Editor-in-Chief of Lifestyle Reviews Rollo Tomassi has been mentioning the late Konrad Jamrozik’s paper on “Estimate of deaths attributable to passive smoking among UK adults: database analysis.” Here is the abstract:
“Objective To estimate deaths from passive smoking in employees of the hospitality industry as well as in the general workforce and general population of the United Kingdom.
Design Calculation, using the formula for population attributable proportion, of deaths likely to have been caused by passive smoking at home and at work in the UK according to occupation. Sensitivity analyses to examine impact of varying assumptions regarding prevalence and risks of exposure.
Setting National UK databases of causes of death, employment, structure of households, and prevalences of active and passive smoking.
Main outcome measures Estimates of deaths due to passive smoking according to age group (< 65 or ≥ 65) and site of exposure (domestic or workplace).
Results Across the United Kingdom as a whole, passive smoking at work is likely to be responsible for the deaths of more than two employed people per working day (617 deaths per year), including 54 deaths in the hospitality industry each year. Each year passive smoking at home might account for another 2700 deaths in persons aged 20-64 years and 8000 deaths among people aged ≥ 65.
Conclusion Exposure at work might contribute up to one fifth of all deaths from passive smoking in the general population aged 20-64 years, and up to half of such deaths among employees of the hospitality industry. Adoption of smoke free policies in all workplaces and reductions in the general prevalence of active smoking would lead to substantial reductions in these avoidable deaths.”
Alas some of the comments in the Rapid Responses do not seem to singing from the same hymn sheet. Best comment is “Overall, the paper must be regarded as speculative and unscientific, adding nothing to the debate on passive smoking.”
“Alastair G Browne, Civil Servant
Response to BMJ, doi:10.1136/bmj.38370.496632.8F (published 2 March 2005)
There are several issues with the above report which make it’s results and conclusion meaningless.
To start off with, the report does not examine the null hypothesis. To quote from the abstract of the report:-
“Objective: To estimate deaths from passive smoking in employees of the hospitality industry as well as in the general workforce and general population of the United Kingdom.”
Here, the assumption is being made that deaths are indeed caused by so-called “passive smoking”. By approaching the subject in this manner, it is not possible to conduct an un-biased study.
The whole paper itself seems to be based upon calculations and estimates, not hard facts. Despite this, this paper is supposedly intended to be taken as fact itself. Would the author care to explain how he has managed to turn assumption into so-called fact without having done any scientific tests himself?
During the introduction, the author suggests that with regard to pubs, bars, nightclubs etc. “…the generation of tobacco smoke is not intrinsic to the process of selling food and drink…” There are plenty who will argue against this. For some people, smoking tobacco goes hand in hand with enjoying alcohol in its various forms. Additionally, on completion of a good meal, there are many who may enjoy rounding it off with a good cigar. Therefore, the smoking of tobacco is far from being “not intrinsic” to these kind of establishments. To view things in any other way is biased.
The author then goes on to state that making these kind of establishments smoke-free would protect staff who work there. Is this not a bit premature to suggest this kind of thing in the introduction? Before any research has been carried out? The assumption has been made here that environmental tobacco smoke (ETS) is harmful to non-smokers. This has been made a-priori and is therefore questionable.
The introduction then goes on to attempt to argue about the economics of the leasure industry vis-a-vis smoking bans. Would the author kindly explain the relevance of this argument in the context of the title of the paper? All this does is set the scene for a biased paper.
Moving on the the method of the experiment, the author talks about percentages of people exposed to ETS. This appears to be justified. However, the author then makes the broad, sweeping statement that smoke exposure in the home provides “…a relatively clear picture of the risks for lung cancer and ischaemic heart disease.” Once again, this is pure supposition. Where is the proven link between ETS and the ailments mentioned? Why is the assumption made that lung cancer and ischaemic heart disease can only be caused by tobacco smoke? What stops other factors such as motor vehicle exhausts, airborne benzine from petrol, or anything else for that matter, from causing these diseases? There is no mention of these other substances in the report. Why not?
