«Smokeless tobacco and oral cancer, the curious history of a fact Carl V. Phillips, MPP PhD Constance Wang, MS PhD Brian Guenzel, BS Christina Daw, ...»
Smokeless tobacco and oral cancer,
the curious history of a "fact"
Carl V. Phillips, MPP PhD
Constance Wang, MS PhD
Brian Guenzel, BS
Christina Daw, MPH
University of Texas School of Public Health
Center for Philosophy, Health, and Policy Sciences, Inc.
Poster, 36th Annual Society for Epidemiologic Research meeting
11 June 2003
Smokeless tobacco (ST) use (which currently refers primarily to snuff dipping)
is widely believed to have health consequences similar to those from smoking, and in particular, to be a major risk factor for oral cancer.
This, it turns out, is not true. What is true:
Smokeless tobacco is so much less harmful than smoking that they should not even be compared, and the claims about oral cancer risk are not well supported.
As a result of the widespread misunderstanding, we are missing a huge opportunity for harm reduction: We should be encouraging tobacco users to use ST rather than cigarettes.
Sweden has seen smoking rates fall dramatically as ST has become more widely used. The result is the expected drop in smoking related diseases, including lung cancer rates less than half that of the U.S.
The Royal College of Physicians  has suggested that it is time to consider such a harm reduction strategy. The number of other public health advocates supporting such a strategy is increasing.
Indeed, ST appears to offer a best-case scenario for harm reduction: Even for the worst plausible health effects of ST, the reduction in risk for life threatening disease compared to smoking is clearly more than 90%, and quite possibly more than 99%.
The research leaves no room for doubt about the reduction in risk. Studies have looked for associations between the use of ST and a variety of cancers and other diseases. The only claim that has stuck is that ST causes oral cancer. Even if the worst plausible claims about that association were true, the risk from ST would still be only about 1% of that from cigarettes, and harm reduction would still be a promising strategy.
But the evidence is far from providing clear support for even the claim of oral cancer risk.
Curious The belief that ST is a major risk factor for oral cancer (OC) is a major impediment to a harm reduction strategy. This should not be – even with a high relative risk for OC, the total absolute risk for ST compared to cigarettes would be small (OC has a low absolute risk; smoking causes many diseases at high absolute levels). Nevertheless, the belief in that high relative risk makes it very difficult to promote harm reduction, so it is worth carefully examining the basis of that belief.
The current best evidence (see below) actually provides strong support for the claim that ST causes very little or zero risk for OC. The strongest scientifically legitimate claim that can be made based on the evidence is something like, "ST might cause OC, and a few of the studies on the topic have found a positive association" or perhaps better, "the evidence does not let us rule out the possibility that there is some small risk of OC from ST."
How has this translated into a universal belief that the risk is large and clearly established?
History In 1986, the U.S. Surgeon General released a report declaring that ST causes OC. The year before, the WHO's International Agency for Research on Cancer (IARC) made a similar claim as part of a broader report. These claims were based on case reports, a single toxicology study (many other toxicology studies found no link), a collection of very underpowered epidemiologic studies (some of which provided weak support for the claim, but most of which had major flaws, such as not controlling for smoking and drinking), and a single adequately powered epidemiologic study [Winn, et al., 1981].
It seems unlikely the Surgeon General and IARC reports would have been issued had it not been for the Winn study. Thus, the official declarations that there is a causal link were effectively based on a single study.
Furthermore, an oft repeated position of government and other public health officials is that the matter was settled in 1986, and there is no need to revisit it, implying that our belief should be based on the one study in perpetuity, whatever other evidence might exist.
The Winn study was conducted in the 1970s based on exposure to a somewhat different product (dry snuff from the early- and mid-20th century, rather than modern moist snuff, which contains much lower levels of the chemicals that are believed to be carcinogenic) and a non-representative population (mostly elderly Appalachian women). But most importantly, it was just one study with clear limitations.
Fact Since the publication of the Surgeon General's report, there has been substantially more new epidemiologic research on the subject than existed in 1986, and it tends to support the claim that there is little or no association. In particular, the two high powered and high quality published studies [Lewin et al., 1998; Schildt et al., 1998], which study an exposure much more relevant to current products than Winn, provide strong support for the lack of an association.
Furthermore, several of the studies that purport to show a positive association do so by picking out a non-representative subgroup with an elevated relative risk, and the Winn article exaggerates its findings through the choice of subgroups [see Phillips's poster, "Publication bias in situ," tomorrow].
It is not necessary to debate the accuracy of how the studies reported their data or the relative quality of different studies: Even if the Lewin and Schildt studies did not exist and the studies with positive findings were taken at face value, it might be that most of the evidence would be seen as supporting the claim, but even then the evidence would be too thin to make definitive statements and declare the matter settled.
"Fact" Yet despite the majority of evidence supporting a conclusion that there is very little or no risk, the claim of a large causal association is taken as fact among experts and lay people. The authors of this poster have told hundreds of people, including educated lay people, clinicians, and public health experts, about what the literature actually says about ST and OC. Only one (a practicing dentist) was aware of the truth; the rest found it quite surprising. Indeed, the authors were unaware of the truth until each started doing research in the area.
The "fact" is clearly well established.
It is not difficult to understand why it is so well established. As anyone who follows current national politics knows, if you repeat a claim often enough, regardless of a lack of evidence to support it, people start to take it as fact. The many voices making the claims about ST and OC create an illusion of overwhelming evidence. Government agencies, health and medical organizations, and advocacy groups present the OC risk as fact and typically suggest that the resulting risk makes the health implications of ST use comparable to those of smoking.
