Prenatal phthalate exposure measurement: A comparison of metabolites quantified in prenatal maternal urine and newborn’s meconium.

Abstract

Phthalates are chemicals suspected to adversely affect fetal neurodevelopment, but quantifying the fetal exposure is challenging. While prenatal phthalate exposure is commonly quantified in maternal urine, the newborn’s meconium may better capture cumulative prenatal exposure. Currently, data on phthalates measured in meconium is sparse. We measured phthalate metabolites in 183 maternal second and 140 third trimester (T2, T3) urine, and in 190 meconium samples collected in an autism enriched-risk pregnancy cohort of 236 mothers. Eleven and eight metabolites were detected in over 90% of urine and meconium samples, respectively. Hydrophilic and hydrophobic metabolites were detected in both biosamples. Most urine phthalate metabolite distributions were similar between T2 and T3. Among metabolites detected in both biosamples, those of di(2-ethylhexyl) phthalate displayed a similar pattern in magnitude across metabolite type. Specifically, T2 creatinine adjusted distribution [median (25%, 75%)] of urine measured mono(2-ethylhexyl-carboxypentyl) (MECPP), mono(2-ethyl-5-hydroxyhexyl) (MEHHP), and mono(2-ethyl-5-oxohexyl) phthalate (MEOHP) were 18.8(11.9, 31.4), 11.8(7.2, 19.1), and 8.9(6.2, 14.2) ng/mg. In meconium these were 16.6(10.9, 23.7), 2.5(1.5, 3.8), and 1.3(0.8, 2.3) ng/g, respectively. Metabolite-to-metabolite correlations were lower in meconium than urine, but patterns were similar. For example, correlation (95% CI) between mono(2-ethylhexyl) phthalate and MECPP was 0.73 (0.66, 0.78), and between MEOHP and MEHHP was 0.96 (0.95, 0.97) in urine as compared to 0.10 (−0.04, 0.24) and 0.31 (0.18, 0.43) respectively in meconium. Correlations between same metabolites measured in urine and meconium were low and differed by metabolite and trimester. Correlation between MEHHP in urine and meconium, for example, was 0.20 (0.008, 0.37) at T3, but 0.05 (−0.12, 0.21) at T2. Our study provides evidence of general population-level prenatal phthalate exposure in a population at high risk for neurodevelopmental disorders and supports the utility of meconium to measure prenatal phthalate exposure but provides little evidence of correlation with exposure measured in prenatal maternal urine.

Graphical Abstract

1. Introduction

¹Phthalates are a group of chemicals extensively used in plastics and personal products and their widespread use causes ubiquitous human exposure on a global scale(Koch et al., 2017). Phthalates are usually grouped into high and low molecular weight (HMWP, LMWP) based on the length of their carbon chain(NRC, 2008). HMWP like di(2-ethylhexyl) phthalate (DEHP) are commonly used as plasticizers in diverse products such as plastic packaging, and vinyl flooring, and have also been detected in processed food(Schettler, 2006; Serrano et al., 2014; Zota et al., 2016). LMWP like diethyl phthalate (DEP) are industrial solvents and are also used as stabilizers in perfumes, nail polish, and other personal care products(ATSDR, 1995, 2001). Since the chemical bond binding phthalates to plastics is not highly resistant to heat or organic solvents(Guart et al., 2011), they can leach into the environment. Human exposure can be through ingestion, inhalation, and dermal absorption, and commonly used phthalates like DEHP, DEP, butylbenzyl phthalate (BBzP), and di-n-butyl phthalate (DnBP) have been detected in national biomonitoring programs(CDC, 2017). Since LMWP like DEP are extensively used in personal care products, there is a higher potential for such phthalate exposure in women of childbearing age(Koniecki et al., 2011). In the US, data from the National Health and Nutrition Examination Survey (NHANES) indicates that concentration of all detected phthalates were higher among females than males(Silva, Barr, et al., 2004).

Once absorbed, phthalate diesters are rapidly metabolized into monoesters. For long carbon chain phthalates, primary monoester metabolites are further oxidized into secondary metabolites. The end products are glucuronidated and eliminated through urine and feces(Frederiksen et al., 2007). These primary and secondary metabolites have been detected in several biosamples including urine and meconium(Arbuckle et al., 2016; Silva, Reidy, et al., 2004). Since most current epidemiological studies assessing prenatal phthalate exposure have measured exposure in maternal urine, most reports on magnitude of exposure and correlation patterns of metabolites are available only within urine. Since several primary and secondary metabolites are products of the same long chained parent diester- for example, as in figure 1, DEHP is metabolized into the primary metabolite mono(2-ethylhexyl) phthalate (MEHP), and several oxidative metabolites, these are expected to be highly correlated. Correlation patterns can also be due to exposure from common sources. Most studies have reported high correlations between the DEHP metabolites and between DnBP, DiBP, and BBzP metabolites(LaRocca et al., 2014; Meeker et al., 2009).

Fig.1.

Metabolism of DEHP with selected metabolites

Meconium is the newborn’s first stool and is usually passed within first 48 hours of life. It does not require invasive collection procedures and is easily collectible but has not been widely used in epidemiological studies to assess prenatal phthalate exposure. Meconium accumulates from approximately 12^th week of pregnancy and contains endogenous and exogenous chemical exposure to the fetus, and so it has potential to provide a summary of cumulative exposure in pregnancy(Bearer, 2003; Shwachman, 1981). Also, as a fetal product, it is proximal to the fetal compartment than maternal urine. Meconium has been used in several studies to measure exposure to pesticides, persistent organic pollutants, endogenous compounds such as androgens(Frey et al., 2017; Ostrea jr. et al., 2009), but currently only few studies have used it to measure prenatal phthalate exposure, and only one epidemiological study- the Plastics and Personal-care products use in pregnancy (P4), evaluated correlation between urine and meconium measured metabolites. In that study of 80 mother/child dyads, eight phthalate metabolites were detected in 60% of 54 meconium samples(Arbuckle et al., 2016) and moderate magnitude (~.35) and statistically significant correlations for three (mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethyl-5-oxohexyl) (MEOHP), and monoethyl phthalate (MEP) ) metabolites between meconium and maternal urine were reported. In addition, metabolites from the same parent diester detected in urine were shown to be highly correlated.

