Early Life Phthalate and Replacement Plasticizer Exposures and Changes in Early Childhood Brain Functional Connectivity and Structural Morphology

Abstract

Phthalates and replacement plasticizers (PRPs) are ubiquitous exposures in daily life across all age ranges. Exposure to phthalates has been linked to changes in cognitive and behavioral development and associated with increased risk of some developmental disabilities. We examined the extent to which early life exposure to PRPs was associated with changes in connection strength of resting state functional networks or impacted structural morphologies of cortical regions of interest that underlie basic and higher order cognitions. We utilized the UNC Chapel Hill enrollment in the Baby Connectome Project, a longitudinal study of normative brain development of children between 2 weeks and 5 years of age. Non-sedated structural and resting state functional magnetic resonance imaging of the brain during natural sleep were obtained longitudinally, along with urine samples that were analyzed for 17 PRP metabolites. Using kernal weighted estimating equations and generalized linear models, we identified multiple PRP metabolites were associated with alterations in within-network connection strengths in the executive control and dorsal attention networks, with directionality often differing between boys and girls. PRP exposure among boys tended to be associated with lower functional connectivity, whereas PRP exposure among girls tended to be associated with higher functional connectivity. Among girls, MiBP metabolite concentrations were also significantly associated with cortical thinning in several regions of interest in the temporal lobe. Our results indicate that exposure to PRPs in early life has a measurable impact on the developmental trajectory of brain maturation, with potentially important differences by child sex.

Graphical Abstract

1. Introduction

Ortho-phthalate diesters (phthalates) and replacement plasticizers (together referred to as PRP hereafter) are ubiquitous exposures in daily life across all age ranges, having numerous sources including toys [1–3], personal [4–6] and baby care products [7–9], building materials [10], enteric coatings on medications [11–13], medical devices [14, 15], and the diet [16, 17], including among other dietary sources breastmilk [18] and formula [9, 19–21]. The main industrial function of phthalates is to make plastics soft and flexible, but they also function as solvents and lubricants, and as fixatives in fragrances, and are as a result widely incorporated into numerous food packaging and processing materials, household, and personal care products [22]. Changes in consumer product formulations have resulted in reduced exposure to some phthalates over time, such as di-2-ethylhexyl phthalate (DEHP) [23]. However newer replacement phthalates, and non-phthalate replacement plasticizers such as di(isononyl) cyclohexane-1,2-dicarboxylate (DINCH) and bis(2-ethylhexyl) terephthalate (DEHTP), have been introduced, with evidence of increasing human exposure over time [23–25].

In the United States and other middle and high income countries, exposure to many phthalates is virtually universal across all ages [26–29] with higher exposure to several phthalates in the youngest ages [29]. We recently reported widespread PRP exposure among infants and children between 2 weeks and 5 years of age, with virtually universal detection of many phthalate metabolites [30]. Moreover, urinary concentrations of monobutyl (MnBP), mono-3-carboxypropyl (MCPP), and monoisobutyl phthalate (MiBP) were significantly higher in the first year of life as compared to older ages [30], suggesting infancy-specific pathways of exposure.

Prenatal phthalate exposure has been extensively studied in relation to childhood cognitive, motor, visual, and behavioral development [31, 32], a literature encompassing dozens of studies in populations spanning multiple continents. However, despite extensive research over an almost 20-year period, there remains uncertainty as to the nature and extent of neurodevelopmental impacts, with systematic reviews noting inconsistencies across studies with respect to association magnitudes, directions, and the presence or absence of sex-specific effects. There is considerably less work on the consequences of postnatal exposure, and few studies with dense sampling of exposure in the earliest years of life. With respect to neurodevelopmental outcomes, the vast majority of the literature utilizes parent-reported behavioral inventories or administered assessments of cognitive function. Challenges in the reliability of parent-reported behaviors or symptoms, as well as constitutional determinants of child performance on administered tests (e.g. having adequate sleep and nutrition), may induce sufficient outcome variability to obscure any, presumably modest, underlying association.

Brain magnetic resonance imaging (MRI) is a safe and non-invasive [33], and sensitive technique to assess the brain structure and function, and when utilized longitudinally, can be used to chart the growth and development of the brain over time [34]. There is a small emerging literature incorporating brain imaging at a single point in time to examine consequences of prenatal and postnatal phthalate exposure [35], but with little overlap in child age or imaging modality to draw strong conclusions as to the presence of a robust signal. These innovative tools may provide novel insights into the neurodevelopmental consequences of exposure during developmental periods where other assessment methods have limited reliability. Using longitudinal structural MRI (sMRI), we recently reported that early life phthalate exposure was associated with changes in global measures of brain growth in the first years of life, with notable differences by child sex [36].

