Rapid Responses to:

LETTERS: Richard Horton Passive smoking: Agreeing the limits of conflicts of interest BMJ 2003; 327: 503-a [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Passive smoking: Non-smoking researchers’ personal values cause conflict of interest.

Erik Nord   (10 October 2003)

Passive Smoking Risks: From Pollution, Not Bias

James L. Repace   (2 December 2003)

Passive smoking: Non-smoking researchers’ personal values cause conflict of interest. 10 October 2003
Erik Nord, Senior Researcher, PhD Norwegian Institute of Public Health, P.O. Box 4404 Nydalen, N-0403 Oslo, Norway Send response to journal: Re: Passive smoking: Non-smoking researchers’ personal values cause conflict of interest.	Editor – Commenting the paper by Enstrom and Kabat [1] on passive smoking, Richard Horton [2] invites the medical community to clarify the issue of conflict of interest. The following story suggests that in research on life style related risk factors, industrial sponsorship is far from the only threat to science. The paper expresses my personal views. A key issue in limiting environmental smoke is whether even low levels of passive smoking significantly increase health risks. If they do, there is a moral case for securing citizens completely smoke free environments in working places and public places of various kinds. If low levels of exposure are harmless, most people will presumably think of compromise solutions as more fair. In Norway, Parliament passed legislation in the spring of 2003 that bans smoking in all pubs, cafes and restaurants from June 2004 [3]. Not even separate smoking rooms will be allowed, on the grounds (a) that waiters will have to enter such rooms from time to time and (b) that even very limited exposures to environmental smoke is ’documented’ to be a significant health hazard. In the Health Minister’s Bill to Parliament, twenty micrograms nicotine per m3 is indicated as an average level of exposure for waiters working in pubs in Norway. One can deduce from other material in the Bill that 5 micrograms/m3 is attainable with various protective measures without banning smoking entirely (the equivalent of smoking less than half a sigarette per day). However, the Bill implies that at an exposure of 5 micrograms/m3 over a whole working life, five in a thousand waiters will die from myocardial infarction. The chain of statements and publications leading to this latter claim may be summarized as follows [4]: A. The Minister of Health ordered a report on passive smoking that would give the necessary scientific basis for proposing a total smoking ban in pubs and restaurants. The assignment was given exclusively to two scientists who, for all their expertise, are known as having strong feelings against smoking and for many years have been heavily engaged in anti-smoking campaigns. B. A diagram in their report showed a linear relationship between exposure to environmental smoke and heart disease, suggesting a level of risk at a nicotine level of 5 micrograms/m3 as indicated above [5]. C. The diagram was reproduced from an article in Risk Analysis in 1998 [6]. Here there is an rough estimate of risk of heart disease at a considerably higher level of exposure (around 20 micrograms/m3) based on observational population data with considerable uncertainty. From this estimate, extrapolation down to zero exposure was done on a simple assumption of linearity. For this assumption there is no justification apart from a reference to an earlier article [7]. D. The earlier article refers the linearity assumption to a 1985 paper by the same authors [8], without further justification. E. The 1985 paper refers the linearity assumption to four publications from around 1980, without further justification [9-12]. F. All these latter papers are about cancer only. They offer biological a priori arguments for assuming linearity in cancer risk. They do not discuss heart disease. G. Even in addressing cancer, the papers are without direct evidence of risk at low levels of exposure – simply because low long term risk is difficult to measure. Instead, they address the hypothesis of a sigmoid relationship between exposure and risk, which is often observed in nature. Such a relationship could include a threshold value for environmental smoke under which exposure