Association of Ambient Air Pollution with Perceived Stress in Pregnant Women: A Prospective Cohort Study

Background: Air pollution is associated with perceived stress in the general population, but its inuence on maternal perceived stress during pregnancy has not been investigated.We aimed to investigate the relationship between air pollution and non-specic perceived stress among pregnant women. Methods: Our analysis included2162 pregnant women who had participated in the cohort for childhood origin of asthma and allergic disease study between 2008 and 2015. Maternal exposures to particulate matter with an aerodynamic diameter < 2.5 µm (PM 2.5 ) and < 10 µm (PM 10 ), as well as to nitrogen dioxide (NO 2 ) and ozone (O 3 ) for each trimesterand the entire pregnancy were determined using land-use regression models. Maternal perceived stress during the third trimester was assessed using the 14-item Perceived Stress Scale (PSS): scores ranged from 0-56 with higher scores indicating increased stress. Linear regression models were applied to estimate associations between PSS scores and each air pollutant, after adjusting for socio-demographic and behavioral covariates. Results: In single-pollutant models,after adjustment, an IQR increase in the whole pregnancy exposure to PM 2.5 and PM 10 and O 3 in the third trimester was related to 0.37 (95% condence interval [CI]: 0.01, 0.74) and 0.55 (95% CI: 0.12, 0.98) and 0.29 (95% CI: 0.05, 0.52) points increase in the PSS score, respectively.This association was more evident in women with child-bearing age and lower levelofeducation, and the association of PM 10 was stronger in thespring season.In multi-pollutant models, exposures to PM 10 and O 3 were associated with higher perceived stress. Conclusion:Our ndings suggest that pregnancy exposure to PM 2.5 , PM 10 and O 3 is positively associated with maternal PSS score during the third trimester.


Introduction
Maternal psychosocial stress is relatively common during the prenatal period [1,2]. Stress perceived by mothers during this period has been associated with adverse outcomes for mothers, including postpartum depression and suicide [3], as well as adverse outcomes for the fetus, including preterm birth [4], low birth weight [5], congenital heart defects [6] and impaired cognitive development of neonate [7]. In the last decades, air pollution has emerged as a potential risk factor for the development of psychosocial stress. Epidemiological studies show associations of air pollution with a range of psychiatric disorders, including cognitive decline [8], anxiety and depression [9,10] and suicide [11]. Experimental research supports these epidemiological associations [12,13], indicating that exposure to air pollution leads to depression-like response in mice [13]. So far, few epidemiological studies have investigated the potential effects of air pollution on maternal psychosocial stress, and more speci cally perceived stress, which is thought to contribute to the development of psychiatric disorders, including depression, in chronic state [14,15].
Psychosocial stress can be manifested in many forms, such as perceived stress, anxiety, and depression.
Previous studies have shown that air pollution is associated with perceived stress [16], mood [17], depressive symptoms [10], anxiety symptoms [9], psychiatric emergency room visits [18] and mental health disorders as summarized in a meta-analysis [19]. Perceived stress, widely assessed using the Perceived Stress Scale (PSS), is a measure of the degree to which situations in one's life are appraised as stressful and has been suggested as a global stress measure [20]. A longitudinal study of older men reported that exposure to air pollution, including particulate matter with an aerodynamic diameter < 2.5 µm (PM 2.5 ) and nitrogen dioxide (NO 2 ), is linked with higher PSS score, particularly in cold months [16].
However, other study that explored associations between personal exposure to NO 2 and PSS in the healthy elderly persons has reported insigni cant results [17].
The pregnant women are especially vulnerable to air pollution due their increased ventilation rate for higher oxygen requirements of the developing fetus and decreased oxygen-binding capability [21]. In addition, brain is susceptible to environmental factors due to high metabolic demands [9]: air pollutants are capable of reaching the brain, potentially by crossing the blood-brain barrier, and can trigger neuroin ammation, leading to pathological changes [22]. To date, only few studies have investigated the relationship between air pollution and maternal psychosocial stress in perinatal period. These studies have mostly focused on depression or emotional stress [23 − 26], while other indicator of psychosocial stress, such as perceived stress, has not been studied on pregnant women. Maternal perceived stress during pregnancy is an important predictor for postpartum depression [27]. Highlighting the relationship between air pollution and the perceived stress will help to understand a possible mechanism that relates air pollution to depression.
The aim of this study was to investigate potential associations between exposure to air pollution during different trimesters and maternal PSS, which was assessed at 36th week of pregnancy, hypothesizing that increased level of air pollution would be associated with higher PSS in our sample of pregnant women.