The infant brain has heightened vulnerability to toxic compounds due in part to developmental modulation of the blood brain barrier that results in higher penetration of toxic chemicals [37]. In addition, pathways responsible for metabolism and excretion of xenobiotics are not yet fully developed [38]. At the same time, the brain undergoes major developmental growth during the first years of postnatal life [39–44], with well-organized developmental processes establishing the foundation for cognitions that govern higher-order behaviors [45]. A large body of evidence supports that the human brain is organized into interconnected networks with correlated activity patterns, and that these networks play unique roles in cognition, sensory activity, and engagement with other brain networks [46]. The objective of the current study was to examine the impacts of early life exposure to PRPs on the development of these cognitions as measured by within-network functional connectivity of resting-state functional networks, as well as impacts on structural morphologies of cortical regions of interest that underlie basic and higher order cognitions.

2. Methods

2.1. Study Design

The Baby Connectome Project (BCP) is a longitudinal study of normative brain development in children between 2 weeks and 5 years of age, details of which have been previously published [47]. Enrollment into the BCP was conducted at the University of Minnesota and the University of North Carolina at Chapel Hill (UNC Chapel Hill). Eligibility criteria included singleton pregnancies delivered at term (37-42 weeks) with normal birthweight and no pregnancy complications. To limit participant attrition and provide a complete picture of the arc of brain development in early life, the BCP employed an accelerated cohort design, with infants entering observation in a staggered pattern and followed forward with longitudinal scans and inventories (Supplemental Figure 1). The current study utilizes enrollment at the UNC Chapel Hill site only, where urine collection was attempted at every participant visit (n = 267). Urine collection methods, sample processing, and storage have been previously described [30]. Participants were excluded from this analysis if they failed to donate urine at any visit (n = 60) or had no usable structural or resting-state functional MRI (n = 46), leaving a final sample size of 161 infants. This research was reviewed and approved by the University of North Carolina Office of Human Research Ethics (IRB #21-2221).

2.2. Measurement of phthalates and replacement plasticizers

Urinary chemical metabolite measurement methods have been previously described [30]. We used highly sensitive triple quadrupole mass spectrometry interfaced with a Vanquish ultra high performance liquid chromatography to perform targeted analysis of 17 metabolites of PRPs, including MnBP, MCPP, MiBP, monoethyl phthalate (MEP), monobenzyl phthalate (MBzP), mono-2- ethylhexyl phthalate (MEHP), mono-2-ethyl-5-hydroxyhexyl phthalate (MEHHP), mono-2-ethyl-5-oxohexyl phthalate (MEOHP), mono-2-ethyl- 5-carboxypentyl phthalate (MECPP), mono-2-carboxymethylhexyl phthalate (MCMHP), monooxononyl phthalate (MONP), monocarboxyisooctyl phthalate (MCOP), monocarboxyisononyl phthalate (MCNP), cyclohexane-1,2-dicarboxylic acid, monohydroxy isononyl ester (MHNCH), cyclohexane-1,2-dicarboxylic acid, monocarboxy isooctyl ester (MCOCH), mono-2-ethyl-5-hydrohexyl terephthalate (MEHHTP), and mono-2-ethyl-5-carboxypentyl terephthalate (MECPTP). In order to account for urinary dilution, specific gravity (SG) was measured using a hand-held refractometer (Reichert TS Meter Model TS400). Urine samples were randomly assigned to analytic batch, with all samples from a given participant analyzed in the same batch. In addition, each analytic batch contained procedural blank samples, two in-house control urine samples, and 2-3 aliquots of pooled urine quality control aliquots. External reference samples from National Institute of Standards and Technology (NIST, SRM 3673) were analyzed in every fourth analytical batch. Values below the limit of detection (LOD) were imputed as LOD/√2. Concentrations were standardized by SG to account for urinary dilution as described in Thistle et al. (2024) [30]. MONP, MCOP, and MCNP were infrequently detected (< 10%) and thus not included in any further analyses. For phthalate metabolites that devolved from the same parent compound (e.g. DEHP, DINCH, DEHTP), we calculated molar sums of metabolites: ΣDEHP = MEHP + MECPP + MEHHP + MEOHP + MCMHP; ΣDINCH = MCOCH + MHNCH; and ΣDEHTP = MECPTP + MEHHTP; in units of μmol/L, by summing SG-adjusted metabolite concentrations (with LOD/√2 imputation) after dividing by the molecular weight. In statistical analyses, concentrations of all metabolites or molar sums were log transformed after investigation of model fit statistics.

2.3. Brain Imaging

Infants and children were imaged during non-sedated sleep on the same 3T Siemens Prisma MRI scanner using a Siemens 32 channel head coil in the Biomedical Research Imaging Center at the UNC Chapel Hill. The imaging protocol included non-sedated sMRI (T1-weighted and T2-weighted), rsfMRI (resting state functional MRI), and diffusion MRI. Images were acquired during natural sleep. Detailed imaging procedures are described in Howell et al. [47].

