Introduction

Particulate matter (PM) is a complex mixture of both solid and liquid particles that spans many sizes and consists of various chemical components, including carbon and organic compounds, and inorganic compounds. The different components come from specific sources and can be grouped into primary PM, which is directly emitted from traffic, industrial processes, and burning waste; and secondary PM, which forms in the atmosphere from chemical reactions. PM with an aerodynamic diameter ≤2.5 μm (PM_2.5) has been shown to have an adverse impact on the human body through direct deposits in the lung and, to some extent, by penetrating the alveolar membranes. These tiny particles activate inflammatory processes by releasing mediators resulting in imbalance in the autonomic nervous system and neuroendocrine pathway [1]. There is an increasing body of literature showing the adverse effects of ambient air pollution, predominantly on respiratory and cardiovascular health. Previous studies have established a strong relationship between ambient PM_2.5 and adverse cardiovascular effects, including elevated blood pressure among the general population [2]; however, their effects on the cardiovascular system during pregnancy, including the occurrence of hypertensive disorders in pregnancy (HDP) remains unclear [3–7].

HDP, which complicates 5% to 10% of pregnancies, is a major cause of maternal and fetal morbidity and mortality [8–10]. During pregnancy, there is an increase in blood volume by 30% to 50% to adapt to the need for increased metabolic demand of the growing fetus [11]. This physiological hypervolemia during pregnancy results in an increased cardiac output. There is also a compensatory increase in heart rate in an effort to increase oxygenation in the face of dilutional hypervolemia. The concomitant increase in the rate of hyperventilation potentially leads to inhalation of more ambient toxic substances during pregnancy, which may contribute to other maternal morbidities, including HDP.

The existing epidemiological findings regarding the association of PM_2.5 exposure with HDP are inconsistent [3–6]. Previous studies have shown both harmful impacts [12–20] and negative or null associations [21,22]. Previous studies examining the relationship between PM_2.5 exposure and preeclampsia also reported inconsistent associations [23–28]. Inconclusive results may be partially due to differences in study populations, sample sizes, methods of exposure assessments, study regions with different PM_2.5 levels and compositions, climate conditions, and failure to differentiate different magnitudes of HDP.

HDP are classified into 4 types based on the severity of clinical features by the American College of Obstetricians and Gynecologists (ACOG) [29]: chronic hypertension, gestational hypertension (GH), preeclampsia (PE), and chronic hypertension with superimposed preeclampsia and eclampsia (E) (S1 Appendix). The mechanisms linking air pollution and HDP subcategories may be different. For example, the pathophysiology of GH is thought to be distinct from that of PE’s in terms of alterations in placental vasculature and angiogenic factors [30]: PE-E involves chronic uteroplacental ischemia from maternal angiogenic imbalance with a changed interaction between vasoactive cytokines, leading to vasospasm [31,32], as opposed to GH. Therefore, examining HDP subcategories separately would be more elucidative about the true associations of the risk factors with HDP and the underlying mechanistic pathways [33].

Additionally, some PM_2.5 constituents and sources may be more harmful than others to health [34,35]. Previous studies found that different PM_2.5 constituents and exposure windows were differently associated with several adverse pregnancy outcomes, such as gestational diabetes mellitus [36] and postpartum depression [37], which is important to develop corresponding interventions targeting the main culprits and critical windows. However, very limited studies have investigated the windows of susceptibility to PM_2.5 and HDP risk [38], and the effects of PM_2.5 chemical compositions [12]; no prior study has explored the HDP risk related to overall mixture effects of PM_2.5.

The objective of this study is to investigate the relationship between maternal residential exposures to PM_2.5 total mass and its constituents (i.e., sulfate, ammonium, nitrate, organic matter, and black carbon) and HDP in a large pregnancy cohort taken from Kaiser Permanente Southern California (KPSC) electronic health record (EHR) data between 2008 and 2017. Specific aims are to (1) examine the associations between HDP and exposure to PM_2.5 mass and its constituents by the severity of the outcomes (GH versus PE-E); and (2) identify windows of susceptibility to PM_2.5 exposure during pregnancy. The primary hypothesis is that maternal exposure to PM_2.5 is associated with an increased risk of developing HDP, while the associations would differ by the severity of HDP, PM_2.5 constituents, and different exposure windows during pregnancy.

