Abstract
Background Long-term exposure to ambient air pollution has been linked to childhood-onset asthma, although evidence is still insufficient. Within the multicentre project Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE), we examined the associations of long-term exposures to particulate matter with a diameter <2.5 µm (PM2.5), nitrogen dioxide (NO2) and black carbon (BC) with asthma incidence in adults.
Methods We pooled data from three cohorts in Denmark and Sweden with information on asthma hospital diagnoses. The average concentrations of air pollutants in 2010 were modelled by hybrid land-use regression models at participants’ baseline residential addresses. Associations of air pollution exposures with asthma incidence were explored with Cox proportional hazard models, adjusting for potential confounders.
Results Of 98 326 participants, 1965 developed asthma during a mean follow-up of 16.6 years. We observed associations in fully adjusted models with hazard ratios of 1.22 (95% CI 1.04–1.43) per 5 μg·m−3 for PM2.5, 1.17 (95% CI 1.10–1.25) per 10 µg·m−3 for NO2 and 1.15 (95% CI 1.08–1.23) per 0.5×10−5 m−1 for BC. Hazard ratios were larger in cohort subsets with exposure levels below the European Union and US limit values and possibly World Health Organization guidelines for PM2.5 and NO2. NO2 and BC estimates remained unchanged in two-pollutant models with PM2.5, whereas PM2.5 estimates were attenuated to unity. The concentration–response curves showed no evidence of a threshold.
Conclusions Long-term exposure to air pollution, especially from fossil fuel combustion sources such as motorised traffic, was associated with adult-onset asthma, even at levels below the current limit values.
Abstract
Long-term exposure to air pollution, especially from fossil fuel combustion sources such as motorised traffic, is associated with the development of asthma in adults, even at levels below the current EU and US limit values and possibly WHO guidelines https://bit.ly/2QW5yA7
Introduction
Asthma is a complex and heterogeneous chronic respiratory disease affecting people of all ages [1]. Although lifestyles and genetic factors play important roles in asthma aetiology [2], environmental exposures are increasingly recognised as likely risk factors [3]. Ambient air pollution is one of the main contributors to morbidity and mortality worldwide [4]. The Global Burden of Disease Study ranked ambient air pollution the sixth most important risk factor for morbidity and mortality globally in 2016, and attributed 7.5% of all deaths to particulate matter with a diameter <2.5 µm (PM2.5) [5]. While PM2.5 levels are decreasing in most developed countries [6], evidence from studies with levels below current limit values suggests that the association with mortality likely has no safe threshold [7–10]. Evidence on morbidity outcomes, including asthma, is more limited.
The association between long-term exposure to air pollution and childhood-onset asthma has been extensively studied, and a recent meta-analysis of 41 studies demonstrated increased risks for nitrogen dioxide (NO2), PM2.5, particulate matter with a diameter <10 µm (PM10) and black carbon (BC) [11]. However, the literature on adult-onset asthma is more limited [12], in part due to the lack of cohorts with information on asthma incidence in adults (supplementary table S1) [13–22]. Of seven cohort studies on long-term exposure to NO2 and adult-onset asthma, all observed positive associations [13, 14, 16–19, 21], with three observing nonsignificant associations [13, 16, 17]. The majority [13, 14, 16, 17], but not all [15], of the studies on PM2.5 suggested positive associations. Two studies reported positive associations between air pollution and asthma incidence in nonsmokers only: one with traffic-related PM10 [20] and the other with ozone (O3) [22]. The studies differed in the definition of adult asthma incidence, with the majority relying on self-reported asthma symptoms, doctor-diagnosed asthma and/or use of asthma medication [15–17, 19–22], while only three used more objective definitions based on first-ever hospital discharge diagnoses [13, 18] or asthma surveillance databases, which combined physician insurance billing with emergency room and hospital visit data [14]. Although the studies on air pollution and adult-onset asthma all come from relatively low air pollution areas, such as Europe [16, 18–21], Canada [14], the USA [15, 17, 22] and Australia [13], few examined the shape of the concentration–response curve in the low exposure range.
