How Would We Know if Vaping is a Gateway to Smoking?
One of the most commonly-repeated claims about vaping is that e-juice and e-cigarettes are a gateway to smoking. In support of these claims, proponents usually cite the CDC National Youth Tobacco Surveys, which merely show an increase in vaping in middle and high school students from 2011 to 2014, or other studies using that data to assess gateway claims specifically. The problem is that even these gateway-specific studies generally look for an association between vaping and smoking, with no information as to whether the smoking or the vaping came first.
If vaping is a gateway to smoking, then it would be plausible that e-cigarettes could have a negative net effect on public health overall. Making some genuine effort to determine whether this is happening is therefore an important priority, even if the available evidence doesn’t suggest any cause for concern.
CASAA has recently opened a research fund for this very purpose, aiming to raise $7,500 (and contributing an equal amount itself) to conduct the study. The new research proposal and the paper by Carl V. Phillips that it’s based on revolve around a crucial question: how would we even know if vaping was a gateway to smoking?
CASAA’s New Research Proposal
The research proposed by CASAA has two parts. First, their goal is to conduct a systematic review of all the evidence that is claimed to show a gateway effect from tobacco products such as e-cigarettes and snus. The analysis of this evidence will use the framework Carl V. Phillips created in his paper, enabling the authors to sort the good evidence from the inconclusive studies or junk science.
The second part is a little different, but would involve working back from the end-result claims that there is some sort of gateway effect and identifying the source of those claims. When the claim is traced back to a study – if that’s even possible – then the study will be evaluated as above. Undoubtedly, this portion of the study will involve identifying numerous statements made which have no basis in fact.
You can donate to the project through CASAA’s Research Fund page, if you’re interested.
Carl V. Phillips’ Paper: Why the Cited Evidence Doesn’t Support a Gateway Effect for Vaping and THR
The paper this is all based on is a little dry to read (which you can do in full for free, if you’re interested), but raises numerous excellent points that can help anybody investigate a gateway claim for themselves.
The paper opens by pointing out that claiming that a low-risk product (like e-cigarettes or snus, but Phillips also cites marijuana as an example) is a “gateway” to a higher-risk one is a useful rhetorical tactic for those hoping to oppose that low-risk behavior. Simply put: even if you don’t have a reason for opposing the low-risk behavior, you can claim that it leads to a high-risk one and oppose it on that basis.
The specific points raised through the paper are well-summarized by a checklist for evaluating gateway claims near the end, so using these as a guide, here’s how to tell if vaping, snus or pretty much anything else can be legitimately called a “gateway” based on the evidence presented:
1 – Is the Research Science or Propaganda?
The first point refers to two basic requirements of any scientific investigation: a definition of what you’re testing and a clear approach to testing for it.
The definition of a gateway seems simple enough – the use of a low-risk product causing the risk of a high risk product – but the web of cause and effect is really more complicated than that. Imagine that e-cigarettes really did cause people to smoke – how do you know somebody who vapes and then smokes wouldn’t have ended up smoking anyway? They might have been heading there anyway, so would it be fair to say vaping caused it just because that’s the path the individual happened to take? Really, a “gateway” must be defined as a product that causes use of another when the end result wouldn’t have happened otherwise.
The second part is pretty complicated – and much of the paper addresses what sort of data would be needed – but effectively boils down to turning the claim into a clear, testable hypothesis.
This would be aided by two other points. Firstly, why would a gateway even exist? If somebody decided to not smoke – probably due to the health risks – and wouldn’t have consumed nicotine if e-cigarettes hadn’t become available, why would that person then electively take the additional risk? There is a potential explanation in the more efficient nicotine delivery from smoking (and the additional chemicals that heighten its addictive properties), but without such a theory to work from, knowing what to look for is challenging.
Secondly, it would be very useful if there was a testimonial of such a progression to draw upon for the same reason. But so far, no such testimonial has surfaced: could it be that it hasn’t happened often enough for anybody to find one?
2 – Is it a Gateway or Harm Reduction?
Many pieces of research cited to support the gateway effect only find an association: for example, smokers are more likely to have vaped than non-smokers. Anybody could identify two possible reasons for that – either the tobacco e-juice caused the smoking or the smoking caused the vaping – but there are others (e.g. neither caused the other). If there’s no effort to determine which is occurring, then it’s a sign you’re not reading an unbiased analysis of the evidence.
Phillips runs through how this could be done with a Stanton Glantz study, looking at e-cigarette use by how many years participants had been smoking, showing that longer-term smokers (who started before vaping was popular) were more likely to have vaped than shorter-term ones, suggesting (but not proving) that it was harm reduction, not a gateway.
3 – Are Other Factors (Sensibly) Controlled for?
