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.

Ambient air pollution has been associated with lung cancer, but the shape of the exposure-response function - especially at low exposure levels - is not well described. The aim of this study was to address the relationship between long-term low-level air pollution exposure and lung cancer incidence. The "Effects of Low-level Air Pollution: a Study in Europe" (ELAPSE) collaboration pools seven cohorts from across Europe. We developed hybrid models combining air pollution monitoring, land use data, satellite observations, and dispersion model estimates for nitrogen dioxide (NO2), fine particulate matter (PM2.5), black carbon (BC), and ozone (O3) to assign exposure to cohort participants' residential addresses in 100 m by 100 m grids. We applied stratified Cox proportional hazards models, adjusting for potential confounders (age, sex, calendar year, marital status, smoking, body mass index, employment status, and neighborhood-level socio-economic status). We fitted linear models, linear models in subsets, Shape-Constrained Health Impact Functions (SCHIF), and natural cubic spline models to assess the shape of the association between air pollution and lung cancer at concentrations below existing standards and guidelines. The analyses included 307,550 cohort participants. During a mean follow-up of 18.1 years, 3956 incident lung cancer cases occurred. Median (Q1, Q3) annual (2010) exposure levels of NO2, PM2.5, BC and O3 (warm season) were 24.2 µg/m3 (19.5, 29.7), 15.4 µg/m3 (12.8, 17.3), 1.6 10-5m-1 (1.3, 1.8), and 86.6 µg/m3 (78.5, 92.9), respectively. We observed a higher risk for lung cancer with higher exposure to PM2.5 (HR: 1.13, 95% CI: 1.05, 1.23 per 5 µg/m3). This association was robust to adjustment for other pollutants. The SCHIF, spline and subset analyses suggested a linear or supra-linear association with no evidence of a threshold. In subset analyses, risk estimates were clearly elevated for the subset of subjects with exposure below the EU limit value of 25 µg/m3. We did not observe associations between NO2, BC or O3 and lung cancer incidence. Long-term ambient PM2.5 exposure is associated with lung cancer incidence even at concentrations below current EU limit values and possibly WHO Air Quality Guidelines.