Exploring diet as a source of plasticizers in pregnancy and implications for maternal second-trimester metabolic health

Abstract

Background and objectives:

Diet plays critical roles in modulating maternal metabolic health in pregnancy, but is also a source of metabolic-disrupting phthalates and their replacements. We aimed to evaluate whether the effects of better diet quality on favorable maternal metabolic outcomes could be partially explained by lower exposure to phthalates/replacements.

Methods:

At 13 weeks gestation, 295 Illinois women (enrolled 2015–2018) completed a three-month food frequency questionnaire that we used to calculate the Alternative Healthy Eating Index (AHEI)-2010 diet quality index. We quantified 19 metabolites, reflecting exposure to 10 phthalates/replacements, in a pool of five first-morning urine samples collected monthly across pregnancy. We measured 15 metabolic biomarkers in fasting plasma samples collected at 17 weeks gestation, which we reduced to five uncorrelated principal components (PCs), representing adiposity, lipids, cholesterol, inflammation, and growth. We used linear regression to estimate associations of diet quality with [1] phthalates/replacements and [2] metabolic PCs, as well as [3] associations of phthalates/replacements with metabolic PCs. We estimated the proportion of associations between diet quality and metabolic outcomes explained by phthalates/replacements using a causal mediation framework.

Results:

Overall, every 10-point improvement in AHEI-2010 score was associated with −0.15 (95% CI: −0.27, −0.04) lower adiposity scores, reflecting lower glucose, insulin, C-peptide, leptin, C-reactive protein, but higher adiponectin biomarker levels. Every 10-point increase in diet quality was also associated with 18% (95%CI: 7%, 28%) lower sum of di-2-ethylhexyl terephthalate urinary metabolites (ΣDEHTP). Correspondingly, each 18% increase in ΣDEHTP was associated with 0.03 point (95% CI: 0.01, 0.05) higher adiposity PC scores. In mediation analyses, 21% of the inverse relationship between diet quality and adiposity PC scores was explained by lower ΣDEHTP.

Conclusions:

The favorable impact of diet quality on maternal adiposity biomarkers may be partially attributed to lower metabolite concentrations of DEHTP, a plasticizer allowed to be used in food packaging materials.

1. INTRODUCTION

Dysregulation of maternal metabolic processes in pregnancy can lead to adverse metabolic conditions, including pre-eclampsia, dyslipidemia, and gestational diabetes mellitus (GDM), which are significant risk factors of poor pregnancy and birth outcomes, along with adverse childhood cardiometabolic and developmental problems that may persist into adolescence and adulthood (Bjorstad et al., 2010; Hack et al., 1995; Markopoulou et al., 2019; Mitchell et al., 2017; Palatianou et al., 2014; Tita et al., 2009; Zhao et al., 2012). Poor metabolic health in pregnancy is also implicated in maternal later-life cardiometabolic disease risk (Dassanayake et al., 2020; Haggerty et al., 2021; Heron, 2018). Therefore, identifying modifiable factors that affect maternal metabolic health in pregnancy could help inform points of intervention to protect maternal and child lifelong health.

Diet contributes to maternal metabolic health in pregnancy (Lain and Catalano, 2007) and modulates chronic inflammation (Bragg et al., 2022; Chen et al., 2021; Hennig et al., 2007; Lecorguille et al., 2021). However, diet can be a chronic source of hormone and metabolic-disrupting chemicals, including some phthalates and their replacements (Pacyga et al., 2019). Because phthalates can be used widely in food packaging materials, during food processing, and in medication/supplement coatings, phthalate exposure in pregnancy is ubiquitous (Serrano et al., 2014; Woodruff et al., 2011). Observational studies link certain dietary behaviors to higher phthalate exposure (Pacyga et al., 2019; Serrano et al., 2014), which is supported by several intervention studies in non-pregnant individuals (Corbett et al., 2022). The chronic nature of maternal dietary phthalate exposure is especially concerning due to observational studies linking some phthalates to increased risk of gestational metabolic conditions (Fisher et al., 2018a; James-Todd et al., 2018; Shaffer et al., 2019b; Soomro et al., 2021), potentially due to higher inflammation and interaction with receptors that mediate glucose and lipid homeostasis (Strakovsky and Schantz, 2018; Veiga-Lopez et al., 2018). Pregnant women may also become increasingly exposed to less well-studied plasticizers, including cyclohexane-1,2-dicarboxylic acid diisononyl ester (DiNCH) and di(2-ethylhexyl) terephthalate (DEHTP) (Bui et al., 2016; Lemke et al., 2021; Lessmann et al., 2016; Pacyga et al., 2022), which may also have adverse metabolic-disrupting effects in pregnancy (Deierlein et al., 2022; Wu et al., 2021). Therefore, it is critical to understand whether high quality diets may be key to reducing exposure to metabolic-disrupting phthalates and their replacements during this sensitive window (Barrett et al., 2015).

In the current study, our overarching goal was to determine whether the favorable impact of maternal diet quality on metabolic status is partially explained by lower exposure to metabolic-disrupting phthalates/replacements, which has not been previously investigated. To accomplish this goal, we discreetly evaluated associations of 1) maternal diet quality with metabolic biomarkers, 2) phthalate/replacement biomarkers with metabolic biomarkers, and 3) maternal diet quality with phthalate/replacement biomarkers.

2. METHODS

2.1. Illinois Kids Development Study (I-KIDS) recruitment and enrollment

This study includes a subset of pregnant women enrolled in I-KIDS, a prospective, longitudinal pregnancy and birth cohort from Champaign-Urbana, Illinois (Pacyga et al., 2021; Pacyga et al., 2023a; Pacyga et al., 2022). Recruitment occurred from two local obstetric clinics. Women were eligible to participate if they were less than 15 weeks pregnant, having a singleton pregnancy, older than 18 but younger than 40 years old, fluent English speakers, not in a high-risk pregnancy, willing to provide a fasting blood sample, living within 30 minutes of the study site, and had no plans to move out of the study radius before their child turned one. Women in the current study enrolled between the years 2015 and 2018 and were followed through delivery. The characteristics of women in this sample reflect the full I-KIDS cohort (Pacyga et al., 2022). The Institutional Review Board at the University of Illinois approved the study, and all women provided written informed consent. The analysis of de-identified specimens at the Centers for Disease Control and Prevention (CDC) laboratory was determined not to constitute engagement in human subjects’ research.

2.2. Collection of maternal sociodemographic and lifestyle information at baseline

Study staff interviewed women about their health, lifestyle, and sociodemographic characteristics at median 13 weeks, which was the first study visit. Women self-reported their annual household income, parity, race and ethnicity, highest educational attainment, and alcohol intake since conception to determine first trimester intake (Pacyga et al., 2023a). Pre-pregnancy body mass index (BMI, kg/m²) was calculated using self-reported pre-pregnancy weight and height.

