Responses to the SCENIHR questions:
Carl V. Phillips
The Questions
[1.0] What are the adverse health effects of smokeless tobacco?
[2.0] What is the addiction potential of smokeless tobacco?
[3.0] Does the available data support the claim that smokeless tobacco may constitute a smoking cessation aid comparable to pharmaceutical nicotine products?
[4.0] What is the impact of smokeless tobacco use on subsequent initiation of smoking?
[5.0] Is it possible to extrapolate the information on the patterns of smokeless tobacco use, smoking cessation and initiation from countries where oral tobacco is available to EU-countries where oral tobacco is not available?
[1.0] What are the adverse health effects of smokeless tobacco?
I have chosen to focus my remarks on issues of interpretation of epidemiologic evidence and related epistemology, my primary area of professional expertise.
Epidemiologic evidence is both fundamentally important (because it is the only way to observe actual health effects on actual people) and notorious (because of study errors, publication biases, and other uncertainties that mean that most conclusions are overstated). This means that some critical analysis is necessary. But the report seems to uncritically accept the conclusions stated by the authors of various studies in their article abstracts, without attempting to critically evaluate the results. Such headline conclusions are often biased or inaccurate portrayals of the results.
It is impossible to go into detail about every study in this comment process, but it is surprising that it was not done as part of the report. There are hundreds of critical analysis that have been written about articles you cite, in letters to the editor, subsequent articles, and other forums, but these were ignored. To note just one particularly well-known example, the Winn et al. (1981) study that is probably the most often cited found a very low relative risk for black women, and so only the famous relative risk of 4.2 for non-black women (usually mis-identified as white women) was emphasized by the authors. Even though you presumably do not think that black people are fundamentally different or do not count, you repeat this biased reporting.
The only critical questioning of the conclusions that study authors presented and emphasized that I could identify in the report was with regard to Schildt et al. (1998). This is the one case I am aware of where the authors stated a conclusion of no elevated risk, but others have published assessments that show that the data can be interpreted to conclude otherwise. (This contrasts with at least a dozen of the studies that claim that modern Western smokeless tobacco (ST) causes disease, where later published assessments show that the data can be interpreted as showing otherwise.) The singling out of this study for critical comment suggests a motivational bias in this report. Part of the comment about Schildt et al. is, "The relatively weak effect of smoking is noteworthy," and yet the assessment of Winn et al. did not mention the strangely low relative risk from smoking reported in that study.
The simple acceptance of whatever happens to have been written is particularly problematic in the case of pancreatic cancer. Epidemiologic publication tends to follow a predictable cycle. Because there are so many possible exposure-disease associations, there is little interest in a paper that says "E did not seem to cause D in our data" until there are articles that claim that E does cause D. Thus, the first few papers published on an exposure-disease association will almost inevitably claim there is an association. It is often the case that a decade of null results will then follow, showing that the original publications were outliers in the inevitable distribution of results across studies. You chronicle exactly this phenomenon in the literature on ST and oral cancer, and come to the conclusion that the science does not support the claim of association. But as recently as 10 years ago, many who looked at only what had been published at the time would have insisted otherwise.
It is currently premature to conclude that ST causes pancreatic cancer, let alone with the definitiveness and finality that appears in the answer to the question. The reasons for this include: the aforementioned incentives and arc in publication; the low relative risks reported in the studies that claim to demonstrate an effect, combined with the failure to adequately control for known pancreatic cancer risk factors; and the critiques that have been written about some of the studies (particularly Bofetta et al. 2005 and Alguacil and Silverman, 2004).
This is not to suggest that we should ignore the possibility that ST cause pancreatic cancer in research and policy. But the evidence of the claim is similar to what existed regarding oral cancer in the 1980s, and the research publication arc has now proven rather embarrassing for those who stated emphatic conclusions back then. In addition, to the extent that tentative conclusions are drawn, a simple qualitative statement of "increased risk" is not terribly informative; the absolute health risk caused by ST, even if there is a relative risk of 1.7 for pancreatic cancer, is quite small.
