Association Between Ambient Air Pollution and Birth Weight by Maternal Individual- and Neighborhood-Level Stressors

Key Points Question When is the sensitive window of ambient air pollution exposure in association with birth weight? Findings In this cohort study of 628 predominately low-income Hispanic women who were pregnant, exposures to particulate matter and nitrogen dioxide in early pregnancy to midpregnancy were significantly associated with lower birth weight, particularly for mothers experiencing higher perceived stress during the prenatal period and living in a neighborhood with a high level of stressors from environmental pollution. Meaning The findings of this study suggest that protecting women who are pregnant from air pollution may improve birth weight, particularly among mothers with high levels of psychological stress or environmental pollution.


Introduction
Newborns with low birth weight face increased risks of neonatal mortality and long-term disease risk. [1][2][3][4][5] In addition to its genetic determinants, 6 birth weight has been associated with prenatal environmental exposures, including ambient air pollution. [7][8][9][10][11][12][13][14][15] Because fetal growth is precisely programmed, 16 specific time windows in pregnancy may be particularly sensitive to the effect of environmental exposures. Such sensitive windows, if well identified, could be a compelling period of targeted intervention and provide a better understanding of environmental influences on fetal development. 17 However, previous findings regarding sensitive windows of air pollution exposure with birth weight have been inconsistent, with each of the 3 trimesters having been identified. [18][19][20][21] Although the whole prenatal period could be a sensitive window to air pollution exposure, erroneous estimation and inconsistencies could also be due to a temporal trend in exposure levels in adjacent time windows. 22 Statistical approaches, such as distributed lag models (DLMs), can better handle temporal trends, thus identifying true sensitive windows. [22][23][24] Applying DLMs is therefore needed to better assess the critical periods of air pollution exposure during pregnancy in association with birth weight.
Robust literature reports an enhanced susceptibility to the adverse effects of prenatal air pollution exposure on birth outcomes in populations with socioeconomic disadvantage and associated stressors. 25,26 Stressors at the individual level, such as psychological perceived stress in women who are pregnant, has been assessed in few longitudinal cohort studies. 27,28 Generally, high levels of prenatal psychosocial stress exposure have been shown to increase children's susceptibility to asthma and neurocognitive underdevelopment in association with air pollution exposures, possibly through altered maternal allostatic load or placental function. 26,29 Moreover, neighborhoodlevel stressors could arise from the total environmental pollution (eg, traffic emissions, water and soil contamination) (hereinafter, cumulative burden), as well as the population's vulnerability to poor health. 30 However, whether individual-level psychological stress and neighborhood-level cumulative burden could be associated with the susceptibility to air pollution exposure with birth weight remains unknown.
In a cohort of predominately low-income Hispanic women who were pregnant, 31 we used DLMs to explore sensitive windows of ambient air pollution exposure in association with birth weight.
Furthermore, we evaluated whether the association between air pollution and birth weight differed by individual-level psychological stress, neighborhood-level cumulative burden, or both.

Study Population
The Maternal and Developmental Risks from Environmental and Social Stressors (MADRES) study is an ongoing prospective pregnancy cohort study started in 2015. 31  As of September 1, 2021, 710 participants had reached delivery and had high-quality birth outcome information abstracted from electronic medical records. We excluded 12 participants who reported smoking during pregnancy and 70 participants who delivered prematurely (<37 gestational weeks). A flowchart describing the final sample size of 628 women is provided in eFigure 1 in the Supplement. Using GPower, version 3.1.9.6, 32 we found our final sample size was sufficient to detect a greater than 4-g difference in birth weight (Cohen f 2 >0.01) with a power of 80% and the type I error of 0.05.

Measurement of Ambient Air Pollution
During prenatal visits, participants completed a questionnaire on residential addresses dated from 1 year before conception through the third trimester, which was further reviewed with study staff to ensure data accuracy. Thereafter, residential addresses and mobility were prospectively collected at each follow-up visit. Geocoded daily residential address histories were assembled for each participant. Daily estimates of 24-hour average nitrogen dioxide (NO 2 ), particulate matter with aerodynamic diameter less than 2.5 μm (PM 2.5 ) and less than 10 μm (PM 10 ), and 8-hour maximal ground-level ozone (O 3 ) were assigned to each participant's residential location from 12 weeks before conception to 36 weeks 7 days during pregnancy, using inverse-distance-squared weighted spatial interpolation from ambient air quality monitoring data (US Environmental Protection Agency Air Quality System), with an average of 4 monitoring stations within 8 to 14 km to each residential address (eTable 2 in the Supplement). The 12 weeks before conception were chosen as a plausible biologically relevant window of exposure that may affect ovulation and the maternal and intrauterine environment close to conception. To reduce impacts from daily fluctuations, we calculated weekly average levels of each pollutant concentration.