Infant-dedicated surface-based MRI processing has been previously described in [48, 49]. Briefly, we used the iBEAT V2.0 toolbox to process brain structural images, reconstruct and parcellate cortical surfaces [50, 51] based on for cortical surface reconstruction, and then parcellated the cerebral cortex into 68 regions of interest following the FreeSurfer Desikan-Killiany atlas [52]. We then calculated surface areas and average cortical thickness of each region of interest. The rfMRI data were preprocessed using FSL [53], aligned with T1 images using boundary-based registration, motion artifacts were removed, and the resultant images were spatially smoothed. For resting state functional network construction, we used the Schaefer 100 whole-brain surface parcellation with 100 cortical regions [54]. A functional connectivity matrix was developed for each scan, by calculating pairwise Pearson’s correlation coefficients for each of the Schafer 100 cortical regions, and then grouped into the Yeo-7 network parcellation (executive control, default mode, dorsal attention, limbic, salience, sensorimotor, visual) [55]. Within network connection strength was calculated as the average correlation from all regions within a functional network. Framewise displacement was also measured as an index of head movement between frames and adjusted to account for scan effects.

2.4. Statistical Analysis

We performed association analyses between toxicants and brain structural and functional measures. All statistical analyses were performed in R version 4.4.0, using baseline packages alongside specific packages noted below. Brain measures included 7 resting-state within-network connectivity and 136 brain structural traits (surface area, cortical thickness for 68 cortical regions), all standardized prior to the above association analysis. Toxicant exposures included 8 measures (MnBP, MCPP, MiBP, MEP, MBzP, and the molar sums for DEHP, DINCH, and DEHTP). Due to the nonlinear age trends for brain measures, we first removed nonlinear age effects from each brain measure prior to association analyses, using the Generalized Additive Mixed Model [56] based on mgcv packages,

$$\mathrm{Y}\_\mathrm{i}\left(\mathrm{t}\right)=f\left(t\right)+b\_i+ϵ\_i\left(t\right),$$ (1)

where t is the age, Y_i(t) represents the brain measure, b_i denotes subject-specific random intercept, f(t) is the fixed effect which was estimated by cubic splines, and ϵ_i (t) is random noise. The residualized brain measures Y^~_i(t) were then obtained by removing fixed effect f(t) from Y_i (t).

Each subject had only a few longitudinal observations, and toxicant measurements and MRI assessments sometimes occurred at mismatched times, which together produced a sparse asynchronous longitudinal design. To estimate population-level concurrent (time-matched) associations without forcing visit alignment, we used the AsynchLong package [57], and fit a continuous-time linear model with kernel weighting [58, 59].

$$Y^{~}\_i\left(\mathrm{t}\right)=\mathrm{X}\_\mathrm{i}\left(t\right)\beta \_X+Z\_i\left(t\right)\beta \_Z+e\_i\left(t\right),E[e\_i(t)│X\_i(t\left)\right]=0,Cov[e\_i(t),e\_i(s)│X\_i(t\left)\right]=\mathrm{\Sigma }(t,s),$$ (2)

where X_i(t) is the exposure trajectory evaluated at time t using each subject’s actual toxicant sampling times, Z_i(t) contains time-varying or time-invariant covariates, and Σ(t,s) represents the time-dependent within-subject correlation. We implemented the Weighted Last Observation Carried Forward (WLV) estimator via AsynchLong, with an identity link, a Gaussian kernel and default bandwidth chosen automatically by the package’s default data-adaptive procedure, which weights observations by their temporal proximity between exposure and outcome measurements. This approach yields coefficient estimates, standard errors, and p-values based on two-sided Wald test that account for the irregular timing of visits. The key assumptions of this model are as follows: 1) the method targets a mean model with time-invariant regression effects; 2) it relies on the premise that the underlying covariate process is sufficiently smooth over time such that measurements closer to the outcome time are more informative than distant measurements, and 3) as with other approaches for irregular follow-up, inference can be less robust if measurement times are strongly informative (e.g. if visit timing depends on unobserved health status) or if covariate trajectories change abruptly relative to time gaps. Automatic bandwidth selection is intended to balance bias-variance in estimating the temporally aligned covariate contribution. All possible exposure-outcome pairs were tested according to Model (2), and the primary models were stratified by infant sex. We adjusted for infant age and maternal education as confounding covariates Z_i(t), which were identified via directed acyclic graphs. In sensitivity analyses, we repeated Model (2) for the entire population, adjusting for but not stratifying by infant sex. Multiple testing correction was performed using the Benjamini–Hochberg false discovery rate (FDR) method to obtain q-values to adjust 544 tests for surface area, 544 tests for cortical thickness, and 56 tests for functional networks, respectively. We also conducted a sensitivity analysis excluding participants older than 3 years, which reduces heterogeneity due to older ages and potentially larger time gaps between visits.

3. Results

We included participants in this analysis if they had at least one successful sMRI or rsfMRI and one successful urine donation concurrently or at any preceding time point (sMRI N = 159 infants, N = 409 scans; rsfMRI N = 129 infants, N = 362 scans). This resulted in a population majority female (male sex = 45.3%). The study population reported majority white race (77.6%), and non-Hispanic ethnicity (89.4%). This was also a highly educated population, with the majority of mothers reporting a college degree or higher educational level (74.5%). Almost half of the study population reported an annual household income of at least $100,000 per year. Consistent with eligibility criteria, infants in this study were full term, and healthy birthweight (Table 1).

Table 1.