Study population
Participants included in this study were enrolled in the Cohort for Childhood Origin of Asthma and Allergic disease (COCOA), a prospective hospital-based birth cohort study conducted in South Korea, details of which have been published previously [28]. Brie y, pregnant women before 26 weeks of gestational age were recruited at ve medical centers and eight public health centers in the Seoul metropolitan area between 2008 and 2015. For this analysis, subjects were limited to those with maternal PSS data. Of the 3102 women recruited, 638 women without PSS data and 86 with multiple records were excluded.
Women with missing information on residential address and preterm births were not included in the study. Some covariates with missing information were also excluded. Supplementary material (Fig. S1) provides further details of the exclusion criteria. The nal study population consisted of 2162 pregnant women. Prior to enrollment, written informed consent was obtained from all study subjects.  10 ), NO 2 and O 3 at maternal residential addresses were estimated using LUR model, as previously described [29,30]. The LUR model used several geographical variables, such as tra c indicators, surrounding-land use, topography and spatial trends, and the nal LUR model included lengths of roads, tra c intensities on nearest roads, total heavy-duty tra c on all roads and a variable representing spatial trends. Model performance was assessed using leave-one-out cross-validation (LOOCV). The models explained 66 − 81% of the variability in the measured PM 2.5 , PM 10 , NO 2 and O 3 levels, and the predicted values tted well with the measured values, as reported in our previous studies [31,32]. The model-adjusted R 2 and LOOCV R 2 for PM 2.5 , PM 10  To examine potential critical exposures during pregnancy, we calculated the average concentrations of PM 2.5 , PM 10 , NO 2 and O 3 for three trimesters and the entire pregnancy: the rst trimester (from 1 to 13 weeks), the second trimester (from 14 to 27 weeks), and the third trimester (from 28 weeks to until birth).

Assessment of maternal perceived stress
Maternal perceived stress was evaluated during the third trimester of pregnancy using the 14-item PSS [20,33], which is a validated measure of stress appraisal and was used to assess the extent to which respondents think their lives have been unpredictable, uncontrollable and overloaded during the previous month. The PSS is a widely used stress appraisal measure and has been veri ed among pregnant women at risk for mental stress (Cronbach's alpha = 0.88) [34]. Each item was rated using a Likert-type ve-point scale that ranges from "never" (0) to "very often" (4). The scores for seven positively stated items in the 14-items scale were reversed, and a total score was obtained by summing all items. The total score ranges from 0 to 56, a high score indicates a high perception of stress. In the current study, the PSS was normal in distribution and had excellent reliability (Cronbach's alpha = 0.884).

Covariates
Potential confounding variables, including maternal age at delivery, maternal education, occupation during pregnancy (yes vs. no), parity (nulliparous vs. parous), smoking history (ever vs. never), family income and drinking during pregnancy (yes vs. no), were ascertained at baseline. The maternal age was divided into two groups: child-bearing age (< 35 years old) and advanced age (≥ 35 years old). Mother's education was categorized into three levels: secondary school, college or university and graduate school. Family income was dichotomized as high (≥ 4 million Korean won per month) or low (< 4 million Korean won per month) [35]. The data on gestational age in weeks was obtained from medical records at delivery. We also obtained medical history of participants according to physicians' diagnoses, including hypertension or diabetes mellitus, asthma, thyroid diseases, malignant tumor and liver diseases. We used children's birth season to indicate a seasonal variation in the perceived stress during the third trimester. The season at delivery was classi ed into spring (March to May), summer (June to August), autumn (September to November) and winter (December to February) according to the climate characteristics in South Korea. Previous studies indicated that these variables were related to psychosocial stress and may affect the estimated association between exposure to air pollution and perceived stress [2,16,36].

