What this study adds
Using a vital statistics cohort (n = 279,051 births), our study examined associations between adverse birth outcomes and hazardous air pollution (HAP) exposures in Portland, OR. Few studies have addressed the effects of HAP exposures during pregnancy, a major shortcoming to the literature that we address in a city with high HAP concentrations and extensive community concern. We documented exposure gradients by sociodemographic characteristics, highlighting a potential environmental inequity. However, we did not observe consistent associations between adverse birth outcomes and cumulative HAP exposures after accounting for sociodemographic characteristics. Some HAP-specific models (e.g., butadiene, cadmium, chromium) demonstrated elevated risks that should be investigated in more depth.
Currently, the United States Environmental Protection Agency (EPA) regulates emission of 187 specific hazardous air pollutants (HAPs) via the Clean Air Act.1 Unlike criteria air pollutants, HAPs are defined as air toxics with known or suspected serious health effects, usually focusing on cancer outcomes.1 However, 105 of these 187 HAPs are known to be associated with health effects other than cancer, including adverse birth outcomes.2 Many areas, such as Portland, OR, attain criteria air pollutant standards but fail to maintain HAP levels that are compliant.3,4 There is no exposure threshold for HAPs that is considered safe for human health per the EPA,1 but there is a dearth of epidemiological research linking HAP exposure to birth outcomes.
Most of the evidence for an association between HAP and adverse pregnancy outcomes come from studies of traffic-related air pollutants. For example, a meta-analysis of 62 studies observed a decrease in term birth weight of 28.1 g (95% confidence interval [CI]: −44.8, −11.5) per 20 ppb increase in nitrogen dioxide (NO2).5 NO2 is a surrogate for the traffic-related air pollution mixture that contains numerous HAPs, such as diesel particulate matter (DPM), polycyclic aromatic hydrocarbons (PAHs), and benzene, toluene, ethylbenzene, and xylene (BTEX).6,7 While there are fewer studies specifically examining nonvehicle air toxic exposures, some studies have observed associations between adverse birth outcomes and nonspecific ambient HAP exposures to PAHs, chromium, and nickel8–10 as well as proximity to industrial sources such as oil and gas development,11–14 coal power plants,15–17 mining activity,18,19 and metal smelters.20 Overall, there are limited studies examining exposure to specific HAPs during pregnancy, especially HAP mixtures from multiple sources.5,21,22
In this study, we examine associations between 19 HAP exposures and adverse birth outcomes (term birth weight, preterm birth, and small for gestational age) for 279,051 births in the Portland Metropolitan Region from 2000 to 2014. Portland, OR, has consistently exceeded hazardous air pollutant (HAP) regulations for decades. Although some studies have addressed community health concerns regarding these emissions,4,23 the potential negative impacts on adverse birth outcomes is a major concern that has yet to be addressed. This analysis builds on a previous study to assess HAP concentrations in Portland and links these HAPs models to birth outcomes using a vital statistics cohort.24 The results of this study will inform the potential impacts of different air toxics and mixtures to adverse birth outcomes, address community concerns, and highlight future research needs.
We obtained vital statistics information from the Oregon Health Authority Center for Health Statistics for all births in the Portland Metropolitan Region with corresponding mothers residential addresses from 2000 to 2014 (n = 314,988 births). Sociodemographic characteristics for each mother, father, and infant were captured from birth certificate records. This analysis was restricted to births with a maternal residence within the city limits of Portland, OR (n = 289,651). We removed implausible observations based on maternal ages (<10 and >65 years old; n = 13), gestational ages (<12 and >45 weeks; n = 125), and birth weights (<100 and >6,500 g; n = 34). In addition, after considering our previous criteria, we removed stillbirths (n = 384) and multiple births (n = 10,044). Our final sample consists of 279,051 live singleton births in the Portland Metropolitan Region.