Characteristics of Participants, University of North Carolina Baby Connectome Project (N = 161¹)

Characteristic	N (%)	Mean (SD)
Male Sex	73 (45.3)
Race
White	125 (77.6)
Black	18 (11.2)
Multi/Other	18 (11.2)
Non-Hispanic Ethnicity	144 (89.4)
Maternal Education
Less than College Degree	39 (24.2)
College Degree	42 (26.1)
Graduate or Professional degree	78 (48.4)
Total Household income (annual)
Less than $50,000	22 (13.7)
$50,000 - < $100,000	45 (28.0)
$100,000 or more	79 (49.1)
Gestational age (days)		276.9 (8.0)
Birthweight (grams)		3509.6 (420.4)

¹Included in this analysis are individuals who had at least one successful sMRI (n = 159) or rsfMRI (n = 129).

PRP metabolites were widely detectable (60-100%) (Table 2). Median concentrations of low molecular weight phthalates MnBP, MiBP and MEP were the highest among phthalates measured. Among replacement plasticizers, DEHTP metabolites, particularly MECPTP, was highly detected in child urines. Correlations among phthalate metabolites and their reproducibility over time have been previously described [30].

Table 2.

Phthalate and Replacement Plasticizer Biomarker Concentrations in the UNC Baby Connectome Project (n = 410 urine samples)

	Parent Compound	% Detect	LOD	Minimum	25th Pctl	50th Pctl	75th Pctl	Maximum	Geometric Mean	Geometric Standard Deviation
MEP ng/mL	Diethyl phthalate	83	4.107	<LOD	5.36	11.05	23.36	1711.44	13.01	3.45
MnBP ng/mL	Di-n-butyl phthalate	100	0.782	1.45	9.09	14.11	24.46	279.15	15.26	2.17
MiBP ng/mL	Diisobutyl phthalate	99	0.469	<LOD	6.68	11.20	22.10	538.80	12.74	2.81
MCPP ng/mL	Di-n-octyl phthalate Di-n-butyl phthalate	87	0.315	<LOD	1.45	2.48	4.50	24.31	2.18	2.98
MBzP ng/mL	Butyl Benzyl phthalate	100	0.093	0.47	2.54	5.08	10.09	186.35	5.43	2.77
MEHP ng/mL	Di(2-ethylhexyl) phthalate	88	0.295	<LOD	0.71	1.38	2.45	29.13	1.31	2.87
MEOHP ng/mL	Di(2-ethylhexyl) phthalate	100	0.092	0.06	2.32	3.72	6.83	91.02	3.98	2.38
MEHHP ng/mL	Di(2-ethylhexyl) phthalate	99	0.063	<LOD	3.43	5.40	9.42	164.01	5.73	2.61
MECPP ng/mL	Di(2-ethylhexyl) phthalate	100	0.012	0.54	4.90	8.88	15.81	260.82	9.32	2.44
MCMHP ng/mL	Di(2-ethylhexyl) phthalate (DEHP)	96	0.033	<LOD	0.69	1.17	2.06	34.85	1.08	3.28
∑DEHP1 μmol/L	NA	NA	NA	0.01	0.04	0.07	0.12	1.72	0.08	2.28
MHNCH ng/mL	Di(isononyl) cyclohexane-1,2-dicarboxylate (DINCH)	66	0.190	<LOD	<LOD	0.64	2.46	616.54	0.73	4.82
MCOCH ng/mL	Di(isononyl) cyclohexane-1,2-dicarboxylate (DINCH)	68	0.025	<LOD	<LOD	0.20	0.71	100.24	0.16	6.80
∑DINCH1 μmol/L	NA	NA	NA	0.00	0.00	0.00	0.01	2.27	<0.01	4.84
MEHHTP ng/mL	Di-(2-ethylhexyl) terephthalate (DEHTP)	98	0.035	<LOD	1.73	3.30	6.76	341.37	3.52	3.79
MECPTP ng/mL	Di-(2-ethylhexyl) terephthalate (DEHTP)	100	0.091	2.43	33.72	58.56	122.39	5149.56	67.36	3.10
∑DEHTP1 μmol/L	NA	NA	NA	0.02	0.12	0.20	0.42	17.86	0.23	3.05

¹We calculated molar sums as follows: ΣDEHP = MEHP + MECPP + MEHHP + MEOHP + MCMHP; ΣDINCH = MCOCH + MHNCH; ΣDEHTP = MECPTP + MEHHTP by summing SG-adjusted metabolite concentrations (with LOD/√2 imputation) after dividing by the molecular weight.

Early life exposure to multiple PRPs was associated with sex-specific alterations in within-network functional connectivity of the executive control and dorsal attention networks in boys and girls (Figure 1, Table 3). Whereas among boys increased exposure tended to be associated with reduced functional connectivity, among girls, increased exposure tended to be associated with increased functional connectivity. Specifically, among boys higher early life exposure to MnBP, MBzP, DINCH and DEHTP was associated with reduced functional connectivity in the executive control network, and higher DEHTP was associated with reduced functional connectivity in the dorsal attention network. Among girls, MCPP, MiBP and DEHP were associated with increased functional connectivity in the executive control and/or dorsal attention networks. Only MBzP was associated with reduced functional connectivity in executive control network among girls.

