Empirical spatial air pollution models have been applied extensively to assess exposure in epidemiological studies with increasingly sophisticated and complex statistical algorithms beyond ordinary ...linear regression. However, different algorithms have rarely been compared in terms of their predictive ability.
This study compared 16 algorithms to predict annual average fine particle (PM2.5) and nitrogen dioxide (NO2) concentrations across Europe. The evaluated algorithms included linear stepwise regression, regularization techniques and machine learning methods. Air pollution models were developed based on the 2010 routine monitoring data from the AIRBASE dataset maintained by the European Environmental Agency (543 sites for PM2.5 and 2399 sites for NO2), using satellite observations, dispersion model estimates and land use variables as predictors. We compared the models by performing five-fold cross-validation (CV) and by external validation (EV) using annual average concentrations measured at 416 (PM2.5) and 1396 sites (NO2) from the ESCAPE study. We further assessed the correlations between predictions by each pair of algorithms at the ESCAPE sites.
For PM2.5, the models performed similarly across algorithms with a mean CV R2 of 0.59 and a mean EV R2 of 0.53. Generalized boosted machine, random forest and bagging performed best (CV R2~0.63; EV R2 0.58–0.61), while backward stepwise linear regression, support vector regression and artificial neural network performed less well (CV R2 0.48–0.57; EV R2 0.39–0.46). Most of the PM2.5 model predictions at ESCAPE sites were highly correlated (R2 > 0.85, with the exception of predictions from the artificial neural network). For NO2, the models performed even more similarly across different algorithms, with CV R2s ranging from 0.57 to 0.62, and EV R2s ranging from 0.49 to 0.51. The predicted concentrations from all algorithms at ESCAPE sites were highly correlated (R2 > 0.9). For both pollutants, biases were low for all models except the artificial neural network. Dispersion model estimates and satellite observations were two of the most important predictors for PM2.5 models whilst dispersion model estimates and traffic variables were most important for NO2 models in all algorithms that allow assessment of the importance of variables.
Different statistical algorithms performed similarly when modelling spatial variation in annual average air pollution concentrations using a large number of training sites.
•Multiple statistical algorithms with very different assumptions were compared.•Despite the difference in modeling frameworks, predictions among the models exhibit generally good agreement.•The use of an external evaluation dataset strengthens evaluation by cross-validation.
Long-term exposure to ambient air pollution has been associated with premature mortality, but associations at concentrations lower than current annual limit values are uncertain. We analysed ...associations between low-level air pollution and mortality within the multicentre study Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE).
In this multicentre longitudinal study, we analysed seven population-based cohorts of adults (age ≥30 years) within ELAPSE, from Belgium, Denmark, England, the Netherlands, Norway, Rome (Italy), and Switzerland (enrolled in 2000–11; follow-up until 2011–17). Mortality registries were used to extract the underlying cause of death for deceased individuals. Annual average concentrations of fine particulate matter (PM2·5), nitrogen dioxide (NO2), black carbon, and tropospheric warm-season ozone (O3) from Europe-wide land use regression models at 100 m spatial resolution were assigned to baseline residential addresses. We applied cohort-specific Cox proportional hazard models with adjustment for area-level and individual-level covariates to evaluate associations with non-accidental mortality, as the main outcome, and with cardiovascular, non-malignant respiratory, and lung cancer mortality. Subset analyses of participants living at low pollutant concentrations (as per predefined values) and natural splines were used to investigate the concentration-response function. Cohort-specific effect estimates were pooled in a random-effects meta-analysis.
We analysed 28 153 138 participants contributing 257 859 621 person-years of observation, during which 3 593 741 deaths from non-accidental causes occurred. We found significant positive associations between non-accidental mortality and PM2·5, NO2, and black carbon, with a hazard ratio (HR) of 1·053 (95% CI 1·021–1·085) per 5 μg/m3 increment in PM2·5, 1·044 (1·019–1·069) per 10 μg/m3 NO2, and 1·039 (1·018–1·059) per 0·5 × 10−5/m black carbon. Associations with PM2·5, NO2, and black carbon were slightly weaker for cardiovascular mortality, similar for non-malignant respiratory mortality, and stronger for lung cancer mortality. Warm-season O3 was negatively associated with both non-accidental and cause-specific mortality. Associations were stronger at low concentrations: HRs for non-accidental mortality at concentrations lower than the WHO 2005 air quality guideline values for PM2·5 (10 μg/m3) and NO2 (40 μg/m3) were 1·078 (1·046–1·111) per 5 μg/m3 PM2·5 and 1·049 (1·024–1·075) per 10 μg/m3 NO2. Similarly, the association between black carbon and non-accidental mortality was highest at low concentrations, with a HR of 1·061 (1·032–1·092) for exposure lower than 1·5× 10−5/m, and 1·081 (0·966–1·210) for exposure lower than 1·0× 10−5/m.