The aim of this study was to investigate the associations of long-term air pollution exposures (PM2.5, NO2, BC and O3) and asthma incidence in adults, and to assess the shape of the concentration–response curves, particularly below current European Union (EU) and US limit values or World Health Organization (WHO) guidelines.
Methods
Study population
Within the Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE) project, individual data from 11 European cohorts were harmonised, pooled and analysed using a secure, remote access server at Utrecht University (Utrecht, The Netherlands). We used data from three cohorts that had information on asthma hospital discharge diagnoses: 1) the Cardiovascular Effects of Air Pollution and Noise in Stockholm (CEANS) study [23], which combined data from four subcohorts: the Stockholm Diabetes Prevention Program (SDPP), the Cohort of 60-year-olds (SIXTY), the Stockholm Screening Across the Lifespan Twin (SALT) study, and the Swedish National Study on Aging and Care in Kungsholmen (SNAC-K); 2) the Danish Diet, Cancer and Health (DCH) study [24]; and 3) the Danish Nurse Cohort (DNC) study, which included two subcohorts from recruitment rounds in 1993 and 1999 [25]. The confounder data from each cohort were collected through questionnaires at cohort recruitment, between 1992 and 2004. More details on the three cohorts can be found in the supplementary material. The study was undertaken in accordance with the Declaration of Helsinki and all three cohorts were approved by the local ethics committees in accordance with national regulations.
Outcome definition
We defined incidence of asthma as the first hospital discharge diagnosis (inpatient, outpatient or emergency room visits for Danish DNC and DCH, and inpatient visits for Swedish CEANS) in participants without asthma diagnoses before baseline. The follow-up period was from 1992–2004 (baseline years) until 2011 (CEANS) or 2015 (DCH and DNC). We used primary discharge diagnoses of asthma with International Classification of Diseases, Ninth Revision (ICD-9) or Tenth Revision (ICD-10) codes 493 or J45–46, respectively.
Exposure assessment
Annual average concentrations of PM2.5, NO2, BC and warm season O3 (April–September; maximum running 8-h averages) for 2010 were estimated at participants’ baseline residential addresses, at 100×100 m spatial resolution, using standardised Europe-wide hybrid land-use regression models [26, 27], described in more detail in the supplementary material. Additionally, we back-extrapolated pollutant concentrations for each year from baseline to the end of follow-up for two available cohorts (CEANS and DCH) for sensitivity analyses.
Statistical analysis
We used Cox proportional hazard models to examine the associations between long-term exposures to air pollution and asthma incidence, with censoring at death, diagnosis of chronic obstructive pulmonary disease (COPD, principal diagnoses with ICD-9: 490–492 and 494–496 or ICD-10: J40–44), emigration or the end of follow-up, whichever came first. Participants with asthma diagnoses at baseline were excluded from the analyses. We included the air pollutants separately as a linear variable and used age as the underlying timescale [28]. The associations with air pollution were estimated through three steps, with an increasing level of adjustment for a priori defined individual- and area-level confounders. Model 1 included age (time axis), sex (strata), subcohort (strata) and cohort baseline year; Model 2 additionally adjusted for individual lifestyles and socioeconomic status: smoking status (never-smoker, ex-smoker and current smoker), smoking duration (years), smoking intensity (linear and squared term; cigarettes per day), body mass index (BMI; categorical variable according to the WHO: <18.5, 18.5–24.9, 25.0–29.9 and ≥30.0 kg·m−2), marital status (single, married/living with partner, divorced and widowed), employment status (employed and other) and educational level (primary school or lower, secondary school and university degree or higher); and Model 3 (main model) additionally adjusted for area-level mean income (continuous variable in EUR), which is at the municipality level in 2001 for DCH and DNC or at the neighbourhood level in 1994 for CEANS. Participants with complete information for all variables in Model 3 were included in analyses.