is harmless. The papers argue that it is impossible in practice to observe such thresholds (for the same reasons as direct measurement of low exposure risk is unfeasible). Linear extrapolation is therefore recommended in order to be on the safe side: A straight line drawn from the steep (middle) part of a sigmoid (’s-shaped’) curve to origo will lie above the lower part of the sigmoid curve and thus provide a margin of safety at lower levels. Estimates based on linear extrapolation are thus not unbiased estimates (’best guesses’) of risk, but deliberately chosen high ( = biased ) estimates. As noted above, the end point of this chain of information is an allegation by the Minister of Health that significant risk of heart disease from passive smoking is ’documented’ even at low levels of exposure. It turns out that the initial information is in fact (a) not evidence in the ordinary sense of the word and (b) not about heart disease. The misleading referencing also occurs with respect to lung cancer, although not as severely. The Ministry of Health implied that at a nicotine level of 5 micrograms/m3, and exposure over a whole working life (40-50 years), five in ten thousand waiters will die from lung cancer. This was presented as an unbiased estimate (’best guess’). As shown above, it is actually a normative safety judgement. I stress that the point I make here concerns the appropriateness of the referencing and thus the validity of the documentation that the Ministry of Health actually used. It is of course possible that there exists other evidence – published perhaps in later years – that could give more support to the Ministry’s claims. But that remains unproven until one has seen the evidence. The Norwegian Parliament and the Norwegian general public related to the information that was given by the Ministry and was entitled to this being correct. It was not. The story suggests that for research into the relationship between life style and health, the issue of ’conflict of interest’ pertains not only to economic interest. Researchers’ personal values may be equally important threats to impartiality in reporting. So perhaps scientific authors in the health and life style field should be obliged to declare attitudinal positions in addition to economic ones. (I, for instance, am an occasional smoker.) For smoking policies, the story suggests that the scientific basis for uncompromising restrictions in pubs and restaurants is not as sound as the Norwegian Parliament was led to believe. Other countries who are planning similar legislation (Ireland, Holland) may perhaps learn from Norway. References 1. Enstrom JE, Kabat GC. Environmental tobacco smoke and tobacco related mortality in a prospective study of Californians, 1960-98. BMJ 2003;326:1057-61. 2. Horton R. Passive smoking. Agreeing the limits of conflict of interest. Letter. BMJ 2003;327:503. 3. Odelstingsproposisjon nr 23 (2002-2003) om røykfrie serveringssteder (Bill to Parliament regarding smokefree pubs etc). Oslo: Ministry of Health 2002. 4. Nord E. Uetterrettelighet i den sentrale helseforvaltning. J Norw Med Ass 2003; 123: 2763-4. 5. Dybing E, Sanner T. Innspill til høringsnotat om røykfrie serveringssteder. Report dated May 21 2002. Oslo: Directorate for Health and Social Affairs. 6. Repace JL, Jinot J, Bayard S, Emmons K, Hammond SK. Air nicotine and saliva cotinine as indicators of workplace passive smoking exposure and risk. Risk Anal 1998;18:7183. 7. Repace JL, Lowrey AH. An enforceable indoor air quality standard for enviromental tobacco smoke in the workplace. Risk Anal 1993;13:463-475. 8. Repace JL, Lowrey AH. A quantitative estimate of nonsmokers lung cancer risk from passive smoking. Environm Int 1985;1 1:3-22. 9. Crump KS, Hoel DG, Langley CH, Peto R. Fundamental carcinogenic processes and their implications for low dose risk assessment. Cancer Res 1976;36:2973-2979. 10. U. S Environmental Protection Agency. National emission standards for hazardus air pollutants. Fed Reg 1979;44:58642-58661. 11. Interagency Regulatory Liaison Group. Scientific bases for identification of potential carcinogens and estimation of risks. J Natl Cancer Inst 1979;63:241-268. 12. Doll, R, Peto R. The causes of cancer. Oxford: Oxford University Press 1981. Competing interests:   I am an occasional smoker.