Statistical analysis
Descriptive statistics were presented as mean, standard deviation (SD), and frequency (%). Spearman's correlation coe cient was used to estimate the correlations among air pollutants during pregnancy.
Linear regression models were used to assess the association between air pollutants averaged over each trimester and the whole pregnancy and PSS score. Point estimate was estimated by interquartile range (IQR) increase in the exposure level for each air pollutant. Model 1 examined the bivariate associations between PSS score and air pollutants measured during the pregnancy. Model 2 adjusted for age, education, occupation and income only, whereas model 3 additionally adjusted for a number of lifestyle and maternal characteristics, including gestational age, smoking, drinking during pregnancy, parity and season at delivery. In model 4, as our primary model, we added an additional control for chronic health conditions, including asthma, thyroid disease, malignant tumor, liver disease and hypertension or diabetes. In addition to continuous variable, we modeled air pollution as categorical variables (quartiles of air pollution exposure) using linear regression for the model 4. Tests for linear trend across quartiles were conducted by using the median value of air pollutant to each quartile and modeling this variable as a continuous variable in separate regression models.
To further explore the potential nonlinear association between maternal air pollution exposure and PSS score, we performed a restricted cubic spline analysis based on linear regression model [37]. The optimal number of knots for the spline analysis was determined by Akaike information criterion (AIC). Nonlinearity was assessed by testing the regression coe cient of nonlinear term, with p for nonlinearity < 0.05 indicating a nonlinear association: the overall association indicated that the regression coe cients of both linear and non-linear terms of the factor were equal to zero.
In addition, strati ed analyses were conducted to examine the effect modi cation by maternal age, parity, education, income and season at delivery. To evaluate the signi cance of effect modi cation on the multiplicative scale, we included an interaction (product) term between air pollution exposure and each of these characteristics in our primary model.
Then several sensitivity analyses were conducted. First, to test the stability of the effects after controlling for other pollutants, we built multi-pollutant models for pregnancy exposure to all pollutants except for PM 2.5 due to its high correlation with PM 10 (Table S1). Second, we reanalyzed the primary model after excluding women with chronic health condition. Third, we reanalyzed the primary model using multiple imputation technique. Finally, we evaluated the robustness of our results to unmeasured cofounding by calculating the E-value [38], which indicates the minimum strength of association that an unmeasured confounder must have to explain away associations between exposure (air pollution) and outcome (PSS score). We reported the E-value for relative risk estimate for air pollution using our primary model. Statistical analyses were conducted using STATA (version 16.0; Stata Corporation). P < 0.05 was considered statistically signi cant. Table 1 presents the summary statistics of the study population. The mean age of pregnant women was approximately 33 years, and the mean PSS score was 20.2 with a standard deviation of 7.2. Among mothers, 1315 (60.8%) were primiparous, 2063 (95.6%) had higher educational level (college or above) and 32 (1.5%) were diagnosed with diabetes or hypertension in pregnancy. Most participants had high income (62.2%), with the majority of women reported to be never smokers (92.2%).  (Table S1). Spearman's correlation coe cients between PM 2.5 and PM 10 during pregnancy were strong (Spearman's correlation coe cient r = 0.82 to 0.91). PM 10 showed a moderate positive correlation with NO 2 (r = 0.30 to 0.54),

