Long-Term Air Pollution, Genetic Susceptibility, and the Risk of Depression and Anxiety: A Prospective Study in the UK Biobank Cohort

Abstract

BACKGROUND: Depression and anxiety are two mental disorders that are often comorbid. However, the associations of long-term air pollution exposure with depression and anxiety remain inconclusive. OBJECTIVE: We conducted a cross-sectional and prospective study to examine the associations of ambient exposure to particulate matter (PM) with a diameter of =2:5 lm (PM2:5), =10 lm (PM10), and 2:5-10 lm (PMcoarse), nitrogen oxides (NOx), and nitrogen dioxide (NO2) with the risk of depression and anxiety in the UK Biobank. METHODS: This study included 398,241 participants from the UK Biobank, 128,456 of whom participated the 7-y online mental health survey. A total of 345,876 individuals were free of depression and anxiety at baseline; of those, 16,185 developed incident mental disorders during a median of 8.7 y of follow-up. Depression and anxiety were assessed using hospital admission records and mental health questionnaires. Associations of air pollution with prevalent and incident mental disorders were examined using logistic regression and Cox regression models, respectively. RESULTS: Elevated levels of the five air pollutants were associated with higher odds of mental disorders at baseline. Levels of four pollutants but not PMcoarse were also associated with higher odds and risks of mental disorders during follow-up; specifically, hazard ratios [HR, 95% confidence interval (CI)] of an interquartile range increase in PM2:5, PM10, NOx, and NO2 for incident mental disorders were 1.03 (95% CI: 1.01, 1.05), 1.06 (95% CI: 1.04, 1.08), 1.03 (95% CI: 1.01, 1.05), and 1.06 (95% CI: 1.04, 1.09), respectively. An air pollution index reflecting combined effects of pollutants also demonstrated a positive association with the risk of mental disorders. HR (95% CI) of incident mental disorders were 1.11 (95% CI: 1.05, 1.18) in the highest quintile group in comparison with the lowest quintile of the air pollution index. We further observed that the associations between air pollution and mental disorders differed by a genetic risk score based on single nucleotide polymorphisms previously associated with genetic susceptibility to mental disorders in the UK Biobank cohort. DISCUSSION: To our knowledge, this research is one of the largest cohort studies that demonstrates an association between mental health disorders and exposure to long-term air pollution, which could be further enhanced by genetic predisposition.

Headnote

Background: Depression and anxiety are two mental disorders that are often comorbid. However, the associations of long-term air pollution expo-sure with depression and anxiety remain inconclusive.

Objective: We conducted a cross-sectional and prospective study to examine the associations of ambient exposure to particulate matter (PM) with a diameter of =2:5 lm (PM2:5), =10 lm (PM10), and 2:5-10 lm (PMcoarse), nitrogen oxides (NOx), and nitrogen dioxide (NO2) with the risk of depres-sion and anxiety in the UK Biobank.

Methods: This study included 398,241 participants from the UK Biobank, 128,456 of whom participated the 7-y online mental health survey. A total of 345,876 individuals were free of depression and anxiety at baseline; of those, 16,185 developed incident mental disorders during a median of 8.7 y of follow-up. Depression and anxiety were assessed using hospital admission records and mental health questionnaires. Associations of air pollution with prevalent and incident mental disorders were examined using logistic regression and Cox regression models, respectively.

Results: Elevated levels of the five air pollutants were associated with higher odds of mental disorders at baseline. Levels of four pollutants but not PMcoarse were also associated with higher odds and risks of mental disorders during follow-up; specifically, hazard ratios [HR, 95% confidence interval (CI)] of an interquartile range increase in PM2:5, PM10, NOx, and NO2 for incident mental disorders were 1.03 (95% CI: 1.01, 1.05), 1.06 (95% CI: 1.04, 1.08), 1.03 (95% CI: 1.01, 1.05), and 1.06 (95% CI: 1.04, 1.09), respectively. An air pollution index reflecting combined effects of pollutants also demonstrated a positive association with the risk of mental disorders. HR (95% CI) of incident mental disorders were 1.11 (95% CI: 1.05, 1.18) in the highest quintile group in comparison with the lowest quintile of the air pollution index. We further observed that the associations between air pollution and mental disorders differed by a genetic risk score based on single nucleotide polymorphisms previously associated with genetic suscepti-bility to mental disorders in the UK Biobank cohort.

Discussion: To our knowledge, this research is one of the largest cohort studies that demonstrates an association between mental health disorders and exposure to long-term air pollution, which could be further enhanced by genetic predisposition. https://doi.org/10.1289/EHP10391

Introduction

Depression and anxiety are common mental health disorders that often co-occur and can broadly impact the longevity by promoting aging-related conditions.1,2 With the rapidly increasing preva-lence of depression and anxiety, as well as their unclear etiology, the task of identifying modifiable risk factors for mental disorders and developing relevant prevention strategies has drawn increas-ing attention.3,4 Recently, mounting numbers of epidemiological studies have examined the associations between exposure to am-bient air pollutants and the risk of depression and anxiety.5-8 However, although the short-term impact of air pollution on mental disorders, especially depression, has been widely studied, results of previous population-based studies on the long-term impact of air pollution exposure remain controversial due to het-erogeneities of study designs.5,6,8 Several studies were cross-sectional or crossover analyses with limited time of follow-up at the city level and focused on only particulate matter (PM) with a diameter of =2:5 lm (PM2:5) and/or =10 lm (PM10).8-10 Other studies implemented insurance data without personal assessment using questionnaires,11 which may underestimate the influence of air pollution on mental health because people with mild syndromes are not inclined to seek insurance services. A study by Newbury et al. employed the data of mental health service use as the surrogate for the prevalence of mental disorders and specifi-cally examined the relapse of disorders, but similar underestima-tion of mild syndromes might exist.12 Therefore, it is worthwhile to study the associations of long-term exposure to ambient air pollution with depression and anxiety in a large independent study with a prospective design and comprehensive assessments of depression and anxiety.

Notably, previous studies addressing the associations between air pollution and the risk of depression or anxiety also mainly focused on the individual effect of each air pollutant.5-8 Considering that am-bient air pollution is a mixture of gaseous pollutants and particles,13 exploring their synergistic effects of on mental disorders is critical to uncover the holistic nature of air pollution on mental health in real-world settings, which, to our knowledge, have not been assessed in a prospective cohort study. Additionally, mental health disorders have been commonly recognized to be influ-enced by genetic and environmental factors.14,15 Recently, evi-dence has suggested that the gene-environment interplay could influence the development of health disorders,16 but the knowl-edge on whether the genetic susceptibility could modify the asso-ciation between the combined exposure to multiple air pollutants and mental disorders is scarce.

Hence, we investigated the associations of five major air pollu-tants, including PM2:5, PM10, PM with diameters of 2:5-10 lm (PMcoarse), and nitrogen oxides (NOx and NO2), with the preva-lence and incidence of depression and anxiety with a prospective design. This study was based on data of ? 0:4 million UK Biobank participants enrolled in a nationwide population-based cohort study in the UK with a median follow-up of 8.7 y. Leveraging the detailed records of air pollution and genetic variations of UK Biobank, we constructed an air pollution index to examine to what extent exposure to multiple air pollutants was associated with a greater burden of depression and/or anxiety and assessed the possi-ble gene-air pollution interplay.

