Smoking: Henry R. Sturman

Lies About Secondhand Smoke

28 November 2003
Henry R. Sturman

It is highly unlikely that secondhand smoke is deadly because the dose received in this manner is extremely low. So why does the Health Board claim that secondhand smoke causes thousands of deaths per year?

——————————

Last week the Health Board presented a report which asserts that secondhand smoke causes at least thousands of deaths each year from heart problems, hundreds from lung cancer and a smaller amount from sudden infant death syndrome. In conflict with these grim conclusions reached by the Board, many studies conclude that nobody dies from secondhand smoke. In 30 studies (mostly in research on lung cancer) 81% of the 177 results found no statistically significant correlation between secondhand smoke and adverse health effects. A positive correlation was found in 13% of the results and a negative correlation was found in 6% of them (which would indicate that secondhand smoke is actually good for you). (Source: Online Library of Drug Policy, BMJ, WHO.) Even a large scale study in 1998 conducted by the World Health Organization, which is fanatically against smoking, could find no connection between secondhand smoke and lung cancer. As a result, the World Health Organization attempted to deep-six the study. (Source: The Economist, March 12, 1998.) By the way, even when a statistically significant correlation is found between secondhand smoke and health damage, that is still not good proof of the conclusion that secondhand smoke is bad for your health. But more about that later.

Everything indicates that the report from the Health Board is a political, and not a scientific, report. Board member Dr. F.W.A. Verheugt said in de Volkskrant [major Dutch newspaper] of November 22nd, that it was only “by coincidence” that the report appeared two days before the Tweede Kamer [the lower house of the Dutch Parliament] met to discuss smoking policy. He was apparently not aware that Dr. J.C. de Jongste, the Chairman of the Board, had already admitted in B&W; [former news/talk show on Dutch public television] on November 17th, that the timing was specially aimed at influencing the Tweede Kamer meeting. Oops.

——————————

Two interest groups play a role in the science of smoking. On one side are the tobacco industry and smokers’ rights organizations, on the other side are anti-smoking organizations and the pharmaceutical industry (which happily sells patches and pills to help people quit smoking). One would hope that an institute such as the Health Board would maintain a distance from such conflicts of interests. But nothing is further from the truth. Board member Paul van Spiegel is currently participating in an evaluation of the so-called RookStopTherapie [Stop Smoking Therapy]. The study is partially financed by the pharmaceutical company GlaxoSmithKline (producers of Nicorette) which set up the website hulpbijstoppennmetroken.nl [a Dutch website to help quit smoking] and which was involved in a bribery scandal in Italy (Source: The Guardian, February 13, 203).

In 1997 van Spiegel published the results of a study which compared the satisfaction of users of two inhaled powders used by asthma patients. Both products are known as excellent drugs, but from the study it appeared that patients found Turbuhaler better than Diskus. Coincidentally, the study was sponsored by Astra, the maker of Turbuhaler. In this type of research it is very easy to produce the desired result. The only thing you have to do is ask patients questions that lay emphasis on the advantages of your own product. A person who participates in this kind of dubious study is of course well qualified to participate in a political report about secondhand smoke.

In 2000 Board member Verheugt was a co-organizer of a medical congress in Vienna for which the pharmaceutical company Pharmacia & Upjohn (owner of the website stoppenmetroken.nu) [stop smoking] was one of the main sponsors. Drug maker Pfizer (a participant in the ‘Partnership to Stop Smoking’) organized a congress in Amsterdam last month. Pfizer donated the participants’ costs to the Sophia Children’s Hospital, where Board Chairman De Jongste is a professor of childhood lung diseases.

——————————

The Health Board’s report is almost entirely based on statistical studies. But De Jongste is no statistician, as his appearance on Twee Vandaag [a Dutch television news program] on November 17th confirms. He said that he places the blame for children with lung symptoms upon the parents if they smoke. But according to his own report the increased risk from secondhand smoke is 20% to 50% for such symptoms.

