by Carl V Phillips
Lee Johnson has two good posts about teenage vaping, with an emphasis on claims that it is a “gateway” to smoking. The first looks at how to scientifically assess whether there is a gateway effect, and the second is mostly about policies that might discourage teen vaping. They are long, but you should read them if you are interested in really understanding the science and politics of vaping (as opposed to those who read blogs like they use Facebook, seeking a minute of reading to affirm their personal biases; since you are here, presumably that is not you).
Lee interviewed me (via email) for the article. We agreed that I would post the full text of my interview here after his story came out. It turns out to be largely redundant with his article because he did such a good job of capturing what I said. The interview follows (his questions in italics), and after that are a few notes I had on the article (TL;DR version of that: When reading Lee’s article, skip over the quotes from Michael Siegel, because they are mostly wrong).
Oh, and the TL;DR version of Lee’s analysis of the gateway effect is summed up by his points:
[W]e can make a very strong argument that the vaping-to-smoking gateway effect is unlikely to be a significant problem. But, as much as we might like to, we’re not in the position to completely discount the possibility either.
[T]he reason we don’t have the studies to answer the question appears to be because few people really even care about the reality of the gateway effect. People opposed to vaping want something vaguely resembling science to point to in support of their rhetoric, and we vapers want something to point at to shut them up. The science doesn’t need to be robust for that, so why go to the effort?
How would you describe the Surgeon General’s treatment of the evidence when it comes to youth vaping?
That is, the authors pretended to believe basically any bit of (what they would consider) bad news that was asserted by anyone. I say “pretended” and “faux” because, while they are not exactly serious scientists, it seems unlikely the authors are really are so dim as to believe all that. This is reinforced by the fact that they did not seem to believe any of the analysis that disputes those claims.
This is standard and predictable for “official” reports (in this field and most others): A political actor decides what the message is going to be, based on political goals, and then commissions a report that is designed to support the thesis. The authors may have a bit of independence, but are rarely at liberty to contradict the chosen message and, in any case, are selected because they have already demonstrated they endorse the politically-motivated conclusions. This applies equally to the official reports where a different decision was made about what the conclusions will be, like the PHE and RCP reports.
Official reports are heavily used by activists on both sides of the fight because other political actors like the work of political actors, but anyone actually seeking the truth should avoid them. (As an aside, it is, of course, the case that the pro-vaping official reports are more scientifically accurate than the anti-vaping ones, but that is because deciding to support the pro-THR position causes one to be closer to the truth, not the other way around. That is, the decision about what the report would say happened to be closer to scientifically accurate; the decision was almost certainly not made because it was scientifically accurate.)
What do you think about the decision to focus the report on youth vaping and almost completely ignore the issue of vaping’s risks relative to smoking?
Anti-tobacco efforts have not been about the actual risks for longer than e-cigarettes have been a thing. To the anti-tobacco extremists who control most of the discourse on the topic, a low-risk tobacco product is a problem, not an improvement. Tobacco controllers prefer high-risk products because people have a compelling reason to not use them. If the products are low-risk and people know that, they have little reason to avoid them. I have written more about this in my series on why there is anti-THR.
Your paper on the evidence cited for gateway effects gives a clear run-down of the issues with the evidence used to support the gateway hypothesis. My impression from reading it is that none of the currently available evidence is sufficient to answer the question one way or another – is that a fair conclusion to draw?
You actually have to step it back a bit and ask “what question?” If the hypothesis is “there has been one gateway case ever”, let alone “there will be a gateway case eventually”, then the answer is undoubtedly “yes”, and no evidence is really needed. There are enough exposed at-risk individuals (i.e., those who could become a gateway case and who try vaping) that this falls into the “if it is not impossible then it is inevitable” zone.
This is in spite of the fact that we should not expect this gateway effect to happen much, for reasons I go into in the paper. Gateway proponents have tricked everyone, including their critics, into not questioning the underlying premise that it is reasonable to expect vaping to often cause smoking. In reality it seems rather unlikely. While this is a little more plausible than the original “cannabis use causes opioid/cocaine/etc. use” version of the gateway claim, it still makes only a tiny bit more sense than claiming that vaping causes risky sexual behavior.
