliquots from participants were analyzed for quality control purposes (Kannan et al., 2021). The coefficients of variation (CVs) for HHEAR QC pools A and B ranged from 6% to 24%, except for mono 2-ethyl hexyl phthalate (MEHP) (50–55%) (Supplementary Table S2). Mean recoveries of analytes from SRM samples ranged from 86% to 119%. A relative percent difference (RPD=|sample result − repeat result|/[sample result + repeat result]*100) was calculated for each duplicate aliquot pair when a metabolite was detected above the LOD in both duplicates, with RPDs ranging from 5% to 26% (Supplementary Table S2).

2.3. Child behavior assessment

The CBCL/1½–5, also known as the CBCL preschool form, is a widely used questionnaire for assessing children’s emotional and behavioral problems, with reliability and validity established across multiple cultures (Achenbach, 2000; Campagna et al., 2023). Caregivers score 99 items on a three-point Likert scale (0 – not true, 1 – somewhat/sometimes true, or 2 – very/often true), and the scores are summed to yield seven syndrome scales. These syndrome scale raw scores are further combined to compute two broadband scales: internalizing scores (derived from anxious/depressed behaviors, emotionally reactive behaviors, somatic complaints, and withdrawn behaviors) and externalizing scores (derived from aggressive behaviors and attention problems). The broadband scale raw scores are normalized by sex and age to generate T-scores (mean = 50, standard deviation [SD] = 10). Higher T-scores indicate more emotional and behavioral problems in children. The internalizing and externalizing T-scores were used in the analyses.

The timing and frequency of CBCL 1½–5 assessments varied by cohort. Among the 2617 mother–child dyads, 1040 children from eight cohorts were administered the CBCL/1½–5 multiple times between ages 1.5–5 years: 739 children had two measurements, 292 had three, and 9 had four. All repeated CBCL/1½–5 measurements were included in the analyses, as the intraclass correlation coefficients (~0.6) indicated moderate reproducibility.

2.4. Covariates

We a priori constructed a directed acyclic graph to identify confounders, precision variables for outcomes, and potential mediators in the pathway from prenatal exposure to phthalates and alternative plasticizers to child emotional and behavioral outcomes (Textor et al., 2016) (Supplementary Fig. S2). The final covariate set included cohort, maternal race/ethnicity (Hispanic, multiple/other races, non-Hispanic Black, non-Hispanic White) as a proxy for structural racism and discrimination, maternal education (high school or less, some college or no/associate’s degree, bachelor’s degree, graduate or professional degree) as a measure of maternal socioeconomic status (SES), maternal marital status (married or living with a partner, single or not living with a partner), maternal age at specimen collection (in years), maternal pre-pregnancy body mass index (BMI; in kg/m²), maternal tobacco use during pregnancy (no, yes), parity (0, 1, ≥2), breastfed ever (no, yes), child sex (female, male), year of birth (in years), and age at CBCL assessment (in years). We excluded two potential mediators (gestational age at birth and pregnancy complications, i.e., gestational diabetes, hypertension, or preeclampsia) from the covariate set.

2.5. Statistical analysis

Prior to the statistical analysis, metabolite concentrations below the LOD were imputed as the LOD divided by the square root of two. To account for urinary dilution, all metabolite concentrations were adjusted for specific gravity (SG) using the following equation (Kuiper et al., 2021): C_sg = C × (SG_median – 1)/(SG – 1), where C_sg is the SG-adjusted concentration, C is the measured concentration, SG_median is the median SG value for each cohort, and SG is the measured SG value. We summarized the distributions of SG-adjusted metabolite concentrations of phthalates and alternative plasticizers and calculated Spearman correlation coefficients among the metabolites detected > 50% of the samples.

As several metabolites originate from common parent compounds and are therefore highly correlated, we computed molar sums (nmol/mL) for DEHP, di-(2-propylheptyl) phthalate (DPHP), and DiNP metabolites using the metabolites detected in over 70% of the samples. The molar sum for DINCH metabolites was also computed because at least one metabolite was detected in over 70% of the samples. The equations used to calculate the molar sums are provided in Supplementary Table S3. The SG-adjusted molar sums and phthalate metabolites detected in more than 70% of the samples were modeled as continuous variables. They were log2 transformed to reduce the influence of outliers and scaled by the interquartile range (IQR) to improve interpretability across metabolites. For metabolites detected in 50% to 70% of the samples, we modeled them as three-level categorical variables: Group 1 (non-detectable levels; reference), Group 2 (lower detectable levels; below the median of detectable values), and Group 3 (higher detectable levels; above the median of detectable values). Metabolites detected in less than 50% of the samples were excluded from the analyses.

To examine associations between individual metabolites and internalizing and externalizing T-scores, we used linear mixed-effects models with a random effect for child, accounting for the repeated CBCL/1½–5 measurements within each child. Given that missing data for all covariates were below 20%, we imputed missing observations using multiple imputation by chained equations including all exposures, outcomes, and covariates. We imputed 100 datasets and combined effect estimates using Rubin’s rules (Rubin, 1987). To assess the robustness of the findings, we performed leave-one-cohort-out analyses by excluding each cohort one at a time to evaluate that cohort’s influence on the overall results.

