Correspondence

Air Pollution and Mortality

N Engl J Med 1994; 330:1237-1238April 28, 1994

Article

To the Editor:

Because of the possibility of residual confounding, relative risks of the order of 1.3 or less, as reported by Dockery et al. in their study of air pollution and mortality (Dec. 9 issue),1 are difficult to interpret in epidemiologic investigations. As is common in epidemiologic studies, Dockery et al. adjusted for age by stratifying subjects into five-year age groups. Because of the strong influence of age on lung cancer and cardiovascular mortality, this stratification system may not be fine enough to detect small relative risks. For example, the mortality rate from lung cancer among smokers increases roughly in association with the seventh power of age2. The average age of the study subjects from Steubenville, Ohio, was 51.6 years, whereas it was 48.4 in the subjects from Portage, Wisconsin. Among smokers, the risk of lung cancer at the age of 51.6 relative to that at the age of 48.4 is roughly (51.6/48.4)⁷, or 1.57. A similar computation for nonsmokers (the mortality rate from lung cancer increases roughly in association with the fourth power of age) yields a relative risk of 1.29. Thus, if the average difference in age reported in Table 1 of the article by Dockery et al. results from the Steubenville subjects' being somewhat older than the other subjects in many of the age strata, then the modest “Steubenville effect” reported in the paper may be fully explained by this difference in age alone.

Inadequate adjustment for smoking status could also contribute to the results reported by Dockery et al. Although information on individual subjects' level of smoking was available, it is not quite clear how this information was used. There is also reason to believe that the relative risk associated with smoking is not constant, as assumed by Dockery et al., but depends on age2,3. Finally, I do not believe that socioeconomic status and occupational exposure were adequately adjusted for in this study.

Even small risks, if associated with the current levels of air pollution in the United States, are of enormous importance to the public health because of the ubiquity of exposure. I believe that the approach used by Dockery et al. is the best one for the detection of small effects. Unfortunately, precisely because of the small size of the effects to be detected, even more careful control of confounding is needed.

Suresh H. Moolgavkar, M.D., Ph.D.
Fred Hutchinson Cancer Research Center, Seattle, WA 98104

3 References

1. 1

   Dockery DW, Pope CA III, Xu X, et al. An association between air pollution and mortality in six U.S. cities. N Engl J Med 1993;329:1753-1759
   Full Text | Web of Science | Medline

2. 2

   Doll R. An epidemiological perspective of the biology of cancer. Cancer Res 1978;38:3573-3583
   Web of Science | Medline

3. 3

   Doll R, Peto R. Cigarette smoking and bronchial carcinoma: dose and time relationships among regular smokers and lifelong non-smokers. J Epidemiol Community Health 1978;32:303-313
   CrossRef | Web of Science | Medline

Author/Editor Response

The authors reply:

To the Editor: Moolgavkar suggests that the reported association between fine-particulate air pollution and mortality in six U.S. cities was the result of confounding due to inadequate control for age. Analysis of the data does not support this assertion. The degree of confounding can be assessed in terms of the stability of the estimated association with and without control of the posited confounder.1 In our data, the estimated mortality-rate ratio associated with the increase in fine-particle concentrations across the six cities and adjusted for sex, smoking status, level of education, and body-mass index but not for age was 1.33 (95 percent confidence interval, 1.14 to 1.54). After adjustment for these factors and for age stratified according to five-year age groups, the mortality-rate ratio was reduced to 1.26 (95 percent confidence interval, 1.08 to 1.47), as originally reported. Stratification of age in one-year age groups plus control for the other factors in the original model increased the estimated mortality-rate ratio to 1.29 (95 percent confidence interval, 1.11 to 1.50). The results for mortality from lung cancer and cardiopulmonary disease were similar. Thus, the use of five-year age categories for stratification was appropriate, as previously reported.

Moolgavkar further suggests that the possibility of residual confounding by some other factor or by inadequate modeling cannot be excluded. We agree that residual confounding is a concern in any epidemiologic study. In our analysis, we explicitly evaluated the stability of the association as covariates were incrementally added to the model. After an adjustment for age and sex, none of the posited confounders had more than a minimal effect on the variable estimates or confidence intervals. Although smoking status, level of education, body-mass index, occupational exposure, and a history of chronic diseases are predictive of mortality, there was no indication that any of these risk factors confounded the association between fine-particulate air pollution and mortality. Moreover, the stability of these estimates suggests that the observed associations were not due to confounding by unknown or unmeasured risk factors or to inappropriately modeled covariates.

Douglas W. Dockery, Sc.D.
Harvard School of Public Health, Boston, MA 02115

C. Arden Pope, III, Ph.D.
Brigham Young University, Provo, UT 84602

1 References

1. 1

   Rothman KJ. Modern epidemiology. Boston: Little, Brown, 1986.
   

Citing Articles (5)

Citing Articles

1. 1

   Sonja Greven. (2011) An Approach to the Estimation of Chronic Air Pollution Effects Using Spatio-Temporal Information. Journal of the American Statistical Association 106:494, 396-406
   CrossRef

2. 2

   Douglas W. Dockery. (2009) Health Effects of Particulate Air Pollution. Annals of Epidemiology 19:4, 257-263
   CrossRef

3. 3

   Klaus-Michael Ochsenkühn, Theopisti Lyberopoulou, G. Koumarianou, Maria Ochsenkühn-Petropoulou. (2008) Ion chromatographic and spectrometric determination of water-soluble compounds in airborne particulates, and their correlations in an industrial area in Attica, Greece. Microchimica Acta 160:4, 485-492
   CrossRef

4. 4

   Krewski, Daniel, , Burnett, Richard T., , Goldberg, Mark S., , Hoover, Kristin, , Siemiatycki, Jack, Abrahamowicz, Michael, , White, Warren H., . (2004) Validation of the Harvard Six Cities Study of Particulate Air Pollution and Mortality. New England Journal of Medicine 350:2, 198-199
   Full Text

5. 5

   Lucas M. Neas. (2000) Fine particulate matter and cardiovascular disease. Fuel Processing Technology 65-66, 55-67
   CrossRef