Skip to main content
NCBI home page
The BMJ logoLink to The BMJ
. 1998 Aug 1;317(7154):344. doi: 10.1136/bmj.317.7154.344

Passive smoking and heart disease

Authors need to analyse the same data

Maurice E LeVois 1,2,1-150, Maxwell W Layard 1,2,1-150
PMCID: PMC1113639  PMID: 9685290

Editor—In their meta-analysis Law et al1-1 reject results we published1-2,1-3 on environmental tobacco smoke and coronary heart disease, using data from the two large cancer prevention studies by the American Cancer Society and the national mortality followback survey. They reject our results because they disagree with our interpretation of data from other studies and because our analysis was funded by the tobacco industry (table 1).

Table 1.

Data rejected by Law et al for their meta-analysis of spousal smoking and mortality from coronary heart disease in cancer prevention study I.1-2 Values are relative risks (95% confidence intervals)

Spousal smoking Men* Women†
Ex 0.95 (0.83 to 1.09) 0.99 (0.93 to 1.05)
Current:
 1-19 0.99 (0.89 to 1.09) 1.04 (0.97 to 1.12)
 20-39 0.98 (0.85 to 1.13) 1.06 (0.98 to 1.15)
 ⩾40 0.72 (0.41 to 1.28) 0.95 (0.78 to 1.15)
Ex or current 0.97 (0.90 to 1.05) 1.03 (0.98 to 1.08)
Pipe/cigar 1.06 (0.99 to 1.14)
Open in a new tab
1-150

7758 deaths among never smokers.  

1-151

7133 deaths among never smokers. 

By excluding our results Law et al discard 16 280 relevant deaths from coronary heart disease with spousal smoking data and retain 6600 cases. They give no hint that their meta-analysis includes under a third of the available published data. The reasons for rejecting so many data should be considered. If Law et al believe we have misrepresented the data, they should analyse the same data from the American Cancer Society and the national mortality followback survey, and report their results.

Law et al’s argument that our data from the second cancer prevention study disagree appreciably with data reported by Steenland et al1-4 is wrong.1-5 They incorrectly compare our results for ever-smoking spouse exposure with Steenland et al’s results for current-smoking spouse exposure. Both studies present comparable results for subjects in the second cancer prevention study who are married to a current smoker. We calculate the relative risk for men to be 1.30 (95% confidence interval 1.11 to 1.51), for women 1.08 (0.89 to 1.32), and for both sexes combined 1.21 (1.07 to 1.37). These results are similar to those reported by Steenland et al (men 1.22 (1.07 to 1.40), women 1.10 (0.96 to 1.27)),1-4 which we have combined to give a relative risk of 1.16 (1.05 to 1.27) for both sexes.

Both sets of analyses report a barely significant association between environmental tobacco smoke and coronary heart disease in men, with a negative dose response. There is no significant association between spousal smoking and death from coronary heart disease in women, nor any sign of a dose response. Nearly twice as many women as men died of coronary heart disease in the second cancer prevention study, which makes the data for women particularly relevant to any meta-analysis.

Law et al’s selective rejection of two thirds of the relevant data on environmental tobacco smoke and coronary heart disease raises additional questions about their interpretation of other data. We have noted significant publication bias in the pooled results on environmental tobacco smoke and coronary heart disease, for example.1-2 In reaching the opposite conclusion, Law et al ignore the significant association between study size and relative risk in the previously published spousal smoking studies as well as the significant difference between published and unpublished results on environmental tobacco smoke and coronary heart disease.

Table 2.

Data rejected by Law et al for their meta-analysis of spousal smoking and coronary heart disease: case-control analysis of data from national mortality followback survey1-3

Spousal smoking Men
Women
Cases Controls Relative risk (95% CI) Cases Controls Relative risk (95% CI)
No 378 783 1.0 459 969 1.0
Yes  97 215 0.97 (0.73 to 1.28) 455 961 0.99 (0.84 to 1.16)
Open in a new tab