2.3. Collection of diet data and calculation of the Alternative Healthy Eating Index (AHEI)-2010

Women completed an adapted and validated-for-pregnancy version of the full-length Block-98 semi-quantitative food frequency questionnaire (FFQ; NutritionQuest, Berkeley, CA) (Bodnar and Siega-Riz, 2002; Boucher et al., 2006; Laraia et al., 2007). The FFQ was completed at median 13 weeks gestation, which captured first trimester diet from approximately conception through this first study visit (Pacyga et al., 2023a). We estimated women’s first-trimester daily average energy intakes and calculated the AHEI-2010 index capturing dietary patterns that predict mortality and chronic disease risk (Chiuve et al., 2012; McCullough et al., 2002) and has been validated for use in pregnancy (Chia et al., 2019; Li et al., 2021). The AHEI-2010 includes six positive components: fruit, vegetables, nuts/legumes, whole grain, docosahexaenoic acid (DHA)/eicosapentaenoic acid (EPA), polyunsaturated fatty acids (PUFAs); and five negative components: red/processed meat, sugar-sweetened beverages (SSBs)/fruit juice, sodium, trans fat, and alcohol. Each food component was scored on a 10-point scale, such that higher scores for positive components and lower scores for negative components represent higher intakes of these foods/nutrients. Alcohol was removed from the total score because pregnant women are guided to not consume alcohol in pregnancy (ACOG, 2015), and the AHEI-2010 assigns moderate alcohol consumption the highest (beneficial) score. Thus, in this study, the AHEI-2010 has a maximum score of 100 with scores closer to 100 reflect better diet quality. For the current study, we considered the total score and the standardized scores of the individual food/nutrient components.

2.4. Quantification of early second trimester plasma metabolic biomarkers

The current study focused on early second trimester metabolic status as this is when the placenta begins to coordinate gestational metabolic adaptations (Grimes and Wild, 2000; Lain and Catalano, 2007; Mor et al., 2017; Tal et al., 2000). We assessed metabolic status via a panel of glucose, lipid, and inflammatory biomarkers described elsewhere (Cinzori et al., 2024). Briefly, at median 17 weeks, women fasted overnight and had their blood drawn for plasma metabolic biomarker assessment. Fourteen metabolic biomarkers were measured at the University of Michigan Diabetes Research Center Clinical Core Chemistry Laboratory using standard manufacturer protocols (Cinzori et al., 2024). Within- and between-assay coefficients of variability of all analytes were less than 6%. Using prior equations, we calculated very low-density lipoprotein (VLDL) and low-density lipoprotein (LDL) cholesterol concentrations (Sampson et al., 2020) along with homeostasis model assessment of insulin resistance (HOMA-IR) (Matthews et al., 1985).

2.5. Quantification of urinary phthalate/replacement metabolite concentrations across pregnancy

Women collected first-morning urines in polypropylene urine cups at median 13, 17, 23, 28, and 34 weeks gestation for chemical biomarker assessment as detailed in prior publications (Pacyga et al., 2021; Pacyga et al., 2022; Pacyga et al., 2023b). All women provided samples during at least three and up to five timepoints. Samples from each gestational timepoint were pooled for each woman. Specifically, 900 μL of urine from the first study visit was added to a 5 mL cryovial tube (Nalgene, Rochester, NY) and stored at −80 °C, and the same volume of fresh urine from subsequent study visits was layered onto frozen urine from the previous visit until the last study timepoint. Pooled urine samples were thawed and vortexed to measure specific gravity using a handheld refractometer (TS400; Reichert Technologies, Depew, NY) prior to shipment, and frozen pooled samples were shipped overnight on dry ice to the CDC laboratory to measure 19 phthalate/replacement metabolite concentrations. These included four di(2-ethylhexyl) phthalate (DEHP) metabolites mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethylhexyl) phthalate (MEHP), and mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), three di-isononyl phthalate (DiNP) metabolites monocarboxyoctyl phthalate (MCOP), mono-isononyl phthalate (MiNP), and monooxononyl phthalate (MONP), two di-n-butyl phthalate (DBP) metabolites mono-n-butyl phthalate (MBP) and monohydroxybutyl phthalate (MHBP), two di-isobutyl phthalate (DiBP) metabolites mono-isobutyl phthalate (MiBP) and mono-hydroxy-isobutyl phthalate (MiHBP), two DiNCH metabolites cyclohexane-1,2-dicarboxylic acid-mono(carboxyoctyl) ester (MCOCH) and cyclohexane-1,2- dicarboxylic acid-monohydroxy isononyl ester (MHiNCH), two DEHTP metabolites mono(2- ethyl-5-carboxypentyl) terephthalate (MECPTP) and mono(2-ethyl-5-hydroxyhexyl) terephthalate (MEHHTP) along with monocarboxynonyl phthalate (MCNP), mono(3- carboxypropyl) phthalate (MCPP), monobenzyl phthalate (MBzP), and monoethyl phthalate (MEP), which are the major metabolites of di-isodecyl phthalate (DiDP), di-n-octyl phthalate (DOP), benzylbutyl phthalate (BBzP), and diethyl phthalate (DEP), respectively. Analyses were performed by on-line solid phase extraction coupled with isotope dilution high performance liquid chromatography-electrospray ionization-tandem mass spectrometry using established CDC protocols. The laboratory uses rigorous quality assurance and control measures and chemical assays have high reproducibility with long-term within- and across-day coefficient of variation less than 14% (Silva et al., 2013; Silva et al., 2007; Silva et al., 2019).

2.6. Statistical analyses

2.6.1. Derivation of the analytic sample and presentation of descriptive statistics

Women in the analytic sample had complete data on first trimester diet, the concentrations of urinary phthalate/replacement metabolites of interest, and all metabolic biomarkers of interest. Our final sample comprised of 295 pregnant women as indicated in Supplemental Figure 1. The primary limitation for the sample size was the exclusion of a large number of women (early in the study) whose chemical data were analyzed prior to the inclusion of DEHTP metabolites in the CDC phthalate panel (Supplemental Figure 1). We described the analytic sample using n (%) or median (25^th, 75^th percentiles). We also described dietary, chemical, and metabolic measures using min, max, and 25^th, 50^th, and 75^th percentiles.