We investigated if associations persisted at low air pollution concentrations by excluding participants exposed to levels above pre-defined cut-off values based on existing EU and US limit values and WHO guidelines: for PM2.5, the EU, US and WHO annual average limit/guideline values are 25, 12 and 10 µg·m−3, respectively; for NO2, the WHO/EU and WHO Health Risks of Air Pollution In Europe [29] annual average limit/guideline values are 40 and 20 µg·m−3, respectively. To evaluate the shape of the concentration–response curves between air pollutants and asthma incidence, we applied natural cubic splines with three degrees of freedom in Model 3 and tested for linearity by comparing it with linear models using the likelihood ratio test. We also performed threshold analyses, in which the pollutants were set to zero for exposures below certain (threshold) values, assuming no effect below the thresholds. The performance of threshold models was evaluated by comparison of the Akaike Information Criterion with the corresponding linear model. We also fitted two-pollutant models in Model 3, in an attempt to account for mutual correlation of pollutants.
We conducted several sensitivity analyses. First, to examine the robustness of using air pollution exposure modelled for 2010, we re-ran Model 3 with 1) time-varying air pollution concentrations, by linking back-extrapolated annual averages for each year from baseline until the end of follow-up for cohorts with complete residential address history (only CEANS and DCH), using 1- or 5-year strata of calendar time to account for secular time trend in asthma incidence and air pollution, and 2) back-extrapolated annual average concentrations at baseline for all cohorts. Second, we estimated associations in Model 3 by separately including each of the three cohorts or by excluding one cohort each time. We also graphically showed the trend of yearly back-extrapolated pollutant concentrations during the follow-up period using the ratio and the absolute difference method in the CEANS (n=19 320) and DCH (n=51 991) cohorts, which had available address history information.
We also performed effect modification by age (<65 and ≥65 years), BMI, smoking status, marital status, employment status, educational level and COPD status at baseline. Effect modification was evaluated by introducing an interaction term into Model 3 and tested by the Wald test.
The results are presented as hazard ratios and 95% confidence intervals. All analyses were performed in R version 3.4.0 (www.r-project.org).
Results
From a total of 106 727 participants from the three cohorts with complete air pollution exposure data (21 986 from CEANS, 56 308 from DCH and 28 433 from DNC), we excluded 821 with asthma diagnoses before the beginning of follow-up and 7580 with missing information on confounders, leaving 98 326 participants for analyses. During a mean follow-up of 16.6 years, 1965 participants developed asthma (table 1). The mean age at baseline was 55.8 years. Participants who developed asthma were more likely to be female, obese, and have higher levels of PM2.5, NO2 and BC at the residence than asthma-free participants. For NO2, all cohorts showed some exceedances of the EU limit value and the WHO recommendation of 40 µg·m−3; for PM2.5, the individual levels in all cohorts complied with the EU limit value of 25 µg·m−3, and the CEANS cohort also complied with the US limit value (12 µg·m−3) and WHO guideline (10 µg·m−3) (figure 1). More details on the characteristics of study participants, air pollution levels in each subcohort and by quintiles of NO2 concentrations are shown in supplementary tables S2, S3 and S4, respectively. We found that participants living in the highest quintiles of exposure to NO2 were more likely to be smokers, single, less educated and have lower income, but of similar age and BMI compared with those living in areas with low NO2 levels (supplementary table S4). We observed that air pollution levels were decreasing during follow-up time (supplementary figure S1). PM2.5, NO2 and BC were generally moderate to highly correlated with each other (Pearson correlation coefficients >0.6), while O3 was negatively correlated with the other pollutants (supplementary table S5). NO2 and BC were highly correlated in all subcohorts (Pearson correlation coefficients 0.67–0.93) except for SNAC-K (Pearson correlation coefficient 0.43).
Characteristics of participants at baseline (1992–2004) and air pollutants for 2010 by adult-onset asthma status
Distribution of annual average of air pollution concentrations by subcohorts for 2010: a) particulate matter with a diameter <2.5 µm (PM2.5), b) nitrogen dioxide (NO2), c) black carbon (BC) and d) ozone (O3). CEANS: Cardiovascular Effects of Air Pollution and Noise in Stockholm; SDPP: Stockholm Diabetes Prevention Program; SIXTY: Cohort of 60-year-olds; SALT: Stockholm Screening Across the Lifespan Twin; SNAC-K: Swedish National Study on Aging and Care in Kungsholmen; DCH: Danish Diet, Cancer and Health; DNC: Danish Nurse Cohort; EU: European Union; WHO: World Health Organization. Red dashed lines in a) and b) indicate different limit/guideline values from the EU, USA and WHO for PM2.5 (US: 12 µg·m−3; WHO: 10 µg·m−3) and NO2 (WHO/EU: 40 µg·m−3; WHO Health Risks of Air Pollution In Europe: 20 µg·m−3), respectively. Boxes indicate median with interquartile range; whiskers extend to the 5th and 95th percentiles.