By contrast, if the hypothesis is “vaping causes everyone who does it to start smoking within a week” then it is pretty easy to disprove with available evidence. If it is “vaping causes many people to become smokers ten years later”, then there it is difficult to imagine any possible evidence that could inform it. I trust you see my point.
So we have to start by specifying what seems to be a reasonable version of the claim We have to do it since the proponents of the claim never actually define what they are claiming. It seems reasonable to go with something like “regular vaping causes 10% of the would-be nonsmokers who do it to become smokers within two years.” Then your conclusion is right: It is theoretically possible to assess this scientifically, but no evidence we actually have should move our prior belief about whether that is true by very much.
What would you say are the biggest issues with the studies we have so far?
The second biggest issue is the one I just alluded to: No one making gateway claims ever specifies what they are actually claiming, in terms of what portion of would-be nonsmoking vapers are caused to be smokers, over what period of time, etc. This is not necessarily a fault with a study itself (unless it was specifically designed to look for a gateway effect, in which case failing to properly define the effect is a fatal flaw in the study). But it is a fatal flaw in any attempt to interpret the study (either as supporting the gateway claim or arguing against it). An analysis that provides good evidence that vaping does not cause 50% of would-be nonsmokers to take up smoking might still support the claim that it causes 10% to do so (or, more likely, provide no useful information about the latter claim).
It is not interesting to claim that there is any gateway effect, just like it is not interesting to claim that an exposure causes any cancers. Epidemiology is a science of measurement, not of detecting the dichotomies that are sometimes interesting in physics or paleontology. A casual reader of epidemiology (a category that includes most epidemiology researchers!) would be forgiven for not realizing this, given how results are often described dichotomously. But magnitude is what matters. Consider the claim “vaping causes cancer, specifically exactly one cancer death ever.” This is more similar — in terms of the science and decision-making — to “vaping is harmless” than it is to “vaping increases cancer risk by 50%”. The same goes for the gateway claim.
On top of that, there is the question of what population is being considered. There are no constants in epidemiology. Any decent Epid 101 class teaches that who, when, and where matter. Vaping might have trivial gateway effects in a population where smoking is relatively normal and easily accessible, but rather larger effects in a hypothetical future populations where vaping is normal and smoking is something very few kids ever consider trying.
The biggest issue is the (often intentional) failure to address the epidemiologic concept of response types. I am working on a paper right now that focuses on this. Basically for dichotomous exposures and outcomes, as with the typical gateway claim, there are four response types: immune (will not have the outcome regardless of the exposure), doomed (will have the outcome regardless), protective (will have the outcome if not exposed, but will not have it if exposed), and causal (will have the outcome if exposed, but not otherwise). A gateway case is properly defined as anyone of the causal type who is exposed.
The unknown proportions of doomed and immune types, in particular the different mix in the exposed and unexposed populations (known as confounding), means that most statistics cited as suggesting a gateway effect offer no such evidence. That is, we can be sure there is an excess of doomed-to-smoke types among vapers (as compared to never-vapers), and there is little doubt that this disparity dwarfs any actual gateway effect. Thus, if response type is not thoroughly measured or controlled for (and it never is), it is simply not possible to detect the gateway effect. Any association between the exposure and the outcome is plausibly explained by confounding; since there is no estimate of how much confounding there is, it is impossible to know if the association represents any causal relationship above and beyond the confounding.
Cutting the other way, if there are any protective types in the population (and there undoubtedly are), that will mask any gateway effect. So if there were no confounding, the association would be a measure of the net of causal types minus protective types. If there is a sufficient proportion of protective types, it would mean the net effect of vaping is to reduce smoking in the particular population. But this does not mean there is not also a gateway effect (i.e., there could still be some exposed causal types).
Anyone who does not understand the previous three paragraphs has little hope of making sense of the evidence about gateway effects. Any real expert on the topic would be able to write those three paragraphs if prompted, something very few of those claiming expertise could do.