Until a year ago the CDC's website contained the blatantly false claim that the answer to "Is smokeless tobacco safer than cigarettes?" is " NO WAY!" [Kozlowski and O'Conner, 2003]. Kozlowski and O'Conner report that (thanks to their challenge), CDC changed this from literally false to merely misleading.
They kept the answer (and its punctuation), but changed the question to "Is smokeless tobacco safe?" Kozlowski and O'Conner go on to report that at the time of their writing (and, as we found in our web search, at the time of this writing), the Substance Abuse and Mental Health Services Administration (part of U.S. DHHS) website still contained the statement "Question: Isn't smokeless tobacco safer to use than cigarettes? Answer: No."
Why do these misleading statements seem to dominate public perception? A reading of the popular, policy, and scientific literature suggests that a few opinion leaders have a pattern of citing only each other, the Surgeon General's report, and Winn, while other organizations cite these sources and the opinion leaders. The illusion of broad and deep evidence, then, all goes back to the findings of a single study and systematically ignores the strong evidence to the contrary.
Study of Information Available on the Internet To examine the extent to which the many available sources of popular information trace their claims to a small number of sources and systematically ignore the larger body of contrary data, we performed a systematic review of popular sources of information. This review looked at websites that implicitly purport to deliver a public service message about the risk of ST and health, in particular OC. While websites do not contain all popularly available
information, we consider this a fairly comprehensive review for several reasons:
Most Americans who would do a proactive search for information on this • topic would probably start with a web search.
• Most organizations who have a stated position on the topic, particularly those actively trying to influence popular opinion, have a web page that reflects their claims. (We could not identify any major players on this topic who do not.) Even if the web search were to miss major sources of popular information, • there is no reason to believe it is unrepresentative of the total body of popular information.
Study Methods We performed a Google search for [tobacco AND cancer AND (smokeless OR snuff OR dip OR spit OR chew OR chewing)], the latter disjunction covering most of the synonyms for "smokeless". We conducted the search on 3 May 2003 and stored the results offline so that we were working from a single permanent list.
The search yielded 124,000 page hits. Google's filter (which eliminates similar pages, including most (but not all) similar multiple page hits at a single domain, as well as many less popular pages that are not good matches to the search terms) narrowed this to 763 hits. Using the filter, an algorithm that we cannot know the exact details of, to narrow the search to a manageable quantity means that we cannot know the exact sampling mechanism.
The exact sampling properties are not important for our purposes, and we have no reason to believe it produced an unrepresentative picture of the information available. (Any unrepresentativeness would have to be quite large to change the results, given how clear they were.) Any unrepresentativeness would not be among the topic results. Since the population includes the top results from the Google search (and far more of them than any sane person would actually read), the sites in our population probably include almost every site someone would find if they did a search. Our mental model of the search process was that we were imitating the information gathering process that a typical educated, nonexpert American might use. Such an individual would likely only look at a few or maybe a few dozen of the first sites coming up on a search engine, so our population is almost certainly a superset of this.
We were interested in public service sites (as implicitly self-defined, without an attempt on our part to judge what is a genuine public service) which state an organization's own opinion about ST and OC. Thus, our protocol eliminated hits that were commercial (selling tobacco products or methods for quitting tobacco); news sites (including newspaper-type sites and clipping services);
search-engine-type sites and others that just provide links; sites from South Asia (because the products dipped there contain other major ingredients have a fundamentally different epidemiology from Western moist snuff); or scientific literature (journals and conference abstracts).
Eliminating these and the double counting from organizations that were duplicated (some domains that showed up more than once on the search list;
local chapters of national organizations; some organizations with multiple websites), left 366 domains.
For each of these, we searched the site (the entire domain, not just the page hit from the search) for statements about the health effects of ST and collected the results. To search within a domain, we used whatever strategies appeared appropriate for that particular site, ranging from looking at every page we could find a link to (for sites that focused on ST and health), to following all relevant text or index links from the page hit from our search (for sites that covered multiple topics), to using internal (site-only) search engines (for large domains with myriad information, such as American Cancer Society (ACS) or NCI). The sites were reviewed as they existed at the time, with the review taking place between 4 May 2003 and yesterday.
We identified any stated sources for those claims, including other organizations and specific documents. We expected to find that, to the extent that sites identified their sources, they would trace back to the aforementioned publications and a few opinion-leading organizations (NCI, ACS), widely cited by other sites.
We also identified any statements that ST was substantially less harmful that smoking or that claims about it causing OC are less than clearly established to estimate the portion of the total that these represent. Additionally, we looked for numbers or other specific scientific claims that could be traced to a specific source. Finally we looked for anything we judged to be an interesting and unexpected pattern of information.
Result: Guilt by association The clearest result of our review is confirmation of the impression that the risk from ST is widely conflated with the risk from cigarettes.
Almost every site had one version or the other of the above "NO WAY!" statements.
Many stated, in one way or another, that ST is as dangerous as (or occasionally even more dangerous than) smoking. This is clearly false.
The vast majority of sites offered a version of the claim that "ST is not a 'safe' alternative to cigarettes." Notwithstanding the lack of compelling evidence about any disease association, this is likely true (almost nothing is perfectly safe, after all). However, when the statement is made without any caveats about how much safer ST really is or about the very thin evidence of any risk from ST all, the message is that ST and cigarettes are comparably dangerous. Making a literally true statement that will inevitably cause the audience to believe something false is functionally and ethically equivalent to making the false statement.