Research report (Health Effects Institute)

INTRODUCTION:Traffic emissions comprise a complex mixture of pollutants, including tailpipe (exhaust), nontailpipe (e.g., brake and tire wear, resuspended road dust), and noise pollution. While regulations have drastically reduced tailpipe emissions, the growing vehicle fleet and miles traveled have contributed to rising levels of nontailpipe and noise emissions. These under-regulated exposures, often rich in toxic transition metals, may pose significant health risks, particularly when noise acts as a co-exposure that may influence traffic-related effects. METHODS:We examined lung function effects associated with nontailpipe exposures in the Children's Health Study (CHS), a cohort of over 1,200 children from eight Southern California communities. Participants underwent lung function testing for forced expiratory volume in 1 second (FEV1) and forced vital capacity (FVC) at three time points between 2008 and 2012. Participant-specific estimates of particulate matter (PM) mass and metal exposures were developed by applying spatiotemporal regression models to measurements of quasi-ultrafine (PM0.2, aerodynamic diameter ≤0.2 µm), fine (PM0.2-2.5), and coarse (PM2.5-10) particles collected at 220 locations across two seasons. We used supervised variable selection of over 150 variables, including meteorology, roadway and traffic characteristics, land use, and dispersion-modeled estimates of tailpipe emissions. Particular attention was paid to PM components with high oxidative potential or those serving as markers for nontailpipe sources. Traffic noise exposures were estimated at a 20-m resolution along roadways using two approaches: (1) mobile measurements of A-weighted equivalent noise levels (LAeq in decibels [dB]) collected in Long Beach and modeled using machine learning; and (2) CadnaA, an acoustic model applied to generate LAeq across all communities. We used mixed effects models to evaluate associations between exposures and lung function, adjusting for covariates and accounting for repeated measures. RESULTS:Spatiotemporal PM models showed strong performance (leave-one-community-out cross-validation coefficient of determination [R2] 0.59-0.82; 10-fold cross-validation R2 0.76-0.92). Incorporating intersection-based hotspots and meteorology substantially improved accuracy. Noise models using extreme gradient boosting (XGB) achieved high accuracy (leave-one-route-out cross-validation R2 = 0.71, root mean square error [RMSE] 4.54 dB; fivefold cross-validation R2 = 0.96, RMSE 1.8 dB), with traffic volume, road proximity, and meteorology as top predictors. LAeq estimates from XGB performed well on smaller roads, while CadnaA performed better on highways. Both approaches captured fine-scale spatial variation in traffic-related noise. Epidemiologically, coarse and fine nontailpipe metals, especially iron (Fe) and copper (Cu) as markers of road dust and brake wear, were consistently and significantly associated with reduced FEV1 and FVC. These associations persisted after adjusting for PM mass, and were stronger than those of tailpipe markers like fine elemental carbon (EC2.5, diesel) and quasi-ultrafine organic carbon (OC0.2, gasoline). Traffic noise was not independently associated with lung function but strongly confounded the OC0.2-FEV1 association (noise-adjusted FEV1 was 33% larger in magnitude), suggesting co-occurrence or shared spatial patterns between noise and tailpipe emissions. In contrast, green space was directly associated with better lung function and acted as a confounder, attenuating nontailpipe PM component associations while strengthening that of OC0.2. CONCLUSIONS:This study highlights the lung function impacts of PM components, the role of traffic noise, and the influence of green space. From an exposure modeling standpoint, a major takeaway is the importance of meteorology and intersection features for PM components, and meteorology, traffic volume, and land use for traffic noise. Nontailpipe metals had the strongest and most consistent associations with reduced lung function, likely via oxidative stress pathways. Noise influenced the associations of tailpipe markers, reflecting a separate stress mechanism. Green space emerged as a protective factor, highlighting its potential to influence adverse associations between traffic-related air pollution and children's respiratory health. As nontailpipe sources grow in importance with increased electric vehicle adoption, mitigation strategies should target brake, tire, and road wear. Regulatory actions, lighter electric vehicle designs, and regenerative braking could help. Simultaneously, expanding green space offers a feasible, co-beneficial intervention to buffer both air and noise pollution and to promote children's respiratory health.

INTRODUCTION:China launched ambitious air pollution regulations in 2013 that have resulted in substantial reductions in concentrations of particulate matter with aerodynamic diameters of 2.5 micrometers or less (PM2.5). The main objective of this project is to analyze whether regulations to control PM2.5 have been associated with declining mortality rates, especially in locations where the regulations caused larger reductions in PM2.5 concentrations. In addressing this main objective, we control for ozone (O3) concentrations. In addition, we examine and analyze available PM2.5 components (that is, the subspecies that make up PM2.5) and related air pollutants in secondary analyses that focused on understanding the influence of specific source sectors and emission control policies on PM2.5 trends. METHODS:We used both observations and model outputs to characterize observed and counterfactual spatiotemporal trends in air quality from 2008 to 2019 across China. We characterized observed PM2.5 and O3 concentrations from 2008 to 2019 across China using published gridded datasets that incorporate observations along with remote sensing and other data to estimate surface concentrations that are spatially and temporally complete. We used the Community Multiscale Air Quality (CMAQ) model to both reproduce observed pollution and to simulate concentrations of PM2.5 and O3 that would have occurred absent regulations from 2008 to 2019. Additional CMAQ scenarios quantified the impacts of a range of source-specific emission control policies on ambient concentrations. In a secondary, supporting analysis, we analyzed observed speciated particulate matter data from three cities using source apportionment methods to infer changes in pollution source influences over time. We analyzed mortality data from two large, representative cohorts maintained by the Chinese Center for Disease Control and Prevention in relation to spatial and temporal variations in PM2.5 and O3 concentrations using two methods. We first analyzed the health data using a traditional Cox proportional hazards model to estimate conventional hazard ratios (HR) for PM2.5 risk. Next, we implemented novel causal methods based on principal stratification to analyze the extent to which regulatory policies implemented starting in 2013 were associated with reduced mortality across China. RESULTS:From 2013 to 2019, we estimate there was a ~45% reduction in PM2.5 as compared to a no-control scenario, but with considerable regional heterogeneity. Associations were observed between PM2.5 and mortality rates in all cohorts, though results were sensitive to model specification in a cohort of elderly subjects. Results of causal models suggested that improvements in PM2.5 since 2013 were associated with increased survival probability in both cohorts. CONCLUSIONS:The scope of air pollution regulations and resulting PM2.5 improvements in China since 2013 provided a unique opportunity for accountability research. Our study provides evidence supporting the health benefits of those policies.

INTRODUCTION:The main goal of this project was to develop the TRACER (TRAcking Community Exposures and Releases) model to assess exposures to air pollutants from unconventional oil and gas development (UOGD) and to inform future health studies. The project's main focus was on the Eagle Ford Shale in south-central Texas, a large oil and gas production region that includes the production of dry gas, wet gas, and oil. This heterogeneity of production types makes the Eagle Ford Shale a microcosm of UOGD sites throughout the United States. The project was later expanded to also include mobile measurements in the Permian Basin and modeling in the Marcellus Shale. Methods:We expanded a model originally designed to predict emissions of methane to pollutants of concern to human health. We coupled the expanded emissions model with dispersion models to evaluate the impacts of UOGD emissions on concentrations of pollutants at a receptor site and regionally, and we compared the performance of different dispersion models. We also coupled the emissions model with a chemical transport model to evaluate the impacts of UOGD emissions on ozone, a secondary pollutant. We conducted targeted stationary and mobile measurements to evaluate emissions from flaring, which were then included as inputs to the TRACER model, and to provide a comprehensive dataset for model evaluation. Finally, we evaluated community exposures in Karnes County, Texas, to pollutants of concern to human health emitted by UOGD, and differences in exposure by income and ethnicity/race. RESULTS:We found that destruction efficiencies and emission ratios from flares are highly variable. Ambient concentrations at receptor sites, and therefore population exposures, have high diurnal variability, with the highest concentrations and exposures observed at night. Elevated concentrations observed at night can be due to both large nonroutine emissions events and routine emissions, coupled with wind speeds and atmospheric stability conditions that are conducive to producing high concentrations. Ambient concentrations are affected by thousands of UOGD sources that are tens of kilometers from receptor sites, and thus, predicting concentrations at these sites requires high computational intensity. We evaluated and compared different dispersion models and found that the CALPUFF dispersion model (using stability classification to predict dispersion parameters) generally performs best but is also the most computationally expensive. Coupling the emissions model with the chemical transport model (Comprehensive Air Quality Model with extensions) revealed that realistic temporal and spatial allocation of nitrogen oxides emissions can result in higher predicted ozone concentrations. Reduced-complexity exposure approaches that include meteorology generally capture salient features (e.g., spatial-temporal variability) found in observations and in more complex models, but concentrations from these approaches have higher bias than CALPUFF. CONCLUSIONS:We show strengths and limitations of multiple methods to assess UOGD source-specific concentration enhancements and spatial-temporal exposure patterns. Combined with spatial differences among population group residences, spatial variability in emissions and dominant wind patterns leads to differences in exposure between racial or ethnic and income groups. All groups experience higher exposure at night than during the day.