Regarding the claim that TSNAs are human carcinogens: It is worth noting that there is no direct evidence of this and it must be inferred from a variety of very indirect sources. Subjecting non-human animals to exceptionally high doses of these chemicals has induced cancer in some cases, but there are few chemicals for which this is not true. We can reason by analogy, considering other nitrosamines where there is convincing epidemiologic evidence of carcinogenicity. But these are only suggestive of possible human carcinogenicity from TSNAs in doses/quantities that are actually experienced.
The only direct evidence that TSNAs in ST are carcinogenic comes from comparing studies of older or non-Western products (which have much higher levels of TSNAs) to studies of modern products. This is the only source of a contrasting exposure to TSNAs, holding most other things (in particular, exposure to smokeless tobacco) equal. Some studies of older or non-Western products show an association with oral cancer, whereas modern studies of modern products do not. Thus, if one believes that the older and non-Western studies sometimes show positive associations because of genuine different effects (and not poorer methodology or publication bias), the reasonable conclusion is that TSNAs cause cancer in high concentrations, but the concentrations in current products do not cause measurable levels of cancer.
It is important to recognize what constitutes evidence of a claim and what is related by tangential. Three is something of a double standard about such inference in the report, wherein social/economic evidence from Sweden is very strongly suggestive of the potential of ST in reducing smoking, but in the report much of that evidence is treated as uninformative because it requires inferences beyond what is directly observed. And yet TSNA exposure from ST use is declared to be carcinogenic, when this requires a quite tenuous inference from indirect evidence.
Finally I note that though the report offered some answers to this question in its most relevant form – i.e., what are the health effects as compared to the existing, popular substitute product, cigarettes – but the summarized answer made the much less informative comparison of the health effects of ST versus the unrealistic scenario of no tobacco use at all.
[2.0] What is the addiction potential of smokeless tobacco?
I have chosen to focus my remarks on issues of interpretation of epidemiologic evidence and related epistemology, my primary area of professional expertise.
[3.0] Does the available data support the claim that smokeless tobacco may constitute a smoking cessation aid comparable to pharmaceutical nicotine products?
In direct response to the answer to this question, it should be noted that it is not standard scientific epistemology to draw comparative conclusions only if a direct comparative study has been conducted. That is, the lead statement that notes that there has not been a randomized trial comparing smokeless tobacco to pharmaceutical products is literally true, but rather disingenuous, since a reader would likely conclude that there has been no clinical trial of smokeless tobacco as a smoking cessation tool, but as you note in your text, there has been such a study (Tilashalski et al. 1998). Moreover, when citing that study, you fail to cite the seven-year followup of it (Tilashalski et al. 2005). These studies, which constitute useful information about the question as asked, suggested that smokeless tobacco is much more effective in aiding smoking cessation than studies of pharmaceutical products suggest that those products are.
More important, however, are two observations that must be made about this question.
First, any assessment of past switching to a highly-reduced-harm alternative to cigarettes must be considered in the context of what consumers believed about different products' health risks. People cannot be expected to change their behavior to improve their health if they do not know that the new behavior is health-improving. In the case of smokeless tobacco, very few people in Europe, outside of Sweden, or in North America know that smokeless tobacco poses approximately 1/100th the risk of smoking. There is a substantial literature about the efforts made by anti-tobacco activists to misinform people about smokeless tobacco, and these have succeeded in convincing most people that smokeless tobacco poses a health risk similar to that from cigarettes (see references below). Thus, past observations about decisions to switch, outside of Sweden, provide very little information about what would happen if people knew the truth.
The only thing we can conclude from observing people who mostly do not know the comparative risk is that however many of them switched products, more people would use smokeless tobacco as an effective means to quitting smoking if everyone were given access to the products and accurate information.