We observed positive associations between PM2.5, NO2 and BC and asthma incidence in all three models, with minor attenuations of estimates from Model 1 to Model 3 (table 2). We observed larger hazard ratios in subsets of participants (Model 3) with PM2.5 levels below 15, 12 and 10 µg·m−3 (table 3). Hazard ratios for NO2 were also slightly higher when only including participants with concentrations below 40, 30 and 20 µg·m−3. Likewise, for BC, the fully adjusted hazard ratios remained increased even below 1×10−5 m−1 (table 3). We did not find any evidence for a threshold for the associations between PM2.5, NO2 and BC and asthma incidence (figure 2), with no evidence of deviation from linearity observed (data not shown), which is also supported by the threshold analyses (supplementary table S6).
Associations between long-term air pollution exposure and adult-onset asthma
Associations between long-term air pollution exposure and adult-onset asthma below various cut-off values based on Model 3#
Estimated concentration–response curves for effects of long-term air pollution exposure on adult-onset asthma: a) particulate matter with a diameter <2.5 µm (PM2.5), b) nitrogen dioxide (NO2), c) black carbon (BC) and d) ozone (O3). Natural cubic splines with three degrees of freedom were fit for air pollutants to evaluate the shape of the associations based on the main model (Model 3). Black solid lines indicate hazard ratio values and black dashed lines indicate their 95% confidence intervals. Red dashed lines show hazard ratios equal to unity indicating no risk attributed to air pollution exposure. Background density histograms show the trends in distributions of air pollutants among participants. Green dashed lines indicate the 5th and 95th percentiles of air pollutant concentrations.
In two-pollutant models, the hazard ratios for NO2 and BC remained unchanged after adjusting for PM2.5, whereas the hazard ratios for PM2.5 were attenuated to below unity when adjusting for NO2 or BC (table 4). In two-pollutant models with O3, the hazard ratios for PM2.5, NO2 and BC were essentially unaffected, while the negative association between O3 and asthma incidence was attenuated to unity.
Two-pollutant models for association between long-term air pollution exposure and adult-onset asthma based on Model 3#
Observed associations were robust when time-varying concentrations were used controlling for time trends (supplementary figure S2 and supplementary table S7) and when restricting participants to subsets of cohorts (supplementary table S9). However, effect estimates of air pollution exposure back-extrapolated to the baseline year were attenuated to unity for PM2.5, and remained unchanged for NO2 and BC (supplementary table S8). The associations of PM2.5, NO2 and BC with asthma incidence were consistently stronger in ex-smokers, unemployed and low-educated participants (supplementary figure S3). O3 also showed a borderline positive association in never-smokers.
Discussion
In this pooled analysis of three cohorts, long-term exposures to PM2.5, NO2 and BC were associated with increased risks of asthma in 98 326 adults from Denmark and Sweden, even at levels below the current EU limit values. The concentration–response curves were steeper at the lower end of the exposure ranges and showed no evidence of a threshold below which air pollution effects were null. The association of asthma with PM2.5 was attenuated to unity in two-pollutant models, while the associations with NO2 and BC remained robust.
