Note: This is a translation/adaptation of a postscript to a Dutch article (translated version) I published in HP/De Tijd (28 November 2003) in which I criticized a report by the Dutch Health Board (the postscript was not published in HP/De Tijd). In this report, which was published on 17 November 2003, it was claimed that passive smoking kills several thousand people per year in Holland through heart disease, several hundred through lung cancer and several dozen through sudden infant death syndrome. The report was clearly politically motivated, as it was published just two days before a discussion on smoking policy in Dutch parliament. The report was shocking. Not so much because of its unwarranted conclusions, but because of its complete lack of even a hint of scientific objectivity. In my critique I listed as many as twenty examples of omission, error and misdirection in the report. I also documented the authors' (most of whom were physicians) utter lack of basic statistical insight and their sponsorship by pharmaceutical companies which finance anti-smoking campaigns (because they hope to sell medication to help people quit smoking). So the fraud which is ETS (environmental tobacco smoke) has even permeated formerly reputable governmental scientific bodies. While the government report received heavy attention in the press, my criticism of it was conveniently ignored by most of the press. Such is the fate of those who would expose politically correct science.
The notion that ETS (also known as passive or second-hand smoke) can kill you is a scientific fraud. Or rather, it's a witch hunt (on smokers) masquerading as science. The main reason it's very unlikely that passive smoking is deadly, is that the dose is extremely low - about one thousandth of the dose a smoker typically receives (source: Covance Laboratories). And indeed, most studies fail to show a statistically significant correlation between passive smoking and disease. For example, a very large study commissioned by the World Health Organization showed no link between passive smoking and lung cancer or heart disease. The results even indicated a possible beneficial effect of passive smoking on the health of children. The reaction of the World Health Organization, which is vehemently anti-smoking, was to attempt a cover-up of the study (source: The Economist, 12 March 1998). In 1992 the EPA did a study in which they claimed to have shown a link between passive smoking and lung cancer. In 1998 a federal judge ruled that the EPA's decision to classify passive smoke as a carcinogen based on that study was invalid. According to the judge the EPA had grossly violated accepted scientific standards in their study. (Source: Washington Post, 19 July 1998.)
In this article I will explain why most of the studies which conclude that passive smoking increases the risk of, for example, lung cancer or heart disease are invalid or fraudulent.
Confounding factors
Some studies do show a statistically significant correlation between passive smoking and lung cancer of heart disease. But a correlation between two variables is not proof of a causal link, because there may be other "confounding factors" which are the cause of the correlation. For example: suppose a study shows a 20-30% increased risk of heart disease for passive smokers (compared to non-smokers who are not passive smokers). On average smokers are from a lower socio-economic class than non-smokers. So, non-smokers with a smoking husband or wife would also be expected to be more common in the lower socio-economic classes. Perhaps people from lower socio-economic classes typically eat less healthy diets than people from higher socio-economic classes. In that case it is possible that passive smokers are more likely to get heart disease not because of passive smoke but because of their food.
Many factors are suspected to have an influence on the risk of heart disease and lung cancer, and there may be hundreds of other unknown factors. Sometimes one tries to correct for some of these factors, but that's only a partial solution as the total influence of confounding factors cannot be known. For that reason epidemiologists have much stricter guidelines for this kind of so-called "multi-factorial analysis" than for "single-factorial analysis" (diseases with only one risk factor, such as polio in the sense that the polio virus is the fundamental cause of polio). A golden rule of multi-factorial analysis is not to accept relative risks lower than two as proof of a causal connection, because of the problem of confounding factors. (A relative risk of two means that the presence of a certain factor makes the probability of something else twice as high.) But even with higher relative risks researchers often have doubts. For political reasons those researchers most lacking in scientific integrity are willing to make an exception to this golden rule in the case of passive smoking. The relative risks assumed by anti-smoke researchers are typically much lower than two, in particular those for lung cancer (typically 1.2) and heart disease (typically 1.2-1.3). But other much higher relative risk are not taken seriously. For example, according to one study the relative risk of diesel emissions for cancer is 2.6. According to another study various relative risks of electromagnetic radiation for health problems are above 3. And reported relative risks of bird keeping (e.g. for lung cancer) vary between 2 and 6.7. Still, these figures are not taken as good evidence of a causal connection, even though they are much larger than figures claimed for passive smoking. (Source).