Second, it is sometimes the case that literally answering a question, rather than responding to its underlying premise, can distract us from what is most useful to know. Asking whether smokeless tobacco provides an alternative to smoking that is comparable to pharmaceutical nicotine products implicitly assumes (a) that the effectiveness of pharmaceutical products presents a worthy target and (b) that comparability is relevant to decision making.
A large body of research shows that the effectiveness of existing pharmaceutical products in smoking cessation is quite poor, usually barely better than unassisted quit rates. It is likely that most advocates of tobacco harm reduction would find it a substantial (and unexpected) disappointment if a population of smokers were educated about the low risks from smokeless tobacco, but their rate of smokeless-tobacco-aided smoking cessation were no higher than the rate of successful quitting using pharmaceutical products.
Moreover, it is not clear why comparability to the effectiveness of pharmaceutical products is relevant to anything. When proposing a new safety feature for cars, we do not ask if its benefits are comparable to those of side air bags. If there are benefits, it does not matter whether they are larger or smaller than the benefits of some other intervention unless the two interventions compete in an either-or manner. If the interventions do not crowd each other out, then from a health perspective the net benefits of the new option add to the benefits that have already been provided by the existing option. There is no reason to expect that smokeless tobacco (whose public health niche would be to provide a long-term source of nicotine in doses comparable to those from smoking for those who do not quit entirely) and existing pharmaceutical products (whose niche is to assist the transition to nicotine abstinence) would compete. Indeed, given the over-hyped messages about the benefits of using pharmaceutical products, it is likely that few smokers who consider switching to smokeless tobacco would not have already tried (and failed with) the pharmaceutical products.
(Note that in this context, competition refers to whether one product might crowd out the health benefits from the other. From a corporate profits standpoint, anything that is effective at getting people to quit smoking is a competitor for pharmaceutical stop-smoking products. Thus, pharmaceutical companies presumably are interested in the comparison that was posed and concerned about smokeless tobacco because of the effect it might have on their profits, but this is not relevant to a public-spirited analysis of health effects.)
It would be possible to create pharmaceutical products that occupied the same niche as smokeless tobacco in smoking cessation efforts. A high-peak-dose, competitively-priced pharmaceutical nicotine products would directly compete with smokeless tobacco in all ways. However, even then the introduction of smokeless tobacco would still have net benefits if there is anyone who would not switch to long-term-substitute pharmaceutical products, but would switch to smokeless tobacco. There would be no measurable health advantage of one product over the other, since there is no reason to believe that the pharmaceutical product would have higher or lower health effects than smokeless tobacco but there is every reason to believe that both would be very low. (Why is that? We have never observed a large population of long-term pharmaceutical nicotine users, so the only scientific basis we have for concluding the risks from it would be low is that the risks from smokeless tobacco are so low. Thus we have absolutely no basis for concluding that the risks would be different.) Therefore even if smokeless tobacco were competing with some hypothetical pharmaceutical product in the future, the competition would only represent a cost to the pharmaceutical companies; there would be no measurable public health cost from the competition.
References
Haddock, C.K., Lando, H., Klesges, R.C., Peterson, A.L., & Scarinci, I.C. (2004). Modified tobacco use and lifestyle change in risk reducing beliefs about smoking. American Journal of Preventive Medicine, 27, 35 41.
O'Connor, R.J., Hyland, A., Giovino, G.A., Fong, G.T., & Cummings, K.M. (2005). Smoker awareness of and beliefs about supposedly less harmful tobacco products. American Journal of Preventive Medicine, 29, 85 90.
Phillips CV, Wang C, Guenzel B. You might as well smoke; the misleading and harmful public message about smokeless tobacco. BMC Public Health 5:31, 2005. http://www.biomedcentral.com/1471 2458/5/31
Smith, S.Y., Curbow, B., Stillman, F.A. (2007). Harm perception of nicotine products in college freshmen. Nicotine &Tobacco Research, 9, 977–982.