Air toxics exposure data
The Oregon Department of Environmental Quality (DEQ) created the Portland Air Toxics Solutions (PATS) dispersion model to better understand air toxic sources and concentrations in Portland. DEQ selected the 2005 inventory year as the base year for this project because it has the most air toxics monitoring data and good emissions inventory data.4 The 2005 inventory estimates emissions based on the amount of specific air toxic generating activity occurring and the air toxics emission rate for that activity. The main emission inventory components included emission factors, activity data, and spatial allocation. Categories of emissions include nonpoint (area) sources, mobile road sources, mobile nonroad sources, and permitted point sources. A full description of the emission inventories has previously been published.4
Briefly, all emissions were spatially allocated within the PATS region for input into a CALPUFF dispersion model. Background concentrations were added to all modeled values to account for transport of regional emissions from outside the PATS study area as well as natural and unidentified emission sources. These background estimates were developed by EPA for the 2002 National-scale Air Toxics Assessment. A 2017 forecast was produced by applying growth factors to the 2005 emissions and then subtracting any emissions controlled by federal and state air toxics regulations. DEQ refined and improved emissions data between the 2005 and 2017 estimates and we use the updated estimates as our primary exposure measures. The locations used to estimate toxic concentrations in the PATS project are the geographic centers, or centroids, of the year 2000 census block groups in the study area. The PATS model yields a single projected estimate of exposure that is spatially varying, but not temporally varying, across the Portland Metropolitan Region.
The PATS dispersion model estimates 19 specific pollutants: 1,3-butadiene, 1,4-para-dichlorobenzene, polycyclic aromatic hydrocarbon (PAH) composite estimate, acetaldehyde, acrolein, arsenic, benzene, cadmium, chromium VI, DPM, ethylbenzene, formaldehyde, lead, manganese, methylene chloride, naphthalene, nickel, perchloroethylene, and trichloroethylene.4 The PAH composite estimate takes into account 15 of the 16 PAHs identified by EPA for Toxic Release Inventory estimates excluding naphthalene, which is separate in the PATS model. For each maternal residence in our study we assign the corresponding PATS estimate for each of these pollutants as a log continuous estimate standardized by dividing the estimate by the interquartile range and as a categorical estimate by taking the geometric interval quintiles. The categorical estimates are derived from the spatial distribution of the PATS estimates across Portland, which allows us to examine specific hotspots of concern within the city that would otherwise be obscured.
In addition to examining each air toxic separately, we examined cumulative HAP exposures. We examined different cumulative metrics, including summing each air toxic, summing cancer potency factor weighted concentrations, as well as summing the highest exposure quintiles of each HAP. This set of composite metrics allows us to examine simultaneously spatial colocations of various HAPs. Because some areas of Portland have modeled concentrations that are statistical outliers, we leverage these multiple metrics to ensure that these hotspots and their potential health effects are not obscured in our models. We hypothesize that these composite metrics are a better representation of how HAPs exposure may affect infant health, so these composite exposures are the primary focus of our analysis.
Neighborhood socioeconomic status data
Neighborhood contextual characteristics were calculated for census tracts from the 2000 and 2010 census (whichever was closest to the infant’s birth year), including percent of households below the poverty line; percent of population that is nonwhite; race; and median household income. We did not include other measures of air pollution exposures (e.g., distance to major roads, known industrial point sources) as these were included in the PATS dispersion model.
We used information in the vital statistics database to determine the outcome for each infant using birth weight and gestational age fields. For term birth weight, we excluded all births with a gestational age less than 37 weeks and evaluated this outcome as a continuous variable. For preterm birth, we included all births with a gestational age less than 37 weeks and examined this outcome as a binary variable. For SGA, we calculated the 10th percentile for birth weight by gestational age and infant sex across our sample and assessed this outcome as a binary variable.
