Congressional Research Service
* The Library of Congress *
Washington, DC 20540-7000

                                                        March 23, 1994

     FROM : Jane G. Gravelle
                 Senior Specialist in Economic Policy
                 Office of Senior Specialists

                 C. Stephen Redhead
                 Analyst of Biomedical Science
                 Science Policy Research Division

     SUBJECT: Discussion of Source of Claims of 50,000 Deaths
                         from Passive Smoking

          This memorandum is in response to your request for information
     on the possible source of an estimated premature 50,000 deaths from
     passive smoking effects.

          This estimate is much larger than the (disputed) estimate of
     3,000 premature deaths from lung cancer due to passive smoking
     effects that was reported in a recent study by the Environmental
     Protection Agency (hereafter EPA Report). [1] The 3,000 estimate is
     the only one mentioned in a recent article on the cause of death in
     the United States [2] and lung cancer is the only passive smoking
     illness that is officially recognized by a government agency.

          The approximately 50.000 number was mentioned in testimony by
     the American Medical Association[3] which states that passive
     smoking may kill as many as 53,000 Americans annually. This
     statement in turn appears to be ultimately traceable to an article
     published in 1988 in Environmental International. [4] This article
     used existing epidemiological studies (statistical studies of
     incidence of disease among human populations) to estimate these
     deaths which included, under one set of calculations, 37,000 lung
     cancer deaths, 12,800 deaths from other cancers, and 37,400 deaths
     from heart disease. (Wells actually reported estimate ranging from
     38,000 to 53,000, with a preferred estimate of 46,000)

          Each issue of Environment International contains an editorial;
     the one in the issue containing the Wells article was directed at
     that article. The editorial indicated that the study received mixed
     reviews from reference (two recommended publication after revision
     and the third recommended against publication on the grounds that
     it was to speculative), but the editors chose to publish the paper
     despite mixed reviews.

          In the following three years there were a series of critiques
     and rejoinders related to this paper. [5] The main criticisms
     related to two points; that the evidence, particularly with respect
     to heart disease, was at odds with information on physical levels
     of exposure, and that there are of a variety of serious problems
     with epidemiological studies. The original article and
     correspondence are enclosed. [6] The following discussion elaborates
     on those issues.

          Generally, there are two ways that one might try to estimate
     the number of premature deaths, if any, from passive smoking. One
     could rely on estimates of physical absorption of the components of
     smoke and then use those estimates to extrapolate based on studies
     of the effects of active smoking. These are called dosimetric
     approaches. Secondly, one might try to directly estimate the
     effects of passive smoking by comparing disease rates of
     individuals who are and are not exposed to passive smoke. These
     studies are the epidemiological one. [7] Typically, they compare the
     rates of disease in nonsmoking women married to husbands who smoke
     and husbands who do not smoke. Obviously, one might be interested
     in the extent to which the estimates derived from these two
     approaches are similar, or are different.

          Consider the lung cancer studies first. Even though these
     premature deaths are only a small part of the total estimate in the
     Wells article, discussing these estimates will clarify some of the
     problems with the Wells estimate.

          The recent EPA study estimated a risk for lung cancer of about
     30 percent from passive smoking based on epidemiological studies.
     This is, according to their analysis of the statistical studies,
     nonsmoking wives of men who smoker have 30 percent more lung cancer
     than nonsmoking wives of men who do not smoke. This risk is, in
     turn, only a tiny fraction of the risk from active smoking
     (probably around 3 percent),[8] and the proportion of
     passive-smoking premature deaths to active smoking premature deaths
     estimated by the EPA is between two and three percent. [9]

          A physical extrapolation approach would tend to yield smaller
     effects. According to the EPA report, measures of cotinine in the
     urine indicate that, overall, passive smokers have about 1/2 of one
     percent of the level of active smokers. [10] Since the number of
     current and former smokers are the same as the number of never-
     smokers, the estimated premature deaths annually from passive
     smoking for never smokers would be about 600 using a linear
     extrapolation. [11] This number is considerably less than the EPA's
     estimate of 2000 (the remaining 1000 estimate was for effects of
     environmental tobacco smoke on former smokers who are about half
     the number of never-smokers). There us also a section in the EPA
     study that discusses extrapolations based on the physical exposure
     to passive smoking; these estimates also tend to be smaller that
     the epidemiological estimates and some are very low.

