Topological modelling of urban air pollution and cognition

Abstract The impact of air pollution on cognitive function has attracted repeated scrutiny, and current results point to a potentially detrimental role. The purpose of this study was to examine this association in geographic space. Cross-sectional, complete observational data from UK Biobank were extracted for the four metropolitan regions of Birmingham, Leeds, Liverpool and Manchester in England, UK, including three pollution indicators and two measures of cognitive performance. A set of additional covariates served to adjust for potential confounders. Spatial analyses for each region and combination of pollution indicator and cognition measure were conducted using mass-univariate linear regression in GeoSPM. Conventional non-spatial Bayesian regression models were used for comparison. A significant interaction between air pollution and cognitive performance was identified in 51 areas based on a two-tailed t-test at p < 0.05 FEW (voxel-level family-wise correction). In 29 of those areas, increased pollution and reduced cognition co-occur, and a pattern of central locations and primary roads emerges, suggesting a potentially harmful effect predicated by geography.