Passive Smoking Risks: From Pollution, Not Bias 2 December 2003
James L. Repace, Visiting Assistant Clinical Professor, Tufts University School of Medicine 101 Felicia Lane, Bowie, MD 20720,U.S.A. Send response to journal: Re: Passive Smoking Risks: From Pollution, Not Bias	PASSIVE SMOKING RISKS: FROM POLLUTION, NOT BIAS Editor – In a misguided attempt to clarify the issue of the influence of conflict of interest on passive smoking studies raised by Richard Horton [1], Nord [2] wrongly asserts that the scientific basis for Norway’s workplace smoking ban is unsound. Nord believes that the basis report prepared by the Health Ministry[3] is rife with “conflict of interest”, and does not support a risk of passive smoking at what he asserts are “low levels.” Nord’s criticism [2] focuses on three of my papers [4-6], referenced in the report [3]. Nord asserts that the Health Minstry report’s heart disease dose-response relationship [4] is “undocumented” and that assumptions concerning linearity of cancer risk [5,6] at what he calls “low levels of exposure” are “unjustified”. Our meta-analysis-based passive smoking heart disease risk model predicts occupational passive-smoking heart disease mortality in the middle of the range estimated by U.S. OSHA for U.S. workers [4]. In our cancer risk exposure-response model, the response is derived from the lung cancer mortality rates in 2 large cohort studies of passive smoking in the U.S. and Japan, while population-average exposure is based upon modeling, and is consistent with exposure and dose measurements for nonsmokers from tobacco smoke air pollution [5,6,7]. In validation, our model was able to quantitatively predict, within 5%, both the risk ratio and risk rate observed in the American Cancer Society’s CPS I passive smoking study (ACS-CPS I) [6]. Our model of passive smoking induced lung cancer risk has been widely accepted, e.g. [8-10]. Thus, the Norwegian report’s [3] assumptions based in part on my work are both well documented and properly justified. On the other hand, the issues raised by Horton and Nord [1,2] were provoked by the massively-criticized [1] passive smoking epidemiology paper of Enstrom and Kabat [11], which analyzes a California subset of the very same ACS-CPS I cohort study. Far beyond any conflict of interest issue, Enstrom and Kabat’s scientific analysis is fatally flawed due to its incorrect definition of passive smoking exposure as a “nonsmoker’s marriage to a smoker” [6]. It has long been known, but less widely appreciated, that studies of passive smoking and lung cancer based only on spousal smoking status and failing to take into account workplace exposures will underestimate risk [6,12]. Enstrom and Kabat [11] claim that in the 1950’s and 1960’s “most female nonsmokers married to neversmokers were not exposed.” This long discredited notion [6] remained demonstrably false even as recently as 1989-1991, when a national probability sample showed that a substantial fraction of U.S. nonsmokers reporting “no home or workplace passive smoking” in fact had passive smoking doses greater than those who did report such exposure [13] (see figure below). This means that even when epidemiological studies of passive smoking account for both home and workplace exposure, they will still suffer underestimated risk and depressed statistical significance unless they correct for exposure in the “unexposed” referent group. Moreover, studies of active smoking and disease which have cavalierly assumed nonsmokers to be an unexposed referent group must also underestimate disease risk. In other words, the health risks of both active and passive smoking are worse than commonly understood. Thus, such basic exposure assessment failures in environmental epidemiology have had far more adverse effects on our understanding of the true magnitude of tobacco-smoke-related disease than the real or imagined conflicts of interest raised by Nord and some critics of Enstrom and Kabat. Finally, are real-world exposures to tobacco smoke pollution “low” in any sense of the word as Nord suggests? Much of the debate on passive smoking has unaccountably ignored human