We examined associations between cumulative and individual PATS exposure measures with term birth weight, preterm birth, and SGA using linear and logistic models. For each exposure metric, we ran models with the exposure as a log continuous variable standardized by the IQR (to account for the skewed distributions of HAP levels) and as a categorical quintile metric by geometric intervals, comparing the highest to lowest quintiles. Fully adjusted models included infant sex, birth month, birth year, maternal and paternal age, maternal and paternal race, maternal and paternal education, insurance status, parity, maternal smoking and alcohol use during pregnancy, gestational and chronic hypertension, gestational and chronic diabetes, maternal weight gain during pregnancy, and Women, Infants, and Children (WIC) service eligibility. The neighborhood socioeconomic measures described above were also included in the fully adjusted models. We coded missing covariates in as a separate category for categorical variables and excluded missing data for continuous variables. Exploratory stratified analyses were conducted to examine potentially important modifying variables, including maternal education, maternal race, maternal ethnicity, delivery payment mechanism (i.e., insurance status), and neighborhood poverty levels.
Table 1 summarizes the demographic characteristics of all births within the Portland city boundary by cumulative air toxics (lowest and highest quintiles of the sum of individual air toxics weighted by toxicity). Compared with babies born in the lowest quintile of cumulative air toxics, babies in the highest quintile experienced lower mean birth weights, higher percentage of SGA, lower maternal education at birth; were more likely to be non-white, Hispanic, and use Medicaid or the Oregon Health Plan; were less likely to be exposed to maternal tobacco smoke in utero, be eligible for WIC, and have gestational diabetes and hypertension; and had lower median household incomes in their neighborhood, higher percentage of people below the poverty line, and higher percentage of nonwhite population in their neighborhood.
Hazardous air pollution exposures
Figure 1 shows the spatial distribution of the composite risk metric as well as individual patterns for select HAP exposures in Portland, OR. Other HAP spatial patterns are included in eFigure 1 (https://links.lww.com/EE/A25). Individual HAPs demonstrate substantial spatial heterogeneity due to differences in major source contributions (e.g., traffic, industry, home heating). eTable 1 (https://links.lww.com/EE/A25) demonstrates the distribution of the individual HAP exposures.
We first examined multiple metrics of cumulative HAPs that were derived from the PATS models (Table 2). These metrics include sum of the HAPs concentrations, sum of HAPs risk weight by cancer toxicity, and sum of the HAPs exposure quintiles, all of which we examined in log continuous and quintile frameworks. Across all metrics, we found unadjusted results that demonstrate an increased risk of adverse birth outcomes, but these risks are fully attenuated after we adjust our models for confounding sociodemographic characteristics. For example, in the composite sum of the HAPs concentrations metric, we find a 36.99 (95% CI: −42.66, −31.32) decrease in unadjusted models, but this association trends toward null in adjusted models (−4.14; 95% CI: −9.64, 1.37). Similar attenuation was observed for the other composite metrics and birth outcomes.
Next, we investigated associations between each of the 19 hazardous air pollutants and birth outcomes. Table 3 summarizes fully adjusted models for the log of an IQR increase and the highest versus the lowest exposure quintiles. In unadjusted models, we observed elevated risks of all three adverse pregnancy outcomes (eTable 2; https://links.lww.com/EE/A25), but most of these associations were attenuated in the fully adjusted models (Table 3), similar to our cumulative HAP measures. In our adjusted models comparing the highest to the lowest quintile of exposures, we observed elevated risks for term birth weight and 1,3-butadiene (−16.86; 95% CI: −29.66, −4.06), cadmium (−31.37; 95% CI: −56.20, −6.54), and chromium VI (−12.70; 95% CI: −23.10, −2.31); and SGA and 1,3-butadiene (1.18; 95% CI: 1.07, 1.30). In our fully adjusted log IQR models, we observed elevated risks for term birth weight and manganese (−4.51; 95% CI: −6.88, −2.13), methylene chloride (−1.99; 95% CI: −3.93, −0.05), and trichloroethylene (−4.24; 95% CI: −8.02, −0.45); preterm birth and methylene chloride (1.02; 95% CI: 1.00, 1.04); and SGA and ethylbenzene (1.03; 95% CI: 1.00, 1.06). We also observed some protective associations for specific HAPs, including for PAH and term birth weight as well as acetaldehyde, benzene, diesel particulate matter, and ethylbenzene and preterm birth.
