Longitudinal Associations between Ambient Air Pollution and Angiogenic Biomarkers among Pregnant Women in the LIFECODES Study, 2006-2008

Abstract

Background: Exposures to ambient air pollution during pregnancy have been linked to adverse pregnancy outcomes such as preeclampsia and fetal growth restriction. Although evidence has shown that women with preeclampsia have higher ratio of soluble fms-like tyrosine kinase 1 to placental growth factor (sFlt-1/PlGF ratio), the potential impact of air pollution on markers of placental growth and function has not been well studied. Objectives: We aimed to examine longitudinal associations between ambient air pollution exposure and angiogenic factors among pregnant women in LIFECODES, a prospective birth cohort and biorepository in Massachusetts in the United States. Methods: P1GF and sFlt-1 were measured among pregnant women using plasma samples collected around 10, 18, 26, and 35 wk' gestation. Women's exposures to ozone (O3), fine particulate matter with aerodynamic diameter <2.5 itm (PM2.5), and nitrogen dioxide (NO2) within 1, 2, 4, and 8 wk prior to each plasma sample collection were estimated based on geocoded residential addresses, and mixed effect linear regression models were fitted to assess their associations with sFlt-1/PlGF ratio, sFlt-1 (ng/mL), and P1GF (pg/mL). Percent changes in outcomes associated with each interquartile range increase in exposures were reported, along with their 95% confidence intervals. Results: A total of 1,066 pregnant women were included. In the multipollutant models, significant associations were observed for increased sFlt- 1/P1GF ratio (PM2.5 3-8 wk' gestation, NO2: 35-39 wk' gestation), elevated sFlt-1 (O3: 26-34 wk' gestation, PM2.5: 3-8 wk' gestation), decreased sFlt-1 (NO2: 4-8 wk' gestation), and decreased P1GF (NO2: 34-39 wk' gestation) after adjusting for sociodemographic status, smoking, drinking, body mass index, parity, history of chronic hypertension, and conception time. Discussion: Exposures to PM2.5 during early pregnancy and exposures to O3 and NO2 during late pregnancy were associated with increased sFlt-1/ P1GF ratio, elevated sFlt-1 and with decreased P1GF, which may be a potential mechanism underlying ambient air pollution's impacts on adverse pregnancy and birth outcomes.

Headnote

Background: Exposures to ambient air pollution during pregnancy have been linked to adverse pregnancy outcomes such as preeclampsia and fetal growth restriction. Although evidence has shown that women with preeclampsia have higher ratio of soluble fms-like tyrosine kinase 1 to placental growth factor (sFlt-1/PlGF ratio), the potential impact of air pollution on markers of placental growth and function has not been well studied.

Objectives: We aimed to examine longitudinal associations between ambient air pollution exposure and angiogenic factors among pregnant women in LIFECODES, a prospective birth cohort and biorepository in Massachusetts in the United States.

Methods: P1GF and sFlt-1 were measured among pregnant women using plasma samples collected around 10, 18, 26, and 35 wk' gestation. Women's exposures to ozone (O3), fine particulate matter with aerodynamic diameter <2.5 itm (PM2.5), and nitrogen dioxide (NO2) within 1, 2, 4, and 8 wk prior to each plasma sample collection were estimated based on geocoded residential addresses, and mixed effect linear regression models were fitted to assess their associations with sFlt-1/PlGF ratio, sFlt-1 (ng/mL), and P1GF (pg/mL). Percent changes in outcomes associated with each interquartile range increase in exposures were reported, along with their 95% confidence intervals.

Results: A total of 1,066 pregnant women were included. In the multipollutant models, significant associations were observed for increased sFlt- 1/P1GF ratio (PM2.5 3-8 wk' gestation, NO2: 35-39 wk' gestation), elevated sFlt-1 (O3: 26-34 wk' gestation, PM2.5: 3-8 wk' gestation), decreased sFlt-1 (NO2: 4-8 wk' gestation), and decreased P1GF (NO2: 34-39 wk' gestation) after adjusting for sociodemographic status, smoking, drinking, body mass index, parity, history of chronic hypertension, and conception time.

Discussion: Exposures to PM2.5 during early pregnancy and exposures to O3 and NO2 during late pregnancy were associated with increased sFlt-1/ P1GF ratio, elevated sFlt-1 and with decreased P1GF, which may be a potential mechanism underlying ambient air pollution's impacts on adverse pregnancy and birth outcomes. https://doi.org/10.1289/EHP11909

Introduction

Gestation is a crucial period that has substantial impacts on the health of both infants and mothers, influenced by a complex combination of genetic, social, and environmental factors.1 A growing body of evidence has linked ambient air pollution exposures during pregnancy with a number of adverse pregnancy and birth outcomes, such as preterm delivery, low birth weight, small for gestational age,2-6 gestational diabetes mellitus, and preeclampsia,7-11 even when the exposure levels are low.12'13 However, the underlying mechanisms of such associations are not well understood. Emerging studies have been focusing on the role of disordered angiogenesis, or the formation of new blood vessels, as one potential mechanism linking ambient air pollution and pregnancy outcomes like preeclampsia and later life risk of cardiovascular disease.4'6'14-17

As a leading cause of maternal and fetal morbidity and mortality,18 preeclampsia has been recognized as a risk factor of many adverse birth outcomes,19 as well as future cardiovascular diseases and diabetes mellitus among both mothers and their offspring.20-22 Preeclampsia is estimated to affect 5%-7% of pregnancies and is associated with a significant number of maternal and fetal deaths globally each year.23 Although the pathophysiological mechanisms of preeclampsia are not fully elucidated, alterations in the circulating maternal angiogenic profile have been implicated.24 Studies have shown that preeclampsia is characterized by increased anti-angiogenic proteins such as soluble fms-like tyrosine kinase 1 (sFlt-1) and concurrently decreased pro-angiogenic proteins such as placental growth factor (P1GF).25 P1GF is a family member of vascular endothelial growth factor (VEGF) produced mainly by placenta and is involved in the processes of embryogenesis and placentation.26 In an uncomplicated pregnancy, concentrations of P1GF stay low in the first trimester and increase with gestation, usually peaking at 26-30 wk and decreasing toward delivery. However, abnormally lower levels of P1GF are usually observed in women whose pregnancies are complicated by preeclampsia in comparison with women with uncomplicated pregnancy at similar gestational age.27-29 sFlt-1 is a splice variant of VEGF receptor 1, which inhibits the interactions between VEGF and its receptors. sFlt-1 increases toward the completion of pregnancy.30 Pregnancies with preeclampsia have significantly higher increases in circulating sFlt-1.31 Excess sFlt-1 is known to be associated with decreased levels of free P1GF in maternal circulation, causing angiogenic imbalance and creating an anti-angiogenic state mathematically expressed as sFlt-1/PlGF ratio.32-34 Previous studies have shown that an increased sFlt-1/PlGF ratio is associated with maternal endothelial dysfunction and usually occurs before the clinical sign of preeclampsia (e.g., hypertension and proteinuria).24 Although evidence for the predictive roles of sFlt-1 and P1GF is mixed, the sFlt-1/PlGF ratio has been proposed as a promising marker by many studies.35 In addition, an increasing number of studies examined dysregulated maternal angiogenic profile in relation to small for gestational age and fetal growth restriction, as well as diabetes mellitus and cardiovascular diseases in later life.36-40

Despite the well-established associations between pregnancy ambient air pollution exposures and preeclampsia,41^44 only two studies have assessed disordered angiogenesis as a potential underlying mechanism,45'46 and neither of them considered the rapid changes in angiogenic biomarker levels across pregnancy. In this study, we leveraged a prospective birth cohort and biore-pository with angiogenic biomarkers repeatedly measured four times across pregnancy to assess their longitudinal associations with three criteria ambient air pollutants that have widely been reported to be associated with preeclampsia,47'48 including ozone (O3), fine particulate matter (PM) with aerodynamic diameter <2.5 urn (PM2.5), and nitrogen dioxide (NO2).

