draft Regulatory Science White Paper #2: The “gateway effect”

by Carl V Phillips

You will recall that a month ago, we posted our first draft white paper, on e-cigarette use by children.  We received, in various forms, the very useful suggest that our original title for the series (position papers or position statements) was misleading and did not do them justice.  Thus, we have renamed the series “Regulatory Science” and just call them white papers.

The term “regulatory science” is a thing right now, and is the language being used by the FDA and others.  It is not clear exactly what it means (and, indeed, it is my personal feeling that almost anytime someone precedes “science” with an adjective like that — pretty much any adjective that does not demarcate a specific area of science like “biological” — the adjective might as well be “junk”). We have chosen to interpret the term as something like: “scientific analysis that is specifically geared toward translating the scientific knowledge base into decision-relevant information”.   That is basically what I taught as a professor, and that turns out to be almost completely absent (perhaps completely now that I am retired from academia ;-) from public health sciences programs.

Anyway, the first white paper has now been reviewed — by you and recruited peer reviewers — and will be finalized and posted soon.  The second and third are done.  The second follows this paragraph.  As before, we are interested in any comments from anyone — peer review type comments especially, but we will certainly consider anything else that might strike you also.  CASAA members are invited to weigh in on the “our position” bit, since you are part of “our” (want to comment on that and are not a member? join – it’s free!).

Consumer Advocates for Smoke-free Alternatives Association

DRAFT Regulatory Science Series #2:  The “gateway effect”

draft – not yet peer reviewed

 Soundbite version

Those who oppose the promotion of tobacco harm reduction frequently claim that low-risk tobacco products will cause people to start smoking, the so-called “gateway effect.”  This claim has been heavily employed in anti-e-cigarette disinformation campaigns.  However, there is absolutely no evidence that such causation occurs.  Data that is cited in support of gateway claims about smokeless tobacco merely demonstrates the obvious point: people who use one tobacco product are more likely to use other tobacco products than those who avoid tobacco entirely.  Moreover, a moment’s thought reveals that such an effect is unlikely:  Why would someone who chose abstinence over smoking, upon discovering that they prefer a third alternative to either of those, be caused to make a change to their least-preferred option, smoking?  Asking that question, rather than making affirmative arguments, is probably the best response to anyone who is claiming that there is a gateway effect.

 Finally, even if someone speculates that there are some gateway cases, these are still clearly dwarfed by the number of would-be smokers who choose a low-risk alternative instead.  Thus the low-risk products still provide large net population health benefits.

 Introduction

A typical fallback position for those who want to condemn a drug or behavior, but cannot credibly argue that it is harmful in itself, is to claim that it is a “gateway” to genuinely harmful behaviors.  In the case of tobacco harm reduction (THR) products, the claim is that they cause some people to smoke.  In the early 2000s, anti-THR activists frequently claimed that smokeless tobacco (ST) was a gateway to smoking, though they seem to have largely dropped the argument because it proved much more effective to just mislead people into believing that ST causes substantial risk in itself.  But the rhetoric has recently been revived as an attack on e-cigarettes because attempts to convince the public they are high-risk are falling flat.

A simple analysis reveals that the gateway claim is fatally flawed.  It must be realized that the gateway claim is that the THR product is causing smoking – that is, smoking would not have occurred if the other product did not exist or were not tried by the user.  Mere precedence – e.g., someone puffing an e-cigarette before they ever tried a cigarette and later becoming a smoker – does not suggest there is causation; the smoking habit probably would have happened had the other product not existed.  When the gateway claim is used in anti-THR rhetoric, there is (presumably intentional) conflation of causation and mere precedence, but the implication is always about causation, and thus “gateway” should be interpreted in terms of causation.

While it is theoretically possible for low-risk tobacco product use to cause smoking, there is no empirical evidence that it ever occurs, and indeed anyone who thinks through the scientific analysis will realize that it is vanishingly unlikely.

Empirical evidence

Simply put, there is no empirical evidence of a gateway from low-risk products to smoking.  What was cited as a evidence of this for ST merely showed that people who were inclined to use tobacco liked to use tobacco.  That is, the conclusions were based entirely on the (clearly true and unsurprising) observation that people who used one tobacco product were more likely to become users of another tobacco product than were lifelong never-users (example).  So, in particular, users of ST were more likely to become smokers than were never-users of tobacco.  But this told us nothing about whether they would have been smokers had they never used ST.