To explore what covariates may be influencing the strong risk attenuations in our models, we implemented incremental models and note that the main decreases in point estimates are from adding socioeconomic, behavioral, clinical, and neighborhood covariates (Table 4). Covariate selection for the incremental models followed a logical progression of covariates determined a priori that we hypothesize would influence the results from most to least. We also considered what covariates have been commonly used in the existing literature on environmental exposures and birth outcomes.
Exploratory stratified models were conducted to determine if associations between HAP exposures and adverse birth outcomes were similar across sociodemographic and neighborhood characteristics using the log sum of continuous HAP concentrations. Within racial and ethnic groups, we observed some large differences for minority populations in terms of HAPs exposures (eTable 3; https://links.lww.com/EE/A25). For example, black mothers are exposed to more methylene chloride than other groups and infants born to Hispanic mothers appear to have higher exposures to 1,3-butadiene, acetaldehyde, benzene, DPM, ethylbenzene, formaldehyde, and methylene chloride. However, in our stratified models (eTable 4; https://links.lww.com/EE/A25), we observed no consistent pattern of associations between our cumulative HAP metrics and adverse birth outcomes for specific populations.
The Environmental Protection Agency (EPA) National Air Toxics Assessment (NATA) shows that Portland’s air exceeds regulatory levels of at least 66 HAPs, 49 of which are known carcinogens.3 There is tremendous community concern regarding past and current air toxic exposure levels and their potential health impacts, especially since a 2016 study using tree moss identified previously unknown hot spots of metals in Portland.25 We examined associations between HAP concentrations and adverse birth outcomes using a population-based birth cohort in Portland, OR, between 2000 and 2014. To accomplish this analysis, we took advantage of a unique city-wide HAP dispersion model that allows us to estimate multiple HAP exposures. Our results show that composite HAPs exposures, and most individual HAP exposures, were not associated with adverse birth outcomes. However, some specific air toxic exposures warrant further research, including 1,3-butadiene, cadmium, and chromium VI.
We did not observe associations between our metrics of combined HAP exposures and adverse birth outcomes, but we note that many of our estimates lack sufficient precision to rule out a small yet meaningful impact of infant health. We used three different methods to capture potential HAP mixtures, including summing individual HAP concentrations, summing cancer toxicity weighted measures, and summing the quintiles of each HAP concentration. Each toxicity metric allows us to capture different aspects of exposure mixtures that may be affecting perinatal health. By summing the individual HAP concentrations, we examined whether the total amount of HAP is a driving pathway toward adverse pregnancy outcomes. In our cancer toxicity-weighted measures, we included the differing toxicological impacts of each HAP (we did not use a developmental toxicity weighting as several of the HAPs do not have weights established22). Finally, the sum of each HAP quintile focused on capturing spatial hot-spots for each HAP, rather than absolute concentrations. For all three measures, we did not observe consistent associations with adverse pregnancy outcomes in fully adjusted models. Within our composite toxicity measures, we examined what industry may be driving the highest exposures via the Toxic Release Inventory and observed a very high number of emission sources located in a small area emitting a wide range of pollutants, including chromium, lead, naphthalene, nickel, toluene, xylene, and zinc (eFigure 1; https://links.lww.com/EE/A25).26 To fully understand the health impacts of local industrial emissions, future studies may need to incorporate mixture modeling techniques such as structural equation models to assess the distinct impacts of simultaneous exposure to multiple HAPs during pregnancy.