2.6.2. Modeling maternal urinary phthalate/replacement biomarker concentrations

To correct for urine dilution (Meeker et al., 2009; Pacyga et al., 2021; Pacyga et al., 2022; Pacyga et al., 2023b), we specific-gravity adjusted phthalate/replacement biomarker concentrations (Pacyga et al., 2021; Pacyga et al., 2022; Pacyga et al., 2023b). Because DEHP, DiNP, DBP, DiBP, DiNCH, and DEHTP are each metabolized into several metabolites that are excreted in urine, we molar summed (acronyms denoted with Σ in the results) the respective metabolite concentrations to provide an estimate of each woman’s exposure in pregnancy. We multiplied the sums by the molecular weight of MECPP, MCOP, MBP, MiBP, MHiNCH, or MECPTP, to express the nmol/mL concentrations in ng/mL. Exposure to parent compounds DiDP, DOP, BBzP, and DEP were estimated using the ng/mL of their major urinary metabolite. Because of the potential bias linked with imputing concentrations below the LOD (Succop et al., 2004), we used instrument-read values for all phthalate/replacement metabolite concentrations. One sample had DiNCH metabolite concentrations of zero, thus added a constant (1.0) to ΣDiNCH prior to ln-transformation (Weiss et al., 2015).

2.6.3. Modeling maternal early second trimester plasma metabolic biomarker concentrations

Because 70% of women had IL-6 levels below the lower assay limit, we removed IL-6 from further analyses. For insulin (16% < lower assay limit) and free fatty acids (0.3% < lower assay limit), we replaced concentrations below the lower assay limit with [plate minimum/√2] and [minimum/√2], respectively. In addition to evaluating each of the 16 metabolic biomarkers individually, we employed principal component analysis (PCA) to identify linear patterns among metabolic biomarkers that may reflect common biological pathways (Hotelling, 1933). HOMA-IR was excluded from PCA due to strong correlation with insulin (r = 0.99). Therefore, a total of 15 metabolic biomarkers were included in PCA. All metabolic biomarkers with right-skewed distributions were ln-transformed and all biomarkers were z-transformed prior to inclusion in PCA. We specified a PCA with a Varimax rotation to create distinct, uncorrelated principal components (PCs) that explained metabolic biomarker variance in the analytic sample. Scree plots of eigenvalues were used to determine the ideal number of PCs that explain the most variability in the data (Pacyga et al., 2022), but also reflect known biology. Metabolic biomarkers with loadings r ≥ 0.4 were considered indicative that the individual PC was representative of a given biological construct. PCA analyses were conducted in SAS 9.4 (SAS Institute Inc, Cary, NC) using PROC FACTOR.

2.6.4. Covariate selection

We used a directed acyclic graph (DAG; Supplemental Figure 2) to guide covariate selection. Splines for continuous or ordinal variables were evaluated to determine the operationalization of these covariates. Multicollinearity among covariates was evaluated using correlations (Pearson or Spearman for continuous and polychoric for categorical variables). All correlations between covariates were less than |0.4|. Final statistical models included educational attainment, race/ethnicity, parity, first-trimester alcohol intake, and fetal sex as categorical variables (see Table 1 for categories and reference groups) and annual household income as an ordinal variable. Race/ethnicity was a proxy for unmeasured structural/cultural factors that may confound our proposed associations. Due to small sample sizes, women who self-reported as non-Hispanic Black, Asian, Native Hawaiian or other Pacific Islander, American Indian or Alaska Native, Multiracial, or Other were classified into one category representing racial/ethnic minority groups. We accounted for gestational age at blood draw when evaluating metabolic biomarkers as the outcome. Average daily caloric intake was included in models when evaluating diet quality as the exposure but was excluded from mediation analyses (section 2.6.6.) due to convergence issues. In models evaluating individual AHEI-2010 components as the exposure, possible confounding from overall diet quality was accounted for by controlling for the AHEI-2010 total score excluding the individual component of interest. Final models did not include pre-pregnancy BMI, similar to other studies (Minguez-Alarcon et al., 2022), given that pre-pregnancy BMI is comorbid with gestational metabolic status and may mediate the associations of interest.

Table 1.

Characteristics of pregnancy women in the analytic sample (n = 295).

Characteristic	Median (25th, 75th percentile)
Age (years)	31.0 (28.0, 33.0)
Pre-pregnancy BMI (kg/m2)	24.9 (21.9, 29.3)
Average daily caloric intake in first trimester1 (kcals)	1523.0 (1155.0, 1788.0)
	n (%)
Race/ethnicity 1
(ref) Non-Hispanic White	240 (81.4)
Other2	55 (18.6)
Educational attainment 1
Some college or less	47 (15.9)
(ref) College graduate or higher	248 (84.1)
Annual household income 1
< $70,000	104 (35.3)
≥ $70,000	191 (64.7)
Parity 1
(ref) Nulliparous	159 (53.9)
Multiparous	136 (46.1)
Alcohol consumption in the first trimester 1
(ref) None	170 (57.6)
Any	125 (42.4)
Fetal sex 1
Female	150 (50.8)
Male	145 (49.2)

¹Covariates included in analyses.

²Includes non-Hispanic Black, Asian, Native Hawaiian or other Pacific Islander, American Indian or Alaska Native, Multiracial, and Others.

2.6.5. Primary linear regression analyses

Unadjusted and covariate-adjusted linear regression models evaluated associations of 1) first trimester diet quality with early second trimester metabolic biomarkers, 2) first trimester diet quality with phthalate/replacement biomarkers, and 3) phthalate/replacement biomarkers with metabolic biomarkers. All exposures and outcomes were parameterized as continuous. Due to having right-skewed distributions, all phthalate/replacement biomarkers and most metabolic biomarkers (except glucose, total cholesterol, LDL cholesterol, and MCP-1) were ln-transformed. The AHEI-2010 total and component scores were not transformed. β estimates and 95% CIs were back-transformed to represent the difference in the outcome for each 10% increase in the exposure. For models evaluating diet quality as the exposure, a 10% increase in diet scores reflects a 10.0- or 1.0-point increase, respectively. Linear regression analyses were performed in SAS 9.4 (SAS Institute Inc, Cary, NC) using PROC GLM.

2.6.6. Primary mediation analyses

Formal mediation analyses using a causal framework (Valeri and Vanderweele, 2013) was conducted to test the hypothesis that phthalate/replacement biomarker concentrations (mediators) partially explain the relationship between maternal early pregnancy diet (exposures) with early second trimester metabolic biomarkers (outcomes). Guided by results of linear regression analyses, mediation analyses were only conducted if 1) all three associations were not null and 2) the direction of association for each of the three individual relationships made sense in the context of the others. Following the approach outlined by Valeri and Vanderweele (Valeri and Vanderweele, 2013), multiple mediation models with an exposure-mediator interaction (no bootstrapping) were then used to confirm the total effect and estimate the natural direct effect (i.e., association between diet quality and metabolic biomarkers), and the natural indirect effect (i.e., association between diet quality and metabolic biomarkers through phthalates/replacements). We also determined the percentage mediated or proportion of the total effect explained by the natural indirect effect. Mediation was performed in SAS 9.4 (SAS Institute Inc, Cary, NC) using PROC CAUSALMED.