Considering its potential advantages, there is a need to evaluate feasibility of measuring phthalate metabolites in meconium and to further explore its ability to capture prenatal phthalate exposure. To elaborate, since phthalate metabolites are currently commonly measured in urine, correlation patterns between metabolites in meconium can be compared to those established in urine in order to evaluate the validity of results. In addition, since chemicals measured in meconium may be cumulative over pregnancy, they are expected to be positively correlated with those measured in urine. Thus, in this study we measured phthalate metabolites in meconium and maternal urine, compared detected correlation patterns in meconium to those in maternal urine, and evaluated correlations with same metabolites measured in meconium and urine.

2. Methods

2.1. Study population

The Early Autism Risk Longitudinal Investigation (EARLI) was a prospective cohort study of pregnant mothers who previously had a child diagnosed with Autism Spectrum Disorder. EARLI enrolled from four sites in Philadelphia, Baltimore, and California (UC Davis, and Northern California Kaiser Permanente) from 2009 to 2012. Eligibility criteria included ability to speak English or Spanish, older than 18 years, living within 2 hours of the study site, and less than 29 weeks gestation. Additional details of EARLI study design are reported elsewhere(Newschaffer et al., 2012). Enrollment interview was conducted at the study clinic where mothers were administered a comprehensive demographic questionnaire, and biosamples collected.

2.2. Maternal urine collection

Mothers were provided with urine collection kits containing a plastic ‘urine hat’ commode specimen collection container and polypropylene storage containers prior to second and third trimester (T2, T3) prenatal home visits, and instructed to collect the entire first morning void of urine and label it with time and date of collection. 183, and 140 urine samples were collected at the T2 and T3 visits, respectively. These samples were kept refrigerated and collected by study personnel within 24 hours. Urine samples were aliquoted and frozen at −20° C at the local lab until batch shipped to the study biorepository and kept in long-term storage at −190° C.

2.3. Meconium collection

Meconium could have been collected at the hospital or at home by parents. As with urine, mothers were provided with a sampling kit for meconium collection. EARLI had agreements with the hospitals to allow collection and storage of meconium if passed at the hospital. If the child was discharged without passing meconium, parents were instructed to scrape all the meconium off the diaper using study provided wooden tongue depressor into the 30 ml polypropylene collection container. 190 meconium samples were collected in the study. Time and date of sample collection was recorded on the container and it was kept frozen in their home freezer until collected by study staff. Samples were kept at −20° C and batch shipped to the same biorepository where it was homogenized, aliquoted and kept in long-term storage at −190° C. Time to meconium collection was estimated using cord blood/placenta collection time and categorized into ≤12 hours, >12 hours, and those with inadequate data to compute time to meconium collection.

The EARLI study was not specifically designed to eliminate external phthalate contamination issues and did not use prescreened equipment in biosample collection or field blanks in collection kits. However, once samples were collected, they were stored in temperature-controlled settings to avoid degradation of the chemical.

2.4. Phthalate metabolite measurement

2.4.1. Urine

Urine phthalate metabolites were measured by the Centers for Disease Control and Prevention (CDC). After enzymatic deconjugation to separate glucuronidated conjugates, total urinary concentrations of 14 phthalate metabolites were quantified using online solid phase extraction coupled to high performance liquid chromatography-electrospray ionization-tandem mass spectrometry (HPLC_ESI-MS/MS), based on modification of a published method(Silva et al., 2013). Limits of detection (LOD) ranged from 0.2 (mono-isononyl phthalate (MiNP)), to 1.2 ng/ml (MEP). Accuracy ranged from 95 to 105%, as calculated from the recovery of three spiking levels. Low-concentration and high-concentration quality control materials were prepared with pooled human urine and analyzed with the standards, reagent blanks and the study samples. The quality control concentrations were evaluated using standard probability rules for each analytic run(Caudill et al., 2008). Creatinine (ng/ml) was also measured by the same lab and we divided metabolite values by creatinine to adjust for urine dilution and report results in ng/mg.

2.4.2. Meconium

Thirteen phthalate metabolites were quantified in EARLI meconium by the Exposure Science Laboratory at the A.J. Drexel Autism Institute (AJDAI). The meconium samples were first pre-treated with phosphoric acid to denature active enzymes and enzymatic deconjugation. The analytes were extracted using liquid-liquid extraction and then analyzed by high-performance liquid chromatography-electrospray ionization tandem mass spectrometry (HPLC-ESI-MS)/High resolution MS modified from a published method(Arbuckle et al., 2016). Laboratory analysts were blinded to sample identity during sample processing and data analysis and samples were randomized into 96-sample blocks for processing and LC-MS analysis. Briefly, approximately 30 mg of meconium was measured in a pre-chilled 4C glass tube, 1 mL of 100 mM H₃PO₄ in LC-MS grade water was added using a glass pipette and then the sample was vortex mixed for 10 minutes to denature esterases. To homogenize the sample, the acid-denatured meconium was bath sonicated for 10 minutes in an ice-bath in a Branson bath sonicator, vortexed another 10 minutes, then neutralized with 50 μL of NaOH (1M) to pH 6.5. Enzymatic deconjugation was conducted by adding 50 μL of β-glucuronidase (5.3 U/μL in 1M ammonium acetate pH 6.5) then incubated at 37 °C for 3 hours. Next, 50 μL of phthalate internal standard mix (Supplementary table 1) in LC-MS grade methanol mix was prepared by a 1:200 dilution from a premade stock. The sample was mixed, then 50 μL of acetic acid, 50 μL saturated NaCl, and 5 mL of methyl tert-butyl ether was added to each tube. The samples were inverted, then vortex mixed for 10 minutes, centrifuged for 5 min at 4 °C using the slow acceleration/deceleration setting, then stored at −80°C to freeze the lower aqueous layer. The upper organic layer was moved to a new glass tube and then evaporated to dryness under nitrogen gas. The sample was resuspended in 50 μL of 95/5 (v/v) water: methanol, transferred to a 96-well autosampler plate, centrifuged 2000 x g for 5 minutes, and then 10 μL was injected for analysis on a Ultimate 3000 UHPLC coupled to a Q Exactive Plus High resolution mass spectrometer operating in negative ion mode alternating between full scan and data independent acquisition. Alongside the samples, calibration curves from 4000 pg/sample to 0.0625 pg/sample were prepared in methanol and hexane stripped pooled meconium from healthy newborns, and quality controls were prepared at spiked levels of 25 replicates of 0 pg/sample, 3 replicates of 2 pg/sample, and 3 replicates of 100 pg/sample added into pooled meconium from health newborns prepared identically to the EARLI samples. Area of each analyte peak corresponding to the most intense ion was integrated in Tracefinder 4.1 (Thermo Fisher, San Diego, CA) and normalized to peak area of the matched stable isotope labeled internal standard or if no commercially available internal standard was found, the internal standard with the closest retention time. All standards and isotope labeled internal standards were from Cambridge Isotope Labs (Andover, MA). Runs were accepted when the quality controls returned Coefficient of variations (CV) under 20% and the replicate un-spiked meconium were under 5% CV consistent with the original method validation. The values were calculated normalized to the wet weight of meconium, and limit of detection (LOD) was 0.0005 ng/g (0.5 pg/g) based on replicate runs of the calibration curve prepared in hexane stripped meconium at 3 times the lowest point where the difference between spiked and experimentally determined amount was under 25%, but for any value where a signal was detectable the laboratory reported this number to the analysts conducting analysis of the data. Table 1 lists the metabolites measured and the biosample of assessment. Ratio of secondary to primary (MEOHP/ MEHHP/ mono-2-ethyl-5-carboxypentyl phthalate (MECPP), to MEHP), and between secondary metabolites (MEHHP/MEOHP, MECPP/MEOHP) of DEHP were computed in both urine and meconium measured phthalates to evaluate differences in maternal and fetal metabolism.