Methods

Results

The descriptive statistics of selected demographic and pregnancy characteristics and air pollution levels are presented in Table 1. In total, 4.8% (17,977/373,905) and 5.0% (19,381/386,361) of eligible pregnancies were GH and PE-E cases, respectively. Among PE-E group, 19,334 cases of PE, 3,054 cases of preeclampsia superimposed upon chronic hypertension, and 283 cases of eclampsia were identified.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of the study population, 2008–2017.

https://doi.org/10.1371/journal.pmed.1004395.t001

Patients with GH were more frequent among older mothers, African American or non-Hispanic white mothers, mothers with college (≤4 years), mothers who live in high-income neighborhoods, ever smokers, smoking or passive smoking mothers, mothers without MediCal/Medicaid insurance, primiparous mothers, and obese mothers; PE-E cases were more frequent among older mothers, African American or Hispanic mothers, mothers with 9 Grade to college <4 years, mothers who live in low-income neighborhoods, ever smoker, smoking or passive smoking mothers, mothers with MediCal/Medicaid insurance, primiparous mothers, and obese mothers (p < 0.05).

Summary statistics and Pearson correlation coefficients between air pollution metrics during the entire pregnancy are presented in Table 2. PM_2.5 total mass was highly correlated with most PM_2.5 chemical constituents, including PM_2.5 organic matter (r = 0.91), nitrate (r = 0.85), black carbon (r = 0.79), and ammonium (r = 0.75), and moderately correlated with PM_2.5 sulfate (r = 0.48). Relatively weaker correlations were noticed between sulfate and other PM_2.5 constituents (r ≤ 0.45). Moderate to strong correlations were observed between other PM_2.5 constituents (nitrate, ammonium, organic matter, and black carbon) (r ≥ 0.53). All the correlation coefficients in Table 2 are statistically significant (p < 0.05).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary statistics and Pearson correlation coefficients between air pollution exposure metrics during the entire pregnancy.

https://doi.org/10.1371/journal.pmed.1004395.t002

Results from Cox regression models are illustrated in Fig 1. No significant associations were observed between the risks of GH and PM_2.5 and its constituents; while the relative risks of PE-E were positively associated with PM_2.5 total mass [HR = 1.07 (95% CI [1.04, 1.10] p < 0.001)] and 2 PM_2.5 constituents: highest for PM_2.5 black carbon [HR = 1.12 (95% CI [1.08, 1.16] p < 0.001)], followed by PM_2.5 organic matter [HR = 1.06 (95% CI [1.03, 1.09] p < 0.001)]. In sensitivity analysis (S4 Appendix), similar results were observed for associations in base models between air pollution and GH or PE-E after further adjusting for pre-pregnancy BMI and primiparity, or applying the discrete time approach models. The PAFs for PE-E corresponding to the standards of the US EPA (9 μg/m³) and the WHO (5 μg/m³) were 6.37% and 12.26%, respectively. For example, 6.37% PE-E cases among our study population could be attributed to PM_2.5 levels above 9 μg/m³. In the multi-pollutant model (Table 3), β coefficients > 0 indicate positive weights of individual constituents; β coefficients < 0 indicate negative weights of individual constituents. The overall mixture coefficient ψ (log HR) from quantile g-computation is the sum of all β coefficients of the exposures of interest. Simultaneous increase in all of the 5 PM_2.5 constituents by a quartile was associated with a higher relative risk of PE-E [HR = 1.05 (95% CI [1.03, 1.07] p < 0.001)], and black carbon gave the greatest contribution of overall mixture effects (70.63%) among all individual constituents, followed by organic matter (25.59%) and sulfate (3.79%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Adjusted HRs and 95% CIs of air pollution during pregnancy associated with GH and PE-E.

