More nonsense about gateway effects – this time from ecig supporters

by Carl V Phillips

It seems that most every researcher or pundit making claims about gateway effects — that e-cigarette use causes some people to then become smokers — has no clue about what evidence would support or contradict such a claim.  It is a truly amazing and sad commentary on what passes for scientific thinking in this realm.

I have already explained at length how Glantz et al.’s claims about having found evidence of a gateway effect are fatally flawed.  But they are actually one step better than two recent claims by e-cigarette supporters that there is no gateway effect.  Glantz basically made a single observation about a statistical correlation that you would, indeed, expect to see if the gateway claim were true.  But you would also expect to see that correlation if either of two other things were true, one of which (confounding by common cause) certainly is true and the other of which (people are employing THR) is far more plausible and better supported than the gateway claim.  Basically the logic was this:  “If it is Monday at noon, there should be light coming in my window. There is light coming in my window. Therefore it is Monday at noon.”  There are obviously many more states of the world where the observation is true and the conclusion is not, so this is terrible reasoning.  I remember learning that in grade school; apparently Glantz, the faculty at UCSF, and the editors and reviewers of the junk journals he writes for missed that day of class in third grade and never took any science courses in college.

But that is not the worst reasoning one could engage in.  At least observing the sun makes it a little more likely it is Monday at noon, since it allows you to rule out many times of the week, and therefore you have increased the probability that it is Monday at noon.  Of course, you have increased the probability that it is midafternoon on Wednesday just as much.  But the observations in this study, a survey of European e-cigarette users, do not even do that much.  That has not stopped e-cigarette touters from claiming that the study shows there is no gateway effect, a claim which has made the rounds in social media.

It is rather baffling.  The authors make no such suggestion in the article or their press release.  There is simply no support for the claim to be found there.  The interpretation seems to trace to this crap article in New Scientist, which makes the claim in the text, or this even worse one in International Business Times which put the claim in the headline.  Naturally, clueless people who know nothing about the topic believe whatever they read, but some people who are repeating this silly claim should know better.  (Here’s a useful little epistemic hint for future reference:  When a news story says a study was published in the non-existent Journal of Tobacco Studies, it is probably not a very good source of information.)

The “reasoning” seems to be that because only 1% of never-smokers in the survey had tried an e-cigarette, e-cigarettes must not be a gateway.  Huh?  By that logic, smoke-free tobacco products do not harm fetuses because less than 1% of pregnant women use them.  (Note: it remains unknown whether or not there is such harm.)  All the observation shows is that if there is a gateway effect, its possible absolute magnitude (so far) is capped by the fact that only a small portion of the at-risk[*] population has had the exposure.  It tells us nothing about whether there is any such effect.

[*Side note 1:  For those who may be confused due to the fact that this phrase is used incorrectly more often than correctly, "at risk" means "anyone who could become a case".  So "at risk of being a gateway case" means "not a smoker" (notice that this means that ex-smokers are at risk of becoming gateway cases, not just never smokers, so leaving them out of the "reasoning" was another error).  Thus, Glantz is at risk of becoming a gateway case, whereas someone who quit smoking using e-cigarettes is not.  The common misuse of the term is incorrectly substituting it for the intended claim that someone is at high risk of becoming a case.]

[Side note 2: I feel a personal need to point out that part of the touting of the study by pro-e-cigarette pundits is that it came out of Harvard.  I beg to differ. The authors all have some affiliation with Harvard School of Public Health, not the real Harvard.  The faculty at HSPH were serious researchers 20 years ago, and there are still a few good people from past eras (though not in the anti-tobacco and other nanny-state units), but is now it is pretty much just a school of public health, with all that implies.  Or as I put it previously,

The study, from a research group calling itself “Harvard School of Public Health”, was widely cited in the popular press.  (Note that while this organization does not seem to be related to the prestigious research institution called “Harvard University”, the allegations in the literature that it is actually a lobbying front group for shadowy government agencies and industry have not been proven, and so cannot be considered a reason to doubt their research.

Resuming….