Questionnaires
Questionnaires present perhaps an even bigger problem than confounding factors, especially if they are used to determine past exposure to passive smoke. Most passive smoke studies are meaningless because of the illegitimate use of "case-control studies". In a passive smoking case-control study one asks people with, say, lung cancer to what extent they were exposed to passive smoke in the past. Those results are then compared to a "control group" without lung cancer. But obviously people with lung cancer are more likely to report passive smoke exposure than those without lung cancer. Many have a desire to blame something (or someone) for their disease. And the very fact of questioning lung cancer patients about passive smoke is suggestive of a certain answer. In addition, these types of guesses are unreliable because it is impossible for people to precisely remember an amount of exposure to passive smoke many years ago. It is also difficult for people to remember how long they lived together with a smoker in the past and how much he or she smoked. And yet, these guesses about passive smoke exposure are often recorded as if they were exact figures. And then one happily goes on to calculate confidence intervals not based on the uncertainty of these figures, but only on the uncertainty associated with the sample size. Applying statistical techniques to nonsense data and then claiming validity of the results is the hallmark of junk science.
Most ETS studies are invalid case-control studies. They do not measure differences in disease incidence between people who are or are not exposed to passive smoke. Instead all they do is measure differences in biased exposure recall between cases with a disease and "controls" without the disease. These studies are said to demonstrate a danger of ETS, but do not and can not show a correlation between disease and measured ETS exposure. They only show a correlation between disease and biased individual recalls of ETS exposure over a life time. Incidentally, since case-control studies do not measure differences in disease incidence, controlling for incidence confounders is unwarranted. But what is striking is that even though such studies are used to draw conclusions about causality, in conflict with all accepted epidemiological and scientific standards, most passive smoking studies still don't find a statistically significant correlation. And those that do only show that people with a disease are more likely to report passive smoke exposure than people without the disease, which is hardly surprising. A small bias in the answers to questionnaires can easily lead to the false conclusion of a significantly increased ETS exposure. An inference of increased risk from such data is obviously unwarranted on logical and common sense grounds.
Two examples
Two examples. Suppose that 30% of non-smokers without lung cancer report passive smoke exposure in the past. Suppose that passive smoking does not increase the risk of lung cancer. In that case we would also expect 30% of non-smokers with lung cancer to report past passive smoke exposure. But suppose that because of their bias this percentage is 35%. The probability of ever getting lung cancer is about 1% for a non-smoker. But for non-smokers with lung cancer who were passive smokers this probability is overestimated by 5/30 or 17% and thus determined to be 1.17% instead of 1%. In addition, the percentage of non-smokers with lung cancer who do not report past passive smoking drops from 70% to 65%. Thus their estimated probability of lung cancer is underestimated by 5/70 or 7% and determined to be 0.93% instead of 1%. The relative risk is then calculated to be 1.17/0.93=1.26. It is typically claimed that the relative risk of passive smoking for lung cancer is 1.2. We have now seen that this figure can be easily reached with just a small bias in the answers about passive smoking and is therefore no evidence of a causal link.
A second example. Again, suppose that passive smoking does not increase the risk of lung cancer. Suppose there is no bias in the answers about passive smoking by nonsmokers with and without lung cancer. Suppose that in both groups 30% reports having been a passive smoker. But suppose that 2% of the smokers with lung cancer incorrectly self-reports having been a passive smoker rather than an active smoker. Furthermore, suppose that 30% of all people smoke and that smokers have a ten-fold risk of lung cancer compared to non-smokers. If smokers and non-smokers had the same chance of lung cancer, then we would also expect 30% of lung cancer patients to be smokers and 70% to be non-smokers. But because the lung cancer risk is ten times higher for smokers, the ratio between smokers and non-smokers with lung cancer will be 300/70 instead of 30/70. So 300 (81%) out of 370 people with lung cancer will be smokers and 70 (19%) will be non-smoker. Of the 70 non-smokers 21 (30%) will report past passive smoking. But because 2% (6) of the smokers claim to be passive smokers rather than smokers, we see 27 non-smoking passive smokers instead of 21. So their chance of lung cancer appears to have increased by 6/21=29%. The relative risk of passive smoking is then determined to be 1.29. Again we see that a small bias in question answers can easily explain a correlation higher than what is typically claimed to be evidence of causality. If both types of bias are present, the total error will be even larger.