Measurement of Birth Weight
Birth weight and newborn sex were abstracted from electronic medical records. Gestational age at birth (median, 39.4 weeks; range, 37.0-42.4 weeks) was estimated based on ultrasonography measurement (87.6%), medical consensus (11.5%), or last menstrual period (0.9%). We calculated sex-specific birth weight for gestational age z scores (BWZ). 33

Covariates
We a priori selected the following modifiers: perceived stress during pregnancy measured by the Perceived Stress Scale (PSS) 34 to represent individual-level psychological stressor, and the cumulative impact score from the California statewide environmental justice screening tool, CalEnviroScreen 4.0 (CES), to represent the neighborhood-level cumulative burden from both total environmental pollution and population vulnerability. 30 The PSS is a widely used, validated 10-item psychological instrument to assess stressors that may possibly influence mental health of the respondent (how the stressor is perceived), 34 which may be more relevant to a list of stressors that will affect one's mental health.  Assessment to identify census tracts in California that are disproportionately burdened by both multiple sources of pollution (eg, air, water, and soil) and high population vulnerability (eg, a higher proportion of poverty, low educational level, prevalence of asthma, and low birth weight). 30 The CES cumulative impact score ranges from 0 to 100, with a higher score indicating a higher cumulative burden, and the score was assigned to each participant based on residential census tract. 30 We chose the California state median to dichotomize a high CES group (Ն50) and a low CES group (<50).

Statistical Analysis
We examined frequencies and proportions to describe the distributions of categorical variables. Birth weight for gestational age z score was described by mean (SD) and tested for differences by covariates using analysis of variance. We used DLMs to examine the association of weekly exposure to each of the air pollutants (ie, PM 2.5 , PM 10 , NO 2 , and O 3 ) with BWZ. 23 The DLMs simultaneously included exposure to each air pollutant's weekly levels from 12 weeks before conception to 36 gestational weeks with a cross-basis, which combined a linear dose-response function and a nonlinear lag-response function with a 3-knot natural cubic spline to constrain correlated weekly exposures. 23 The number and location of knots for each air pollutant were chosen based on the Akaike information criterion and visual inspections. 23 Distributed lag models were adjusted for the aforementioned potential confounders and weekly temperature. Effect size estimates are presented as changes in BWZ per each IQR increase in air pollutant (ie, 4 μg/m 3 for PM 2.5, 12 μg/m 3 for PM 10

Results
The study included 628 pregnant women and their newborns (mean [SD] BWZ, −0.08 [1.03]). The    All results were from distributed lag models adjusted for weekly temperature, maternal age, educational level, maternal prepregnancy body mass index, maternal ethnicity and nativity, cohabitation status, birth order, and enrollment time. Outcome estimation was based on per IQR increases in each air pollutant: particulate matter with aerodynamic diameter less than 2.5 μm (PM 2.5 ), 4 μg/m 3 ; particulate matter with aerodynamic diameter less than 10 μm (PM 10 ), 12 μg/m 3 ; nitrogen dioxide (NO 2 ), 11 ppb; and ozone (O 3 ), 15 ppb. Error bars indicate 95% CI. 9 to 14 gestational weeks was associated with birth weight (−13.5 g; 95% CI, −15.6 to −11.5 g). There was no association of PM 10  In stratification analysis by PSS (Figure 2), we observed inverse associations of PM 2.5 , PM 10 , and NO 2 with birth weight in the high-PSS group (n = 132), but no associations in the low-PSS group (n = 411). In stratification analysis by CES (Figure 3), we observed associations of PM 2.5 and PM 10 with birth weight in the high-CES group (n = 396), but no associations in the low-CES group (n = 230). We further stratified the analysis by the combined groups of PSS and CES scores (Figure 4). In the high-PSS, high-CES group (n = 82), we found wide sensitive windows of PM 2.5 and PM 10  weight. An IQR increase in NO 2 exposure in each week from 23 to 26 weeks also was associated with birth weight (63.2 g; 95% CI, 47.8 to 78.5 g). Some sparse significant differences in BWZ by PM 10 in the low-PSS, high-CES group and by NO 2 in the high-PSS, low-CES group were also seen, both within the window from 5 to 10 gestational weeks. There were no associations between O 3 and birth weight.