Results
whereas NO 2 showed a moderate negative correlation with O 3 (r = − 0.46 to − 0.53). Spearman correlations of each air pollutant with itself across three trimesters are presented in Table S2. The levels of air pollution for second trimester were strongly correlated with the levels for the entire pregnancy (r = 0.80 to 0.88).       Table 3. + p-value < 0.1. * p-value < 0.05.
Spline analyses showed that increase in the mean PM 10 concentration during the second trimester and O 3 concentration during the third trimester were associated with higher PSS score (p for overall association = 0.039 and 0.006, respectively) ( Fig. 1B and D): PM 2.5 during the second trimester showed only marginal level of signi cance for overall association (p for overall association = 0.053) (Fig. 1A). Spline analyses showed linear dose-response function of PM 2.5 and PM 10 concentrations during the second trimester (p for nonlinear = 0.743 and 0.999, respectively). In contrast, the association between maternal O 3 exposure during the third trimester and PSS score was nonlinear (p for nonlinear = 0.024).
The overall association between NO 2 during the rst trimester and PSS score was insigni cant (p for overall = 0.364) and did not signi cantly deviate from linearity (p for nonlinear = 0.519) (Fig. 1C), which was consistent with estimates from the other models (Fig. S2). Figure 2 shows evidence of effect modi cation for the relationship between air pollution and PSS by season. Signi cant interactions were observed between spring and exposure to PM 10 during the rst trimester and the entire pregnancy (p for interaction < 0.01); positive associations between PSS score and PM 10 were signi cant in spring, but not in other seasons. Inverse association between NO 2 exposure during the third trimester and PSS was observed in autumn, but the interaction term was not signi cant.
Analyses of effect modi cation by maternal age, parity, education and income are reported in the supplementary material (Fig. S3). In age-strati ed models, we found a signi cant association between PM 10 exposure during the rst and second trimesters and the entire pregnancy and O 3 exposure during the third trimester and PSS score in pregnant women with child-bearing age, but not in women with advanced age. There was a signi cant interaction between O 3 exposure during the third trimester and age group on the PSS score (p for interaction < 0.05), while the interaction for PM 10 exposure in the second trimester and the entire pregnancy was marginal level of signi cance (p for interaction < 0.1). For education, signi cant positive associations between PSS score and exposure to PM 2.5 during the second trimester and the entire pregnancy were only observed for women with less than graduate education (p for interaction < 0.05). For parity and income, the associations of PM 2.5 , PM 10 and O 3 with PSS were greater for primiparous women and women with low income, but p-values for interaction terms were not signi cant (p > 0.05).
In sensitivity analyses, the relationships between maternal exposure to PM 10 during the second trimester and the entire pregnancy and O 3 during the third trimester and PSS score remained signi cant after adding other pollutants into the primary models (multi-pollutant models). The positive relationships between maternal exposure to air pollution and PSS score were not materially changed after excluding mothers with chronic health condition (Table S4). The ndings were also similar when examining the association using the multiple imputation technique (Table S5). The E-values with lower 65% CI and relative risks for the primary model are reported in the supplementary material (Table S6). The results suggest that our conclusions are overall robust to unmeasured confounding bias.