To investigate associations between a metabolite mixture and internalizing and externalizing T-scores, we employed quantile-based g-computation. This approach estimates the joint effect of simultaneously increasing all exposures in the mixture and does not assume directional homogeneity, allowing metabolites to have positive or negative weights (Keil et al., 2020). For the quantile–based g–computation models, we divided each metabolite into four ordered categories. If a metabolite was detected in ≥75% of samples, its concentrations were categorized into quartiles. if detection was below 75%, values below the LOD were assigned to the lowest category, and the remaining values at or above the LOD were split into three equal-sized groups. The overall effect, represented by psi (ψ), was pooled across imputed datasets, and median weights were calculated to reflect the relative contribution of each exposure to the mixture effect.

To evaluate which syndrome scales might be driving the broadband scale associations, we ran secondary analyses limited to exposure–outcome pairs that showed meaningful associations for internalizing and externalizing T-scores. Given the truncation and pronounced skewness of the syndrome scale T–scores, we instead analyzed the raw syndrome scale scores, treating them as count data. We fitted covariate-adjusted zero inflated negative binomial models and presented the results as count ratios (CRs) by exponentiating the estimates.

Considering the sex-specific effects of phthalates as endocrine disruptors (Jankowska et al., 2021), we tested whether child sex modified the associations between individual phthalate metabolites and internalizing and externalizing T-scores. We performed sex-stratified analyses and assessed the interaction between sex and each metabolite added to the model. As sex-dependent confounding was observed, we fitted an augmented product term model, which included not only the interaction term between sex and each metabolite but also interaction terms between sex and all covariates (Buckley et al., 2017).

As an exploratory analysis, we also evaluated effect modification by the COI at birth, an index of overall neighborhood factors, given that neighborhood conditions may influence the harmful effects of environmental exposures (Barrett and Padula, 2019; Payne-Sturges et al., 2023). We geospatially linked the maternal residential address at birth to the census tract-level COI, calculated by averaging scores across three domains: education, health and environment, and social and economic (Acevedo-Garcia et al., 2020). The nationally normed z-score ranged from 0 to 1, with higher values indicating greater neighborhood opportunity, and was subsequently categorized into tertiles (low, medium, and high). For the effect modification analysis, we stratified the analyses between individual phthalate metabolites and internalizing and externalizing T–scores by each COI tertile and assessed the interaction between COI and each metabolite by additionally including the interaction term in the model.

A cut-off for statistical significance was a p-value of less than 0.05 for the main and stratified analyses, and less than 0.1 was used to indicate statistically significant effect modification using p-values from the interaction term model. We did not apply multiple comparisons adjustments, as avoiding such adjustments can reduce errors in interpretation by focusing on actual observations of nature (Rothman, 1990). All statistical analyses were conducted using R version 4.4.4 (R Foundation for Statistical Computing, Vienna, Austria).

3. Results

3.1. Sociodemographic characteristics and CBCL/1½–5 internalizing and externalizing T-scores

Pregnant women participating in this study were 45% non-Hispanic White, 24% non-Hispanic Black, 21% Hispanic, and 11% multiple or other races (Table 1). A majority of the mothers held a bachelor’s degree or higher (51%), were married or cohabitating (76%), and had at least one previous pregnancy (55%). Most did not experience pregnancy complications (87%) and abstained from tobacco use during pregnancy (93%). The median (25th, 75th percentiles) maternal age at prenatal urine collection was 31 (26, 34) years. Children in the study were 51% male, and 54% were born in or after 2016, with a median (25th, 75th percentiles) gestational age at birth of 39 (38, 40) weeks Most mothers initiated breastfeeding (88%). Prenatal maternal urine samples were collected between 2007 and 2021. The largest cohort, Conditions Affecting Neurocognitive Development and Learning in Early Childhood (CANDLE), collected samples in the earlier years (2007–2011), while the second- to fifth-largest cohorts collected samples in the later years (2014–2021) (Supplementary Table S4). The mean (SD) CBCL/1½–5 T-scores across all observations (n = 3967) were 44 (11) for internalizing behaviors and 45 (10) for externalizing behaviors, with no substantial variation across cohorts. The sociodemographic characteristics of the included mother–child dyads in this study were generally similar to those of the excluded dyads (n = 11,809), except that there were higher proportions of non-Hispanic White and Black pregnant women, lower proportions of Hispanic women, and those with higher education levels among the included group (Supplementary Table S5).

Table 1.

Sociodemographic characteristics and CBCL/1½–5 T-scores among 2617 ECHO mother-child dyads.