Table 1.

Phthalate diesters, metabolites & biosample for assessment

Parent Phthalate	Phthalate Metabolite	Abbreviation	Biosample	Primary/Secondary metabolite
DBP	mono-n -butyl	MBP	Urine/ Meconium	Primary
DBP	mono-3-hydroxy-n-butyl	MHBP	Urine	Secondary
DiBP	mono-2-hydroxy-isobutyl	MHiBP	Urine	Secondary
DiBP	mono-isobutyl	MiBP	Urine/Meconium	Primary
DEP	monoethyl	MEP	Urine/Meconium	Primary
DCHP	monocyclohexyl	MCHP	Meconium	Primary
BBzP	monobenzyl	MBzP	Urine/Meconium	Primary
DiDP	mono carboxyisononyl	MCNP	Urine	Secondary
DiNP	mono carboxyisooctyl	MCOP	Urine	Secondary
DOP/DBP, others	mono-3-carboxypropyl	MCPP	Urine	Secondary
DEHP	mono(2-ethyl-5-carboxypentyl)	MECPP	Urine/Meconium	Secondary
DEHP	mono(2-ethyl-5-hydroxyhexyl)	MEHHP	Urine/Meconium	Secondary
DEHP	mono(2-ethylhexyl)	MEHP	Urine/Meconium	Primary
DEHP	mono(2-ethyl-5-oxohexyl)	MEOHP	Urine/Meconium	Secondary
DiNP	mono-isononyl	MiNP	Urine/Meconium	Primary
DiPP	mono-isopentyl	MiPP	Meconium	Primary
DOP	mono-n-octyl	MOP	Meconium	Primary
DPP	mono-n-pentyl	MPP	Meconium	Primary

DBP, di-n-butyl phthalate; DiBP, di-isobutyl phthalate; DEP, di-ethyl phthalate; DCHP, di-cyclohexyl phthalate; BBzP, Butyl-benzyl phthalate; DiDP, di-isodecyl phthalate; DiNP, di-isononyl phthalate; DOP, di-n-octyl phthalate; DEHP, di(2-ethylhexyl) phthalate; DPP, di-n-pentyl phthalate; DiPP, di-isopentyl phthalate

2.5. Covariates

Socio-demographic data was collected through face-to-face surveys administered during the enrollment visit. To describe the study population, continuous demographic variables like maternal age were summarized using means and standard deviation, and categorical data on maternal race/ethnicity, education, income, parity, pre-pregnancy BMI, study site, birth year, and child sex were summarized using frequencies. We used study site, child sex, and birth year of the child as potential effect modifiers of the correlation between the metabolites assessed in urine and meconium.

2.6. Statistical Analysis

Machine generated values of metabolites less than LOD but higher than zero were left unchanged, but zero values were set to $\frac{LOD}{\sqrt{2}}$. (Hornung et al., 1990). Since distributions were skewed, phthalate metabolites were summarized using geometric means (GM) and 95% confidence intervals (CI). We compared the GM (95% CI) of individual metabolite distributions to the GM (95% CI) of metabolite distributions from the P4 study (meconium); the 2011–12 National Health and Nutrition Examination Survey (NHANES) biomonitoring study and the Health Outcomes and Measures of the Environment (HOME) study, a Cincinnati based pregnancy cohort (urine). We used Spearman correlations to evaluate correlation of phthalate metabolites within urine and meconium. Considering the possibility that phthalate levels in meconium might differ between those children born preterm, term, and post term, we evaluated these correlations by gestational age categories < 38, 38 - <40, and > 40 weeks. In addition, using (1-correlation) as a measure of dissimilarity, we conducted agglomerative hierarchical clustering with complete linkage using correlations of the nine metabolites measured in common between urine and meconium samples to further explain the correlations(Johnson et al., 2002). This technique uses the dissimilarity index to combine metabolites that are similar. No structure is imposed on the data, but natural patterns are allowed to develop into cohesive groups. The method starts with each individual metabolite and builds up, with complete linkage assuring maximum distance between groups. The absolute magnitude of metabolite levels measured in a liquid (urine) are not directly comparable to those measured in a non-liquid (meconium), but correlation between the levels may be computed to gauge the transfer of exposure from maternal to the fetal systems. Therefore, we used Spearman correlations to assess relationship between the same metabolites measured in urine and meconium and computed these correlations stratified by study site, child sex, and child year of birth. We calculated intraclass correlation coefficients (ICC) for all urine metabolites using random effects regression models to evaluate the reproducibility of phthalate metabolites over pregnancy. Since metabolite distributions were positively skewed, we used natural-log transformation to better approximate the normal distribution in these models. Finally, we used the Kruskal-Wallis test to compare the distribution of the meconium measured metabolites by time to meconium collection.