INTRODUCTION:Obesity is a major public health concern because it increases the risk of numerous diseases, including cardiovascular disease and type 2 diabetes. Ambient and near-roadway air pollution has been associated with childhood obesity risk, independent of diet and physical activity. However, the biological mechanisms underlying these relationships remain unclear. Based on our previous work and existing literature, we hypothesized that exposure to air pollutants alters the developing infant gut microbiome and fecal metabolome, with implications for childhood obesity risk. In this study, we aimed to determine whether prenatal or early-life exposure to ambient air pollution and near-roadway air pollution is associated with the gut microbiome and fecal metabolome during the first 2 years of life. METHODS:Our analysis had two components, both of which examined participants from the Southern California Mother's Milk Study, a Latino cohort in which we collected detailed information regarding maternal and child health during the first 24 months of life. Residential-based estimates of exposure to ambient particulate matter (particulate matter ≤2.5 µm and ≤10 µm in aerodynamic diameter: PM2.5 and PM10, respectively), nitrogen dioxide (NO2), and ozone (O3), as well as near-roadway air pollution (NOx), were modeled using residential address histories. High-throughput metagenomics and metabolomics were performed on stool samples collected at 1, 6, 12, 18, and 24 months of age. Overall, our sample included 207 unique individuals with gut microbiome data and 127 unique individuals with fecal metabolomics data. In the first analysis component, we examined the cross-sectional associations of pre- and postnatal exposure to ambient and near-roadway pollutants with the infant gut microbiome and fecal metabolome at 1, 6, 12, 18, and 24 months of age. In the second analysis component, we examined the longitudinal associations of pre- and postnatal exposure to air pollutants with the trajectory of the developing infant gut microbiome and fecal metabolome. RESULTS:Our findings indicate that exposure to air pollutants during prenatal and postnatal periods is associated with significant changes in the developing gut microbiome and its metabolic output, as evidenced by perturbations in the fecal metabolome. These molecular alterations were evident in both cross-sectional and longitudinal analyses. The results suggest that early-life exposure to air pollution can disrupt the developmental trajectory of the gut microbiome, potentially leading to changes with substantial health implications. These findings underscore the importance of mitigating air pollution exposure during critical developmental periods to protect and promote gut health and overall well-being in infants. CONCLUSIONS:We identified gut microbes and fecal metabolites associated with early-life exposure to air pollution. Many of these markers of gut bacterial composition and function have been linked to childhood obesity. These findings contribute to our understanding of mechanisms underlying the obesogenic effects of air pollutants in early life. Future work in this cohort will include integrated mixture and multi-omics analyses to explore the joint impact of air pollution exposure on the gut microbiome and fecal metabolome.

INTRODUCTION:As of December 2023, more than 6.9 million people globally had died from COVID-19, including more than 1.165 million deaths in the United States. It is estimated that approximately 18.8 million people in the United States have experienced post-acute COVID-19 conditions, also known as post-acute sequelae of SARS-CoV-2 (PASC) or long COVID, in the first 3 years after the pandemic. Although some initial cases of long COVID have resolved, with the ongoing incidence of COVID-19, roughly 17.8 million persons in the United States continue to suffer from long COVID at the time of this writing.1-3 Preliminary evidence early in the COVID-19 pandemic suggested that exposure to air pollution increased the likelihood of contracting COVID-19 and worsened outcomes for those who became ill. The validity of these findings was uncertain, however, as few studies used highly accurate exposure models incorporating individual-level data on patient characteristics and risk factors. Although the COVID-19 public health emergency has ended, the disease continues to pose substantial risks to individual and population health. At the time of this writing, nearly 35,000 individuals per week are hospitalized with COVID-19 in the United States, and the weekly number of COVID-19-related deaths ranges from 900 to 1,400.4. METHODS:In this study, we investigated relationships between ambient air pollution and COVID-19-related outcomes, including incidence, severity, mortality, and long COVID conditions. We used advanced models to estimate exposures, incorporating numerous air pollutants, particle species, and wildfire emissions. We used administrative COVID-19 data and several cohorts of patients from a large health system, and each was formed to evaluate different hypotheses. Daily air pollution exposures for Southern California were estimated with high spatial and chemical resolution, using a combination of land use regression and chemical transport models for the years 2016, 2019, and 2020. Exposure variables included ozone (O3), nitrogen dioxide (NO2), fine particulate matter (PM) ≤2.5 μm in aerodynamic diameter (PM2.5mass), ultrafine PM ≤0.1 μm in aerodynamic diameter (PM0.1), and major sources or chemical components of PM in each size fraction. Exposures for multiple study populations were investigated using statistical analysis methods to test for associations with COVID-19-related outcomes, including the following. • COVID-19 cases (N = 773,374) and deaths (N = 14,311), by age, race, and sex, for 308 ZIP codes in Los Angeles County between June 19 and January 3, 2021. A negative binomial regression was performed for both individual and multiple ambient air pollutants to evaluate their associations with COVID-19 incidence and mortality. • Patients with COVID-19 who were admitted to Kaiser Permanente Southern California (KPSC) hospitals between June 1, 2020, and January 30, 2021 (N = 21,415). Cox proportional hazards models were used to evaluate associations between ambient air pollutant exposure and COVID-19 mortality. A subset was of KPSC patients with COVID-19 who received care exclusively in KPSC hospitals (N = 15,978). A multistate survival model was used to examine how air pollution affects the transition to recovery or deterioration to more severe COVID-19 states (e.g., intensive care admission or death). A subset was of KPSC patients with COVID-19 who maintained membership with KPSC for 1 year after hospital discharge (N = 12,634). We combined a set of 45 diagnoses of post-acute sequelae of SARS-CoV-2 (PASC) into categories based on organ systems and then studied a subset of these PASC categories that could be affected by air pollution, including cardiac, cardiometabolic, pulmonary, and neurological conditions. Logistic regression was used to evaluate associations between 30-day air pollution exposure before hospital admission and PASC conditions diagnosed at 3 months and 12 months post-discharge. RESULTS:PM0.1, O3, NO2, and PM2.5 elemental carbon exposures were identified as risk factors for COVID-19 incidence and mortality in the general population of Los Angeles County. Air pollution exposures were also significantly associated with COVID-19 mortality in the cohort of hospitalized KPSC patients, controlling for other individual health risks. Incremental increases equivalent to the interquartile range for several pollution exposure concentrations were significantly associated with increased mortality, including PM2.5 mass (hazard ratio [HR], 1.12), PM0.1(HR, 1.06), PM2.5 nitrate (HR, 1.12), PM2.5 elemental carbon (HR, 1.07), PM2.5 on-road diesel (HR, 1.06), and PM2.5 on-road gasoline (HR, 1.07). Humidity and temperature in the month of diagnosis were significant negative predictors of COVID-19 mortality and negative modifiers of the air pollution effects. Results of the multistate analysis were consistent with these findings and further suggested that O3, NO2, and PM2.5 each were associated with deteriorating health states. Increased PM2.5 concentration was associated with increased risk of deterioration to both intensive care admission (HR, 1.16) and death (HR = 1.11). Effects of O3 were similar to those of PM2.5, but O3 also affected the transition from recovery to death (HR, 1.24). Several air pollutants - particularly O3, PM0.1, and PM2.5 nitrate - were significantly associated with several long COVID outcomes, including cardiac, cardiometabolic, and pulmonary conditions. CONCLUSIONS:Broadly, we concluded that several common air pollutants are associated with COVID-19 incidence, mortality, and progression to more severe states of illness, including long COVID conditions. Air pollution is a modifiable environmental risk factor that could be altered to improve the prognosis of COVID-19, thereby also reducing the public health impacts of coronaviruses now and in the future. This is particularly important for preventing long COVID, as evidence suggests that PASC conditions can occur even in vaccinated individuals. Given that 10% to 30% of individuals with COVID-19 will experience some form of PASC, which can have lifelong debilitating effects,5 the importance of addressing modifiable environmental risk factors, such as air pollution, cannot be underestimated. A recent Lancet editorial noted that societal investment in understanding the pathogenesis of long COVID and preventive measures has lagged well behind the levels needed to effectively treat and mitigate this complex disease.6 Our research focused mostly on hospitalized patients, but it also included one study on the general population effects. The results of both analyses were generally concordant, although our most important findings likely apply only to patients hospitalized with COVID-19.

INTRODUCTION:FRONTIER aimed to provide a robust and comprehensive evaluation of the impact of maternal exposure to traffic-related air pollution (TRAP) on fetal growth. Toward this aim it (1) disentangled the effects of noise as a co-exposure; (2) identified the relevant window(s) of vulnerability for this impact; (3) evaluated its modification by household and neighborhood level socioeconomic status (SES), stress, physical activity, and the timing of conception and delivery in relation to the COVID-19 pandemic lockdown; (4) elucidated the role of placental function as an underlying mechanism; and (5) explored the potential of urban tree canopies and green spaces to mitigate it. METHODS:FRONTIER established a new pregnancy cohort of 1,080 pregnant women in Barcelona, Spain - Barcelona Life Study Cohort (BiSC). Fetal growth was characterized by anthropometric measures at birth, together with ultrasound-based trajectories of fetal development. We developed an innovative exposure assessment framework integrating objective data on time-activity patterns with dispersion, land use regression, and hybrid models, with campaigns of personal and home-outdoor air pollution monitoring to estimate maternal exposure levels as well as inhaled dose of black carbon (BC), nitrogen dioxide (NO2), fine particulate matter (PM2.5), and PM2.5 copper, iron, and zinc in the main microenvironments for pregnant women (home, workplace, and commuting routes). We also assessed maternal exposure to noise by integrating measurements at participants' homes and outdoors using noise monitors with modeled microenvironmental noise levels, data on noise sensitivity, annoyance, and protections against noise. We developed single- and multipollutant models to evaluate the impact of TRAP exposure and inhaled dose on fetal growth while also correcting for the exposure measurement error. We further evaluated the modification of associations by SES, stress (cortisol levels and perceived stress), physical activity (objective and subjective measures), their mitigation by urban greenness and canopy volume, and their mediation by Doppler ultrasound measures of placental function. RESULTS:We found that higher pregnancy exposure to NO2, BC, PM2.5, and PM2.5 copper and iron contents, particularly at home and all microenvironments combined, were generally associated with lower birth weight, higher risk of small for gestational age (SGA), and a decelerated trajectory of fetal growth, although some of these associations were not statistically significant. These associations appeared to be stronger for mothers with higher SES and those with higher objective measures of psychological stress. For the COVID-19 pandemic and physical activity, as effect modifiers, and urban greenness and canopy cover, as effect mitigators, we observed mixed patterns. In multipollutant models that include different measures of exposure to noise in addition to TRAP, the associations between TRAP and fetal growth generally remained consistent with those that we observed in our main analyses. We found two potential windows of vulnerability for the association of TRAP with fetal growth: one at the end of the first trimester and the beginning of the second trimester, and another at the end of the third trimester. Finally, we observed that a small proportion of the associations between PM2.5 and fetal growth could be mediated through the impact of these pollutants on placental function (i.e., umbilical artery pulsatility index). CONCLUSIONS:Exposure to TRAP is associated with impaired fetal growth.