We observe some positive associations for individual HAP exposures, but the majority of our models indicate no associations. We present a graphical version of our results in Figure 2, where we demonstrate the relationships between our point estimates and their two-sided P values. Labeled HAPs reflect exposures that are statistically significant for at least one outcome. Although we find many point estimates suggestive of an association between HAP exposure and adverse infant health outcomes, Figure 2 shows that many of these associations do not demonstrate statistical significance and no clear pattern between HAP exposures and adverse birth outcomes emerge.
There is little existing literature for direct comparison to our results, however, some of our specific HAPs have been examined in other studies. One previous study27 shows that increases in maternal blood manganese were associated with decreased birth weight, which was also observed in our study (4.51 g [95% CI: −6.88, −2.13] decrease per log IQR increase in manganese exposure). Another study shows that each μg/m3 unit increase in maternal benzene exposure is associated with a birth weight decrease of 16.5 g (95% CI: 17.6, 15.4).28 Although we did not find associations among benzene and birth weight, we do note an elevated risk of SGA among mothers with higher benzene exposure. In addition, a meta-analysis of traffic-related air pollution studies, using NO2 as a surrogate marker, observed a decrease in term birth weight of 28.1 g (95% CI: −44.8, −11.5) per 20 ppb increase in nitrogen dioxide (NO2).5 We did not observe associations with DPM, although the resolution of the PATS model does not capture fine-scale spatial variability of air toxics associated with the highest roadway gradients. We also observed protective associations with composite PAH exposure, which has been shown in a previous study.29 We do not know what biological explanation may exist for these inverse results, though both results could be due to residual confounding.
For most HAPs, we observed consistently higher HAP exposures for lower socioeconomic and minority groups, and large unadjusted associations between HAPs and adverse birth outcomes, but no consistent increases in adverse birth outcomes in fully adjusted models. This attenuation was due primarily to the inclusion of clinical and neighborhood characteristics, as shown in the incremental models illustrated in Table 4. The large disparities in HAP exposures between ethnicity status highlight the well-recognized patterns of higher air toxic exposures for minority populations, but the distribution of HAPs appears to be less unequal than in other urban populations (eTable 3; https://links.lww.com/EE/A25).2,30,31 In models restricted to these populations, we did not observe consistent adverse impacts of HAP exposures on adverse birth outcomes. Some of the attenuation observed in our adjusted models may also be due to the rapid gentrification of the Portland Metropolitan Region, which has been extensively documented.32–35 Since 2000, 25.4% of census tracts in Portland have met criteria for undergoing gentrification.36 These swift sociodemographic neighborhood changes may be altering longer-term exposure patterns as well as how traditional risk factors are affecting adverse pregnancy outcomes.
Several limitations of our study should be considered when interpreting our results. First, we rely on surrogate measures of HAP exposure during pregnancy that are based on models applied to mother’s residential address at time of delivery. The PATS CALPUFF dispersion model is based on available emissions data in 2005 (and subsequently updated) and model accuracy is driven by the quality of these emissions data. In fact, the PATS estimates do not account for some previously unknown emissions sources, such as two artisan glass factories.37 The PATS model also represents a snapshot of HAP in Portland and does not produce monthly or yearly estimates to match pregnancy periods. In addition, the PATS model does not capture fine-scale pollution gradients (e.g., 100s of meters around roadways or industrial sources) that may be important for HAP exposures. Nevertheless, the PATS model represents a unique resource to simultaneously examine multiple HAP exposures from different sources, with many of these sources being spatially stable over time. Second, we only had maternal address at time of delivery, and we do not necessarily know where the mother resided during pregnancy if they moved. Third, because we are exploring broad categories of HAPs and adverse birth outcomes, we ran over 200 models in our analyses, which puts us at a high probability of some results being due to chance instead of a true association. We interpret our models with caution due to a multiple comparisons issue, but we note that our conclusions are drawn from overall patterns of results and not individual associations. Fourth, many of our risk estimates yield wide confidence intervals that lack precision, thus we cannot dismiss the notion that some of our maternal exposures may have a meaningful, albeit small, impact on infant health. Fifth, residual confounding from unmeasured confounding factors cannot be ruled out in this analysis, similar to other birth cohort studies. However, we included a wide range of individual and geographic covariates to account for as much unmeasured variation as possible. We also control for secular trends in birth outcomes via birth year and month covariates in all models.