What would be required to conclusively answer the question of whether vaping is a gateway to smoking?
Well, “conclusively” might be a bit much to ask, unless the question is one of the extreme versions I have suggested. But as I lay out my previous paper, the best chances of measuring the effects lies in either natural experiments or really good propensity scores.
An example of a natural experiment would be to take advantage of vaping being a fad among teenagers that has (or at least had) very uneven uptake. If someone were to identify high schools with high vaping rates and match them to demographically very similar schools with low rates, and then compare smoking rates, we could estimate the net effect (causal cases minus protective cases). That would not exactly measure the gateway effect because of that problem I just explained. But it would offer a measure of the net of causal minus protective types in the exposed population and this net seems to be what most political debates focus on (though a deeper dive into the moral philosophy suggests that saying protective and causal cases just cancel each other out is not an easily defensible position).
A propensity-score based approach would be to create the best available estimator of whether someone is destined to smoke (asking lots of questions about preferences, behaviors, and exposures) that is independent of vaping exposure, and then see if vaping exposure is associated with smoking among many who otherwise appear destined to not smoke. Since no propensity score is perfect, this would probably require some fairly sophisticated statistics as well as a serious stepwise approach.
I realize that is fairly opaque. Which is the real point: Using propensity measures or tracking natural experiments would require a lot more work than the standard practice of throwing together a half-assed survey about behaviors and a few demographic questions, and administering it to a convenient population. It would require skills that few who study tobacco use have. It would also require serious scientific honesty, because it is trivially easy to pretend to be seeking the truth using methods like that, but really just be dressing up a politically determined conclusion.
Why do you think that research hasn’t been completed?
That has answers at several levels. First, we are talking about public health research here, where there is basically no incentive to do decent science. The rewards (employment, funding, news coverage) are the same for doing a simplistic junk science study as for putting in the effort to do a complicated genuinely truth-seeking study. It is no more difficult to publish junk than it is to publish good work in public health. Indeed, it might be easier. Furthermore, most of the people doing this work do not even have the capability to figure out how to do things right. We are talking about people who have never once in their careers properly piloted a survey, reported a meaningful sensitivity analysis of their model, or consulted an expert before just charging in to concoct an ad hoc research project. They are less likely to figure out an innovative and careful way to answer a tough question than kids doing a high school science project.
Second, the anti-tobacco researchers are worse than average, in terms of scientific skills, even compared to the low standards of public health. Moreover, they know they can get anything published and into the news (which is how they really earn their pay) if it has the right conclusions, regardless of how absurd their analysis is.
Third, for this particular area of research, pretty much no one actually cares about seeking the truth. Anti-tobacco activists want to say there is a gateway effect, to denigrate e-cigarettes. Pro-ecig activists want to say there is no substantial gateway effect, to fight back. (That was being charitable, by the way; the typical claim is that there is no gateway effect, which is absurd.) Both groups cite statistics that, for the most part, do not support their claim at all. This is a rare area of the tobacco wars where critics of tobacco control are almost as unscientific as the tobacco controllers.
I suggested in my recent year-end post that the debate around e-cigarettes (like the debate about other tobacco products) has substantially entered a post-truth era: the rhetoric in the fight is largely unconstrained by science, and the “research” is just designed to produce rhetoric. It is actually not that bad for the technology research, where industry researchers seek the truth to try to figure out how to make or use the products better (in particular, less harmful), and to honestly assess the risk. But the gateway issue is totally post-truth. No one in that particular fight is interested in the truth, just the message. Finding out the truth might be inconvenient.
If someone does become curious about the truth, they probably quickly figure out that the necessary research is not worth the effort. We can be confident there is nonzero gateway effect, but also that the magnitude must be small enough that it will be lost in the noise. That is, given the problems I discussed in the context of response types, is hard to imagine that any plausible research could measure what is undoubtedly (for reasons I explain in my paper) a very small gateway effect precisely enough to distinguish it from zero or from double the actual rate. So it is not really worth doing. Of course, those who wish to argue there is no large gateway effect could do the legitimate research, because it would undoubtedly reinforce their message. But since they are content to cite research that does not actually support their claim, and get away with that just fine, why would they bother?