          These are potential problems with both methods. The physical
     extrapolation method used above assumes a linear relationship
     between the incidence of a disease and exposure. Based on evidence
     from the pattern for active smoking, however, a linear method may
     not be correct. The evidence, however, tends to suggest that such
     an adjustment would reduce the estimates based on physical
     exposure. There is some evidence that disease rises with square of
     the exposure or even with higher powers in the case of lung
     cancer. [12] If the disease were to rise with the square of
     exposure, then the estimate based on cotinine levels would be only
     3 people rather than 600 people. Of course, it is possible that the
     disease rises less than proportionally with exposure. [13] At the
     same time, it is possible that there is a threshold which is so
     small that individuals are not exposed to experience health damage.
     It is also possible that continue is not the best measure of
     exposure; some exposure measures show larger and some show smaller

          Problems also occur with epidemiological studies. It is always
     possible that relationships found with human population studies are
     due to chance, even in a perfectly designed study, and while there
     are statistical methods that assign probabilities of error, they
     are still probabilities. In addition, the precision of the specific
     estimates is always in question. [14]

          There are also some specific problems that have been
     identified with passive smoking studies. First, the measures of
     exposure are based on interviews with subjects or their relatives
     and are subject to considerable error. Secondly, the studies might
     be picking up the effects of active smoking, through
     misclassification of never-smokers as smokers. That is some
     individuals who identify themselves as those who never smoked are
     actually current smokers or former smokers and they may be more
     likely to be married to smokers. Indeed, corrections were made for
     this effect in the EPA report, but it is difficult to know whether
     they are accurate. Finally, these studies do not or cannot fully
     control for "confounders" -- other factors that might be
     responsible for the effect that are simply unrelated with marriage
     to a smoker. For example, smokers tend to be less concerned in
     general about health risks and engage in other behaviors (e.g.
     diet, lack of preventive health care) that might be shared with
     their spouses and that may be the cause of the health effect.

          The EPA chose the epidemiological studies as a basis of their
     approach, but they nevertheless relied on the cotinine measures for
     several aspects of their estimates (such as extrapolating from the
     effects on spouses of smokers to the population in general.

          While the estimates from at least some of the epidemiology
     studies are significantly larger than the estimates of physical
     exposure, these results are not magnitudes apart. The same cannot
     be said, however, for the Wells estimates of deaths from heart
     disease. Using the same type of linear physical extrapolation would
     result in 700 deaths from coronary disease for never smokers, and
     perhaps another 350 for former smoker, with a total of about a
     1000. The portion of the Wells' 53,000 estimated from the
     epidemiological studies, even for several years ago, is 37,000, a
     number that is enormously larger.

          This large estimate occurs because the epidemiological studies
     show a very high risk estimate for passive smoking relative to
     active smoking for heart disease and not for lung cancer. For
     example, Wells indicates a 30 percent additional risk for heart
     disease for males and a 20 percent rise for females, as compared to
     a 70 percent risk for smokers. These relative risks are enormous
     compared to both the dosimetric ratios and to the epidemiological
     results for lung cancer. Note that although the risk ratios are not
     that different from lung cancer, the absolute risk estimates are
     much larger. The risk of lung cancer for nonsmokers is very low,
     and any percentage of a small number is still a small number. The
     risk for heart disease is much larger initially, and therefore any
     significant percentage change in the risk is larger, Put another
     way, even the epidemiological studies of lung cancer produced
     passive-smoking related deaths of less than 3 percent of active
     smoking related deaths, while the heart disease studies produced
     estimates that were 26 percent of active smoking related deaths.

          The biological plausibility of passive smoking effects on
     cardiovascular disease has been the subject of some discussion.
     Both Wells and Sleenland refer to a 1991 review article by Glantz
     and Parmley that suggested that passive smoking may, in
     experimental studies, promote the formation of plaques in blood
     vessels, increase the tendency of blood platelet cells to aggregate
     and form clots, and reduce the oxygen-carrying capacity of the
     blood. [15] There are limited data to support or refute these
     hypotheses. For example, while some studies of nonsmokers found
     that passive smoking appears to promote platelet aggregation,
     parallel studies of active smokers have not consistently shown any
     effect on platelet function. [16]

          Among the chemical components in passive smoke, carbon
     monoxides and nicotine are the most likely to adversely effect
     cardiovascular performance. Carbon monoxide binds tightly to
     hemoglobin and diminishes oxygen transport in the blood stream.
     Nicotine acts in the brain and throughout the body, promoting the
     release of adrenalin and increasing heart rate and blood pressure.
     Although these effects might impair performance, exposures from
     passive smoke are generally thought to be at concentrations below
     those at which physiological changes would occur in healthy
     persons. [17]

          The most likely explanation of these large risks from passive
     smoking epidemiological studies for heart disease is the absence of
     control for other factors. [18] There are many important concerns
     about health risks in general. In general, studies do not, and
     perhaps cannot, control for many of those factors. If smokers;
     wives share in these behaviors, the relationships found in the
     epidemiological studies are spurious.

          The Wells estimate of passive smoking deaths from cancers
     other than lung cancer is even larger relative to active smoking
     deaths than is the case of heart disease--about 50 percent. Again,
     these cancers are influenced by many other factors and the same
     general criticisms can be made about these epidemiological
     estimates as in the case of heart disease.

          In sum, this analysis suggests that the Wells estimates are so
     high relative to measures of physical exposure that they seem
     implausible. It also suggests that the absence of controls or the
     inability to control for other factors may be a major problem in
     relying on epidemiological estimates of the health effects of
     passive smoking.


     [1] Environmental Protection Agency. Respiratory Health Effects
     of Passive Smoking: Lung Cancer and Other Disorders.  December

     [2] J.Michael McGinnis and William F. Fooge.  Actual Causes of
     Death in the United States. Journal of the American Medical
     Association. November 10, 1993, pp. 2207-2212.