An extra risk of 20% means that in a large group of families with non-smoking parents in which 100 children normally develop lung symptoms, the number of children would increase to 120 if the parents had been smokers. Smoking by parents would then account for an extra 20 cases of children with lung symptoms. The 20 extra cases represent 17% of the total group of 120. It follows then that 83% of the children with lung symptoms would have appeared even if the parents had not smoked. In a similar manner an increased risk of 50% means that 67% of the cases would not come from parents who smoked. Therefore, even if we believe the report 67 to 83 percent of the smoking parents had nothing to do with the lung symptoms of their children.

——————————

Van Spiegel exhibited the same lack of statistical insight in Trouw [a Dutch daily newspaper] on November 18th. He asserted that in his practice he treats women with lung cancer who never smoked but got the disease because their husbands smoked. It is almost laughable to see how the authors of the report fumble to find a compromise between scientific honesty and working towards the desired conclusions. Accompanying this article is a list of 20 examples of omissions, mistakes and deceptions. The authors list the most important criteria to judge the strength of evidence suggesting a causal link, but in practice rarely follow them. In individual paragraphs lip service is paid to the uncertainty of the results, but typically stern conclusions follow anyway. The writers cherry-pick studies that support their views, give weak criticism of a few studies not in agreement with their conclusions, and cite the conclusions of various studies incorrectly.

——————————

There is a simple reason why secondhand smoke is unlikely to cause death: The dose is extremely low. Measurements show that someone who breathes secondhand smoke daily at home and at work inhales roughly 1/1000th of the amount of smoke that a smoker does (source: Covance Laboratories). We can see that the order of magnitude of this number is believable with the following hypothetical example. Suppose that a smoker and a non-smoker sit in a room that is 4 x 3 x 2.5 meters large [roughly 1,000 cubic feet]. Suppose that the smoker inhales half of the smoke himself and that the other half goes into the air while the cigarette smolders between pulls. The concentration of smoke is dependent upon the amount of ventilation, but let us suppose that half of all the smoke from the cigarette remains in the room for 10 minutes. Further let’s assume 12 breaths per minute of 0.5 liter each. Suppose, finally, that the non-smoker is present half of the time that the smoker smokes a cigarette. With these assumptions it can be calculated that the non-smoker inhales 1/1,000th of the smoke that the smoker does.

Most numbers indicate that active smokers have approximately a 10 times greater chance of developing lung cancer than non-smokers. Therefore, as an example, if a large group of non-smokers would normally develop 100 cases of lung cancer per year, then the number of cases if the group were smokers would be 1000. That is an increase of 900%. If the risk is proportional to the amount of smoke inhaled, then the increase in risk for a person exposed to secondhand smoke is 1/1,000th of that number or about 0.9%. A non-smoker has around a 1% chance of developing lung cancer. But wait: The increase in risk of 0.9% does not mean that the risk of developing lung cancer goes from 1% to 1.9%, but that the chance increases by 0.9% of 1 %, which is 0.009%. That means that the chance that someone who is exposed to secondhand smoke develops lung cancer increases from about 1% to about 1.009% – a negligibly small effect. And if smoke is only dangerous above a certain threshold dose (which is often the assumption with poisonous substances), then the extra risk could actually be 0% instead of 1%. The higher risk assumed in the report of at least 20% completely unrealistic compared to this. Theoretically 20% is possible if the damage caused by smoke is not proportional to the amount of smoke. But with dangerous substances one generally assumes that such an extreme deviation from a proportional relationship is highly unlikely. Moreover, even if the purported increase in risk as stated in the report of 20% is correct, that does not mean that the risk of lung cancer increases from 1% to 21% for a passive smoker, but rather that it increases from 1% to 1.2% – not really something to lose sleep over.

