Short-term effects of air pollution on mortality in the cities of Rouen and Le Havre, France, 1990-1995.

In this study, the authors examined the short-term effects of ambient air pollution on mortality across 2 French cities: Rouen and Le Havre. In Poisson regression models, which controlled for day-of-week effects, the authors used nonparametric smoothing to control for temporal trend, weather, and influenza epidemics. In Rouen, an interquartile range increase of 60.5-94.1 microg/m3 of ozone was associated with an increase of 4.1% (95% confidence interval = 0.6, 7.8) of total mortality. Daily variations in sulfur dioxide (interquartile range increase = 17.6-36.4 microg/m3) were also associated with an 8.2% increase (95% confidence interval = 0.4, 16.6) in respiratory mortality. An increase of 6.1% (95% confidence interval = 1.5, 10.9) of cardiovascular mortality was also observed with an interquartile range increase of nitrogen dioxide (i.e., 25.3-42.2 microg/m3). With respect to Le Havre, an interquartile range increase in daily levels of sulfur dioxide (11.3-35.6 microg/m3) was associated with an increase of approximately 3% (95% confidence interval = 0.8, 5) of cardiovascular mortality. For particulate matter less than or equal to 13 microm in diameter (interquartile increase = 21.5, 45.4 microg/m3), an increase of 6.2% (95% confidence interval = 0.1, 12.8) was observed. The estimates of pollutant effects and their standard deviations were slightly affected by the degree of smoothing temporal variations in this study. When low collinearity was present, the 2-pollutant models provided acceptable estimates of pollutant effects. They suggested that the ozone effect was independent of the Black Smoke effect, and that the effects of sulfur dioxide and nitrogen dioxide were unlikely to be confounded by ozone concentrations. However, high collinearity leads to large estimates of the pollutant coefficient variances and, therefore, leads to inaccurate estimates of pollutant effects. The analysis of the contributory effects of different pollutant mixtures requires further investigation in those instances in which high collinearity between pollutants is present.