Methods

Study Design and Population

Study design of UK Biobank was reported in detail previously.17 Briefly, UK Biobank is an ongoing prospective study with 502,536 participants recruited in 2006-2010 at the age of 37-73 y old (baseline survey) with multiple follow-ups. At baseline sur-vey, participants were asked to attend 22 examination centers located in the UK to provide information on their lifestyle and health, and their biological samples were collected. As shown in Figure S1, in this study, we included 398,241 participants with available data on mental health status, air pollution, and genetic variants at baseline to analyze the association between air pollution and prevalent mental disorders. Two subsets were subse-quently used to conduct further analyses; accordingly, from 2016 to 2017, a subset with 128,456 participants reported their detailed mental health status via an online 7-y survey and were used to evaluate the prospective associations between baseline air pollution and the detailed syndromes of depression or anxiety at the 7-y survey. A second subset with 345,876 participants who were free of depression or anxiety at baseline was employed to assess the association between air pollution and the incidence of mental disorders diagnosed by the 7-y survey and/or hospital records (see details below). UK Biobank research has received the approval from the North West Multicenter Research Ethical Committee. Written informed consents were provided by all participants.

Assessment of Depression and Anxiety

The assessments of depression and anxiety were conducted using both linked hospital admission records (UK Biobank Data-Fields: 41202 and 41204) and mental health questionnaires at baseline and follow-up. Participants were considered as depression or anxiety positive if either record was positive. For hospital records, participants were classified as cases if they had either an ICD-9/ 10 primary or secondary diagnosis for depression (ICD-9: 311; ICD-10: F32-F33) and/or anxiety (ICD-9: 300; ICD-10: F40-F41). Mental disorder was defined as participants with depression and/or anxiety. The end date of follow-up was 31 December 2018. For participants who were free of mental disorders at baseline, especially for those included in both subset analyses (n = 116,039), their follow-up time was estimated based on the dates of hospital records of incident mental disorder, the dates of the 7-y survey with a positive classification of depression/anxi-ety, or the date of censoring, whichever occurred first.

For questionnaire screening, at baseline, depression and anxiety symptoms were assessed only by Patient Health Questionnaire (PHQ)-4 questionnaire.18 Participants were required to rate, on a four-point Likert scale from 0 (not at all) to 3 (nearly every day), their response to four items: a) "frequency of depressed mood in the last 2 weeks" (UK Biobank Data-Field: 2050), b) "frequency of unenthusiasm/disinterest in the last 2 weeks" (UK Biobank Data-Field: 2060), c) "frequency of tenseness/restlessness in the last 2 weeks" (UK Biobank Data-Field: 2070), and d) "frequency of tiredness/lethargy in the last 2 weeks" (UK Biobank Data-Field: 2080). Total score ranged from 0 to 12, and a score of >6 was considered emotional disorder positive. A total score of >3 for items 1 and 2 was considered as positive for depression, and a total score of >3 for items 3 and 4 was considered as positive for anxiety based on the reported criteria.18

At the 7-y survey, mental health status was assessed using PHQ-9 (an updated PHQ-4)19 and Generalized Anxiety Disorder (GAD)-7 questionnaires20 between 2016 and 2017 with the same four-point Likert scale from 0 (not at all) to 3 (nearly every day). PHQ-9 consists of 9 items for thoughts and feelings in the last

2 wk: a) "recent thoughts of suicide or self-harm" (Suicidal idea-tion; UK Biobank Data-Field: 20513), b) "trouble falling or stay-ing asleep, or sleeping too much" (Sleeping problems; UK Biobank Data-Field: 20517), c) "recent changes in speed/amount of moving or speaking" (Psychomotor changes; UK Biobank Data-Field: 20518), d) "recent feelings of inadequacy" (Feelings of inadequacy; UK Biobank Data-Field: 20507). e) "recent feelings of tiredness or low energy" (Fatigue; UK Biobank Data-Field: 20519), f) "recent feelings of depression" (Depressed mood; UK Biobank Data-Field: 20510), g) "recent trouble con-centrating on things" (Cognitive problems; UK Biobank Data-Field: 20508), h) "recent poor appetite or overeating" (Appetite changes; UK Biobank Data-Field: 20511), and i) "recent lack of interest or pleasure in doing things" (Anhedonia; UK Biobank Data-Field: 20514). GAD-7 consists of seven items for thoughts and feelings in the last 2 wk: a) "recent inability to stop or control worrying" (Worrying control; UK Biobank Data-Field: 20509), b) "recent restlessness" (Restlessness; UK Biobank Data-Field: 20516), c) "recent trouble relaxing" (Lack of relaxation; UK Biobank Data-Field: 20515), d) "recent easy annoyance or irri-tability" (Irritability; UK Biobank Data-Field: 20505), e) "recent worrying too much about different things" (Generalized worrying; UK Biobank Data-Field: 20520), f) "recent feelings of foreboding" (Foreboding; UK Biobank Data-Field: 20512), and g) "recent feelings of nervousness or anxiety" (Anxiety feeling; UK Biobank Data-Field: 20506). Any item with a score of >1 was considered as positive for this symptom. A PHQ-9 or GAD-7 total score of >10 was considered as depression or anxiety symptoms positive according to the corresponding criteria,19,20 respectively.

Exposure Assessments

As previously described,21 the 1-y moving average concentrations of PM2.5, PMcoarse , PM10, NO2, and NOx of UK Biobank were cal-culated based on a Land Use Regression (LUR) model developed by the ESCAPE project. LUR models calculate the annual moving average concentrations of air pollutants using the predictor variables retrieved from the GIS variables, including land use, traffic, and topography by a 100 m × 100 m resolution. Participants' ambi-ent air pollution concentrations were then assigned according to their residential coordinates in the 100 m × 100 m grid cells. Previous reports using leave-one-out cross-validation demonstrated good model performance for PM2.5, PM10, NO2 and NOx (R2 of cross-validation = 77%, 88%, 87%, and 88%, respectively) and a moderate performance for PMcoarse (cross-validation R2 = 57%) in the southeast England area, where a majority of the participants came from (London/Oxford). Therefore, participants living in northern England and Scotland (n = 41,313; Figure S1) were elimi-nated from the PM analyses. Details of the ESCAPE LUR models have been described previously.22,23 The LUR estimates of PM were valid within a 400-km area from Greater London for participants living in this area. The 1-y moving average ambient concentrations of PM2.5 , PMcoarse , and NOx were collected in 2010 only, but because 1-y moving average concentrations of NO2 and PM10 were available for several years (2005, 2006, 2007, and 2010 for NO2 and 2007 and 2010 for PM10), their averaged values were included in our analysis.