     [3] Statement of the American Medical Association, Health and
     the Environmental Subcommittee, House Committee on Energy and
     Commerce, Rn Adverse Health Effects of Exposure to Environmental
     Tobacco Smoke, July 21, 1993.

     [4] A. Judson Wells, An Estimate of Adult Mortality in the
     United States from Passive Smoking. Environmental International.
     Vol. 14, No.3 1988, 249-265.

     [5] Letters from Alan W. Katzanstein, Peter M. Lee and Larry
     Holcoim criticizing the Wells results; a clarifying letter from
     Takoahi Hirayima, a rebuttal to Katzanstein and Lee from James L.
     Repace and Alfred H. Lowrey, and a response from Wells were
     published in 1990 (Vol. 16, No. 2, pp 175-190.  In 1991, a letter
     from Stanton A. Glantz criticizing Lee was published along with
     Lee's reply (Vol. 17, no. 1, pp. 88-91).  Later in 1991, a
     response of Lee to the 1990 letters of Repace and Lowery, and
     Wells, a letter from Muin J. Khoury clarifying a point raised in
     Lee's letter, a joint letter from Glantz and Lee clarifying an
     issue raised earlier in the year, and a response from Repace and
     Lowrey, and Wells to Lee's letter were published (Vol. 17, no. 4,
     370-387).  In 1992 (Vol. 18, No. 3, pp. 315-317, 321-325) another
     letter from Lee and response from Wells was published.

     [6] A more recent paper of the same general type as the Wells
     paper has been published on heart disease.  See Kyle Sleenland,
     Passive Smoking, and the Risk of Heart Disease, Journal of the
     America Association, Vol. 267, No. 1, January 1, 1992.

     [7] Of course, the original estimates of the effects of active
     smoking on disease are based on epidemiological studies in large
     part as well, but there are some problems that occur in passive
     smoking epidemiological studies that are not as serious in active
     smoking studies.

     [8] The risk of lung cancer in smokers and ex-smokers depends
     on intensity, duration, and in the case of ex-smokers, time
     elapsed since quitting.  Passive smoking would involve three
     percent of the risk of active smoking if there is a ten fold
     active smoking risk (i.e. smokers have an additional estimated
     risk of lung cancer that is ten times the disease rate of non-
     smokers) which is typical of current estimates of the risk for
     women as reported in the 1989 Surgeon General's Report (Reducing
     the Health Consequences of Smoking.  U.S. Department of Health
     and Human Services, DDHS Publication No. (CDC) 89-8411).  In
     general the estimates of deaths from passive smoking, the EPA
     actually used the additional risk (of wives married to smokers as
     compared to wives married to non-smokers) in the U.S. passive
     smoking studies, which was about 20 percent.  If studies from all
     countries are considered, the estimated risk from these studies
     was 30 percent.

     [9] The estimated attributable deaths from lung cancer due to
     active smoking are 118,000.  See C. Stephen Redhead, Mortality
     and Economic Costs Attributable to Smoking and Alcohol Abuse,
     Congressional Research Service Report 93-SPR, April 20, 1993.

     [10] EPA Report, pp. 3-43.

     [11] To extrapolate, multiply the percentage of cotinine by
     the ratio of ever smokers to never smokers, and by the number of
     deaths attributable to active smoking.

     [12] Surgeon General's Report, p. 44.

     [13] The argument has been made for a relationship in which
     passive smoking can have large effects relative to active smoking
     for some specific events in laboratory settings, which is largely
     attributed to increased sensitivity of some nonsmokers.  See
     Stanton Glantz and William Parmley, 1991, pp. 1-12.  Circulation
     is a publication of the American Heart Association.

     [14] For a discussion of the methods used by the EPA in
     combining existing studies and the findings of studies published
     later, see Appendix A in Cigarette Taxes to Fund Health Care
     Reform, by Jane G. Gravelle and Dennis Zimmerman, Congressional
     Research Service Report 94-214. March 8, 1994.

     [15] Glantz and Parmley (1991).

     [16] Samet, J.M., Environmental Tobacco Smoke, In
     Environmental Toxicants, ed, M. Lippman, New York Van Rostrand
     Reinhold, 1992.

     [17] U.S. Department of Health and Human Services, The Health
     Consequences of Involuntary Smoking, 1986, Surgeon General's
     Report, DHHS Publication Number (CDC) 87-8938.

     [18] This position is taken by Gary L. Hubert, Robert E.
     Brockie, and Vijay K. Mahajan in a paper written for the layman;
     Passive Smoking and Your Heart, Consumers Research, vol. 75.,
     April 1992, pp. 13-19, 32.  These authors consider the results in
     the Wells and Sleenland studies biologically implausible, and
     also note that six of the nine epidemiological studies show
     causes of heart disease (e.g. that, lack of exercise, lack of
     preventive health cure) that may be engaged in by smokers.  That
     is, there is much evidence that smokers tend to be less
     relative risks that are in excess of risks estimated for active
     smokers and that most have very few controls for other factors
     that might effect heart disease.