HRs and 95% CIs were calculated for per IQR increment for each air pollutant. Model adjusted for maternal age, race/ethnicity, education, block group household income, smoking and passive smoking status during pregnancy, insurance type, season, and year of infant birth. Zip code was fitted as a random effect. CI, confidence interval; GH, gestational hypertension; HR, hazard ratio; IQR, interquartile range; PE-E, preeclampsia-eclampsia.

https://doi.org/10.1371/journal.pmed.1004395.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted HRs and 95% CIs of preeclampsia-eclampsia associated with 1 quartile increase in PM_2.5 mixture during pregnancy based on quantile-based g computation.

https://doi.org/10.1371/journal.pmed.1004395.t003

In the time window analyses during gestational months 1 to 8 (Fig 2 and S5 Appendix), the GH group showed negative associations with exposure to PM_2.5 total mass at month 1 and months 7 to 8, positive associations at month 3, and nonsignificant associations for other gestational months. On the other hand, elevated risks of PE-E were associated with PM_2.5 exposure in months 2 to 6, with the highest risk identified at month 4 [HR = 1.05 (95% CI [1.03, 1.07] p < 0.001)]. In terms of PM_2.5 constituents, only exposure to PM_2.5 sulfate (months 1 to 3) and organic matter (months 3 to 6) were associated with higher GH risks. For PE-E, windows of increased risks occurred during late first trimester and second trimester for all PM_2.5 constituents of interest.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Monthly associations between maternal exposure to PM_2.5 during pregnancy and hypertensive disorders of pregnancy.

N = 373,905 for GH cohort; N = 386,361 for PE-E cohort. HRs and 95% CIs were calculated for per IQR increment for each air pollutant. Models adjusted for maternal age, race/ethnicity, education, household income, smoking and passive smoking status during pregnancy, parity, insurance type, season, and year of infant birth. CI, confidence interval; GH, gestational hypertension; HR, hazard ratio; IQR, interquartile range; PE-E, preeclampsia-eclampsia.

https://doi.org/10.1371/journal.pmed.1004395.g002

Subgroup analysis (S6 Appendix) was conducted stratified by maternal age, race/ethnicity, household income, smoking status during pregnancy, pre-pregnancy BMI, and primiparity. No substantial increased GH risks associated with PM_2.5 were observed in stratified analyses. In the PE-E analyses, positive associations were significantly stronger among Hispanic and African American mothers and mothers who lived in low- to middle-income neighborhoods (p < 0.05), but no significant evidence of heterogeneity for other potential effect modifiers.

Discussion

This is a large retrospective cohort study of 386,361 pregnant women residing in Southern California from 2008 to 2017. Our results show that exposures to PM_2.5 total mass, organic matter, and black carbon were associated with an increased risk of PE-E, but not with GH. In multi-pollutant models, PM_2.5 black carbon contributed the most to the overall mixture effects among all PM_2.5 constituents. The late first trimester and second trimester are likely the most influential time windows for air pollution and PE-E. Furthermore, Hispanic and African American mothers and mothers who live in low- to middle-income neighborhoods were associated with higher risks for PE-E on exposure to PM_2.5.

To the best of our knowledge, it is the first study to examine the associations between PM_2.5 constituents and mixtures and GH and PE-E separately in a large and diverse population in regions with relatively low air pollution levels. Two recent studies in China showed positive associations between PM_2.5 total mass and GH with null association for PE-E [12,14]. Interestingly, our study observed opposite results of positive associations between PM_2.5 and PE-E with null associations for GH. Factors that may partially explain the inconsistent results include differences in exposure levels, exposure assessment methods, criteria for HDP diagnosis, and study region and population. For example, the PM_2.5 concentrations in those studies were approximately 4 to 5 times (around 50 μg/m³) that of our study. They estimated air pollution exposure based on a hospital’s address rather than individual maternal address that may lead to more exposure misclassification. Furthermore, in those studies, the outcome was measured with physician-filled questionnaires instead of using diagnostic codes, which might introduce recall bias. Our findings of positive association of PE-E with exposure to PM_2.5 are supported by a Swedish study with similar air pollution levels as that of our study (OR = 1.35 (95% CI [1.11, 1.63]) per 5 μg/m³ increment of PM_2.5) [57]. A meta-analysis also reported that exposure to PM_2.5 enhanced the risk of PE, but not for GH [5]. Our finding of positive associations between PE-E and PM_2.5 black carbon is also consistent with previous studies [12,57].