Methods

Study Population

Women ages 18 y or older who had prenatal care before 15 wk' gestation and intended to give birth at Brigham and Women's Hospital (BWH) were eligible and enrolled in LIFECODES, an ongoing prospective birth cohort and repository based at BWH in Boston, Massachusetts. BWH is a tertiary urban teaching hospital of Harvard Medical School, with an ethnically and economically diverse catchment population. Written informed consent was obtained from each participant, who then completed questionnaires on demographic and medical history at baseline. Nonfasting samples of blood and urine were collected at baseline and three subsequent study visits, along with clinically relevant pregnancy characteristics. More details of the LIFECODES study can be accessed elsewhere.49 Participants in this study were a subset of women enrolled between 2006 and 2008 in LIFECODES who had angiogenic biomarkers measured (n = 1,602). We focused on those with a singleton pregnancy (n= 1,480) and further excluded those with unavailable angiogenic biomarkers at all four visits (n = 293; 19.8%) or unsuccessfully geocoded addresses (n= 121; 8.2%). A total of 1,066 (66.5%) gestations were included in this subset analysis, as shown in Figure S1. This study was approved by the institutional review board at the Mass General Brigham HealthCare System (2006P000395 and 2009P000810).

Assessment of Angiogenic Biomarkers

Concentrations of circulating P1GF and sFlt-1 were measured in plasma samples collected at four visits across pregnancy.49 The median [interquartile range (IQR)] gestation week for each of the four visits is 9.7 (8.4-11.4), 17.9 (17.0-18.7), 26.0 (25.0-26.9), and 35.1 (34.4-35.9), respectively. A histogram showing the number of participants for each corresponding week of gestation by study visits is presented in Figure S2. Among the 1,066 included participants, 815 (76.4%) pregnant women had angiogenic biomarkers measured at each of the four visits, 234 (22.0%) individuals had measures at three visits, 16 (1.5%) had measures at two visits, and 1 (0.1%) had only one measure from one visit. Both P1GF and sFlt-1 were measured with prototype ARCHITECT immunoassays (Abbott Laboratories). The P1GF immunoassay measures unbound P1GF (i.e., free form of P1GF-1), with a lower limit of detection (LLOD) of 1 pg/mL and an upper limit of quantification of (ULOQ) of 1,500 pg/mL. The sFlt-1 immunoassay measures total sFlt-1 (i.e., both free and bound sFlt-1), with a LLOD of O.lOng/mL and an ULOQ of 150ng/mL. In addition, sFlt-1/ P1GF ratio was also calculated, which is considered better than a single biomarker with improved prognostic value for preeclampsia in previous studies.50-52

Assessment of Ambient Air Pollution Exposure

Data on daily ambient O3 (ppb), PM2.5 (ug/m3) and NO2 (ppb) at 1 km spatial resolution were obtained from the National Aeronautics and Space Administration's (NASA) Socioeconomic Data and Applications Center (SEDAC, 03: https://doi.org/10.7927/a4mb-4t86; PM2.5: https://doi.org/10.7927/0rvr-4538; N02: https://doi.org/ 10.7927/f8eh-5864).53-55 An ensemble-based approach was developed to estimate daily concentrations of O3 (8 h maximum), PM2.5 (24 h average), and NO2 (1 h maximum) on 1-km grid cells for the contiguous United States by integrating multiple machine learning algorithms (e.g., neural network, random forest, and gradient boosting) and a wide range of predictors (e.g., satellite data, meteorological models, land use models, and chemical transport models).56-58 Overall model performance assessed by cross-validated i?-squared values for daily concentrations of O3, PM2.5, and NO2 were reported as 0.90, 0.86, and 0.79, respectively. These exposures were spatio-temporally linked to pregnant women based on their geocoded residential addresses collected before delivery. Average exposures were calculated within 1-, 2-, 4-, and 8-wk periods prior to each study visit. These exposure windows were selected in order to be comparable with previous studies and to explore potential acute effects of ambient air pollution at different time windows.45

Covariates

Questionnaires were administered at the enrollment, which collected information on potential covariates, including sociodemo-graphic factors (e.g., maternal age, race, ethnicity, educational attainment, and insurance status) and behavioral factors (e.g., cigarette smoking and alcohol use during pregnancy). Racial identities (i.e., Caucasian, Black, South Asian, East Asian, Native American/Pacific Islander, more than one race, and other) and ethnicity (i.e., Hispanic or not) were assessed by self-report. We categorized race and ethnicity into four groups: non-Hispanic White, non-Hispanic Black, Hispanic, and Others, a group which included South Asian, East Asian, Native American/Pacific Islander, more than one race, and other race or ethnicity, given the small sample size in each of these subgroups. Health history (e.g., history of chronic hypertension and nulliparity), prepregnancy body mass index [BMI (kg/m2), calculated by height and weight before pregnancy], and gestational age (confirmed by ultrasound scanning before 15 gestation wk) were abstracted from medical records. Other covariates include season and year of conception. Potential confounders were selected based on directed acyclic graphs (DAGs) presented in Figure S3. In addition, information on pregnancy complications and outcomes were also obtained from medical records accessed after birth, including gestational hypertension (i.e., systolic blood pressure >140 mmHg, or diastolic blood pressure >90 mmHg at second to fourth study visits with negative urine protein test), preeclampsia (i.e., gestational hypertension with urine protein >300 mg/24 h or protein/creati-nine >0.2), gestational diabetes mellitus [i.e., 2 or more abnormal values on the 3 h oral glucose tolerance test (OGTT)], and preterm delivery (i.e., <37 wk of gestation). Preeclampsia was further categorized into early onset (i.e., occurred prior to 34 wk' gestation) vs. late onset (i.e., occurred in or after 34 wk' gestation) and with vs. without severe features, based on documented gestational age at disease onset. Health information and clinical data were reviewed and validated by a panel of two Maternal-Fetal Medicine-certified physicians.