In the US, it seems that the potential risks of vaping are put front and center, while the very real benefits are rarely mentioned. Do you think that the desire to discourage youth vaping is to blame (at least partially) for this bias?
No doubt pursuit of that goal makes the disinformation somewhat worse. But I suspect it would not be much different even if that were not an issue. Tobacco controllers want to discourage all tobacco product use, regardless of its benefits, and they control almost all the money and media in this space. Even if kids were magically prevented from ever vaping, the disinformation campaigns and attacks on adults’ liberty would still exist.
Is there an inherent conflict between discouraging non-smoking teens from vaping and encouraging adult smokers to switch, or is there a way to do both?
It is an inherent conflict. It is simple economics. If we increase the net quality of the products (better devices, attractive flavors, affordable prices) or the perceived quality by increasing people’s knowledge (knowing it is low-risk), it makes it more attractive for everyone. If we lower the net quality or hide information it is going to make it less attractive for everyone.
Of course, if there were a way to make the products differentially appealing, that could create a wedge between these (though this would not be sufficient to say it should be done). So, for example, if the myth that particular flavors are more appealing to teens were true, that would offer ways to drive a small wedge. But there is not even a small wedge apparent to those of us in the reality-based world, let alone a big one. It turns out that teenagers are remarkably similar to people.
What, if any, actions do you think should be taken to discourage non-smoking teens from vaping?
That, of course, is an ethical question, not a scientific one. Every political discussion is predicated on the assumption they should be discouraged from vaping. This is far from self-evident, but no serious analysis of the question ever takes place. The most obvious argument against the assumption is that there are enough protective types among teenagers that the resulting reduction in smoking initiation will outweigh any downsides. A bolder, but still prima facie legitimate, argument is that it is not obvious there are sufficient downsides of the behavior to warrant any actions, at least not beyond just banning sales to minors. Bolder still is the argument that the benefits of nicotine for a subset of teenagers are quite large, and they should be using it (note the explosion in the use of nasty prescription stimulants over the period where teenage smoking decreased). Oh, and your question implies the assumption that smoking teens should not be discouraged, which also does not go without saying. Obviously a lot of people seem to feel otherwise (though they rarely even attempt to articulate a normative argument for their position).
My point here is not to endorse or dispute any of those possible positions, but to note that there is no serious discussion of what the basis for that “should” should even be.
One thing that is clear is that, at worst, teenage vaping is just one of hundreds of little things that kids might think are a good idea but their elders might realize are not. It is hard to imagine any knowledgeable parents being nearly as worried about their kids vaping as about them driving dangerously, getting pressured into ill-advised sex, drinking, not trying in school, smoking weed, not exercising, and numerous other hazards of adolescence. Actual vaping (as opposed to just trying an e-cigarette a few times) is also far less common that those more serious hazards. The breathless swooning about teenage vaping, which is responsible for the concocting of gateway claims, is all a product of a small group of anti-tobacco fanatics who consider it a moral affront that anyone (adult or teenager) uses the products.
It is no different from those who take moral offense at gay people or women not covering their heads in public. The “problem” is considered much of a problem only because some people chose to declare it a problem. Unfortunately, modern Western liberalism has a habit of insisting we “respect” people’s choices, even when those choices are to try to restrict others’ behavior. So the fanatics are given the benefit of the doubt instead of being told (like you would tell any child who is trying to impose his preferences on others) “your dislike of this is your problem, not ours.” Vaping advocates would be wise to stop conceding almost all the fanatics’ claims and assumptions and then try to fight within the tiny space they have left themselves.
My notes on Lee’s article:
The major problem with the article was presenting the analysis from Michael Siegel as if it were scientifically valid. Siegel is a dedicated and energetic supporter of vaping (though not of THR), and very likely offers value to the cause, but his attempts at scientific analysis are very frequently out-and-out wrong. He got his start writing about tobacco control’s environmental tobacco smoke myths and calling out their whole-cloth claims, and he is on solid ground when he sticks to that. Now I usually just avoid reading him to avoid the urge to grade his work, but this forced it in front of me.