Characteristics	Frequency (%)a
Maternal race/ethnicity
Non-Hispanic White	1170 (45%)
Non-Hispanic Black	621 (24%)
Hispanic	553 (21%)
Multiple/other racesb	<280 (11%)
Missing	<5
Maternal education
Less than high school, high school degree, GED or equivalent	783 (31%)
Some college, no/associate’s degree, trade school	462 (18%)
Bachelor’s degree	640 (25%)
Masters, professional, or doctorate degree	677 (26%)
Missing	55
Maternal marital status
Married or living with a partner	1935 (76%)
Single or not living with a partner c	618 (24%)
Missing	64
Parity
0	1132 (45%)
1	766 (30%)
≥2	617 (25%)
Missing	102
Pregnancy conditions (diabetes, hypertension, or preeclampsia)
No pregnancy conditions	2260 (87%)
Reported one of the pregnancy conditions	345 (13%)
Missing	12
Maternal tobacco use during pregnancy
No	2273 (93%)
Yes	163 (7%)
Missing	181
Child sex
Female	1273 (49%)
Male	1344 (51%)
Year of birth
2007	37 (1%)
2008	107 (4%)
2009	172 (7%)
2010	188 (7%)
2011	195 (8%)
2012	70 (3%)
2013	104 (4%)
2014	163 (6%)
2015	164 (6%)
2016	191 (7%)
2017	215 (8%)
2018	344 (13%)
2019	541 (21%)
2020	<120 (<5%)
2021	<10 (<1%)
Child Opportunity Index tertile
Low	770 (33%)
Medium	767 (33%)
High	792 (34%)
Missing	288
Ever breastfed
No	270 (12%)
Yes	1993 (88%)
Missing	354
Sample collection trimester
First	192 (7%)
Second	1599 (61%)
Third	826 (32%)
Characteristic	Mean (SD)
Pre-pregnancy BMI (in kg/m2; missing = 131)	27 (7)
Maternal age at specimen collection (in years)	30 (6)
Gestational age at birth (in weeks)	39 (2)
CBCL/1½–5 T-scores (all observations; n = 3967)
Internalizing	44 (11)
Externalizing	45 (10)

Abbreviations: BMI, body mass index; CBCL/1½–5, Child Behavior Checklist for Ages 1½–5; GED, general educational development; SD, standard deviation.

^aPercentage was calculated without missing observations.

^bOther races include Asian, native Hawaiian or other Pacific Islanders, American Indian or Alaska native, and multiple races.

^cThis category includes single, widowed, separated, divorced, never married, partnered, and not living together.

3.2. Phthalate/alternative plasticizer metabolite concentrations in prenatal maternal urine samples

Fourteen individual phthalate metabolites and a composite analyte of two metabolites were detected in greater than 70% of the samples: five DEHP metabolites, two DPHP metabolites, two DiNP metabolites, monoethyl phthalate (MEP), a composite of MnBP and MiBP (MnBP/MiBP), mono-carboxy isononyl phthalate (MCiNP), mono-benzyl phthalate (MBzP), mono (3-carboxypropyl) phthalate (MCPP), and mono(7-carboxyheptyl) phthalate (MCHPP) (Table 2). Three metabolites of an alternative plasticizer DINCH were detected in 47% to 67% of the samples. The highest median SG-adjusted urinary concentrations were observed for MEP (37.4 ng/mL), MnBP/MiBP (21.9 ng/mL), mono (2-ethyl-5-hydroxyhexyl) phthalate (MEHHP) (7.8 ng/mL), mono-(2-ethyl-5-carboxypentyl) phthalate (MECPP) (7.2 ng/mL), MBzP (6.3 ng/mL), mono-carboxy isooctyl phthalate (MCiOP) (5.0 ng/mL), and mono (2-ethyl-5-oxohexyl) phthalate (MEOHP) (4.1 ng/mL), all of which were detected in over 95% of the samples. Compared with other cohorts, three cohorts (CANDLE, Early Autism Risk Longitudinal Investigation, and Markers of Autism Risk in Babies – Learning Early Signs), which collected samples in earlier years, had higher median concentrations for DEHP metabolites, DiNP metabolites, MBzP, and MCHPP (Supplementary Table S6).

Table 2.

Specific gravity-adjusted urinary concentrations of phthalate and alternative plasticizer metabolites among 2617 ECHO pregnant participants.

Parent compound	Metabolites a	LOD (ng/mL)	% >LOD	Percentile (ng/mL)
				25th	50th	75th
Phthalates
Di-2-ethylhexyl phthalate (DEHP)	Mono (2-ethyl-5-hydroxyxhexyl) phthalate (MEHHP)	0.07	>99%	4.19	7.75	15.8
	Mono-(2-ethyl-5-carboxypentyl) phthalate (MECPP)	0.35	99%	3.89	7.23	14.8
	Mono (2-ethyl-5-oxohexyl) phthalate (MEOHP)	0.05	>99%	2.21	4.05	8.05
	Mono-2-(carboxymethyl) hexyl phthalate (MCMHP)	0.18	82%	0.51	1.36	3.32
	Mono 2-ethyl hexyl phthalate (MEHP)	0.09	72%	<LOD	1.16	3.57
	∑DEHP b			12.8	23.3	48.3
Di-(2-propylheptyl) phthalate (DPHP)	Mono-2-(propyl-6-carboxy-hexyl)-phthalate (MPCHP)	0.05	99%	0.32	0.58	1.13
	Mono-2-(propyl-6-oxoheptyl)-phthalate (MPOHP)	0.04	84%	0.11	0.27	0.65
	Mono-2-(propyl-6-hydroxy-heptyl)-phthalate (MPHHP)	0.06	31%	<LOD	<LOD	0.16
	∑DPHP b			0.58	1.03	1.95
Di-iso-nonyl phthalate (DiNP)	Mono-carboxy isooctyl phthalate (MCiOP)	0.05	98%	2.32	5.00	13.4
	Monohydroxy-iso-nonyl phthalate (MHiNP)	0.12	74%	<LOD	0.36	1.13
	∑DiNP b			2.52	5.33	14.0
Diethyl phthalate (DEP)	Monoethyl phthalate (MEP)	0.08	>99%	16.3	37.4	109
Dibutyl phthalate (DBP), di-iso-butyl phthalate (DIBP)	Composite of mono-n-butyl phthalate and mono-isobutyl phthalate (MnBP/MiBP)	0.15	99%	12.1	21.9	39.6
Di-iso-decyl phthalate (DIDP)	Mono-carboxy isononyl phthalate (MCiNP)	0.06	98%	0.43	0.87	1.78
Butylbenzyl phthalate (BBzP)	Mono-benzyl phthalate (MBzP)	0.08	96%	2.32	6.33	16.7
Di-n-octyl phthalate (DnOP)	Mono (3-carboxypropyl) phthalate (MCPP) c	0.06	89%	0.45	0.88	1.74
Di-n-octyl phthalate (DnOP)	Mono(7-carboxyheptyl) phthalate (MCHPP) c	0.23	84%	0.53	1.38	4.15
Dimethyl phthalate (DMP)	Mono-methyl phthalate (MMP)	0.32	63%	<LOD	1.42	3.26
Diheptyl phthalate (DHPP)	Mono-2-heptyl phthalate (MHPP)	0.05	62%	<LOD	0.13	0.41
Di-n-hexyl phthalate (DnHP)	Mono-hexyl phthalate (MHxP)	0.04	56%	<LOD	0.07	0.17
Di-iso-propyl phthalate (DiPP)	Mono-isopropyl phthalate (MiPP)	0.10	27%	<LOD	<LOD	0.17
Alternative plasticizer
Di-iso-nonyl-cyclohexane-1,2-dicarboxylic acid	Cyclohexane-1,2-dicarboxylic acid mono hydroxyisononyl ester (MHNCH)	0.04	67%	<LOD	0.12	0.37
(DINCH)	Cyclohexane-1,2-dicarboxylic acid mono(oxo-isononyl) ester (MONCH)	0.05	61%	<LOD	0.09	0.21
	Cyclohexane-1,2-dicarboxylic acid mono carboxyisooctyl ester (MCOCH)	0.06	47%	<LOD	<LOD	0.18
	∑DINCH b			0.16	0.31	0.74

