The notion that nicotine vaping leads people to smoking—known as the “gateway” theory—is a common refrain among those who seek to restrict access to vapes in the name of youth protection, including some scientists. It persists, amid overwhelming evidence that vaping is a “gateway” out of smoking for large numbers of people—and even though data from numerous countries show population-level correlations between increases in vaping and decreases in youth smoking.
Arielle Selya, PhD, is a behavioral scientist with a background in adolescent substance use and addiction. She authored a recent paper titled “The ‘Gateway’ hypothesis: evaluation of evidence and alternative explanations.”
The original gateway hypothesis, she noted, described use of legal substances (like tobacco and alcohol) leading to illicit substances, often starting with cannabis and then progressing to “harder” drugs. Concerns about cannabis being a “gateway drug” influenced marijuana policy for decades.
“The real question is: What would these youths have done in a hypothetical world without e-cigarettes? We call this the counterfactual.”
That original gateway hypothesis has been “debunked,” Dr. Selya pointed out, but the version applied to vapes is very much alive—despite exhibiting similar flaws. Our interview has been edited for length and clarity.
Kiran Sidhu: If some youth use electronic nicotine delivery systems (ENDS) and then go on to smoke, why isn’t that enough to prove the gateway theory?
Arielle Selya: Even though this is the time sequence that’s necessary for gateway—first using ENDS, and then later initiating smoking—this alone isn’t enough to prove the gateway hypothesis. This is because some of these youth might have gone on to smoke cigarettes anyway, even if they never used e-cigarettes.
Fundamentally, the real question is: What would these youths have done in a hypothetical world without e-cigarettes? We call this the counterfactual.
We can’t know for sure what would have happened in a counterfactual world without e-cigarettes, but there are ways of making an informed guess—based on youths’ other characteristics, or on population-level trends in smoking. Simply having the right time sequence (using ENDS first and then going on to smoke) doesn’t answer the counterfactual question, so it can’t show that e-cigarettes cause some youth to smoke.
KS: In your recent paper, you say that the problem of “confounding” is a “severe weakness of the evidence for the gateway explanation.” What do you mean by that?
AS: The main type of evidence put forward for the gateway hypothesis comes from surveys of youth that find that e-cigarette use is correlated with cigarette smoking. In other words, when you look at all youth, the ones who use e-cigarettes are more likely to also smoke cigarettes.
But this doesn’t mean e-cigarette use causes these youth to smoke. The ones who use no substances are very different from those who use multiple substances or products, when you look at things like home environment, peer group, mental health factors or personality traits.
For example, some youth have a greater tendency to take risks or experiment with different things, and this might explain why some youth try both [vapes and cigarettes] and others stay away from both.
That’s what I mean by the problem of confounding: These studies compare apples to oranges, and it’s not necessarily the case that e-cigarette use caused these youth to smoke. Many of them might have gone on to try cigarettes regardless of whether they used e-cigarettes first.
KS: You talk about “common liability”—where “youth who have certain characteristics were already predisposed to use both ENDS and cigarettes.” What’s the weight of evidence supporting this idea versus the gateway theory?
AS: I think even proponents of the gateway hypothesis would agree that there are some shared risk factors that partly explain both e-cigarette use and cigarette smoking. But there’s disagreement on how to handle this in the analysis and separate the (hypothesized) causal effect of e-cigarettes from the pre-existing tendency to use nicotine-containing products.
The most common way to handle this is with statistics, by adjusting for some of these shared risk factors or “common liability” factors. But there are so many common liability factors: Lee et al. identified about three dozen, and found that most studies only adjust for a handful.
“The more common liability factors are adjusted for, the weaker the association between e-cigarette use and smoking becomes.”
Even worse, there are common liability factors that we don’t even have information on—e.g., most surveys don’t collect genetic data—so it is difficult or impossible to adequately separate common liability from a possible gateway effect using conventional statistics approaches.
But the more common liability factors are adjusted for, the weaker the association between e-cigarette use and smoking becomes—to the point where in some cases, it’s no longer significant, meaning that common liability factors completely explain the association, and there’s no evident causal effect of e-cigarettes.
This results in nearly all gateway studies having the same fundamental flaw, of confounding by common liability factors. This is one reason why some researchers like Marcus Munafò have called for evaluating different kinds of evidence.
For example, beyond finding that youth who vape are more likely to also smoke, the gateway hypothesis would predict more youth smokers in the population as e-cigarette use increases. But that’s not what we see. Youth smoking rates are falling faster after e-cigarettes became available.
In my view, it seems there is only one type of evidence given in support of gateway—i.e., surveys of youth showing a correlation between vaping and smoking—but there are at least three types of evidence for common liability. There’s survey data where common liability largely or fully explains the apparent gateway association; there’s evidence showing shared genetic predisposition; and there’s the population trends showing that e-cigarette uptake is accompanied by more rapid declines in smoking. That’s three separate converging types of evidence in support of a common liability explanation.
KS: How much of a role does genetic predisposition play in youth smoking and vaping?
AS: There are probably many things that influence a predisposition to smoking and vaping. I am not an expert on genetic research, but there was a recent paper by Khouja et al. that concluded genetics explains part of the association.