Of course, there are lots of things that can impact somebody’s chance of becoming a smoker, so researchers try to control for “confounders” in order to “separate the signal from the noise,” so to speak. For example, having friends who smoke is linked to higher odds of smoking yourself, so if somebody both had friends who smoked and vaped, it could be the former that caused the smoking (or at least contributed) and not solely vaping, so you need to take its effect into account. The only problem is that this isn’t always done particularly well. In some cases, he argues, they basically just throw whatever variables they have into the mix and hope for the best. He uses race as an example of one such attempt to control for important variables: there’s no real reason to think it would play a significant role, but it’s often controlled for anyway.
An example of a good set of factors to control for would be based on propensity for smoking, looking at things such as their propensity for risk taking, the area they grew up in and the career path they’re on. If a sensible set of demographic and personality traits is chosen, the pre-existing risk of later smoking would be properly controlled for and the researchers could then focus on the impact of using the low-risk product alone. When this was done for smokeless tobacco, the researchers found no evidence of a gateway.
4 – Would the Data Make Sense if the Hypothesis Was False?
This is a simple one: if you’re looking at some evidence which “supports” the gateway model, it’s hardly compelling if the evidence is also explained by the opposite hypothesis. A perfect example of this is the numerous studies finding a correlation between vaping and smoking: the findings are often equally well explained if smokers are vaping to reduce their risk, so it can’t really be seen as supporting a gateway to smoking – at least not without further evidence.
5 – Do They Find Out Which Came First?
This is an obvious point – it’s unlikely to be a gateway effect if people smoke first and then try vaping – but it has a little more to it than you may think. It’s entirely possible that somebody tried smoking but didn’t like it, took up vaping afterwards, and then their new-found enjoyment of nicotine led them to try smoking (for a more efficient dosing mechanism, for example). Similarly, somebody could be interesting in consuming nicotine any way they can, and in this case, just because they happened to encounter an e-cig first doesn’t really make it a gateway. With these caveats in mind – which basically mean you can’t be certain based on which came first alone – this is still a strong indication of whether it’s a gateway case or a harm reduction one.
6 – Is it the Model or Reality?
Phillips rightfully describes the hunt for the gateway effect as a “highly politicized” issue, and this can have serious impacts on what researchers do and what they report. The simplest example is which other variables they attempted to control for (as in point 3): if you remove a potentially irrelevant factor, such as race, from your model, do the results stay the same? If you change your definition of what constitutes having been a smoker, does it affect the findings? The problem here is that researchers may settle on the model that produces the result they want and basically pretend that using this model was always their plan. If this happens, even peer reviewers won’t be able to tell that’s what they’ve done. The clearest indication of reliability is when the authors present numerous “cuts” at the data and show how this would affect the outcome. An example of this is shown in the paper, where the study from Stanton Glantz was analyzed in numerous ways, all of which were presented, and all of which showed that the harm reduction hypothesis was more likely than the gateway, based on the data.
Why the Research to Date Doesn’t Support the Gateway Hypothesis
The Glantz paper above (available in full for free) is a perfect example of a study that was claimed to offer evidence of a gateway, but could really do no such thing. They used the 2011 and 2012 NYTS data, and their findings can basically be summarized as “there was an association between vaping and smoking.”
Looking through the individual points above doesn’t help the study:
- No effort was made to define the gateway effect (it wasn’t even mentioned), and it wasn’t formed into a clear hypothesis.
- They didn’t consider the alternative, harm reduction hypothesis at all.
- Age and gender (two relevant factors) were controlled for, but they also included race (which is of questionable value) and made no effort to control for propensity or other confounding variables.
- The data would certainly make sense if the opposite were true: smokers would be more likely to vape if they were vaping to reduce the harm associated with their habit.
- There were no questions about which product came first.
- There was no presentation of the effect of using different models or adjusting definitions.
This study was a pretty bad one, admittedly, but the overall problems persist in almost all other studies into the gateway effect. One of the best which addressed vaping did take into account the order of use – and showed no obvious gateway effect – but again the research didn’t try to control for pre-existing risk through propensity scores.
Conclusion – Try Harder to Spot a Gateway
So far, the main limitation of research into the gateway effect can be summarized as not trying enough. It doesn’t take a scientist to tell you that smokers also having vaped doesn’t really tell you anything about whether one caused the other, but this simple point limits most of the available studies. The upshot is: if you want to claim vaping is a gateway, we need to make some serious effort to determine if it’s true. CASAA’s upcoming study will undoubtedly be good news for the industry.
- Phillips, C. V. (2015) Gateway Effects: Why the Cited Evidence Does Not Support Their Existence for Low-Risk Tobacco Products (and What Evidence Would)
- CASAA: CASAA Research Fund Launches Fundraising for Study of Gateway Claims
- Dutra, L. and Glantz, S. (2014) E-cigarettes and conventional cigarette use among US adolescents: A cross-sectional study