Correspondence
Passive Smoking and Coronary Heart Disease
N Engl J Med 1999; 341:697-700August 26, 1999
- Article
To the Editor:
In their meta-analysis of passive smoking, He et al. (March 25 issue)1 analyzed 10 cohort and 8 case–control studies and concluded that nonsmokers exposed to environmental tobacco smoke had an overall relative risk of coronary heart disease of 1.25. I want to point out several problems with their analysis. For the cohort studies they analyzed, each adjusted relative risk shown in Figure 1 of their article is higher than the corresponding crude relative risk that can be calculated from the data given in the figure. For example, in the study by Garland et al.2 the crude relative risk can be calculated as 3.5, and the relative risk reported by He et al. is 14.9, whereas the adjusted relative risk was reported by Glantz and Parmley as 2.7.3 However, the most dramatic difference occurs in the study by Steenland et al.4 for which the crude relative risk is 0.54 and the relative risk reported by He et al. is 1.2.
It is often instructive to compare the crude relative risk with the adjusted relative risk to ascertain the influence of the adjustment. With use of the meta-analytic methods of He et al., the overall crude relative risk is 0.84 for the 10 cohort studies. Thus, the conclusion as to whether exposure to passive smoke is harmful or helpful appears to depend on an adjustment process that is often imprecise and ambiguous.
Interpretation of the case–control studies may be even more difficult. Because in a case–control study the relative risk cannot be calculated directly, the odds ratio is used as a surrogate when the disease is rare. However, if the disease is not rare in the particular group being studied (even if it is rare in the general population), then the odds ratio overestimates the actual relative risk.5 This can yield an exaggerated effect. Furthermore, in a case–control study, what is actually estimated is the relative probability of exposure, given that a person has heart disease. Since heart disease has multiple causes, it is not logical to argue a relative probability (relative risk) of heart disease given that a person is exposed to a particular risk factor. Therefore, the case–control studies should be excluded from the meta-analysis, or at least the cohort and case–control studies should be analyzed separately.
Of course, these considerations would not be relevant if the reported effect of passive smoking were large. It is because the effect is so small that these issues must be taken into account in the final interpretation.
5 ReferencesGeorge D. Swanson, Ph.D.
California State University, Chico, CA 95929-03301
He J ,Vupputuri S ,Allen K ,Prerost MR ,Hughes J ,Whelton PK . Passive smoking and the risk of coronary heart disease -- a meta-analysis of epidemiologic studies. N Engl J Med 1999;340:920-926
Full Text | Web of Science | Medline2
Garland C ,Barrett-Connor E ,Suarez L ,Criqui MH ,Wingard DL . Effects of passive smoking on ischemic heart disease mortality of nonsmokers: a prospective study. Am J Epidemiol 1985;121:645-650
CrossRef | Web of Science | Medline3
Glantz SA ,Parmley WW . Passive smoking and heart disease: epidemiology, physiology, and biochemistry. Circulation 1991;83:1-12
Web of Science | Medline4
Steenland K ,Thun M ,Lally C ,Heath C Jr . Environmental tobacco smoke and coronary heart disease in the American Cancer Society CPS-II cohort. Circulation 1996;94:622-628
Web of Science | Medline5
Zhang J ,Yu KF . What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998;280:1690-1691
CrossRef | Web of Science | Medline
To the Editor:
He et al., in their Methods section, erroneously assume that the data presented by Steenland et al.1 concern American Cancer Society Cancer Prevention studies I and II, when in fact they concern only study II. Because of this error, results from study I were not included in their meta-analysis, a serious omission in view of the large number of cases of heart disease and the lack of relation with passive smoking seen in that study.