2.6.7. Identification of notable findings

Throughout the results, we pointed out any associations were P < 0.05 and P < 0.10. However, notable results were identified based on the precision around but also the strength/direction of the effect estimates to identify consistent patterns across the evaluated associations (Wasserstein and Lazar, 2016). Analyses did not account for multiple comparisons based on prior recommendations (Rothman, 1990), but we acknowledge the potential for spurious results.

3. RESULTS

3.1. Characteristics of the analytic sample

Participant characteristics are presented in Table 1. Women had a median age of 31.0 years, pre-pregnancy BMI of 24.9 kg/m², and average daily energy intake in first trimester of 1522.8 kcals. Most women were non-Hispanic White (81%), college educated (84%), had annual household incomes ≥ $70,000 (65%), and did not consume alcohol in the first trimester (58%). For a little over half of the women, this was their first child, and around half of women were carrying a male fetus. Median (min, max) AHEI-2010 was 51.7 points out of 100 (25.6, 79.4) (Table 2). Women in this analytic sample had median phthalate/DiNCH biomarker concentrations (Supplemental Table 1) similar to, but DEHTP biomarker concentrations higher than those reported in same-age U.S. women (Pacyga et al., 2022).

Table 2.

Distribution of maternal first trimester AHEI-2010 and its individual components (n = 295).

Index component	Min	25th percentile	Median	75th percentile	Max
Total score (max 100 pts)	25.64	44.50	51.70	59.33	79.39
Positive components 1
Vegetables (max 10 pts)	0.56	2.84	4.30	6.08	10.00
Fruit (max 10 pts)	0.03	2.55	4.58	7.60	10.00
Whole grains (max 10 pts)	0.07	0.87	1.80	2.86	8.96
PUFA (max 10 pts)	3.12	5.74	6.70	8.02	10.00
EPA and DHA (max 10 pts)	0.07	1.09	2.08	4.40	10.00
Nuts and legumes (max 10 pts)	0.14	2.13	4.77	9.07	10.00
Negative components 2
SSBs and fruit juice (max 10 pts)	0.00	0.00	4.11	8.04	10.00
Trans fat (max 10 pts)	3.76	7.27	8.00	8.49	10.00
Red and processed meat (max 10 pts)	0.00	5.80	7.27	8.50	10.00
Sodium (max 10 pts)	0.00	4.38	6.52	8.71	10.00

¹Foods beneficial for health: higher scores indicate higher intakes of these foods.

²Foods that should be consumed in moderation: higher scores indicate lower intakes of these foods.

AHEI-2010, Alternative Healthy Eating Index-2010; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; SSB, sugar-sweetened beverage.

3.2. Distribution of metabolic biomarker levels and resulting PCs

Distributions of metabolic biomarkers can be found in Table 3. PCA identified five PCs that explained 66% of the variability in metabolic biomarker concentrations (Supplemental Figure 3). PC 1 loaded positively on glucose, insulin, C-peptide, leptin, and CRP, and negatively on adiponectin – which we considered the adiposity component (Supplemental Table 2). PC 2 loaded positively on VLDL cholesterol, triglycerides, and free fatty acids (lipids component). PC 3 loaded positively on total, HDL, and LDL cholesterol (cholesterol component). PC 4 loaded positively on MCP-1 and TNF-α (inflammation component). PC 5 loaded positively on IGF-1 (growth component).

Table 3.

Distribution of maternal early second trimester plasma metabolic biomarker concentrations (n = 295).

	Lower limit1	% ≥ Lower limit	Min	25th percentile	Median	75th percentile	Max
Adiposity
Glucose (mg/dL)	11.9	100	58.00	75.00	79.00	84.00	108.00
Insulin (pg/mL)3	58.0	84.4	4.74	75.80	202.00	617.00	11737.00
HOMA-IR	calculated	calculated	0.02	0.41	1.12	3.67	67.30
C-peptide (pg/mL)	24.0	100	401.00	894.00	1146.00	1503.00	5336.00
Leptin (pg/mL)	27.0	100	1008.00	8501.00	15024.00	24188.00	89896.00
Adiponectin (ng/mL)	1.0	100	1580.90	8477.60	11384.80	15511.60	41580.40
CRP (mg/L)	0.05	100	0.18	2.68	5.39	9.61	44.29
Lipids
VLDL cholesterol (mg/dL)	calculated	calculated	8.61	16.22	20.72	26.34	52.23
Triglycerides (mg/dL)	11.9	100	54.00	96.00	119.00	151.00	317.00
Free fatty acids (mmol/L)2	0.07	99.7	0.08	0.29	0.39	0.51	1.57
Cholesterol
Total cholesterol (mg/dL)	33.4	100	129.00	194.00	220.00	242.00	308.00
HDL cholesterol (mg/dL)	7.3	100	43.00	61.00	71.00	80.00	127.00
LDL cholesterol (mg/dL)	calculated	calculated	43.30	106.03	122.78	142.38	209.75
Inflammation
MCP-1 (pg/mL)	6.0	100	26.21	82.34	105.00	133.00	245.00
TNF-α (pg/mL)	0.3	100	0.80	3.10	4.10	5.20	13.00
Growth
IGF-1 (ng/mL)	14.4	100	44.40	99.70	118.00	144.00	311.00

¹Lower assay limits according to protocols were the minimum detectable concentration (MDC) for glucose, C-peptide, leptin, total cholesterol, HDL cholesterol, triglycerides, free fatty acids, insulin, MCP-1, and TNF-α; the limit of detection (LOD) for adiponectin and IGF-1; and lower limit of detection (LLOD) for CRP.

²Insulin concentrations for n=46 were imputed using the equation: (plate minimum/√2).

³Free fatty acid concentrations for n=1 were imputed using the equation: (minimum/√2).

CRP, C-reactive protein; HDL, high-density lipoprotein cholesterol; HOMA-IR, homeostasis model assessment of insulin resistance; IGF-1, insulin-like growth factor 1; LDL, low-density lipoprotein cholesterol; LLOD, lower limit of detection; LOD, limit of detection; Max, maximum; MCP-1, monocyte chemoattractant protein-1; MDC, minimum detectable concentration; TNF-α, tumor necrosis factor alpha; VLDL, very low-density lipoprotein.