The increased risk of heart disease of 20% to 30% concluded by the Health Board is even more absurd. Whether or not active smoking causes an increased risk of heart problems is in itself controversial, but according to the most pessimistic researchers smoking raises the chance of heart disease by 100%. Since a passive smoker receives only 1/1000th of the amount of smoke that a smoker receives, an estimate of only 0.1% follows for the increase in risk from secondhand smoke.

——————————

Why do some studies seem to show a connection between secondhand smoke and lung cancer or heart disease? If you do enough research you can always find something that will coincidentally give a statistically significant result. In addition, misrepresentation occurs when researchers publish positive (hoped for) results but not negative ones. Further, the methods used are not always of good quality. Meta-analyses (combinations of different individual studies) can be manipulated by selecting inclusion criteria so that only studies with the desired results are included. Question lists used to determine after the fact, for example, how much victims of lung cancer were exposed to secondhand smoke are unreliable. It is difficult for someone to accurately assess the amount of secondhand smoke he was exposed to in the past.

But the most important problem is that of so-called confounding variables. A correlation between two variables is still not evidence that one is caused by the other. A study once showed that families who lived in houses where storks regularly sat on the roofs had more children on average than families who lived in houses less visited by storks. It seems though that the storks were not the reason for more children. It was just that larger families, on average, had bigger houses with more roof space for storks.

Therefore in epidemiology one generally does not interpret relative risks lower than two as evidence for a causal relationship. A relative risk of two means that a certain factor makes the risk of something twice as high. But most of the relative risks in the Health Board’s report are much lower than two, namely 1.2 for lung cancer and 1.2-1.3 for heart disease.

——————————

The Board was correct about one thing. Smoking is annoying and can lead to irritation of the eyes and air passages, among other things. This is sufficient reason for non-smokers to request smokers not to smoke around them. It is not necessary to turn smokers into murderers. Those who do are frauds.

——————————

Twenty examples of omissions, mistakes and misrepresentations in the report “Health Damage from Secondhand Smoke”
1 All numbers in the report have large margins of uncertainty. That is an indication that the results can be the product of coincidence in combination with selective publication or citing. The authors completely ignore this point.

2 The Board offers no numbers concerning the amount of secondhand smoke a person is typically exposed to. This is understandable since the fact that the exposure is in the order of 1/1000th of that of a smoker makes all conclusions in the report a priori absurd.

3 As Elsevier [a Dutch weekly news magazine] reported last week, the Board says nothing about a recent article in Science concerning the theory that poisonous substances in small concentrations might have a beneficial effect.

4 Trouw [a Dutch daily newspaper] reported on November 19th, that a mathematical error was found in the computation of supposed yearly number of sudden infant death syndrome cases caused by secondhand smoke. The computed number should have been ten cases instead of tens of cases.

5 The authors failed to report that they were sponsored by pharmaceutical companies that finance anti-smoking campaigns.

6 In the conclusion it is stated that the risk of low birth weight babies is approximately 20 to 40 percent higher if the future mother smoked or was exposed to secondhand smoke (page 8). This suggests an equivalent risk for smoking and exposure to secondhand smoke, which is extraordinarily improbable given the enormous difference in dose.

7 The authors all but admitted that their method was biased. They say that they mostly used meta-analyses (rather than individual studies) because they all pointed in the same direction (page 15).

8 In addition it is not correct that all of the meta-analyses come to the same conclusion. Even in their own report the Health Board mentioned a meta-analysis with a confidence interval of 95% that did not point in the desired direction (see point 10 below). (A confidence interval gives an indication of the accuracy of a certain number. For example, if the estimated value of a relative risk is 1.35 and has a 95 percent confidence interval of 1.13 – 1.62, then there is a 95% chance (with certain statistical assumptions) that the relative risk actually is between 1.13 and 1.62 and a 5% chance that the relative risk is lower than 1.13 or higher than 1.62. In reality, the uncertainty of the computed relative risks is much greater than the intervals suggest due to the methodological problems mentioned above.)