INTRODUCTION:Since 2015, thousands of rural and peri-urban villages across Beijing and northern China have been treated by a household Clean Heating Policy (CHP) that banned household coal burning and subsidized the costs of electric heaters and electricity. Whether this large-scale policy was successful in improving air quality and health remains an important and unresolved question. We estimated the effects of the CHP policy on air quality and cardiopulmonary health in Beijing villages and quantified how much of the policy's effects on health were mediated by changes in air pollution and indoor temperature. METHODS:In winter 2018-2019, we enrolled 1,003 participants in 50 Beijing villages that were eligible for, but not currently treated by, the CHP and followed them over four consecutive winter data collection waves. In waves 1, 2, and 4, we administered questionnaires and measured participants' anthropometrics, blood pressure (BP), airway inflammation (fractional concentration of exhaled nitric oxide [FeNO]), and 24-hour personal exposure to fine particulate matter (particulate matter ≤2.5 μm in aerodynamic diameter [PM2.5]). Fasting whole blood samples were obtained at clinic visits in waves 1 and 2 for analysis of glucose, lipid profile, and markers of inflammation and oxidative stress. We attempted to contact all prior participants in each follow-up wave. If a previously enrolled participant was not at home or refused subsequent participation, staff first tried to randomly recruit an eligible participant from the same household. If this was not possible, village guides helped field staff to enroll a new participant from a new household using the same sampling procedures as the baseline. Wintertime outdoor PM2.5 was measured in all four waves, and wintertime indoor PM2.5 was measured in waves 2, 3, and 4. Indoor temperature was measured in all waves. The PM2.5 filters were analyzed for their mass, black carbon (BC), and chemical composition, which were used for source apportionment. To estimate the impacts of the policy, we used a difference-in-differences design that accommodated the staggered rollout of the CHP. We used "extended" two-way fixed effects models and marginal effects to quantify the effect of the policy on air pollution and health outcomes. We further evaluated whether villages treated by the policy in different years responded differently to the policy and whether the observed health impacts of the policy were mediated through changes in air pollution or home (indoor) temperature. RESULTS:We enrolled a total of 1,438 participants from 1,236 households during our four study waves. At baseline (wave 1), the mean participant age was 60 years old (standard deviation [SD] = 9.2), 60% of participants were female, and most participants (63%) worked in agriculture. Geometric mean personal exposures to PM2.5 were twice as high as outdoor PM2.5 (72 vs. 36 µg/m3), and the main source contributors were local and transported dust, regional and domestic coal and biomass burning, and secondary pollutants. By waves 2, 3, and 4, there were cumulative totals of 10, 17, and 20 villages (of 50 total) exposed to the CHP. Uptake and adherence to the policy were high: among villages treated in wave 2, the proportion of households using heat pumps and coal heaters, respectively, changed from 3% and 97% in wave 1 to 94% and 3% in wave 4, with similar clean energy transitions in villages exposed to the policy in later waves. Marginal effects derived from multivariable extended two-way fixed effects models showed that exposure to the policy increased wintertime indoor temperature by 1° to 2°C and reduced indoor seasonal PM2.5 by approximately 20 µg/m3. Treatment by the policy also reduced contributions to PM2.5 from solid fuel sources, including household coal burning, and improved BP (~1.5 mm Hg lower systolic BP [SBP] and diastolic BP [DBP]) and self-reported respiratory symptoms (~8 percentage point reduction in any symptoms). There was notable heterogeneity in effects across treatment cohorts, with larger benefits to indoor PM2.5 and health in villages treated in earlier years relative to later years. In the mediation analysis, indoor PM2.5 and indoor temperature explained most of the total effect of the policy on SBP and roughly half of the total effect on DBP, but this did not explain improvements in self-reported respiratory symptoms. We did not find evidence of meaningful effects of the policy on outdoor or personal exposure to PM2.5 or on biomarkers of inflammation and oxidative stress. CONCLUSIONS:In this comprehensive field-based assessment of a large-scale household energy policy in Beijing, we observed high fidelity and compliance with the CHP. Exposure to the policy reduced BP and self-reported chronic respiratory symptoms, and the effects for BP were mediated by reductions in indoor PM2.5 and improvements in home temperature, providing empirical evidence that clean household energy policies can provide population health benefits.

INTRODUCTION:Most existing epidemiological evidence on the health effects of air pollution has focused on single-pollutant analyses, although recent research has increasingly emphasized estimating the effects of multiple exposures simultaneously. In this report, we used causal inference methodology to highlight four impediments to analyses with multiple exposures: (1) there is little information in the data to estimate effects typically of interest, (2) the effects of air pollution mixtures can be heterogeneous, (3) exposure assessment using an individual's home location can be problematic when daily mobility takes them to areas of different exposure levels, and (4) bias due to unmeasured confounding. The objectives of this report were to address these four concerns through the development of rigorous statistical methodology and to provide a corresponding case study that examines the health effects of air pollution in the Medicare cohort in the United States. METHODS:The statistical methodology developed in this report improves the analysis of environmental mixtures in two distinct ways. First, our results highlight inherent difficulties, which require careful consideration in any study of the health effects of multiple exposures. Second, we developed a statistical methodology that broadens the scope of questions that can be answered in analyses of air pollution mixtures and can increase the policy relevance of evidence obtained from epidemiological studies using multiple exposures. Additionally, we illustrated the aforementioned approaches in a nationwide study of the health effects of air pollution in the US Medicare population, extending the existing evidence on the health effects of air pollution within this cohort. RESULTS:In specific aim 1, we found that quantities typically targeted in studies with multiple exposures are difficult to estimate from the observed data alone, as they frequently rely on model-based extrapolation, which can provide unreliable findings. We presented alternative strategies that provide policy-relevant evidence of health effects, while avoiding issues caused by extrapolation. In specific aim 2, we found that the adverse effects of particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5) components are heterogeneous and that these effects are more pronounced in areas with lower socioeconomic status. Specific aim 3 studied the mobility of individuals and found that ignoring mobility can bias health effects, although typically toward the null of no exposure effect. Incorporating mobility in the Medicare cohort did not lead to substantially different findings; however, accounting for mobility tended to increase the magnitude of estimated health effects. In specific aim 4, we developed a methodology for assessing robustness of health effects to unmeasured confounding bias and found that there is robust evidence overall of a harmful effect of pollution on public health. CONCLUSIONS:Our studies provide strong evidence of air pollution effects on public health, and our methodology gives new insights into key issues about this effect. Specifically, the effects of air pollution are heterogeneous and affect certain subgroups of the population more than others, and these effects are moderately robust to unmeasured confounding bias. Future studies can incorporate the ideas and approaches developed in this report to address important questions in analyses with multiple exposures.

INTRODUCTION:Community concerns about the potential health effects of energy development have grown in recent years. This project evaluated the links between unconventional oil and gas development (UOGD) and potential water contamination in Beaver, Greene, and Washington counties of southwestern Pennsylvania (SW PA). This region, with its long history of hydrocarbon development, including coal mining and conventional oil and gas development, has many overlapping sources of potential contamination. Additionally, it is one of the most active UOGD regions globally. As the study progressed, we extended many of our statistical investigations of groundwater in SW PA to the entire state. METHODS:We used statistical analysis to isolate the influences of geogenic and anthropogenic processes on groundwater chemistry and to identify potential linkages between UOGD and water contamination using a groundwater chemistry dataset of over 7,000 samples in SW PA, each with approximately 40 reported chemical analytes. We primarily targeted contamination by salt species found in brines. We conducted six community focus groups in the tri-county region during the summers of 2022 and 2023, which helped identify areas of community concern and interpret our preliminary findings. The focus groups highlighted wastewater mismanagement as a key area of community concern, which we examined in our geoscience analysis. Where possible, we also extended our statistical analysis to the entire state (28,609 groundwater quality analyses) so we could assess the effect of different land uses and geology on water quality. RESULTS:Across the SW PA region, we observe small but statistically significant increases in barium (Ba) and strontium (Sr) in groundwater within 1 km of UOGD, with higher concentrations associated with greater proximity to and density of unconventional oil and gas (UOG) wells. Statistical inferences from the groundwater data point to spills of briny wastewaters on UOG well pads as the likeliest explanation for these increases. For example, Ba and Sr have an even stronger relationship with the locations of spill-related violations at UOG well pads. We found a statistically significant increase in salt concentrations near wastewater impoundments that are no longer in operation because of reprimands by the state regulator and environmental violations. These relationships persist even after better controlling for other geogenic and anthropogenic salt sources using a fixed-effects model. The information gathered from the focus groups suggests that communities are most concerned about potential radiation exposure from UOGD wastewater management, which may increase cancer risks. The geoscientific analysis does not reveal evidence across the region of increased concentrations of species associated with radiation risks in groundwater related to UOGD. This lack of evidence is partly because few groundwater analyses measure or detect radium, the biggest source of radiation in Pennsylvania groundwater. CONCLUSIONS:Our results suggest that the statistically significant increases in salts associated with UOGD are likely due to wastewater spills or leaks from impoundments rather than hydraulic fracturing itself. Our inference that wastewater spills and leaks from impoundments are the most likely mechanism related to increases in brine concentrations aligns with community concerns about wastewater management. This research, along with other previous or ongoing studies, documents that contamination is localized in areas we refer to as "hotspots." Therefore, although geospatial analysis shows extremely small regional increases in brine salt concentrations in groundwater near UOGD, we conclude these increases are due to numerous, well-distributed spill and leak incidents across the shale play, despite their localized impact. The increases in brine salt concentrations in groundwater samples were never observed to be above contamination levels that pose risks for human health according to US Environmental Protection Agency guidelines. However, in areas with dense UOGD, our analysis indicates that some toxic species could be of local concern, given dissolved species ratios and Cl levels in the wastewaters generated through oil and gas development (known as produced water) in Pennsylvania. This result is predicated on assumptions about the average species concentrations in produced waters, the spatial density of UOG wells, and the locations of hotspots. High ionic strength wastewater released into groundwater could also induce secondary mobilization of hazardous species like radium via cation exchange. To address public concerns, additional groundwater testing, especially for radium, should be conducted in identified hotspots, near problematic impoundments, or near spills.

