The data used were collected form an ongoing longitudinal birth cohort study in the city of Ezhou, Hubei Province, China. Pregnant women were invited to participant in the study at their first antenatal examination (<12 weeks of gestational) in Ezhou Maternal and Child Health hospital from November 2018 to November 2019. The women were eligible for enrollment if the following criteria were met: (1) resident of Ezhou city, (2) willingness to complete questionnaires at first antenatal examination, (3) willing to urine samples and serum during regular prenatal care visit, and (4) willingness to give birth to their babies in the hospital where the study was being conducted.
A total of 568 pregnant women donated their urine and serum samples at their first antenatal examination (8.2±2.4 weeks), and 498 participants complete the survey questionnaire. Thirty-five pregnant women had participant in our previous intervention study (Wu et al., 2021), resulting in 463 pregnant women for analysis. This study was approved by the Ethics Committee of Hubei University of Chinese Medicine. Written informed consent was obtained from all participants.
We used a self-administered questionnaire, which was developed based on a previous questionnaire used in the US, Peru, and Democratic Republic of Congo and further adapted to the setting in China (Bai et al., 2016). It has also been demonstrated validation in our previous study (Wu et al., 2017). The questionnaire was divided into three sections: sociodemographic information, living conditions, and daily lifestyle. Socio-demographic information included age, education (college, high school, middle school or below), house income (<5000$, 5000-10000$, >10000$), first born status (yes, no), and occupation (unemployed, employed). Living environment included building type (reinforced concrete, brick-wood), distance from the road (<30, 30-100, >100 m), average living space (<20, 20-40, >40 m2), kitchen fuel (goal, natural gas, wood), decorating materials (bricks, wall paints, tiles), second-hand smoke (yes, no). Lifestyle included frequency of diary product consumption (<2, 2-5, >5 times/week), puffed food consumption (<2, 2-5, >5 times/week), take out (<2, 2-5, >5 times/week), eat fruits (<5, >5 times/week), eat vegetable (<5, >5 times/week), water source (tap water, well water, pond water), frequency of use plastic bag (<2, 2-5, >5 times/week), frequency of cosmetic use (<2, 2-5, >5 times/week), frequency of use electronic products (<2, 2-5, >5 hours/day), physical activity (<1, 1-2, >2 hours/day) and traffic type (bus, car, walk).
Phthalate metabolite measurements
The participant provided the first morning urine sample in 30-mL brown glass bottles. Samples were divided into aliquots and stored at -80°C until analysis was performed. These included the following: mono-isobutyl phthalate (MiBP) and mono-methyl phthalate (MMP) from DMP; mono-ethyl phthalate (MEP) from DEP; mono-n-butyl phthalate (MnBP) from di-n-butyl phthalate; mono-n-octyl phthalate (MOP) from DnOP; mono-benzyl phthalate (MBzP) from butyl benzyl phthalate, and mono (2-ethylhexyl) phthalate (MEHP) from DEHP. The analyses also involved three secondary oxidation metabolites of DEHP: namely mono-(2-ethyl-5-hydroxyhexyl)-phthalate (MEHHP), mono-(2-ethyl-5-oxohexyl) phthalate (MEOHP), and mono-(2-ethyl-5-carboxypentyl) phthalate (MECPP). Then, we used ΣDEHP represents the sum of the molar concentrations of MEHP, MEHHP, MEOHP, and MECPP.
The metabolites were analyzed in the School of Laboratory Medicine, Hubei University of Chinese Medicine using liquid chromatography-mass spectrometry (LC-MS/MS; Agilent, USA), according to the method for urine phthalates measurement described in Specht et al. (Specht et al., 2015). The detailed method was presented in our previous study (Wu et al., 2018). The calibration curve covered the range 0.100–200 ng/mL, and each batch of samples analyzed involved blank and quality controlsamples. The intra- and inter-day relative standard deviations were 11.7% and 13.2%, respectively.
The metabolite concentrations were corrected for the urine dilution using specific gravity (SG) as follows: Pc = P [(SGM–1)/(SG–1)], where Pc is the SG-adjusted urine concentration (ng/mL), P is the measured metabolite concentration, SG is the specific gravity of the urine sample, and SGM is the median SG of the samples for the studied population (Upson et al., 2013). The SG of each sample was measured using a handheld refractometer (PAL10-S; Atago, Tokyo, Japan) at room temperature. For PAE concentrations below the limit of detection (LOD), a value equal to was imputed.
Thyroid hormone measurement
Serum sample were collected at first antenatal examination. The samples were frozen at -20°C until shipped overnight on dry ice to the analytical laboratory. Blood plasma was analyzed for thyroid stimulating hormone (TSH), free triiodothyronine (FT3), free thyroxine (FT4), total triiodothyronine (TT3), total thyroxine (TT4) and thyroglobulin (TG) at Ezhou Maternal and Child Health hospital by automated chemiluminescence immunoassay (Roche, Germany). The intra-assay coefficients of variation (CV) for all hormones ranged from 3.2% (for FT3) to 9.8% (for TSH), and inter-assay CV were ranged from 2.4% (for TT3) to 7.4% (for FT4).
Descriptive statistics were used to describe our study population characteristic with the expression of a number (%) or mean±SD. Geometric means and selected percentiles were calculated to describe the distributions of urinary phthalate metabolites (and SG-adjusted) and thyroid hormones.
We fitted generalized estimating equation (GEE) models, using an identity link and exchangeable correlation, to explore the characteristic that predict phthalate exposure (Reeves et al., 2019). We initially fitted single predictor model for each phthalate biomarker concentration including all single covariate. Multivariable regression models were fitted considering all predictors with a p value less than 0.20 from the bivariate GEE models. We used a backward selection approach to select a final, parsimonious model for each phthalate biomarker where all variables significant at p <0.10 level were retained. We calculated predicted means and 95% confidence intervals (CI) for each variable based on the final parsimonious model, with all covariates held at their means. For categorical variables, we calculated the mean at each level of the variable, for continuous variables, we calculated means at the midpoint value of each quartile.
To make our results from these models including In-transformed continuous biomarkers and/or outcome more interpretable, we transformed regression coefficients to percent changes (and associated 95% confidence intervals) in hormones concentration in relation to the interquartile range (IQR) increase in urinary biomarker concentrations. Liner Mixed models was used for associations between phthalate exposure and thyroid hormones (Aker et al., 2018). Models included covariates that were significantly associated with one or more thyroid hormones as well as one or more urinary phthalate metabolites. All analyses were performed using R version 3.5.3 (http://www.r-project.org).