I know this is harsh, but after a decade of trying to help Siegel get things right (and to not write about what he does not understand, and to spend more time reading rather than writing so he could see what he does not understand) I have reached the conclusion that it is not going to happen. Certainly nothing short of saying this will get his attention. (I kid myself with that: he is not likely to read this. If he read what I wrote he would avoid many of his errors.) So the only reasonable course of action seems to be to discourage others from thinking he is a good source of scientific analysis, whatever other contributions he might make. What he writes sounds sciencey, of course, which is why people who want to believe his conclusions are often fooled into accepting his analysis, and why it offers useful propaganda value. It is the same phenomenon with Glantz and the partisans who want to believe him. (That comparison is not random, by the way: Siegel apprenticed his science (such as it is) with Glantz.) But the present exercise was truth-seeking, not propaganda.
Most of Siegel’s errors in Lee’s piece should be immediately evident to anyone who reads my analysis. For example, he claims (yet again) that “smoking rates have continued to plummet. This is not consistent with the hypothesis that vaping is a gateway to smoking.” If you read my paper, this post, or just Lee’s excerpts from my interview, it will be obvious why that is wrong. (In fairness to Lee, he was following the accepted role of the reporter, recounting the results of multiple interviews despite the fact one shows the other is wrong.)
Oh, and the first quote from Siegel is him banging on, yet again, about how e-cigarettes are Not Tobacco!!!, claiming this is the main reason why the Surgeon General’s report on teen vaping was wrong. Even setting aside the fact that he is clearly wrong to call this wrong, his obsession with this argument has reached the point of seeming deranged. (There are lots of ways to categorize things, based on different criteria for different purposes. A choice of how to categorize, therefore, cannot be wrong unless it is absurd, like calling e-cigarettes a kind of suspension bridge. Obviously it is not absurd to categorize e-cigarettes based on their role in society and active ingredient, even ignoring the fact that they do, in fact, contain part of the tobacco plant.) At best, this comes across like a six-year-old who just learned that tomatoes are a fruit when categorized anatomically, and smugly telling his parents he cannot finish his vegetables because he only has fruit left.
In fairness, this is not childish smugness and probably not derangement, but rather it is politics-as-therapy (a phenomenon that rarely produces good outcomes). That is, his self-image is as a tobacco controller, and so he cannot come to grips with the fact that his primary public identity is as a cheerleader for something that most people (quite reasonably) call a tobacco product. So he wants to change reality to deal with his personal conflict.
Anyway, moving on…. One point that I did not make in my interview (though it is in my paper), is the observation that being a never-smoker is not a necessary condition for being a gateway case, as is implied in the article and frequently implied or stated elsewhere. Someone who smoked a few cigarettes, or even a lot, might still be a causal type — someone who is destined to be a nonsmoker in the absence of vaping, but destined to become a smoker if she takes up vaping. Indeed, it is not even clear that someone who has already smoked some is less likely to be a gateway case than someone who has never smoked. Yes, the previous smoking suggests she is more likely to have been doomed to smoke. But it also suggests she is more likely to be on the cusp of being a smoker or becoming abstinent, increasing the chance that an exposure like vaping is enough to keep her from becoming abstinent.
Thus, all the analyses that assume only never-smokers can be gateway cases — and thus insist there cannot be many gateway cases because there are relatively few never-smoking vapers — are wrong. Indeed (and this may be the first time I have published this observation), given that there are so many more ever-smoking teenagers who vape or trial an e-cigarette, ignoring the ever-smokers means ignoring most of the at-risk individuals (i.e., potential gateway cases). Someone seeking to deny the gateway claim might gerrymander their definition of “gateway” (which is not a well-defined scientific term) to require that the vaping cause someone to smoke their first cigarette, but this is clearly not what is meant when concern is expressed about the claim. Of course, it is worth reiterating that there is no reason to believe there are many gateway cases at all, but if the goal is estimate the effect then obviously we should not be looking only at the pristine tip of the iceberg.