3. Results

EARLI study sample recruitment was almost equally distributed between the four study sites. As reported in Table 2, most (73%) children were born in 2011 and 2012. Mothers were on average 33.8 (4.6) years old, mostly White, well educated, and had relatively high household income. Although the study sample consisted of 236 mothers, the number or biosamples available for analysis depended on women who completed study specific visits: resulting in different sample sizes (T2 urine, N=183, T3 urine, N=140, meconium N=190) available for analysis. A detailed comparison of sociodemographic and metabolite levels between the full study sample and different analysis samples is provided in Supplementary Table 3. For both T2 and T3 urine samples, 11 of 14 metabolites were detected in over 90% of samples, and 8 of 13 were detected in over 90% of meconium samples (Supplementary Table 2). Of metabolites measured in both biosamples, only MEHP and MiNP were not detected at or above 90% in both samples. Among all urine metabolites with values below LOD, only less than 5% of values were replaced by $\frac{LOD}{\sqrt{2}}$, but among meconium metabolites used in this article, ~ 20% of values for MEHP and MiNP were replaced by $\frac{LOD}{\sqrt{2}}$.

Table 2.

Demographic characteristics, N=236

Variable	N (%)
EARLI Study site
Drexel	61 (25.8)
Johns Hopkins	53 (22.5)
Kaiser Permanente	72 (30.5)
UC Davis	50 (21.2)
Child birthyear
2009	6 (2.5)
2010	53 (22.5)
2011	89 (37.7)
2012	84 (35.6)
2013	2 (0.9)
missing	2 (0.8)
Child Sex- Male	121 (51.1)
Maternal Age (mean, SD)	33.8 (4.6)
Pre-Pregnancy BMI
Underweight	3 (1.3)
Normal	97 (40.9)
Overweight	58 (24.5)
Obese	70 (29.5)
Missing	8 (2.4)
Race/ethnicity
NH White	128 (54.2)
NH Black	27 (11.4)
NH Asian	33 (13.9)
NH Other	8 (3.4)
Hispanic	38 (16.1)
Missing	2 (0.9)
Maternal Education
High School or less	29 (12.3)
Some college/ Tech School	67 (28.4)
Bachelors or more	137 (58.1)
Missing	3 (1.3)
Annual Household Income
<49 K	61 (25.9)
50– < 75 K	34 (14.4)
75– <100 K	37 (15.7)
100– <200 K	78 (33.1)
> 200 K	17 (7.2)
Missing	9 (3.8)
Parity: 2 +	130 (55.1)

EARLI, Early Autism Longitudinal Risk Investigation; NH, Non-Hispanic; BMI, Body Mass Index

Table 3 summarizes the median and selected percentiles of urine and meconium phthalate metabolites. Except for mono-carboxyisooctyl phthalate (MCOP), whose distribution shifts considerably lower from T2 to T3 (not statistically significant, Wilcoxon signed-rank test p-value 0.47), all other urine metabolite distributions are similar at both timepoints. The relatively hydrophilic primary metabolites of LMWP (MEP, MBP, MiBP) dominated among all metabolite distributions in both T2 and T3 urine samples except for high levels of MCOP, the secondary metabolite of di-isononyl phthalate (DiNP). The ICC ranged from a low of 0.15 for MiNP, between 0.2 to 0.3 for the DEHP metabolites, between 0.4 to 0.5 for DiBP and DnBP metabolites, to a high of 0.63 and 0.65 for MEP and MBzP, respectively. To compare exposure levels here to the US population and other pregnancy cohorts, Supplementary figures 1 and 2 display the creatinine adjusted GM (95% CI) for metabolites from EARLI; the HOME study, a Cincinnati based pregnancy cohort that enrolled participants between 2003 to 2006 (Percy et al., 2016); and the 2011–12 NHANES female sample. For some metabolites with data from all three sources available, NHANES GM-CI overlap considerably with EARLI study intervals. HOME study estimates for all metabolites except MiBP are higher than EARLI or NHANES. The estimates of mono-isobutyl phthalate (MiBP), MEHP, MEOHP, MECPP, MBP and mono-carboxyisononyl phthalate (MCNP) were higher for EARLI T2 measures, but lower than NHANES levels for T3- MEHP, MECPP and MCNP. EARLI study had the lowest values of MEP and is included in the plot with a separate axis due to its high value and variability in other samples. The Table 3 also reports the results for metabolites measured in meconium. Similar to maternal urine, both hydrophilic and hydrophobic metabolites were detected in meconium, and the metabolites of MiBP, MEP and MBP dominated among all detected metabolite distributions. The levels of the secondary metabolites of DEHP followed the same pattern (MECPP > MEHHP > MEOHP) in meconium as well. The Supplementary figure 3 compares the GM (95% CI) of EARLI with the P4 study. The GMs are in general much lower than results from the P4 study. Although all detected metabolites in EARLI were higher than P4, the only secondary metabolites available in common (MEOHP, MEHHP) had similar variance and were closer to P4 values. Primary metabolites in EARLI, however, were much higher and more variable.

Table 3.

Median and selected percentiles of metabolites in urine (LOD and creatinine adjusted), by trimester (ng/mg), and in meconium (ng/g)

