Passive smoking

Editor—Owing to the charged atmosphere surrounding the issue of passive smoking, our paper provoked strong reactions on bmj.com. The most disturbing reactions have come from the enforcers of political correctness who pose as disinterested scientists but are willing to use base means to trash a study whose results they dislike. They have no qualms about engaging in personal attacks and unfounded insinuations of dishonesty rather than judging research on its merits.¹ The resulting confusion has misled many readers and diverted attention from the facts of the study.

Table 1.

Self reported total exposure to environmental tobacco smoke among female never smokers in the California cohort of the cancer prevention study I by smoking status of spouse (taken mainly from tables 4 and 5 of full paper)

	History of regular exposure to cigarette smoke from others in work or daily life as of 1999 (%)
Smoking status of spouse	None	Light	Moderate	Heavy
Low exposure:
Married to a never smoker as of 1959	61.7	24.3	10.9	3.1
Married to a never smoker as of 1972	63.6	23.9	9.7	2.8
Never married to a smoker as of 1999	76.7	16.1	5.3	1.9
High exposure:
Married to a smoker of 40+ cigarettes per day as of 1959	16.2	12.5	47.5	23.8
Exposed 40+ years to a smoking spouse as of 1999	14.1	20.5	44.3	21.1

Since 15 May Michael Thun of the American Cancer Society has led a campaign to discredit our study, including his letter above. However, almost every sentence in his letter is misleading, and he disregards key information in the full version of our paper. Contrary to the title of his letter, we have presented an accurate analysis of the California cohort of the cancer prevention study I (CPS I), not disinformation, and it comes from the University of California, Los Angeles, and the State University of New York, Stony Brook, not the tobacco industry.

Anyone who reads the full version of the paper and our response to the reviewers of 9 January² will see that in fact we provided detailed evidence that refutes the claim that our study is “fatally flawed because of misclassification of exposure.” Contrary to Thun's unsubstantiated assertion that “tobacco smoke was so pervasive in the United States in the 1950s and 1960s that virtually everyone was exposed, at home, work, or in other settings,” the table shows that most female never smokers married to never smokers were not exposed. It also shows that 1959 spousal smoking was strongly related to self reported total exposure to environmental tobacco smoke as of 1999, in spite of the misclassification of exposure that occurred over 40 years.

Thun also attempts to minimise our recontact of survivors in 1999. Instead of the 2% and 5% he cites, we obtained 1999 responses from 8.7% (3094/35561) of the subjects alive on 1 January 1960, from 35.6% (3094/8693) of the subjects known to be alive as of 31 December 1998, and from about 45% of the subjects who actually received the questionnaire (see table 1 and text of full paper). In addition, we have shown in tables 2 and 3 that the 1999 respondents were reasonably representative of the 1959 subjects. Thun claims that “misclassification of exposure is compounded because no information was collected on the smoking status of the spouse between 1972 and 1999,” but he completely ignores table 9. This table clearly shows that results for coronary heart disease for follow up periods of 6, 7, and 13 years, when exposure misclassification would be minimised, were the same as the results in tables 7 and 8 for follow-up periods of 26 and 39 years.

Furthermore, although Thun is in a position to check our results by analysing the data from CPS I, he has yet to identify a single error. His attack should be seen for what it is—an attempt to discredit work that is at variance with the position he is committed to. However, the evidence for the health effects of passive smoking is neither as consistent nor as iron clad as Thun wants to portray it. Rather, the widely accepted evidence is the result of selective reporting of data and, when necessary, attempts to suppress divergent data. Our paper provides a prime example of these tactics.

Horton has posed serious questions regarding the issues of conflict of interest and the difficulty of determining the credibility of research findings, particularly those that involve tobacco industry funding. We suggest four things be done for controversial papers such as ours. Firstly, the integrity of the authors should be thoroughly and fairly investigated. In our case, we both have a substantial record of accomplishment in conducting relevant epidemiologic studies and, until now, we have never had our professional integrity challenged. Secondly, full disclosure should be made regarding conflicts of interest, as has been done with our paper. We want to make clear that the tobacco industry played no part in our paper other than providing the final portion of the funding. The tobacco industry never saw any version of our paper before it was published, never attempted to influence the writing of the paper in any way, and did not even know the paper was being published until it became public. In addition, we have never testified on behalf of the tobacco industry, never owned any stock in the tobacco industry, never been employees of the tobacco industry, and would never have accepted tobacco industry funds if there had been any other way to conduct this study. However, full disclosure must be required of all authors and organisations. In particular, what are the competing interests of Thun, and where does the American Cancer Society get its funds? Thirdly, and most importantly, the integrity of the underlying data must be thoroughly and fairly investigated. The best way to resolve questions about the validity of research findings is through independent examination of the underlying data, something that is now required in principle by the Data Quality Act for US studies with public policy implications.³ Fourthly, journals must be willing to publish and discuss controversial findings, as long as they meet the criteria of good science.

Regarding the comments of the working group of the International Agency for Research on Cancer (IARC), we have not claimed that our study changes the weight of the worldwide evidence on environmental tobacco smoke and lung cancer, but it does change the US evidence. When our results are included, meta-analysis of US results on environmental tobacco smoke and lung cancer among both men and women yields a summary relative risk of about 1.10 for ever/never exposure, which is just on the border of statistical significance. Our results have an even greater impact with regard to environmental tobacco smoke and coronary heart disease, where meta-analysis of US results, which constititute most of the evidence, yields summary relative risks of about 1.05 for current/never and ever/never exposure. The end of our response to the reviewers summarises the relative risks for environmental tobacco smoke and coronary heart disease by exposure status for all US cohort studies.² Because of our findings, we conclude that “the association between exposure to environmental tobacco smoke and coronary heart disease and lung cancer may be considerably weaker than generally believed.” Finally, we do not think the weak association with lung cancer means that environmental tobacco smoke “causes” lung cancer, and we certainly do not think that this issue is “resolved scientifically.”

In response to Glaser and Milne, we have used a standard method of analysis for prospective cohort data: Cox proportional hazards regression based on the SAS PHREG program.⁴ All results have been properly adjusted for age at entry, which is by far the strongest risk factor for death. Tables 7 and 8 show that confounding variables such as education have virtually no effect on the relative risks. Too much is being made of statistical fluctuations in tables 2 and 3. For a fair evaluation of our study, it must be put in perspective with all other similar studies, which has not yet been done.

Finally, we too are in favour of the strongest possible protections for non-smokers. However, the attempt to suppress any divergent results because of their possible effect on public policy can only harm science in the long run. In a rational society, there are ample grounds for regulating involuntary exposure to tobacco smoke without manipulating scientific results. What is most dangerous is the willingness to distort the truth to defend one's position, claiming all along that science and righteousness are on one's side.