Statistical Analyses

Baseline characteristics of included participants and distributions of maternal plasma sFlt-1/PlGF ratio with concentrations of sFlt-1 and P1GF at each study visit were described. Exposures to O3,

PM2.5, and NO2 averaged during different exposure windows prior to each study visit were calculated. Analysis of variance (ANOVA) tests were used for comparisons, and p-values <0.05 were considered statistically significant. Levels of sFLT-1 below the LLOD (n = 5) were imputed as the LLOD (i.e., O.lng/mL) divided by the square root of 2.59 Maternal sFlt-1/PlGF ratio and concentrations of sFlt-1 and P1GF were log-transformed to account for their skewed distribution. To account for multiple study visits of each participant during pregnancy, linear mixed effect models were fitted, adjusting for maternal age (continuous), race or ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Others), education (high school or less, some college, and college or higher), smoking and alcohol use during pregnancy (yes/no), prepregnancy BMI (continuous), insurance (private insurance/HMO, self-paid or Medicaid/Mass Health, and none), nulliparity (yes/no), history of chronic hypertension (yes/no), and season and year of conception. Missing data were imputed using the chained equations method with the "mice" package in R,60 and a single data set was imputed, given minimal impact of the imputation procedure due to small fraction of missingness. To account for the nonlinear relationship between gestational age and angiogenic biomarkers, we included a cubic polynomial function of gestational age in the model, with the degree determined by the Akaike information criterion (AIC). In addition, to allow time-varying associations between air pollution and angiogenic biomarkers, we included interactions between air pollution exposures and gestational age in the model and obtained gestation week-specific associations. Multipollutant models were fitted by including all three air pollutants simultaneously to account for the potential confounding by coexposures. We reported results as percent changes in original untransformed levels of angiogenic factors (i.e., by multiplying coefficients for log-transformed outcomes by 100) for each IQR increase in exposures, along with their 95% confidence intervals (CIs). Pair-wise Pearson correlations were calculated for average exposures during different exposure windows at each study visit. We further conducted a number of sensitivity analyses to examine the robustness of our findings. To assess potential nonlinear exposure-response relationships, we used cubic polynomial functions of the exposures with the degree determined by the AIC in the fully adjusted multipollutant models. In addition, we fitted single-pollutant models and unadjusted multipollutant models to examine the potential impacts of adjusting for coexposures and potential confounders. To examine the potential impacts of pregnancy complications and preterm delivery, we further conducted two sets of analyses among women without complications (i.e., chronic hypertension, preeclampsia, gestational hypertension, and gestational diabetes mellitus) and those who delivered preterm, respectively. To examine the potential impacts of the seasonal differences in exposures on the associations, we conducted stratified analyses by season of conception [i.e., warm season (i.e., May-September) vs. cold season (i.e., October-April)]. Moreover, self-identified race or ethnicity was included as a potential confounder to account for a portion of social relations and differences in stress, though we also conducted sensitivity analyses excluding race and ethnicity as a covariate to examine whether such correction is biologically meaningful.61-64 All analyses were conducted in R (version 4.1.1; R Core Development Team).

Results

Table 1 shows the distributions of baseline characteristics among included women who were enrolled between 2006 and 2008 in the LIFECODES study. The 1,066 participants included in this study had a mean maternal age of 31.9 y (?5.6 y) and a mean prepregnancy BMI of 25.3 ( ?5.7). The present analysiscomprises 61.4% non-Hispanic White participants (n = 655), 13.9% non-Hispanic Black participants (n = 148), 13.1% Hispanic participants (ra = 140), and 11.5% participants of other races/ethnicities (n = 123). Most of the participants had college or higher education (ra = 727; 69.5%), had private insurance or were enrolled in the HMO (n = 818; 79.3%), and did not smoke (n = 1,012; 94.9%) nor use alcohol during pregnancy (n = 1,003; 95.9%). Nearly half of them were nulliparous (" = 459; 43.1%) and 5.3% of them had a history of chronic hypertension (n = 56). Prevalence of pregnancy complications and preterm delivery were also examined, including preeclampsia with different onset (early onset: n= 16; 1.5%, late onset: " = 67; 6.3%) and severity (with severe features: " = 20; 1.9%, without severe features: " = 66; 6.2%), gestational hypertension (n = 52; 4.9%), gestational diabetes mellitus (n = 82; 7.7%), and preterm delivery (n = 113; 10.6%). Participants included in this study had demographic characteristics similar to those who were excluded, and they were more likely to have complications such as with preeclampsia, gestational hypertension, and gestational diabetes (Table SI).

Table 2 presents the maternal plasma sFlt-1/PlGF ratio and concentrations of sFlt-1 and P1GF at each of the four visits across pregnancy. There were statistically significant differences in sFlt-1/PlGF ratio, sFlt-1, and P1GF between visits (p< 0.001). Specifically, the highest median sFlt-1/PlGF ratios were observed at the first study visit. Median levels of sFlt-1 peaked at the fourth visit (i.e., median gestation week of 35), whereas median concentrations of P1GF were observed to peak at the third visit (i.e., median gestation week of 26). The average exposure levels of O3, PM2.5, and NO2 during different exposure windows at each study visit are shown in Table 3. Significant differences in exposures between visits were observed, except for O3 exposures within 1-wk prior to each visit and NO2 exposures during all the exposure windows.

Figures 1, 2, and 3 show the longitudinal associations between air pollution exposures and angiogenic biomarkers across the entire pregnancy from the fully adjusted multipollutant models. Detailed effect size estimates and 95% CIs were shown in Excel Table SI.

Figure 1 presents the gestation week-specific associations of O3 with angiogenic biomarkers. Significant associations between O3 within 8 wk prior to visits and elevated sFlt-1 at 26-34 wk' gestation were observed. Marginally negative associations were also observed for sFlt-1 at 15-21 wk' gestation with O3 exposures within 1, 2, and 4 wk prior to visits. No statistically significant association was found between O3 and sFlt-1/PlGF ratio or P1GF.

Figure 2 shows the associations for PM2.5. PM2.5 exposures within 2 and 4 wk prior to visits were significantly associated with increased sFlt-1/PlGF ratio at 3-8 wk' gestation. In addition, exposures to PM2.5 within 1 , 2, and 4wk prior to visits were significantly associated with increased sFlt-1 at 3-8 wk' gestation. Nostatistically significant association was observed during the remainder of pregnancy or between PM2.5 and P1GF at any period.

Figure 3 shows the results for NO2 exposures. We found that NO2 exposures within 1 wk prior to visits were positively associated with sFlt-1/PlGF ratio at 35-39 wk' gestation, and negative associations were observed between NO2 exposures within 4 and 8 wk prior to visits and sFlt-1 at 3-8 wk' gestation. In addition, NO2 exposures within 1 wk prior to visits, but not other lags, were significantly associated with decreased P1GF at 34-39 wk' gestation.

Pair-wise Pearson correlations between average exposures during different exposure windows prior to each study visit are presented in Figure S4. Weak positive correlations were observed between O3 and PM2.5 and between PM2.5 and NO2. Weak to moderate negative correlations were observed between O3 and NO2.

Potential nonlinear exposure-response relationships were also examined. Figure S5 shows the exposure-response relationships by gestation week. Figures S6, S7, and S8 show the results at gestation weeks, with significant associations observed in previous models (Figures 1-3). Significant S-shaped nonlinear associations between PM2.5 within 2 wk prior to visits and sFlt-1/PlGF ratio and sFlt during early pregnancy (i.e., 3-9 wk' gestation) were observed. In addition, significant S-shaped exposure-response relationships were also observed between NO2 exposures within 1 wk prior to visits and sFlt-1/PlGF ratio and P1GF at 36-39 wk' gestation. No other significant associations were observed when potential nonlinear exposure-response relationships were considered.

