Disentangling associations between multiple environmental exposures and all-cause mortality: an analysis of European administrative and traditional cohorts

Abstract

Background: We evaluated the independent and joint effects of air pollution, land/built environment characteristics, and ambient temperature on all-cause mortality as part of the EXPANSE project.

Methods: We collected data from six administrative cohorts covering Catalonia, Greece, the Netherlands, Rome, Sweden, and Switzerland and three traditional cohorts in Sweden, the Netherlands, and Germany. Participants were linked to spatial exposure estimates derived from hybrid land use regression models and satellite data for: air pollution [fine particulate matter (PM_2.5), nitrogen dioxide (NO₂), black carbon (BC), warm season ozone (O₃)], land/built environment [normalized difference vegetation index (NDVI), distance to water, impervious surfaces], and ambient temperature (the mean and standard deviation of warm and cool season temperature). We applied Cox proportional hazard models accounting for several cohort-specific individual and area-level variables. We evaluated the associations through single and multiexposure models, and interactions between exposures. The joint effects were estimated using the cumulative risk index (CRI). Cohort-specific hazard ratios (HR) were combined using random-effects meta-analyses.

Results: We observed over 3.1 million deaths out of approximately 204 million person-years. In administrative cohorts, increased exposure to PM_2.5, NO₂, and BC was significantly associated with all-cause mortality (pooled HRs: 1.054, 1.033, and 1.032, respectively). We observed an adverse effect of increased impervious surface and mean season-specific temperature, and a protective effect of increased O₃, NDVI, distance to water, and temperature variation on all-cause mortality. The effects of PM_2.5 were higher in areas with lower (10th percentile) compared to higher (90th percentile) NDVI levels [pooled HRs: 1.054 (95% confidence interval (CI) 1.030-1.079) vs. 1.038 (95% CI 0.964-1.118)]. A similar pattern was observed for NO₂. The CRI of air pollutants (PM_2.5 or NO₂) plus NDVI and mean warm season temperature resulted in a stronger effect compared to single-exposure HRs: [PM_2.5 pooled HR: 1.061 (95% CI 1.021-1.102); NO₂ pooled HR: 1.041 (95% CI 1.025-1.057)]. Non-significant effects of similar patterns were observed in traditional cohorts.

Discussion: The findings of our study not only support the independent effects of long-term exposure to air pollution and greenness, but also highlight the increased effect when interplaying with other environmental exposures.

Keywords: NDVI; administrative cohorts; air pollution; all-cause mortality; ambient temperature; external exposome; traditional adult cohorts.