Maternal urine phthalate metabolite exposure and miscarriage risk: a nested case–control study of the Zunyi Birth Cohort

Phthalates (PAEs) are widespread persistent organic pollutants and endocrine disruptors. However, the associations between PAE exposure and the risk of miscarriage in humans are unclear, and an insufficient number of studies have evaluated the possible threshold or dose-dependent effects of first trimester PAE exposure on miscarriage risk. Our research measured the levels of mono-methyl phthalate (MMP), mono-ethyl phthalate, mono-isobutyl phthalate, MiBP mono-butyl phthalate (MBP), mono-octyl phthalate, mono-benzyl phthalate, mono(2-ethylhexyl) phthalate, mono(2-ethyl-5-oxohexyl) phthalate, and mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP) in maternal urine collected in early gestation between 150 pregnancies ending in miscarriage and 150 pregnancies with live birth. We also estimated the odds ratios (ORs) and 95% confidence intervals (CIs) for miscarriage and each PAE as a continuous variable or quartile. A restrictive cubic splines was used to assess dose-dependent effects after controlling for maternal characteristics (e.g., age, educational level). we identified monotonically increasing dose-dependent effects of MEHHP and MMP on the risk of miscarriage. The largest effect estimates were approximately threefold higher for the highest MBP (OR = 2.57; 95% CI = 1.32–5.01) or MMP quartile (OR = 3.57; 95% CI = 1.82–7.00) and two-fold higher for the highest MEHHP quartile (OR = 2.12; 95% CI = 1.10–4.11). Our research preliminarily obtained possible thresholds of MBP, MEHHP, and MMP which were 18.07, 2.38, and 0.80 µg/g Cr for the risk of miscarriage, respectively. First-trimester exposure to MBP, MEHHP, and MMP exceeding certain thresholds increases the risk of miscarriage.


Introduction
Phthalates (PAEs) are plastic plasticizers widely used in the plastic industry in medicinal products, food packaging, personal care products, cosmetics, and painting materials, as well as coatings for oral medications (Qureshi et al. 2016). As environmental endocrine disruptors, PAEs have estrogenic activity, which can result in reproductive and developmental toxicity (Katsikantami et al., 2016). A review has reported that PAEs can induce alterations in puberty and fertility disorders in men and women (Hlisníková et al. 2020). Jiang et al. reported that PAEs are easily released into the environment because of their non-covalent interactions with substances, and they possibly impair ovarian function (Jiang et al. 2021a, b). An epidemiological study has reported the associations between PAE exposure with miscarriage, recurrent pregnancy loss, and missed miscarriage Radke et al. 2019;Zhang et al. 2020).
Substantial evidence links PAE exposure to miscarriage at the levels of physiology, animal experiments, and human epidemiology. Extravillous trophoblast (EVT) invasion is a key event in embryonic development (Jauniaux and Burton 2005). Gao Fumei et al. provided evidence that mono(2-ethylhexyl) phthalate (MEHP) can inhibit human EVT invasion via the PPARγ pathway, implicating female reproductive toxicity in the occurrence of pregnancy loss (Gao et al. 2017a). PAEs have been found to be associated with increases of inflammation biomarker levels (IL-6 and IL-10) and oxidative stress . Their adverse effects on endocrine function are likely to change the circulating levels of hormones, such as reductions of the production of estradiol and progesterone, which are responsible for maintaining pregnancy, and also lead to an imbalance between pro-oxidant and anti-oxidant levels, resulting in miscarriage (Jiang et al. 2021b). Animal studies observed that exposure to specific PAEs significantly decreases embryo survival, increases the incidence of resorption, reduces the number and size of litters, and increases the abortion rate in rats (Jukic et al. 2016;Schmidt et al. 2012). A large number of epidemiological studies also identified PAE metabolites that are strongly related to miscarriage based on measurements in the urine, plasma, and hair of pregnant women (Messerlian et al. 2016;Mu et al. 2015;Toft et al. 2012;Yi et al. 2016).
The aforementioned results were inconsistent for some metabolites, and current researches did not evaluate the possible threshold of PAE exposure that increases the risk of miscarriage. Most studies did not focus on first-trimester PAE exposure, and the findings were based on cross-sectional or case-control studies, which cannot prove cause and effect. We conducted a nested case-control study based on the Zunyi Birth Cohort using maternal urine samples obtained in the first trimester and evaluated the associations between prenatal exposure to nine PAEs and the risk of miscarriage. We evaluated the possible threshold and dosedependent effects of PAE exposure on miscarriage and verified the consistency of PAE exposure and miscarriage from multiple aspects to fill a gap in this underdeveloped field.