Abbreviations: LOD, limit of detection.

^aPhthalate metabolites detected in less than 25% of samples were excluded from the table: monohydroxy-iso-decyl phthalate (MHiDP, LOD = 0.16); mono-isodecyl phthalate (MiDP, LOD = 0.06); mono-propyl phthalate (MPrP, LOD = 0.01); mono-n-octyl phthalate (MOP, LOD = 0.07); mono-pentyl phthalate (MPeP, LOD = 0.04); mono-ethyl terephthalate (METP, LOD = 0.08); mono-tert-butyl terephthalate (MTBTP, LOD = 0.08); mono-benzyl terephthalate (MBzTP, LOD = 0.09); phthalic acid (PA, LOD = 5.00).

^bThe molar sums were calculated by dividing the concentration of each analyte by its molecular weight, summing them, and then multiplying the molar sum by the molecular weight of a primary analyte (i.e., MECPP, MPOHP, MHiNP, and MONCH) to express the molar sum on the same units as other analytes.

^cMCPP and MCHPP are primarily metabolites of DNOP but can also be metabolites of other high molecular weight phthalates.

Most SG-adjusted urinary phthalate metabolite concentrations exhibited weak to moderate positive correlations with each other (Spearman correlation coefficients [r_sp] = 0.16 to 0.69), except for mono-2-(propyl-6-oxoheptyl)-phthalate (MPOHP) and mono-methyl phthalate (MMP), which showed little correlations (r_sp = −0.15 to 0.15) (Supplementary Fig. S3). Stronger correlations were observed among the four DEHP metabolites (MEHHP, MECPP, MEOHP, and MEHP) (r_sp = 0.63 to 0.97), between mono-2-(propyl-6-carboxy-hexyl)-phthalate and MCiNP (r_sp = 0.88), and between MCHPP and monohydroxy-iso-nonyl phthalate (r_sp = 0.83). The three DINCH metabolites were moderately correlated with each other (r_sp = 0.64 to 0.73) but showed weak or little correlations (both positive and negative) with most phthalate metabolites (r_sp = −0.31 to 0.28).

3.3. Associations between phthalate and alternative plasticizer metabolites and CBCL/1½–5 internalizing and externalizing scores

An IQR increase in MBzP concentrations was associated with a 0.53-point (95% confidence interval [CI]: 0.05, 1.00) increase in externalizing T-scores (Fig. 1 and Supplementary Table S7). For mono-hexyl phthalate (MHxP), being in Group 3, compared with Group 1, was associated with a 1.23-point (95% CI: 0.35, 2.12) increase in externalizing T-scores, while Group 2 showed a marginal increase relative to Group 1 (β_ext = 0.77, 95% CI: −0.04, 1.58). Additionally, an IQR increase in ∑DEHP (β_ext = 0.50, 95% CI: −0.01, 1.01) and membership in Group 3 for mono-2-heptyl phthalate (MHPP) (versus Group 1; β_ext = 0.94, 95% CI: −0.01, 1.88), were marginally associated with higher externalizing T-scores. Conversely, an IQR increase in ∑DiNP and MCHPP was associated with lower internalizing T-scores, respectively (β_int = −0.45, 95% CI: −0.91, 0.02 for ∑DiNP; β_int = −0.76, 95% CI: −1.22, −0.30 for MCHPP). Other phthalate metabolites, as well as the molar sum of an alternative plasticizer DINCH, were not associated with behavioral scores. A metabolite mixture of phthalates and the alternative plasticizer DINCH was associated with lower internalizing T-scores (ψ = −0.70, 95% CI: −1.44, 0.05), largely driven by MCHPP, but not with externalizing T-scores (ψ = 0.29, 95% CI: −0.43, 1.00) (Supplementary Table S8).