1 ReferencesKenneth W.E. Denson, D.Phil., F.R.C.Path.
Thame Thrombosis and Haemostasis Research Foundation, Thame OX9 3NY, United Kingdom1
Steenland K ,Thun M ,Lally C ,Heath C Jr . Environmental tobacco smoke and coronary heart disease in the American Cancer Society CPS-II cohort. Circulation 1996;94:622-628
Web of Science | Medline
To the Editor:
Bailar's editorial1 on the article by He et al.2 deserves rebuttal. Most of Bailar's concerns about meta-analyses of passive smoking and coronary heart disease have been addressed elsewhere in the literature. Specifically, results in male and female patients are sufficiently homogeneous to allow their combination.3,4 The various studies can be, and were, given meaningful quality scores.3,4 The questions of misclassification of smoking status and exposure status have been adequately dealt with, as has publication bias.3 Adjustment has been made for potential confounders,2-4 and positive trends with dose were found for 16 of 22 studies without the inclusion of the nonexposed group.4
Bailar is also worried that the pooled relative risk of coronary heart disease associated with passive smoking is large as compared with the risk associated with active smoking. The ratio of excess risks, active to passive, is about (1.93–1)÷(1.24–1), or 3.9, when the risk for active smokers is measured against that for nonexposed persons who have never smoked. This is not unusual for an anatomical site that is not in direct contact with tobacco smoke and for which the dose–response curve is convex upward.5 For comparison, the best study we have on breast cancer and active and passive smoking 6 found an active-to-passive ratio of excess risks of only (3.0–1)÷ (2.3–1), or 1.5. Although the data are fewer, similar low ratios appear to hold for other noncontact cancer sites, such as the cervix, liver, and brain, and for lymphoma and leukemia. Probably these ratios are low because so many of the entities in environmental tobacco smoke that cause coronary heart disease and cancer are in the vapor phase7; they therefore are deposited more completely in the lung and are harder to clear than particle deposits. The toxins must be cleared into the body fluids, where they can circulate to distant sites, such as the heart or breast.
Bailar appears to prefer a “thoughtful review of the usual type.”1 If so, he should read the review by Kritz et al.,8 which covers the same ground as the study by He et al.2 and others,3,4 but without the meta-analysis. Of course, Kritz et al. also conclude that environmental tobacco smoke causes coronary heart disease.
8 ReferencesA. Judson Wells, Ph.D.
5 Ingleton Cir., Kennett Square, PA 193481
Bailar JC III . Passive smoking, coronary heart disease, and meta-analysis. N Engl J Med 1999;340:958-959
Full Text | Web of Science | Medline2
He J ,Vupputuri S ,Allen K ,Prerost MR ,Hughes J ,Whelton PK . Passive smoking and the risk of coronary heart disease -- a meta-analysis of epidemiologic studies. N Engl J Med 1999;340:920-926
Full Text | Web of Science | Medline3
Wells AJ . Passive smoking as a cause of heart disease. J Am Coll Cardiol 1994;24:546-554
CrossRef | Web of Science | Medline4
Wells AJ . Heart disease from passive smoking in the workplace. J Am Coll Cardiol 1998;31:1-9
CrossRef | Web of Science | Medline5
Law MR ,Morris JK ,Wald NJ . Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence. BMJ 1997;315:973-980
CrossRef | Web of Science | Medline6
Morabia A ,Bernstein M ,Heritier S ,Khatchatrian N . Relation of breast cancer with passive and active exposure to tobacco smoke. Am J Epidemiol 1996;143:918-928
Web of Science | Medline7
Wells AJ . An estimate of adult mortality in the United States from passive smoking: a response to criticism. Environ Int 1991;17:382-385
CrossRef | Web of Science8
Kritz H ,Schmid P ,Sinzinger H . Passive smoking and cardiovascular risk. Arch Intern Med 1995;155:1942-1948
CrossRef | Web of Science | Medline
To the Editor:
Many readers would dispute Bailar's conclusion that “we still do not know . . . whether exposure to environmental tobacco smoke increases the risk of coronary artery disease.” The evidence available to mid-1997 on this topic was reviewed by Australia's National Health and Medical Research Council (NHMRC).1 The NHMRC considered 22 analyses from 16 studies; 17 of the 22 analyses indicated some increase in the risk of coronary events among nonsmokers with exposure to environmental tobacco smoke, and in 8 of them the results were statistically significant. Rather than undertake a quantitative meta-analysis, the NHMRC summarized the data in terms of a median relative risk and corresponding interquartile range. The median estimate of 1.24 (interquartile range, 1.02 to 1.62) is entirely consistent with the pooled estimate of 1.25 (95 percent confidence interval, 1.17 to 1.32) derived by He et al. and was supported by findings of excess risks of mortality from all causes in seven of eight prospective studies of passive smoking.