Discussion
In the predominately low-income Hispanic population that comprises the MADRES cohort, we observed robust associations of lower birth weight with NO 2 , PM 2.5 , and PM 10 exposures in the window from early pregnancy to midpregnancy. The inverse associations were generally stronger, and the sensitive windows were generally wider within subgroups with either high individual-level perceived stress, high neighborhood-level cumulative burden, or both. We found no evidence for associations of air pollution exposure during the preconception period with birth weight nor for prenatal O 3 exposure with birth weight.
A robust literature exists reporting the inverse association of ambient PM 2.5 and PM 10 exposure with birth weight, and our results are consistent with these overall findings. In a meta-analysis of 17 studies, increases in PM 2.5 exposure during the entire pregnancy were associated with lower birth weight with an effect size similar to ours. 38 We found exposure to PM 10 from 5 to 17 gestational weeks was associated with lower birth weight in women who had high perceived stress, lived in a neighborhood with a high cumulative burden, or both, but not in the overall cohort, suggesting outcomes of PM 10 may be sensitive to population characteristics that have also been noted in previous studies. 10,20 Other studies, including a meta-analysis, reported inconsistent associations of NO 2 with birth weight. 14,39 We found an association of NO 2 exposure during the sensitive window from 9 to 14 gestational weeks with lower birth weight in the overall cohort that was strengthened among the subgroup with high perceived stress but attenuated toward the null in the low perceived stress subgroup. These findings together suggest a critical role for stressors at individual and neighborhood levels in determining the associations between sensitive windows of various air pollutants and birth weight.
Previous studies aiming to identify associations between sensitive windows of PM 2.5 exposure and birth weight have identified inconsistent sensitive windows ranging from early to late pregnancy. [9][10][11] In the present study, our identification of early pregnancy to midpregnancy as a sensitive window is consistent with other studies. [8][9][10]14,15 Registry-based studies based in China 12,13 and Texas 21 have also used DLMs to identify sensitive windows in finer periods (eg, by week or month), and generally have found the exposure window of 20 to 35 gestational weeks to be

High PSS
Associations among 411 participants with low PSS and associations among 132 participants with high PSS scores. Associations were adjusted for weekly temperature, maternal age, educational level, prepregnancy body mass index, ethnicity and nativity, educational level, cohabitation status, birth order, and enrollment time point. Outcome estimation was based on per IQR increases in each air pollutant: particulate matter with aerodynamic diameter less than 2.5 μm (PM 2.5 ), 4 μg/m 3 ; particulate matter with aerodynamic diameter less than 10 μm (PM 10 ), 12 μg/m 3 ; nitrogen dioxide (NO 2 ), 11 ppb; and ozone (O 3 ), 15 ppb.    High PSS, High CES Sample sizes in subgroups were 150 for low PSS and CES scores, 260 for low PSS scores and high CES scores, 49 for high PSS scores and low CES scores, and 82 for high PSS and CES scores. Associations were adjusted for weekly temperature, maternal age, educational level, prepregnancy body mass index, ethnicity and nativity, marital status, and birth order. Effect size estimation was based on per IQR increases in each air pollutant: particulate matter with aerodynamic diameter less than 2.5 μm (PM important, which partly overlaps with our findings. Early pregnancy to midpregnancy is the critical period of organogenesis and functional initiation that lays the foundations for fetal growth, such as the beginning of hematopoiesis, brown fat formation, and thyroid hormone secretion. 16 Therefore, exposure to air pollution during early pregnancy could affect fetal and placental development and eventually lead to lower birth weight, particularly for mothers with high psychosocial stress.

JAMA Network Open | Environmental Health
Both individual-level psychosocial stress and neighborhood-level cumulative burden could amplify the inverse association of PM 2.5 , PM 10 , and NO 2 exposure with birth weight. Our findings are consistent with studies that found neighborhood-level socioeconomic deprivation is associated with prenatal exposure to air pollution and offspring health outcomes, including birth weight. 25,26,40,41 Among a handful of studies that have measured individual-level psychological stress, susceptibility to asthma, poor lung function, or accelerated biological aging associated with air pollution exposures was greater in a higher stress group than in the lower stress group. 27,28 To our knowledge, our study is the first to report that individual-level maternal perceived stress may increase susceptibility to changes in birth weight associated with exposure to PM 2.5 , PM 10

Strengths and Limitations
A strength of our study is the use of DLMs to identify sensitive windows of exposure and the use of individual-and neighborhood-level stressors to investigate outcome modification.
The study has limitations, including lack of information on the source and constituents of particulate matter. Neighborhood-burden CES cumulative impact score is only available in California, but similar measures, such as the Environmental Justice Screening, are available for all US states. 46 Sample size was reduced in subgroup analyses, especially in the case of PSS missingness (13.5%). We excluded preterm births owing to the missingness of exposure before 37 weeks, which could potentially introduce collider bias. We attempted to evaluate such bias by comparing exposure and covariates between preterm and full-term infants (eTable 5 in the Supplement) and by controlling for known covariates associated with both preterm status and birth weight. Although we cannot rule out the possibility of collider bias, we believe such bias would be nonsignificant. Our analyses involve many statistical tests; thus, some of the associations, especially those not consistently seen in subgroup analyses, might be identified by chance.

Conclusions
In this cohort study, ambient PM 2.5 , PM 10 , and NO 2 during sensitive windows primarily in early pregnancy to midpregnancy were associated with lower birth weight. Wider sensitive windows and greater changes were seen in children born to mothers with high perceived stress or who lived in a health-disparity neighborhood with a high burden from environmental pollution and population vulnerability.