Among HDP types (S1 Appendix), HDP with isolated elevated blood pressure with no systemic involvement is GH, and HDP with organ involvement or systemic manifestations (i.e., proteinuria, involvement of kidney, elevated liver enzymes, or decreased platelet count) constitute PE, that also includes superimposed preeclampsia, and hemolysis, elevated liver enzymes, and low platelet count (HELLP) syndrome, whereas, preeclampsia with neurological manifestations is considered as eclampsia. In our analyses, we observed that exposures to PM_2.5 total mass, black carbon, and organic matter are positively associated with the severe categories in the HDP spectrum (i.e., PE-E), while the milder form of HDP (i.e., GH) may not be affected as such (S4 and S5 Appendix). The differential associations between air pollution and HDP subcategories may be explained by different mechanisms involved with mild and severe varieties of HDP. Increased expression of cytochrome P-450 and induction of stress response enzymes by PM_2.5 have been documented [58]. The cytochrome P-450 liver enzymes induction pathway could result in rapid clearing of vaso-constrictive cytokines from the system [59]. Therefore, in cases of mild HDP, PM_2.5 may lead to vasodilatation by inducing breakdown of vasoconstrictive factors and thus not contributing to the risk for GH. The mechanistic pathway for PE-E on the other hand has been suggested to be the result of compromised trophoblast invasion by placental vasculatures, release of placental vasoactive substances leading to marked vaso-constriction and resulting placental hypoxia [60]. PM_2.5 exposure has been linked to endothelial dysfunction [61]. An imbalance between vasoactive substances, including vasoconstrictive factors (e.g., Thromboxane A2 and endothelin) and vasodilator factors (e.g., prostacyclin and nitric oxide), have been demonstrated as potential mechanisms of PE-E [60]. Therefore, exposure to the ambient pollution could trigger an increased production of placental vasoconstrictive factors that would overwhelm the cytochrome P-450 enzymes induction system leading to more severe categories of HDP, explaining the positive associations of pollution exposure with PE-E. Previous studies also documented elevated placental biomarker 3-nitrotyrosine and hypomethylated leptin promoters in placental tissue on exposure to PM_2.5 and black carbon [62,63], which has been linked to high-risk pregnancies such as preeclampsia [64]. These mechanistic pathways are shown in S7 Appendix.

It should be noted that PE was more common among smoking mothers in our study population, which is different from previous studies reporting a protective effect of smoking on preeclampsia risk [65]. It has been proposed that combustion products in tobacco (e.g., carbon monoxide) [66] might be responsible for the protective effect of smoking on PE due to vasodilation. However, the inverse association between smoking and PE might be controverted due to multiple sources of bias, including eligibility criteria, losses to follow-up of women potentially at risk, misclassification, competing events, or incorrect adjustment in previous studies [55]. Further studies are warranted to extricate bias when evaluating the smoking-preeclampsia paradox.

Although several studies examined the associations between air pollution and HDP, studies explored different exposure windows are insufficient; findings are not consistent among studies, including the first, second, or third trimester [4–6]. Overall, the first and the second trimester were more susceptible to PM_2.5 exposure in our study. Specifically, we found gestational month 3 as the critical exposure period for GH, and months 2 to 6 as the critical exposure period for PE-E. Our results are partially consistent with findings from another study in Southern California, indicating that first trimester exposure to PM_2.5 was associated with increased odds of HDP [38]. However, this study investigated the PM_2.5-HDP relationship among Hispanic population without considering separate HDP outcomes and had a small sample size (n = 298). To date, further research to identify the susceptible windows of PM_2.5 is still needed given the inconclusive results among previous studies.