INTRODUCTION:Rapid growth in unconventional oil and gas development (UOGD), driven by improvements in directional drilling and hydraulic fracturing technologies, has made the United States the world's largest producer of oil and natural gas. Rapid UOGD expansion in Colorado's Denver-Julesburg (DJ) Basin, often intersecting with a growing population, has elevated local concerns about noise and air quality impacts. Independent near-pad observations of hazardous air pollutants (HAPs) exposure and noise are uncommon, particularly during well drilling and completions, where evolving technologies and practices continue to alter emissions. The chief goals of this study were to identify potential impacts of UOGD HAPs emissions on nearby communities, characterize A- and C-weighted noise impacts from UOGD operations, and contextualize air and noise pollution measurement and modeling results at health-relevant temporal and spatial scales. Our focus was on the DJ Basin, where a changing regulatory environment has driven innovation to better protect human health and the environment, and past observations facilitate examining whether new operational practices have reduced impacts. METHODS:Air quality and noise impacts of UOGD operations conducted by three major DJ Basin operators were studied at four large well pads at three locations across the basin. Studied operations included well drilling, hydraulic fracturing, coiled tubing/millout operations, flowback, and early production. Methane and speciated volatile organic compound (VOC) measurements were made at near-pad and background locations to characterize increases in concentrations associated with specific UOGD activities. Air dispersion modeling was conducted to assess impacts as a function of distance from well pad operations. A- and C-weighted noise levels were monitored at multiple locations. RESULTS:Near-pad concentration increases of UOGD-related VOCs were most prominent on short timescales, with the most concentrated plumes often one to two orders of magnitude more concentrated than regional background concentrations but typically persisting for less than an hour at a given location. Benzene and, during drilling operations, n-nonane levels came closest to their respective chronic noncancer inhalation health guideline values (HGVs) but remained well below these levels. One-hour benzene levels exceeded 9 parts per billion by volume (ppbv), the acute HGV, on several occasions during different operational phases. Analyses of VOC emissions from specific operational practices provide the first estimate of speciated VOC emission rates from coiled tubing/millout operations, revealed increased emission of C8-C10 n-alkanes from use of low-odor synthetic drilling muds, and indicated that implementation of closed-loop, tankless fluid-handling systems can reduce flowback emissions of benzene by more than 98% relative to other, recent green completion operations. A TRACER (TRAcking Community Exposures and Releases) UOGD preproduction emission model was developed to provide stakeholders with a new tool to forecast HAPs and other VOC emissions from planned well drilling and completion operations and to test strategies to reduce air quality impacts. Noise parameters for A- and C-weighted noise consistently exceeded Colorado Energy and Carbon Management Commission (ECMC) thresholds for chronic noise at the minimum compliance point (350 feet) during all activities monitored. At the maximum compliance point (1,975 feet), A-weighted noise no longer exceeded ECMC chronic thresholds during drilling and coiled tubing/millout activities and only sporadically exceeded ECMC chronic thresholds during hydraulic fracturing. However, C-weighted noise consistently exceeded ECMC chronic thresholds during drilling and hydraulic fracturing and sporadically exceeded ECMC chronic thresholds during coiled tubing/millout operations. CONCLUSIONS:Consistent with prior work in the DJ Basin, noise from UOGD operations poses a concern during well drilling and completions, while the greatest air pollutant exposure risk is associated with acute exposure to benzene. This finding highlights the importance of high time-resolution monitoring to document risk near UOGD preproduction operations. Increased utilization of improvements to preproduction operational practices, including the use of grid-powered, electrified drill rigs, synthetic drilling muds, and closed-loop, tankless fluid-handling systems, has reduced air quality and some noise impacts around the DJ Basin UOGD, while Colorado's implementation of a 2,000-foot setback distance has increased protection from HAPs exposure for those living near new well pads. UOGD operations in many other basins are not conducted using the same emission-reducing operational practices, and residents elsewhere often do not enjoy the protection of a 2,000-foot setback; our analyses find considerably higher risk at 500 feet.

Introduction:Unconventional oil and gas development (UOGD3) has enabled the exploration of previously inaccessible or uneconomic oil and gas resources in shale rock, resulting in thousands of extraction sites across the landscape, many near people's homes. Human exposure to air pollution and noise related to these activities poses a health risk. This study focused on characterizing air pollutants, greenhouse gas emissions, airborne radioactivity, and noise associated with UOGD in two shale production basins. Methods:During one year of stationary air monitoring in Loving, New Mexico, in the western part of the Permian Basin (PB), we characterized the magnitude, frequency, and duration of UOGD-related emissions at temporal scales from 1 minute to seasonal. Continuous monitoring was performed for meteorological variables, near-surface ozone, nitrogen oxides (NOx), sulfur dioxide, hydrogen sulfide, 20 speciated volatile organic compounds (VOCs) in the ethane to octane volatility range, and noise. Airborne radioactivity was measured in the gas and particle phases. Source apportionment was performed using nonnegative matrix factorization (NMF). To disentangle sound frequencies, we developed spectrograms and conducted machine learning regression to analyze sound sources and relationships to air pollutants. A network of passive hydrocarbon samplers that collected weekly measurements of 15 hydrocarbons was established throughout populated regions of both the PB and Eagle Ford Shale (EFS) areas to measure regional pollutant concentrations and their spatial gradients around UOGD. Visible Infrared Imaging Radiometer Suite (VIIRS) Nightfire (VNF) data were acquired to quantify gas flaring activity throughout the region during our field measurement period. VNF flares and estimated flare gas volume were linked to the stationary air quality measurements to examine associations. Results:Most of the monitored primary air pollutants showed high variability, with frequent concentration spikes that exceeded background mole fractions by up to three orders of magnitude. The frequency of concentration spikes, their maximum mole fractions, and averaged metrics (hourly, 8-hour, 24-hour, and annual) were higher than those reported from mostly urban comparison sites. Near-surface ozone measurements confirmed that this area of the PB is in nonattainment of the ozone national ambient air quality standard (NAAQS), exceeding the 70 ppb threshold more than 30 days of the year, and that this nonattainment is driven by UOGD-related emissions of VOCs and NOx. The highest ozone values occurred during conditions of high temperatures, dry air, and slow advection of air masses across the PB from the south-southeast. VOC monitoring showed dominant impacts from saturated hydrocarbons, and associated NMF analysis identified five emission sources: a UOGD-related hydrocarbon source, two gas-flaring sources, a general combustion source, and a road transportation/traffic source. These sources correspond well with identified spatial distributions of surface well pad activity, gas flaring during the study period, and road traffic in the area. Radioactivity measurements displayed diurnal and seasonal changes consistent with prior work; elevated levels were observed from the north-northwest sector. Airborne radioactivity correlated with aromatic VOCs, petroleum hydrocarbon alkanes, and CO2. Our noise measurements were dominated by low frequencies (<100 Hz), which were most strongly associated with CO2. The passive sampler network identified northwest-to-southeast increases in ambient hydrocarbon concentrations, in line with the identified UOGD and traffic density in the area. Benzene levels across the network at times exceeded health-based reference values and were significantly higher than benzene levels recorded during air monitoring in large Texas metropolitan areas. Average hydrocarbon levels were significantly correlated with the well density surrounding each site in both shale basins. Conclusions:Our extensive air quality research in the Permian-Delaware Basin revealed, at times, extraordinarily high levels of air pollution, including air toxics such as benzene, and frequent high ozone days in violation of the ozone NAAQS. Our analyses show that the overwhelming amount of this pollution is due to UOGD activities, including emissions from production and storage, gas flaring, and truck traffic.

INTRODUCTION:Large-scale epidemiological studies investigating long-term health effects of air pollution can typically only consider the residential locations of the participants, thereby ignoring the space-time-activity patterns that likely influence total exposure. People are mobile and can be exposed to considerably different levels of air pollution or air pollution mixtures when inside versus outside, commuting, recreating, or working. Neglecting these mechanisms in exposure assessment may lead to incorrect distributions of exposure over the population, which may, subsequently, lead to incorrect exposure-health relations in epidemiological studies. In this study, we investigated whether a more sophisticated mobility-enhanced exposure assessment would lead to different exposure predictions and health effect estimates compared with using a residential-based exposure. METHODS:Agent-based modeling (ABM3) was used to model mobility patterns in Switzerland and the Netherlands based on travel survey information. Hourly air pollution surfaces of nitrogen dioxide (NO2) and fine particulate matter (PM2.5) developed separately for the Netherlands and Switzerland, for weekdays and weekends, were overlaid with the ABM data to extract exposures. These air pollution exposures were assigned to two adult cohorts in Switzerland - the Swiss Cohort Study on Air Pollution and Lung and Heart Diseases in Adults (SAPALDIA) and the Swiss National Cohort (SNC) - and the European Prospective Investigation into Cancer study adult cohort in the Netherlands (EPIC-NL). Exposures were assigned based on (1) residential address location only (residential-based) and (2) residential and work address locations plus mobility (mobility-enhanced). In the case of SAPALDIA, known work address locations were available and additionally used. Associations with health outcomes (natural and cardiovascular mortality, coronary and stroke events, blood pressure, and lung function) in the three cohorts were investigated. To evaluate the performance of the ABM, we collected GPS readings from 489 participants in Switzerland and 189 participants in the Netherlands in tracking campaigns. The participants recorded GPS readings, using both a wearable GPS recording device and a mobile phone app while also recording their time-activity in the app diary. RESULTS:We successfully developed mobility-enhanced exposures for over 3 million participants, including an assessment of uncertainty. We found a good agreement between exposures estimated with the app and the GPS tracker, supporting the scalability of the approach. We evaluated the ABMs with GPS and time-activity data collected independently in tracking campaigns that included almost 700 participants from selected areas in the two countries. For these participants, the exposures based on GPS measurements versus those derived from ABM showed a moderate to good agreement (R2 = 0.52-0.81). Within the three cohorts, when compared with exposure based on only the residential location, the mobility-enhanced exposure showed very high correlations (R > 0.95). Finally, the epidemiological analyses revealed very small differences in the associations across health outcomes for the different exposure estimates (mortality in SNC; cardiovascular morbidity and mortality in EPIC-NL; and lung function and blood pressure in SAPALDIA) within the three cohorts. In SAPALDIA, where the work address was known for a subset of individuals, a further comparison using the estimated work address in the ABM indicated little difference in mobility-enhanced exposures. CONCLUSIONS:Our results suggest that the assessment of air pollution exposure at the residential address in epidemiological studies generally does not lead to substantial bias in health effects estimates. If time-activity patterns in other study areas differ greatly from the patterns analyzed in our study, differences between residential and activity-enhanced exposures may be larger. Despite the good agreement between residential and work locations, exposure research should continue to strive toward improving exposure assessment in large-scale epidemiological studies to minimize exposure misclassification.