Tilashalski K, Rodu B, Cole P. Seven year follow up of smoking cessation with smokeless tobacco. J Psychoactive Drugs. 2005 Mar;37(1):105 8.
Tilashalski K, Rodu B, Cole P. A pilot study of smokeless tobacco in smoking cessation. Am J Med. 1998 May;104(5):456 8
[4.0] What is the impact of smokeless tobacco use on subsequent initiation of smoking?
The question of whether the existence or popularity of smokeless tobacco might cause an increase or decrease in smoking is of critical importance since, due to the very minor direct health effects of using smokeless tobacco, this is the only potential source of substantial health impacts. The question is phrased without reference to whether the impact is an increase or decrease in smoking initiation, though the answer emphasizes only a potential increase. This may not be entirely appropriate since, in addition to the expectation that some current smokers would switch to a lower risk product, the culturally accepted use of smokeless tobacco might cause some people who otherwise have started smoking to never do so, but only use smokeless tobacco instead. However, in a culture where the overwhelmingly dominant form of tobacco use is cigarette smoking, there are likely to be few people who are inclined to use tobacco regularly who do not meet the "100 cigarettes ever" threshold for being called a smoker. Thus, even when smokeless tobacco provides a substitute for long term use of cigarettes, it may prevent few nicotine users from crossing the "ever smoker" threshold (which is not to say that it will prevent them from smoking so much that it kills them, which it likely will in many cases).
In assessing whether there is a "gateway" effect, it is important to clarify what causation means in this context. For the gateway concept to have meaning we have to consider cause and effect (or "impact", as phrased in the question). Smokeless tobacco availability causing smoking initiation means that someone who would never have smoked if smokeless tobacco were not available becomes a smoker when smokeless tobacco is available (probably after trying smokeless tobacco and switching, though this need not be the mechanism). It is only people who fit that causal scenario that are prevented from smoking by maintaining a ban on smokeless tobacco.
There is no reason to believe that there is any substantial population that fits that description. Someone with an inclination to use nicotine and who is willing to risk the health impact of smoking will smoke if there is no other satisfying source of nicotine available. (The persistence of smoking in spite of extreme anti-smoking efforts in some jurisdictions suggests that about 20% of the population finds nicotine sufficiently appealing to meet this description. It is very unlikely that such individuals will never even try nicotine, and thus could be kept a nonsmoker by hiding the appeal of nicotine.) Who, then, might use smokeless tobacco but would not use tobacco at all if it were not available? This must be people who like nicotine, but not enough to suffer the health consequences of smoking. But why would such a smokeless tobacco user switch to cigarettes, which still have the health risk that they were trying to avoid? Most anyone who would be willing to do that would probably have just smoked in a world that lacked smokeless tobacco.
The logic of this assessment is important because simplistic empiricism is of little value when it is ignored. For example, if someone would have taken up smoking anyway then the fact that they used smokeless tobacco first is not a case of causation, though many of the claims about a gateway effect simply look at the order of use. Nor is it surprising that people who use one product are more likely to use the other product, either in the past, the future, or concurrently; this just reflects the fact that some people quite like to consume nicotine while others do not. The U.S. studies that are cited in the report provide no actual evidence of a gateway effect. All just provide evidence that the same people who are inclined to use nicotine from one source are more inclined than others to later use nicotine from another source.
The report and the answer to this question emphasize cultural differences (between Sweden and other countries) in arguing that we cannot extrapolate the data we have. But much more important than cultural difference are knowledge differences. If there is currently a gateway effect in North America, there is no reason to expect that it would persist if consumers were given honest information. North American consumers are currently the victim of misinformation about smokeless tobacco, and generally believe its use is at least as unhealthy as smoking. Thus, if after becoming a nicotine user they discover they like smoking a bit more than smokeless tobacco use, they see no reason to not switch. I have dubbed this the "you might as well smoke" message to smokeless tobacco users, and it is undoubtedly responsible for killing some people who would not have switched to smoking if they knew the truth.