exposure to air pollutants. Actual physical measurements of air quality show conclusively that 90%-95% of human exposure to harmful respirable particulate pollution and carcinogenic polycyclic aromatic hydrocarbons in restaurants, bars, casinos, and other venues is eliminated by smoking bans [7,14]. Exposure to these air pollutants has been linked to increased risk of respiratory and cardiovascular disease, as well as cancer [14]. Taken alone, the air pollution exposure reductions afforded by indoor smoking bans result in major public health gains for a very small investment, and provide more than ample justification for the smoke-free workplace laws recently passed in 6 U.S. States, plus those impending in Norway and Ireland in 2004. James Repace, MSc., Health Physicist Visiting Assistant Clinical Professor, Tufts University School of Medicine, and REPACE ASSOCIATES, Inc. Secondhand Smoke Consultants 101 Felicia Lane Bowie, MD 20720, U.S.A. Email: repace@comcast.net; WEBSITE: www.repace.com Disclosure: This work is supported by the Flight Attendant Medical Research Institute’s Distinguished Professor Award. 1. Horton R. Passive smoking. Passive smoking: Agreeing the limits of conflict of interest. Letter. BMJ 2003;327:503. 2. Passive smoking: Non-smoking researchers’ personal values cause conflict of interest. Nord E. BMJ (10 October 2003). 3. Dybing E, Sanner T. Innspill til høringsnotat om røykfrie serveringssteder. Report dated May 21 2002. Oslo: Directorate for Health and Social Affairs. 4. Repace JL, Jinot J, Bayard S, Emmons K, Hammond SK. Air nicotine and saliva cotinine as indicators of workplace passive smoking exposure and risk. Risk Anal 1998;18:7183. 5. Repace JL, Lowrey AH. An enforceable indoor air quality standard for enviromental tobacco smoke in the workplace. Risk Anal 1993;13:463-475. 6. Repace JL, Lowrey AH. A quantitative estimate of nonsmokers lung cancer risk from passive smoking. Environm Int 1985;1 1:3-22. 7. Repace JL, Lowrey AH. Indoor Air Pollution, Tobacco Smoke, and Public Health. SCIENCE 208: 464-474 (l980). 8. Zitting A, Husgafvel-Pursiainen K, Rantanen J. Environmental tobacco smoke – a major preventable cause of impaired health at work. Scandinavian Journal of Work, Environment, & Health (Vol. 28, supplement 2, 2002). 9. Samet JM, Wang SS. Environmental tobacco smoke. Ch. 10, in: Environmental Toxicants, Human Exposures and their Health Effects. M. Lippmann, Ed. John Wiley & Sons, New York, 2000. 10. Weiss ST. Passive smoking, what is the risk? Am Rev Resp Dis 133:1-3, 1986. 11. Enstrom JE, Kabat GC. Environmental tobacco smoke and tobacco related mortality in a prospective study of Californians, 1960-98. BMJ 2003;326:1057-61. 12. Johnson KC, Repace J. Lung cancer and passive smoking - Turning over the wrong stone. BMJ 321 (7270) 1221 (2000). 13. Pirkle JL, Flegal KM, Bernert JT, Brody DJ, Etzel RA, Maurer KR. Exposure of the population to environmental tobacco smoke: the third national health and nutrition examination survey, 1988 to 1991. JAMA 275: 1233- 1240 (1996). 14. Repace JL. 2003. An Air Quality Survey of Respirable Particles and Particulate Carcinogens in Northern Delaware Hospitality Venues Before and After a Smoking Ban. www.repace.com. Recent Reports. Figure 1 (adapted from Pirkle et al.) Nonsmokers’ serum cotinine in a U.S. national probability sample during 1989-1991 [8]. Most subjects reporting no environmental tobacco smoke (ETS) exposure (light shading) are in fact exposed, and most of these have doses significantly overlapping (intermediate shading) those subjects who do report ETS exposure (dark shading). Epidemiological studies relying on self-reporting of passive smoking status will underestimate the risks of passive smoking when comparing “exposed” nonsmokers to “unexposed” nonsmokers, unless ETS exposure of the “unexposed” referent group is accounted for, which very few studies have done. Competing interests: The author is a secondhand-smoke consultant and has served as an expert witness in litigation. He has received numerous awards from public health organizations for his work on secondhand smoke, as listed in his CV at www.repace.com.