INTRODUCTION:Epidemiological studies often make use of exposure data that is collected in opportunistic and logistically convenient ways. And, while exposure assessment is fundamental to environmental epidemiology, little is known about what exposure assessment study designs are optimal for health inference. The objective of this project was to advance our understanding of the design of exposure assessment measurement campaigns and evaluate their impact on estimating the associations between long-term average air pollution exposure and cognitive function. This feeds into the broader goal of advancing understanding of air pollution exposure assessment design for application to epidemiological inference. METHODS:We leveraged data from the Adult Changes in Thought (ACT3) Air Pollution study (ACT-AP) to characterize exposures for over 5,000 participants from the ongoing ACT cohort. This is a population-based cohort of urban and suburban elderly individuals in the greater Puget Sound region drawn from Group Health Cooperative, now Kaiser Permanente, starting in 1994. Participants were routinely followed with routine biennial visits until dementia incidence, drop-out, or death. Extensive health, lifestyle, biological, and demographic data were also collected. The outcome measure used in this report is cognitive function at baseline based on the Cognitive Abilities Screening Instrument derived using Item Response Theory (CASI-IRT). The IRT transformation of the CASI score improves score accuracy, measures cognitive change with less bias, and accounts for missing test items. Health association analyses were based on 5,409 participants with both a valid CASI score and who had lived in the mobile monitoring region during at least 95% of the 5 years prior to baseline. We used 5-year average exposures that accounted for residential history. Exposure data came from two distinct exposure assessment campaigns carried out by the ACT-AP study: a campaign using low-cost sensors (2017+) that supplemented existing regulatory monitoring data for fine particles (PM2.5, 1978+) and nitrogen dioxide (NO2, 1996+), and a year-long multipollutant mobile monitoring campaign (2019-2020). The evaluation of the added value of low-cost sensor data relied on a combination of regulatory monitoring data and other high-quality data from research studies, calibrated 2-week low-cost sensor measurements from over 100 locations, which were mostly ACT cohort residences, and a snapshot campaign that measured NO2 using Ogawa samplers. Predictions were at a 2-week average time scale, used a suite of ~200 geographic covariates, and were obtained from a spatiotemporal model developed at the University of Washington. The Seattle mobile monitoring campaign collected a combination of stationary roadside and on-road measurements of ultrafine particles (UFPs, four instruments), black carbon (BC), NO2, carbon dioxide (CO2), and PM2.5. Visits were temporally balanced over 288 drive days such that all sites were visited during all seasons, days of the week, and most hours of the day (5 a.m. to 11 p.m.) approximately 29 times each. For the on-road measurements, we divided the driving route into 100-meter segments and assigned all measurements to the segment midpoint. Predictions used the same suite of geographic covariates in a spatial model fit using partial least squares (PLS) dimension reduction with universal kriging (UK-PLS) to capture the remaining spatial structure. We reported model performance metrics for both the spatial and spatiotemporal models as root mean squared error (RMSE) and mean squared error (MSE)-based R2. The reference observations for the spatiotemporal model were low-cost sensor measurements at home locations (with performance metrics averaged over their entire measurement period to approximate spatial contrasts), and for the spatial model, the reference observations were the all data long-term averages at stationary roadside locations. Using various approaches to sample data from these two exposure monitoring campaigns, we determined the impact on exposure prediction and estimates of health associations using two confounder models and 5-year average exposure predictions for cohort members at baseline developed from the alternative campaigns. For the low-cost sensor data, we evaluated temporally or spatially reduced subsets of low-cost sensors, as well as a comparison of the low-cost sensor versus snapshot campaigns for NO2. For the mobile monitoring data, we considered designs focused on the stationary roadside and on-road data separately. We reduced the stationary roadside data temporally by restricting seasons, times of day, or days of week for the campaign, while also considering a reduced number of visits using balanced sampling, as well as a set of unbalanced visit designs. We also reduced the on-road data spatially and temporally to assess the importance of spatially or temporally balanced data collection. In addition, we considered the impact of incorporating temporal adjustment to account for temporally unbalanced sampling, as well as plume adjustment to account for on-road sources. For each design, we evaluated prediction model performance using the all data stationary roadside observations (mobile campaign) or the measurements at homes (low-cost sensor campaign) as reference observations to ensure consistency in reported performance metrics. We also used long-term average exposures estimated from these alternative campaigns in health association analyses under two different confounder models that were adjusted by potentially confounding variables: Model 1 adjusted for age, calendar year, sex, and educational attainment; Model 2 included all Model 1 variables with the addition of race and socioeconomic status. Furthermore, using the stationary roadside data, we applied parametric and nonparametric bootstrap methods to account for Berkson-like and classical-like exposure measurement error for the UFP exposure in confounder model 1. In a separate methods-focused aim, we developed and applied advanced statistical methods using the stationary roadside mobile monitoring data. To evaluate possible improvements in exposure model performance, we applied tree-based machine learning algorithms that also account for residual spatial structure, and compared these to UK-PLS. This led to the development of a variable importance metric that uses a leave-one-out approach to evaluate the change in predictions across various user-specified quantiles. The variable importance metric produces covariate-specific averages that reflect how the predictions, on average, vary across different quantiles of each covariate. This serves as an intuitive measure of the contribution of this covariate to the predicted outcome. A key idea in this variable importance approach is to reuse the trained mean model across all locations and to refit the covariance model in a leave-one-out manner. In separate work to address dimension reduction for multipollutant prediction, we extended classical principal component analysis (PCA) and a recently developed predictive PCA approach to optimize performance by balancing the representativeness in classical PCA with the predictive ability of predictive PCA. We called the new method representative and predictive PCA, or RapPCA. Finally, we characterized the various exposure assessment campaigns in terms of the value of their information as quantified by cost. We calculated costs, focused predominantly on staff days of effort, for various exposure assessment designs and compared these to exposure model performance statistics. RESULTS:We found that air pollution exposure assessment design is critical for exposure prediction, and also impacts health inference. We showed that a mobile monitoring study with stationary roadside sampling that has at least 12 visits per location in a balanced and temporally unrestricted design optimizes exposure model performance while also limiting costs. Relative to weaker alternatives, a balanced and temporally unrestricted design has improved accuracy and reduced variability of health inferences, particularly for confounder model 1. To address temporal balance, it is important that the exposure sampling in mobile monitoring campaigns cover all days of the week, most hours of the day, and at least two seasons. The popular temporally restricted business-hours sampling design had the poorest performance, which was not improved by adjusting for the temporally unbalanced sampling approach. We found similar patterns using on-road data, though the findings were weaker overall. For the alternative exposure campaign that supplemented regulatory monitoring data with low-cost sensor data, while the exposure prediction model performances improved with the inclusion of the low-cost sensors, there was little notable impact on the health inferences, and the costs were steep. Given that the supplementary exposure assessment data were sparse relative to the existing regulatory monitoring data, and that the low-cost sensor data collection used a rotating approach due to the limited number of sensors (i.e., low-cost sensor measurements were not collected using a balanced design), it was much more challenging to develop deep insights from this exposure assessment approach. Finally, we found that leveraging spatial ensemble-learning methods for prediction did not improve exposure prediction model performances or alter health inferences. The new multipollutant dimension-reduction we developed, RapPCA, had the best predictive performance and also minimized the prediction error in comparison with both classical and predictive PCA. CONCLUSIONS:This project has shown that there should be greater attention to the design of the exposure data collection campaigns used in epidemiological inference. Based on the multiple investigations conducted, many of which focused on UFPs, we found that exposure predictions with better performance statistics resulted in health association estimates that were generally more consistent with those obtained using the "best" exposure model predictions (the model with all data included), although the pattern of health estimates was often less conclusive than the pattern of prediction model performances. Furthermore, we found that it is possible to design air pollution exposure assessment studies that achieve good exposure prediction model performance while controlling their relative cost. We developed strong recommendations for mobile monitoring campaign design, thanks to the well-designed and comprehensive Seattle mobile monitoring campaign. Insights from supplementing regulatory monitoring data with low-cost sensor data were less compelling, driven predominantly by a data structure with sparse and temporally unbalanced supplementary data that may not have been sufficiently comprehensive to demonstrate the impacts of alternative designs. Broadly speaking, better exposure assessment design leads to better exposure prediction model performance, which in turn can benefit estimates of health associations. We did not find that leveraging advanced statistical methods (specifically, spatial ensemble-learning methods for prediction) improved exposure prediction model performances. This finding is not consistent with the conclusions reached by other investigators, and may have been due to the already sophisticated UK-PLS approach we used by default, and in particular its application in conjunction with the large number of covariates that we considered in the PLS model, such that the contribution of any single covariate was approximately linear. In other words, it is reasonable to believe that in the presence of the large set of covariates we considered, each can contribute an approximately linear association with the pollutant being modeled, such that the potential added value of the spatial Random Forest approach is not observed in the model fit. Other settings with a smaller number of possible covariates available may lead to different conclusions and suggest greater added value of the application of a spatial Random Forest approach. We based our approach on leveraging the extensive air pollution exposure assessment and outcome data available from the ACT-AP study. Thus, we sampled from the existing air pollution data to evaluate exposure assessment designs that were subsets of those data. Then, conditional on each of these designs, we evaluated subsequent health inferences, which focused on cognitive function at baseline using the CASI-IRT outcome. The magnitude and uncertainty of these health association estimates were dependent upon the associations evident in the ACT cohort, and the insights we were able to develop are conditional on the strengths and weaknesses of these data. Specifically, while we observed some larger impacts on health association estimates of more poorly performing exposure models relative to the complete all data exposure model, such as the business-hours design from a mobile monitoring campaign, many of the differences were small and did not deviate meaningfully from the health association estimate obtained from the "best" exposure model. The degree of impact on the epidemiological inference depended on the magnitude of the health association estimate from the "best" exposure model and the width of its confidence interval. Future investigations should replicate and expand upon these findings in other settings, including application to new cohorts and exposure assessment data, as well as in simulation studies, which provide an alternative approach to using real-world data to evaluate a constellation of exposure models. However, while knowledge of the assumed underlying truth is an important strength of simulation studies, it is challenging to capture real-world complexity meaningfully in simulation studies. Our foray into applying advanced machine-learning methods to improve exposure predictions produced the surprising result that our default UK-PLS approach for spatial prediction produced similar performance metrics to spatial ensemble-learning methods. Future evaluations that assess smaller subsets of exposure covariates will allow determination of the relative exposure model performance benefits of UK-PLS versus spatial ensemble-learning methods, and provide insights into the possible reason that our conclusions differ from others in the literature.