3.3. Associations of maternal first trimester AHEI-2010 with early second trimester metabolic biomarkers

Overall, the AHEI-2010 total score was associated with lower adiposity and higher cholesterol PCs, but was not associated with the lipids, inflammation, or growth components (Figure 1 and Supplemental Table 3). Specifically, each 10-point improvements in diet quality were associated with −0.15 (95% CI: −0.27, −0.04) lower adiposity PC scores (Figure 1), which was primarily due to higher intake of fruits and nuts/legumes but lower intake of trans fat (Supplemental Table 3). Consistently, better diet quality, reflected as both total AHEI-2010 score and as individual components, was also associated with lower levels of individual adiposity biomarkers (Supplemental Tables 4 and 5). Additionally, each 10-point improvement in diet quality was associated with 0.12 (95% CI: 0.00, 0.23) higher cholesterol PC scores (Figure 1 and Supplemental Table 3) and higher individual cholesterol biomarkers (Supplemental Table 7), due to higher intakes of whole grains and EPA/DHA and lower intakes of trans fat and red/processed meat (Supplemental Table 3 and 7). Several individual AHEI-2010 components were also associated with lipid, inflammation, and growth PC scores and/or individual metabolic biomarkers (Supplemental Tables 3, 6, and 8), which generally persisted after additionally accounting for overall diet quality (Supplemental Tables 3, 4, 7, 8).

Figure 1. Relationships of maternal first trimester diet with early second trimester metabolic PCs.

Data are presented at the percent difference in metabolic PCs scores for each 10-point or 1-point increase in the AHEI-2010 or its individual components, respectively. Covariates in linear regression models included race/ethnicity, annual household income, parity, first-trimester alcohol intake, fetal sex, average total daily caloric intake in first trimester, and gestational age at blood collection for metabolic biomarker analysis. AHEI-2010, Alternative Healthy Eating Index-2010; CI, confidence interval; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; PUFA, polyunsaturated fatty acids; SSB, sugar-sweetened beverage.

3.4. Associations of phthalate/replacement biomarkers with early second trimester metabolic biomarkers

Biomarkers of most phthalates/replacements, except MCNP, MCPP, and ΣDiNCH, were associated with altered metabolic PC scores (Figure 2 and Supplemental Table 9). Each 10% increase in ΣDEHTP concentrations was associated with 0.02-point higher adiposity scores (95% CI: 0.01, 0.03), 0.01 point higher lipids scores (95% CI: 0.00, 0.02), and −0.01 point lower growth scores (95% CI: −0.02, 0.00). ΣDiNP and MBzP were associated with higher adiposity PC scores, ΣDEHP was associated with higher lipids PC scores, whereas MBzP, ΣDBP, and ΣDiBP were associated with higher inflammation PC scores. Conversely, ΣDBP was associated with lower adiposity PC scores, and MEP was associated with lower lipid, cholesterol, and growth PC scores. Associations of phthalates/replacements biomarkers with individual metabolic biomarkers generally reflected PCA findings (Supplemental Tables 10–14).

Figure 2. Relationships of maternal urinary phthalate/replacement biomarkers with early second trimester metabolic PCs.

Data are presented as the difference in metabolic PC scores for each 10% increase in urinary phthalate/replacement biomarker concentration. Covariates in linear regression models included race/ethnicity, annual household income, parity, first-trimester alcohol intake, fetal sex, and gestational age at blood collection for metabolic biomarker analysis. CI, confidence interval; MBzP, monobenzyl phthalate; MCNP, monocarboxynonyl phthalate; MCPP, mono-(3-carboxypropyl) phthalate; MEP, monoethyl phthalate; ΣDiNP, sum of di(isononyl) phthalate metabolites; ΣDEHP, sum of di(2-ethylhexyl) phthalate metabolites; ΣDBP, sum of di-n-butyl phthalate metabolites; ΣDiBP, sum of di-iso-butyl phthalate metabolites; ΣDiNCH, sum of di(isononyl) cyclohexane-1,2-dicarboxylate metabolites; ΣDEHTP, sum of di(2-ethylhexyl) terephthalate metabolites.

3.5. Associations of maternal first trimester AHEI-2010 with phthalate/replacement biomarkers

AHEI-2010 total score was only associated with ΣDEHTP, such that each 10-point improvement in diet quality was associated with −17.90% (95% CI: −27.78, −6.68) lower ΣDEHTP concentrations (Table 4), likely due to higher consumption of nuts/legumes and lower consumption of SSBs/fruit juice and trans fat. When dietary components were modeled individually, lower intake of red/processed meat was associated with higher ΣDBP and ΣDiBP but lower ΣDiNP, whereas lower intake of SSBs/fruit juice was associated with lower MBzP concentrations, and higher intake of EPA/DHA scores were associated with lower ΣDiNP concentrations. Most associations remained after accounting for maternal diet quality when modeling each individual diet factor, except the association between trans fat and ΣDEHTP was attenuated (Supplemental Table 15).

Table 4.

Associations of maternal first trimester diet with phthalate/replacement biomarker concentrations.