Of course, the fabled cotinine test is cited. Bearing in mind that cotinine levels may be raised by the subject eating potatoes, what measures were taken to ensure that this was not a confounding factor?
The calculations performed by the author carry on the assumption that ETS is solely responsible for the ailments listed in the results table. To make such an assumption is completely unfounded and produces results that are invalid.
It is basic scientific procedure that to measure cause and effect, one must first establish and isolate the causes that lead to the effects in question. No attempt has been made to do this in this instance. The fact that a percentage of “smoking-related diseases” may in fact be caused by factors other than ETS has not been considered. The author cites previous work as being definitive with regard to ETS. If one reads these reports, one will find that they are far from definitive. Many of them have similar assumptions to those made in this paper. Additionally, many of these papers make suggestions or suppositions within their summaries. Once again, would the author please explain how he managed to turn these into fact without further work?
To summarise, this paper is biased and scientifically invalid. It draws upon material produced from other sources and rather than analysing them, accepts them as hard fact—even though it may not have been the case that the authors of the sources in question intended for their papers to be taken as fact. The act of taking potentially false information, then applying statistical analysis to it, results in a completely meaningless set of figures.
Alastair G Browne MSc BEng(Hons)”
“Michael J. McFadden, Writer/Researcher/Activist
Two points raised in the Rapid Responses immediately prior to this come together with a question I have written about for several years. USDHHS (U.S. Dept. of Health and Human Services) has classified Ethyl Alcohol as a carcinogen. To be true, they have only classified it as such when it is “consumed,” presumably in liquid form, but alcohol is a very volatile liquid. (1)
A cigarette emits roughly a half milligram of active Class A carcinogens with the most significant in terms of weight being benzene at 3/10ths of a milligram. A standard martini releases roughly one full gram of the Class A carcinogen ethyl alcohol into the air in the space of an hour: an amount equal to 2,000 cigarettes. You can see this for yourself most clearly if you pour a large shot (48 grams) of grain alcohol into a martini glass and set it someplace ventilated and safe for two days. When you come back it will be gone. If the cat didn’t drink it the alcohol went into the air and was breathed and ingested by any who wandered through the room during that period.
Some might claim that DHHS specified “consumption” of alcohol in order to rule out any airborne effects, but to say that mucosal cancers from liquid alcohol do not imply mucosal cancers from evaporated alcohol makes an absolute mockery of the old “tar in acetone painted on mouse skin” proofs that medical scientists were so fond of in the 1950s and 60s.
As Dr. Lee, in the Response immediately above this one, points out: “Logic dictates that if cigarette smoke is harmful when inhaled into the lungs of smokers then the same smoke when inhaled into the lungs of non- smokers will also be harmful. To argue otherwise would be foolhardy.” People like myself argue that the dilution of that smoke, particularly in modern venues with far better ventilation than generally reflected in epidemiological studies based on exposures stretching back 30 or 40 years, make a huge difference.
Nonsmokers in well designed and ventilated bars and restaurants would normally inhale no more than a few micrograms of active Class A carcinogenic material from cigarettes. In exceptionally well designed and ventilated venues the total amount would probably be measurable only in nano- and picograms. The alcohol case is clearly far stronger: nondrinkers would be likely to inhale milligrams rather than mere micrograms in drinking allowed venues… particularly if smoking is banned and ventilation levels reduced.
Of course there’s no massively funded Antialcohol lobby to run huge epidemiological studies on passive drinking and such studies would be very difficult to design. Non-drinking bar workers could be compared to non- drinking pool-hall workers (both groups would be exposed to similar amounts of smoke thereby removing that as a variable) but such population pools aren’t very large. However, as Luc Bonneux pointed out above, risk management often involves making decisions “not supported by hard epidemiologic data.”
He also points out that risk management principles dictate that “risks are to be reduced to levels as low as possible. The lowest level possible is easy to identify: apply the law, ban smoking in all public places.” Following those principles, and following the thinking and reasoning of Dr. Lee, one would have to accept the necessity of banning alcoholic drinks in restaurants and perhaps even in bars.
