Political Statistics An Economist and Statistician Examines the Scientific Standards Used in Secondhand-Smoke Studies
Mark Edward Stover
From the Print Edition:
Bill Cosby, Autumn 94
"It is now proved beyond doubt that smoking is one of the leading causes of statistics."
-- Fletcher Knebel, Reader's Digest, December 1961
The antismoking hysteria in the United States continues. No longer confined to exhortations aimed at smokers to quit, smoking bans in public venues and the workplace are increasingly common. The belief that secondhand smoke represents a public-health hazard provides support for these restrictions. However, the argument that environmental tobacco smoke--as the scientific literature refers to it--is harmful remains open to question.
Much of the concern stems from the supposed similarities between mainstream smoke, the smoke inhaled by the smoker, and second-hand smoke. Yet there exist important differences. Secondhand smoke consists primarily not of mainstream smoke but of sidestream smoke. This is the smoke emitted by smoldering tobacco between puffs. Mainstream smoke is hotter and much more concentrated. It is thus physically and chemically different from secondhand smoke. The diluted nature of sidestream smoke means that the exposure levels are unequal. James Enstrom, M.D., professor of epidemiology at UCLA, estimates that exposure to secondhand smoke is equivalent to consuming only a few cigarettes per year. There are no studies that examine the health effects of such low levels of smoking.
Whether secondhand smoke causes lung cancer has been the subject of various studies. In most statistical analysis, if the probability that the effect is due to chance is less than 5 percent, statisticians accept the hypothesis that there is a cause-and-effect relationship. However, the majority of secondhand-smoke studies fail to meet this criterion of statistical significance. Even among those that show an increased risk, the level of increase is too small--most scientific standards would dismiss the relationship.
These studies concentrate on the incidence of lung cancer among nonsmokers exposed to secondhand smoke. In Environmental Tobacco Smoke and Morality, published in 1992 (Karger Press, Basel, Switzerland), a review of 12 studies of workplace exposure to secondhand smoke found no increased risk of lung cancer among nonsmokers. Workplace exposure is arguably the most relevant to public-health authorities. Nevertheless, the Environmental Protection Agency examined pregnant women whose husbands did or did not smoke at home. This assumes that people who live with smokers experience greater exposure to secondhand smoke than people who do not. There is no measure of actual exposure. The EPA's December 1992 report, "Respiratory Health Effects of Passive Smoking: Lung Cancer and Other Disorders," reviewed 11 such studies conducted in the United States. Only one found a statistically significant increase in lung cancer among the wives of smokers. In fact, three actually observed a decrease in lung-cancer rates.
As reported in the July 31, 1992, issue of Science, analysts at the EPA then "engage in some fancy statistical footwork." They used a procedure known as meta-analysis to pool the results of the different studies. Given the heterogenous and diverse nature of the studies, one may question whether the application of this technique is appropriate. However, this procedure revealed only a 1.19 to 1 increase in lung cancer among the nonsmokers married to smoking husbands, which is not considered significant.
In its November 1992 issue the American Journal of Public Health published another comparison of lung-cancer rates among women whose spouses were both smokers and nonsmokers. Among women married to men who smoked a low-to-moderate number of cigarettes, this study found a 1 to 0.7 ratio of risk, or a very small decrease in risk. (Most scientific studies demand a minimum 3 to 1 ratio of risk to prove a causal relationship and to rule out coincident circumstances.) Among women married to heavy smokers, a 1.3 to 1 ratio of risk was found--a minimal increase. At the very least these results do not suggest that secondhand smoke causes lung cancer.
This literature is also fraught with a number of potential methodological problems. One of the most important concerns is smoking-status misclassification. The identification of former smokers as individuals who have never smoked can alter the results of a test for risk from secondhand-smoke exposure. Also socioeconomic differences in households where smoking occurs may further explain any increase in risk. Attempts to control for such problems are difficult.
The evidence concerning the health effects of secondhand smoke is weak and contradictory. A disinterested review finds that is does not warrant concerns for the public health and certainly does not justify the extreme public-health measures proposed by the EPA and in Congressional committees. The actions of the agency in this regard represent another example of government meddling. The ultimate aim is to reduce smoking by making it inconvenient. It is unfortunate that many authorities believe that this goal justifies a misrepresentation of the facts.
You must be logged in to post a comment.