Research subjects
Pregnant women recruited in this study were obtained from the ongoing Zunyi Birth Cohort. After providing written informed consent, study participants completed an electronic questionnaire during face-to-face interviews. The study eligibility criteria included natural conception, singleton pregnancy, and a live fetus after birth. Pregnant women were included at 0-13 +6 weeks of gestation and excluded if they presented with serious chronic diseases and infectious diseases, such as cancer, chronic cardiovascular and cerebrovascular diseases, chronic renal failure, and HIV infection. The participants' urine was collected in the first trimester. Information on miscarriage and gestational age was obtained from each hospital's discharge registry. We excluded 26 GDM (Fig. S1) cases and finally established a nested case-control study with 150 cases ended in miscarriage from the hospital's birth registry and matched these pregnancies in a 1:1 ratio with pregnancies that ended in singleton live births (controls). The cases and controls were individually matched in a ratio on age (± 5 years). This study was ethically reviewed by the Affiliated Hospital of Zunyi Medical University (Batch No.: KLL-2019-006).

Measurement of PAE metabolites
Urine samples were analyzed to determine the concentrations of nine PAE metabolites by high-performance gas chromatography-tandem mass spectrometry (7010b, Agilent, Santa Clara, CA, USA). These metabolites were mono-methyl phthalate (MMP), mono-ethyl phthalate (MEP), mono-isobutyl phthalate (MiBP), mono-butyl phthalate (MBP), monooctyl phthalate (MOP), mono-benzyl phthalate (MBzP), MEHP, mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), and mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP). Spot urine specimens were collected in polypropylene tubes in the obstetrical clinics and transferred to the Zunyi Medical University Laboratory (Zunyi, Southwest China) for analysis. Urine samples were divided into 1-mL aliquots and stored at − 80 °C until analysis. The urinary creatinine concentration was measured by the laboratory of the Affiliated Hospital of Zunyi Medical University using an automatic urine analyzer (AU5800, Beckman, Brea, CA, USA). Details of the method for detecting the concentrations of metabolites in urine are presented in the Supplementary Information. Briefly, 1.5 mL of urine was added into a 10-mL-centrifuge tube. Then, 1 mL of sodium acetate solution buffer was added to each urine sample and standard sample, and 10 μL of internal standard solution and 10 μL of β-glucuronidase/sulfatase were added to fully hydrolyze the sample. Excess MgSO 4 ·7H 2 O was then added to reach solution supersaturation. Next, 1 mL of n-hexane + diethyl ether (4:1) was added to each tube. The tubes were then centrifuged, and the relevant supernatants were transferred to new 5-mL-centrifuge tubes. This step was repeated twice, and the supernatants were dried twice under a nitrogen stream. A silylation reagent [N, O-bis (trimethylsilane) trifluoroacetamide:trimethylchlorosilane = 99:1] was then added to the 5-mL-centrifuge tubes containing the supernatants. After complete derivatization, the samples were analyzed after allowing the tubes to cool to 37 °C.

Statistical analyses
The limit of detection (LOD) of PAE metabolites was assumed to correspond numerically to threefold the signalto-noise ratio. Notably, in our analyses, we replaced the PAE metabolite concentrations that were below the LOD with the relevant LOD divided by the square root of 2. The urine Crconcentration was used to correct for urine dilution using the following formula: urine Cr-corrected PAE metabolite concentration (μg/g Cr) = PAE metabolite concentration (μg /L)/ [urine Cr-concentration (μg/L) × 113 (g/mol) × 10 − 6153]. We calculated the Cr-adjusted urine PAE metabolite concentration distributions (geometric mean and percentile values). We used conditional logistic regression model to estimate the odds ratio (OR) and 95% CI for the relationship between PAE exposure and miscarriage. Next, we categorized PAE levels into quartiles according to the distribution among controls, and the lowest quartile served as the reference group to evaluate the linear dose-response relationship. The correlations of PAE exposure with miscarriage risk were also analyzed using the unmatched miscarriage population (n = 150) and control population (n = 298) in this cohort. We performed multiple sensitivity analyses to examine the robustness of the association between PAE exposure and miscarriage. First, we tested the homogeneity of the association among subpopulations by exploring the interaction between PAE metabolites and subgroup indicators (e.g., age). Moreover, we fitted a generalized additive model including cubic splines for the continuous PAE level (µg/g Cr) to evaluate potential non-linear relationships and visualize exposure-outcome responses.