Our results on PM2.5 and asthma incidence are in line with those from two studies that also used objective asthma incidence definitions, based on cohorts of 1.1 million adults in Toronto, ON, Canada (HR 1.02 (95% CI 1.00–1.04) per 3.2 µg·m−3) [14] and 100 084 adults in Sydney, Australia (HR 1.08 (95% CI 0.89–1.30) per 1 μg·m−3) [13], as well as with two studies with self-reported asthma, with 23 704 participants in six ESCAPE cohorts (OR 1.04 (95% CI 0.88–1.23) per 5 μg·m−3) [16] and 50 884 females from the US Sisters Cohort (OR 1.20 (95% CI 0.99–1.46) per 3.6 μg·m−3) [17]. In contrast, the American Nurses’ Health Study did not detect an association between PM2.5 and self-reported asthma (HR 0.90 (95% CI 0.73–1.12) per 10 μg·m−3) [15]. Our findings of an association between NO2 and asthma incidence are generally in line with existing evidence. In studies using objective asthma definitions, HRs for NO2 ranged from 1.03 (0.88–1.19) per 5 μg·m−3 in the Sydney cohort [13] and 1.03 (95% CI 1.02–1.05) per 4.1 ppb (around 7.7 μg·m−3) in the Toronto cohort [14] to 1.10 (95% CI 1.01–1.20) per 5.8 μg·m−3 in the DCH cohort [18]. Results for NO2 from studies with self-reported asthma also suggest positive associations, with ORs of 1.10 (95% CI 0.99–1.21) per 10 μg·m−3 in six ESCAPE cohorts [16], 1.12 (95% CI 0.96–1.30) per 5.8 ppb (∼10.9 μg·m−3) in the US Sisters Study [17], 1.43 (95% CI 1.02–2.01) per 10 μg·m−3 in the European Community Respiratory Health Survey study [21] and 1.54 (95% CI 1.00–2.36) per 10 μg·m−3 in a Swedish cohort [19]. Additionally, our finding of an association between BC and asthma incidence is consistent with the ESCAPE finding [16]. We did not observe an association of O3 with asthma, overall, but found that it might increase risks of asthma in nonsmokers, which is in line with an earlier finding by McDonnel et al. [22].
Our findings provided solid evidence that air pollution affects asthma below current limit values and guidelines. This study is based on cohorts from Denmark and Sweden, with some of the lowest air pollution levels in Europe. The findings of this study agree with the majority of the literature on air pollution and adult-onset asthma, which comes from areas with low to moderate PM2.5 levels in Europe [16, 18–20], Canada [14], the USA [15, 17] and Australia [13] (supplementary table S1).
Our findings of attenuated PM2.5 effects in two-pollutant models with NO2 or BC can be difficult to interpret and require further exploration. Differential measurement error may complicate the interpretation of two-pollutant models [30]. The pollutant with the lowest measurement error may show the most consistent association in two-pollutant models. After adjustment for NO2, the significant single-pollutant hazard ratio for PM2.5 was reduced to unity, whereas the association with NO2 remained robust after adjustment for PM2.5. Given that the correlation between PM2.5 and NO2 was moderate and the width of the confidence interval was only modestly increased in the two-pollutant models, we did not interpret the reduction of the hazard ratio for PM2.5 as merely an artefact related to multicollinearity. The association with NO2 might reflect direct effects of NO2 or related particles emitted at combustion, such as BC and ultrafine particles (UFPs; particulate matter with a diameter <0.1 µm). We also did not interpret the reduction of the PM2.5 hazard ratio as implying that particles had no effect in our setting, as adjustment for NO2 also adjusted for particles from the same sources as NO2, including motorised traffic and other combustion sources. Only two studies to date examined two-pollutant models with PM2.5 and NO2. The Toronto cohort study found that the association with PM2.5 was robust to additional adjustment for NO2, although notably, associations with NO2 were stronger, both in single- and in two-pollutant models [14], suggesting independent effects of both pollutants. Furthermore, the Toronto study, as the first and only study to have examined the role of UFPs on asthma incidence, found no association with UFPs, providing some support for a direct effect of NO2 on asthma [14]. Our results are in line with the finding in the ESCAPE study, where, comparable to our PM2.5 results, the hazard ratio for PM10 (highly correlated with PM2.5) was attenuated to below unity with NO2 included in a model [16].