Measurements of Covariates

To control for potential bias resulting from covariates that were reported to be related to mental health or air pollution,6,11,24,25 we included the following, reported at baseline: age (year; contin-uous), sex (male/female; categorical), body mass index (BMI, kg/m2; continuous), race (based on UK Biobank Data-Field 1657 "self-reported ethnic group," categorized into White, Black, Asian, and Other; Other included those reported "White and Black mixed" and "other ethnic group" to the question), smoking status (current/former/never; categorical), healthy alcohol intake status (yes/no; categorical), healthy physical activity status (yes/ no; categorical), years of education (=10 y and >10 y; categori-cal), forced respiratory volume in the first second (FEV1, liter; continuous), forced vital capacity (FVC, liter; continuous), Townsend deprivation index (continuous), live in urban area (yes/no; categorical), and prevalent hypertension, coronary heart disease (CHD), and diabetes (yes/no; categorical). Height and weight were measured by nurses during the baseline assessment visit, and BMI was calculated by dividing weight (kilograms) by the square of height (meters). Healthy alcohol intake was defined as: <28 g=day for male and <14 g=day for female. Healthy physi-cal activity status was defined as: =150 min=wk moderate or =75 min=wk vigorous or 150 min/wk mixed (moderate + vigorous) activity. The activity was assessed using the Metabolic Equivalent Task (MET) minutes based on the short International Physical Activity Questionnaire (IPAQ).26 FEV1 and FVC were measured at baseline via spirometry. Townsend deprivation index ranged from -6:3 to 10.2 and was constructed based on the percentages of four key variables in the UK Biobank according to previously reported algorithm,27 unemployment, overcrowded household, non-car ownership, and non-home ownership. The percentages for each area were based on postcodes and census data at the year of baseline visit, with each participant assigned percentages corresponding to the postcode of their home dwell-ing but not based on individual-level factors. The percentages of unemployment and overcrowded household were normalized by ln (percentagevalue+1) because the two were skewed. The four percentages were then z-scored and summarized as the Townsend deprivation index. A higher index indicates a higher level of deprivation. The history of hypertension, CHD, and diabetes was based on self-reported information and medical records. Because all covariates were missing <2% (Table 1), we imputed continuous variables with sex-specific mean values and implemented a missing indicator approach for the categori-cal variables.28

Genetic Risk Scores for Mental Disorders

Details of the genotyping, data imputation, and quality control in UK Biobank have been previously reported.17,29 Briefly, ?10% of participants were genotyped using the Applied Biosystems UK BiLEVE Axiom Array by Affymetrix (807,411 sites), with the remaining participants being genotyped using the Applied Biosystems UK Biobank Axiom Array (825,927 sites), both of which were specifically designed for the UK Biobank with 95% shared sites. Phasing and imputation of the remaining 5% unique single-nucleotide polymorphisms (SNPs) were conducted using SHAPEIT3 and IMPUTE2 with both the merged UK10K and 1000 Genomes Phase 3 reference panel and the Haplotype Reference Consortium (HRC) reference panel (i.e., array imputa-tion). SNPs after the array imputation were considered for building a genetic risk score (GRS). We directly used the imputed genetic data without further quality control.29 Based on the avail-ability of imputed SNPs, we created a GRS for mental disorders using 37/44 depression-related SNPs and 9 out of 10 unique anxiety-related SNPs based on the two largest genome-wide association studies (GWASs).30,31 Each SNP was recoded as 0, 1, or 2 according to the number of risk alleles, and missing SNP values of individuals were imputed by corresponding mean values (i.e., individual imputation). To avoid the bias from nonlinearity, this GRS based on multiple genes was calculated and weighted by summing up after multiplying the logarithmic odds ratios (ORs) of each included SNP reported by the selected GWASs on depression or anxiety with the number of risk alleles and then di-vided by the total effect size. The GRS ranged from 0.85 to 2.70. The detailed SNPs are reported in Table 2.

Construction of Air Pollution Index

As a post hoc analysis, to account for the combined effects of multiple air pollutants on mental disorders, we created an air pollution index by excluding PMcoarse due to its weak associa-tions with mental disorders using a modified algorithm reported by a previous study21 (for details, see the "Results" section). This index is a continuous weighted score by adding concentrations of the air pollutants, weighted by the effect esti-mates (b coefficients) on the odds of mental disorders at baseline or the risk of incident mental disorders at the follow-up in the crude model, which adjusted for age and sex only. The equation of the index was: Air pollution index = (b[PM2.5] × PM2.5 + b[PM10] × PM10 + b[NO2] × NO2 + b[NOx] × NOx)/ (b[PM2.5] + b[PM10] + b[NO2] + b[NOx]) × 4, and was also categorized into five groups by quintiles.

To avoid upweighting the positive associations retrieved from this study, because the coefficients for the construction of the index were from the same cohort, we validated it with two approaches. First, we conducted a 10-fold cross-validation analysis that was conducted in a previous air pollution-related study based on the UK Biobank.32,33 UK Biobank data were randomly divided into 10 batches. One of the batches was testing data, and the other nine were the training data in each run. Logistic and Cox regression models were applied to obtain regression coefficients to build the air pollution index in the training data, and the constructed index was then tested in the testing data. After being repeated 10 times, the ORs and HRs were then polled using a fixed-effect meta-analysis to obtain the comprehensive ORs and

HRs.34 Similar estimates suggested that this air pollution index was robust in the present study (Table S1). Furthermore, we gen-erated a quantile score without taking the prior weights of each air pollutant into consideration. Ambient concentration within the fourth quartile of each air pollutant except PMcorase was defined as "extreme" exposure to corresponding pollutants. This score was then determined as the sum of "extreme" exposed air pollu-tants.Thisscoreandtheairpollutionindexwerehighly correlated with each other (Table S2), suggesting that the bias from up-weighting the positive associations in the construction of air pollution index was quite limited.

Statistical Analyses

Sociodemographic and lifestyle factors and mental health out-comes for all participants, participants with the mental health survey data at the 7-y survey, and participants who were free of depression or anxiety were summarized using descriptive statistics.

We first examined the associations of the five air pollutants with the total PHQ-4 score using mixed-effect linear regression models and the odds of mental disorders, depression, and anxiety at baseline using logistic regression models to evaluate the cross-sectional association between air pollution and mental disorders. We adjusted for age, sex, BMI, race, smoking status, healthy alcohol intake status, healthy physical activity status, years of education, FEV1 , FVC, Townsend deprivation index, residence in an urban area, and prevalent hypertension, CHD, and diabetes, and we controlled for the examination center as a random effect in the model to account for the potential clustering bias from health examinations. Yielded estimates were reported as changes per one interquartile range (IQR) increase in the ambient concen-tration of each air pollutant. Two forms of air pollution index for baseline analysis were built accordingly, and the corresponding dose-response curves of air pollutants and the index with PHQ-4 score were further assessed by restricted cubic spline regression models controlling for previously described covariates. The 5th, 50th, and 95th percentiles of each air pollutant and air pollution index were selected as knots.

Then, in the subset with 7-y survey data, we examined the prospective associations between the baseline air pollution levels and the odds of mental disorders and corresponding symptoms of each disorder (logistic regression) and the scores of PHQ-9 and GAD-7 (linear regression) at the 7-y survey. Models adjusted for covariates described in the previous section and the status of mental disorders (yes/no) at baseline.

