Mortality and air pollution: lessons from statistics

Cross sectional studies which attempt to link persistent geographic differences in mortality rates with air pollution are reviewed. Some early studies are mentioned and detailed results are given for seven major contemporary studies, two of which are still in the publication process. Differences among the studies are discussed with regard to statistical techniques, trends in the results over time (1959 to 1974), and interpretation and use of the results. The analysis concludes that there are far too many problems with this technique to allow causality to be firmly established, and thus the results should not be used for cost benefit or policy analysis