The report from the NHMRC also examined the relation between passive smoking and coronary heart disease in light of the criteria proposed by Hill2 and concluded that “all the evidence put together is reasonably coherent.”1 Like Bailar, the NHMRC drew attention to the relatively large excess risk of coronary heart disease associated with passive smoking as compared with the risk attendant on active smoking, but it cited evidence that platelet function in nonsmokers is particularly sensitive to exposure to environmental tobacco smoke.3
Whatever the limitations of meta-analysis, the abundant evidence that passive smoking causes harm to health can no longer be ignored.
3 ReferencesKonrad Jamrozik, M.B., B.S., D.Phil.
Graham A. Colditz, M.D., Dr.P.H.
Harvard Medical School, Boston, MA 021151
NHMRC Working Party. The health effects of passive smoking: a scientific information paper. Canberra, Australia: NHMRC, 1997.
2
Hill AB . The environment and disease: association or causation? Proc R Soc Med 1965;58:295-300
Medline3
Davis JW ,Shelton L ,Watanabe IS ,Arnold J . Passive smoking affects endothelium and platelets. Arch Intern Med 1989;149:386-389
CrossRef | Web of Science | Medline
To the Editor:
In his editorial, Bailar uses several flawed arguments. His concern about reporting bias might be appropriately applied to the case–control studies, but not to the cohort studies (10 of the 18 studies analyzed). Moreover, when the two types of studies were analyzed separately by He et al., the conclusions were strikingly similar. The possibility of publication bias was also raised. This is often a valid criticism of meta-analyses, but it appears misdirected in this instance. Specifically, of the 18 studies reviewed, 7 found a significant association and 11 did not. If anything, the bias here is likely to be very small. In addition, random reporting errors, also mentioned by Bailar, are likely to attenuate and not spuriously strengthen these associations.
Bailar considers the range of relative risk of about 1.0 to 2.2 in He et al.'s study to be “very small.” He believes that uniformity in results is “not necessarily good.” Such comments are only his opinion, and not facts. Finally, Bailar criticizes the use of a “multiplicative model,” but meta-analysis must use the models of the original studies and may not choose others. Bailar does not show in any way that the particular model in this case produces a bias.
Valentin Fuster, M.D., Ph.D.
Philip Greenland, M.D.
Kiang Liu, Ph.D.
American Heart Association, New York, NY 10029To the Editor:
Bailar questions whether a 25 percent increase in the incidence of coronary heart disease associated with passive smoking, as reported by He et al., is compatible with the generally reported increase of about 75 percent among active smokers. Bailar finds it more plausible that the added risk of lung cancer from passive smoking may be about 2 percent of the risk associated with active smoking. That view ignores the fact that the mechanisms causing lung cancer and coronary heart disease are different. Tobacco smoke induces lung cancer through chemical carcinogenesis, whereas none of the mechanisms of coronary heart disease resemble chemical carcinogenesis. These different underlying mechanisms may explain how environmental tobacco smoke causes a larger increase in the incidence of coronary heart disease than in the incidence of lung cancer.
Michael Rabinoff, D.O., Ph.D.
UCLA Neuropsychiatric Institute and Hospital, Los Angeles, CA 90024-1759Author/Editor Response
The authors reply:
To the Editor: Swanson's estimation of crude relative risks for the 10 prospective studies presented in our meta-analysis is erroneous because he used the number of persons rather than person-years of exposure as the denominator for his calculations. Contrary to his claim, there is not a large discrepancy between crude and adjusted relative risk estimates.