INTRODUCTION:Determining whether chronic exposure to air pollution contributes to observed disparities in COVID-19 outcomes requires integrating multiple determinants of patient vulnerability to COVID-19, given the complex interactions that contribute to health disparities. Exposure to adverse social and structural factors heightens vulnerability to environmental exposures, potentially resulting in increased risk of unfavorable COVID-19 outcomes. Additionally, as populations are often exposed to various co-occurring adverse factors in the setting of disinvested neighborhoods and communities, examining such factors individually may not be sufficient to fully understand how they may modify the effects of air pollutant exposures. In an effort to explain COVID-19-related disparities observed in New York City (NYC3), this study aimed to estimate the effect of chronic air pollutant exposures on the risk of COVID-19 morbidity and mortality and to determine whether these effects vary by neighborhood-level vulnerability as defined by social and structural factors. METHODS:We used harmonized electronic health record (EHR) data from five healthcare systems in NYC to derive a study population of hospitalized or emergency department (ED) patients diagnosed with COVID-19 from March 1, 2020, through February 28, 2021, who had a NYC zip code of residence. To reduce potential selection bias, we also constructed a subset of the study population restricted to patients with residential zip codes in the typical catchment area of the hospitals affiliated with the EHR data repository. We estimated air pollutant concentrations for fine particulate matter (PM2.5), nitrogen dioxide (NO2), black carbon (BC), and ozone (O3) by using zip code-level 11-year averages based on data from the 2009-2019 New York City Community Air Survey. For each pollutant, we constructed Cox proportional hazards models to estimate the hazards of fatality (i.e., dying from COVID among individuals with COVID) and hospital length of stay. Additionally, for each pollutant, we constructed Poisson regression models to estimate RRs (RRs) for acute respiratory distress syndrome (ARDS), pneumonia, mechanical ventilation, and dialysis during hospitalization and risk of hospitalization among ED patients. Models were adjusted for age, sex, body mass index, smoking status, history of chronic disease, and a neighborhood environmental vulnerability index (NEVI). Interaction terms were used to evaluate effect modification between pollutant exposures and the NEVI metric. Additionally, we conducted supplementary analyses to determine the joint effects of air pollution and pre-existing chronic diseases and whether those relationships varied by NEVI tertile. To supplement the fatality analysis, we conducted an excess mortality analysis among the full urban population using all-cause mortality data from public health records for 2015-2020. Sensitivity analyses were performed to evaluate the effect of selection bias. RESULTS:Exposures to NO2, PM2.5, and BC were positively associated with risks of ARDS, pneumonia, and dialysis, whereas O3 exposure was inversely associated with these morbidity outcomes, likely because of the strong inverse correlation between O3 and NO2. Conversely, we observed an unexpected inverse association between exposures to NO2, PM2.5, and BC and risks of fatality and mechanical ventilation. We observed statistically significant effect modification by NEVI for some of the associations between NO2, PM2.5, O3, and BC exposures and risks of ARDS, pneumonia, and dialysis. In areas with greater environmental vulnerability (i.e., higher NEVI metrics), there were generally stronger positive associations between air pollutant exposures and the risk of hospitalization among ED patients and risks of ARDS, pneumonia, and dialysis among hospitalized patients. Exposures to NO2, PM2.5, and BC were generally negatively associated with the risk of fatality, even in areas with higher NEVI metrics. Most positive associations between air pollution and COVID-19 outcomes were limited to the initial phase of the pandemic, except for the risk of hospitalization, which was positively associated with NO2, PM2.5, and BC exposures throughout the study period. Even after accounting for the NEVI metric and pre-existing chronic disease, racial disparities persisted in the effect of air pollution on risks of pneumonia and hospitalization, with the largest RRs among Black and Hispanic populations. Results of the all-cause mortality analysis also showed no evidence of greater excess mortality in areas with higher levels of air pollution. The greatest excess mortality was observed in areas with high NEVI metrics, regardless of air pollutant exposures. CONCLUSIONS:When limiting to individuals in the hospital's typical catchment areas, the observed positive associations between air pollutant exposures and COVID-19-related morbidities such as ARDS, pneumonia, and use of dialysis were strongest in areas with higher neighborhood-level environmental vulnerability. Inverse associations between air pollutant exposures and severe outcomes like death and use of mechanical ventilation were unexpected findings that highlighted challenges in examining such associations at the population level in NYC.

INTRODUCTION:Cohort studies have been widely used to estimate the effects of long-term exposure to air pollutants on health outcomes. The nature of the exposure (i.e., personal exposure to outdoor-generated pollution) and the large number of participants in cohorts preclude measuring individual exposure longitudinally. Thus, surrogate measures, such as exposure models, are increasingly used in epidemiological studies to estimate individualized long-term exposures. We evaluated whether increasingly detailed estimates of long-term individual exposure in large-scale studies yield better estimates of the health effects of exposure to outdoor air pollution. We utilized several personal exposure measurement campaigns, which were implemented before the start of MELONS, the uniquely dense monitoring network and surrogate measures previously developed for London. METHODS:Data from 344 participants in four personal measurement campaigns, two measuring particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5), nitrogen dioxide (NO2), and ozone (O3), and two measuring black carbon, covering 12,901 person days during 2015-2019, were used. The total personal exposure measurements were separated into exposures from outdoor and indoor sources by estimating appropriate infiltration factors and behaviors. The exposures were extrapolated from the measurement period per subject (from a few days to >9 months) to annual exposures, taking ambient concentration, infiltration, and behavior variability into account. These annual exposures were defined as true exposures, although it is acknowledged that several assumptions involved in their estimation introduce uncertainty. Surrogate measures of exposure were assigned based on the nearest fixed-site monitor to the residence or the prediction from combined dispersion, machine learning, and land use regression models at the participants' residence. The models were adjusted for age-group and area-specific time-activity patterns based on a large survey. Measurement errors (MEs) were calculated between "true" and surrogate exposures and used as input in a simulation study to investigate the resulting bias in health effect estimates, using total mortality as a health outcome. We estimated the amount of classical and Berkson error in the ME. In addition, we tested, in several theoretical error scenarios, the effectiveness of two correction methods: simulation extrapolation (SIMEX) and regression calibration (RCAL). Finally, we applied the different surrogate exposure methods using data from the UK Biobank London cohort (~62,000 subjects) to assess associations with several mortality and morbidity outcomes in Cox regression models adjusted for multiple covariates and applied correction methods. RESULTS:Exposure to outdoor-generated pollution accounted for at least 50% of total personal exposure, even in subjects spending almost all of their time indoors. We found large MEs, possibly due not only to the nature and uncertainty of using surrogate measures but also to several uncertainties incorporated in the "true" exposure assessment. The resulting bias in health effect estimates from ME was large and almost always toward the null (i.e., the health effects are underestimated, sometimes by as much as 100%). Larger total ME and larger proportion of classical ME led to more underestimation of effects. SIMEX and RCAL were effective methods for bias correction. Furthermore, the different scale (magnitude) of measurement of surrogate exposure estimates of ambient concentrations introduced additional systematic ME, which was addressed by expressing the effects per interquartile range and not per fixed increment of the pollutant. The application to the UK Biobank cohort data showed hazard ratios above 1 for a few outcomes and surrogate exposures, which were corrected, leading to larger estimated effects. CONCLUSIONS:Our results underline the importance of exposure to ambient air pollution ME in estimating health effects and the difficulty in obtaining an accurate estimate of the "true" personal exposure to outdoor-generated pollutants. The common use of surrogate measures of exposure introduces ME, which can be substantial and largely classical, leading to a large underestimation of effects on health. Researchers should consider correcting for ME when reporting results from epidemiological studies on the effects of long-term air pollution exposures and plan ahead by designing appropriate validation studies.