Metabolite	Minimum	25th %	Median	75th %	Maximum
URINE
MBP	1.17	9.64	15.25	24.00	112.83
	1.84	10.07	15.42	24.03	217.61
MBzP	0.32	4.18	6.67	14.37	120.76
	0.47	3.41	6.87	15.81	183.35
MCNP	0.79	2.29	3.80	7.80	522.78
	0.50	2.14	3.52	6.65	131.25
MCOP	2.23	11.25	26.77	71.09	1508.76
	2.10	9.87	18.21	56.05	512.36
MCPP	0.26	1.39	2.48	5.22	844.20
	0.25	1.20	2.35	6.00	1406.24
MEHP	0.10	1.41	2.88	4.83	378.48
	0.13	1.22	2.28	4.77	75.40
MECPP	3.56	11.92	18.75	31.43	1563.42
	3.15	10.16	16.54	30.79	674.89
MEHHP	1.49	7.17	11.84	19.06	1227.14
	0.68	6.40	11.08	17.89	626.48
MEOHP	1.09	6.15	8.86	14.22	908.86
	0.79	5.52	8.95	14.65	468.93
MEP	3.92	18.52	33.46	70.54	17248.55
	2.47	17.44	34.12	68.95	2279.36
MHBP	0.16	0.76	1.23	2.05	11.34
	0.20	0.81	1.23	2.27	20.67
MHiBP	0.32	2.26	3.60	5.30	85.80
	1.03	2.31	3.42	5.35	163.80
MiBP	0.96	6.66	9.91	14.85	162.06
	2.34	6.57	10.41	15.89	443.47
MiNP	0.13	0.72	1.30	2.99	235.52
	0.11	0.59	1.06	2.97	106.78
MECONIUM
MBP	0.0004	12.93	21.21	31.45	235.22
MBzP	1.41	7.54	10.88	17.66	334.02
MCHP	0.003	0.004	0.53	2.57	40.54
MECPP	0.06	10.90	16.58	23.68	266.62
MEHHP	0.0004	1.49	2.54	3.81	57.19
MEHP	0.0004	5.66	10.92	21.01	458.94
MEOHP	0.0004	0.77	1.26	2.25	20.47
MEP	0.0004	2.66	33.84	233.98	14566.38
MiBP	0.15	23.25	39.31	78.40	2404.00
MiNP	0.0004	0.29	2.01	7.93	844.57
MOP	0.0004	0.0004	0.0004	2.62	392.55
MPP	0.0004	0.35	1.18	3.38	229.42
MiPP	3.85	55.84	128.72	322.46	7275.03

MBP, mono-n-butyl phthalate; MBzP, mono-benzyl phthalate; MCNP, mono-carboxynonyl phthalate; MCOP, mono-carboxyoctyl phthalate; MCPP, mono-3-carboxypropyl phthalate; MEHHP, mono(2-ethyl-5-hydroxyhexyl) phthalate; MECPP, mono(2-ethyl-5-carboxypentyl) phthalate; MeHP, mono(2-ethylhexyl) phthalate; MEOHP, mono(2-ethyl-5-oxohexyl) phthalate; MEP, mono-ethyl phthalate; MHBP, mono-3-hydroxy-n-butyl phthalate; MiBP, mono-isobutyl phthalate; MHiBP, mono-2-hydroxy-isobutyl phthalate; MiNP, mono-isononyl phthalate; MCHP, mono-cyclohexyl phthalate; MiPP, mono-isopentyl phthalate; MOP, mono-n-octyl phthalate; MPP, mono-n-pentyl phthalate

3.1. Correlation between metabolites measured in urine

Figure 2 is a heatmap of correlations between different phthalate metabolites measured in urine samples. Since there was low variation over time in most metabolites over pregnancy, the graph was created for correlations in the average of levels in a subject’s available urine samples. There were high correlations (0.73 to 0.96) between DEHP metabolites and moderate to high correlations (0.29 to 0.85) between DnBP, DiBP, and BBzP metabolites. Cluster analysis conducted on the 9 urine metabolites showed expected clustering of DEHP metabolites, a DnBP, and DiBP metabolite cluster. MBzP, MiNP, and MEP did not cluster well with other metabolites.

Fig. 2.

Heat map of Spearman correlations between phthalate metabolites measured in urine, averaged over trimester

3.2. Correlation between metabolites measured in meconium

Figure 3 is the heatmap of correlations between metabolites measured in meconium. Correlations between DEHP metabolites ranged from 0.09 for MEHP and MEOHP, 0.16 for MEHP and MEHHP, to a high of 0.31 for correlation between MEOHP and MEHHP. The pattern was somewhat similar to the P4 study, where correlation between MEHP and MEOHP was 0.25, MEHP and MEHHP was 0.21, and MEHHP with MEOHP was 0.94. Only primary metabolites of DiBP and DnBP were detected in EARLI meconium and these exhibited moderate correlation of 0.39. Moderately high correlation was also observed between MEHP and MBzP metabolites measured in meconium, whereas no correlation was detected between these in urine samples. Replicating the cluster analysis on the 9 metabolites showed a cluster of secondary metabolites of DEHP, and a cluster of the primary metabolites of DiBP and DnBP with MiNP, but the primary metabolite of DEHP clustered with MBzP. As with urine samples, MEP did not group with any other metabolite. Interestingly, correlations between metabolites in meconium differed by gestational age of birth. Although levels of phthalate metabolites did not differ by gestational age, the correlation between DEHP metabolites were much closer to that in urine (MEOHP: MEHHP- 0.73, MECPP: MEHHP-0.61, MEOHP: MECPP-0.49) among the 23 children born preterm, than term births (Supplementary Figures 4-6).

Fig. 3.

Heat map of Spearman correlations between phthalate metabolites measured in meconium

3.3. Correlations between same metabolites in urine at T2 and T3

Spearman correlations between the same metabolites at T2 and T3 urine (N=95) differed by metabolite and ranged from a low of 0.19 (MiNP), 0.32–0.39 for the DEHP metabolites, to 0.64 for MBzP and a high of 0.72 for MEP.

3.4. Correlations between same metabolites in urine with those in meconium

Correlations between same metabolites in urine with meconium were computed separately for urine metabolites measured at T2, T3, and trimester averaged measures. Metabolites measured in meconium were not highly correlated with urine measured metabolites regardless of whether trimester-specific or average measure was used. Strength of correlations varied from a low of −0.18 (−0.33, −0.02) for MEOHP at T2 to 0.19 (0.001, 0.36), and 0.20 (0.008, 0.37) for MiBP and MEHHP measured at T3. As displayed in Figure 4, all other 95% CI crossed zero. We also computed correlations in subgroups created by year of childbirth, study site, and child’s sex (Supplementary Figures 7-9). There were no discernable patterns or differences in the study site stratified correlations, but in birth year stratified analysis, low to moderate positive statistically significant correlations were detected between urine and meconium measured MBP, MECPP, MEHHP, MEP and MiBP metabolites among those samples for children born in 2012. There were no differences in correlations by child sex, except for a low positive correlation 0.26 (0.05, 0.44) for MEHHP measured in urine and meconium from the female babies.