Completing the analysis, the people who are true gateway cases in a population that receives honest information are those few who: -would not have smoked if that were the only source of nicotine, -but use nicotine when a much less harmful form is available (which, incidentally, could be pharmaceutical as much as smokeless tobacco), -though they still like cigarettes better, -and then decide that nicotine/tobacco has become so much more appealing that the risk from cigarettes is now worth the marginal benefit that cigarettes offer above that from smokeless tobacco.
That is quite a conjunction. It is basically impossible for us to empirically determine how many people would fit that description, and so the studies that get cited are largely a smokescreen for empirical ignorance. But the logic of the situation is much more compelling than the observational evidence: it is difficult to imagine very many true gateway cases once people knew the comparative risks.
[5.0] Is it possible to extrapolate the information on the patterns of smokeless tobacco use, smoking cessation and initiation from countries where oral tobacco is available to EU-countries where oral tobacco is not available?
If interpreted quantitatively, the answer to this question seems accurate but relatively uninteresting. That is, we cannot predict how many smokers will switch to smokeless tobacco, and in what communities, by a particular time. There are too many economic, sociologic, and educational issues that are left as unknowns in a scenario that merely describes the changing availability of smokeless tobacco.
However, if interpreted qualitatively, the answer given to this question in the report borders on scientific nihilism, suggesting a belief that we can never predict anything that we have not already seen (a attitude that translates into not predicting anything in public health since the social situation always differs). Fundamental and empirically well-established economic theory predicts that when a much-lower-cost substitute for a product is offered, many consumers of the existing product will switch to it. In this case, the much lower cost is the health cost, not the purchases price. The existing product has a roughly 1-in-3 chance of substantially hastening a user's death, while the substitute product has about 1/100th that risk; it is difficult to imagine a deeper discount.
Of course it is always reassuring to have at least one direct empirical observation to back a theory, no matter how compelling that theory is, and no matter how well supported it is by analogous experiences. This is the value of the Swedish experience, as a confirmation of something that we have every reason to believe is true. Epistemically, this is quite different from the Swedish data being the only reason to believe that people would switch upon learning of the comparative risks.
If we had just a few empirical findings, based on limited data, and not backed by any theory, then it would be appropriate to say "this is not really enough to go on; there might be something odd about this population". For example, that would be the appropriate response to the epidemiologic claims that smokeless tobacco causes pancreatic cancer, even though it does not cause cancer at more proximate sites. But a single example that shows that an overwhelmingly convincing theory is borne out is a good reality check, and so it does not need to be extrapolated to be informative. The Swedish phenomenon is confirmation that a near-universally accurate economic prediction applies in the particular case of smokeless tobacco. While switching patterns in Sweden now may have social forces driving them, before smokeless tobacco became socially popular, first-movers were apparently motivated by the lower cost.
Again, social factors – including cultural history of tobacco use, the social role of smoking, attitudes toward environmental tobacco smoke and other perceptions of what is clean and dirty, and most importantly education about and attitude toward health risks – will clearly influence the pace at which smokers switch products. In particular, if no one is willing or able to educate the public about the fact that Western-style smokeless tobacco poses about 1/100th the risk of smoking, then consumers will not realize that the product is much less costly and have no reason to switch. The rate at which the experience in other places comes to replicate the experience in Sweden is not some uncontrollable feature of the populations; it is largely within the control of the same authorities who currently discourage/ban smokeless tobacco.
In summary, the answer to "is it possible to extrapolate" is, perhaps, "no", if we pretend that we know nothing about human behavior other than the observations about Swedish tobacco use. But the more informative answer is that we do not need to extrapolate, because other sources of knowledge provide much more convincing evidence that that experience can be replicated if people are given accurate information.
|
Back to SCENIHR Page |