INTRODUCTION:COVID-19 lockdowns led to considerable reductions in air pollutant emissions worldwide, providing a unique opportunity to examine the impacts of reduced air pollution on mortality. This project aimed to quantify changes in nitrogen dioxide (NO2) and fine particulate matter (PM2.5) concentrations due to COVID-19 lockdowns, estimate associations between short-term exposures to these air pollutants and mortality rates, and calculate the attributable changes in mortality in four regions that implemented lockdowns but were mildly affected by the pandemic in early 2020, including Jiangsu Province, China; California, USA; Central and Southern Italy; and Germany. METHODS:To account for meteorological impacts and air pollution time trends, we used a machine learning-based meteorological normalization technique and the difference-in-differences approach to quantify changes in NO2 and PM2.5 concentrations due to lockdowns in early 2020. Using daily air pollution and mortality data from 2015 to 2019, we applied interactive fixed effects models (a causal modeling approach) to estimate associations between day-to-day changes in PM2.5 and NO2 concentrations and all-cause, natural-cause, and cardiovascular mortality rates before the pandemic in each region. Finally, using the quantified air pollution changes and the estimated air pollution-mortality relationships, we calculated the changes in mortality that were attributable to air pollution changes due to the lockdowns. RESULTS:We found that meaningful improvements in air quality occurred during the lockdowns in early 2020 in Jiangsu, China; California, USA; and Central and Southern Italy, with smaller magnitudes of reduction in PM2.5 compared to NO2. We observed no significant reduction in NO2 and a small increase in PM2.5 in Germany. After controlling for unmeasured spatial and temporal confounders, we detected statistically significant associations between short-term increases in PM2.5 and NO2 concentrations and increases in daily all-cause, natural-cause, and cardiovascular mortality rates in all four study regions from 2015 to 2019. Specifically, we determined that lockdown-induced reductions in NO2 resulted in avoiding 1.41 (95% empirical confidence interval [eCI]: 0.94-1.88), 0.44 (95% eCI: 0.17-0.71), and 4.66 (95% eCI: 2.03-7.44) deaths per 100,000 people in Jiangsu, China; California, USA; and Central and Southern Italy, respectively. Mortality benefits attributable to PM2.5 reductions in these regions also were statistically significant, albeit of a smaller magnitude, and resulted in avoiding 0.16 (95% eCI: 0.04-0.29), 0.23 (95% eCI: 0.03-0.43), and 0.91 (95% eCI: 0.09-1.78) deaths per 100,000 people in Jiangsu, China; California, USA; and Central and Southern Italy, respectively. In Germany, the mortality benefits attributable to NO2 changes were not statistically significant (mortality change of -0.11; 95% eCI: -0.25 to 0.03 deaths per 100,000 people), and an observed increase in PM2.5 was associated with an increase in mortality of 0.35 (95% eCI: 0.22-0.48) deaths per 100,000 people during the lockdown. CONCLUSIONS:Using a causal modeling approach, this study contributes to the growing body of evidence that short-term exposures to PM2.5 and NO2 are associated with increased all-cause and cause-specific mortality rates. In areas mildly affected by the COVID-19 pandemic, lockdowns in early 2020 generally improved air quality and led to health benefits, especially in association with NO2 reductions, with notable heterogeneity across regions. This study underscores the importance of accounting for local characteristics when policymakers adapt successful emission control strategies from other regions.

INTRODUCTION:Assessment of long-term exposure to outdoor air pollution remains a major challenge for epidemiological studies. One of these challenges is characterizing fine-scale spatial variation of the ambient concentrations of key traffic-related air pollutants - including ultrafine particles (UFPs), black carbon (BC), and nitrogen dioxide (NO2). Epidemiological studies have used widely different approaches to address these challenges, including empirical land use regression (LUR) models based on fixed-site routine or targeted monitoring, low-cost sensor networks, mobile monitoring, and deterministic dispersion models. Little information is available about the relative performance of these different approaches for assessing long-term exposure to traffic-related air pollution. Different methods may result in heterogeneity in health effect estimates from epidemiological studies applying different exposure-assessment approaches. The Specific Aims of the study. 1. Develop long-term ambient air pollution exposure estimates for selected epidemiological studies based on low-cost sensors, mobile and fixed-site monitoring, and deterministic dispersion modeling. 2. Compare different exposure assessment methods in terms of their ability to predict spatial variation of long-term average concentrations using external validation data. 3. Compare different exposure assessment methods in terms of air pollution effect estimates in selected epidemiological studies. We assessed UFPs, NO2, BC, and particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5). METHODS:We evaluated annual average air pollution concentrations across the Netherlands using a suite of different exposure models, which differed in modeling approach (empirical LUR, deterministic dispersion models) and monitoring data used (low-cost sensors, mobile monitoring, nationwide and Europewide routine monitoring, and study-specific targeted monitoring). For empirical models, we tested three model development algorithms: supervised linear regression (SLR), Random Forest, and least absolute shrinkage and selection operator (LASSO). The predictions of the models were compared at 20,000 addresses across the Netherlands. The performance was also tested on external validation data, which were obtained from a new campaign (2021-2023) and existing data from different years, allowing assessment of how well recent models predict past air pollution exposure. Epidemiological analyses in three cohort studies were conducted to compare health effect estimates of the different exposure models. We assessed associations of air pollution in a national administrative cohort with natural-cause and cause-specific mortality, in a cohort study that had detailed lifestyle data with natural-cause mortality and incidence of stroke and coronary events, and in a mature birth cohort with lung function and asthma incidence. RESULTS:Exposure predictions at residential sites from the dispersion model and the Europewide hybrid LUR models were available for multiple years in the period 2010-2019. For these models, exposure predictions of different years in the period 2010-2019 were highly correlated for BC, NO2, and PM2.5 (Correlation coefficient R > 0.9). Consistently, the year of the exposure model did not affect the presence of an association with mortality and morbidity. Small differences in hazard ratios (HR) were related to exposure contrast for different years. The HR for the association of NO2 with natural-cause mortality was 1.026 (95% confidence interval [CI]: 1.022-1.031) for the 2010 exposure estimate and 1.030 (1.024-1.035) for the 2019 exposure estimate of the Europewide LUR model, expressed per 10 µg/m3. The exposure models generally resulted in highly to moderately correlated exposure predictions at residential sites across the Netherlands (R > 0.7 for BC, NO2, and UFPs; R > 0.5 for PM2.5). The predicted level of exposure and exposure contrast could differ substantially between models and algorithms within models; for example, the interquartile range (IQR) for BC for each of the various models at the 20,000 residential locations ranged between 0.1 and 2.2 µg/m3. Mobile monitoring studies generally resulted in modestly higher BC concentrations and exposure contrasts compared to other exposure models. Small differences were found between the different models in explaining the spatial variation of air pollution concentrations at the new and existing validation sites. Models explained historical exposure patterns at external sites covering more than 10 years moderately well, especially for BC (R > 0.7) and NO2 (R > 0.7), and moderately so for UFPs (R > 0.5). Most models predicted the small concentration contrasts of PM2.5 relatively poorly. Consistent with the high correlation of the different exposure models, the application of these models generally resulted in similar conclusions on the presence of associations with natural-cause, respiratory, and lung cancer mortality in the large nationwide cohort, and with asthma incidence and lung function in the birth cohort. However, the effect estimates differed substantially; for example, the HR for natural-cause mortality in the nationwide administrative cohort for a 1 µg/m3 increase in BC ranged from 1.01 (95% CI: 0.99-1.02) to 1.09 (1.07-1.10). For the outcomes with small effect estimates and the smaller cohort studies, differences in conclusions related to the exposure assessment method were more distinct. Differences in exposure assessment may contribute substantially to the observed heterogeneity of effect estimates in systematic reviews of epidemiological studies. High heterogeneity was indicated by the commonly used heterogeneity measure I2, where the value was above 80% for a meta-analysis of the different effect estimates for natural-cause mortality in the nationwide cohort. Validation of long-term exposure models for the nonroutinely monitored pollutants BC and especially UFPs was challenging, despite generally successful monitoring. The new external validation monitoring campaign resulted in rather unstable estimates of the long-term average spatial contrast, both across sites and where affected by temporal variation, especially for BC and PM2.5. No consistent differences were found in the model performance of SLR, Random Forest, and LASSO, both in internal cross-validation of model building and on external validation sites not used in model building. Exposure predictions from the three algorithms were generally highly correlated and resulted in similar associations with health. However, for individual models, occasionally large differences were found in exposure contrast, validation statistics, and associations with mortality and morbidity outcomes. There was little benefit in using low-cost sensors for NO2 and PM2.5. The addition of low-cost sensor data did not improve NO2 estimates in models that combined dispersion model estimates and data from the national monitoring network data. CONCLUSIONS:The main conclusions of the project. • Exposure predictions of BC, NO2, and PM2.5 for different years between 2010-2019 were highly correlated, documenting stable spatial contrast patterns. Consistently, the year of the exposure model did not affect the presence of an association with mortality and morbidity outcomes. • Models explained historical exposure patterns at external sites covering more than 10 years moderately well, especially for BC. • Different exposure models generally resulted in highly to moderately correlated exposure predictions. The predicted level of exposure and exposure contrast could differ substantially between models. Small differences were found between the different models in explaining spatial variation at validation sites. • Application of different exposure models resulted in similar conclusions about the presence of associations with health outcomes, but effect estimates differed substantially in magnitude between individual exposure models. No consistent differences in effect estimates were found between groups of mobile, dispersion, and fixed-site LUR models. • Differences in exposure models may therefore contribute substantially to the observed heterogeneity of effect estimates in systematic reviews of epidemiological studies. Factors that explained some of the heterogeneity of effect estimates included the performance of the model at external validation sites and the predicted exposure contrast. • Exposure predictions from the three algorithms were generally highly correlated and resulted in similar associations with health. No consistent differences were found in their model performances.