Finally, a point about that question of whether there is an inherent conflict between encouraging adult smokers to switch and discouraging nonsmoking teenagers from vaping (setting aside the debatable — but never actually analyzed — question of whether there even should be aggressive interventions to do the latter). My point is that apart from the trivial expedient of forbidding sales to minors, along with the horrible option of making possession or use status offenses, there is no apparent action that affects teenagers that does not affect adults in the same direction (though the magnitude may differ). Lee characterized the answer to the question as “a matter of perspective”, but sorry, no — this is a matter of science.
There is no theoretical reason that there could not be some other ways to affect teenagers without affecting adult smokers, but the reality is that none exist. Consider the options that are addressed in the article. After declaring there is no inherent conflict, Siegel offers only sales bans and “marketing restrictions” as methods. It is standard tobacco control rhetoric to claim that there is a well-defined activity called “marketing to children”, that can be restricted. But to anyone in the real world, this is obvious nonsense. Marketing to 35-year-olds cannot be done without also marketing to 16-year-olds. As I noted, teenagers are a lot like people. Any restriction on “marketing to children” beyond banning advertising on the Disney Junior network will also constrain marketing to adult smokers. In fact, I picked that example because Disney Junior is probably the network I watch the most, so even that restriction would reduce outreach to adults. Again, there might be differences in magnitude, but not in sign.
Of course, the savvy tobacco controllers probably realize this is nonsense (though their useful idiots believe it) and know they are just using it as an excuse to ban all marketing, signage, and visibility, and to thereby create stigma. Needless to say, eliminating all marketing that children see (“verify you are 21 by entering your credit card information and scanning your driver’s licence so we can show you our ad”) is eliminating marketing.
Another policy addressed in the article that would not have the claimed differential effect is punitive taxation. The myth here (stated in the article) is that adult smokers are less price sensitive than teen trialers. But this is largely nonsense also. There are exceptions — one appears in the article — of adult smokers who will pay a lot extra for something that helps them quit smoking. Some buy ridiculously priced NRT, after all. But for everyone else, the price of e-cigarettes is a huge potential incentive or disincentive. The same concept applies to punitive taxes on cigarettes: It is a myth that teenagers care more about price. The reality is that for the vast majority of families with Western levels of wealth, kids are more willing and able to spend freely on such things than their parents.
Adults with $20 in their pocket think “I can take the kids for pizza on Friday” or “I can pay off some of my credit card debt”, while teenagers with $20 think “what entertainment can I buy for this? Today.” This is even more true for raising the price of e-cigarettes than it is for discouraging smoking with punitive taxes. An adult smoker assessing the price difference thinks “I will pay (or save) this much from switching, which adds up to X every year”, while a teenager wanting to try something today will not even notice a dollar difference in the purchase price. The difference in response to price that the tobacco controllers think is adult-vs.-teenager is actually “reasonably wealthy person” vs. “impoverished person”. Of course, public health people assume for themselves an in loco parentis attitude toward poor people, treating them like teenagers (part of the reason such people voted to let Paul Ryan take away their medical care, because they are so angry at the paternalistic elites — but that is another story). But that does not actually mean that the people they treat like children are the actual children.
The final example given is the standard myth about flavors. If there really were a group of flavors whose existence causes some teenagers to vape (a dubious claim in the first place) but that causes no adult smokers to switch (or stay switched), they could be banned without affecting incentives for adult smokers. If there really were flavors that had more effect on teenager choices, then they could be banned to have more effect on teenagers than on adults, but there would still be a reduction in appeal to some adults. (Side note: Notice that the conditions are the very strong claim that the flavors’ existence causes vaping or switching, not merely that they have differential appeal.) Again, those of us in the reality based world know there are no such flavors apparent. But the real point is my original one: Anything that lowers the quality will affect the appeal to both teenagers and adults, in the same direction, even if it were one of these mythical reductions in quality that affects teenagers a lot more than adults.
Ok, done. This is already too long, so no conclusion statements.