Furthermore, the Cox proportional hazards model was used to assess the prospective associations between baseline air pollution levels and the incidence of mental disorders during the follow-up among participants free of depression and anxiety at baseline. Corresponding air pollution index for the joint association was constructed accordingly, and its dose-response relationships with the risk of incident mental disorders, depression, and anxiety were also assessed by restricted cubic spline regression. The 5th, 50th, and 95th percentiles of the air pollution index were selected as knots.

Finally, to evaluate whether the genetic predisposition of mental disorders may modify the association of air pollution index with the odds of mental disorders at baseline and the risk of mental disorders at the follow-up, we tested the gene-air pollu-tion interaction by employing the interaction terms of the air pol-lution index with the GRS of mental disorders in the models for the prevalence and incidences of depression and anxiety. In the case that interaction effect did not meet criteria for statistical sig-nificance, we generated a category variable based on the quintiles of air pollution index, and dichotomized GRS (by median) was further generated to show the joint association of both factors with mental disorders. For all primary analyses, we used the E-values and the lower confidence interval to indicate the robust-ness to unmeasured confounding.35 This E-value could test the robustness of an association between exposure and outcome and to evaluate evidence for causation. The unmeasured confounders needed to be associated with both air pollution and mental disorders with a risk ratio of the E-value to fully explain the observed associations between air pollutants and mental disorders.

Additionally, three sensitivity analyses were performed to validate the robustness of our primary findings. We first tested the estimates of air pollutants in relation to incident mental disorders in a model adjusting for PM2:5, PM10, NOx, and NO2 to examine the copollutant cofounding. Then, we examined the associations between air pollutants and the concurrent incidence of depression and anxiety. Last, we conducted another analysis using the primary model in participants who self-reported at the baseline survey that they had been living at the same baseline address for more than 5 y to evaluate whether the mobilization may affect the primary findings.

SAS version 9.4 TS1M7 (SAS Institute Inc.) was used to clean data and conduct analyses. A two-sided p-value <0:05 was considered statistically significant.

Results

Participants' Characteristics and Distributions of Air Pollutants

Characteristics of the study population and the subset of par-ticipants free of mental disorders at baseline were similar with respect to most variables but slightly different in the subset of people with 7-y survey data (Table 1). For the whole popula-tion, participants' age [mean ± standard deviation (SD)] was 56:6 ± 8:1 y and ? 95% (n= 379,373) were White. About 35% (n= 217,167) and 55% (n=139,671) of participants were former and never smokers, respectively. A majority of participants had healthy physical activity (70.4%, n= 280,132), >10 y of educa-tion (66.1%, n= 263,065), and lived in an urban area (84.4%, n= 336,257). Half reported a healthy daily intake of alcohol (n =198,922). About 55% (n=220,764), 6% (n =22,136), and 5% (n = 20,288) had prevalent hypertension, CHD, or diabetes diagnosed by doctors, respectively, according to self-reported information and medical records. The average PHQ-4 score was 1:60 ± 2:10, and about 13% (n= 52,365), 7% (n=25,715), and 11% (n = 41,783) were with prevalent mental disorders, depression, and anxiety, respectively, according to the PHQ-4 score definition or hospital records. The subset with 7-y survey data had a higher proportion of people with >10 y of education (79.4%, n = 101,925) than did the total cohort or subset free of mental disorders at baseline; about 10% (n= 12,417), 4% (n =5,368), and 8% (n =10,032) of the participants in this subgroup had prevalent mental disorders, depression, and anxiety, respectively. The 345,876 participants that were free of mental disorders at baseline were followed for a me-dian time of 8.7 y, and 16,185 of them (4.68%) developed mental disorders (ndepression = 10,854; nanxiety = 8,257).

Average concentrations of air pollutants at baseline for the total population were 9:97± 1:06 lg=m3 (IQR= 1:27) for PM2:5, 6:42±0:90 lg=m3 (IQR=0:79) for PMcoarse, 19:24±2:02lg=m3 (IQR= 2:33) for PM10, 43:68± 15:54 lg=m3 (IQR= 16:53) for NOx, and 29:01±9:17lg=m3 (IQR=10:87) for NO2. Ambient concentrations of all air pollutants were highly correlated with each other (Table 3).

Associations of Air Pollution with the Odds of Mental Disorders at Baseline and 7-y Survey

As shown in Table 4, after adjusting for potential covariates, the ambient levels of four air pollutants except PMcoarse showed strong associations with the PHQ-4 score and the odds of mental disorders. Particularly, an IQR increment of PM2:5 level was associated with a 0.051-unit [standard error ðSEÞ = 0:005] increase in PHQ-4 score, and with the ORs (95% CI) of 1.07 (95% CI: 1.06, 1.09), 1.08 (95% CI: 1.07, 1.10), and 1.07 (95% CI: 1.05, 1.08) for mental disorders, depression, and anxiety, respectively. To further estimate the combined associations of the five air pollutants with mental health at baseline, we constructed an air pollution index based on the coefficients of the five pollu-tants for the odds of mental disorders at baseline retrieved from a crude model adjusting for sex and age only to avoid overadjust-ment bias in the primary analysis models that controlled for all potential covariates. Both continuous and categorical forms of the index were associated with a higher PHQ-4 score and odds of mental disorders in the fully adjusted model. For instance, the ORs of mental disorders were 1.04 (95% CI: 1.00, 1.08), 1.11 (95% CI: 1.07, 1.14), 1.15 (95% CI: 1.11, 1.19), and 1.16 (95% CI: 1.12, 1.20) for second-fifth quintile groups, respectively, in com-parison with the first quintile of the air pollution index (Table 4). Except PMcoarse, the other four air pollutants and corresponding air pollution index demonstrated monotonic increasing dose-response relationships with the PHQ-4 score (Figure 1).

A subset of 128,456 participants took the 7-y online mental health survey; therefore, we examined the prospective associa-tions of ambient air pollution with the odds of mental disorders and symptoms of depression and anxiety at the 7-y survey. Figure 2 (Table S3) shows that baseline PM2:5, PM10, NOx, and

NO2 were positively associated with most depression and anxiety symptoms, odds of mental disorders, and the scores (total, PHQ-9, and GAD-7 scores) of mental health status at the 7-y survey. The baseline concentration of PMcoarse was not associated with any symptoms or scores. This finding suggests that long-term ex-posure to PM2:5, PM10, NOx , and NO2 was associated with the prevalence of mental disorders.

