Trends in relative risk estimates for the association between air pollution and mortality in The Netherlands, 1992-2006

Background: Daily variations in the levels of air pollution are well known to be associated with daily variations in mortality counts. Given the large number of time-series studies, there is little need for simple replication of these results in additional locations. However, additional analyses of time-series data might be useful in elucidating remaining questions on the role of air pollution on mortality. Objectives: Because of ongoing issues related to causality, changing toxicity, the difficulty in isolating the independent effects of individual pollutants, the availability of new methods to detect effect thresholds, and questions about the extent to which effects are restricted to frail members of the population, additional analyses of time-series data might be helpful in addressing these issues. We show an example where additional time-series analyses can be helpful in elucidating specific questions in the field of air pollution epidemiology. Methods: We analysed daily mortality and air pollution data using Poisson regression in generalised additive models. Air pollution data for the overall period 1992-2006 and for four different periods were analysed to assess the overall risk estimates for the whole period and to assess variability over time for the different effect estimates. Results and Conclusion: We found some statistically significant upward trends, but this was only the case for a few associations without a consistent pattern over the cause-specific deaths. Whether these findings are consistent over time or whether our findings are merely the result of statistical chance can only be elucidated by continuation of monitoring of the relative risks over time in the future. Although these results may indicate that both photochemical and particulate matter air pollution might have become more toxic, the lack of a clear pattern in the results makes these conclusions speculative. (C) 2010 Elsevier Inc. All rights reserved.