Figure S9 shows results from the fully adjusted single-pollutant models. In comparison with the main findings from the multipollutant models, consistent results were observed, except that the associations between NO2 exposures within 1 wk prior to visits and sFlt-1/PlGF ratio and P1GF were only marginally significant.

Figure S10 shows the results from the unadjusted multipollutant models. In comparison with the main analyses, consistent results were observed, except for the marginally significant associations observed between O3 within 8 wk prior to visits and sFlt-1 at 26-34 wk' gestation.

Figure Sll shows the results from sensitivity analyses focusing on women without pregnancy complications (i.e., chronic hypertension, preeclampsia, gestational hypertension, or gestational diabetes mellitus), which are generally consistent with findings from the main analyses, except for the statistically significant positive associations between O3 within 8 wk prior to visits and sFlt-1/PlGF ratio at 26-33 wk' gestation, as well as between O3 within 4 wk prior to visits and sFlt-1 at 24-33 wk' gestation. Similar findings were observed in the sensitivity analyses focusing on women without preterm delivery, as shown in

Figure S12. No significant difference was found between PM2.5 or NO2 and the outcomes.

Figure S13 shows the results from the stratified analyses by season of conception (i.e., warm season vs. cold season). Among women with conception in cold seasons (i.e., October-April, re = 395), O3 and PM2.5 exposures within all exposure windows were significantly associated with higher sFlt-1/PlGF ratio and elevated sFlt-1 at 3-8 wk' gestation. O3 exposures within 8 wk prior to visits were also significantly associated with higher sFlt-1/P1GF ratio at 26-34 wk' gestation, and O3 exposures within all exposure windows were significantly associated with elevated sFlt-1 at 27-34 wk' gestation. Among women with conception in warm seasons (i.e., May-September, ra = 671), significant associations between O3 exposures within all exposure windows and higher sFlt-1/PlGF ratio at 36-39 wk' gestation, as well as between PM2.5 exposures within 2 and 4 wk prior to visits and higher sFlt-1/PlGF ratio and elevated sFlt-1 at 11-17 wk' gestation, were observed. In addition, exposures to O3 within 1 and 2 wk prior to visits were significantly associated with decreased P1GF at 33-39 wk' gestation. Significant negative associations between PM2.5 within 2 wk prior to visits as well as NO2 within 1, 2, and 4 wk prior to visits and P1GF during late pregnancy were also observed.

Figures S14 shows the results from sensitivity analyses without adjusting for race or ethnicity, and no significant difference from the main findings was observed.

Discussion

This study examined longitudinal associations between ambient air pollution (i.e., O3, PM2.5, and NO2) and angiogenic biomarkers by leveraging an ongoing prospective birth cohort and biorepository in Massachusetts. In summary, we found that ambient air pollution was associated with increased sFlt-1/PlGF ratio (i.e., for O3, 26-34 wk' gestation; for PM2.5: 3-8 wk' gestation; for NO2: 35-39 wk' gestation) and elevated sFlt-1 (i.e., O3: 26-34 wk' gestation, PM2.5: 3-8 wk' gestation), even among women with uncomplicated pregnancies. We also found negative associations between NO2 and P1GF, especially during late pregnancy (34-39 wk' gestation) in women with conception in warm seasons.

To our knowledge, only a few studies have been conducted to examine ambient air pollution and angiogenic biomarkers during pregnancy.45'46 One was a cross-sectional study in Brazil (n = 131), with O3 and NO2 exposures passively measured by personal monitors during the first trimester, that found that higher NO2 was associated with increased sFlt-1/PlGF ratio, with no associationobserved for O3. Another study examined exposures to fine PM with aerodynamic diameter <10 urn (PM10) and NO2 during the first and second trimester in The Netherlands with a large sample size (n = 7,801) and accounted for residential histories when performing spatiotemporal linkages. It found that higher PM10 and NO2 were associated with lower P1GF and sFlt-1 in the second trimester.45 Complementary to previous studies, the repeatedly measured plasma angiogenic biomarkers in our study enabled us to longitudinally examine their time-varying associations with air pollution exposures. It is important to note that the LIFES CODES study contains rich maternal information that allowed us to adjust for many potential confounders. In addition, modeled ambient air pollution data were generated by an ensemble-based approach incorporating multiple machine learners and predictor variables, with fine spatiotemporal resolutions and high model performance.56-58 We identified several critical pregnancy periods associated with higher sFlt-1/PlGF ratio and elevated sFlt-1, including O3 exposures in early (i.e., 3-8 wk' gestation among women with conception in cold seasons) and late pregnancy (i.e., 26-34 wk' gestation among women with conception in cold seasons, and 36-39 wk' gestation among women with conception in warm seasons), PM2.5 exposures in early pregnancy (i.e., 3-8 wk' gestation), and NO2 exposures in late pregnancy (i.e., 35-39 wk' gestation). Although the first missed menstrual period or even implantation can occur after 3 wk' gestation (except for pregnancies with in vitro fertilization), 3-8 wk' gestation is usually considered as the embryonic period. It has been suggested to be a susceptible window to many hazardous exposures in previous studies.65'66 These associations were also significant among women without pregnancy complications. Consistent with our findings, a recent animal study in rats also showed significantly increased maternal circulating sFlt-1 associated with O3 exposures in both early and late gestation.67

These findings suggest that ambient air pollution may potentially influence an anti-angiogenic state during early and late pregnancy. Increased sFlt-1 is known to provide alternative receptors of P1GF and VEGF, and it often comes with declines in P1GF that usually occur 5 wk before the onset of preeclampsia.68

In addition, studies indicated that pregnant women with intermediate angiogenic imbalance may be at higher risk of preterm delivery and recurrent spontaneous abortion, independent from preeclampsia69'70; however, the underlying mechanisms of this angiogenic imbalance among women with preeclampsia remains unclear. Previous in vitro studies observed such imbalance in trophoblast cell migration with the presence of reduced oxygen tension,71'72 and there is evidence showing that ambient air pollution may cause changes in oxygen saturation reflected by particulate-related pulmonary vascular inflammation.73'74 Our study associates ambient air pollution with angiogenic imbalance during early and late pregnancy, which may provide implications for future interventions in preeclampsia and other adverse maternal and birth outcomes. On the other hand, similar to observations in the previous study,45 our results also suggest air pollution may potentially promote a pro-angiogenic profile during the second trimester. The observed differential effects of air pollution on sFlt-1 corresponding to different stages of gestation may also reflect the nature of alterations in levels of angiogenic proteins in reaction to pregnancy.75'76 Our findings also indicate that potential underlying mechanisms may differ between air pollutants: We observed that O3 exposures have impacts on angiogenic biomarkers during both early and late pregnancy (depending on season of conception), whereas PM2.5 exposures mostly affect angiogenic biomarkers in early pregnancy, and NO2 exposures mainly have impacts in late pregnancy. These observations are consistent with previous studies identifying critical exposure windows for pregnancy complications such as preeclampsia, hypertensive disorders of pregnancy, and gestational diabetes mellitus.77-81 An interesting aspect is that findings from our study suggest seasonal variations in associations between ambient air pollution and angiogenic biomarkers. Early pregnancy may be a susceptible period of air pollution for women with conception in cold seasons, whereas late pregnancy may be a susceptible period for those with conception in warm seasons. The seasonal variations might be attributable to the compositional heterogeneities of ambient air pollutants in different seasons. For example, PM2.5 is a complex mixture with large spatiotemporal heterogeneities, with its chemical compositions varying significantly by seasons and regions and differentially associated with adverse pregnancy outcomes.79

This study has several strengths. First, we leveraged a large prospective birth cohort and biorepository with a diverse study population. More important, multiple longitudinally measured angiogenic biomarkers enabled us to examine time-varying associations across pregnancy. Several limitations need to be noted. First, potential exposure misclassifications exist because air pollution exposures were estimated based on women's residential addresses collected before delivery. Data on residential history or time-activity pattern may have improved the measurement but were not available.82 Second, ambient air pollution is a complex mixture with heterogeneous compositions that contribute to the risk differentially. Future studies are warranted to consider different chemical compositions and examine their associations with angiogenic biomarkers. Third, we do not have enough power to implement mediation analyses or distributed lag models.83 Future studies with larger sample sizes are needed to confirm our findings and to better examine and identify critical windows of exposure.

