We are pleased to present in this issue a special supplement on ozone and mortality (see pages 427–468). The unusual history of this supplement deserves explanation. In August 2004, the journal received 3 independent submissions. Each was a meta-analysis of published studies on ozone and mortality. One of the cover letters provided this background:
In the summer of 2003, the US Environmental Protection Agency decided to fund 3 separate research teams to investigate a controversial and unresolved question—whether the epidemiological literature to date suggested a relationship between ozone exposure and risk of premature death. The EPA currently does not include this relationship in regulatory impact analyses, and changing or confirming this approach could significantly influence air pollution policy in the United States and elsewhere.
The 3 teams received identical working databases of studies from EPA and communicated briefly with major points of clarification, but conducted the analyses separately. The goal was to see whether differences in analytical methods or subjective decisions by the researchers would lead to similar or different conclusions. The 3 groups distributed copies of the manuscripts to one another in July 2004 to ensure similarity of frameworks, but no attempt was made to edit manuscripts to reconcile differences in findings.
We were intrigued on several counts. One is the creative approach taken by EPA in commissioning 3 independent analyses. Any consensus that would emerge from 3 separate analyses is bound to carry considerable weight. However, nearly as interesting is the experiment itself: what happens when separate teams of researchers tackle a single (if highly complex) epidemiologic analysis? What does this experiment tell us about epidemiologic analysis itself?
Of course, these questions would be moot if the papers1–3 themselves were not well done. In keeping with the independent development of these papers, we solicited reviews from 3 separate sets of experts who (without knowing of the other papers) ranked each manuscript highly enough for publication on its own merits. Our thanks to reviewers Bert Brunekreef, Aaron Cohen, Frank Gilliland, Stephanie London, Dana Loomis, Duncan Thomas, and Erich Wichmann for their part in this experiment.
Having accepted the papers for publication, we proceeded to invite commentaries4,5 from 2 distinguished epidemiologists. David Bates4 (whose career spans the half-century from London's “killer fogs” to recent air pollution policies) puts these papers into a public health context. He concludes that the need to reduce ambient ozone exposures is now “urgent.” Steven Goodman5 reflects on the analytic process and comments on what may be a nearly-inevitable bias in the accumulation of single-site studies. As he also notes, these ozone analyses demonstrate our unwarranted confidence in “confidence intervals”—brackets that capture statistical uncertainty but not the host of other factors that blur the truth.
We thank EPA for covering the publishing costs of this special supplement. This funding allows us to bring these papers to press with highest priority.
1. Bell ML, Dominnici F, Samet JM. A meta-analysis of time-series studies of ozone and mortality with comparison to the National Morbidity, Mortality, and Air Pollution Study. Epidemiology. 2005;16:436–445.
2. Ito K, DeLeon SF, Lippmann M. Associations between ozone and daily mortality: analysis and meta-analysis. Epidemiology. 2005;16:446–457.
3. Levy JI, Chemerynski SM, Sarnat JA. Ozone exposure and mortality: an empirical Bayes meta-regression analysis. Epidemiology. 2005;16:458–468.
4. Bates DV. Ambient ozone and mortality [Commentary]. Epidemiology. 2005;16:427–429.
5. Goodman SN. The methodologic ozone effect [Commentary]. Epidemiology. 2005;16:430–435.
© 2005 Lippincott Williams & Wilkins, Inc.