Logic
But the virtual reality of passive smoking studies is sometimes even stranger than this. Instead of classifying non-smokers as passive smoker or non-passive smoker, one often estimates a level of exposure based on long questionnaires. Everybody is exposed to passive smoke now and then, so there is no clear border between passive smokers and non-passive smokers. The degree of exposure is measured by asking questions such as "How many cigarettes did your husband/wife/father/uncle/etc. smoke per day 10, 20...50 years ago?" and "How many hours per day were you exposed to that smoke 10, 20...50 years ago?" Surely one must have a remarkable memory to be able to give meaningful answers to such questions. In the most extreme case, and usually only with the help of questionable meta-analyses, one finds that the exposure to passive smoke is about 20% higher in non-smokers with lung cancer than in non-smokers without lung cancer. Sometimes one then draws the incredulous conclusion that a 100% relative exposure to passive smoke does not cause lung cancer, while a 120% relative exposure is the cause of all cases of lung cancer in non-smokers. Other researchers conclude from these data that passive smoking increases the risk of lung cancer by 20%, a conclusion bereft of any logical foundation. For the 20% refers to a difference in exposure and not to a number of extra cases of lung cancer. In fact the risk differential is not 20% but 100%: 0% of the "control group" has lung cancer and 100% of the cases. From this a relative risk of infinity would follow - which is clearly absurd.
Prospective studies
Because of the problems with questionnaires such case-control studies are completely invalid. Prospective studies (or cohort studies), such as the very large study by van Enstrom and Kabat (which did not find evidence that passive smoking causes lung cancer or heart disease), are more reliable. In a prospective study one follows the same group of people for many years (up to several decades). That way, one can record the smoking and passive smoking habits of people before some of them get lung cancer. Many years later one looks at which people in the group developed lung cancer. The results are more reliable than with a case-control study, because the answers about smoking and passive smoking are not biased by people already having lung cancer. In addition, the questionnaires are more reliable because one asks people about their current exposure to passive smoke at various points in time, which is easier to report accurately than exposure 10, 20...50 years ago.
But even prospective studies of passive smoking are very unreliable, even though they are more reliable than case-control studies. Answers to questions about passive smoke exposure are still presumptive and inexact. And if only a small proportion of smokers report being a passive smoker rather than an active smoker, this can still lead to a large error in the results. Furthermore, one often does not even correct for known confounding factors of which there are some three dozen reported in the literature for lung cancer, and several hundred for cardiovascular diseases. Besides, immaterial positive results are still no good evidence of a causal link because one can never completely correct for confounding factors. If passive smoking were really dangerous, the claimed risk increase is so small that the only way to demonstrate it would be to do a so-called randomized controlled study - the only type of study that is relatively free of the effect of confounding factors. In such a study one would randomly divide participants into two groups: one exposed and one not exposed to passive smoke. Observing the differences in disease incidence after two decades or more might resolve the question, but because of the vanishing conjectural risk it would take an inordinately large number of subjects to reach statistical efficiency, and such a study would be virtually impossible. As things stand now, the only possible conclusion is that if risks are connected to passive smoking, they are not demonstrable or measurable.