See Editorial by Smith, Education and debate p 333

References

  • 1-1.Law MR, Morris JK, Wald NJ. Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence. BMJ. 1997;313:973–980. doi: 10.1136/bmj.315.7114.973. . (18 October.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 1-2.LeVois ME, Layard MW. Publication bias in the environmental tobacco smoke/coronary heart disease epidemiologic literature. Regul Toxicol Pharmacol. 1995;21:184–191. doi: 10.1006/rtph.1995.1023. [DOI] [PubMed] [Google Scholar]
  • 1-3.Layard MW. Ischemic heart disease and spousal smoking in the national mortality followback survey. Regul Toxicol Pharmacol. 1995;21:180–183. doi: 10.1006/rtph.1995.1022. [DOI] [PubMed] [Google Scholar]
  • 1-4.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. doi: 10.1161/01.cir.94.4.622. [DOI] [PubMed] [Google Scholar]
  • 1-5.LeVois ME, Layard MW. Letter to the editor re: Steenland et al. Circulation. 1997;96:2086–2087. [Google Scholar]
BMJ. 1998 Aug 1;317(7154):344.

Evidence on passive smoking and heart disease needs re-evaluation

Peter Lee 1,2-150

Editor—Law et al estimate that there is a 30% excess risk associated with spousal smoking on the basis of 6600 deaths from heart disease in 19 epidemiological studies.2-1 They distort the estimate by omitting results that are based on 15 000 deaths from heart disease from the first cancer prevention study by the American Cancer Society. That study reported on spousal smoking and lung cancer in 1981.2-2 Ten years ago the society told me that the study showed no effect on heart disease, but it never published its findings. Law et al reject published analyses of this study by LeVois and Layard2-3 because they are tobacco consultants and their analyses of the second cancer prevention study cited different relative risks from those reported by the American Cancer Society.2-4 Both analyses actually reported similar findings—a small decrease in risk for spouses who were former smokers and a small, non-dose related, increase for spouses who were current smokers. The “difference” arose because LeVois and Layard followed precedent in concentrating on the index “spouse ever smoked” whereas the society emphasised results for “spouse current smoker.” For the first cancer prevention study the risk was not increased for either index. If Law et al wish to reject analyses by tobacco consultants, they should have analysed the study themselves in order to avoid publication bias.

Law et al dismiss smoking misclassification bias as unimportant, ignoring recent evidence that smokers who deny smoking have a risk of heart disease that is 4.0 times higher (95% confidence interval 1.8 to 9.1) than that of smokers who admit to smoking.2-5 Their consideration of confounding relates only to a few of the many risk factors for heart disease.

Law et al say that the risks from passive smoking and smoking one cigarette a day are similar—an acute response to effects on platelets. This is based on a questionable backward linear extrapolation from data on risk by amount smoked (ignoring the known relative risk of 1.0 in non-smokers) and unjustified reliance on results from one single study of platelet aggregation in 10 subjects exposed to passive smoking in a hospital corridor.

This argument implies virtually no dose-response relation between risk in non-smokers and the extent of their exposure to environmental smoke, but Law et al never consider the dose-response data for passive smoking. The table shows that the results from the studies tend to fall into two groups—small studies reporting a moderate association and marked dose response and large studies reporting essentially no overall association or dose response. Neither group of results is consistent with the authors’ theories.

Table.

Risk of heart disease in relation to spousal smoking, overall and by exposure level

Study* Sex Spousal exposure index Spousal smoking relative risk† (95% confidence interval) Exposure levels‡ Relative risk by exposure level Significance of trend§
Small studies (totalling 1815 cases)
Hirayama18 F Ever 1.16 (0.94-1.43) 0 1-19 ⩾20 1.00 1.10 1.31 Yes
Svendsen21 M Current 1.61 (0.96-2.71) 0 1-19 ⩾20 1.00 1.20 1.75 No
He22 F Ever 1.50 (0.63-3.60) 0 1-20 ⩾21 1.00 2.30 6.86 Yes
Hole23 F Ever 1.65 (0.79-3.46) None Low High 1.00 2.09 4.12 Yes
LaVecchia27 M+F Current 1.21 (0.57-2.53) 0 1-14 ⩾15 1.00 1.13 1.30 No
Tunstall-Pedoe29 M+F Current 1.37 (1.07-1.75) None Little Some Lot 1.0 1.2 1.5 1.6 Yes
Kawachi32 F Current 1.71 (1.03-2.84) None Occasional Regular 1.00 1.58 1.91 Yes
Ciruzzi33 M+F Ever 1.43 (0.90-2.00) 0 1-20 ⩾21 1.00 1.27 1.41 Yes
Jackson2-2 M Current 1.06 (0.39-2.91) None Low High 1.00 1.30 0.90 No
F Current 3.75 (1.15-12.19) None Low High 1.00 2.10 7.50 Yes
Large studies (totalling 20 099 cases)
Cancer prevention study I36 M Current 0.98 (0.91-1.06) 0 1-19 20-39 ⩾40 1.00 0.99 0.89 0.72 No
F Current 1.04 (0.99-1.09) 0 1-19 20-39 ⩾40 1.00 1.04 1.06 0.95 No
Cancer prevention study II31 M Current 1.22 (1.07-1.40) 0 1-19 20 ⩾21 1.00 1.33 1.17 1.09 No
F Current 1.10 (0.96-1.27) 0 1-19 20 21-39 ⩾40 1.00 1.15 1.07 0.99 1.04 No
Layard35 M Ever 0.97 (0.73-1.28) 0 1-14 15-34 ⩾35 1.00 0.76 1.07 0.92 No
F Ever 0.99 (0.84-1.16) 0 1-14 15-34 ⩾35 1.00 0.85 1.15 1.06 No
Open in a new tab
2-150

