[Long-term exposure to air pollution and natural mortality: variations related to the use of different exposure indicators in the cohorts of BIGEPI project].

Abstract

Objectives: appropriate assessment of exposure to air pollution is crucial for the estimation of adverse effects on human health, both in the short and long term. Within the BIGEPI project, different indicators of long-term exposure to air pollution, in association with mortality by cause, were tested within the Italian longitudinal metropolitan studies (LMS). This allowed an evaluation of differences in effect estimates using the different exposure indicators.

Design: closed cohort.

Setting and participants: subjects aged >=30, who took part in the 2011 census, residents in 5 cities (Turin, Bologna, Rome, Brindisi and Taranto).

Main outcome measures: at the time of enrolment, residential exposure levels to particulate matter <=10 μm (PM10), PM <=2.5 μm (PM2.5), nitrogen dioxide (NO2) and ozone (O3) for the period April-September (O3 warm season) were obtained from models at different spatial resolutions, from 1x1km to 200x200m (from the BEEP project) to 100x100m (ELAPSE project). In addition, locally developed models were used in each area (FARM photochemical model at 1x1-km for the cities of Rome, Taranto and Brindisi, Land-Use Regression (LUR) model for the city of Turin, PESCO model for Bologna). Cox proportional hazards models were applied to assess the association between exposure to air pollution (assessed using different exposure indicators) and natural mortality, adjusting for both individual and area covariates.

Results: the exposure levels derived by the different models varied between pollutants, with differences between the averages ranging from 3 to 20% for PM10, from 1 to 23% for PM2.5, and from 3 to 28% for NO2; the results for O3 were more heterogeneous. A total of 267,350 deaths from natural causes were observed. There is low heterogeneity in the effect estimates calculated from different environmental models, while there is greater variability in average exposure values, with different behaviour depending on the model and the characteristics of the area investigated. Differences are more pronounced where local risk factors are relevant, e.g., in industrial cities, thus suggesting the need of considering industrial exposure separately from other sources.

Conclusions: the numerous heterogeneities in the data used make it difficult to draw conclusions about the comparisons studied. Nevertheless, this study suggests that different approaches to the assessment of environmental exposure should be evaluated depending on the national or local level of interest, also according to the specifities of the investigated areas.