In theory, evidence like this could show whether there is a gateway effect by controlling for propensity to smoke.  That is, if many former ST users became smokers even though they were not the type of people who normally become smokers, this would suggest a gateway effect (and if it did not occur, it would be empirical evidence that there was no such measurable effect).  None of the researchers or activists making gateway claims seemed to have even seriously attempted this, nor even to have understood that it was necessary.  One response reanalyzed the data from one of the naive pro-gateway claims that merely observed that those who used one tobacco product were more likely to use another; using with controls for propensity, it showed that most of the claimed association disappeared.  Another study used propensity based matching and did not detect a causal relationship between ST use and later smoking.  However, it is difficult to imagine creating a sufficiently accurate “propensity to smoke” score that could detect the signal amidst the noise of the obvious correlation between liking one tobacco product and liking the other, and the modeling would offer the analyst so many degrees of freedom that it would be easy to bias the results for political purposes.  Thus, there is basically no possibility that this line of research can distinguish between zero gateway effect and a small gateway effect, though it can clearly rule out a large effect.

(Aside: The mere fact that switching from ST to smoking was about as common as the other way around in historical U.S. data is disturbing from a public health perspective, even though it does not support the gateway claim.  This pattern is easily explained by the fact that most Americans were (and largely still are) victims of the anti-THR disinformation campaign that convinced them that smokeless is just as high risk as smoking, so they might as well smoke.  The good news is that willingness to switch products means that THR is likely once someone fully understands and internalizes the truth.)

There have been several studies that provide affirmative observations that are generally contrary to gateway claims about ST.  They found that few cases of ST use were even candidates for the gateway claim, that people who initiated tobacco use with ST were less likely to smoke than those who initiated with cigarettes, and that in Sweden (a population where ST use is common) there was far more switching from smoking to ST than the other way around.

The evidence from Sweden shows that if there is a gateway effect, it is utterly dwarfed by the THR effect (that is, even if the gateway affects a few people, far more would-be smokers choose ST, so the net effects are overwhelmingly positive).  Sweden has a high and long-standing use of THR products (in particular, snus) and has by far the lowest smoking prevalence of any country where smoking was ever popular.  Moreover, the comparison of men (whose use of snus is more prevalent and occurred earlier in time) and women shows an inverse association: male smoking rates dropped markedly as they took up snus while female smoking rates only followed the steady downward trend observed elsewhere (see, e.g., reference).  If snus were causing many users to smoke, this would not be the case.  (No analogous observation is possible about e-cigarettes because they are not sufficiently popular in any population that a gateway effect would impact population statistics.)

Neither the above-cited studies, nor the Swedish data, nor any conceivable data can rule out that there are some gateway cases.  A few cases would simply not be detectable, and thus cannot be ruled out.  But they do show that there are either few or none.

In the current discussion of e-cigarettes, there are not even any observations about people trying e-cigarettes first and becoming smokers, let alone support for the claim that it is a gateway.  The claims about e-cigarettes are just made up from whole cloth (example).

There are no apparent studies of possible gateway effects at the individual level (for example, testimonials of smokers who report that they eschewed smoking until they became a user of a low-risk alternative and then found themselves drawn to smoking).

Scientific analysis

The empirical evidence fails to demonstrate any gateway effect and shows that it is small if it exists.  The argument that it is non-existent, or close do that, can be found in the logic.  It is absurd to even think that low-risk tobacco/nicotine products would cause smoking.  Consider the combination of traits and choices that would have to be true of someone for someone to be a “gateway” case:

1) He would not smoke if not exposed to the low-risk alternative (otherwise smoking would have occurred anyway, so the other exposure is not causing it).  That is, abstinence is preferred to smoking, perhaps because of the health risks or perhaps just a lack of taste for tobacco use.

2) He chooses to use the low-risk product.  That is, the low-risk product is preferred to abstinence, so he has a preference for tobacco use so long as the risk is low.