Sure, it might hurt their businesses a bit. Fancy upscale restaurants would lose their profits on $100 bottles of wine and the clientele frequenting bars would probably suffer an initial dip until people got used to enjoying each others’ company over glasses of vegetable juice or soda pop. However, as has been stated consistently by smoking ban proponents, any possible risk to health is first and foremost: nonpartakers should not be forced to partake anything over a zero- tolerance exposure to any potentially harmful chemical, and the removal of drug use from the visual environment of children and teenagers will make them less likely to become addicted themselves.
We should remember that underage drinkers account for nearly 20% of alcohol consumption (2) while underage smokers account for just 3% to 5% of the cigarette market (3). Would removing alcohol from bars, restaurants, TV commercials, movies, and sports events while quintupling its taxes cut underage drinking by 400%? Antismoking advocates feel such measures are effective for tobacco so they should be similarly effective for alcohol, true?
Do I really think alcohol should be banned from bars and restaurants, from movies and TV, from sports events and wedding champagne toasts? Of course not: the risk levels of nondrinkers, if evaluated honestly in studies not funded by pressure groups would be below consideration by any rational person. And anyone who did not want to be exposed to such fantastical wisps of risk could simply avoid establishments where alcohol was allowed.
Exactly the same argument can be made about smoking and smoking bans.
Michael J. McFadden
Author of “Dissecting Antismokers’ Brains”
(1) National Toxicology Program. DHHS. 11th Report on Carcinogens http://ntp.niehs.nih.gov/ntp/roc/eleventh/known.pdf
(2) Foster, S.E., R.D. Vaughn, W.H. Foster and J. A. Califano, Jr. 2003. Alcohol consumption and expenditures for underage drinking and adult excessive drinking. JAMA 289 (8): 989-95
(3) Viscusi, W. Kip. “Smoke and Mirrors….” The Brookings Review, Winter 1998 Vol. 16 No. 1, pp. 14-19
I am a member of several Free Choice organizations, and have absolutely no financial interests with Big Tobacco, Big Hospitality, or any other player in this arena other than as a customer and as the author of a book in the field.
Competing interests: I am a member of several Free Choice organizations, and have absolutely no financial interests with Big Tobacco, Big Hospitality, Big Pharma, or any other player in this arena other than as a customer and as the author of a book in the field.”
“Gio B. Gori, Principal Scientists, The Health Policy Center
In the enumeration of UK deaths attributable to ETS, Konrad Jamrozik – professor of evidence based healthcare – makes a number of arguable assumptions. (1) Of these, none is more consequential than stipulating that the claimed epidemiologic risks of ETS exposure are based on valid evidence.
A number of critics favorable and unfavorable to ETS risk claims have voiced concern for some 30 years about the uncertainties attending epidemiologic studies of passive smoking, with special reference to the central role of exposure recall. Likely because not much could be done about the latter, the issue remained dormant and was eclipsed by sideline interests in recall bias, smoker misclassification, matching bias, confounders, publication bias, statistical issues, and meta-analysis. Still, some notion of the potential dimensions of recall uncertainties is crucial and long overdue. Recently, I chanced to examine the questionnaire used to interview participants in the largest multi-country IARC/WHO European ETS study of 1998. (2) The editors of BMJ or anyone interested in evidence based findings ought to analyze this document critically, for it is claimed to be among the best if not the best of ETS questionnaires, and is intended to generate the essential input to risk estimates, namely an integral of ETS exposure over the lifetime of each study subject.
What and how many cigarettes, pipes, or cigarillos did mom or dad smoke daily in my presence during my childhood and youth? What about uncle Joe and granddad? What and how many cigarettes, pipes, or cigarillos did my spouse(s) smoke in my presence during a lifetime? How airtight were my homes? How smoky were different workplaces during my lifetime? Can I recall my precise diet of last week? How much better is it than throwing dice?
