May 14, 1999
Mr. Frank Burgos Editorial Page Editor
Philadelphia Daily News
400 N. Broad Street
Philadelphia, Pennsylvania 19130-4015Dear Mr. Burgos:
A commentary by Jeff Jacoby that appeared in the May 12, 1999, issue of the Philadelphia Daily News challenges the science used by the Centers for Disease Control and Prevention (CDC) to estimate how many people die from cigarette smoking each year in the United States. The commentary was based on an article by Robert Levy and Rosalind Marimont in the latest issue of Regulation magazine, which contains numerous errors about the harm of smoking and the risks of secondhand smoke. We would like to take this opportunity to clarify the method by which CDC estimates smoking-related deaths.
First, Levy and Marimont claim that the government counts as a smoking-related death all smokers who die from a certain disease, even if they had other risk factors for that disease. This is not true. For each disease, CDC attributes only a percentage of the deaths as being due to smoking, based on the best medical science. For example, for heart disease, CDC estimates that the proportion of deaths due to smoking is only 16 percent for persons age 65 and older. For lung cancer, in which the authors acknowledge smoking to be a "high risk factor," CDC considers only 83 percent of the deaths as being smoking-related. [CDC attributable risk calculations are typically based on odds ratios from case-control epidemiologic studies -- hardly the "best medical science." While case-control epidemiology may be useful for identifying high risks of rare diseases -- like smoking and lung cancer -- it is notoriously unreliable for low risks of common diseases -- like smoking and heart disease. For example, epidemiologic studies report about the same size statistical association for smoking/heart disease as for secondhand smoke/heart disease. But if the smoking/heart disease association was true, it should be much higher as smokers are also exposed to their own secondhand smoke.]
The authors also stress that other risk factors must be statistically controlled for if the impact of a single factor like smoking is to be reliably determined. We agree and conducted a careful analysis to examine that very issue. Our findings concluded that controlling for other risk factors changed the proportion of deaths attributed to lung cancer by only one to two percent, and the proportion of deaths from heart disease by less than one percent--hardly the huge impact alleged by the authors. [Pretty vague and probably untrue. For example, heart disease has many risk factors and probably no study is based on a complete set of risk factors. Excessive smoking is more than a risky habit - - it's a risky lifestyle. Smokers tend to eat poorer quality diets, drink more, get less exercise, etc. Keep in mind, the largest study-ever of heart disease, the World Health Organization's MONICA, failed to find associations with traditional risk factors like high cholesterol, hypertension and smoking.]
The authors also claim that many of the deaths from tobacco are not premature deaths. However, studies that have followed smokers and nonsmokers for many years have found that smokers are three times more likely to die between the ages of 45 and 64 and two times more likely to die between the ages of 65 and 84 than those who have never smoked. Although a certain proportion of smoking-related deaths occur among older Americans, the fact is that 33 percent of non-smokers live to age 85, compared with only 12 percent of smokers. [For actuarial purposes, "premature" deaths are those that occur before attaining life expectancy. Any other definition is so speculative as to be flaky. Men who die after about 72 and women who die after about 78 are not premature deaths -- even if they smoked.]
Finally, the authors say that smoking-related deaths estimated by CDC are not real deaths, but "computer-generated phantom deaths" using non-representative populations to calculate risk. In 1989, the State of Oregon asked physicians to report on death certificates whether tobacco use contributed to the death. Between 1989 and 1996, physicians reported that tobacco contributed to 20 percent of Oregon deaths, the exact percentage of deaths attributed to smoking over the same time period using CDC's method. The CDC estimate and the Oregon death certificate data differed in their cumulative estimates of the number of smoking-attributable deaths for the eight years by only 61 deaths -- a difference of about one tenth of one percent. This real-life experience provides strong evidence that the statistical methods used by CDC provide an accurate calculation of the real deaths occurring daily in the United States that are caused by tobacco use. [Unverified reports by lone physicians are hardly conclusive or ipso facto reliable. That CDC can point to a coincidence with one set of statistics is probably just that.]
Cigarette smoking and other tobacco use is the single most-studied health risk factor in the history of medicine. Scientific facts support our estimate that each year, more than 400,000 deaths in this country are prematurely caused by smoking-related diseases. [There is no question that excessive smoking is a significant health risk. Pushing bogus statistics on an unsuspecting public does not make this fact more compelling.]
Sincerely yours,
Michael P. Eriksen, Sc.D.
Director
Office on Smoking and Health
National Center for Chronic Disease Prevention and Health Promotion
Comments on this posting?
Click here to post a public comment on the Trash Talk Bulletin Board.
Click here to send a private comment to the Junkman.