Associations of Air Pollution with the Incidence of Mental Disorders

The fully adjusted Cox regression model showed the associations of baseline air pollution levels with the incidence of mental disorders for the 345,876 participants free of mental disorders at baseline during a median follow-up of 8.7 y (Table 5). PM10 and NO2 showed strong associations with the risk of incident mental disorders. An IQR increment of PM10 level at baseline was associated with 6% higher risk of mental disorders (95% CI: 1.04, 1.08), 5% higher risk of depression (95% CI: 1.02, 1.07), 7% higher risk of anxiety (95% CI: 1.04, 1.10). An IQR increase in NO2 was associated with similar risks of mental disorders and each disorder but was associated with relatively lower risks with respect to PM2:5 and NOx. PMcoarse level was not associated with any of the health outcomes. The air pollution index was positively associated with the risk of mental disorders. For instance, HRs (95% CI) of the risk of mental disorders were 1.06 (95% CI: 1.00, 1.12), 1.07 (95% CI: 1.01, 1.13), 1.10 (95% CI: 1.04, 1.16), and 1.11 (95% CI: 1.05, 1.18) for second- fifth quintiles, respectively, in comparison with the lowest quintile of the air pollution index (Table 5). Dose-response relationships between the air pollution index and the risk of mental disorders demonstrated monotonically increasing pat-terns (Figure 3).

Sensitivity Analyses

In the model that simultaneously included the four pollutants (except PMcoarse), NOx and NO2 remained associated with incident mental disorders, whereas PM2:5 became null (p = 0:06) and PM10 demonstrated attenuated estimates (Table S4). Furthermore, considering that depression or anxiety may exac-erbate mental health impairment and thus cause the other, we assessed the association of air pollution with concomitant depression and anxiety at follow-up. A total of 2,926 partici-pants had both incident depression and anxiety during the follow-up. PM2:5 (HR=1:02; 95% CI: 1.00, 1.04, p= 0:040), PM10 . (HR= 1:04; 95% CI: 1.01, 1.09, p = 0:016), NOx (HR=1:03; 95% CI: 1.01, 1.07, p=0:009), and NO2. (HR= 1:06; 95% CI: 1.01, 1.10, p = 0:0007) were statistically associated with the concomitant incidence, and the highest quintile of the air pollution index was associated with a 13% higher risk of the concomitant incidence than the lowest but was not significant (Table S5, pQ5 of air pollution index = 0:05, 95% CI: 1.00, 1.28). Another sensitivity analysis of the 321,090 par-ticipants living at the address for more than 5 y also demon-strated robust associations of the air pollution index with prevalent mental disorders (OR =1:05; 95% CI: 1.04, 1.07, p< 0:0001) and the incidence of mental disorders (HR= 1:03; 95% CI: 1.01, 1.06, p= 0:0015), which suggests that the mobil-ity of participants may affect our main findings minimally (Table S6).

Impact of Genetic Susceptibility on the Associations between Air Pollution and Mental Disorders

We tested whether the genetic background could modify the associ-ations between air pollution and mental disorders. The GRS we gen-erated was associated with the PHQ-4 score (Coefficient = 0:024, SE = 0:003), odds of mental disorders at baseline (OR = 1:03; 95% CI: 1.02, 1.04), and the incident mental disorders during the follow-up (HR = 1:04; 95% CI: 1.02, 1.06) (Table S7). However, the inter-action between GRS and the air pollution index was not signifi-cantly associated with prevalence of mental disorders at baseline (Coefficient of interaction term = - 0:0001, SE = 0:0050, p-interaction = 0:84) or their incidence (Coefficient of interaction term = 0:011, SE = 0:0085, p-interaction = 0:19) during the follow-up. We also noted that estimates of the air pollution index were essentially unchanged in mutual adjustment models (Table S8). These findings suggested that both the GRS and air pollution index could be inde-pendently associated with the prevalence and risk of mental disorders.

Given that no interaction was observed, we tested the joint effect of both air pollution and genetic susceptibility using the quintiles of air pollution index and the dichotomized GRS. We observed that participants with higher GRS and air pollution index had higher odds of mental disorders and risk of mental disorders (Figure 4; Table S9). Participants with higher GRS and the fifth quintile of air pollution index had ?20% higher odds of mental disorders at baseline (OR= 1:20; 95% CI: 1.15, 1.26) and about 15% greater risk of incident mental disorders during the follow-up (HR= 1:15; 95% CI: 1.06, 1.23) than participants with the low GRS and first quintile of the air pollution index.

Discussion

In this prospective study using the UK Biobank cohort, we eluci-dated the individual and joint associations of five major air pollu-tants with depression and anxiety, the two most common mental disorders. We identified that baseline exposures to PM2:5, PM10, NOx, and NO2 were not only associated with the prevalent mental disorders at baseline but were also associated with the risk of mental disorders during a median follow-up of 8.7 y. Combined exposure of the four pollutants represented by an air pollution index was robustly associated with elevated risks of incident mental disorders, depression, and anxiety during the follow-up separately. For the results of primary analyses demonstrated in Table 4 and Table 5, the relatively large E-values indicate that they were robust. We found that such adverse effects of air pollution could be enhanced by the genetic susceptibility to mental disorders despite the fact that no robust interaction between the air pollution index and the GRS was observed.

To the best of our knowledge, this work is the one of the larg-est population-based cohort studies that provides evidence on the association between air pollution and mental disorders, especially among older adults. Our study suggests that an IQR increase in long-term exposure to PM2:5 was associated with more than 8% increases in the odds of depression at baseline and 1% increases in its risks. The IQR of PM2:5 was 1:27 lg=m3 in our study, which is higher than the findings of a systematic review and meta-analysis by Braithwaite et al. reporting 10% higher pooled odds of depression per 10 ug=m3 PM2:5.8 A potential explanation is the approaches to access depression of the selected publications in this review; we noted that the majority of studies used data of self-reported medical records and/or emergency visits, which only captured the information of severe or intense depression, because individuals with light or moderate depression syndromes may ignore their mental status and may not visit doctors. Moreover, the yielded risk of incident depression in our partici-pants free of mental disorders at baseline could be ?8% per 10 lg=m3 increment in PM2:5 level, which was much lower than the findings of Liu et al.'s meta-analysis (18%).5 This difference could be explained by the characteristics of the population of the UK Biobank, which is a volunteer cohort, and participants are likely healthier with fewer self-reported health disorders and live in less socioeconomically deprived areas than the general population.36 Both may limit the effect of air pollution on incident mental disorders because healthier participants and a better community could provide better health resilience to mitigate health risks.37 Reported associations between PM10 and depres-sion have been similar in previous studies,24 which is in line with our findings. Furthermore, our study found relatively stronger long-term positive associations of NOx and NO2 with depression than PM2:5, which was consistent with the study of Bakolis et al. showing that NO2 and NOx had relatively stronger associations with common mental disorders than PM2:5.10 In comparison with depression, the impact of air pollution on anxiety has been much less frequently reported. Our study also reported the associations of long-term exposure not only to PM2:5 but also to PM10, NOx, and NO2 with anxiety. Except the study of Power et al., in which the authors found a positive impact of PM2:5 on prevalent anxiety,38 most of the small number of existing studies only explored the associations of short-term exposure to air pollutants with anxiety and yielded null findings.7

In our study, estimates of the associations of NOx. and NO2 with mental disorders were higher than that of PM2:5 even in the model controlling for all pollutants simultaneously (Table S4), which is in line with previous findings10,12 and suggests that PM2:5 may not be the largest contributors to the development of mental disorders, though NO2 shares some sources with PM2:5. Such robust effects of long-term NOx and NO2. exposure were in agreement with the associations of short-term exposure of NO2 with depression in a recent meta-analysis.6 PMcoarse was the only pollutant that demonstrated a nonlinear association with the mental health PHQ-4 score at baseline. It was not associated with the incidence of mental disorders or most depression and anxiety symptoms during the follow-up, which agrees with the previous statement on the health effects of PMcoarse,39 suggesting that the health implications of PMcoarse remain less characterized than those for PM2:5. Due to such heterogeneities of the asso-ciations between various air pollutants and mental health, we attempted to depict the combined associations of relevant air pollutants by constructing the air pollution index. It was associated with the risk of mental disorders and the comorbidity of depression and anxiety and further suggested that it could be featured as a more comprehensive measure of the effects of multiple air pollutants on mental health. A similar algorithm has been implemented in previous epidemiological studies not only including environmental exposures,21,40 but also covering lifestyle and nutrition.41 Such a simple score may make the results of epidemiological studies much easier for nonpro-fessionals to interpret and understand.