Conclusions Leveraging a large ongoing prospective birth cohort and biorepo-sitory, we found that exposures to air pollution were associated with higher sFlt-1/PlGF ratio and elevated sFlt-1 concentrations in early and late pregnancy. Findings from this study provide insights into the underlying mechanisms of preeclampsia or other adverse birth outcomes in relation to ambient air pollution. Acknowledgments The authors thank the participants and field staff at Brigham and Women's Hospital. This work was funded, in part, by the Intramural Research Program and the National Institute of Environmental Health Sciences (ZIAES103321). Initial funding for the recruitment of the birth cohort was provided by Abbott Diagnostics (9MZ-04-06N03). This work was also supported in part by the National Heart, Lung, and Blood Institute under award no. K01HL153797. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

References

References 1. Miranda ML, Maxson P, Edwards S. 2009. Environmental contributions to disparities in pregnancy outcomes. Epidemiol Rev 31:67-83, PMID: 19846592, https://doi.org/10.1093/epirev/mxp011.

2. Quraishi SM, Hazlehurst MF, Loftus CT, Nguyen RHN, Barrett ES, Kaufman JD, et al. 2022. Association of prenatal exposure to ambient air pollution with adverse birth outcomes and effect modification by socioeconomic factors. Environ Res 212(Pt E):113571, PMID: 35640705, https://doi.Org/10.1016/j.envres. 2022.113571.

3. Nyadanu SD, Dunne J, Tessema GA, Mullins B, Kumi-Boateng B, Lee Bell M, et al. 2022. Prenatal exposure to ambient air pollution and adverse birth outcomes: an umbrella review of 36 systematic reviews and meta-analyses. Environ Pollut 306:119465, PMID: 35569625, https://doi.Org/10.1016/j.envpol.2022. 119465.

4. Lavigne E, Yasseen AS, Stieb DM, Hystad P, van Donkelaar A, Martin RV, et al. 2016. Ambient air pollution and adverse birth outcomes: differences by maternal comorbidities. Environ Res 148:457-466, PMID: 27136671, https://doi.org/10. 1016/j.envres.2016.04.026.

5. Klepac P, Locatelli I, Korosec S, Kiinzli N, Kukec A. 2018. Ambient air pollution and pregnancy outcomes: a comprehensive review and identification of environmental public health challenges. Environ Res 167:144-159, PMID: 30014896, https://doi.Org/10.1016/j.envres.2018.07.008. 6. Zhu X, Liu Y, Chen Y, Yao C, Che Z, Cao J, et al. 2015. Maternal exposure to fine particulate matter (PM2.5) and pregnancy outcomes: a meta-analysis. Environ Sci Pollut Res Int 22(5):3383-3396, PMID: 25163563, https://doi.org/10.1007/ s11356-014-3458-7.

7. Zheng Y, Wen X, Bian J, Lipkind H, Hu H. 2021. Associations between the chemical composition of PM2.5 and gestational diabetes mellitus. Environ Res 198:110470, PMID: 33217440, https://doi.Org/10.1016/j.envres.2020.110470. 8. Nobles CJ, Williams A, Ouidir M, Sherman S, Mendola P. 2019. Differential effect of ambient air pollution exposure on risk of gestational hypertension and preeclampsia. Hypertension 74(2):384-390, PMID: 31230552, https://doi.org/10. 1161/HYPERTENSI0NAHA.119.12731.

9. Zhang H, Wang Q, He S, Wu K, Ren M, Dong H, et al. 2020. Ambient air pollution and gestational diabetes mellitus: a review of evidence from biological mechanisms to population epidemiology. Sci Total Environ 719:137349, PMID: 32114225, https://doi.Org/10.1016/j.scitotenv.2020.137349. 10. Mandakh Y, Rittner R, Flanagan E, Oudin A, Isaxon C, Familari M, et al. 2020. Maternal exposure to ambient air pollution and risk of preeclampsia: a population-based cohort study in Scania, Sweden. Int J Environ Res Public Health 17(5):1744, PMID: 32155988, https://doi.org/10.3390/ijerph17051744.

11. Malmqvist E, Jakobsson K, Tinnerberg H, Rignell-Hydbom A, Rylander L 2013. Gestational diabetes and preeclampsia in association with air pollution at levels below current air quality guidelines. Environ Health Perspect 121 (4):488 493, PMID: 23563048, https://doi.org/10.1289/ehp.1205736. 12. Melody S, Wills K, Knibbs LD, Ford J, Venn A, Johnston F, et al. 2020. Adverse birth outcomes in Victoria, Australia in association with maternal exposure to low levels of ambient air pollution. Environ Res 188:109784, PMID: 32574853, https://doi.Org/10.1016/j.envres.2020.109784.

13. Chen G, Guo Y, Abramson MJ, Williams G, Li S. 2018. Exposure to low concentrations of air pollutants and adverse birth outcomes in Brisbane, Australia,2003-2013. Sci Total Environ 622-623:721-726, PMID: 29223898, https://doi.org/ 10.1016/j.scitotenv.2017.12.050.

14. Vadillo-Ortega F, Osornio-Vargas A, Buxton MA, Sanchez BN, Rojas-Bracho L, Viveros-Alcaraz M, et al. 2014. Air pollution, inflammation and preterm birth: a potential mechanistic link. Med Hypotheses 82(2):219-224, PMID: 24382337, https://doi.Org/10.1016/j.mehy.2013.11.042.

15. Kulkarni AV, Mehendale SS, Yadav HR, Kilari AS, Taralekar VS, Joshi SR, et al. 2010. Circulating angiogenic factors and their association with birth outcomes in preeclampsia. Hypertens Res 33(6):561-567, PMID: 20224572, https://doi.org/ 10.1038/hr.2010.31.

16. Conroy AL, McDonald CR, Gamble JL, Olwoch P, Natureeba P, Cohan D, et al. 2017. Altered angiogenesis as a common mechanism underlying preterm birth, small for gestational age, and stillbirth in women living with HIV. Am J Obstet Gynecol 217(6):684.e1-684.e17, PMID: 29031892, https://doi.Org/10.1016/j.ajog. 2017.10.003.

