Fine particles and mortality.

In an interesting paper in a recent issue (vol 17, issue 14) of the journal, Enstrom examined the association between fine particulate matter (PM) pollution and mortality in a cohort of elderly Californians. The analyses used proportional hazards regression and after adjusting for age, sex, cigarette smoking, and other potential confounders, Enstrom concluded, “These epidemiologic results do not support a current relationship between fine particulate pollution and total mortality in elderly Californians, but they do not rule out a small effect, particularly before 1983.” Enstrom’s analyses were based on a subcohort of individuals enrolled in the first Cancer Prevention Study (CPS I) conducted by the American Cancer Society (ACS). Enstrom’s conclusion is consistent with the conclusions of a cohort study among veterans conducted by Lipfert et al. (2000), but is at odds with the results from analyses of the second ACS cohort (CPS II) by Pope and others (Pope et al., 1995, 2002; Krewski et al., 2000), which reported statistically significant associations between fine particulate pollution and mortality. Every epidemiological study has weaknesses and limitations and, undoubtedly, both proponents and skeptics of the ‘fine particles cause death’ thesis will find much to criticize in the studies that do not support their conclusions. These discrepant results raise an important question, however. Can contemporary epidemiological and statistical tools reliably detect miniscule risks, particularly with strong risk factors as potential confounders? All the cohort studies referred to above use proportional hazards modeling for data analyses. But is proportional hazards really the appropriate tool for these analyses? First, it is highly unlikely that proportionality of hazards would hold over the entire period of time covered by these studies. Statistical tests for departures from proportionality of hazards have low power. Enstrom states that, in his analyses, these tests failed to reject proportionality of hazards. However, his finding of a higher relative risk associated with fine particles over the period 1973–1982 is inconsistent with proportionality of hazards over the entire