Even a cursory perusal of the IARC questionnaire reveals that most questions can be leading, and that answers can only be vague guesses open to multifold error. Moreover, risk calculations utilize indexes of exposure that are extruded from a combination of several answers, whereby the error of composite indexes is the amplified sum of the errors of their components, as defined by standard rules. (3)
It is a common sense notion and the fundamental principle of metrology that a measurement is not a measurement unless coupled to an estimate of its uncertainty. (3) Inexplicably however, individual recalls of exposure are written down as precise digits, and are later used in estimates of risk and statistical significance under the delusory assumption that such digits have no margin of uncertainty. In reality, and although unknown and unknowable, the uncertainties of individual ETS exposure recalls jump in the face as obviously real, and their composite weight can easily exceed the claimed 20-50% risk excesses attributed to ETS exposure.
We are facing the ugly prospect that the entire epidemiologic literature on ETS is in fact a gross delusion. Should the flagrant inconsistency of the various results surprise? As gatekeepers and purveyors of evidence based information, the editors of BMJ have an obligation to clarify this conundrum to their readers. It could not be dismissed with the unworthy retort that I might be an industry spokesman, for these are verifiable facts an not opinion. Wisely, the editors have noted that “passive smoking is accepted as a cause of fatal disorders”. Indeed it may be accepted, but is it so? Is BMJ prepared to make a similar allowance for creationism?
1.Jamrozik K. Estimate of deaths attributable to passive smoking among UK adults: database analysis. BMJ, doi:10.1136/bmj.38370.496632.8F, March 2, 2005.
Competing interests: I have occasionally consulted with the tobacco industry. I have also been deputy director of the Division of Cancer Cause and Prevention and director of the Smoking and Health Program of the US National Cancer Institute. My interest is truth, or its closest possible approximation.”
“Peter N Lee, Consultant Statistician
EDITOR – Whereas previous estimates of risk from passive smoking have been limited to nonsmokers, Jamrozik produces much higher estimates of deaths by including deaths in smokers. He notes that the joint effect of active and passive smoking has not been examined epidemiologically and argues, not unreasonably, that an additive model is appropriate. However his calculations, which extrapolate relative risk estimates derived from studies of nonsmokers to the whole population – smokers included – actually involve an implausible multiplicative model. The huge difference this makes can be seen by assuming that, compared to unexposed nonsmokers, passively exposed nonsmokers have a relative risk of lung cancer of 1.24 and unexposed smokers have a relative risk of 20. Under the multiplicative model that Jamrozik actually uses, the relative risk of a smoker exposed to passive smoking would be 20 x 1.24 = 24.8, with 4.8/24.8 = 19.4% of deaths in this group attributed to passive smoking. Under the additive model that he argues for, but does not actually use, the relative risk would be 20.24, the percentage of deaths attributed to passive smoking being much lower, at 0.24/20.24 = 1.2%.
There are other technical problems with Jamrozik’s analysis. He assumes that, because 85% of adults aged 20-64 work, 85% of deaths in adults of this age occur in workers, clearly incorrect in view of the well -known “healthy worker effect”. He also assumes without any support that at home and at work exposure are independent, again leading to overestimation of the risk. Failure properly to take age into account is also a potential problem. Among adults of working age, is the average age of workers in the hospitality industry really the same as that of the whole population, as implicitly assumed?
The whole calculation is only relevant if the relative risk estimates used for lung cancer, ischaemic heart disease (IHD) and stroke actually represent causal effects of passive smoke exposure. For lung cancer, my colleagues and I1 recently concluded that most, if not all, of the association is an artefact due to various sources of bias, based on a detailed analysis which recent reviews2,3 have considered, but have not refuted. For IHD, there are also problems of interpretation4, with the largest studies showing little or no association.5,6
For stroke, the relative risk estimate is based on only seven studies, and overlooks evidence from as many other studies,7-13 most of which find little or no relationship. Until a full review of the evidence taking proper account of potential sources of bias and confounding has been published, it remains unclear whether any causal effect exists.
Overall, the paper must be regarded as speculative and unscientific, adding nothing to the debate on passive smoking.
Peter Lee Independent Consultant in Epidemiology and Statistics 17 Cedar Road, Sutton, Surrey SM2 5DA PeterLee@pnlee.demon.co.uk
Competing interests : Long-term consultant to the tobacco industry”