17. Darling AM, McDonald CR, Conroy AL, Hayford KT, Liles WC, Wang M, et al. 2014. Angiogenic and inflammatory biomarkers in midpregnancy and small-for-gestational-age outcomes in Tanzania. Am J Obstet Gynecol 211(5):509.e1- 509.e8, PMID: 24881826, https://doi.Org/10.1016/j.ajog.2014.05.032.

18. World Health Organization. 2005. The World Health Report 2005: Make Every Mother and Child Count Geneva, Switzerland: World Health Organization.

19. Backes CH, Markham K, Moorehead P, Cordero L, Nankervis CA, Giannone PJ, et al. 2011. Maternal preeclampsia and neonatal outcomes. J Pregnancy 2011(2011):214365, PMID: 21547086, https://doi.org/10.1155/2011/214365. 20. Staff AC, Dechend R, Redman CWG. 2013. Review: preeclampsia, acute athero-sis of the spiral arteries and future cardiovascular disease: two new hypotheses. Placenta 34 Suppl:S73-S78, PMID: 23246096, https://doi.Org/10.1016/j. placenta.2012.11.022.

21. Mosca L, Benjamin EJ, Berra K, Bezanson JL, Dolor RJ, Lloyd-Jones DM, et al. 2011. Effectiveness-based guidelines for the prevention of cardiovascular disease in women-2011 update: a guideline from the American Heart Association. Circulation 123(11):1243 1262, PMID: 21325087, https://doi.org/ 10.1161/CIR.0b013e31820faaf8.

22. Weissgerber TL, Mudd LM. 2015. Preeclampsia and diabetes. Curr Diab Rep 15(3):9, PMID: 25644816, https://doi.org/10.1007/s11892-015-0579-4. 23. Rana S, Lemoine E, Granger JP, Karumanchi SA. 2019. Preeclampsia: pathophysiology, challenges, and perspectives. Circ Res 124(7):1094 1112, PMID: 30920918, https://doi.org/10.1161/CIRCRESAHA.118.313276.

24. Lam C, Lim K-H, Karumanchi SA. 2005. Circulating angiogenic factors in the pathogenesis and prediction of preeclampsia. Hypertension 46(5)1077-1085, PMID: 16230516, https://doi.Org/10.1161/01 .HYP.0000187899.34379.b0. 25. Rasmussen LG, Lykke JA, Staff AC. 2015. Angiogenic biomarkers in pregnancy: defining maternal and fetal health. Acta Obstet Gynecol Scand 94(8):820-832, PMID: 25753566, https://doi.org/10.1111/aogs.12629.

26. Khalil A, Muttukrishna S, Harrington K, Jauniaux E. 2008. Effect of antihypertensive therapy with alpha methyldopa on levels of angiogenic factors in pregnancies with hypertensive disorders. PLoS One 3(7):e2766, PMID: 18648513, https://doi.org/10.1371/journal.pone.0002766. 27. Chappell LC, Duckworth S, Seed PT, Griffin M, Myers J, Mackillop L, et al. 2013. Diagnostic accuracy of placental growth factor in women with suspected preeclampsia: a prospective multicenter study. Circulation 128(19):21212131, PMID: 24190934, https://doi.org/10.1161/CIRCULATIONAHA.113.003215.

28. Chau K, Hennessy A, Makris A. 2017. Placental growth factor and pre-eclampsia. J Hum Hypertens 31(12):782-786, PMID: 29115294, https://doi.org/10.1038/jhh. 2017.61. 29. Knudsen UB, Kronborg CS, von Dadelszen P, Kupfer K, Lee S-W, Vittinghus E, et al. 2012. A single rapid point-of-care placental growth factor determination as an aid in the diagnosis of preeclampsia. Pregnancy Hypertens 2(1 ):8 15, PMID: 26104984, https://doi.0rg/IOIOI6/j.preghy.2OII.O8.li7.

30. Lehnen H, Mosblech N, Reineke T, Puchooa A, Menke-Mollers I, Zechner U, et al. 2013. Prenatal clinical assessment of sFit-1 (soluble fms-like tyrosine kinase-1)/PIGF (placental growth factor) ratio as a diagnostic tool for preeclampsia, pregnancy-induced hypertension, and proteinuria. Geburtshilfe Frauenheilkd 73(5):440-445, PMID: 24771924, https://doi.org/10.1055/s-0032-1328601. 31. Palmer KR, Tong S, Kaitu'u-Lino TJ. 2017. Placental-specific sFLT-1: role in preeclamptic pathophysiology and its translational possibilities for clinical prediction and diagnosis. Mol Hum Reprod 23(2):69-78, PMID: 27986932, https://doi.org/10. 1093/molehr/gaw077.

32. Herraiz 1, Simon E, Gomez-Arriaga PI, Martinez-Moratalla JM, Garcia-Burguillo A, Jimenez EAL, et al. 2015. Angiogenesis-related biomarkers (sFlt-1/PLGF) in the prediction and diagnosis of placental dysfunction: an approach for clinical integration. Int J Mol Sci 16(8)19009-19026, PMID: 26287164, https://doi.org/10. 3390/ijms160819009. 33. Kim S-Y, Ryu H-M, Yang J-H, Kim M-Y, Han J-Y, Kim J-0, et al. 2007. Increased sFlt-1 to PIGF ratio in women who subsequently develop preeclampsia. J Korean Med Sci 22(5):873-877, PMID: 17982238, https://doi.Org/10.3346/jkms.2007.22.5.873.

34. De Vivo A, Baviera G, Giordano D, Todarello G, Corrado F, D'anna R, et al. 2008. Endoglin, PIGF and sFlt-1 as markers for predicting pre-eclampsia. Acta Obstet Gynecol Scand 87(8):837-842, PMID: 18607829, https://doi.org/ 10.1080/00016340802253759.

35. Agrawal S, Cerdeira AS, Redman C, Vatish M. 2018. Meta-analysis and systematic review to assess the role of soluble FMS-like tyrosine kinase-1 and placenta growth factor ratio in prediction of preeclampsia: the SaPPPhirE study. Hypertension 71(2):306-316, PMID: 29229743, https://doi.org/10.1161/ HYPERTENSIONAHA.117.10182.

36. Taylor RN, Grimwood J, Taylor RS, McMaster MT, Fisher SJ, North RA, et al. 2003. Longitudinal serum concentrations of placental growth factor: evidence for abnormal placental angiogenesis in pathologic pregnancies. Am J Obstet Gynecol 188(1 ):177-182, PMID: 12548214, https://doi.org/10.1067/mob.[phone].

37. Benton SJ, Hu Y, Xie F, Kupfer K, Lee S-W, Magee LA, et al. 2012. Can placental growth factor in maternal circulation identify fetuses with placental intrauterine growth restriction? Am J Obstet Gynecol 206(2):163.e1-7-163.e7, PMID: 22055338, https://doi.Org/10.1016/j.ajog.2011.09.019.

38. Lenderink T, Heeschen C, Fichtlscherer S, Dimmeler S, Hamm CW, Zeiher AM, et al. 2006. Elevated placental growth factor levels are associated with adverse outcomes at four-year follow-up in patients with acute coronary syndromes. J Am Coll Cardiol 47(2):307-311, PMID: 16412852, https://doi.Org/10.1016/j.jacc. 2005.08.063.