Our analysis provides a broad examination of 19 HAPs and adverse birth outcomes in a large population-based birth cohort in Portland, OR, a city with consistently high HAP concentrations that result from a diverse set of emission sources. Our results show that cumulative HAPs exposures, and most individuals HAP exposures, were not associated with adverse birth outcomes, while some specific air toxic exposures warrant further research, including 1,3-butadiene, cadmium, and chromium VI.
Conflicts of interest statement
The authors declare that they have no conflicts of interest with regard to the content of this report.
We thank Oregon Vital Statistics for providing the birth outcome data and the Oregon Department of Environmental Quality for providing the Portland Air Toxics Study (PATS) modeling data.
1. US EPAWhat are Hazardous Air Pollutants
?US EPAAvailable at: https://www.epa.gov/haps/what-are-hazardous-air-pollutants
. Published December 3, 2015. Accessed 21 April 2018
2. US EPAEnvironments & Contaminants: Hazardous Air Pollutants2016Washington, D.CUnited States Environmental Protection Agency50–57
3. Koberstein P. Study: Portland Air Among Worst in Nation.The Portland TribuneAvailable at: http://koin.com/2016/03/02/study-portland-air-among-worst-in-nation/
. Published March 2, 2016. Accessed August 29, 2017
4. Armitage S. Portland Air Toxics
Solutions Committee Report and Recommendations2012Portland, ORState of Oregon Department of Environmental Quality
5. Stieb DM, Chen L, Eshoul M, Judek S. Ambient air pollution, birth weight
and preterm birth
: a systematic review and meta-analysis.Environ Res2012117suppl C100–111
6. Kampa M, Castanas E. Human health effects of air pollution.Environ Pollut2008151362–367
7. Richmond-Bryant J, Owen RC, Graham S, et al. Estimation of on-road NO2
ratios, and related roadway gradients from near-road monitoring data.Air Qual Atmosphere Health201710611–625
8. Choi H, Jedrychowski W, Spengler J, et al. International studies of prenatal exposure to polycyclic aromatic hydrocarbons and fetal growth.Environ Health Perspect20061141744–1750
9. Laurent O, Hu J, Li L, et al. Sources and contents of air pollution affecting term low birth weight
in Los Angeles County, California, 2001–2008.Environ Res2014134488–495
10. McDermott S, Salzberg DC, Anderson AP, Shaw T, Lead J. Systematic review of chromium
and nickel exposure during pregnancy and impact on child outcomes.J Toxicol Environ Health A2015781348–1368
11. McKenzie LM, Guo R, Witter RZ, Savitz DA, Newman LS, Adgate JL. Birth outcomes and maternal residential proximity to natural gas development in rural Colorado.Environ Health Perspect2014122:412–417
12. Casey JA, Savitz DA, Rasmussen SG, et al. Unconventional natural gas development and birth outcomes in Pennsylvania, USA.Epidemiology201627163–172
13. Whitworth K, Marshall A, Symanski E. Maternal residential proximity to unconventional gas development and perinatal outcomes among a diverse urban population in Texas.PLoS One201712e0180966
14. Hill EL. Shale gas development and infant health: evidence from Pennsylvania.J Health Econ201861134–150
15. Ha S, Hu H, Roth J, Kan H, Xu X. Associations between residential proximity to power plants and adverse birth outcomes
.Am J Epidemiol2015182215–224
16. Yang M, Bhatta RA, Chou S-Y, Hsieh C-I. The impact of prenatal exposure to power plant emissions on birth weight