Denson expresses concern because we excluded the American Cancer Society Cancer Prevention Study I from our meta-analysis. We did this because LeVois and Layard1 failed to provide details of their analytic approach. Moreover, their findings in the data from the Cancer Prevention Study II conflict with Steenland et al.'s more rigorous assessment.2
In his editorial, Bailar criticizes both the methods and conclusions of our meta-analysis. We are surprised that he believes meta-analysis does not yield a more reliable conclusion than “a thoughtful review of the usual type,” because it is generally considered to be the most objective means to gain an overview of evidence from different studies. This is especially true in the identification of a small increase in relative risk that may have important public health implications. Bailar suggests that the relation between passive smoking and coronary heart disease may reflect the desire of sick patients to identify a cause for their illness (recall bias), but this would not explain the findings of the 10 prospective studies in which exposure was assessed before the onset of disease. Bailar worries about the failure of journals to publish negative findings, but our analysis did not suggest this bias, and the inclusion of three unpublished studies did not change the estimate of relative risk.
Bailar asserts that the consistent pattern of association we observed in studies with different designs, methods, and populations may “suggest consistency in bias rather than consistency in real effects.” Most epidemiologists believe that consistency suggests the presence of a causal association.
The relative risk (or odds ratio) is a widely used measure of association. Elsewhere, Bailar states that “findings must be expressed on a common scale (often as odds ratios)” in a meta-analysis.3 Bailar's suggestion that dose–response relations be studied without the inclusion of an unexposed reference group is interesting but unconventional.
There are many potential mechanisms by which passive smoking may result in coronary heart disease, and they are different from those by which cigarette smoking may cause lung cancer. Why should we expect the same pattern of relative risk in association with active and passive smoking for lung cancer as we do for coronary heart disease?
3 ReferencesJiang He, M.D., Ph.D.
Janet Hughes, Ph.D.
Paul K. Whelton, M.D.
Tulane University School of Public Health and Tropical Medicine, New Orleans, LA 701121
LeVois ME ,Layard MW . Publication bias in the environmental tobacco smoke/coronary heart disease epidemiologic literature. Regul Toxicol Pharmacol 1995;21:184-191
CrossRef | Web of Science | Medline2
Steenland K ,Thun M ,Lally C ,Heath C Jr . Environmental tobacco smoke and coronary heart disease in the American Cancer Society CPS-II cohort. Circulation 1996;94:622-628
Web of Science | Medline3
Bailar JC III . The promise and problems of meta-analysis. N Engl J Med 1997;337:559-561
Full Text | Web of Science | Medline
Author/Editor Response
I am grateful to these correspondents for raising several issues that need discussion. However, their conclusions are mistaken. In my editorial, I did not deny that there is a relation between passive smoking and coronary heart disease, but I noted that the evidence presented to support a relation is not convincing. It is likely that such a relation exists, but more work will be needed to confirm it, and still more to estimate its strength with much precision.
We must examine evidence that seems to support a favored hypothesis with even more care than we would examine evidence against it. Single-minded dedication to a specific proposition may be useful once other work has clearly shown a need for action, but not before. Furthermore, a well-informed scientist can come up with a plausible explanation for almost any set of research findings. To describe a mechanism by which environmental tobacco smoke might increase coronary artery disease is not to show that it operates in the real world.
Exposure–response relations for toxic agents (excluding many carcinogens) are generally concave upward — that is, the effects of successively smaller exposures decrease more rapidly than the dose itself, and often something close to a threshold may be found at low doses. The levels of exposure to specific constituents of environmental tobacco smoke are not fully understood, but I do not know of any for which exposures among nonsmokers are as high as one third of those among smokers. This is further reason for caution in concluding that an increase in risk induced by environmental tobacco smoke among nonsmokers is one third or more of the excess risk among smokers.
Other evils of environmental tobacco smoke are well known, and even without coronary artery disease there is strong reason to protect the nonsmoking public. I understand the urge to “pile it on,” perhaps in the hope of generating stronger action sooner, and there may be reasons related to public health and public policy for taking action before the evidence is complete. Those reasons are not advanced by overstatement.
John C. Bailar, III, M.D., Ph.D.
University of Chicago, Chicago, IL 60637