Superscript numbers are references as given in Law et al2-1; only studies providing dose-response data are included.  

2-151

Adjusted for age and confounding variables as far as presented by authors’ estimates from data by level of exposure where appropriate.  

2-152

Cigarettes/day except for Hole, Tunstall-Pedoe, Kawachi, and Jackson studies.  

2-153

Significance of positive trend at P<0.05. 

The conclusion that passive smoking is an important cause of heart disease is premature. The data require further evaluation.

References

  • 2-1.Law MR, Morris JK, Wald NJ. Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence. BMJ. 1997;315:973–980. doi: 10.1136/bmj.315.7114.973. . (18 October.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2-2.Garfinkel L. Time trends in lung cancer mortality among nonsmokers and a note on passive smoking. J Natl Cancer Inst. 1981;66:1061–1066. doi: 10.1093/jnci/66.6.1061. [DOI] [PubMed] [Google Scholar]
  • 2-3.LeVois ME, Layard MW. Publication bias in the environmental tobacco smoke/coronary heart disease epidemiologic literature. Regul Toxicol Pharmacol. 1995;21:184–191. doi: 10.1006/rtph.1995.1023. [DOI] [PubMed] [Google Scholar]
  • 2-4.Steenland K, Thun M, Lally C, Heath C. Environmental tobacco smoke and coronary heart disease in the American Cancer Society CPS-II cohort. Circulation. 1996;94:622–628. doi: 10.1161/01.cir.94.4.622. [DOI] [PubMed] [Google Scholar]
  • 2-5.Suadicani P, Hein HO, Gyntelberg F. Mortality and morbidity of potentially misclassified smokers. Int J Epidemiol. 1997;26:321–327. doi: 10.1093/ije/26.2.321. [DOI] [PubMed] [Google Scholar]
BMJ. 1998 Aug 1;317(7154):344.

BMJ should encourage open debate of available evidence

Ruth Dempsey 1

Editor—As a principal scientist for Philip Morris I am aware that many readers will take the views expressed in this letter somewhat cynically. I appeal to readers’ scientific objectivity and urge them to give due consideration to the issues raised below.

Law et al claim a potential relation between environmental tobacco smoke and ischaemic heart disease.3-1 They exclude the largest datasets available on the subject because they have “been published by Layard and LeVois, consultants to the tobacco industry.”3-1 These data represent about twice as many ischaemic heart disease events as the other 19 studies added together. Law et al also mislead readers into believing that LeVois and Layard’s analyses are inconsistent with the rest of the evidence. This is not the case: several of the 19 studies they selected for analysis report similar findings to those of LeVois and Layard. The results of one of the three studies published by LeVois and Layard3-2 are not inconsistent with the findings of an independent analysis of the same data by the American Cancer Society.3-3 A straightforward comparison shows mostly similarities—the only apparent differences are caused by the emphasis the different authors give to different reference groups.

This type of exclusion is symptomatic of the one-sided nature of the debate surrounding tobacco. There are other scientific flaws in Law et al’s study. Unfortunately, even though many other scientists with no particular affection for the tobacco industry may share my views, few will speak out on this issue for fear of being branded as sympathisers of the industry. This is a disservice to science and erodes one of the fundamental tools of scientific investigation—the opportunity to debate.