Fig. 1.

Associations of prenatal maternal urinary metabolite concentrations of phthalates and the alternative plasticizer DINCH with CBCL/1½–5 internalizing and externalizing T-scores among 2617 ECHO mother-child dyads. Point estimates indicate beta coefficients, and error bars indicate 95% confidence intervals. Numeric data on beta coefficients and 95% confidence intervals are provided in Supplementary Table S7. Continuous exposure variables (i.e., specific gravity-adjusted molar sums and phthalate metabolites detected in > 70% of the samples) were log2-transformed and scaled by interquartile range. For categorical exposure variables (i.e., phthalate metabolites detected in 50–70% of the samples), Groups 1, 2, and 3 represent non-detectable levels, lower detectable levels (below the median of detected values), and higher detectable levels (above the median of detected values), respectively. Adjusted models included cohort, maternal race/ethnicity, maternal education, maternal marital status, maternal age at specimen collection, maternal pre-pregnancy body mass index, maternal tobacco use during pregnancy, parity, ever breastfed, and child sex, year of birth, and age at CBCL/1½–5 assessment. Abbreviations: CBCL/1½–5, Child Behavior Checklist for Ages 1½–5; CI, confidence interval; DEHP, di-2-ethylhexyl phthalate; DINCH, di-iso-nonyl-cyclohexane-1,2-dicarboxylic acid; DiNP, di-iso-nonyl phthalate; DPHP, di-(2-propylheptyl) phthalate; ECHO, Environmental influences on Child Health Outcomes; IQR, interquartile range; MBzP, mono-benzyl phthalate; MCHPP, mono(7-carboxyheptyl) phthalate; MCiNP, mono-carboxy isononyl phthalate; MCPP, mono (3-carboxypropyl) phthalate; MEP, monoethyl phthalate; MHPP, mono-2-heptyl phthalate; MHxP, mono-hexyl phthalate; MMP, mono-methyl phthalate; MnBP/MiBP, composite of mono-n-butyl phthalate and mono-isobutyl phthalate.

From the leave-one-cohort-out analyses, the associations of MBzP and MHxP Group 3 with higher externalizing T-scores remained robust (Supplementary Table S9). However, the associations of ∑DEHP and MHPP Group 3 with higher externalizing T-scores substantially moved toward the null when the Illinois Kids Development Study (IKIDS) cohort was excluded. Similarly, the associations of MCHPP and ∑DiNP with lower internalizing T-scores disappeared after excluding the CANDLE cohort.

Secondary analyses of syndrome scale raw scores revealed that the associations of MBzP and MHxP Group 3 with higher externalizing scores might be contributed by both the aggression and attention scales (CR = 1.03 − 1.04 per IQR increase in MBzP; CR = 1.13 − 1.14 for MHxP Group 3) (Supplementary Table S10). For ∑DEHP, MHPP Group 3, and MHxP Group 2, the aggression scale appear to drive the association with externalizing scores (CR = 1.03 per IQR increase in ∑DEHP; CR = 1.10 for MHPP Group 3; CR = 1.13 for MHxP Group 2). The associations of ∑DiNP and MCHPP with lower internalizing scores were likely driven by the anxiety/depression scale (CR = 0.97 per IQR increase in both) and, for MCHPP, additionally by the somatic complaints scale (CR = 0.98).

3.4. Effect modification analyses

Associations between MMP Group 2 and MHxP Groups 2 and 3 with externalizing T-scores were modified by child sex (interaction p < 0.09) (Fig. 2 and Supplementary Table S7). Among female children, these metabolites were associated with higher externalizing T-scores (β_ext = 0.64, 95% CI: −0.41, 1.69 for MMP Group 2; β_ext = 1.74, 95% CI: 0.60, 2.88 for MHxP Group 2; β_ext = 2.48, 95% CI: 1.26, 3.71 for MHxP Group 3), while male children showed null associations (β_ext = −0.68, 95% CI: −1.79, 0.43 for MMP Group 2; β_ext = −0.09, 95% CI: −1.25, 1.07 for MHxP Group 2; β_ext = 0.17, 95% CI: −1.11, 1.46 for MHxP Group 3). In contrast, males showed a positive association between MnBP/MiBP and externalizing T-scores (β_ext per IQR increase = 0.54, 95% CI: −0.08, 1.16), while females showed a null association (β_ext = −0.16, 95% CI: −0.71, 0.39) (interaction p < 0.10). Additionally, the associations of ∑DEHP, MBzP, and MHPP Group 3 with higher externalizing T-scores were statistically significant only among male children (β_ext = 0.85, 95% CI: 0.11, 1.59 for ∑DEHP; β_ext = 0.83, 95% CI: 0.14, 1.52 for MBzP; β_ext = 1.82, 95% CI: 0.44, 3.20 for MHPP Group 3), although no effect modification by sex was observed (interaction p > 0.12). We observed no sex-dimorphic associations between metabolites and internalizing T-scores.

Fig. 2.