Other problems
Another problem, mentioned in the book "Murder a Cigarette" (page 78), by Ralph Harris en Judith Hatton, is that lung cancer is not always correctly diagnosed. For example, one study indicates that 33% of the cases diagnosed as lung cancer, are in fact not cases of lung cancer. Another study concluded that heavy smokers have a 90% probability of having their lung cancer correctly diagnosed, while for non-smokers that probability is only 62%. If smokers with lung cancer have a higher chance of a correct diagnosis than non-smokers with lung cancer, then it also seems possible that non-smokers with lung cancer who are passive smokers are more likely to be correctly diagnosed than non-smokers who are not passive smokers. And that can be a source of error in the results of passive smoking studies. Another potential cause of biased results based on questionnaires is that not all patients may be willing to participate. As said, some patients may find it rewarding to be able to blame their illness on something (or someone). Therefore those lung cancer patients who have been exposed to passive smoke may be more likely to participate in the study.
Contradiction
Another point mentioned in "Murder a Cigarette" (page 79), is that various conclusions from various studies simply don't add up. Or rather, they add up to too much. According to one conclusion 90% of lung cancer cases are caused by active smoking. According to various other studies 30% of lung cancer cases are caused by radon, 40% by occupational hazards and at least 30% by an unhealthy diet. Adding all that up explains 190% of all cases of lung cancer - a contradiction. And we haven't even added the percentages for lung cancer cases supposedly caused by diesel emissions, bird keeping and race (American Indians appear to be less likely than whites to develop lung cancer, while Chinese people appear to be more likely). Apparently there is also a gene which has a large influence on the chance of developing lung cancer. If there is already so much uncertainty about, for example, the risk of smoking, then one might expect even more uncertainty about passive smoking. For the dose is much and much lower for passive smoking. And so the effects, if any, will probably be much smaller and even more difficult to measure.
Meta-analyses
This article, published in the Journal of the American Medical Association, looks at the correlation between the results of meta-analyses and review studies in the area of passive smoking and the affiliation of the authors. 106 publications were examined and 64% of those were found to conclude that passive smoking is unhealthy. However, of the 29% of studies in which one or more of the authors was affiliated with the tobacco industry, only 6% concluded that passive smoking is unhealthy. And of the other 71% of studies as many as 87% concluded that passive smoking is unhealthy.
A broad definition of affiliation was used. For example, if one of the authors had ever been involved in a passive smoking study sponsored by a tobacco company, or had visited at least two seminars sponsored by a tobacco company, the study was classified as tobacco affiliated, even if the study under consideration was not sponsored by a tobacco company.
A large defect in this study is that the authors did not look at affiliation with anti-smoking organizations, government agencies biased against smoking and the pharmaceutical industry (which is largely anti-smoking because they hope to sell medication to help people quit smoking). As mentioned, 29% of the studies were affiliated with pro-smoking organizations. If the other 71% were all affiliated with anti-smoking organizations, two conclusions follow:
It is puzzling that the authors conclude that the inconsistency of the results is largely due to publications written by authors affiliated with tobacco. That is a non sequitur. They only established that there is a huge inconsistency in the results and that the results largely depend on the affiliation of the authors. But they did not establish whether this is due to incorrect conclusions of authors biased for smoking or incorrect conclusions of authors biased against smoking. In fact, there is even some indication in the figures that it is more likely the inconsistency is due to anti-smokers than that it is due to pro-smokers: the percentage of anti-smokers going against their bias (13%) is larger than the percentage of pro-smokers going against their bias (6%). If the fraction of pro-smoke researchers who are objective is the same as the fraction of anti-smoke researchers who are objective, this points to the pro-smoke researchers being more likely to be right. Especially if we consider that the agreement among authors affiliated with the tobacco industry may well be much higher than 94%, because of the very broad definition of affiliation used.
In the 1950s and 1960s the tobacco industry played an ugly role by downplaying the health risks of smoking. Now, in the area of passive smoking, roles are reversed. The tobacco industry correctly concludes that passive smoking is probably not unhealthy, and certainly not deadly, while the anti-smoke industry plays the opposing role by claiming that it is deadly. Smoke can be irritating to non-smokers. That is reason enough for non-smokers to request that smokers don't smoke in their immediate vicinity. But people who accuse smokers of being killers are frauds.