Biological mechanisms underlying the association between ambient air pollution exposure and mental disorder remain incon-clusive. Inflammation and oxidative stress are two commonly suggested biological pathways. Neuroinflammation caused by air pollution may deplete the brain serotonin, dysregulate the hypothalamus-pituitary-adrenal axis, and interfere with the production of neurons in the dentate gyrus of the hippocampus.42 Oxidative stress further contributes to the degradation of dopami-nergic neurons, which is likely involved in the neuropathology of depression and additionally intimately impairs mitochondrial function, and enhances excitotoxicity and nitric oxide toxicity.43 Moreover, neuroimaging-based epidemiological studies revealed that long-term exposure to air pollution may be associated with changes in brain structures and functioning in white matter, corti-cal gray matter, and basal ganglia,25 which may intensify the pro-gression of mental disorders. Additionally, the indirect impact of air pollution on human health status, including residential dissat-isfaction,44,45 poor self-perceived health and life quality,46 frailty,47 as well as increased risk of major chronic diseases, such as cardiovascular diseases, respiratory diseases, dementia, and cancers,16,48,49 may be taken into account because they could trigger the incidence of depression and anxiety by bringing increasing psychological burden, especially for older adults. A more intriguing aspect is that we found that the combined associ-ations of major air pollutants with mental health could be moder-ately enhanced by a GRS based on previous established SNPs of mental disorders. A plausible explanation of the observed lack of interaction between the air pollution index and GRS could be that parts of the SNPs we used to build the GRS were related to the de-velopment of neuro systems (e.g., RBFOX1 and PTPRT) or regula-tion of intercellular signaling in the brain (e.g., PDE4B),50-52 which may not directly interact with neuroinflammation or oxida-tive stress that may result from long-term exposure to air pollution. Nevertheless, given the risk that loci of mental health may soon be revealed by emerging genome analyses in the future, other genetic variants that play critical roles in modifying the risks of mental health resulting from air pollution may be found.

Our study has several strengths, including the large sample size, the detailed and repeated measurements of mental health status, and data available on a wide range of potential covariates and genetic variants. Several limitations are worthwhile to note when interpreting the findings. First, as a volunteer cohort, partic-ipants in the UK Biobank cohort were likely healthier than the general population, which may limit the impact of air pollution, because relatively healthier participants may suffer from less psy-chosocial pressure. Early-life experience of mental health problems may also alter a person's tendency to live in bad neighborhoods, which often have worse air quality.53 Future studies with more detailed records of mental health status since childhood are war-ranted to explore this research question. Furthermore, the measure-ment bias of air pollution must be acknowledged. Our air pollution data were mostly only a single measurement of the annual average outdoor air pollution level in 2010 without any further measures during the follow-up. The baseline visit of the UK Biobank was conducted from 2006 to 2010, which prohibited our further explora-tion on the lag or short-term (<1 month) effects of air pollution on depression and anxiety. Also, because people, especially older adults, spent most of their time indoors,54 individual exposure to each air pollutant may differ from that indicated by the outdoor am-bient concentrations we examined. Additionally, the impacts of the components of PM2:5 and other major air pollutants such as ozone and sulfur dioxide on depression and anxiety are worth investigation in future studies. For the air pollution index, although it was internally validated in our study as introduced in the "Methods" section, future studies are required to use this approach in other independent cohorts. Additionally, given only about one-third participants joined in the 7-y survey and we observed a larger proportion of people with mental disorders in this subset than the whole population without any form of mental disorders at baseline, the real-world burden of depression and anxiety during the follow-up could be worse than that based on the current information. Last, because the precise timing of the onset of mental disorders was lacking, both conditions together may weaken the observed associations of air pollutants with the risk of mental disorders in our study by causing potential underesti-mation of incident mental disorders and overestimation of cen-soring time.

In conclusion, our study is one of the largest cohort studies investigating the associations of long-term exposure to major air pollutants with the risk of depression and anxiety in ~ 0.4 million people. Our findings established that air quality may be a modifiable factor that could be targeted to reduce the mental health burden in modern society. Beyond this, our study not only underpinned the combined associations of air pollutants with mental disorders using a novel air pollution index, but also highlighted a potential dose-response in the develop-ment of depression and anxiety under long-term air pollution exposure. More efforts are warranted to reveal the whole landscape of the impact of environmental exposures to additional air pollutants on mental health and to explore the underlying bi-ological mechanisms in a multidisciplinary approach.

Acknowledgments

Data are available in a public, open-access repository per application (https://www.ukbiobank.ac.uk/). This research has been conducted using the UK Biobank Resource under Application Number 44430. The UK Biobank data are available on application to the UK Biobank.

X.G. was supported by grants from the National Key Research and Development Program of China (No. 2022YFC3702704), the China Center for Disease Control Key Laboratory of Environment and Population Health (2022-CKL-03), and Peking University (BMU2021YJ044). The authors thank C Chen for the language assistance.

References

References

1.Verhoeven JE, Révész D, van Oppen P, Epel ES, Wolkowitz OM, Penninx BW. 2015. Anxiety disorders and accelerated cellular ageing. Br J Psychiatry 206(5):371-378, PMID: 25657360, https://doi.org/10.1192/bjp.bp.114.151027.

2.Révész D, Verhoeven JE, Milaneschi Y, Penninx BW. 2016. Depressive and anxiety disorders and short leukocyte telomere length: mediating effects of meta-bolic stress and lifestyle factors. Psychol Med 46(11):2337 2349, PMID: 27266474, https://doi.org/10.1017/S0033291716000891.

3.Cuijpers P, Beekman AT, Reynolds CF 3rd. 2012. Preventing depression: a global priority. JAMA 307(10):1033 1034, PMID: 22416097, https://doi.org/10. 1001/jama.2012.271.

4.Beaudreau SA, O'Hara R. 2008. Late-life anxiety and cognitive impairment: a review. Am J Geriatr Psychiatry 16(10)790-803, PMID: 18827225, https://doi.org/ 10.1097/JG P.0b013e31817945c3.