Associations of prenatal maternal urinary metabolite concentrations of phthalates and the alternative plasticizer DINCH with CBCL/1½–5 internalizing and externalizing T-scores, stratified by child sex. Point estimates indicate beta coefficients, error bars indicate 95% confidence intervals, and shaded areas indicate a p-value for interaction term between child sex and phthalate metabolites < 0.1. Numeric data on beta coefficients and 95% confidence intervals as well as p-values for interaction terms are presented in Supplementary Table S7. Continuous exposure variables (i.e., specific gravity-adjusted molar sums and phthalate metabolites detected in > 70% of the samples) were log2-transformed and scaled by interquartile range. For categorical exposure variables (i.e., phthalate metabolites detected in 50–70% of the samples), Groups 1, 2, and 3 represent non-detectable levels, lower detectable levels (below the median of detected values), and higher detectable levels (above the median of detected values), respectively. Adjusted models included cohort, maternal race/ethnicity, maternal education, maternal marital status, maternal age at specimen collection, maternal pre-pregnancy body mass index, maternal tobacco use during pregnancy, parity, ever breastfed, and child year of birth and age at CBCL/1½–5 assessment. CBCL/1½–5, Child Behavior Checklist for Ages 1½–5; CI, confidence interval; DEHP, di-2-ethylhexyl phthalate; DINCH, di-iso-nonyl-cyclohexane-1,2-dicarboxylic acid; DiNP, di-iso-nonyl phthalate; DPHP, di-(2-propylheptyl) phthalate; mono-benzyl phthalate; IQR, interquartile range; MCHPP, mono(7-carboxyheptyl) phthalate; MCiNP, mono-carboxy isononyl phthalate; MCPP, mono (3-carboxypropyl) phthalate; MEP, monoethyl phthalate; MHPP, mono-2-heptyl phthalate; MHxP, mono-hexyl phthalate; MMP, mono-methyl phthalate; MnBP/MiBP, composite of mono-n-butyl phthalate and mono-isobutyl phthalate.

The COI modified associations between several phthalate metabolites and behavioral scores, with effects varying by group (Supplementary Fig. S4 and Table S11). Among children in the high-COI group (more neighborhood opportunity), MHPP Group 3 was associated with higher externalizing T-scores (β_ext = 2.65, 95% CI: 1.01, 4.29), while children in the low-COI group showed null but inverse associations (β_ext = −0.44, 95% CI: −2.33, 1.45; interaction p = 0.02). MEP was associated with higher externalizing T-scores among children in the medium-COI group (β_ext = 0.98, 95% CI: 0.05, 1.91), whereas it was associated with lower T-scores among the low-COI group (β_ext = −0.44, 95% CI: −1.35, 0.47; interaction p = 0.02). Conversely, inverse associations were observed in the medium-COI group between MMP Group 2 and externalizing T-scores (β_ext = −0.97, 95% CI: −2.33, 0.40) and between MHPP Group 2 and internalizing T-scores (β_int = −1.69, 95% CI: −3.19, −0.19), while null but positive associations were found in the low-COI group (interaction p = 0.03–0.09).

4. Discussion

This large-scale study, conducted in a geographically and sociodemographically diverse U.S. population, provided robust evidence that higher prenatal urinary concentrations of MBzP and MHxP were associated with small to modest increases in externalizing behaviors, reflected in more aggression and attention problems, in children aged 1.5–5 years. DEHP metabolites and MHPP Group 3 (higher detectable levels) were linked to more externalizing behaviors, likely driven by aggression. However, these associations lacked robustness and seemed to be driven by one specific cohort based on the results of the leave-one-cohort-out analyses. Similarly, the associations between DiNP metabolites and MCHPP with fewer internalizing behaviors appeared to be driven by anxiety/depression and, respectively, influenced by a different specific cohort. However, these two specific cohorts influencing the associations did not substantially differ from the overall study sample in these metabolite concentrations, behavioral T-scores, or the child’s age at assessment. Additionally, MCHPP contributed substantially to the marginal inverse association between the metabolite mixture and internalizing behaviors. Child sex modified several associations: males exhibited a positive association between MnBP/MiBP and externalizing behaviors, while females showed a null association. In contrast, adverse associations of MMP Group 2 (lower detectable levels) and MHxP with externalizing behaviors were stronger in females than males. The evidence for effect modification by a measure of neighborhood opportunity was inconsistent. Notably, no associations were found between prenatal exposure to the alternative plasticizer DINCH and child behaviors.

Many prior epidemiological studies have linked gestational exposure to certain phthalates with increased behavioral problems, particularly externalizing behaviors, in young children, often in a sex-specific manner. Consistent with our findings, several studies reported adverse associations between prenatal MBzP and behavioral scores, while MHxP and DiNP metabolites were less studied. For example, the ECHO-PATHWAYS study, which analyzed 1694 mother–child dyads from three U.S. cohorts, reported that MBzP, MHxP, and a phthalate mixture were associated with higher total problem scores assessed by the CBCL/1½–5, while MCiOP (a DiNP metabolite) was linked to lower scores (Barrett et al., 2024). The associations were more pronounced among males, although the statistical significance of the interaction was not evaluated. This study quantified phthalate metabolites in a different laboratory during a separate time period, using different analytical methods, and did not quantify alternative plasticizer metabolites. Additionally, two of the three cohorts in the ECHO-PATHWAYS study (CANDLE and Global Alliance to Prevent Prematurity and Stillbirth [GAPPS]) were part of our study (31% of our analytic sample). The cohort-specific analyses for the remaining third cohort in the ECHO-PATHWAYS study similarly reported an association between MBzP and externalizing behaviors at ages 4–5 years in males only, although the researchers did not measure MHxP and DiNP metabolites (Day et al., 2021). They observed that a phthalate mixture in late pregnancy, weighted with MBzP and MCPP, was linked to more externalizing behaviors in males but not in females. Additionally, prenatal MBzP concentrations were associated with higher scores for externalizing behaviors and attention-deficit hyperactivity disorder (ADHD) symptoms, particularly among males, in studies of preschool children in Canada and Denmark (Andreasen et al., 2023; England-Mason et al., 2020). Together with our findings in the ECHO Cohort, these studies indicate a consistent pattern linking prenatal exposure to butylbenzyl phthalate (BBzP) to more externalizing behaviors in young children, especially in males.