To disentangle the possible effect of each PAE, we constructed adjusted pollutant models as follows: Model A was adjusted for ethnic, education, maternal profession, marital status, and maternal pre-pregnancy BMI; Model B was adjusted for all variables in Model A plus the smoking status of the pregnant woman and smoking status of the pregnant woman's spouse/partner; and Model C was adjusted for all variables in Model B plus parity status, history of miscarriage, and history of abortion.
Data analysis and graphing were performed using R (version 4.1.1; R Development Core Team) and IBM SPSS Statistics software version 17.0 (IBM, Chicago, IL, USA). The R package included "rms," "glmnet," and "ggplot2".

Results
The accuracy and precision of the study were evaluated by analyzing eight replicates of the quality control samples. The accuracy ranged from 80.45 to 112.59%, and the precision ranged from 3.15 to 8.37%. The LOD and limit of quantitation for PAEs are presented in Table S1. The subsequent analyses did not include MOP, MBzP, and MEOHP because their detection rates were lower than 50%. The medians and interquartile ranges of the nine maternal urine PAE concentrations in the case and control samples are also presented in Table S1. The calibration curves displayed good linearity (R 2 > 0.997). The recoveries of the nine PAE metabolites ranged from 80.45 to 112.59% (Table S1). These demonstrated that the study achieved satisfactory accuracy and precision. Table 1 describes the characteristics of pregnant women in the case and control groups. In the case group, 12.7% of pregnant women were older than 30 years. In addition, most women were of Han ethnicity (97.3%); most had a junior high school or lower education (47.3%); most were unemployed (45.3%); most were married (88.7%); and most were of normal weight (70.0%). In total, 2.7% of women in the case group smoked before pregnancy; 45.3% of their spouses/partners smoked; 59.3% were multiparous; 8.0% had a history of miscarriage; 82% had no history of abortion; 16.7% had 1-3, abortions; and 1.3% had more than three abortions. The details are presented in Table 1.
The concentration of each detected PAE metabolite was compared between cases and controls (Fig. 1). Wilcoxon's rank sum text illustrated that MEHHP (p = 0.020), MBP (p = 0.041), and MMP levels (p = 0.001) were significantly higher in the case group. The levels of the other metabolites did not significantly differ between the groups.
In the analyses of PAE quartiles, the results illustrated that higher MBP (p < 0.03), MEHHP (p = 0.040), and MMP quartiles (p < 0.01) were linked to significantly higher risks of miscarriage versus the reference. Specifically, the highest MBP quartile was linked to an approximately threefold higher risk of miscarriage (OR = 2.57; 95% CI = 1.32-5.01); the highest MEHHP quartile was associated with a twofold higher risk (OR = 2.12; 95% CI = 1.10-4.11); and the highest MMP quartile was linked to an approximately fourfold higher risk (OR = 3.57; 95% CI = 1.82-7.00). A monotonic increase of ORs according to the PAE quartile was also observed for MEHHP and MMP. Further adjustment using Model B slightly attenuated the association for the second and highest quartiles of MBP, all quartiles of MEHHP, and the third and fourth quartiles of MMP but strengthened the association for the fourth quartile of MBP. Adjustment using Model C attenuated the association for the second and fourth quartiles of MBP and all quartiles of MMP, whereas this adjustment slightly strengthened the association for the third quartile of MBP and the third and fourth quartiles of MEHHP. No clear linear response was observed for the other PAEs (Table 3; Fig. S2). We also performed the aforementioned analysis in the unmatched population. The estimated relationships between PAE quartiles and miscarriage risk were basically consistent with the aforementioned results excluding those for MEP, as the highest MEP quartile was linked to a 3.99-fold higher risk of miscarriage versus the reference (Table S3).
We examined whether the estimated association between PAE exposure and miscarriage differed among subpopulations (Fig. 2). The results illustrated that the estimated associations did not vary significantly among subgroups for most population characteristics and subgroups including maternal age, ethnicity, education, profession, and pre-pregnancy BMI.