Figure 1. Associations between PRPs and Within-Network Connection Strength of Dorsal Attention and Executive Control Networks (Beta, 95% CI).

Associations of phthalates and replacement plasticizers (PRPs) and within-network connection strength, adjusted for child age, maternal education, and framewise displacement, and stratified or adjusted for sex. Statistically significant (p < 0.05) associations have confidence intervals that exclude the vertical line.

Table 3.

Phthalate and Replacement Plasticizer Associations and Within-Network Functional Connectivity in non-sedated rsfMRI, UNC Baby Connectome Project (n = 129)

	Executive Control β (95% CI)		Default Mode β (95% CI)		Dorsal attention β (95% CI)		Limbic β (95% CI)
	Girls	Boys	Girls	Boys	Girls	Boys	Girls	Boys
MEP	0.00 (−0.22, 0.21)	−0.16 (−0.41, 0.10)	0.04 (−0.13, 0.21)	−0.13 (−0.36, 0.11)	−0.03 (−0.20, 0.14)	−0.15 (−0.42, 0.12)	0.07 (−0.14, 0.28)	−0.05 (−0.34, 0.24)
MnBP	−0.01 (−0.19, 0.17)	−0.19 (−0.40, 0.02)	0.12 (−0.07, 0.31)	−0.11 (−0.34, 0.12)	−0.06 (−0.23, 0.11)	−0.01 (−0.24, 0.22)	0.08 (−0.05, 0.22)	−0.09 (−0.36, 0.18)
MiBP	0.06 (−0.15, 0.27)	−0.04 (−0.25, 0.17)	0.11 (−0.09, 0.30)	0.12 (−0.07, 0.32)	0.15 (0.00, 0.30)*	0.16 (−0.03, 0.35)	0.06 (−0.10, 0.22)	−0.10 (−0.26, 0.07)
MCPP	0.15 (−0.03, 0.34)	−0.04 (−0.24, 0.16)	−0.01 (−0.16, 0.15)	0.00 (−0.14, 0.14)	0.07 (−0.11, 0.25)	−0.04 (−0.28, 0.20)	0.14 (−0.03, 0.30)	−0.08 (−0.29, 0.14)
MBzP	−0.19 (−0.41, 0.02)	−0.16 (−0.35, 0.02)	−0.11 (−0.32, 0.09)	−0.09 (−0.31, 0.13)	−0.02 (−0.18, 0.14)	−0.01 (−0.21, 0.19)	−0.08 (−0.32, 0.15)	−0.16 (−0.38, 0.06)
∑DEHP	0.16 (−0.02, 0.34)	−0.14 (−0.39, 0.11)	0.14 (−0.03, 0.31)	−0.04 (−0.25, 0.16)	0.15 (0.04, 0.27)**	0.08 (−0.24, 0.40)	0.05 (−0.11, 0.21)	−0.09 (−0.31, 0.12)
∑DINCH	0.29 (−0.13, 0.71)	−0.18 (−0.34, −0.01)*	0.02 (−0.22, 0.25)	−0.01 (−0.19, 0.16)	0.06 (−0.22, 0.34)	0.04 (−0.22, 0.30)	0.08 (−0.15, 0.30)	−0.05 (−0.27, 0.16)
∑DEHTP	0.05 (−0.13, 0.24)	−0.26 (−0.41, −0.10)^	−0.10 (−0.37, 0.17)	−0.03 (−0.25, 0.19)	0.07 (−0.14, 0.29)	−0.35 (−0.59, −0.11)**	−0.07 (−0.27, 0.13)	−0.15 (−0.37, 0.07)
	Salience β (95% CI)		Sensorimotor β (95% CI)		Visual β (95% CI)
	Girls	Boys	Girls	Boys	Girls	Boys
MEP	−0.07 (−0.26, 0.11)	0.05 (−0.12, 0.22)	−0.09 (−0.32, 0.14)	0.08 (−0.12, 0.27)	−0.01 (−0.25, 0.23)	−0.03 (−0.31, 0.25)
MiBP	−0.10 (−0.26, 0.07)	−0.08 (−0.30, 0.13)	0.06 (−0.11, 0.23)	−0.02 (−0.26, 0.22)	−0.02 (−0.25, 0.21)	−0.11 (−0.30, 0.08)
MnBP	−0.01 (−0.24, 0.22)	−0.13 (−0.32, 0.06)	0.20 (0.02, 0.38)*	0.04 (−0.30, 0.39)	0.02 (−0.13, 0.17)	−0.11 (−0.34, 0.11)
MCPP	−0.01 (−0.13, 0.11)	0.02 (−0.15, 0.19)	0.06 (−0.17, 0.30)	0.04 (−0.21, 0.28)	0.06 (−0.12, 0.23)	−0.01 (−0.20, 0.17)
MBzP	−0.03 (−0.18, 0.12)	0.01 (−0.20, 0.21)	0.04 (−0.18, 0.27)	−0.05 (−0.37, 0.27)	−0.21 (−0.51, 0.09)	0.02 (−0.22, 0.27)
∑DEHP	0.04 (−0.17, 0.25)	−0.06 (−0.29, 0.16)	0.12 (−0.12, 0.36)	0.12 (−0.25, 0.48)	−0.01 (−0.21, 0.19)	−0.14 (−0.36, 0.07)
∑DINCH	−0.09 (−0.24, 0.07)	0.01 (−0.21, 0.23)	−0.12 (−0.43, 0.20)	−0.01 (−0.25, 0.22)	−0.06 (−0.22, 0.09)	0.16 (−0.03, 0.35)
∑DEHTP	0.11 (−0.09, 0.31)	−0.34 (−0.50, −0.18)^^	0.04 (−0.15, 0.24)	0.08 (−0.28, 0.44)	0.12 (−0.13, 0.37)	0.00 (−0.26, 0.25)