Limited studies, including ours and the ECHO-PATHWAYS study, have evaluated the association between MHxP and children’s emotional and behavioral outcomes, as MHxP has not been measured in most other research. In several U.S. studies, MHxP was detected in a higher proportion of pregnant or preconception participants (>75%) (Barrett et al., 2024; Nobles et al., 2023) compared to our study (56%), while a study in New York City reported a comparable detection frequency (54%) (Cowell et al., 2023). The parent compound of MHxP, di-n-hexyl phthalate (DnHP), has been restricted by the U.S. Consumer Product Safety Commission for use in children’s toys and childcare articles due to its potential contribution to the cumulative risk from other anti-androgenic phthalates (Lioy et al., 2015). Additionally, DnHP is classified as a reproductive and developmental toxicant under the European Union Classification, Labelling and Packaging Regulation (Dekant and Bridges, 2016). Studies in rats have shown that DnHP induced thyroid hyperactivity and histological changes, which can adversely affect neurodevelopment (Bereketoglu and Pradhan, 2022; Boas et al., 2006; Howarth et al., 2001). Therefore, further research investigating the neurobehavioral impacts of DnHP is needed to confirm our findings.

Epidemiological evidence for other phthalate metabolites is relatively less consistent. While our study observed prenatal DEHP exposure was associated with higher externalizing behavior scores in males but not in females, it was more strongly linked to higher scores in females than in males in Canadian children aged 2 years (Dewey et al., 2023). DEHP metabolites were associated with higher internalizing behavior scores in 2-year-old males but not in females in Israel (Cohen-Eliraz et al., 2023). Additionally, several studies reported associations for low-molecular-weight phthalates that we did not observe. For instance, MnBP, MiBP, MEP, and MMP were associated with more externalizing behaviors only among males aged 3–4 years in Canada (England-Mason et al., 2020) and among children aged 4–9 years in the U.S., with stronger associations in males (Engel et al., 2010). Prenatal MnBP concentrations were also associated with more internalizing behaviors among French males at age 3 years but not at 5 years (Philippat et al., 2017). These inconsistencies may be attributed to variations in phthalate metabolite concentrations, influenced by geographic regions and differences in sample collection periods and timing, as well as diverse sociodemographic characteristics across cohorts.

Externalizing behaviors found to be adversely associated with phthalate exposure in our study capture a broad range of behavioral problems, and as such, underlying biological processes may be difficult to elucidate. Nevertheless, experimental and epidemiological evidence suggests that in utero phthalate exposure may disrupt the neuroendocrine system (Lucaccioni et al., 2021; Rock and Patisaul, 2018). Phthalate exposure during pregnancy can interfere with thyroid hormone activity (Ghassabian and Trasande, 2018; Kim et al., 2019), which plays a critical role in fetal brain development (Bernal, 2005; Calza et al., 2015). In rodent studies, prenatal DEHP exposure altered neuroendocrine gene expression in the neonatal hypothalamus (Gao et al., 2018); reduced proliferation and neurogenesis in the fetal brain, leading to abnormal neuronal distribution in the newborn brain (Komada et al., 2016); and influenced the lipid metabolome in the fetal brain (Xu et al., 2007). Additionally, phthalates may disrupt the neurotransmitter system by affecting both inhibitory and excitatory neurotransmitters, which are essential for regulating neurodevelopmental processes (Minatoya and Kishi, 2021; Miodovnik et al., 2014; Rock and Patisaul, 2018). For instance, perinatal DEHP exposure has been shown to affect levels of the inhibitory neurotransmitter gamma-aminobutyric acid and the excitatory neurotransmitter aspartate in prepubertal rats (Carbone et al., 2012), as well as alter levels of monoamine neurotransmitters, such as dopamine, norepinephrine, and serotonin, in offspring (Kaimal et al., 2024; Yirun et al., 2021).

Prenatal exposure to alternative plasticizers, such as DINCH and DEHTP, is also capable of disrupting the endocrine system (Morales-Grahl et al., 2024; Yu et al., 2023) and have been linked to altered thyroid functions in pregnant women (Cathey et al., 2019; Derakhshan et al., 2021). As epidemiological evidence on the effects on child behavior is limited, we have expanded the literature review to include broader neurodevelopmental outcomes. Consistent with our findings, a previous epidemiological study in France reported that prenatal DINCH exposure was not linked to behavioral scores in 2-year-old children (Guilbert et al., 2021). While DEHTP was not quantified in our study, prenatal exposure to this alternative plasticizer was associated with poorer adaptive and cognitive functions at 6 months of age (Park et al., 2023) and higher overall problem scores at ages 4–6 years (Colicino et al., 2021), especially in males. Given the increasing exposure to these alternative plasticizers (Domínguez-Romero et al., 2023; Jiang et al., 2023; Kasper-Sonnenberg et al., 2019), further large-scale epidemiological studies are warranted to examine their associations with child behaviors, particularly focusing on compounds other than DINCH.