Chemical biomarker	ΣDEHP	ΣDiNP	MCNP	MCPP	MBzP
Index component	%Δ (95% CI)	%Δ (95% CI)	%Δ (95% CI)	%Δ (95% CI)	%Δ (95% CI)
Total score	2.35 (−5.41, 10.74)	−7.19 (−16.07, 2.64)	−1.42 (−8.03, 5.67)	0.57 (−7.65, 9.52)	−6.05 (−16.54, 5.75)
Vegetables	1.58 (−1.99, 5.28)	0.46 (−4.04, 5.17)	−0.36 (−3.45, 2.84)	2.27 (−1.61, 6.29)	−2.78 (−7.86, 2.59)
Fruit	0.45 (−2.26, 3.24)	−2.41 (−5.76, 1.07)	−0.96 (−3.31, 1.46)	−0.52 (−3.43, 2.47)	−2.63 (−6.55, 1.45)
Whole grains	−2.40 (−7.99, 3.53)	2.04 (−5.40, 10.06)	0.58 (−4.52, 5.96)	1.60 (−4.68, 8.30)	−4.08 (−12.23, 4.82)
PUFA	−2.02 (−6.92, 3.14)	2.52 (−4.00, 9.49)	−1.44 (−5.80, 3.12)	0.37 (−5.05, 6.11)	−4.52 (−11.61, 3.13)
EPA/DHA	−1.13 (−4.29, 2.13)	−3.89 (−7.79, 0.17)#	−0.94 (−3.73, 1.94)	−1.24 (−4.65, 2.29)	1.34 (−3.49, 6.42)
Nuts/legumes	1.31 (−1.35, 4.05)	−1.08 (−4.41, 2.36)	0.09 (−2.24, 2.48)	0.96 (−1.91, 3.92)	0.77 (−3.20, 4.90)
SSBs/fruit juice	0.59 (−1.64, 2.87)	−0.52 (−3.34, 2.38)	0.49 (−1.47, 2.50)	−1.10 (−3.47, 1.33)	−2.97 (−6.17, 0.34)#
Trans fat	3.69 (−3.78, 11.73)	−1.79 (−10.76, 8.09)	3.04 (−3.52, 10.05)	4.77 (−3.35, 13.58)	−3.97 (−14.18, 7.47)
Red/processed meat	0.78 (−3.33, 5.08)	−6.83 (−11.63, −1.78)*	−1.75 (−5.29, 1.92)	0.21 (−4.21, 4.83)	2.46 (−3.77, 9.09)
Sodium	1.33 (−4.53, 7.56)	0.30 (−7.08, 8.26)	−0.34 (−5.44, 5.04)	1.28 (−5.05, 8.03)	5.99 (−3.08, 15.92)
Chemical biomarker	MEP	ΣDBP	ΣDiBP	ΣDiNCH	ΣDEHTP
Index component	%Δ (95% CI)	%Δ (95% CI)	%Δ (95% CI)	%Δ (95% CI)	%Δ (95% CI)
Total score	−1.56 (−11.49, 9.48)	5.26 (−2.16, 13.25)	3.64 (−5.77, 13.99)	−2.03 (−9.94, 6.59)	−17.90 (−27.78, −6.68)*
Vegetables	0.66 (−4.09, 5.64)	2.09 (−1.25, 5.55)	1.19 (−3.09, 5.67)	0.14 (−3.62, 4.04)	−2.36 (−7.96, 3.58)
Fruit	−1.01 (−4.6, 2.71)	2.01 (−0.55, 4.64)	1.81 (−1.50, 5.23)	1.34 (−1.59, 4.34)	−2.31 (−6.63, 2.20)
Whole grains	−4.21 (−11.54, 3.72)	1.61 (−3.83, 7.35)	−4.32 (−10.90, 2.75)	0.39 (−5.76, 6.93)	−3.76 (−12.70, 6.10)
PUFA	−1.99 (−8.55, 5.04)	−1.13 (−5.75, 3.73)	−2.69 (−8.55, 3.53)	1.80 (−3.64, 7.55)	−3.82 (−11.63, 4.69)
EPA/DHA	−1.56 (−5.78, 2.84)	−0.49 (−3.46, 2.57)	−0.44 (−4.27, 3.55)	−0.46 (−3.86, 3.07)	0.43 (−4.82, 5.97)
Nuts/legumes	1.62 (−1.97, 5.35)	0.87 (−1.60, 3.40)	2.31 (−0.93, 5.66)	−0.18 (−2.99, 2.71)	−6.26 (−10.24, −2.09)*
SSBs/fruit juice	−0.28 (−3.26, 2.78)	−0.71 (−2.75, 1.38)	−1.72 (−4.34, 0.97)	−1.88 (−4.19, 0.49)	−6.26 (−9.60, −2.80)*
Trans fat	−7.60 (−16.44, 2.17)	5.52 (−1.57, 13.11)	6.97 (−2.25, 17.06)	−0.95 (−8.56, 7.30)	−10.32 (−20.69, 1.41)#
Red/processed meat	1.23 (−4.31, 7.09)	4.93 (0.98, 9.03)*	4.90 (−0.23, 10.30)#	−1.77 (−6.05, 2.71)	−0.80 (−7.40, 6.28)
Sodium	2.92 (−5.03, 11.55)	3.30 (−2.28, 9.19)	5.29 (−2.02, 13.14)	−1.09 (−7.20, 5.43)	−1.00 (−10.29, 9.25)

Data are presented as the percent difference in urinary phthalate/replacement biomarker concentrations for every 10 point or 1 point increase in the AHEI-2010 total score or individual components, respectively. Models account for race/ethnicity, annual household income, parity, alcohol intake in the first trimester, fetal sex, and average total daily caloric intake. AHEI-2010, Alternative Healthy Eating Index-2010; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; MBzP, monobenzyl phthalate; MCNP, monocarboxynonyl phthalate; MCPP, mono-(3-carboxypropyl) phthalate; MEP, monoethyl phthalate; PUFA, polyunsaturated fatty acids; SSB, sugar-sweetened beverage; ΣDiNP, sum of di(isononyl) phthalate metabolites; ΣDEHP, sum of di(2-ethylhexyl) phthalate metabolites; ΣDBP, sum of di-n-butyl phthalate metabolites; ΣDiBP, sum of di-iso-butyl phthalate metabolites; ΣDiNCH, sum of di(isononyl) cyclohexane-1,2-dicarboxylate metabolites; ΣDEHTP, sum of di(2-ethylhexyl) terephthalate metabolites.

^*P<0.05,

^#P<0.1.

3.6. Associations of AHEI-2010 with metabolic biomarkers mediated by phthalates/replacements

Only associations of AHEI-2010 total score and individual components with the adiposity PC were mediated by ΣDEHTP (Figure 3). Specifically, lower ΣDEHTP concentrations explained 21% of the total association between better diet quality and lower adiposity PC scores. When dietary components were modeled individually, lower ΣDEHTP concentrations explained 25% of the total association between higher nut/legume consumption and lower adiposity PC scores and 15% of the total association between lower trans fat intake and lower adiposity PC scores. Associations of AHEI-2010 total score with C-peptide, leptin, adiponectin, and CRP, nuts/legumes with leptin and CRP, SSBs/fruit juice with leptin and adiponectin, and trans fat with C-peptide, leptin, adiponectin, and CRP were partially mediated by ΣDEHTP; other tested relationships showed no evidence of being meaningful (Supplemental Table 16).

Figure 3. Associations of AHEI-2010 (A) total score, (B) nuts/legumes, (C) trans-fat with adiposity component mediated by ΣDEHTP urinary concentrations.

Reported data represent the difference in adiposity PC scores for each 10 point or 1 point increase in AHEI-2010 total score or individual component, respectively. ΣDEHTP was ln-transformed, and an interaction between the AHEI-2010 or individual component and ΣDEHTP was included. Models included the following covariates: race/ethnicity, educational attainment, annual household income, parity, alcohol intake in the first trimester, fetal sex, and gestational age at blood draw. AHEI, Alternative Healthy Eating Index; ΣDEHTP, sum of di(2-ethylhexyl) terephthalate metabolites.

4. DISCUSSION

4.1. Summary of findings

As hypothesized, associations of better maternal diet quality or some individual diet components with favorable metabolic status were partially (up to 25%) explained by lower biomarker concentrations of the plasticizer DEHTP. These findings provide further evidence that diet is a modifiable lifestyle factor that can be targeted to decrease exposure to phthalates, with potential short- and long-term benefits for maternal and child health.