In the PAE metabolite exposure spline model, the risk of miscarriage was higher for higher PAE concentrations. As presented in Fig. 2, the miscarriage risk in relation to MIBP increased starting at a concentration of approximately 5.80 µg/g Cr, and there was a sharper increase from 5.80 to approximately 25.0 µg/g Cr. The estimated OR also gradually increased for MEP starting at a concentration of 2.84 µg/g Cr, whereas the risk decreased at higher MEP concentrations (15.0 µg/g Cr). The risk continued to increase along with increasing MEP concentrations, and the threshold for a higher risk of miscarriage was 0.80 µg/g Cr. The risk sharply increased for MEHHP from a concentration of 2.38 µg/g Cr to approximately 15.0 µg/g Cr before decreasing slowly at higher concentrations.

Discussion
In this research, we explored the associations between urinary PAE levels and miscarriage risk in southwest China. The results revealed that PAE exposure was widespread  Li et al. 2018a, b;Polinski et al. 2018), and the inconsistency with other data may be partially attributable to differences in maternal characteristics, study design, sample sizes, and regions.  According to our findings, the concentrations of MBP, MEHHP, and MMP were significantly higher in cases than in controls. A similar study regarding the relationship between PAE exposure and missed miscarriage also found that the concentrations of these three substances were significantly higher in the case group than in the control group, whereas MEP and MIBP concentrations did not significantly differ between the groups (Yi et al. 2016). The lack of significant  Mu et al. (2015), but our unmatched (n = 449) study revealed a significant difference in MEP levels between the groups. Further research indicated that each one-unit increase in pregnant urinary MMP concentrations increased the risk of miscarriage by 90%, which exceeded the finding by He et al. that increases of the log-transformed urinary MMP concentration increased the risk of miscarriage by 49% (J. He et al. 2021). Our results remained stable after model correction using variables reflective of the condition of pregnant women, such as the history of miscarriage, parity, and the smoking status. These factors were reported in previous studies as confounding factors affecting the risk of miscarriage (Adde et al. 2021;Cai et al. 2019;Keramat et al. 2021;Qi et al. 2014;Ushie et al. 2018). The obtained results were more stable and reliable because of the adjustment of various levels and the sensitivity analysis.
The study results revealed that higher MMP, MBP, and MEHHP concentrations were linked to an increased risk of miscarriage in the matched cohort, whereas higher MEP concentrations were linked to an elevated risk of miscarriage in the unmatched cohort (Fig. 3). Previous researchers Fig. 2 Odds ratios (ORs) and 95% confidence intervals (CIs) for miscarriage, according to continuous phthalate concentrations with cubic splines stratified urinary PAE metabolite into quartiles, and the third (OR = 2.21; 95% CI = 1.06-4.60) and fourth quartiles of MMP (OR = 2.85; 95% CI = 1.34-6.05) were associated with a higher risk of missed abortion versus the lowest quartile (He et al. 2021). Messerlian et al. also found that for ΣDEHP (MEHP, MEHHP, and MEOHP), the RRs (95% CI) were 2.3 (0.63-8.5), 2.0 (0.58-7.2), and 3.4 (0.97-11.7) in Q2, Q3, and Q4, respectively, versus Q1 (Messerlian et al. 2016). Gao et al. reported that women with higher creatinine-normalized concentrations of MBP and MEP were at increased risk of pregnancy loss (Gao et al. 2017c). In prior research, MEHP was associated with a higher risk of pregnancy loss (OR = 2.9; 95% CI = 1.1-7.6), whereas an inverse association was unexpectedly identified between MEHP and clinical pregnancy loss (OR = 0.17; 95% CI = 0.03-0.95) (Toft et al. 2012). After adjustments for maternal age, education, parity, and gestational weeks at the time of urine collection, Gao Hui et al. observed that higher creatinine-normalized concentrations of MEP, MBP, MEOHP, and MEHHP were significantly associated with an increased risk of clinical pregnancy loss (Gao et al. 2017b (Mu et al. 2015;Qureshi et al. 2016). These findings were consistent with the results for MBP and MMP in matched population and MEP in the unmatched population. Simultaneously, evidence from rat and zebrafish models suggested that exposure to PAEs reduced the number and sizes of litters; decreased the odds of embryo survival or liveborn offspring; and increased the incidence of abortions, post-implantation loss, and intrauterine absorption (Chen et al. 2021;Li et al. 2018a, b). Although the current research results are not completely consistent, most of the findings confirmed that exposure of PAEs increased the risk of miscarriage. The inconsistent results might be attributable to PAE exposure estimation in different trimesters, as most of the aforementioned studies, excluding those of Gao et al., did not assess exposure levels in the first trimester. We also hypothesized that the discrepancy was partially attributable to differences in maternal characteristics, study design, sample size, and co-linearity among PAE metabolites. For instance, previous studies were conducted in three different countries (Denmark, USA, and China), and both cohort and case-control study designs were employed (Messerlian et al. 2016;Peng et al. 2016;Toft et al. 2012). Existing studies were mainly conducted in developed areas, and the exposure sources and levels could be dramatically different from the findings in pregnant women in rural or undeveloped areas. These findings should be verified by additional scientific and prospective large-sample studies.