39. Staff AC, Dechend R, Pijnenborg R. 2010. Learning from the placenta: acute atherosis and vascular remodeling in preeclampsia-novel aspects for atherosclerosis and future cardiovascular health. Hypertension 56(6):1026-1034, PMID: 20956732, https://doi.org/10.1161/HYPERTENSIONAHA.110.157743.

40. Ak G, Buyukberber S, Sevinc A, Turk HM, Ates M, Sari R, et al. 2001. The relation between plasma endothelin-1 levels and metabolic control, risk factors, treatment modalities, and diabetic microangiopathy in patients with Type 2 diabetes mellitus. J Diabetes Complications 15(3):150 157, PMID: 11358684, https://doi.org/10.1016/s1056-8727(01)00137-4.

41. Wang Q, Zhang H, Liang Q, Knibbs LD, Ren M, Li C, et al. 2018. Effects of prenatal exposure to air pollution on preeclampsia in Shenzhen, China. Environ Pollut 237:18-27, PMID: 29466771, https://doi.Org/10.1016/j.envpol.2018.02.010. 42. Goin DE, Sudat S, Riddell C, Morello-Frosch R, Apte JS, Glymour MM, et al. 2021. Hyperlocalized measures of air pollution and preeclampsia in Oakland, California. Environ Sci Technol 55(21):14710-14719, PMID: 34648281, https://doi.org/10.1021/ acs.est.1c02151.

43. Weber KA, Yang W, Lyons E, Stevenson DK, Padula AM, Shaw GM. 2021. Greenspace, air pollution, neighborhood factors, and preeclampsia in a population-based case-control study in California. Int J Environ Res Public Health 18(10):5127, PMID: 34066190, https://doi.org/10.3390/ijerph18105127.

44. Zhao B, Wang M, Song W, Ma H, Meng H, Qi M, et al. 2022. Increasing risk of gestational diabetes and preeclampsia associated with long-term exposure effects of air pollution. Air Qual Atmos Health 15(6):917-928, https://doi.org/10. 1007/sl1869-022-01165-4.

45. van den Hooven EH, Pierik FH, de Kluizenaar Y, Hofman A, van Ratingen SW, Zandveld PYJ, et al. 2012. Air pollution exposure and markers of placental growth and function: the generation R study. Environ Health Perspect 120(12):1753-1759, PMID: 22922820, https://doi.org/10.1289/ehp.1204918.

46. Hettfleisch K, Carvalho MA, Hoshida MS, Pastro LDM, Saldiva SRDM, Vieira SE, et al. 2021. Individual exposure to urban air pollution and its correlation with placental angiogenic markers in the first trimester of pregnancy, in Sao Paulo, Brazil. Environ Sci Pollut Res Int 28(22):28658-28665, PMID: 33544347, https://doi.org/10.1007/s11356-021-12353-7.

47. Pedersen M, Stayner L, Slama R, Sorensen M, Figueras F, Nieuwenhuijsen MJ, et al. 2014. Ambient air pollution and pregnancy-induced hypertensive disorders: a systematic review and meta-analysis. Hypertension 64(3):494-500, PMID: 24935943, https://doi.org/10.1161/HYPERTENSIONAHA.114.03545. 48. Hu H, Ha S, Roth J, Kearney G, Talbott E0, Xu X, et al. 2014. Ambient air pollution and hypertensive disorders of pregnancy: a systematic review and metaanalysis. Atmos Environ (1994) 97:336-345, PMID: 25242883, https://doi.org/10. 1016/j.atmosenv.2014.08.027.

49. McElrath TF, Lim K-H, Pare E, Rich-Edwards J, Pucci D, Troisi R, et al. 2012. Longitudinal evaluation of predictive value for preeclampsia of circulating angiogenic factors through pregnancy. Am J Obstet Gynecol 207(5):407.e1-7-407.e7, PMID: 22981320, https://doi.Org/10.1016/j.ajog.2012.08.010. 50. Staff AC. 2011. Circulating predictive biomarkers in preeclampsia. Pregnancy Hypertens 1(1):28-42, PMID: 26104229, https://doi.Org/10.1016/j.preghy.2010.10.012. 51. Sunderji S, Gaziano E, Wothe D, Rogers LC, Sibai B, Karumanchi SA, et al. 2010. Automated assays for sVEGF R1 and PIGF as an aid in the diagnosis of preterm preeclampsia: a prospective clinical study. Am J Obstet Gynecol 202(1 ):40.e1-7^0.e7, PMID: 19762001, https://doi.Org/10.1016/j.ajog.2009.07.025.

52. Levine RJ, Lam C, Qian C, Yu KF, Maynard SE, Sachs BP, et al. 2006. Soluble endoglin and other circulating antiangiogenic factors in preeclampsia. N Engl J Med 355(10):992-1005, PMID: 16957146, https://doi.org/10.1056/NEJMoa055352.

53. Di Q, Wei Y, Shtein A, Hultquist C, Xing X, Amini H, et al. 2022. Daily and Annual N02 Concentrations for the Contiguous United States, 1-km Grids, v1 (2000-2016). Palisades, New York: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/f8eh-5864.

54. Di Q, Wei Y, Shtein A, Hultquist C, Xing X, Amini H, et al. 2021. Daily and Annual PM2.5 Concentrations for the Contiguous United States, 1-km Grids, v1 (2000-2016). Palisades, New York: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/0rvr-4538.

55. Requia WJ, Wei Y, Shtein A, Hultquist C, Xing X, Di Q, etal. 2021. Daily 8-Hour Maximum and Annual 03 Concentrations for the Contiguous United States, 1-km Grids, v1 (2000-2016). Palisades, New York: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/a4mb-4t86.

56. Requia WJ, Di Q, Silvern R, Kelly JT, Koutrakis P, Mickley LJ, et al. 2020. An ensemble learning approach for estimating high spatiotemporal resolution of ground-level ozone in the contiguous United States. Environ Sci Technol 54(18):11037-11047, PMID: 32808786, https://doi.org/10.1021/acs.est.0c01791.

57. Di Q, Amini H, Shi L, Kloog I, Silvern R, Kelly J, et al. 2019. An ensemble-based model of PM2.5 concentration across the contiguous United States with high spatiotemporal resolution. Environ Int 130:104909, PMID: 31272018, https://doi.Org/10.1016/j.envint.2019.104909.

58. Di Q, Amini H, Shi L, Kloog I, Silvern R, Kelly J, et al. 2020. Assessing N02 concentration and model uncertainty with high spatiotemporal resolution across the contiguous United States using ensemble model averaging. Environ Sci Technol 54(3):1372-1384, PMID: 31851499, https://doi.org/10. 1021/acs.est.9b03358. 59. Hornung RW, Reed LD. 1990. Estimation of average concentration in the presence of nondetectable values. Appl Occup Environ Hyg 5(1):46-51, https://doi.org/10. 1080/1047322X.1990.10389587.

60. van Buuren S, Groothuis-Oudshoorn K. 2011. Mice: multivariate imputation by chained equations in R. J Stat Softw 45(3):1-67, https://doi.org/10.18637/jss. v045.i03. 61. Vyas DA, Jones DS, Meadows AR, Diouf K, Nour NM, Schantz-Dunn J, et al. 2019. Challenging the use of race in the Vaginal Birth After Cesarean section calculator. Womens Health Issues 29(3):201-204, PMID: 31072754, https://doi.org/10. 1016/j.whi.2019.04.007.