3) Having adopted the low-risk product, he then prefers smoking and switches to it.

 A moment’s thought reveals how absurd that combination is.  The low-risk product is preferred to abstinence which is preferred to smoking.  But using the low-risk product does not cause the user to decide he does not like it (and go back to abstinence) or to become a dedicated user of that product.  For some mysterious reason, it somehow causes him to switch to his least-preferred alternative.  Why would this be?  His preference pattern probably reflects the fact that he is concerned about the health effects, even though he likes to use tobacco/nicotine (though it might just be that he just does not like smoke).  Why, exactly, would using the low-risk product cause him to forget about this?

The reason the absurdity of the claim is not immediately obvious to all observers is that tobacco control has convinced many people of their “demonic possession” theory of tobacco use.  That is, contrary to ample evidence, and in contrast with all other consumer choices, they imply that people do not choose to use tobacco or choose among tobacco products because of preferences.  It just sort of happens.  Since people’s actions, according to this story, have nothing to do with preferences or volitions, they must be controlled by demons.  Since demons can behave in completely arbitrary ways, any behavior pattern is thus possible, even when it is clearly absurd if based on an analysis of preferences.

Of course, it might be that someone could craft a non-absurd story of how the gateway works.  But to our knowledge, none of those who assert there is such an effect have ever done so.  We suspect such a story would involve an appeal to a some concept of addiction, but invoking that concept would not explain why an increasing desire for the low-risk product would result in it having less appeal than the previously dispreferred product; some plausible reason for that would be necessary.

The failure of THR opponents to try to produce a mechanism for the gateway effect may be because one part of any plausible explanation does not work in their favor.  When people do not know that smoke-free products pose orders-of-magnitude lower risk than smoking, the barrier for switching in the unhealthy direction is much lower. The aforementioned studies about ST use and smoking in the USA, a population that mistakenly believed the two products posed similar risks, show a similar pattern of switching in both directions.  While it is still the case that most or even all would have still smoked in the absence of ST, some of the switching in the unhealthful direction might have been causal.  Ignorance of the comparative risk alone still does not explain why someone who would have never smoked would be caused to smoke, but it frees that story from one enormous hurdle:  The user is still switching to what was once his least-preferred option, but he does not know that this entails an enormous increase in health risks because he has been tricked into believing that, from the health perspective, he might as well smoke.

In short, if there are gateway cases, much of the explanation for them is the mistaken belief that the low-risk alternative is actually high risk.  The fact that the same activists who purport to worry about a gateway effect also spread such disinformation suggests that they are not genuinely worried about the gateway effect.  It is also worth noting that the claims refer only to non-pharmaceutical tobacco/nicotine products, though it is equally plausible (which is to say, equally implausible) that nicotine gums, lozenges, and patches could be a gateway to smoking.  Yet those who claim to fear the gateway effect usually support aggressive touting of these products.

It follows from this that accurate communication about low-risk alternatives cannot create a gateway effect.  Even if availability of those product could create some gateway effect (contrary to both the empirical evidence and logic that suggests otherwise), accurate communication about the risks would tend to reduce this.  Thus, it is nonsense to even suggest that informing people about the low risk of available products could cause a gateway effect.

Our position

If anyone who would not have otherwise smoked is caused to smoke, that is unfortunate.  If THR were causing this to happen to a substantial degree, it might be a cause for concern even though the net population effects would still be beneficial.  Ethically, it is debatable whether population benefits always trump uncompensated individual costs.  But given that it is difficult to imagine why it would happen at all, and there is nothing to suggest it is happening to a measurable degree, this abstract debate is moot.

It is clear from their other behaviors that those who make claims about a gateway effect are not genuinely concerned, but are merely following a pattern of making every anti-THR claim they can think of.  Their willingness to mislead about comparative risks, and to express no concern about NRT availability and marketing, make clear that they do not really believe what they are saying.

If there is any gateway effect, the best way to reduce it is to clearly communicate to low-risk product users how much higher risk smoking is.  There is no plausible way such communication could increase the effect.

Based on everything we know, this is simply a non-issue.  Unless those who claim to worry about it are able to produce some argument or evidence, rather than just making vague assertions, there is nothing more that can be said about it.