The exact biological mechanisms of how exposures to air pollution promote the development of asthma in adults are not known. Current understanding suggests that NO2, an airway irritant which has been linked to airway inflammation and airflow limitation in animal models [31], may be both a causal agent responsible for asthma development and a proxy for traffic-related PM2.5 or UFPs, which can deposit in the respiratory tract and the lung alveoli causing oxidative stress, inflammation and other biochemical changes related to asthma [32]. NO2 is emitted together with traffic-related particulate matter mainly in the ultrafine range, which contributes minimally to total PM2.5 mass but could contribute significantly to the development of asthma with large particle numbers and surface area, through high pulmonary deposition, causing oxidative stress and inflammation in tracheobronchial and alveolar regions [33]. However, the only previous cohort study with data on PM2.5, UFPs and NO2 and adult-onset asthma reported the strongest associations with NO2, and only weak associations with PM2.5 and UFPs, and found that the significant positive association for UFPs attenuated to null in a two-pollutant model with NO2, supporting the idea of the independent effect of NO2 on asthma development [14]. We presented novel observations of enhanced hazard ratios in ex-smokers, unemployed and low-educated participants for PM2.5, NO2 and BC, as well as in never-smokers for O3. Earlier studies found little evidence for effect modification by education [13], smoking status [13, 15, 16, 18], age or BMI [13, 14, 16], although two studies reported associations between traffic-related PM10 [20] and O3 [22] and asthma in never-smokers only. These results suggest possibly higher susceptibility of nonsmokers and participants with lower socioeconomic status to the effects of air pollution on asthma.
Adult asthma is a chronic disease with a complex phenotype and recurring symptoms that makes it difficult to diagnose and identify a precise time of onset. Asthma definitions based on self-reports from respiratory disease surveys are subject to recall bias, resulting in looser definitions and likely an overestimation of true burden [34, 35]. In this study we benefited from objective definitions based on hospital discharge diagnoses from nationwide hospital registers in Denmark and Sweden. Asthma incidence rates defined by hospital discharge diagnoses may underestimate true asthma burden, as not all asthma patients require hospital contact, and thus an asthma hospital discharge diagnosis typically represents a point of disease progression to a more severe stage or exacerbation. It is appealing as it presents a well-characterised asthma definition, typically confirmed by objective measurements of lung function and reversible airflow obstruction, as standard procedures in Danish and Swedish hospitals. The specificity of asthma diagnoses in the Danish Hospital Discharge Register was found to be as high as 0.98, validating their use in epidemiological studies [36].
The main strengths of our study include pooled analyses of three large prospective cohorts with objective assessments of asthma incidence, detailed individual- and area-level information on major confounders, standardised assessments of air pollution exposure, and long follow-up periods. We most likely have a low sensitivity but high specificity for adult-onset asthma by using hospital discharge diagnoses. A limitation of our study is that our exposure assessment methods relied solely on residential exposures with no information on work addresses, commuting habits or personal time–activity patterns. Finally, our study lacks data on familial histories of asthma and allergy, pet ownership, and environmental tobacco smoke, which may be confounders or effect modifiers.
Conclusions
Our results suggest that long-term exposure to air pollution, especially from fossil fuel combustion sources such as motorised traffic, is associated with the development of adult-onset asthma, even at levels below the current EU and US limit values and WHO guidelines, calling for stricter air quality regulation as an important tool for asthma prevention.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-03099-2020.SUPPLEMENT
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-03099-2020.Shareable
Acknowledgements
Research described in this article was conducted under contract to the Health Effects Institute (HEI), an organisation jointly funded by the US Environmental Protection Agency (EPA) (assistance award number R-82811201) and certain motor vehicle and engine manufacturers. The contents of this article do not necessarily reflect the views of HEI, or its sponsors, nor do they necessarily reflect the views and policies of the EPA or the motor vehicle and engine manufacturers. The authors would also like to thank all participants in the CEANS, DCH and DNC cohort studies, and the respective study teams (the ELAPSE project) for their hard work and effort. Thanks to Niklas Andersson (Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden) for the work of harmonising complicated data on covariates between the four subcohorts in CEANS.
Footnotes
This article has an editorial commentary: https://doi.org/10.1183/13993003.00064-2021
This article has supplementary material available from erj.ersjournals.com
Data sharing: This is an observational cohort study and requests for data can be made by contacting the corresponding author.