4.2. Better maternal diet was associated with favorable adiposity and cholesterol outcomes

Diet indices are powerful tools for evaluating the roles of whole diets in numerous chronic diseases (Pacyga et al., 2023a). Findings from the current study support prior limited literature (Chia et al., 2019; Raghavan et al., 2019), demonstrating that better diet quality was associated with favorable adiposity (lower) and cholesterol (higher) scores. Although counterintuitive, cholesterol biomarkers are expected to increase by 25–50% during pregnancy (Grimes and Wild, 2000; Lain and Catalano, 2007) to support the formation and maintenance of cell membranes and serve as a precursor for gestational sex-steroid hormones (Lain and Catalano, 2007). Of note, our analytic sample is comprised of generally healthy women and excluded those with preexisting conditions. Thus, while prior studies evaluated associations of maternal diet with clinical metabolic outcomes (e.g. GDM or poor glucose tolerance) (Lindsay et al., 2022; Raghavan et al., 2019), we evaluated a large panel of metabolic biomarkers to highlight potential sub-clinical metabolic dysregulations. Given that normal pregnancy adaptations and fetal development depend on dynamic metabolic changes, substantially more needs to be understood about the roles maternal diet plays in regulating metabolic health from conception through birth.

4.3. Phthalate and replacement biomarkers were associated with poor metabolic outcomes

In the current study, some phthalate biomarkers were associated with less favorable adiposity (ΣDiNP, MBzP), lipid (MEP), and inflammation scores (MBzP, ΣDBP, ΣDiBP). Notably, the sum of urinary biomarkers of DEHTP, a plasticizer replacement, was associated with less favorable gestational metabolic health, including higher glucose, insulin, C-peptide, leptin, and CRP levels along with lower adiponectin and HDL cholesterol levels. Most prior observational studies conducted during pregnancy have identified associations of ortho-phthalates with higher odds of GDM (Chen et al., 2023; Fisher et al., 2018b; Gao et al., 2021; James-Todd et al., 2022; Liang et al., 2022; Martinez-Ibarra et al., 2019; Shaffer et al., 2019a; Wang et al., 2023; Zhang et al., 2017; Zukin et al., 2021), but few studies (if any) extensively measured other biomarkers of adiposity (e.g. C-peptide, leptin, adiponectin, IGF-1) or lipid homeostasis (e.g. VLDL cholesterol, free fatty acids). Consistent with current findings, other studies have shown that higher urinary concentrations of ortho-phthalate biomarkers (ΣDiNP, MBzP) in pregnancy were associated with higher second-trimester glucose concentrations (Fisher et al., 2018a; Gao et al., 2021; Shaffer et al., 2019a). However, findings have been inconsistent in other studies focusing on glucose (Gao et al., 2021; James-Todd et al., 2018; James-Todd et al., 2016; Liang et al., 2022; Robledo et al., 2015; Vuong et al., 2021; Wang et al., 2023; Zukin et al., 2021), cholesterol or triglycerides (Jia et al., 2015; Minguez-Alarcon et al., 2022; Vuong et al., 2021), and inflammatory biomarkers (Ferguson et al., 2014; Ferguson et al., 2015; Lee et al., 2023) – likely due to differences in blood collection (fasting vs non-fasting), phthalate/replacement matrix for exposure assessment (gold standard urine vs. blood), urine collection (number and timing of samples), and geographical and temporal trends of phthalate/replacement exposure. Importantly, no studies to date have assessed associations of replacement plasticizers with metabolic health during pregnancy. The current findings warrant concerns about regrettable substitution since prior limited studies suggest that compared to ortho-phthalates, exposure to DEHTP (a tere-phthalate) may also adversely affect maternal metabolism in pregnancy (Deierlein et al., 2022; Wu et al., 2021), possibly via similar mechanisms of action (Philips et al., 2017; Rajkumar et al., 2022; Sheikh et al., 2016; Veiga-Lopez et al., 2018). This is of concern, as DEHTP exposure may be increasing (CDC, 2021; Lessmann et al., 2016; Qu et al., 2022; Runkel et al., 2022), including among pregnant women (Pacyga et al., 2022).

4.4. Better maternal diet was associated with lower phthalate biomarker concentrations

Due to the abundance of phthalates and their plasticizer replacements along the food processing chain and in food packaging materials, diet has become an important source of phthalates in the general population (Serrano et al., 2014; Woodruff et al., 2011). To supplement food monitoring studies, our study and others have leveraged data from pregnancy observational studies to evaluate dietary predictors of phthalate/replacement exposure (Pacyga et al., 2019). Unfortunately, findings from prior observational studies in pregnant populations have been mixed, with little consistency related to specific dietary sources of phthalates (Pacyga et al., 2019). Further, most studies have not assessed dietary determinants of replacements such as DiNCH and DEHTP. To our knowledge, this is one of the first studies in pregnant women to evaluate whether AHEI-2010 total score and individual components predict urinary phthalate/replacement biomarker concentrations – identifying a link between diet quality and DEHTP. Specifically, healthier dietary patterns (along with higher intakes of nuts and legumes, but lower intakes of SSBs, fruit juice, and trans fat) were associated with lower urinary DEHTP biomarker concentrations, which has not been previously reported.

Similarly, healthier dietary behaviors (higher intakes of EPA/DHA, lower intakes of red/processed meat or SSBs/fruit juice) were also associated with lower concentrations of ΣDiNP and BBzP, which is generally consistent with prior food monitoring and pregnancy observational studies (Pacyga et al., 2019; Serrano et al., 2014). Unexpectedly, we also observed that lower intake of red or processed meat was associated with higher concentrations of DBP and DiBP biomarkers. Food monitoring studies have detected these phthalates, albeit in low concentrations, in meat products via food packaging materials (Pacyga et al., 2019); however, DBP and DiBP have also been detected in certain medication and supplement coatings (Kelley et al., 2012), so women who eat less meat may be more likely to take supplements to increase nutrients they lack from their diets (Koivuniemi et al., 2022). Additional well-designed, large-scale studies could evaluate AHEI-2010 along with other diet indices to confirm our findings.

4.5. Better maternal diet with favorable adiposity markers was explained by lower DEHTP

Our most salient finding was that 15–25% of the association between better diet and more favorable adiposity scores was due to higher diet quality being associated with lower DEHTP biomarker concentrations. Our findings suggest that in addition to the known anti-inflammatory and antioxidant properties of healthier diets, healthier diets may also promote metabolic heath by decreasing exposure to metabolic-disrupting chemicals. To our knowledge, no other prior observational studies formally assessed chemical exposures as potential mediators of the relationships between diet and the health of pregnant women. Several intervention studies in non-pregnant adults evaluated the health impacts of dietary interventions that reduce chemical exposures, as reviewed by Corbett et al (Corbett et al., 2022). For example, a randomized crossover trial in adults ≥60 years reported higher bisphenol A urinary concentrations in response to consuming beverages in cans (compared to glass) acutely increased systolic blood pressure over the one week study period (Bae and Hong, 2015). However, most other studies reviewed in Corbett et al. did not investigate whether the intervention was linked to changes in health outcomes (Corbett et al., 2022). Despite the evidence that small changes in dietary habits can reduce exposure to metabolic disrupting chemicals, changing dietary behaviors is often challenging, especially beyond the intervention. However, evidence suggests that women may be willing to make changes if they see tangible improvements in their own health and their child’s health (Forbes et al., 2018). Therefore, observational studies such as ours could support the development of targeted interventions that evaluate whether reductions in chemical exposures improve maternal metabolic health in pregnancy, and ultimately the long-term health of children and mothers