5.Liu Q, Wang W, Gu X, Deng F, Wang X, Lin H, Guo X. 2021. Association between particulate matter air pollution and risk of depression and suicide: a systematic review and meta-analysis. Environ Sci Pollut Res Int 28(8):9029-9049, PMID: 33481201, https://doi.org/10.1007/s11356-021-12357-3.

6.Fan SJ, Heinrich J, Bloom MS, Zhao TY, Shi TX, Feng WR, et al. 2020. Ambient air pollution and depression: a systematic review with meta-analysis up to 2019. Sci Total Environ 701:134721, PMID: 31715478, https://doi.org/10.1016/j. scitotenv.2019.134721.

7.Dickerson AS, Wu AC, Liew Z, Weisskopf M. 2020. A scoping review of non-occupational exposures to environmental pollutants and adult depression, anxiety, and suicide. Curr Environ Health Rep 7(3):256-271, PMID: 32533365, https://doi.org/10.1007/s40572-020-00280-7.

8.Braithwaite I, Zhang S, Kirkbride JB, Osborn DPJ, Hayes JF. 2019. Air pollu-tion (particulate matter) exposure and associations with depression, anxiety, bipolar, psychosis and suicide risk: a systematic review and meta-analysis. Environ Health Perspect 127(12):126002, PMID: 31850801, https://doi.org/10. 1289/EHP4595.

9.Petrowski K, Bührer S, Strauß B, Decker O, Brähler E. 2021. Examining air pol-lution (PM10), mental health and well-being in a representative German sample. Scientific reports 11(1):18436, PMID: 34531408, https://doi.org/10.1038/s41598-021-93773-w.

10.Bakolis I, Hammoud R, Stewart R, Beevers S, Dajnak D, MacCrimmon S, Broadbent M. 2021. Mental health consequences of urban air pollution: pro-spective population-based longitudinal survey. Soc Psychiatry Psychiatr Epidemiol 56(9):1587-1599, PMID: 33097984, https://doi.org/10.1007/s00127-020-01966-x.

11.Khan A, Plana-Ripoll O, Antonsen S, Brandt J, Geels C, Landecker H, et al. 2019. Environmental pollution is associated with increased risk of psychiatric disorders in the US and Denmark. PLoS Biol 17(8):e3000353, PMID: 31430271, https://doi.org/10.1371/journal.pbio.3000353.

12.Newbury JB, Stewart R, Fisher HL, Beevers S, Dajnak D, Broadbent M, et al. 2021. Association between air pollution exposure and mental health service use among individuals with first presentations of psychotic and mood disorders: retrospective cohort study. Br J Psychiatry 219(6):678-685, PMID: 35048872, https://doi.org/10.1192/bjp.2021.119.

13.Kampa M, Castanas E. 2008. Human health effects of air pollution. Environ Pollut 151(2): 362-367, PMID: 17646040, https://doi.org/10.1016/j.envpol.2007.06.012.

14.Tsuang MT, Bar JL, Stone WS, Faraone SV. 2004. Gene-environment interac-tions in mental disorders. World Psychiatry 3(2):73-83, PMID: 16633461.

15.Robert Plomin, DeFries JC, Knopik VS, Neiderhiser JM. 2013. Behavioral Genetics (6th edition). New York, NY: Worth Publishers. https://doi.org/10.1007/ s10519-013-9598-6.

16.Peters A, Nawrot TS, Baccarelli AA. 2021. Hallmarks of environmental insults. Cell 184(6):1455-1468, PMID: 33657411, https://doi.org/10.1016/j.cell.2021.01.043.

17.Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. 2015. UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 12(3):e1001779, PMID: 25826379, https://doi.org/10.1371/journal.pmed.1001779.

18.Stanhope J. 2016. Patient Health Questionnaire-4. Occup Med (Lond) 66(9):760- 761, PMID: 27994085, https://doi.org/10.1093/occmed/kqw165.

19.Kroenke K, Spitzer RL, Williams JB. 2001. The PHQ-9: validity of a brief depres-sion severity measure. J Gen Intern Med 16(9):606-613, PMID: 11556941, https://doi.org/10.1046/j.1525-1497.2001.016009606.x.

20.Spitzer RL, Kroenke K, Williams JB, Löwe B. 2006. A brief measure for assess-ing generalized anxiety disorder: the GAD-7. Arch Intern Med 166(10):1092- 1097, PMID: 16717171, https://doi.org/10.1001/archinte.166.10.1092.

21.Wang M, Zhou T, Song Y, Li X, Ma H, Hu Y, et al. 2021. Joint exposure to vari-ous ambient air pollutants and incident heart failure: a prospective analysis in UK Biobank. Eur Heart J 42(16):1582-1591, PMID: 33527989, https://doi.org/10. 1093/eurheartj/ehaa1031.

22.Eeftens M, Beelen R, de Hoogh K, Bellander T, Cesaroni G, Cirach M, et al. 2012. Development of Land Use Regression models for PM2.5, PM2.5 absorbance, PM10 and PMcoarse in 20 European study areas; results of the ESCAPE project. Environ Sci Technol 46(20):11195-11205, PMID: 22963366, https://doi.org/10.1021/ es301948k.

23.Beelen R, Hoek G, Vienneau D, Eeftens M, Dimakopoulou K, Pedeli X, et al. 2013. Development of NO2 and NOx land use regression models for estimating air pollution exposure in 36 study areas in Europe-The ESCAPE project. Atmos Environ 72:10-23, https://doi.org/10.1016/j.atmosenv.2013.02.037.

24.Vert C, Sánchez-Benavides G, Martínez D, Gotsens X, Gramunt N, Cirach M, et al. 2017. Effect of long-term exposure to air pollution on anxiety and depres-sion in adults: a cross-sectional study. Int J Hyg Environ Health 220(6):1074- 1080, PMID: 28705430, https://doi.org/10.1016/j.ijheh.2017.06.009.

25.de Prado Bert P, Mercader EMH, Pujol J, Sunyer J, Mortamais M. 2018. The effects of air pollution on the brain: a review of studies interfacing environmen-tal epidemiology and neuroimaging. Curr Environ Health Rep 5(3):351-364, PMID: 30008171, https://doi.org/10.1007/s40572-018-0209-9.

26.Craig CL, Marshall AL, Sjöström M, Bauman AE, Booth ML, Ainsworth BE, et al. 2003. International physical activity questionnaire: 12-country reliability and va-lidity. Med Sci Sports Exerc 35(8):1381-1395, PMID: 12900694, https://doi.org/10. 1249/01.MSS.0000078924.61453.FB.

27.Townsend P, Phillimore P, Beattie A. 1988. Health and Deprivation: Inequality and the North: New York, NY: Routledge.

28.Cummings P. 2013. Missing data and multiple imputation. JAMA Pediatr 167(7):656-661, PMID: 23699969, https://doi.org/10.1001/jamapediatrics.2013. 1329.

29.Bycroft C, Freeman C, Petkova D, et al. 2017. Genome-wide genetic data on- 500,000 UK Biobank participants. bioRxiv:166298.

