Estimation of long-term average exposure to outdoor air pollution for a cohort study on mortality.

Recent prospective cohort studies have suggested that long-term exposure to low levels of particulate matter (PM) air pollution is associated with increased mortality due to, especially, cardio-pulmonary disease. Exposure to ambient air pollution was estimated mostly as city average concentrations, assuming homogenous exposure within the city. We used an ongoing cohort study - The Netherlands Cohort Study (NLCS) on diet and cancer - to investigate the relationship between traffic-related air pollution and mortality. The baseline data collection took place in 1986. A study was conducted to develop methods for exposure assessment and evaluate the contrast in exposure to air pollution within the cohort. Assessment of long-term exposure to two traffic-related air pollutants, Black Smoke (BS) and Nitrogen Dioxide (NO(2)), consisted of separate estimation of regional background, urban background, and local traffic contributions at the home address. Interpolation of concentration data from a routine monitoring network was used to estimate the regional background concentration. A regression model relating degree of urbanization to air pollution was used to allow for differences between different towns/neighborhoods of cities. Distance to major roads was calculated to characterize local traffic contributions, using a Geographic Information System (GIS). Interpolation resulted in reasonably precise regional background estimation when distant sites were not used and distance squared was used as the weight. Cross-validation showed that prediction errors were about 15% of the range in regional background concentration. Urban and local scales contributed significantly to the contrast within the cohort. Prediction errors for estimating the urban background were about 25% of the range in background concentrations. When the developed model was applied to the study cohort, there was substantial contrast in estimated exposure to BS and NO(2). About 90% of the study population lived 10 years or more at its 1986 home address - supporting the use of the estimated concentration at the 1986 address as a relevant exposure variable.