Author contributions: The study was conceptualised and designed by Z.J. Andersen, G. Hoek, B. Brunekreef, P. Ljungman and S. Liu. G. Hoek and B. Brunekreef are Principal Investigators of the ELAPSE project. Statistical analysis and drafting of the manuscript was conducted by S. Liu. Z.J. Andersen helped in drafting the manuscript. J.T. Jørgensen and U.A. Hvidtfeldt prepared the individual cohort data for the analyses. G. Hoek, B. Brunekreef, J. Chen and M. Strak coordinated the ELAPSE project, helped in preparing pooled data for analyses and provided support with the access to pooled cohort data. S.P. Rodopoulou, E. Samoli and K. Katsouyanni contributed with the statistical analysis strategy and scripts for the statistical analyses. K. de Hoogh worked for the exposure assessment. All authors have read and revised the manuscript for important intellectual content, and contributed to the interpretation of the results. All authors have approved the final draft of the manuscript.
Conflict of interest: S. Liu has nothing to disclose.
Conflict of interest: J.T. Jørgensen has nothing to disclose.
Conflict of interest: P. Ljungman has nothing to disclose.
Conflict of interest: G. Pershagen has nothing to disclose.
Conflict of interest: T. Bellander has nothing to disclose.
Conflict of interest: K. Leander has nothing to disclose.
Conflict of interest: P.K.E. Magnusson has nothing to disclose.
Conflict of interest: D. Rizzuto has nothing to disclose.
Conflict of interest: U.A. Hvidtfeldt has nothing to disclose.
Conflict of interest: O. Raaschou-Nielsen has nothing to disclose.
Conflict of interest: K. Wolf has nothing to disclose.
Conflict of interest: B. Hoffmann has nothing to disclose.
Conflict of interest: B. Brunekreef has nothing to disclose.
Conflict of interest: M. Strak has nothing to disclose.
Conflict of interest: J. Chen has nothing to disclose.
Conflict of interest: A. Mehta has nothing to disclose.
Conflict of interest: R.W. Atkinson has nothing to disclose.
Conflict of interest: M. Bauwelinck has nothing to disclose.
Conflict of interest: R. Varraso has nothing to disclose.
Conflict of interest: M-C. Boutron-Ruault has nothing to disclose.
Conflict of interest: J. Brandt has nothing to disclose.
Conflict of interest: G. Cesaroni has nothing to disclose.
Conflict of interest: F. Forastiere has nothing to disclose.
Conflict of interest: D. Fecht has nothing to disclose.
Conflict of interest: J. Gulliver has nothing to disclose.
Conflict of interest: O. Hertel has nothing to disclose.
Conflict of interest: K. de Hoogh has nothing to disclose.
Conflict of interest: N.A.H. Janssen has nothing to disclose.
Conflict of interest: K. Katsouyanni has nothing to disclose.
Conflict of interest: M. Ketzel has nothing to disclose.
Conflict of interest: J.O. Klompmaker has nothing to disclose.
Conflict of interest: G. Nagel has nothing to disclose.
Conflict of interest: B. Oftedal has nothing to disclose.
Conflict of interest: A. Peters has nothing to disclose.
Conflict of interest: A. Tjønneland has nothing to disclose.
Conflict of interest: S.P. Rodopoulou has nothing to disclose.
Conflict of interest: E Samoli has nothing to disclose.
Conflict of interest: D.T. Kristoffersen has nothing to disclose.
Conflict of interest: T. Sigsgaard has nothing to disclose.
Conflict of interest: M. Stafoggia has nothing to disclose.
Conflict of interest: D. Vienneau has nothing to disclose.
Conflict of interest: G. Weinmayr has nothing to disclose.
Conflict of interest: G. Hoek has nothing to disclose.
Conflict of interest: Z.J. Andersen has nothing to disclose.
Support statement: This work is supported by the Health Effects Institute (4954-RFA14-3/16-5-3) and a grant from the China Scholarship Council (201806010406). SALT and TwinGene are substudies of The Swedish Twin Registry (STR) which is managed by Karolinska Institutet and receives additional funding through the Swedish Research Council under grant 2017-00641. Funding information for this article has been deposited with the Crossref Funder Registry.
- Received August 11, 2020.
- Accepted November 17, 2020.
- Copyright ©ERS 2021