Fig. 4.

Spearman correlations and 95% CI between same phthalate metabolites measured in urine and meconium, by trimester. Note: T2, Trimester 2; T3, Trimester 3; Tav, Averaged over 2^nd and 3^rd trimester

3.5. Ratio of DEHP metabolites

Analyzing the ratio of secondary to primary (MEHHP, MEOHP, MECPP / MEHP) and among the secondary (MEHHP / MEOHP) metabolites may help understand the differences in metabolism of DEHP among this autism enriched-risk sample. The median of the individual ratio of MEHHP/MEOHP, MECPP/MEOHP, and MECPP/MEHP in T2 urine was 1.3, 2.0, and 7.3, respectively. In meconium, these ratios were 2.0, 12.2, and 1.5, respectively.

4. Discussion:

The utility of meconium to quantify prenatal phthalate exposure in epidemiological studies is currently under-investigated. In our sample we detected both primary metabolites of LMWP like DEP, and primary and secondary metabolites of HMWP like DEHP (MEHP, MEOHP, MEHHP, MECPP), which are at opposite ends of the polarity axis. There are few published studies of phthalate measurement in meconium to which we can compare our results. The analytical chemistry study that developed the method to quantify phthalates in meconium used samples from five babies born during 2004–05 in Atlanta, GA, USA. They reported GM(95% CI) for MECPP, MEOHP and MEHHP that were 17.3 (2.9, 100.6), 1.9 (0.4, 9.2), and 2.3 (0.5, 10.7) ng/g respectively(Kato et al., 2006) – results similar to those observed here, however, other epidemiologic samples that have published on phthalates in meconium have yielded disparate findings. The P4 study conducted in Ottawa, Canada, reported levels several times lower in magnitude than our sample(Arbuckle et al., 2016). For example, GM (95% CI) for MEHHP and MEOHP were 0.41 (0.32, 0.52), and 0.12(0.08, 0.20) ng/g, respectively. These are the only other published data available from North America. However, there have been three published investigations from China reporting on meconium measurement of MBP and MEHP. In a sample of infants born at the Shanghai Medical center for Maternal and Child Health (SMC) between 2005–06, Zhang et al reported MBP and MEHP of 1700 (1200, 2400), and 2900 (1800, 4400) $\frac{\mu g}{g}$ respectively(Zhang et al., 2009). MBP and MEHP measured in infants born at SMC in 2011 were much lower at 101.7 (80.7, 171.9) and 163.8 (90.9, 276.2) $\frac{\mu g}{g}$ respectively(Xie et al., 2015). However, a third study among infants born at SMC in 2011 reported median meconium MEHP of approximately 3800 $\frac{\mu g}{g}$ (Li et al., 2013). All three studies used the same analytical lab to analyze samples. This suggests that although there is an expected decrease in phthalate exposure over time, there is substantial variation in exposure within city, or, perhaps, a lack of reliability in measurement approaches. Variability in reported levels across North America and China is also notable; however, since concentration of phthalates in the built environment and in food packaging material probably varies across countries and by economic development levels, these differences are perhaps less surprising. In summary, even with high variability of measured metabolites, since both polar and non-polar metabolites, primary and secondary metabolites with varying elimination half-lives were detected in meconium by both EARLI and P4 studies, meconium is a useful matrix to measure prenatal phthalate exposure.

The levels of some urine phthalate metabolites in our study were similar to the phthalate exposure levels of the female US population, but most estimates of exposure from the HOME sample were higher than our sample. This is not unexpected since the HOME cohort enrolled at least 3 years prior to EARLI, and NHANES data have been used to show temporal decreases in certain phthalates (mono-n-butyl phthalate (MBP), monobenzyl phthalate (MBzP), DEHP metabolites), but increases on others(Zota et al., 2014). The gestational phthalate exposure profile of urine measured metabolites were similar to other enriched autism risk birth cohorts(Shin et al., 2018) as well as other pregnancy cohorts(Factor-Litvak et al., 2014; Percy et al., 2016) in the US. In all cases MEP had highest concentrations, followed by MBP, MiBP and MBzP. The ICC of metabolites between T2 and T3 were similar in magnitude and pattern to the enriched autism risk cohort mentioned above(Shin et al., 2019) suggesting that exposure from personal care sources (DEP, DnBP, di-isobutyl phthalate (DiBP)) may be more reproducible than exposure from dietary sources (DEHP, DiNP) in this sample. The pattern in the levels of the secondary metabolites of DEHP in our sample was identical across all the above studies. Although we observed these same patterns among meconium measured DEHP secondary metabolites, the primary metabolite MEHP was higher than MEHHP and MEOHP in meconium. In addition, the levels of the LMWP and primary metabolites of the HMWP detected in meconium were similar to or in some cases higher in absolute magnitude than their counterparts in urine, which may indicate potential sample contamination- an issue that we explore in more detail below.

4.1. Correlation between metabolites in meconium

The strength of correlations observed between phthalate metabolites measured in our meconium samples was much weaker than that observed in urine samples from mothers of the same subjects. Currently, to our knowledge there are no other published reports including similar comparisons. Investigators from the P4 study have shared their unpublished data with us (Arbuckle, TE, unpublished results, personal communication, Dec 13, 2018), and analysis show generally similar differences in relative magnitude of correlations in meconium. However, in P4 data, correlation between the two secondary metabolites of DEHP (MEHHP and MEOHP) with the primary metabolite MEHP were each less than 0.25 while the correlation between secondary metabolites was as strong (0.94) as in maternal urine. In EARLI, correlation between the primary to secondary metabolites were similarly low, but correlation between MEHHP and MEOHP was only 0.32. The pattern of lower correlation between primary and secondary metabolites, but higher correlation between secondary ones, however, was similar between EARLI and P4. In EARLI, we saw a stronger positive correlation between MEHP and MBzP in meconium (0.47) than in urine (0.07) and this pattern was similar to P4, where urine correlation was 0.29 compared to 0.49 in meconium. Since both these are primary metabolites that could be elevated in meconium due to external contamination, the results should be considered with caution. The P4 study used pre-screened collection materials, phthalate free inserts in diapers to avoid external contamination, but EARLI was not specifically designed to avoid external phthalate contamination. In our study, we cannot rule out external contamination of meconium, but similarity in pattern with P4, where there was no correlation between meconium and infant urine samples suggests that contamination with child urine may not be an issue.