Evidence suggested that PAEs affect the occurrence of miscarriage via multiple mechanisms. First of all, Several PAE metabolites have been reported to reduce the production of estradiol and progesterone in the ovaries through receptor-mediated signaling pathways, such as cAMP and peroxisome proliferator-activated receptor signaling to change the circulating levels of hormones responsible for maintaining pregnancy (Lovekamp-Swan  Treinen et al. 1990;Wang et al. 2021). Meanwhile, Lovekamp-Swan and Davis found that DEHPinduced decrease in aromatase mRNA and protein levels could reduce the conversion of testosterone to estradiol; low estradiol and progesterone levels are associated with miscarriage in humans (Lovekamp-Swan and Davis 2003). The blockade or absence of progesterone signaling also leads to reproductive problems including pregnancy failure (Sheikh 2016). It is well accepted that normal maternal levels of hormones, especially estrogen and progesterone, are important factors for maintaining pregnancy (San Lazaro Campillo et al. 2019). Therefore, the adverse effects of PAE metabolites on endocrine function are likely to change the circulating levels of hormones responsible for maintaining pregnancy, resulting in a less favorable uterine milieu for implantation and placentation and potentially leading to miscarriage. In addition, PAEs are associated with an increase in inflammation biomarker levels and oxidative stress, which activate uterine natural killer cells Ferguson et al. 2015a, b). There have been evidence that uterine natural killer cells regulate angiogenesis in the non-pregnant endometrium and play a role in implantation and early pregnancy (Larsen et al. 2013). Finally, another hypothesis suggested PAEs decreased the levels of angiogenic factors such as placental growth factor and soluble fms-like tyrosine kinase-1 and reduced placenta log interspersed nuclear element-1 methylation, which disrupted placental development and function during gestation (Ferguson et al. 2015a, b). Therefore, the aforementioned evidence biologically supports the hypothesis that some PAEs could have adverse effects on the risk of miscarriage.
Our study had a number of strengths. First, to our knowledge, no prior study evaluated the possible threshold of PAE exposure that increases the risk of miscarriage. Our research examined the possible thresholds and dose-dependent effects of PAE exposure on miscarriage. Second, we controlled for a wide range of potential confounders using three models, and sensitivity analysis results suggested that results were robust. Finally, this study focused on the association between first-trimester PAE exposure and miscarriage.
The present study also had some limitations. A previous study mentioned that urinary PAE measurements have low reproducibility and sensitivity throughout pregnancy (Fisher et al. 2015). Our study did not conduct repeated-measures analysis, and thus, the findings may not accurately reflect the clinical situation. In addition, this study did not detect blood PAE concentrations to confirm the results. Therefore, further studies with larger cohorts are needed to confirm these findings before final conclusions can be drawn.

Conclusion
To summarize, higher maternal urinary concentrations of three PAE metabolites in the first trimester, namely MBP, MMP, and MEHHP, are consistently associated with an increased risk of miscarriage. We also identified monotonically increasing dose-dependent effects of MEHHP and MMP concentrations on the risk of miscarriage. The effect estimates were approximately threefold higher risks for the highest MBP or MMP quartile and an approximately two-fold higher risk for the highest MEHHP quartile. Our study preliminarily obtained possible thresholds for the risk of miscarriages, and the corresponding concentrations of MBP, MEHHP, and MMP were approximately 18.07, 2.38, and 0.80 µg/g Cr, respectively.