62. Eneanya ND, Kostelanetz S, Mendu ML. 2021. Race-free biomarkers to quantify kidney function: health equity lessons learned from population-based research. Am J Kidney Dis 77(5):667-669, PMID: 33583623, https://doi.Org/10.1053/j.ajkd. 2020.12.001. 63. Umek W, Fischer B. 2020. We should abandon "race" as a biological category in biomedical research. Female Pelvic Med Reconstr Surg 26(12):719-720, PMID: 33136609, https://doi.org/10.1097/SPV.0000000000000979.

64. Vyas DA, Eisenstein LG, Jones DS. 2021. Hidden in plain sight-reconsidering the use of race correction in clinical algorithms. Obstet Gynecol Surv 76(1):5-7, https://doi.org/10.1097/01.ogx.0000725672.30764.f7.

65. Musachio EAS, de Freitas Couto S, Poetini MR, Bortolotto VC, Dahleh MMM, Janner DE, et al. 2021. Bisphenol A exposure during the embryonic period: insights into dopamine relationship and behavioral disorders in Drosophila mel-anogaster. Food Chem Toxicol 157:112526, PMID: 34461193, https://doi.org/10. 1016/j .f ct.2021.112526.

66. Grandjean P, Bellinger D, Bergman A, Cordier S, Davey-Smith G, Eskenazi B, et al. 2008. The faroes statement: human health effects of developmental exposure to chemicals in our environment. Basic Clin Pharmacol Toxicol 102:73-75, PMID: 18226057, https://doi.Org/10.1111/j.1742-7843.2007.00114.x.

67. Garcia M, Salazar R, Wilson T, Lucas S, Herbert G, Young T, et al. 2021. Early gestational exposure to inhaled ozone impairs maternal uterine artery and cardiac function. Toxicol Sci 179(1 ):121-134, PMID: 33146391, https://doi.org/10. 1093/toxsci/kfaa164.

68. Maynard SE, Min J-Y, Merchan J, Lim K-H, Li J, Mondal S, et al. 2003. Excess placental soluble fms-like tyrosine kinase 1 (sFltl) may contribute to endothelial dysfunction, hypertension, and proteinuria in preeclampsia. J Clin Invest 111(5):649-658, PMID: 12618519, https://doi.org/10.1172/JCI17189.

69. Hoffmann J, Ossada V, Weber M, Stepan H. 2017. An intermediate sFlt-1/ PIGF ratio indicates an increased risk for adverse pregnancy outcome. Pregnancy Hypertens 10:165-170, PMID: 29153672, https://doi.Org/10.1016/j. preghy.2017.08.003.

70. Pang L,WeiZ, Li 0, Huang R, Qin J, Chen H, et al. 2013. An increase in vascular endothelial growth factor (VEGF) and VEGF soluble receptor-1 (sFlt-1) are associated with early recurrent spontaneous abortion. PLoS One 8(9):e75759, PMID: 24098721, https://doi.org/10.1371/journal.pone.0075759.

71. Shore VH, Wang TH, Wang CL, Torry RJ, Caudle MR, Torry DS, et al. 1997. Vascular endothelial growth factor, placenta growth factor and their receptors in isolated human trophoblast. Placenta 18(8):657-665, PMID: 9364601, https://doi.org/10.1016/s0143-4004(97)90007-2.

72. Ahmed A, Dunk C, Ahmad S, Khaliq A. 2000. Regulation of placental vascular endothelial growth factor (VEGF) and placenta growth factor (PIGF) and soluble

Flt-1 by oxygen-a review. Placenta 21 Suppl A:S16-S24, PMID: 10831117, https://doi.org/10.1053/plac.1999.0524.

73. Luttmann-Gibson H, Sarnat SE, Suh HH, Coull BA, Schwartz J, Zanobetti A, et al. 2014. Short-term effects of air pollution on oxygen saturation in a cohort of senior adults in Steubenville, Ohio. J Occup Environ Med 56(2):149-154, PMID: 24451609, https://doi.org/10.1097/JOM.0000000000000089.

74. DeMeo DL, Zanobetti A, Litonjua AA, Coull BA, Schwartz J, Gold DR, et al. 2004. Ambient air pollution and oxygen saturation. Am J Respir Crit Care Med 170(4):383-387, PMID: 15142869, https://doi.org/10.1164/rccm.200402-2440C.

75. Hirashima C, Ohkuchi A, Arai F, Takahashi K, Suzuki H, Watanabe T, et al. 2005. Establishing reference values for both total soluble Fms-like tyrosine kinase 1 and free placental growth factor in pregnant women. Hypertens Res 28(9):727-732, PMID: 16419645, https://doi.org/10.1291/hypres.28.727.

76. Hendrix M, Bons J, van Haren A, van Kuijk S, van Doom W, Kimenai DM, et al. 2020. Role of sFlt-1 and PIGF in the screening of small-for-gestational age neonates during pregnancy: a systematic review. Ann Clin Biochem 57(1):44-58, PMID: 31762291, https://doi.org/10.1177/0004563219882042.

77. Daniel S, Kloog I, Factor-Litvak P, Levy A, Lunenfeld E, Kioumourtzoglou M-A, et al. 2021. Risk for preeclampsia following exposure to PM25 during pregnancy. Environ Int 156:106636, PMID: 34030074, https://doi.Org/10.1016/j. envint.2021.106636.

78. Hu H, Ha S, Xu X. 2017. Ozone and hypertensive disorders of pregnancy in Florida: identifying critical windows of exposure. Environ Res 153:120-125, PMID: 27940104, https://doi.Org/10.1016/j.envres.2016.12.002.

79. Zheng Y, Bian J, Hart J, Laden F, Wen TS-T, Zhao J, et al. 2022. PM25 constituents and onset of gestational diabetes mellitus: identifying susceptible exposure windows. Atmos Environ 1994) 291:119409, PMID: 37151750, https://doi.org/10. 1016/j.atmosenv.2022.119409.

80. Cao W, Dong M, Sun X, Liu X, Xiao J, Feng B, et al. 2020. Associations of maternal ozone exposures during pregnancy with maternal blood pressure and risk of hypertensive disorders of pregnancy: a birth cohort study in Guangzhou, China. Environ Res 183:109207, PMID: 32050130, https://doi.Org/10.1016/j.envres. 2020.109207.

81. Olsson D, Mogren I, Forsberg B. 2013. Air pollution exposure in early pregnancy and adverse pregnancy outcomes: a register-based cohort study. BMJ Open 3(2):e001955, PMID: 23386578, https://doi.org/10.1136/bmjopen-2012-001955.

82. Brokamp C, LeMasters GK, Ryan PH. 2016. Residential mobility impacts exposure assessment and community socioeconomic characteristics in longitudinal epidemiology studies. J Expo Sci Environ Epidemiol 26(4):428-434, PMID: 26956935, https://doi.org/10.1038/jes.2016.10.

83. Gasparrini A. 2014. Modeling exposure-lag-response associations with distributed lag non-linear models. Stat Med 33(5):881-899, PMID: 24027094, https://doi.org/10.1002/sim.5963.