More misleading science from the tobacco industryBMJ 2003; 327 doi: https://doi.org/10.1136/bmjusa.03070002 (Published 19 November 2003) Cite this as: BMJ 2003;327:E237
- Michael J Thun, vice president, epidemiology and surveillance research ()
From BMJ USA 2003;July:352
The article by Enstrom and Kabat (p 369) is the latest in a long series of publications funded by the tobacco industry that report little or no relationship between environmental tobacco smoke (ETS) and disease.1 The current study has an aura of legitimacy because it is drawn from the American Cancer Society (ACS) Cancer Prevention Study I (CPS-I), a landmark prospective study of the hazards of active smoking,2 and because the analyses are based on nearly 40 years of data. Despite these apparent strengths, the study by Enstrom and Kabat is uninformative and its conclusions are exaggerated.
Indeed, the negative conclusions were entirely predictable from the outset because of the flawed way in which exposure to ETS was classified. CPS-I collected no information on ETS exposure other than the smoking status of the spouse. The study began in 1959, an era in which secondhand smoke was pervasive; virtually everyone was exposed at work, in social settings, or in other activities of daily living. Enstrom and Kabat therefore could not identify a comparison group of “unexposed” persons. Their analyses essentially compare nonsmokers married to a smoking spouse to nonsmokers with other sources of ETS exposure. This potential for misclassification was further exacerbated in the final 26 years of follow-up after 1972, when no updated information on the smoking status of the spouse was collected. Many of the spouses who reported smoking at the start of the study would have quit, died, or ended the marriage, yet the surviving partner was still classified as “exposed” in the analysis. Enstrom and Kabat could not distinguish persons with continuing exposure to ETS from those with past exposure, or persons with multiple sources of ETS exposure from those exposed only to spousal smoking.
This is not the first time that tobacco industry consultants have published misleading reports about ETS exposure based on flawed analyses of ACS cohort studies. In 1995, LeVois and Layard reported no association between ETS exposure and coronary heart disease (CHD) mortality either in CPS-I or in CPS-II, an even larger ACS cohort study begun in 1982.3 ACS repeatedly communicated to these researchers and to Enstrom that CPS-I was not informative for evaluating ETS because of the deficiencies in the exposure data.4 Enstrom and Kabat lament the exclusion of the earlier CPS-I report from meta-analyses of ETS exposure and CHD, but they do not discuss the methodologic limitations that led to its exclusion. Nor do the authors cite the reanalysis of the CPS-II data by Steenland et al,5 the conclusions of which do not support their views or the position of the tobacco industry.
The reviewers at the BMJ may have been persuaded that the resurvey of 681 subjects in 1999 ensured the validity of the exposure data, or that the size of the CPS-I cohort provided adequate statistical power to evaluate the relationship between ETS and any disease end point. However, respondents to the 1999 resurvey comprised only 2% of the original 35 561 married lifelong nonsmokers who originally enrolled and 5% of those who were followed after 1972. A close examination of the seemingly extensive data on the 1999 respondents provided in Tables 2 through 6 (available at http://bmj.com/cgi/content/full/326/7398/1057) reveals that the data are actually sparse; these tables include many empty cells or percentages based on fewer than five subjects.
The number of deaths from lung cancer in CPS-I is considerably smaller than in numerous other ETS studies. Enstrom and Kabat based their analysis on married, lifelong nonsmokers drawn from the 10% of subjects who live in California. Because lung cancer is rare in lifelong nonsmokers, the relatively small number of observed deaths in this subgroup gives rise to wide 95% confidence intervals that in fact overlap with relative risk estimates cited in meta-analyses by the International Agency for Research on Cancer (IARC),6 the Environmental Protection Agency (EPA),7 and others.8–9
The article selectively omits mention of studies and reviews of ETS published after 1999, as well as reports that addressed the tobacco industry's standard criticisms of the scientific literature on ETS, such as the California EPA report.10 No mention is made of the 2002 report by IARC that concluded that involuntary exposure to tobacco smoke is carcinogenic.6 Enstrom and Kabat minimize the extent of the published literature linking ETS exposure to lung cancer. Recent reviews8–9 list 51 relevant studies, many of which collected information on ETS exposure from other sources, as well as from smoking by the spouse. These studies had been conducted during time periods and in countries where the potential for exposure misclassification was substantially less than in CPS-I. Enstrom and Kabat incorrectly claim that ETS exposure data in CPS-I are comparable in quality to those of most other studies.
ACS does not endorse or agree with the methods and conclusions of the study by Enstrom and Kabat. ACS once collaborated with Enstrom, beginning in 1989, to enable him to extend follow-up of the California subgroup of CPS-I from 1972 forward. The acknowledgments at the end of the article thank two former ACS vice presidents for this assistance but fail to mention that ACS terminated its collaboration with Enstrom several years ago because of his unwillingness to address fundamental methodologic problems with this and other analyses. At no point did Enstrom inform ACS that he had communicated with Philip Morris about the potential value of the CPS-I follow-up in 199011 or that he had applied for and received funding from Philip Morris in 1997.12
Sadly, the forum in which the study by Enstrom and Kabat will be most influential is not the scientific world—most scientists will recognize the study for what it is—but in the battle for public support in communities that are considering more stringent regulations on smoking in public spaces. Misleading publicity appears repeatedly in newspapers and other community publications wherever clean air laws are being debated.13 For example, the methodologically flawed study by LeVois and Layard was recently cited in an op-ed piece in a Seattle newspaper in March 2003,14 eight years after it had been published in an obscure journal.3 The Enstrom and Kabat paper represents an even more valuable public relations resource for tobacco companies. They can refer not only to the ACS as the data source for the study but also to the prestigious BMJ as the journal in which it was published.
Papers p 369
Competing interests None declared.