Would the BMJ be willing to try to rectify some of the damage done to the scientific process and encourage correspondence from independent scientists whose views may be contrary to those of Law et al? Perhaps a direct invitation from the BMJ’s editor would diminish the reluctance of such scientists to participate in the debate.

The issues raised by such papers have important implications for public health. It is therefore vital that they have a sound scientific basis. I believe that there are scientifically valid grounds for doubting the interpretation of the evidence given by Law et al. A fear of repercussions affecting anyone who voices an opinion that could be favourable to the tobacco industry seems to be preventing this process from taking place. We should all be striving to open up the debate, allowing sound scientific principles to decide on the best interpretation of the available evidence. I hope that the BMJ will play its part in facilitating this.

References

  • 3-1.Law MR, Morris JK, Wald NJ. Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence. BMJ. 1997;315:973–980. doi: 10.1136/bmj.315.7114.973. . (18 October.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3-2.LeVois ME, Layard MW. Publication bias in the environmental tobacco smoke/coronary heart disease epidemiological literature. Regul Toxicol Pharmacol. 1995;21:184–191. doi: 10.1006/rtph.1995.1023. [DOI] [PubMed] [Google Scholar]
  • 3-3.Steenland K, Thun M, Lally C, Heath Jr C. Environmental tobacco smoke and coronary heart disease in the American Cancer Society CPS-II cohort. Circulation. 1997;95:2374–2379. doi: 10.1161/01.cir.94.4.622. [DOI] [PubMed] [Google Scholar]
BMJ. 1998 Aug 1;317(7154):344.

There must be better uses for money spent on vilifying passive smoking

K W E Denson 1

Editor—Law et al used a statistical method in a paper that was in press and assessed the increased risk of ischaemic heart disease due to confounding by low dietary intake of fruit and vegetables as only 3%.4-1 Le Marchand et al calculated this as 13% (relative risk of 1.30 reduced to 1.15).4-2

It is well established that active smokers have lower intakes of fruit and vegetables and a higher intake of saturated and unsaturated fats, and that non-smokers who are exposed to environmental tobacco smoke share many of the dietary and lifestyle risk factors of smokers.4-3 Four studies have shown an inverse relation between exposure to environmental tobacco smoke and intake of β carotene or fruit and vegetables, and a study of 4018 spouse pairs showed that the β carotene concentrations were highly correlated (r=0.46, P=0.0001).

It is difficult to allow for confounding in two variables, such as diet and exposure to environmental tobacco smoke, that are correlated,4-4 but the relative risks for ischaemic heart disease in people with low serum vitamin concentrations (for example, 1.56 for low to high concentrations of β carotene, and 1.96 for low to high concentrations of β carotene and vitamin C4-5) are much higher than those reported for exposure to environmental tobacco smoke. The data from studies cited by Law et al did not allow for confounding by intake of saturated and unsaturated fats or by exercise, two elementary strong risk factors for ischaemic heart disease.4-1

We know the score on active smoking, but surely the money spent on these attempts at vilifying passive smoking could be better spent in other areas of the NHS.

I have no partisan interests. I am not funded by the tobacco companies, and this unit is funded out of my own pocket.

References

  • 4-1.Law MR, Morris JK, Wald NJ. Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence. BMJ. 1997;315:973–980. doi: 10.1136/bmj.315.7114.973. . (18 October.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4-2.Le Marchand L, Wilkens LR, Hankin JH, Haley NJ. Dietary patterns of female nonsmokers with and without exposure to environment tobacco smoke. Cancer Causes Control. 1991;2:11–16. doi: 10.1007/BF00052356. [DOI] [PubMed] [Google Scholar]
  • 4-3.Thornton A, Lee PN, Fry J. Differences between smokers, ex-smokers, passive smokers and non-smokers. J Clin Epidemiol. 1994;47:1143–1162. doi: 10.1016/0895-4356(94)90101-5. [DOI] [PubMed] [Google Scholar]
  • 4-4.Leon DA. Failed or misleading adjustment for confounding. Lancet. 1993;342:479–481. doi: 10.1016/0140-6736(93)91599-h. [DOI] [PubMed] [Google Scholar]
  • 4-5.Eichholzer M, Stahelin HB, Gey KF. Inverse correlation between essential antioxidants in plasma and subsequent risk to develop cancer, ischaemic heart disease and stroke respectively: 12 year follow up of the prospective Basel study. EXS. 1992;62:398–410. doi: 10.1007/978-3-0348-7460-1_38. [DOI] [PubMed] [Google Scholar]
BMJ. 1998 Aug 1;317(7154):344.