: evidence from a Pennsylvania power plant located upwind of New Jersey.J Policy Anal Manag J Assoc Public Policy Anal Manag201736557–583
17. Casey JA, Karasek D, Ogburn EL, et al. Retirements of coal and oil power plants in California: association with reduced preterm birth
among populations nearby.Am J Epidemiol20181871586–1594
18. Ahern MM, Hendryx M, Conley J, Fedorko E, Ducatman A, Zullig KJ. The association between mountaintop mining and birth defects among live births in central Appalachia, 1996–2003.Environ Res2011111838–846
19. Ahern M, Mullett M, Mackay K, Hamilton C. Residence in coal-mining areas and low-birth-weight outcomes.Matern Child Health J201115974–979
20. Berkowitz Z, Price-Green P, Bove FJ, Kaye WE. Lead exposure and birth outcomes in five communities in Shoshone County, Idaho.Int J Hyg Environ Health2006209123–132
21. Ferguson KK, Chin HB. Environmental chemicals and preterm birth
: biological mechanisms and the state of the science.Curr Epidemiol Rep2017456–71
22. Backes CH, Nelin T, Gorr MW, Wold LE. Early life exposure to air pollution: how bad is it?Toxicol Lett201321647–53
23. Oregon Health Authority : New Soil, Cancer, Urine Test Data Show Low Risk for Portland Residents : External Relations Division : State of OregonAvailable at: http://www.oregon.gov/oha/ERD/Pages/New-Soil-Cancer-Urine-Data-Shows-Low-Risk.aspx
. Accessed March 2, 2018
24. Tam BN, Neumann CM. A human health assessment of hazardous air pollutants
in Portland, OR.J Environ Manage200473131–145
25. Donovan G, Jovan S, Gatziolis D, Burstyn I, Michael Y, Monleon V. Using an epiphytic moss to identify previously unknown sources of atmospheric cadmium
pollution.Sci Total Environ55984–93
26. US EPATRI Data and Tools.US EPAAvailable at: https://www.epa.gov/toxics-release-inventory-tri-program/tri-data-and-tools
. Published March 3, 2013. Accessed May 24, 2018
27. Zota AR, Ettinger AS, Bouchard M, et al. Maternal blood manganese levels and infant birth weight
28. Zahran S, Weiler S, Mielke HW, Pena AA. Maternal benzene exposure and low birth weight
risk in the United States: a natural experiment in gasoline reformulation.Environ Res2012112139–146
29. Padula AM, Noth EM, Hammond SK, et al. Exposure to airborne polycyclic aromatic hydrocarbons during pregnancy and risk of preterm birth
30. Stewart JA, Mitchell MA, Edgerton VS, VanCott R. Environmental justice and health effects of urban air pollution.J Natl Med Assoc201510750–58
31. Vinikoor-Imler LC, Davis JA, Meyer RE, Messer LC, Luben TJ. Associations between prenatal exposure to air pollution, small for gestational age
, and term low birthweight in a state-wide birth cohort.Environ Res2014132132–139
32. Breyer B, Voss-Andreae A. Food mirages: geographic and economic barriers to healthful food access in Portland, Oregon.Health Place201324131–139
33. Goodling E, Green J, McClintock N. Uneven development of the sustainable city: shifting capital in Portland, Oregon.Urban Stud Plan Fac Publ Present201536504–527
34. Hagerman C. Shaping neighborhoods and nature: urban political ecologies of urban waterfront transformations in Portland, Oregon.Cities200724285–297
35. Monroe Sullivan D, Shaw SC. Retail gentrification and race: the case of Alberta Street in Portland, Oregon.Urban Aff Rev201147413–432
37. Allen P. Bullseye Glass Monitoring Report.Available at: https://www.oregon.gov/deq/filterdocs/bullseyemodelingreview.pdf
Published June 2017. Accessed September 2017