INTRODUCTION:Emerging evidence suggests that ambient air pollution may affect the developing brain and contribute to an increased risk of mental health problems. However, most studies have focused on prenatal or early postnatal periods of exposure, with less attention given to the dynamic neurodevelopment period of early adolescence. Moving forward, it is necessary to consider additional periods of exposure, such as adolescence, and the biological mechanisms that may drive potential neurotoxicological effects. This project aimed to investigate whether 1-year exposure to ambient fine particulate matter (PM2.5) and nitrogen dioxide (NO2) at 9-10 years of age was associated with (1) concurrent prefrontal white matter connectivity at ages 9-10 years and (2) emotional health problems at ages 9-10 years as well as 1 year later. Lastly, we hypothesized that poor prefrontal white matter connectivity might be an intermediate marker (i.e., mediator) for the association between 1-year ambient exposure and mental health outcomes. METHODS:We leveraged data from the multisite, nationwide Adolescent Brain Cognitive Development Study (ABCD Study; N = 11,880), with cross-sectional data on diffusion-weighted imaging at 9-10 years (baseline visit) and longitudinal emotional health outcomes at 9-10 (baseline visit) and 10-11 years (1-year follow-up). Based on residential addresses at ages 9-10 years, novel hybrid spatiotemporal exposure models were applied to estimate 1-year average ambient exposure to PM2.5 and NO2. Diffusion tensor imaging (DTI) was used to measure white matter microstructure in tracts that innervate the prefrontal cortex. Emotional behavioral problems were measured based on caregiver reports using the Child Behavioral Checklist (CBCL). Mixed-effect two-pollutant models were fit using both PM2.5 and NO2 and adjusted for the study site, several potential sociodemographic and lifestyle characteristics, and magnetic resonance imaging (MRI) precision variables when necessary. For emotional health outcomes, longitudinal models included interaction terms for pollutant-by-time for both pollutants. Sensitivity analyses were conducted that also accounted for the number of years the child resided at the residential address, as well as adjusting for prenatal PM2.5 and NO2 exposures. RESULTS:The final analytic sample included 7,546 participants with DTI data and 9,334 participants with emotional behavior data. The annual exposures to PM2.5 and NO2 across 21 study sites were 7.66 μg/m3 [1.72-15.90 μg/m3] and 18.61 ppb [0.73-37.94 ppb], respectively. Annual exposure to PM2.5 was found to be significantly related to prefrontal structural connectivity, including fractional anisotropy (FA) in the right superior longitudinal fasciculus and widespread differences in mean diffusivity (MD) in the corpus callosum, bilateral uncinate fasciculus, left cingulum-hippocampal region, left anterior thalamic radiation, and left superior longitudinal fasciculus. The observed associations between PM2.5 and MD were negative and nonlinear, with greater decreases in MD seen at higher exposure levels. Annual exposure to NO2 was found to have significant, negative linear associations with FA in the right anterior thalamic radiation, left uncinate fasciculus, and corpus callosum. In terms of emotional behavior, 1-year PM2.5 annual exposure was related to slightly less internalizing, anxiety/depression, and aggression problems at the 1-year follow-up. Similarly, 1-year NO2 annual exposure was related to slightly less internalizing and total problems at the 1-year follow-up. Although some of these associations were statistically significant, small parameter estimates suggest these noted effects on emotional outcomes may not be of clinical importance. Given the later findings, the required conditions to test mediation formally were not met. CONCLUSIONS:Our analyses indicate that white matter microstructure is uniquely associated with annual exposure to PM2.5 and NO2 at ages 9-10 years. Against our hypotheses, annual exposure was not related to more emotional problems at ages 9-10 years or after a 1-year follow-up period. These findings suggest air pollution exposure levels below US national ambient air quality standards may have important implications for child white matter development and add to the literature suggesting neurotoxicity at low exposure levels of air pollution may be critical to include in the continuing review and risk assessment for the National Ambient Air Quality Standard.

INTRODUCTION:In the United States, billions of dollars have been spent implementing interventions to reduce traffic-related air pollution (TRAP). These interventions are usually regulatory actions focused on reducing tailpipe emissions. However, they also include local programs to reduce traffic congestion and excess vehicle emissions, such as electronic tolls and roadway capacity improvements. Few health studies have empirically evaluated the direct impact of air pollution exposure reductions from these emission regulations and congestion reduction programs; no studies have examined infant health, an important population health outcome linked to air pollution exposures. OBJECTIVE:Assess changes in birth outcomes for all recorded births in Texas from 1996 to 2016 associated with (1) long-term cumulative regulatory improvements of motor vehicle emissions and resulting TRAP change and (2) local congestion reduction programs that may yield localized TRAP changes over shorter time periods. METHODS:We used Vital Statistics data in Texas from 1996 to 2016 (n = 8.1 million recorded births; n = 6,158,518 births analyzed after exclusions). We calculated diverse traffic-related exposure measures using residential addresses at the time of delivery. We implemented research triangulation methods using different study design and analysis approaches to test our primary hypotheses on the effects of long-term cumulative regulatory improvements and local congestion reduction programs on birth outcomes. RESULTS:Traffic-related exposure measures (nitrogen dioxide [NO2] air pollution, traffic volume, congestion) were consistently associated with adverse birth outcomes over the 20-year study period. This finding is supported by an analysis of pregnant individuals living upwind versus downwind of the same major road, where living downwind within 500 m was associated with an 11.6-g decrease (95% CI: -18.01, -5.21) in term birth weight. For all pregnant individuals, NO2 exposures decreased 59% from 1996 to 2016, while the total vehicle miles traveled (VMT) within 500 m of residential addresses (VMT500m) remained relatively stable. We observed marked differences in TRAP exposure for pregnant individuals by sociodemographic characteristics. While levels of air pollution disparities reduced in absolute terms over the 20 years, relative disparities persisted, and large differences in traffic levels remained. The magnitude of associations between VMT500m and adverse birth outcomes decreased for term low birth weight (-60%, OR in 1996: 1.08, OR in 2016: 1.03 for the highest vs. lowest quintile) and preterm (-65%) and very preterm (-61%) births, but not for term birth weight. A direct analysis of congestion exposure for 2015-2016 births, measured for all roadways in Texas using connected device data, showed that congestion was associated with decreased term birth weight, background traffic, and TRAP levels. When we examined local projects designed to reduce congestion as a natural experiment and applied a difference-in-differences (DiD) study design, we found little evidence that the implementation of tolling projects was associated with improved birth outcomes. For roadway construction projects, we observed increased congestion during construction and decreased congestion post-construction. This dynamic translated into increased odds of term low birth weight (OR 1.19; 95% CI: 1.05, 1.36) for pregnant individuals living within 300 m during construction but no consistent improvements in birth outcomes post-construction. CONCLUSIONS:TRAP is an important environmental health and justice issue that affects pregnancy. Our results provide some evidence supporting that cleaning up the vehicle fleet was more impactful at decreasing adverse pregnancy outcomes than local programs aimed at reducing congestion.

INTRODUCTION:We conducted the HERMES study to address the role of source-specific air pollution and the independent effects of air pollution, noise, and green space as well as the identification of susceptible subgroups defined by sociodemographic characteristics, stress conditions, and comorbidity in relation to cardiometabolic health. We studied three cohorts, a chemistry transport model (CTM) system, a noise model, a high-resolution land use map, and Danish registries on health and sociodemographic variables at individual and small-area levels. METHODS:Using Danish registries we defined a cohort of about 2 million persons living in Denmark. We also used data from the Danish National Health Survey (DNHS) (n = 246,766) and the Diet Cancer and Health - Next Generations cohort (DCH-NG) (n = 32,851). The Danish registries provided sociodemographic data at individual and small-area levels and allowed identification of medical diagnoses, comorbidity, and financial stress. The other two cohorts included information on lifestyle habits and measurements of blood pressure and biomarkers. We used Cox models for analyses of associations between exposures and type 2 diabetes, myocardial infarction (MI), and stroke. For analyses of interactions, we used both Cox and Aalen models and multivariate linear regression models for the analyses of air pollution and biomarkers. RESULTS:Air pollution concentrations correlated well with measurements. Analyses of associations between air pollution and type 2 diabetes, MI, and stroke adjusted for individual and area-level sociodemographic variables showed that further adjustment for individual lifestyle had minimal effect on the risk estimates. All four air pollutants were associated with a higher risk of each of the three endpoints. The local traffic contribution to air pollution seemed more important for risk of type 2 diabetes than the contribution from all other sources combined, whereas for MI and stroke, the contribution from all other sources seemed most important. The most consistent interaction was a stronger association between air pollution and type 2 diabetes, MI, and stroke among those with comorbidity. For MI and stroke, we found several interactions on the absolute scale that could not be detected on the relative scale. In multiexposure analyses, we found that particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5) was most important for cardiovascular diseases, and ultrafine particles (UFPs) were most important for type 2 diabetes. We also found that noise and lack of green space were associated with all three endpoints. Analyses of the DCH-NG cohort showed associations between exposure to air pollution and higher concentrations of non-high-density lipoprotein, lower concentrations of high-density lipoprotein, and higher blood pressure. The contribution to air pollution from sources other than local traffic seemed mainly responsible for these associations. CONCLUSIONS:We found that PM2.5, UFPs, elemental carbon (EC), and nitrogen dioxide (NO2) were all associated with type 2 diabetes, MI, and stroke in single-pollutant models. However, in multiexposure analyses that included noise and green space, only UFPs for type 2 diabetes and PM2.5 for MI and stroke remained associated, suggesting that these are the main air pollutants responsible for increasing the risk of cardiometabolic disease. Noise and lack of green space were also associated with cardiometabolic diseases in multiexposure models. We found that air pollution from local traffic was most important for risk of type 2 diabetes, whereas air pollution from other sources was most important for the risk of MI and stroke, which could relate to different air pollution mixtures and/or different biological pathways. Associations between air pollution and type 2 diabetes, MI, and stroke were consistently stronger among individuals with comorbidity, indicating higher susceptibility to negative air pollution effects in this subpopulation. The results of the interaction analyses showed that higher risk estimates among those of low socioeconomic status could be detected when estimating absolute risk but not when estimating relative risk, indicating that the best picture of effect modification is provided when expressed by both relative and absolute risk. The biomarker study showed expected associations between exposure to air pollution and blood lipid levels and blood pressure.