^*p < 0.05;

^**p < 0.01;

^{^}q < 0.5;

^{^^}q < 0.01

Adjusted for infant age, maternal education and framewise displacement (for scan effects), and stratified by infant sex.

While directionality of associations often differed between boys and girls, there were also occasions where they were aligned, for example, MBzP and functional connectivity of the executive control network, and MiBP and functional connectivity of the dorsal attention and default mode networks (Table 3). In models adjusted for but not stratified by sex, increased MBzP was significantly associated with reduced functional connectivity in the executive control network overall (β = −0.16, 95% CI −0.31, −0.01) (Figure 1). Although MiBP was associated with increased functional connectivity in the dorsal attention and default mode networks among boys and girls at roughly the same magnitude, the overall associations adjusted for sex were suggestive but not statistically significant (β_DAN = 0.10, 95% CI −0.05, 0.25, β_DM = 0.11, 95% CI −0.01, 0.24) (Figure 1).

There were sporadic associations between PRPs and functional connectivity of other networks, some highly significant (e.g. DEHTP and reduced salience among boys, MnBP and increased sensorimotor among girls, MnBP and increased visual among boys), however there were fewer detectable patterns (Table 3). Volcano plots of all PRP associations with functional networks are included in Supplemental Figure 2. We conducted a sensitivity analysis, restricting the population to scans among children less than 3 years of age (Supplemental Figure 3). Associations were in generally highly concordant with the total population, albeit less precisely estimated due to the overall smaller sample size. However, some new associations were revealed or disparities by sex sharpened, including MnBP and MiBP and executive control, and MiBP and DEHP with default mode.

We also examined early life PRP metabolite concentrations with cortical surface area and thickness within 68 regions of interest (Figures 2 and 3, Supplemental Table 1). Among girls, we found that MEP was associated with significantly smaller surface areas of the left pars triangularis (β=−0.24, 95% −0.39,−0.09), right superior parietal cortex (β=−0.33, 95% −0.50,−0.17), and left lateral orbitofrontal cortex (β=−0.28, 95% −0.42,−0.14), after accounting for multiple testing (q-value < 0.05). Additionally, MiBP was associated with significantly larger surface area for the left rostral anterior cingulate (β=0.32, 95% CI 0.15,0.50), after accounting for multiple testing. Although not quite meeting our multiple testing threshold, surface areas of several other regions were highly associated with DEHP, MiBP, DEHTP, MnBP or MCPP respectively at p < 0.01 (Figure 2, Supplemental Table 1).

Figure 2. Early Life Phthalate and Replacement Plasticizers in relation to Cortical Surface Area in 68 Regions of Interest.

Regions associated with metabolites with false discovery rate adjusted q-values < 0.05 are bolded, and are colored in dark blue or dark red respectively. All other associations are p < 0.01 and appear as light blue or pink. Dark blue/light blue regions depict associations leading to reduced surface area, pink/red regions depict associations leading to increased surface area. Purple regions include a mix of increased and decreased surface area associations.

Figure 3. Early Life Phthalate and Replacement Plasticizers in relation to Cortical Thickness in 68 Regions of Interest among Females.

Regions associated with metabolites with false discovery rate adjusted q-values < 0.05 are bolded and are colored in dark blue. Light blue associations are p < 0.01. Shaded blue regions depict associations leading to reduced cortical thickness. There were no significant associations among girls leading to increased cortical thickness.

There were fewer significant region of interest associations with surface areas among boys, however DEHP was significantly associated with larger surface area of the left inferior parietal (β=0.34, 95% CI 0.16, 0.51), and MiBP was significantly associated with smaller surface area of the right caudal anterior cingulate (β=−0.28, 95% CI −0.44, −0.12), after accounting for multiple testing. Other regions were highly associated with MBzP, DEHTP, or MnBP respectively, at p < 0.01 (Figure 2).