Since the expression of uridine glucuronyl transferase (UGT) is low in the fetus compared to adults, and nonexistent in the case of some isoforms(de Wildt et al., 1999), metabolites detected in fetal matrixes like meconium may be due to placental transfer. As reported in other studies assessing the proportions of types of DEHP metabolites, we also saw an elevated proportion of the carboxylated metabolite MECPP compared to MEHP in meconium. As the DEHP metabolite with longest delayed excretion time(Koch et al., 2006), MECPP is more likely to be present in the maternal system and thus is more likely to enter the fetal compartment. Due to differences in elimination half-lives of different phthalates, and possible metabolite differences in rate of placental transfer(Mose et al., 2007), the proportion of metabolites transferred may be low and their correlation patterns may be weaker than in the maternal compartment, as measured in maternal urine. Correlation between same metabolites measured in urine differed by metabolite, but the highest value was 0.7 (MEP). Therefore, it is possible that the correlation across matrixes may not exceed this value. However, this explanation does not account for differences that we detected in strength of correlations by gestational age. Since we did not observe any increase in absolute levels of measured metabolites among the preterm it is unlikely that an influx of phthalates during the preterm delivery caused higher correlations.

4.2. Correlations between same metabolites in urine with those in meconium.

Strength of correlations between any given metabolite measured in maternal urine with that same metabolite measured in meconium were very low in EARLI, with few statistically significant findings. We detected significant positive correlations for several metabolites assessed for children born in 2012. We did not change any of our protocols in sample collection or storage, and these results did not differ by study site. EARLI urine and meconium samples have been stored at below −70° C, temperature at which urine phthalate metabolites have been previously shown to be stable for long periods of time(Samandar et al., 2009). Thus, we consider these results to be due to chance than any systematic underlying process. The P4 study, the only published study to report these associations found significant correlations for MEHHP (0.35), MEOHP (0.35), and MEP (0.37). In EARLI, since MEP was highly variable and did not exhibit any correlations with other metabolites within urine as well as meconium, the lack of correlation between urine and meconium MEP was not surprising. The variability in urine phthalates was similar between EARLI and P4, but EARLI meconium phthalates exhibited higher variability than P4. This may have contributed to lower correlation in MEHHP, however, the lack of correlation between urine and meconium measured MEOHP was unexpected.

4.3. Ratio of DEHP metabolites

The ratio of DEHP metabolites have been previously used to evaluate differences in metabolism between adults and children(Enke et al., 2013). The ratio MEHHP/MEOHP in our sample was somewhat similar to reports in other studies using urine samples from pregnant women(Arbuckle et al., 2016; Enke et al., 2013), but the ratios MECPP/MEOHP and MECPP/MEHP were higher in EARLI than in Enke at al. In meconium samples we report only the ratio of the secondary metabolites of DEHP due to the potential for sample contamination affecting primary metabolites. The ratio MEHHP/MEOHP in our sample was lower than the 3.08 reported in P4, but similar to P4, the magnitude of the ratio was higher in meconium samples than in maternal urine. Our results of MECPP/MEOHP in meconium are also aligned with the high ratios observed in newborn’s first urine, lending support to the hypothesis that fetal metabolism is different from maternal metabolism, and the fetus may preferentially excrete HMWP as secondary (carboxylated) metabolites. Since the glucuronidation elimination pathway has been suggested to be compromised among some autistic children(Stein et al., 2013), the higher magnitude of the ratio (MECPP/MEOHP) in EARLI meconium as compared to newborn urine values in Enke et al. (12.2 vs. 6.4) is perhaps, not surprising. However, this observation highlights the importance of the need for additional research into differences in fetal metabolism of phthalates in these vulnerable populations.

5. Strengths and Limitations.

This study has several limitations. One concern in biosample measurement of phthalates is the possibility of external contamination due to the ubiquitous phthalate. In EARLI, sample collection procedures were not specifically designed to avoid phthalate contamination from the diaper or through collection process. Though we can avoid this issue in urine by measuring metabolites than parent diesters, the contamination problem is potentially more severe in meconium, since it may contain esterases capable of converting external contaminant phthalates to their primary metabolites, which would be indistinguishable from biologically created ones(Kato et al., 2006). As reported in Table 4, the ratio of meconium to urine phthalate metabolites is higher than 1 for all primary metabolites in EARLI as opposed to those in P4, whereas it is lower than 1 (and similar to P4) for secondary metabolites. This phenomenon, coupled with high relative variability of EARLI meconium phthalates, suggests the possibility of external contamination of meconium samples. Since only primary metabolites seem to be affected, and as these could be due to hydrolysis of the parent phthalate, we speculate that external contamination from the diaper may be responsible rather than infant urine. In both urine and meconium samples, we can avoid this issue in analysis stage by focusing on oxidative phthalate metabolites which can only be created in-vivo. In our study, the only secondary metabolites measured in meconium are derivatives of DEHP, and we detected expected correlation patterns between them, and, compared to other meconium metabolites, relatively robust correlations. In addition, correlations of the same secondary metabolites between matrixes do not display substantially different patterns than those for primary metabolites. Thus, we do not expect our conclusions to be altered by sample contamination.

Table 4.