Long-term exposure to concentrations of PM2·5 and NO2 lower than current annual limit values was associated with non-accidental, cardiovascular, non-malignant respiratory, and lung cancer mortality in seven large European cohorts. Continuing research on the effects of low concentrations of air pollutants is expected to further inform the process of setting air quality standards in Europe and other global regions.
Health Effects Institute.
•Air pollution, traffic noise and lack of green space have been associated with diabetes in analyses mainly focusing on one or two environmental factors at a time.•We aimed to investigate if air ...pollution, road traffic noise and green space are independent risk factors of type 2 diabetes.•In a multi-pollutant analysis, ultrafine particles, NO2, noise at both most and least exposed façade and two proxies of lack of green space were all associated with higher risk of type 2 diabetes.•The cumulative risk estimate of the multi-pollutant analysis was much higher than the risk estimate of any single pollutant.
Air pollution, road traffic noise and lack of greenness coexist in urban environments and have all been associated with type 2 diabetes. We aimed to investigate how these co-exposures were associated with type 2 diabetes in a multi-exposure perspective.
We estimated 5-year residential mean exposure to fine particles (PM2.5), ultrafine particles (UFP), elemental carbon (EC), nitrogen dioxide (NO2) and road traffic noise at the most (LdenMax) and least (LdenMin) exposed facade for all persons aged > 50 years living in Denmark in 2005 to 2017. For each air pollutant, we estimated total concentrations and traffic contributions. Based on land use maps, we estimated proportion of green and non-green space within 150 and 1000 m of all residences. In total, 1.9 million persons were included and 128,358 developed type 2 diabetes during follow-up. We performed analyses using Cox proportional hazards models, with adjustment for individual and neighborhood-level sociodemographic co-variates.
In single-pollutant models, all air pollutants, noise and lack of green space were associated with higher risk of diabetes. In two-, three- and four-pollutant analyses of the air pollutants, only UFP and NO2 remained associated with higher diabetes risk in all models. LdenMax, LdenMin and the two proxies of green space remained associated with diabetes in two-pollutant models of, respectively, noise and green space. In a multi-pollutant analysis, we found hazard ratios (95 % confidence intervals) per interquartile range of 1.021 (1.005; 1.038) for UFP, 1.012 (0.996; 1.028) for NO2, 1.022 (1.012; 1.033) for LdenMin, 1.013 (1.004; 1.022) for LdenMax, and 1.038 (1.031; 1.044) and 1.018 (1.010; 1.025) for lack of green space within, respectively, 150 m and 1000 m, and a cumulative risk index of 1.131 (1.113; 1.149).
Air pollution, road traffic noise and lack of green space were independently associated with higher risk of type 2 diabetes.
•We investigated PM2.5 components and mortality in the Danish population.•Eight PM2.5 components were assessed at population addresses by air pollution models.•Sulfate and SOA particles showed robust ...associations with natural cause mortality.•Elemental carbon and dust particles were associated with respiratory disease mortality.
Ambient fine particulate matter (PM2.5) causes millions of deaths every year worldwide. Identification of the most harmful types of PM2.5 would facilitate efficient prevention strategies.
The aim of this study was to investigate associations between components of PM2.5 and mortality in a nation-wide Danish population.
Our study base was Danes born 1921–1985 and aged 30–85 years, who were followed up for mortality from 1991 to 2015. We included 678,465 natural cause mortality cases and selected five age, sex and calendar time matched controls to each case from the study base. We retrieved the address history of the study population from Danish registries and assessed five-year average concentrations of eight PM2.5 components using deterministic Chemistry-Transport Models air pollution models. We estimated mortality rate ratios (MRRs) by conditional logistic regression and adjusted for socio-demographical factors at individual and neighborhood level.
Single pollutant models showed the strongest associations between natural cause mortality and an interquartile increase in sulfate particles (SO4−-) (MRR: 1.123; 95 % CI: 1.100–1.147 per 1.5 µg/m3) and secondary organic aerosol (SOA) (MRR: 1.054; 95 % CI: 1.048–1.061 per 0.050 µg/m3). Two-pollutant models showed robust associations between SO4−− and SOA and natural cause mortality. Elemental carbon and mineral dust showed robust associations with higher respiratory and lung cancer mortality.
This nation-wide study found robust associations between natural cause mortality and SO4−− particles and SOA, which is in line with the results of previous studies. Elemental carbon and mineral dust showed robust associations with higher respiratory and lung cancer mortality.