The other common error about what constitutes evidence of no gateway, most recently committed by Siegel, is suggesting that declining smoking rates, in the context of increasing e-cigarette usage, must mean there is no gateway.  It is a fundamentally different error, though the implications are similar.  The error is a simple misunderstanding of magnitudes, statistics, and the passage of time. Even though e-cigarette use is increasing, its absolute level is still very small among those at risk of becoming gateway cases.  If every single at-risk e-cigarette user is a gateway case, the impact would not be visible in population smoking statistics.  The total number of at-risk e-cigarette users constitutes a smaller portion of the population than the statistical error in smoking prevalence surveys, to say nothing of what it would take to sort it out of the real changes in smoking rates over time.  Moreover, since “destined to start smoking due to her e-cigarette use” is not the same as “has started smoking”, the gateway cases would mostly not even be observable yet.

On top of that, e-cigarettes are causing some people to quit smoking.  So even if we had a perfect measure of smoking prevalence, and the only thing changing it over time was the effects of e-cigarettes, we would still see a decline in smoking rates even if there were a gateway effect (unless, of course, it were already so big in absolute magnitude that it exceeded the THR effect).  Thus, the “reasoning” here is an epic fail for at least three different reasons.

The reason that the implications of the two different errors described above are similar is that their proponents are really saying “the impact of the gateway effect is apparently small so far, so don’t worry about it.”  This is not a very effective argument.  There are very good reasons to believe there is not and never will be a gateway effect, or at least extremely little of it, but neither of the above are among them.  Effectively, those erroneous arguments invite the (valid) retort, “sure, we are not seeing it yet, but the way things are going, it will show up, and so we need to stop e-cigarettes before it does.”

The conclusion that there is no reason to expect a gateway effect is correct.  But making bad arguments in favor of a valid conclusion almost always sets up your opponent with a good argument in response, and that is certainly the case here. Moreover, if we are going to start trafficking in bad arguments, we are going to encourage and validate lying.  The other side will inevitably win the argument under those rules of engagement, because they are waaaay better at it.

4 responses to “More nonsense about gateway effects – this time from ecig supporters

  1. Pingback: More Nonsense About Gateway Effects – This Time From Ecig Supporters | The Spinfuel Network

  2. The claim isn’t as erroneous as you say here Carl. To be a gateway the product needs to convert large numbers of people in a direction. The fact that so far their is no evidence at all of any gateway effect is substantially the same as saying that ecigs are not a gateway. I’m not sure what your point is, that stats dont tell the full story? that we need to be carefully when using stats to support claims? Stats are never more than a snapshot of a small portion, a sampling, they can indicate trends, nothing more. So far they show no trend towards this gateway effect. Could they develop in this direction? probably, if marketed and designed to do so.
    Should we just let the gateway claim stand without showing it lack of evidential suport?
    This is now a PR war, if harm reduction is going to be achieved it will be achieved first in the PR area.

    • Carl V Phillips

      Um, no. To be a gateway to smoking, it needs to cause at least one person who would not have otherwise been a smoker to be a smoker. If you want to put some adjectives or qualifiers with it (“a common gateway” “a major gateway”) then magnitude starts to matter, but such a weaker claim is not what anyone I was referring to was making.

      And, no, my point is not about statistics but about scientific reasoning. There was little of it in evidence in any of the claims I quoted. Statistics never speak for themselves, so whoever interprets them need to think.

      There is a PR war, and it is one where we will fare much worse in if we are shown to be dishonest. However, it turns out that the narrow question of a gateway is of far more interest to the US FDA than the angry mobs, and they are very unlikely to be swayed by propaganda.

      • Carl V Phillips

        I will add some substance to the dealing the adjectives. If your claim is “there may well a gateway effect but its effects are not very widespread because (so far!) relatively few at-risk individuals have been exposed”, then the exposure statistics reported from the European speak to the point. (On the other hand, someone might reply that 1% of the population possibly being enticed into smoking is a big number, not a small one.) If the claim is “there may well be a gateway effect, but it is smaller than the THR effect”, then a time series of statistics about smoking prevalence *could* address the point (I say “could” because in the particular example, they do not because of the other matters I talked about). But in neither of these cases does the evidence support the claim “there is no gateway effect”.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s