Authors’ reply

M R Law 1, J K Morris 1, N J Wald 1

Editor—We excluded the analyses of the tobacco industry consultants Layard and LeVois5-1,5-2 for reasons given in our paper,5-3 and for the following reasons.

(1) They included never smokers married to former smokers (about half the “relevant deaths”), substantially inflating study size but diluting risk since the risk in former smokers, let alone that in their spouses, is not materially increased.5-4

(2) The correct analysis of the second cancer prevention study (never smokers married to current smokers) by Steenland,5-4 which we included, showed an increased risk of heart disease of 20% in men and women combined (P=0.006). LeVois and Layard confirmed this (21% increase, P=0.003), but it was “washed out” in their main analysis by their inclusion of the large number of never smokers married to former smokers. In men these had an implausible significantly reduced risk, a surprising inconsistency from the Steenland analysis of the same data.5-4

(3) Only LeVois and Layard analysed the first cancer prevention study. Their result in never smokers married to current smokers (relative risk 1.03) is statistically inconsistent (P=0.01) with their own result from the second cancer prevention study (1.21), which they did not point out. One must be right and the other wrong. The appropriate analysis is not to take an “average” of the first cancer prevention study and the others but to exclude the result that is inconsistent with all the others, which is what we did.

(4) In the national mortality followback survey—smoking histories on people who had died of lung cancer and their spouses were sought from relatives. This methodology has not been validated and is subject to substantial reporting error and dilution of effect.

Publication bias can be rejected; one must invoke some 300 unpublished studies to explain the association.5-3 Lee’s cited fourfold difference in risk of heart disease between smokers who do and do not admit to smoking cannot reasonably be accepted when the difference in risk between genuine smokers and non-smokers is less than twofold.5-3 The published data in the large cohort studies of smoking and heart disease (figure 2 in our paper5-3) confirm a non-linear dose response. Evidence that low dose exposure to tobacco smoke has a pronounced effect on platelet aggregation comes not from one study but also from six experiments in 158 non-smokers who smoked one or two cigarettes.5-3

Dempsey’s appeal for scientific objectivity is welcome; would that it were followed by her employers. As the tobacco industry has never acknowledged that active smoking causes heart disease or lung cancer, it cannot credibly comment on the lower dose exposure of passive smoking.

Denson believes that confounding, particularly with exercise and dietary fat, is important. Both are excluded by the negligible difference in serum cholesterol and blood pressure between non-smokers living and not living with smokers.5-3 In general, confounding is unlikely to explain most of the association because the factors have neither strong enough associations nor sufficient exposure differences to account for the 30% excess risk. For example, few older people in Western countries exercise sufficiently to reduce their risk by 30%; it is unlikely that all non-smokers married to non-smokers (but none married to smokers) would exercise sufficiently.

References

  • 5-1.LeVois ME, Layard MW. Publication bias in the environmental tobacco smoke/coronary heart disease epidemiological literature. Regul Toxicol Pharmacol. 1995;21:184–189. doi: 10.1006/rtph.1995.1023. [DOI] [PubMed] [Google Scholar]
  • 5-2.Layard MW. Ischaemic heart disease and spousal smoking in the national mortality followback survey. Regul Toxicol Pharmacol. 1995;21:180–183. doi: 10.1006/rtph.1995.1022. [DOI] [PubMed] [Google Scholar]
  • 5-3.Law MR, Morris JK, Wald NJ. Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence. BMJ. 1997;315:973–980. doi: 10.1136/bmj.315.7114.973. . (18 October.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5-4.Steenland K, Thun M, Heath C. Environmental tobacco smoke and coronary heart disease. Circulation. 1997;96:2087–2088. doi: 10.1161/01.cir.94.4.622. [DOI] [PubMed] [Google Scholar]

Articles from BMJ : British Medical Journal are provided here courtesy of BMJ Publishing Group

RESOURCES