Cortical thinning of several regions was also strongly associated with PRP exposure, particularly among girls (Figure 3). MiBP was strongly associated with cortical thinning in the left superior temporal gyrus, left inferior temporal gyrus, right banks of the superior temporal sulcus, right middle temporal gyrus, left precentral gyrus, left entorhinal cortex, and the right precuneus regions after accounting for multiple testing (Supplemental Table 1). Many of these regions are located in the temporal lobe, and play an important role in the Default Mode, Dorsal Attention, sensorimotor, and limbic networks (Supplemental Table 2). MiBP, MnBP, MCPP, and DEHTP were also associated with cortical thinning in other regions at p < 0.01. There were no FDR-adjusted significant associations with cortical thickness among boys, however MBzP, DEHP and MCPP were associated with increased thickness of the right pars opercularis, right rostral anterior cingulate cortex, and right isthmus cingulate cortex at p < 0.01 (Supplemental Table 1). We provide volcano plots for regions of interest analyses in Supplemental Figures 4 and 5.

4. Discussion

In this longitudinal study of early life exposure to PRPs and brain development, we report novel associations between infant and early childhood exposure and functional connectivity in the executive control and dorsal attention networks, as well as altered growth of cortical structures that underlie behavioral and cognitive development. We found that multiple phthalates are associated with alterations in functional connectivity that are sex specific. Increased MCPP, MiBP and DEHP concentrations among girls was associated with higher within-network connectivity of the executive control and dorsal attention networks, whereas among boys, increased MnBP, MBzP, DINCH and DEHTP exposure tended to be associated with lower within-network connectivity in the executive control and dorsal attention networks. We furthermore reported FDR-significant associations of MiBP with cortical thinning in multiple regions of interest among girls, several of which were in the temporal lobe.

The brain undergoes rapid growth and development in the first years of life, measuring at 36% of adult volume at 2 weeks of age, and growing to 83% of adult volume by 2 years of age [39]. During this time, a series of well-organized brain developmental processes establish the foundation for complex neural circuitries that govern various cognitions. While quantitative measures of brain anatomical attributes provide insights into neural substrates underlying the rapid brain size increase during early infancy, resting-state functional magnetic resonance imaging (rsfMRI) is a powerful non-invasive tool capable of characterizing the maturation of brain functional networks which underlie early brain cognitive development [45, 60–65]. By leveraging longitudinal imaging of children, we are further able to examine impacts on trajectories of brain growth and function.

Our study in particular highlights sex-specific associations of several PRPs with changes in functional connectivity of the executive control and dorsal attention networks, two prefrontal cortex networks critical to cognitive control [46]. It should be noted that neither increased nor decreased within-network connection strength is inherently adverse, particularly in this setting of normative brain development. Rather, temporal patterns in connection strength reflect underlying developmental processes, which are complex and have been shown to vary by functional network [48]. Some have suggested that increased within-network connectivity may indicate that maturation processes are underway, whereas decreased connectivity could suggest functional specialization [48, 66]. As such changes in connectivity in early life may reflect alterations in the developmental pacing of these networks, potentially accelerating or delaying the maturation of cognitions that are undergoing hierarchical development. It is notable that when the study population was restricted to children less than three years of age, most of the associations remained directionally consistent, however in some cases the magnitude and significance drastically changed (e.g. MiBP and default mode and executive control). These differences may suggest underlying heterogeneity in age-specific associations. Larger study populations will be required to interrogate heterogeneity in associations by age.

The executive control network (sometimes referred to as the central executive network, or the frontoparietal network) is the key brain network responsible for the completion of goal-directed tasks including decision making, planning, problem solving, as well as working memory [46, 67–69]. Not surprisingly, executive control network functional connectivity has been directly correlated with executive function performance [70, 71]. Executive functions are often impaired in the setting of developmental disorders, such as autism spectrum disorder (ASD) and attention-deficit hyperactivity disorder (ADHD) [72–75]. Altered connectivity patterns, both hyper- and hypo-connectivity, in the executive control, default mode, and salience networks have been linked with ADHD behaviors, although patterns of connection within and across networks have differed by study [76], and sex differences in the neural underpinnings of ADHD have been previously reported [77]. Prenatal phthalate exposure has also been associated with problems with executive functioning [78–80], ADHD and ASD symptoms [79, 81–88], as well as clinical diagnosis of ASD [89], and ADHD [90, 91], often but not always demonstrating sex-specific patterns of association. In this study we report a possible underlying correlate for these associations- namely, that phthalates interfere with the development of the executive control network, as demonstrated by reduced within-network functional connectivity in relation to increasing exposure to MnBP, MBzP, DINCH and DEHTP among boys. However, among girls, only MBzP showed the same pattern of reduced connection strength. In contrast, MCPP, MiBP and DEHP were all associated with increased connection strength in the executive control network among girls. Potential mechanism underlying these observed sex differences in associations of phthalates with brain functional network connection strengths are at this point speculative. However, it should be noted that multiple phthalates and replacement plasticizers have been associated with changes in circulating sex steroid concentrations [92, 93]¸ sex steroids have been shown to exert profound effects on brain structural and functional development [94], and sex-specific patterns in the functional organization of the brain have been previously reported in infants [95] and adults [92, 93, 96]. Therefore, it is conceivable that PRPs induce changes in circulating sex steroids, which alter sex-specific patterns of brain development, resulting in the observed sex-differences in the relationship of PRPs with brain functional and structural measures. Future studies are needed to describe long-term implications of these sex-specific patterns of functional connectivity on cognitive and behavioral outcomes.