Comparison of urine (ng/mg creatinine) and meconium (ng/g) phthalate metabolites in EARLI vs. P4

		EARLI	P4 study
Phthalate	T2 urine/Meconium	Median (25 %, 75 %)	Median (25 %, 75 %)
Primary Metabolites
	Urine	15.3 (9.6, 24)	19 (8.5, 37.5)
MBP	Meconium	21.2 (12.9, 31.5)	2.1 (0.9, 4.5)
	Urine	6.7 (4.2, 14.4)	8.5 (3.4, 23.5)
MBzP	Meconium	10.9 (7.5, 17.7)	0.36 (0.2, 1.04)
	Urine	2.9 (1.4, 4.8)	2.6 (1.2, 5.4)
MEHP	Meconium	10.9 (5.7, 21.0)	0.64 (0.4, 2.1)
	Urine	33.5 (18.5, 90.5)	27 (1.0, 75)
MEP	Meconium	33.8 (2.7, 233.9)	1.1 (0.68, 2.1)
	Urine	9.9 (6.7, 14.9)	6.7 (3.8, 11.9)
MiBP	Meconium	39.3 (23.3, 78.4)	Not detected
	Urine	1.3 (0.7, 3.0)	0.2 (0, 0.3)
MiNP	Meconium	2.0 (0.3, 7.9)	0 (0,0)
Secondary Metabolites
	Urine	18.8 (11.9, 31.4)	10.7 (7.1, 15.9)
MECPP	Meconium	16.6 (10.9, 23.5)	Not detected
	Urine	11.8 (7.2, 19.1)	13.1 (7.7, 21.5)
MEHHP	Meconium	2.5 (1.5, 3.8)	0.37 (0.21, 0.8)
	Urine	8.9 (6.2, 14.2)	8.3 (4.9, 13.2)
MEOHP	Meconium	1.26 (0.8, 2.3)	0.12 (0.1, 0.3)

EARLI, Early Autism Risk Longitudinal Investigation; P4, Plastics and Personal-Care Products use in Pregnancy; T2, trimester 2; MBP, mono-n-butyl phthalate; MBzP, mono-benzyl phthalate; MEHHP, mono(2-ethyl-5-hydroxyhexyl) phthalate; MECPP, mono(2-ethyl-5-carboxypentyl) phthalate; MEHP, mono(2-ethylhexyl) phthalate; MEOHP, mono(2-ethyl-5-oxohexyl) phthalate;; MEP, mono-ethyl phthalate; MiBP, mono-isobutyl phthalate; MiNP, mono-isononyl phthalate.

To evaluate meconium metabolite differences by sample collection at the hospital vs. home, we compared the distribution of the metabolites between samples collected within 12 hours of birth, > 12 hours since birth, and those who did not have adequate data to calculate the time to sample collection (Supplementary Table 4). Except for minor differences in the distribution of mono-n-pentyl phthalate, there were no statistically significant differences in the metabolite distributions, indicating that time to meconium collection did not affect metabolite levels. A related limitation is our inability to differentiate the meconium samples potentially contaminated with neonatal urine. We did not collect neonatal urine and thus are unable to estimate and adjust for the potential contamination. However, if in fact there was mixing of child urine and meconium in many samples, this would have led to an increase in the levels of both primary and secondary metabolites. In our comparison of metabolites with the P4 study (Table 4), we observed only slightly higher levels of the secondary metabolites as opposed to the substantial increase in primary metabolites. Therefore, although we are not able to eliminate the issue of contamination of meconium with child urine, we believe that it is unlikely to have occurred in a majority of our meconium samples.

Another limitation is that we only measured creatinine as opposed to specific gravity (SG) in urine samples, so we were constrained to creatinine adjustment for urine dilution. Several studies have reported that SG is superior to creatinine adjustment in populations undergoing physiological changes such as pregnancy, but advantages are dependent on variation in creatinine- over time, or over physiological characteristics of the individuals(MacPherson et al., 2018). In our data, there was low variation of creatinine by maternal BMI, and since creatinine adjusted phthalate metabolites are expected to be highly correlated with phthalate excretion rate based on urinary flow rate(MacPherson et al., 2018), we do not expect significant changes in results due to the procedure.

In maternal urine, the proportion of samples with detectable levels are similar to reports from other pregnancy cohort studies(Percy et al., 2016; Shin et al., 2018). Differences in magnitude of metabolites and distribution over pregnancy compared to other pregnancy cohorts may be explained due to changes in phthalate use over time(CDC, 2017), possible differences in patterns of personal care product use or diet between study populations. Similarly, differences in magnitude of metabolites between EARLI women and female NHANES samples may be explained by differences in age distributions. EARLI mothers had a median (IQR) age of 34 (22, 44), whereas the NHANES female data that we used for comparison is reported as an aggregate for all females over 6 years of age. Since personal care product use is a major driver of exposure, some of this disparity in exposure level could be due to age-specific differences in their use. Similar to other studies(Arbuckle et al., 2016; LaRocca et al., 2014; Meeker et al., 2009), we noted high correlation between DEHP metabolites and moderately high correlation between DnBP and DiBP metabolites. Results of our clustering procedure were similar to those from the P4 study except for inclusion of MBzP in the DnBP cluster (Arbuckle et al., 2016). Finally, since we quantified 14 metabolites in urine at two trimesters and 13 metabolites in meconium, we conducted a large number of tests and thus some of our results may be due to chance. However, the similarity of correlation patterns across metabolites and consistency of results with the P4 study suggests otherwise.

The main strength of this study is the use of a relatively large sample of women and children from a population at high risk for neurodevelopmental problems. We were able to measure a wide spectrum of metabolites in maternal urine at two timepoints in pregnancy and in meconium. We replicated the chemical analysis methods of the P4 study in our samples and detected similar phthalate metabolite correlation patterns within meconium.

5. Conclusions

Meconium is a useful matrix to measure prenatal phthalate exposure. Correlation patterns between metabolites of the same phthalate, or those with similar applications, seen in urine were also observed in meconium, but with significantly lower magnitude. These correlation patterns were stronger in meconium of babies born preterm. Correlations between same metabolites measured in urine and meconium were very low and did not change by trimester of urine collection. Phthalate metabolites measured in meconium may be good to assess fetal exposure, but due to their rapid clearance from the body, a single assessment in maternal urine from at a discrete point in pregnancy may not be expected to correlate well with fetal exposure. Future studies comparing meconium measured phthalates to urine measured metabolites measured in within- trimester pooled maternal urine samples may help to align the exposure window captured in meconium. Our expectation for future studies would be that correlation patterns observed in urine samples will be detectable in meconium as well, but with decreased magnitude. Any unexpected correlations between metabolites in meconium have to be evaluated in the light of possible external contamination.

Highlights.

• High and low molecular weight phthalate metabolites were detected in meconium.

• Meconium measured metabolites from the same parent phthalate were correlated.

• Low correlations observed between same metabolites in maternal urine & meconium.