Dose
At the beginning of this article I pointed to a study which found that the typical dose of smoke a passive smoker receives is about one thousandth of that of a typical smoker. That the order of magnitude of this figure is believable, can be seen in the following hypothetical example. Suppose a smoker and a passive smoker are together in a room of 4 by 3 by 2.5 meters. Suppose the smoker inhales half of the smoke while the other half leaves the cigarette directly via the side stream. The smoke concentration depends on the ventilation, but let's suppose that half of all the smoke of the cigarette stays in the room for ten minutes. Furthermore, let's assume twelve breaths a minute of 0.5 litre of air each. Finally, suppose the passive smoker is present half of the times that the smoker smokes. With these assumptions it can be calculated that the passive smoker receives one thousandth of the smoke dose of the smoker.
Before I present the calculation, first some explanation about this example. To make the calculation easier I assumed a constant concentration of smoke during ten minutes. Of course that's not realistic. Normally the smoke concentration in a room will slowly increase during the smoking of the cigarette. And when the smoker finishes his cigarette the smoke concentration will slowly decrease because of ventilation. My example is equivalent to the imaginary situation that the cigarette is completely smoked within 0 seconds, after which the relative concentration of smoke in the room linearly decreases from 100% to 0% over a time span of 10 minutes. From that it follows that on average half of all the smoke of the cigarette is in the room during 10 minutes, and so I took that as an assumption. Of course these are all hypothetical assumptions, but I think the calculation is useful in that it shows the order of magnitude of the figure 1/1000 is credible.
Now the calculation. Suppose the smoker could inhale half of the cigarette within one breath. So he inhales the smoke, sucks in some extra air so that he breathes in a total of 0.5 litre smoke/air, and then exhales. (The other half of the smoke enters the room directly from the side stream of the cigarette.) Now, we assume that the 0.5 litre which the smoker exhales is distributed evenly over the space of the room. Given the dimensions of the room the total volume of air is 30X40X25=30,000 litres. So the 0.5 litre exhaled by the smoker is diluted 60,000 times. So if the passive smoker takes a 0.5 litre breath he receives a dose of smoke 60,000 lower than that of the smoker. But during the ten minutes the smoke is in the room he takes 12X10=120 breaths rather than one. So we must multiply the relative dose of 1/60,000 by 120 and then find a relative dose of smoke of 1/500. But we had assumed that only half of all smoke stays in the room. So we should divide this figure by two and arrive at 1/1000. But we only looked at the smoke exhaled by the smoker. Given our assumption that half of the smoke enters the room directly from the cigarette, we should multiply by two and again arrive at a relative dose of 1/500. Finally, because of our assumption that the passive smoker is present in half of the cases when the smoker smokes a cigarette, we must again divide by two. That brings us back to a relative dose of 1/1000.
But suppose the smoker doesn't smoke half of the cigarette in one puff, but in ten. That doesn't change anything about the amount of smoke the passive smoker inhales, for the amount of smoke entering the room stays the same. It also doesn't change anything about the dose of smoke the smoker receives. For it makes no difference whether he gets a single breath of highly concentrated smoke, or ten breaths which are each ten times diluted. No matter in how many breaths the smoker smokes his cigarette, the result of 1/1000 remains unchanged. And if the smoker breaths in, say, 1 litre of air during each inhalation instead of 0.5 litre, nothing changes either. Again, the amount of smoke the passive smoker receives doesn't change, because the amount of smoke entering the room stays the same. And the amount of smoke the smoker receives also stays the same. He inhales twice as much smoke/air but that is compensated by the fact that the dilution of the smoke with air is also doubled.
Conclusions
There is no reason to assume passive smoking is deadly or even bad for one's health. These are the main reasons for that conclusion:
Links
Murder a cigarette
Passive smokers inhale six cigarettes a year
ETS: The triumph of fraud over truth
Passive smoke and disease: an incredible story
The money trail
Why Review Articles on the Health Effects of Passive Smoking Reach Different Conclusions
Environmental tobacco smoke and tobacco related mortality in a prospective study of Californians, 1960-98
Environmental tobacco smoke and coronary heart syndromes: absence of an association.
Environmental tobacco smoke and ischaemic heart disease: a case study in applying causal criteria
A prospective study of passive smoking and coronary heart disease
A little poison can be good for you
Forces International
Email: henry@sturman.net