The dorsal attention network directs voluntary control of visuospatial attention [97, 98], essentially holding attention on tasks that are underway. We found MiBP (in boys and girls) and DEHP (in girls) to be positively associated with functional connectivity of the dorsal attention network. Hyper-connectivity within various regions of the dorsal attention network has been correlated with trait anxiety or anxiety symptoms [99, 100]. In contrast, DEHTP was negatively associated with functional connectivity of the dorsal attention network among boys in our data. Reduced dorsal attention network connectivity has been associated with higher levels of ADHD symptoms [89, 90]. Epidemiologic studies have previously correlated prenatal phthalate exposure with externalizing [79, 101–104] and internalizing symptoms [102–108], that were sometimes sex-specific. Follow-up studies will be needed to ascertain the extent to which impacts of phthalates on cognitive development or behavior are attributable to changes in functional connectivity of dorsal attention network or other networks involved in cognitive control.

Gradual cortical thinning is a normative component of brain maturation, and is driven by normal synaptic pruning, changes in neuronal size, glial cell density, and vasculature [109]. Although not statistically significant, the suggestive association of higher MiBP concentrations with increased functional connectivity in the default mode network in boys and girls is interesting, in light of the FDR-significant associations of MiBP with cortical thinning in regions of interest relevant to the default mode network, sensorimotor, and limbic networks among girls, including the left inferior temporal, left superior temporal, right banks of the superior temporal sulcus, right middle temporal gyrus, and right precuneus. Previous research has reported negative association of prenatal MiBP exposure with total gray and white matter volume among girls at age 10 in The Generation R Study [110]. Future studies are needed to examine the significance of these early life changes in cortical thickness with respect to cognitive and behavioral developmental measures.

This study has both limitations and strengths. Within the scope of typical environmental epidemiology studies, our sample size is small, which limits power, particularly for analyses stratified by sex, and prevents us from examining age-specific hypotheses. However, neuroimaging studies are often small because scanning costs are high, with typical sample sizes in neuroimaging studies measured in the dozens rather than the hundreds or thousands of participants. With infants and young children, scans are conducted during non-sedated natural sleep, and therefore inability of the child to fall asleep or stay asleep in the scanner frequently results in missed scans, and excessive movement during the scan results in further drop out due to motion artifacts. Nonetheless, our study is significantly strengthened by its longitudinal design, with both repeated urine sample and repeated MRI collection over time. The majority of our subjects had at least two longitudinal scans and urines collected, with as many as 7 longitudinal MRIs and 6 longitudinal urines over the course of follow-up. Although due to sporadic missing data due to a failed MRI or failed urine collection, these visits may not be perfectly temporally aligned. Longitudinal measures allow us to capture variability in phthalate exposure over time, as well as the longitudinal trajectories of maturation in structural and functional characteristics of brain development. However, we acknowledge that approximately 25% of the study population had only 1 spot urine sample available for analysis, which will introduce exposure misclassification due to the short biological half-lives of these compounds. Because timing of urine collection was not standardized, it is possible that one source of variability in biomarker concentrations may be temporal proximity of urine collection to mealtimes or use of babycare products.

In order to account for the sparse and sometimes asynchronous exposure and outcome data, we used a specialized analysis approach designed for sparse, asynchronous, longitudinal measures [111]. This approach down-weights values that are far in time from the current response (essentially favoring more temporally coincident exposure and outcome measurements), producing a weighted average association of exposure with imaging outcomes in early life, overcoming challenges posed by limited observations at specific ages. However, future and larger studies will be needed to investigate age-specific windows of vulnerability. This study was designed to test the hypothesis that phthalate and replacement plasticizer exposure in early life is associated with changes in brain structure and function, which necessitates performing many statistical tests. We included FDR correction, particularly for region of interest analyses, in order to reduce the possibility of false positives. However, correction for multiple testing also increases the possibility of type 2 errors (false negatives). We attempted to balance these error trade-offs by relying on FDR evidence for region of interest analyses where the number of tests is quite large, but more flexibly examining patterns of associations across networks for rsfMRI analyses, where the number of tests involved was comparatively small.

Finally, it is important to acknowledge that this population over-represents individuals of high education and income and has limited inclusion of racial/ethnic minorities. While PRP exposure is widespread in the general population, racial/ethnic minorities tend to be more highly exposed to some phthalates, and parental income and education may have a mitigating effect on any adverse impacts of toxic exposures. However adverse neurodevelopmental impacts of phthalate exposures have been observed in a variety of settings, not limited to racial/ethnic minorities or low-income populations.

In conclusion, we report novel associations between early life exposure to PRPs and changes in brain functional connectivity patterns, particularly in the executive control and dorsal attention networks. We also report associations of MiBP with cortical thinning among girls. These results indicate that exposure to phthalates and replacement plasticizers in early life has a measurable impact on biomarkers of brain structural and functional development, that may underlie observational associations with child cognitive and behavioral development.

Highlights.

• PRP metabolites associated with executive control and dorsal attention networks

• Directionality of association with functional connectivity differed by child sex

• MiBP associated with cortical thinning among girls in multiple regions of interest