Spatial contrasts in air pollution and the effects on lung function in children

Abstract

Within the framework of the ESCAPE project (European Study of Cohorts for Air Pollution Effects), we aimed to characterize and explain spatial contrasts in ambient air pollution within and between European study areas. Following a standard protocol, project partners selected “street” and “background” measurement sites in 36 study areas, and measured using identical equipment. NO2 and NOX measurements were available from 40 different sites in each area (80 in The Netherlands/Belgium and Catalunya). PM2.5 (particles <2.5 µm), PM2.5 absorbance (the “blackness” of PM2.5 filters; an indicator for soot), PM10 (particles <10 µm) and PMcoarse (PM10 minus PM2.5) measurements were available for 20 of those areas, for a subset of 20 sites per area (40 in The Netherlands/Belgium and Catalunya). Significant contrasts between street and background concentrations were seen for all pollutants, but most clearly for NO2 and NOX. The measured concentrations formed the basis for the development of land use regression (LUR) models for each study area. LUR models for PM2.5, PM2.5 absorbance, PM10 and PMcoarse were developed by evaluating which environmental factors (represented by variables derived from a Geographic Information System (GIS)) were best able to explain the spatial variability in pollution concentration. Derived LUR models explained a moderate to high portion of variance for all pollutants using predictors related to traffic intensity, population density and nearby industrial and harbour sites. LUR models cannot accurately predict the high concentrations measured in urban canyons: narrow streets with (relatively) tall buildings. Using 3-dimensional building data and GIS, we derived four indicators for this canyon effect, among which the SkyView Factor (the total fraction of sky visible from the measurement site). Basic LUR models were developed based on NO2 and NOX data from 132 sites in the Netherlands. Although consideration of the canyon indicators increased the explained variance only modestly, the P10-P90 range of the SkyView Factor explained substantial concentration differences for both NO2 (5.6 μg/m³)and NOX (10.9 μg/m³). Particulate matter is a heterogeneous mixture of particles which differ in size, shape and composition and hence –presumably- toxicity. We estimated long-term exposure to eight elemental constituents of particulate matter (copper, iron, potassium, nickel, sulphur, silicon, vanadium, zinc) at the homes of 4659 children involved in five birth cohorts. FEV1 (forced expiratory volume in the first second of exhalation) was determined when the children were 6 or 8 years old. After combining cohort-specific findings into a meta-analysis, we found a small decrease in FEV1 associated with exposure to nickel and sulphur. However, heterogeneity was larger for the various PM constituents than for PM10 and PM2.5 mass, and low FEV1 was more consistently associated with PM mass. Recent advancements in the application of LUR models have targeted the concentration data, the predictor variables, the (traditionally typical) multivariable model structure and the performance evaluation. We identify the current challenge to disentangle the contribution of different pollutants to adverse health effects, which requires close collaboration between geographical study areas, between toxicologists and epidemiologists, but also between those who develop LUR models and those who apply them in epidemiology.