[Short-term effects of air pollution in Italy: risk heterogeneity from 1996 to 2005]

Abstract

Background: in the epidemiological literature heterogeneity of short-term effects of air pollutants among different populations is widely documented. It is commonly attributed to air pollutants characteristics, modality of exposure, individual susceptibility or pollutant-temperature interaction. The case of a time trend in effect size is questionable. In the present paper we compare results of Italian studies on short-term effects of air pollutants based on data for the calendar years 1996-2005. Study objective is to evaluate time stability of risk estimates taking into account of between cities heterogeneity.

Materials and methods: daily series of deaths for natural causes and daily series of average air pollutants concentrations were obtained from the MISA and EpiAir datasets. Effect estimates were obtained by Bayesian random effect meta-analysis to cope with between city heterogeneity.

Results: there was no difference in effect estimates using the case crossover or Poisson regression approach on the time period 2001-2005 and the ten Italian cities of the EpiAir study. Using the MISA dataset (time period 1996-2002) and the same statistical approach (Poisson regression with seasonality regression spline) we compared the overall effect estimates selecting different subset of italian cities. The EpiAir cities-subset showed higher risk estimates either for PM10 or NO2. Last, considering the same subset of cities we found an increase in percent variation of mortality for natural causes for 10 microg/m3 of PM10 from 0,36% (CI 95% 0.1;0.8) in the period 1996-2002 to 0.66% (0.4;0.9) in the period 2001-2005. For NO2 we found respectively a change from 0.72% (0.3;1.1) to 1.12% (0.5;1.6).

Conclusion: the results of the EpiAir study are not immediately comparable to the MISA results because the the set of cities included in the two studies is different, the statistical approach is different, and the calendar time period is different. The present analysis shows that considering the subset of cities for which we have data for both time periods and using the same statistical approach the short-term effect of air pollutants on natural mortality is increasing.