As both health and economic burdens increase with severity of HDP, more research on the associations between environmental exposure and reproductive health is needed. We used the PAF to assess the public health impact of exposure to PM_2.5 during pregnancy. Our results showed that a 6.37% absolute reduction of PE-E cases (N = 1,235) in our study population would occur if PM_2.5 exposure levels were reduced to 9 μg/m³ based on the US EPA air quality standards. Women with PE-E require prolonged hospitalization often including emergent cesarean section, which significantly increase health care costs. In 2003, the average (direct and indirect) PE management cost was estimated at $11,208 per woman in the United States [67]. PE has been recognized as one of the independent risk factors of future cardiovascular diseases and stroke, it also affects other vital organs, psychological health, and fetal health [68], thus contributing to the long-term social and financial impact of the health care system [69]. Although everyone is affected by air pollution, the risk may disproportionately be higher among women of certain socioeconomic backgrounds, such as women living in poverty and the minority populations in our study [70], subsequently increasing the burden of HDP. Shen and colleagues reported relatively larger effect estimates of PM_2.5 exposure on HDP among underweight women (BMI <18.5) compared to their counterparts, but the differences between subgroups were not statistically significant [12]. Mobasher and colleagues found that PM_2.5 exposure was significantly associated with HDP among non-obese women (BMI <30), but not obese women [38]. However, our results did not reveal any heterogeneity between the obese and non-obese subgroups. Further, identifying the main sources of air pollution associated with HDP is also important to initiate actions targeting this highly modifiable risk factor (e.g., reducing fuel emissions) to mitigate those adverse effects on reproductive health and future generations.

The main strengths of this study include the large and diverse study population in Southern California; comprehensive and detailed demographic information and medical records from KPSC EHRs allowed us to control for a number of potential confounders in our analysis; the use of time-to-event models to examine the relationship between HDP and air pollution exposure, and identify critical exposure windows; major PM_2.5 chemical components from well-validated air pollution models; the use of an innovative statistical method to estimate the joint effects of air pollution mixtures; and consider residential changes during pregnancy, which may enhance the accuracy for the air pollution exposure assessments in this study. Furthermore, while most previous studies used combined GH and PE as the umbrella diagnosis of HDP, our study addressed the gap in the literature to link the effect of air pollution and different expressions of HDP and to understand corresponding pathophysiology.

There are certain limitations in this study. First, air pollution exposure was estimated only for residential outdoors without considering other exposure sources which could also contribute to the observed associations (e.g., indoor exposures, occupational exposures), or activity patterns due to data unavailability, which may lead to exposure misclassification and bias estimated associations in either direction. Future studies utilizing personal air pollution sampling or blood biomarkers for exposure to the pollutants would help alleviate exposure misclassification. Second, identification of disease status solely based on ICD diagnostic codes may lead to potential outcome misclassification [71]. Third, the dates of HDP diagnosis were collected from the EHR based on the routine screening during patients’ prenatal visits, which may not actually coincide with the start of the disease process. Nonetheless, the KPSC EHRs include repeated measurements from multiple time points during the entire pregnancy (7 ± 4 times) to minimize the potential bias in the estimates of time-varying exposures and outcomes. Moreover, although several covariates were adjusted, some HDP-related factors were unavailable in our analysis, such as family or personal history of HDP. In addition, further exploration of the severity and different subtypes of HDP (e.g., early-onset PE versus late-onset PE, HELLP syndrome, which is not available in our database) would give additional insights of the effects of air pollution on HDP and mechanisms. Given the complexity of air pollution and related environmental exposures, further research is also warranted to examine the effect of other air pollutants (e.g., nitrogen oxides and sulfur oxides), environmental factors (e.g., traffic noise and meteorological factors) [15,72,73], and their joint effect on HDP. Finally, animal studies examining molecular changes from air pollution exposure may shed light on the causation pathway and pathogenesis of GH and PE.

In conclusion, we found that exposures to PM_2.5 total mass, organic matter, and black carbon were associated with an increased risk for PE-E, but not for GH. The main effect of increased PE-E risk was driven by PM_2.5 black carbon (71%) and PM_2.5 organic matter (26%). The susceptible window of air pollution exposure associated with PE-E is gestational months 2 to 6. Hispanic and African American mothers and mothers who live in low- to middle-income neighborhoods may be more vulnerable to air pollution on HDP risk.

Supporting information

Acknowledgments

The authors thank the patients of Kaiser Permanente for helping to improve care through the use of information collected through our electronic health record systems.