They calculated polygenic risk scores, which is a way of looking at multiple specific genes that collectively are associated with initiating smoking, and they found that these polygenic risk scores predicted which youth went on to smoke and vape—and importantly, predicted the two behaviors about equally well. So this does seem to show some genetic predisposition in smoking and vaping, though we don’t yet know how much this contributes to overall predisposition compared to other non-genetic influences.
“Youth smoking was already declining before e-cigarettes became available, but smoking rates declined faster after e-cigarettes appeared.”
KS: Some youth who start with ENDS do go on to smoke. But it can also go the other direction—youths switching from smoking to vaping, which is a good thing as vapes are noncombustible and therefore safer. What does your research say about this?
AS: I talked about the counterfactual—that we don’t know which youth would have tried cigarettes in a hypothetical world without e-cigarettes. As you point out, another possible counterfactual scenario is that more youth could have smoked if e-cigarettes were not available.
We call this the diversion hypothesis—that in youth with a predisposition to use nicotine-containing products, e-cigarette use might divert some of them from ever trying cigarettes.
My own research on population-level smoking trends supports the diversion hypothesis, at least on average at the population level. Youth smoking was already declining before e-cigarettes became available, but smoking rates declined faster after e-cigarettes appeared.
This is consistent with the idea that e-cigarettes are displacing cigarette smoking to an extent. Although it’s not conclusive proof—since e.g., there could be other explanations for the more rapid decline—this pattern of declining smoking rates after e-cigarettes became available is remarkably consistent across countries. Not just in the US, but in Canada, New Zealand, Germany and other countries.
Similar trends have occurred in Japan and South Korea after heated tobacco products became available, and in Norway and Sweden when snus became available. All show a population-level shift away from combustible tobacco to less-harmful noncombustible products, which is favorable for public health.
I should note that theoretically, both diversion and gateway could be happening at the same time for different youth. It’s possible that there is a gateway effect for some youth, but the population trends show that this must be minimal. On the population level as a whole, the groups that have greater uptake of e-cigarettes have proportionally larger declines in cigarette smoking.
KS: You work for Pinney Associates, which consults for the vape company Juul; and Juul partially supported your recent paper. Would you say the impact of your research in the public debate is somewhat lessened by that connection?
I certainly think that some people immediately dismiss my work based on my affiliation. I am always happy to engage with people on the merits of my work, but in some cases the criticism doesn’t even get that far.
On the other hand, I also know that not having industry affiliation is no guarantee that my work will be embraced. Before I became affiliated with industry, I spent 10 years in academia. I had initially bought into the gateway hypothesis as a junior faculty member, but I changed my mind from my own analyses. I published frequently for several years, and most of my papers were in support of common liability and diversion explanations.
However, I felt that my work was being ignored by academic colleagues, despite my attempts to reach out to those with different opinions, and I feel that my stance on e-cigarettes was one factor limiting my ability to get external funding, which was impeding my success as an academic. Because of these challenges, I believed my work would have more impact on the real world if I worked in industry. Even if more people dismiss my work because of my affiliations, I wasn’t having a huge impact even before consulting to industry because of the polarization in the field.
“Gateway concerns are still very prevalent. I think to get wider acceptance of the existing evidence, we first need to address the polarization in the field.”
KS: What’s the likelihood of the gateway theory, at least as a headline description of youth trends, being finally put to bed? How can that happen?
AS: I think I am seeing a growing recognition that smoking rates have fallen to all-time lows and that the fears of the gateway hypothesis leading to growing youth smoking haven’t borne out. I’ve also seen the causal language about gateway claims being toned down and [people] be more careful about describing associations rather than making unsubstantiated causal claims. So I think there is some progress, but gateway concerns are still very prevalent.
I think to get wider acceptance of the existing evidence, we first need to address the polarization in the field. My colleague Joe Gitchell and I just wrote a commentary in the journal Addiction, calling for “adversarial collaborations” on the gateway question, where well-respected evidence-based researchers such as those affiliated with Cochrane may be in a position to engage both sides of the debate, starting with setting definitions.
For example, Conde et al. found that most gateway studies define cigarette smoking as ever smoking once in your lifetime, or smoking at all in the past month—but these definitions mostly capture experimentation, which isn’t likely to pose health harms if it doesn’t persist, as I discuss in another commentary with Riccardo Polosa and Martino Ruggieri. On the other hand, a pattern of escalating e-cigarette use followed by adding cigarettes could be concerning.
Coming to an agreement on definitions, analyses, and even types of evidence ahead of time could help the field come to a consensus.
Photograph (cropped) by Rubin Starset via Flickr/Creative Commons 2.0
The Influence Foundation, which operates Filter, has received unrestricted grants from Juul Labs, Inc, and personal donations from Joe Gitchell. Both The Influence Foundation and the Center of Excellence for the Acceleration of Harm Reduction, founded by Riccardo Polosa, have received grants from Global Action to End Smoking (formerly the Foundation for a Smoke-Free World). Filter’s Editorial Independence Policy applies.