Discussion
In this prospective cohort study of pregnant women, we investigated the association of exposure to air pollution with maternal PSS. We found that maternal exposures to PM 2.5 and PM 10 during the rst and second trimesters and the entire pregnancy and to O 3 during the third trimester were signi cantly associated with higher perceived stress rating, as measured during the third trimester, in the singlepollutant models. The associations between PM 10 exposure during the second trimester and the entire pregnancy and O 3 during the third trimester and PSS remained signi cant in the multi-pollutant models.
In addition, we found the evidence of effect modi cation for the air pollution-perceived stress associations by season, maternal age and education.
Experimental research provides biological plausibility for potential associations of air pollution with stress [39]. Yokota et al. showed that exposure to diesel exhaust particles during pregnancy decreased serotonin in mouse, which may lead to depressive behaviors [13]. Furthermore, epidemiological studies reported the association of air pollution with speci c perceived stress, such as perceived air quality [40], as well as with non-speci c perceived stress [16], where perceptions of air quality were unlikely to change due to the range of air pollution levels. In our study, we aimed to detect changes in non-speci c perceived stress levels, as intermediate steps in the associations observed between exposure to air pollution and depressive symptoms. We focused on pregnant women as there is evidence of high susceptibility for the effects of air pollution among pregnant women. So far, only four studies have investigated the effects of air pollution on psychosocial stress during the perinatal period in a pregnant population [23 − 26]. One of these studies investigated the relationship between air pollution and maternal stress, and found that maternal exposures to PM 2.5 , sulphur dioxide (SO 2 ), and NO 2 were associated with higher scores on Global Severity-Indices, indicating higher levels of emotional stress [26].
In strati ed analyses, we found a signi cant positive effect of PM 10 on PSS score in the spring season, while the effects were not statistically signi cant in other seasons. This nding is consistent with a previous study reporting a higher risk estimate of the link between PM 10 and suicide in the spring season in Korea [11]. The average PM 10 concentration is relatively high in Seoul during the spring season [41], so pregnant women may be more likely to perceive the effect of air pollution in the spring season and the effect may be more apparent in this season. However, some studies reported a stronger association between particulate air pollution and maternal perceived stress or emotional stress in the cold season [16,26], which is inconsistent with our results. The inconsistency is possibly due to the seasonal and geographical variations of air pollutants and different weather conditions and lifestyles [42].
Furthermore, we estimated a stronger positive association of PM 2.5 and PM 10 with PSS score in pregnant women with a lower socio-economic status indicator, particularly in women with lower level of education.
In Korea, air pollution is known to be spatially associated with lower socio-economic status, including low income and lower educational attainment [43,44]. Additionally, individuals with lower socio-economic status may be more likely to have higher perceived stress [45] and may be more susceptible to the effects of air pollution. Thus, a combination of greater pollution exposure and larger in uence of lower socioeconomic status on perceived stress may explain the stronger associations between particulate air pollution and PSS score among pregnant women in our study. In addition, in our study, women with childbearing age showed a signi cant association between air pollution exposure and perceived stress, but not women with advanced age. Young pregnant women are more likely to develop stress-related disorders than those with advanced age [46] that may contribute to the increased vulnerability to air pollution.
Although the underlying biological mechanisms by which air pollution exposure may lead to altered perceptions of stress are yet to be elucidated, our ndings are biologically plausible. Research has demonstrated that exposure to particulate matter increased in ammation in the brains of mice [47]. Chronic in ammation in the brain leads to the formation of reactive oxygen species and oxidative stress, which are reported among the hypothesized biological pathways of mental disorders [48,49]. Pregnant women are reported to be more vulnerable to oxidative stress than the general population due to their altered physiology and increased energy requirements [50]. Maternal air pollution exposure may also lead to stress through the activation of the hypothalamic-pituitary-adrenal axis [51], which is associated with stress and stress-related disorders [52,53].
The major strengths of this study were the prospective design to investigate the role of air pollution on the maternal perceived stress, use of validated models for assessment of exposure to air pollution and adjustment for multiple confounders. In addition, we showed effect modi cation of the association between air pollution and PSS by participant's characteristics, providing insight into susceptibility. Our ndings were robust to multiple sensitivity analyses. Multi-pollutant models enabled us to conclude that both PM 10 and O 3 exposure are associated with higher perceived stress independently from each other.
We believe that our study is the rst to investigate the relationship between air pollution and maternal perceived stress during pregnancy. However, this study had several limitations. First, air pollution estimates were based on maternal residential address, and geospatial data on maternal residential history, movement patterns, and time spent indoors and outdoors were unavailable. This may bias the model of air pollution exposure, particularly if women with perceived stress have different local mobility patterns. Previous research indicated that maternal mobility during pregnancy is usually limited and generally restricted to residential areas [54]. Air pollution tends to be more homogenous across the local areas, which may have little effect on our exposure estimates and was unlikely to have changed our ndings. Second, although we controlled for several important potential confounders, it is possible that other factors may confound the relationship between air pollution exposure and PSS scores, such as noise pollution, meteorologic conditions and exposure to indoor pollutants. However, we evaluated Evalues to see the strength of the relationship a hypothetical unmeasured confounder would have with the air pollution and PSS score to fully account for our results, suggesting that unmeasured confounders might not have a major in uence on the association. Third, perceived stress was measured using selfreports and was assessed only once during pregnancy, a single measurement of the perceived stress might not present maternal stress status throughout pregnancy. Finally, while the homogenous makeup of the COCOA study in the Korean population provided excellent internal validity, its generalizability in other populations may be limited.

Conclusions
Our ndings suggested that maternal exposures to PM 2.5 , PM 10 , and O 3 were related to higher levels of perceived stress among pregnant women. Pregnant women with child-bearing age and lower level of education may be more susceptible to air pollution. In addition, the association of PM 10 with PSS was stronger in spring. Future studies are warranted on how air pollution may lead to higher perceived stress, and the potential role of perceived stress for the association between air pollution and affective disorder.

Consent for publication
Not applicable.

Competing interests
The authors have nothing to disclose.

Availability of data and materials
The datasets generated and/or analyzed during the current study are not publicly available, but are available from the corresponding author on reasonable request.  Adjusted difference in PSS score per IQR increase in air pollutants strati ed by season. Analyses were adjusted for maternal age, education, smoking, occupation, drinking during pregnancy, parity, gestational age, income, asthma, thyroid disease, malignant tumor, liver diseasesand hypertension or diabetes. + p for interaction < 0.1. *p for interaction < 0.05. **p for interaction < 0.01.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. additional le1.docx