4.6. Strengths and limitations

This study had several notable strengths and limitations. First, the current study may have been underpowered, particularly in mediation analyses. However, as we have shown previously (Pacyga et al., 2022; Pacyga et al., 2023b), women in this study reflect the full parent study. Also, most women had measurable concentrations of DiNCH and DEHTP metabolites, which allowed us to evaluate dietary determinants and metabolic-disrupting potential of less studied replacement chemicals. Second, maternal first trimester diet data were collected using an FFQ and not a 24-hour recall. Although FFQs are prone to imprecision due to recall bias and misreporting, they are easy to administer, inexpensive, and reflect patterns of long-term food intake in pregnant individuals (Boucher et al., 2006; McGowan et al., 2014; Venter et al., 2006). Third, the FFQ assessed usual dietary intakes over a three-months period, while urinary phthalate metabolite concentrations reflect exposure within 24–48 hours of collection, due to the non-persistence in the body and episodic day-to-day exposures. However, we assessed urinary phthalate/replacement metabolite concentrations in pooled samples of up-to-five first-morning urine samples collected across pregnancy, which better estimates usual exposure across pregnancy than biomarker concentrations from a single sample (Casas et al., 2018; Rosen et al., 2023; Shin et al., 2019; Vernet et al., 2019). Fourth, AHEI-2010 may not be best suited for identifying dietary sources of chemical exposures because it does not consider food processing/packaging, which is how many phthalates/replacements are likely introduced into the food supply. However, this is one of the first studies to validate the AHEI-2010 against urinary biomarkers of phthalates/replacements, which is critical given that the AHEI is commonly used to evaluate associations of dietary patterns with pregnancy and birth outcomes (Chia et al., 2019). Future studies could consider investigating other indices, such as an ultra-processed food index (Menichetti et al., 2023), which may be a better predictor of phthalate/replacement exposure, but also adverse metabolic health.

Fifth, it may be challenging to generalize our findings to other pregnant populations because women in our study are mostly healthy and highly educated. Nevertheless, women in our study had phthalate biomarker concentrations similar to same-age women from NHANES. Sixth, as with most observational studies, residual confounding is possible. For example, physical activity (which was not assessed in our study) may influence associations of MEP with lower cholesterol and lipids scores because physical activity affects metabolism and may increase exposure to MEP via use of personal care products (Aniansson et al., 2016). However, we accounted for pertinent health, lifestyle, and sociodemographic factors and informed covariate selection using a priori consideration and prior studies. Certain unmeasured factors may also already be inherently accounted for in our study design given the healthy, homogenous nature of our sample. Finally, pregnant women have daily exposure to many chemicals beyond the ones we investigated in this study, and future studies could consider accounting for confounding by co-exposures (Kortenkamp and Faust, 2018).

5. CONCLUSION

In the current study of healthy, highly-educated U.S. pregnant women, associations of better first trimester maternal diet quality with favorable early second trimester adiposity biomarkers were partially explained by lower urinary DEHTP biomarker concentrations. Because, to our knowledge, no other studies have formally investigated whether phthalates/replacements mediate associations of maternal diet quality with maternal metabolic health, future studies are warranted to corroborate our findings, incorporate mixtures approaches to account for co-exposures, and investigate the lasting metabolic consequences of gestational phthalate/replacement exposure from diet.

HIGHLIGHTS.

• Better diet quality was associated with favorable adiposity biomarkers in pregnancy

• Better diet was associated with lower urinary DEHTP metabolite levels (ΣDEHTP)

• Lower maternal ΣDEHTP was associated with favorable adiposity biomarker scores

• ΣDEHTP mediated up to 25% of the association between maternal diet and adiposity

Abbreviations:

AHEI

Alternative Healthy Eating Index

BBzP

benzylbutyl phthalate

BMI

body mass index

CDC

Centers for Disease Control and Prevention

CI

confidence interval

CRP

C-reactive protein

DAG

directed acyclic graph

DEHTP

di(2-ethylhexyl) terephthalate

DEP

diethyl phthalate

DHA

docosahexaenoic acid

DiDP

di-isodecyl phthalate

DiNCH

cyclohexane-1,2-dicarboxylic acid diisononyl ester

DOP

di-n-octyl phthalate

EPA

eicosapentaenoic acid

FFQ

food frequency questionnaire

GDM

gestational diabetes mellitus

HDL

high-density lipoprotein

HOMA-IR

homeostasis model assessment of insulin resistance

I-KIDS

Illinois Kids Development Study

IL-6

interleukin-6

LDL

low-density lipoprotein

MBP

mono-n-butyl phthalate

MBzP

monobenzyl phthalate

MCOCH

cyclohexane-1,2-dicarboxylic acid-mono(carboxyoctyl) ester

MCNP

monocarboxynonyl phthalate

MCOP

monocarboxyoctyl phthalate

MCP-1

monocyte chemoattractant protein-1

MCPP

mono(3- carboxypropyl) phthalate

MECPP

mono(2-ethyl-5-carboxypentyl) phthalate

MECPTP

mono(2- ethyl-5-carboxypentyl) terephthalate

MEHHP

mono(2-ethyl-5-hydroxyhexyl) phthalate

MEHHTP

mono(2-ethyl-5-hydroxyhexyl) terephthalate

MEHP

mono(2-ethylhexyl) phthalate (MEHP)

MEOHP

mono(2-ethyl-5-oxohexyl) phthalate

MEP

monoethyl phthalate

MHBP

monohydroxybutyl phthalate

MHiNCH

cyclohexane-1,2- dicarboxylic acid-monohydroxy isononyl ester

MiBP

mono-isobutyl phthalate

MiHBP

mono-hydroxy-isobutyl phthalate

MiNP

mono-isononyl phthalate

MONP

monooxononyl phthalate

PC

principal component

PCA

principal component analysis

PUFAs

polyunsaturated fatty acids

SSBs

sugar-sweetened beverages

ΣDBP

sum of di-n-butyl phthalate metabolites

ΣDEHP

sum of di(2-ethylhexyl) phthalate metabolites

ΣDEHTP

sum of di-2-ethylhexyl terephthalate urinary metabolites

ΣDiBP

sum of di-isobutyl phthalate metabolites

ΣDiNCH

sum of cyclohexane-1,2-dicarboxylic acid diisononyl ester metabolites

ΣDiNP

sum of di-isononyl phthalate metabolites

TNF-α

tumor necrosis factor-alpha

VLDL

very low-density lipoprotein

DATA SHARING

Data described in the manuscript, code book, and analytic code will not be made available because the authors do not have permission to share data.