30.Meier SM, Trontti K, Purves KL, Als TD, Grove J, Laine M, et al. 2019. Genetic variants associated with anxiety and stress-related disorders: a genome-wide association study and mouse-model study. JAMA Psychiatry 76(9):924-932, PMID: 31116379, https://doi.org/10.1001/jamapsychiatry.2019.1119.

31.Wray NR, Ripke S, Mattheisen M, Trzaskowski M, Byrne EM, Abdellaoui A, et al. 2018. Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression. Nat Genet 50(5):668-681, PMID: 29700475, https://doi.org/10.1038/s41588-018-0090-3.

32.Simon RM, Subramanian J, Li MC, Menezes S. 2011. Using cross-validation to evaluate predictive accuracy of survival risk classifiers based on high-dimensional data. Brief Bioinform 12(3):203-214, PMID: 21324971, https://doi.org/ 10.1093/bib/bbr001.

33.Fan M, Sun D, Zhou T, Heianza Y, Lv J, Li L, et al. 2020. Sleep patterns, genetic susceptibility, and incident cardiovascular disease: a prospective study of 385 292 UK biobank participants. Eur Heart J 41(11):1182-1189, PMID: 31848595, https://doi.org/10.1093/eurheartj/ehz849.

34.Nikolakopoulou A, Mavridis D, Salanti G. 2014. How to interpret meta-analysis models: fixed effect and random effects meta-analyses. Evid Based Ment Health 17(2):64, PMID: 24778439, https://doi.org/10.1136/eb-2014-101794.

35.VanderWeele TJ, Ding P. 2017. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med 167(4):268-274, PMID: 28693043, https://doi.org/10.7326/M16-2607.

36.Fry A, Littlejohns TJ, Sudlow C, Doherty N, Adamska L, Sprosen T, et al. 2017. Comparison of sociodemographic and health-related characteristics of UK Biobank participants with those of the general population. Am J Epidemiol 186(9):1026-1034, PMID: 28641372, https://doi.org/10.1093/aje/kwx246.

37.Wulff K, Donato D, Lurie N. 2015. What is health resilience and how can we build it? Annu Rev Public Health 36:361-374, PMID: 25581148, https://doi.org/10. 1146/annurev-publhealth-031914-122829.

38.Power MC, Kioumourtzoglou MA, Hart JE, Okereke OI, Laden F, Weisskopf MG. 2015. The relation between past exposure to fine particulate air pollution and prevalent anxiety: observational cohort study. BMJ 350:h1111, PMID: 25810495, https://doi.org/10.1136/bmj.h1111.

39.Adar SD, Filigrana PA, Clements N, Peel JL. 2014. Ambient coarse particulate matter and human health: a systematic review and meta-analysis. Curr Environ Health Rep 1(3):258-274, PMID: 25152864, https://doi.org/10.1007/s40572-014-0022-z.

40.Park SK, Tao Y, Meeker JD, Harlow SD, Mukherjee B. 2014. Environmental risk score as a new tool to examine multi-pollutants in epidemiologic research: an example from the NHANES study using serum lipid levels. PLoS One 9(6): e98632, PMID: 24901996, https://doi.org/10.1371/journal.pone.0098632.

41.Arouca A, Moreno LA, Gonzalez-Gil EM, Marcos A, Widhalm K, Molnár D, et al. 2019. Diet as moderator in the association of adiposity with inflammatory bio-markers among adolescents in the HELENA study. Eur J Nutr 58(5):1947-1960, PMID: 29948222, https://doi.org/10.1007/s00394-018-1749-3.

42.Mithaiwala MN, Santana-Coelho D, Porter GA, O'Connor JC. 2021. Neuroinflammation and the kynurenine pathway in CNS disease: molecular mechanisms and therapeutic implications. Cells 10(6):1548, PMID: 34205235, https://doi.org/10.3390/cells10061548.

43.Bhatt S, Nagappa AN, Patil CR. 2020. Role of oxidative stress in depression. Drug Discov Today 25(7):1270-1276, PMID: 32404275, https://doi.org/10.1016/j. drudis.2020.05.001.

44.Liu S, Ouyang Z, Chong AM, Wang H. 2018. Neighborhood environment, resi-dential satisfaction, and depressive symptoms among older adults in residen-tial care homes. Int J Aging Hum Dev 87(3):268-288, PMID: 28942658, https://doi.org/10.1177/0091415017730812.

45.von Lindern E, Hartig T, Lercher P. 2016. Traffic-related exposures, constrained restoration, and health in the residential context. Health Place 39:92-100, PMID: 26995669, https://doi.org/10.1016/j.healthplace.2015.12.003.

46.Nguyen HA, Anderson CAM, Miracle CM, Rifkin DE. 2017. The association between depression, perceived health status, and quality of life among individ-uals with chronic kidney disease: an analysis of the National Health and Nutrition Examination Survey 2011-2012. Nephron 136(2):127-135, PMID: 28249290, https://doi.org/10.1159/000455750.

47.Vaughan L, Corbin AL, Goveas JS. 2015. Depression and frailty in later life: a sys-tematic review. Clin Interv Aging 10:1947-1958, PMID: 26719681, https://doi.org/10. 2147/CIA.S69632.

48.Dominski FH, Lorenzetti Branco JH, Buonanno G, Stabile L, Gameiro da Silva M, Andrade A. 2021. Effects of air pollution on health: a mapping review of sys-tematic reviews and meta-analyses. Environ Res 201:111487, PMID: 34116013, https://doi.org/10.1016/j.envres.2021.111487.

49.Gao X, Coull B, Lin X, Vokonas P, Spiro A, Hou L, Schwartz J. 2021. Short-term air pollution, cognitive performance, and nonsteroidal anti-inflammatory drug use in the Veterans Affairs Normative Aging Study. Nat Aging 1(5):430-437, PMID: 34841262, https://doi.org/10.1038/s43587-021-00060-4.

50.Tibbo AJ, Baillie GS. 2020. Phosphodiesterase 4B: master regulator of brain signaling. Cells 9(5):1254, PMID: 32438615, https://doi.org/10.3390/cells9051254.

51.Lee JR. 2015. Protein tyrosine phosphatase PTPRT as a regulator of synaptic formation and neuronal development. BMB Rep 48(5):249-255, PMID: 25748173, https://doi.org/10.5483/BMBRep.2015.48.5.037.

52.Wamsley B, Jaglin XH, Favuzzi E, Quattrocolo G, Nigro MJ, Yusuf N, et al. 2018. Rbfox1 mediates cell-type-specific splicing in cortical interneurons. Neuron 100(4):846-59.e7, PMID: 30318414, https://doi.org/10.1016/j.neuron.2018.09.026.

53.Fuller R, Landrigan PJ, Balakrishnan K, Bathan G, Bose-O'Reilly S, Brauer M, et al. 2022. Pollution and health: a progress update. Lancet Planet Health 6(6): e535-e547, PMID: 35594895, https://doi.org/10.1016/S2542-5196(22)00090-0.

54.World Health Organization. 2010. WHO Guidelines for Indoor Air Quality: Selected Pollutants. Copenhagen, Denmark: Regional Office for Europe, World Health Organization.