Epidemiologic studies have linked transportation noise to increased morbidity and mortality, particularly for cardiovascular outcomes. However, studies investigating metabolic outcomes such as ...diabetes are limited and have focused only on noise exposures estimated for the loudest residential façade.
We aimed to examine the influence of long-term residential exposure to transportation noise at the loudest and quietest residential façades and the risk for type 2 diabetes.
Road traffic and railway noise exposures (Lden) at the most and least exposed façades were estimated for all dwellings in Denmark during 1990-2017. Aircraft noise was estimated in 5-dB categories. Ten-year time-weighted mean noise exposures were estimated for
individuals
of age. From 2000 to 2017, 233,912 incident cases of type 2 diabetes were identified using hospital and prescription registries, with a mean follow-up of 12.9 y. We used Cox proportional hazards models adjusting for individual- and area-level covariates and long-term residential air pollution. The population-attributable fraction (PAF) was also computed.
Hazard ratios (HRs) and 95% confidence intervals (CIs) for type 2 diabetes in association with 10-dB increases in 10-y mean road traffic noise at the most and least exposed façades, respectively, were 1.05 (95% CI: 1.04, 1.05) and 1.09 (95% CI: 1.08, 1.10). Following subsequent adjustment for fine particulate matter particulate matter
in aerodynamic diameter (10-y mean), the HRs (CIs) were 1.03 (95% CI: 1.03, 1.04) and 1.08 (95% CI: 1.07, 1.09), respectively. For railway noise, the HRs per 10-dB increase in 10-y mean exposure were 1.03 (95% CI: 1.02, 1.04) and 1.02 (95% CI: 1.01, 1.04) for the most and least exposed façades, respectively. Categorical models supported a linear exposure-outcome relationship for road traffic noise and, to a lesser extent, for railway noise. Aircraft noise
was associated with a 1-4% higher likelihood of type 2 diabetes compared with those who were unexposed. We found road traffic and railway noise associated with a PAF of 8.5% and 1.4%, respectively, of the diabetes cases.
Long-term exposure to road, railway, and possibly aircraft traffic noise was associated with an increased risk of type 2 diabetes in a nationwide cohort of Danish adults. Our findings suggest that diabetes should be included when estimating the burden of disease due to transportation noise. https://doi.org/10.1289/EHP9146.
This study aims to estimate the mortality risk associated with air pollution in a Swedish cohort with relatively low exposure. Air pollution models were used to estimate annual mean concentrations of ...particulate matter with aerodynamic diameter ≤ 2.5 µm (PM2.5), primary emitted carbonaceous particles (BC/pOC), sea salt, chemically formed particles grouped as secondary inorganic and organic aerosols (SIA and SOA) as well as ozone (O3) and nitrogen dioxide (NO2). The exposure, as a moving average was calculated based on home address for the time windows 1 year (lag 1), 1–5 years (lag 1–5) and 1–10 years (lag 1–10) preceding the death. During the study period, 1151 cases of natural mortality, 253 cases of cardiovascular disease (CVD) mortality and 113 cases of respiratory and lung cancer mortality were observed during 369,394 person-years of follow-up. Increased natural mortality was observed in association with NO2 (3% 95% CI −8–14% per IQR) and PM2.5 (2% 95% CI −5–9% for an IQR increase) and its components, except for SOA where a decreased risk was observed. Higher risk increases were observed for CVD mortality (e.g., 18% 95% CI 1–39% per IQR for NO2). These findings at low exposure levels are relevant for future decisions concerning air quality policies.
We developed Europe-wide models of long-term exposure to eight elements (copper, iron, potassium, nickel, sulfur, silicon, vanadium, and zinc) in particulate matter with diameter <2.5 μm (PM2.5) ...using standardized measurements for one-year periods between October 2008 and April 2011 in 19 study areas across Europe, with supervised linear regression (SLR) and random forest (RF) algorithms. Potential predictor variables were obtained from satellites, chemical transport models, land-use, traffic, and industrial point source databases to represent different sources. Overall model performance across Europe was moderate to good for all elements with hold-out-validation R-squared ranging from 0.41 to 0.90. RF consistently outperformed SLR. Models explained within-area variation much less than the overall variation, with similar performance for RF and SLR. Maps proved a useful additional model evaluation tool. Models differed substantially between elements regarding major predictor variables, broadly reflecting known sources. Agreement between the two algorithm predictions was generally high at the overall European level and varied substantially at the national level. Applying the two models in epidemiological studies could lead to different associations with health. If both between- and within-area exposure variability are exploited, RF may be preferred. If only within-area variability is used, both methods should be interpreted equally.
Abstract
Background
Only few epidemiological studies have investigated whether chronic exposure to air pollution from different sources have different impacts on risk of diabetes. We aimed to ...investigate associations between air pollution from traffic versus non-traffic sources and risk of type 2 diabetes in the Danish population.