Children in low-opportunity neighborhoods often experience limited access to educational and health resources, reduced social support, and higher stress levels, which may increase their susceptibility to the neurodevelopmental effects of environmental exposures (Acevedo-Garcia et al., 2020; Barrett and Padula, 2019; Dickerson et al., 2023; Payne-Sturges et al., 2023). In our study, we observed both expected positive associations and unexpected inverse associations for several metabolites in low-opportunity neighborhoods. One possible explanation for the mixed findings is that chronic stress and other chemical and non-chemical stressors may mask the impact of phthalates (i.e., the impact of non-chemical stressors may be greater and more salient than the effect of chemical exposures), making the associations appear weaker. For example, the ECHO-PATHWAYS study observed antagonistic interactions between a phthalate mixture and prenatal stressful life events on child behavior (Barrett et al., 2024). Future studies on prenatal phthalate exposure and neurodevelopment should consider the influence of non-chemical stressors, such as neighborhood factors and prenatal maternal stress.

This study is not without limitations. First, we relied on a single spot or first morning urine sample per participant to characterize prenatal exposure to phthalates and alternative plasticizers. Due to the poor to moderate reproducibility of phthalate metabolites (Roggeman et al., 2022), concentrations from a single sample may not accurately reflect exposure levels throughout pregnancy. Second, the CBCL/1½–5 assessment may have been influenced by the language used for assessment, caregiver education level and sex, and child age (Zheng et al., 2024). Additionally, emotional and behavioral problems at a young age, especially internalizing behaviors such as anxiety, depression, and somatic complaints, may be difficult to capture and can change as the child develops (Basten et al., 2016; Kristensen et al., 2010; Shaw et al., 2005). Although we adjusted for the relevant variables and incorporated all repeated CBCL/1½–5 assessments in the analyses, these adjustments may not fully address potential limitations, which could contribute to the weak associations observed for internalizing behaviors. Furthermore, the findings may be specific to young children and may not generalize to other age groups. Future studies should investigate whether the associations we observed between prenatal phthalate exposure and behaviors, as well as the sex differences, persist at older ages. Third, the mother–child dyads in this study were not fully representative of the participating cohorts with respect to several demographic characteristics, such as maternal race/ethnicity and education level, which may limit the generalizability of our findings. Generalizability may also be constrained by exposure differences. Compared with women in comparable NHANES cycles, our pregnant participants showed lower geometric mean of MECPP, MCiOP, MCiNP, and MCPP concentrations but higher MEHHP, MEP, and MBzP concentrations in earlier cycles (2007–2014), with levels converging in later cycles (2015–2018) (Supplementary Table S12). More refined methods that account for cohort–level differences and geographic and temporal trends are needed to compare exposure distributions with other populations.

Despite these limitations, this study has notable strengths. It is the largest study to date to examine a wide range of phthalate and alternative plasticizer metabolites in prenatal urine samples in relation to child behavioral outcomes. A large number of samples collected from 13 ECHO cohorts were analyzed by a single laboratory using established protocols within a short period, minimizing measurement errors. Children’s emotional and behavioral outcomes were repeatedly assessed using a valid and reliable tool to capture fluctuations in outcomes. Furthermore, the study benefitted from a geographically and sociodemographically diverse study population, which included groups with varying phthalate and alternative plasticizer exposure levels during pregnancy and diverse behavioral profiles in children. The ECHO Cohort’s rich data enabled us to not only account for various covariates but also explore neighborhood SES as a potential effect modifier. Although the evidence for effect modification by neighborhood SES was weak and inconsistent across metabolites, its potential influence on the neurodevelopmental effects of prenatal phthalate exposure warrants further investigation.

5. Conclusion

In this large, geographically, and sociodemographically diverse sample, gestational exposure to BBzP and DnHP may be associated with a small-to-modest increase in externalizing behaviors but not internalizing behaviors in young children. Along with findings from the previous ECHO-PATHWAYS study, this is among the first studies to suggest potential adverse effects of DnHP exposure on child behaviors, which warrants further investigation in other populations. There was some evidence of associations of gestational exposure to DEHP and diheptyl phthalate with more externalizing behaviors and of gestational exposure to DiNP and di-n-octyl phthalate with fewer internalizing behaviors. However, these findings were not consistently replicated across cohorts and appeared to be influenced by specific cohorts, suggesting the need for additional research to confirm these relationships. Our data also indicate that prenatal exposure to DINCH, an alternative plasticizer, may not be associated with child behaviors. However, previous toxicological and epidemiological findings suggest that other alternative plasticizers not quantified in this study may be associated with child behaviors. With the increasing use of these novel plasticizers, future studies should explore their neurodevelopmental effects using more recently enrolled cohorts, ideally with repeatedly collected prenatal samples.

Supplementary Material

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envint.2025.109647.

Data availability

Select de-identified data from the ECHO Program are available through NICHD’s Data and Specimen Hub (DASH). Information on study data not available on DASH can be found on the ECHO study DASH webpage.