9 Suggestive: “Moreover components of the smoke and its metabolites can be demonstrated, for example, in the urine of those exposed to secondhand smoke, a sign that the body was actually exposed.” (Page19.) With modern measurement techniques one can demonstrate the presence of the tiniest amounts, but that means nothing in and of itself because whether something is poisonous or not depends on the dose.

10 Manipulative: The report mentions a 1992 meta-analysis from the American EPA (Environmental Protection Agency) which concluded that the chance of lung cancer in women whose husbands smoke increased by 19%. Almost everywhere in the report a confidence interval of 95% was reported, but this time the Board mentioned a margin for the relative risk of 1.04 – 1.35 with a confidence interval of 90%. This is understandable, because with a confidence interval of 95% the lower margin would be less than 1, a result consistent with the theory that secondhand smoke does not increase the chance of developing lung cancer. (Page 20.)

11 The Board did not report that the meta-analysis from the EPA only included 11 studies while at the time there were a minimum of 30 studies available (Source: Murder a Cigarette by Ralph Harris and Judith Hatton). If you choose the best 11 out of 30 studies and still cannot find a statistically significant result, that is not a very good argument for the theory that secondhand smoke causes lung cancer.

12 The Board neglected to mention that an American federal judge voided the decision of the EPA to classify secondhand smoke as a cancer-causing agent based on this study and concluded that the researchers had violated accepted scientific standards.

13 The Health Board did mention a very extensive study that failed to demonstrate that secondhand smoke causes lung cancer, but simply repeated vague arguments against the research from a letter in the British Medical Journal, without mentioning the source (Source: de Volkskrant, November 22). The Health Board did not respond to the detailed defense of the researchers nor did it say what that defense was. The Board stated only that it did not find the defense convincing with no further comment. (Page 21)

14 In addition, the Health Board claimed that the researchers concerned concluded that the risks from secondhand smoke were greatly exaggerated. That is not true. Hans van Maanen reported in de Volkskrant on November 22, that the researchers concluded that there is no evidence for any risk from secondhand smoke.

15 Moreover the relative risk found for lung cancer in spouses of smokers in the above named study was cited incorrectly as 0.75 (0.42 – 1.35). That is the number for men and not for persons exposed to secondhand smoke in general.

16 Concerning the connection between secondhand smoke and forms of cancer other than lung cancer: “In epidemiologic research a statistically significant risk was found in some cases and not in others.” (Page 22.) Why does the Board mention these conflicting results here, but not when the subject is lung cancer or heart disease?

17 The report names a meta-analysis that concluded that mothers who smoke during pregnancy cause a 2.08 times higher risk of sudden infant death syndrome. A different meta-analysis concluded a risk of 2.98. The two confidence intervals (1.83 – 2.38 and 2.51 – 3.54) do not even overlap, from which it follows that at least one of the two studies is defective. The report remains completely silent on this issue. (Page 25.)

18 Further, a relative risk for sudden infant death syndrome of 1.94 was given for secondhand smoke exposure after birth. That is misleading, because this number is based on a comparison of mothers who did or did not continue to smoke after giving birth. The authors did not report that the researchers could not find direct evidence from exposure to secondhand smoke only after the birth.

19 The report references a publication which, according to the Board, claims that the degree to which secondhand smoke exposure by babies after birth has a negative influence on cognitive abilities and behavior, cannot be well judged. This suggests, unjustly, that damage exists but that we simply do not know how much. However, the publication concerned says that there is almost no evidence of such damage.

20 A paragraph about airway injury in children exposed to secondhand smoke states: “The Board shares the vision of various researchers that the relationships mentioned in this paragraph have a causal character.” That is strange, because the Board mentions confidence intervals of 0.95 – 1.53 and 1.08 – 2.04 for the relative risk of secondhand smoke for different forms of middle ear infections. Since both lower limits are very close to 1, this is very weak evidence for a causal relationship.