Methods
We estimated long-term exposure to traffic and non-traffic contributions of particulate matter with a diameter <2.5 µg (PM2.5), elemental carbon (EC), ultrafine particles (UFP) and nitrogen dioxide (NO2) for all persons living in Denmark for the period 2005–17. In total, 2.6 million persons aged >35 years were included, of whom 148 020 developed type 2 diabetes during follow-up. We applied Cox proportional hazards models for analyses, using 5-year time-weighted running means of air pollution and adjustment for individual- and area-level demographic and socioeconomic covariates.
Results
We found that 5-year exposure to all particle measures (PM2.5, UFP and EC) and NO2 were associated with higher type 2 diabetes risk. We observed that for UFP, EC and potentially PM2.5, the pollution originating from traffic was associated with higher risks than the non-traffic contributions, whereas for NO2 similar hazard ratios (HR) were observed. For example, in two-source models, hazard ratios (HRs) per interquartile change in traffic UFP, EC and PM2.5 were 1.025, 1.045 and 1.036, respectively, whereas for non-traffic UFP, EC and PM2.5, the HRs were 1.013, 1.018 and 1.001, respectively.
Conclusions
Our finding of stronger associations with particulate matter from traffic compared with non-traffic sources implies that prevention strategies should focus on limiting traffic-related particulate matter air pollution.
We investigated associations of smoking and coronary heart disease (CHD) by age.
Data came from the Pooling Project on Diet and Coronary Heart Disease (8 prospective studies, 1974-1996; n = 192,067 ...women and 74,720 men, aged 40-89 years).
During follow-up, 4326 cases of CHD were reported. Relative to never smokers, CHD risk among current smokers was highest in the youngest and lowest in the oldest participants. For example, among women aged 40 to 49 years the hazard ratio was 8.5 (95% confidence interval CI = 5.0, 14) and 3.1 (95% CI = 2.0, 4.9) among those aged 70 years or older. The largest absolute risk differences between current smokers and never smokers were observed among the oldest participants. Finally, the majority of CHD cases among smokers were attributable to smoking. For example, attributable proportions of CHD by age group were 88% (40-49 years), 81% (50-59 years), 71% for (60-69 years), and 68% (≥ 70 years) among women who smoked.
Among smokers, the majority of CHD cases are attributable to smoking in all age groups. Smoking prevention is important, irrespective of age.
Objective To test the hypothesis that postmenopausal women who increase their alcohol intake over a five year period have a higher risk of breast cancer and a lower risk of coronary heart disease ...compared with stable alcohol intake.Design Prospective cohort study.Setting Denmark, 1993-2012.Participants 21 523 postmenopausal women who participated in the Diet, Cancer, and Health Study in two consecutive examinations in 1993-98 and 1999-2003. Information on alcohol intake was obtained from questionnaires completed by participants.Main outcome measures Incidence of breast cancer, coronary heart disease, and all cause mortality during 11 years of follow-up. Information was obtained from the Danish Cancer Register, Danish Hospital Discharge Register, Danish Register of Causes of Death, and National Central Person Register. We estimated hazard ratios according to five year change in alcohol intake using Cox proportional hazards models.Results During the study, 1054, 1750, and 2080 cases of breast cancer, coronary heart disease, and mortality occurred, respectively. Analyses modelling five year change in alcohol intake with cubic splines showed that women who increased their alcohol intake over the five year period had a higher risk of breast cancer and a lower risk of coronary heart disease than women with a stable alcohol intake. For instance, women who increased their alcohol intake by seven or 14 drinks per week (corresponding to one or two drinks more per day) had hazard ratios of breast cancer of 1.13 (95% confidence interval 1.03 to 1.23) and 1.29 (1.07 to 1.55), respectively, compared to women with stable intake, and adjusted for age, education, body mass index, smoking, Mediterranean diet score, parity, number of births, and hormone replacement therapy. For coronary heart disease, corresponding hazard ratios were 0.89 (0.81 to 0.97) and 0.78 (0.64 to 0.95), respectively, adjusted for age, education, body mass index, Mediterranean diet score, smoking, physical activity, hypertension, elevated cholesterol, and diabetes. Results among women who reduced their alcohol intake over the five year period were not significantly associated with risk of breast cancer or coronary heart disease. Analyses of all cause mortality showed that women who increased their alcohol intake from a high intake (≥14 drinks per week) to an even higher intake had a higher mortality risk that women with a stable high intake.Conclusion In this study of postmenopausal women over a five year period, results support the hypotheses that alcohol intake is associated with increased risk of breast cancer and decreased risk of coronary heart disease.