July 29, 2025

expert reaction to an observational study comparing teen smoking rates and vaping use across three UK birth cohorts over 50 years

An observational study published in Tobacco Control looks at teen smoking rates in three UK birth cohorts before and after vape use.

Dr John Robins, Research Associate at the Nicotine Research Group, King’s College London (KCL), said:

“This study uses data from three large longitudinal surveys and compares the prevalence of cigarette smoking among 16-17 year olds in 1974, 1986, and 2018. The headline finding that youth smoking rates have declined from 33% to 12% over this period is robust and reassuring, but the conclusions they draw about the role of e-cigarette use on the likelihood of smoking in 2018 are not justified.

“The authors divide up the 2018 cohort into three categories of youth – those who have never vaped, those who previously or experimentally vaped, and those who currently vape. They then contrast the predicted probabilities of current smoking among each of these three subgroups with the totality of youth in the 1974 and 1986 cohorts, and find that the predicted probability of smoking among the current vapers is higher than in those who don’t currently vape in 2018, and higher than in all youth in 1986. This comparison is highly misleading in and of itself, but especially so in the context of the vastly lower rate of smoking in recent years. Of course, the probability of smoking is highest amongst those who currently vape – psychoactive substance use does not exist in a vacuum and young people who use one type of drug are more likely to use another, due to constellation of genetic, individual  and environmental risk factors which result in a common vulnerability to substance use. In simple terms, “people who try stuff try other stuff”. Moreover, here we are not even talking about two different substances, we’re talking about nicotine with two different routes of administration, one profoundly damaging and responsible for the death of up to half of users, and the other a much less harmful alternative. I imagine that the picture would be similar if one were run the same analysis with, say, nicotine patches instead of e-cigarettes.

“In fairness, the authors do caution against any causal interpretation of their findings but then go on to claim that “it may be that youth who would have historically been considered ‘low risk’ of combustible cigarette use become ‘at risk’ through their nicotine exposure by experimenting with or using e- cigarette”, but make no mention of the alternative simple explanation that those youth who are predisposed to use nicotine are using nicotine, and that the overall low prevalence of smoking among youth these days is in part due those youth picking up the e-cigarette instead of smoking. Comparing the “high-risk” group in 2018 to the “all-risk” group at earlier timepoints is not a fair comparison. Whilst the authors do adjust for some sociodemographic and behavioural risk factors which might make someone “high-risk” for smoking, they are far from exhaustive (e.g. no adjustment is made for levels of baseline levels of impulsivity and sensation-seeking).

“Curiously absent from the paper is a predicted probability for smoking in the whole 2018 cohort. A very crude estimate by taking the weighted sum of the averages across the three 2018 groups gives an overall average predicted probability of smoking in 2018 youth of approximately 9.5%; much lower than the 30% and 22% predicted in the 1974 and 1986 cohorts. Yet the authors make no mention of this.

“An analogy of heroin use and methadone substitution treatment comes to mind. If you were to run the equivalent analysis of predicted probabilities of heroin use and compared a cohort of methadone users to a general population cohort, then of course the predicted probability of heroin use would be much higher in the methadone group than the general population group. Should we then say, well methadone treatment is a risk factor for heroin use in people who would otherwise be low risk, and methadone is the problem? Of course not. They are not comparable populations.”

Prof Lion Shahab, Professor of Health Psychology and Co-Director of the UCL Tobacco and Alcohol Research Group, University College London (UCL), said:

“This study tracks smoking rates among 16–17-year-olds across three birth cohorts—1958, 1970, and 2001—and finds a steady decline from 33% in 1974 to 12% in 2018. However, for the 2001 cohort, the study also examines smoking rates by vaping status. It finds that smoking was much higher (33%) among those who currently vaped, compared to just 1.5% among those who never vaped.

“While the authors acknowledge the association is not causal, their language implies otherwise, suggesting that vaping may be reversing progress in reducing youth smoking. This interpretation is problematic for two key reasons.

“First, the study does not establish the direction of the relationship—whether vaping leads to smoking or vice versa. It’s equally plausible that young people who smoke are more likely to try vaping. Second, even if vaping preceded smoking, this does not prove causality. Young people who vape and smoke often share underlying risk factors—a concept known as “common liability.” These include traits (e.g., sensation-seeking) or circumstances (e.g., stressful family environment) that predispose them to risky behaviours.

“The study attempts to adjust for some of these factors but does not account for “propensity to smoke.” In fact, one of the strongest predictors of smoking—parental smoking—was influential in earlier cohorts but not in the 2001 group. This suggests that vaping may be acting as a proxy for smoking risk in this cohort, making comparisons across time misleading.

“In short, while the findings are interesting, they do not support the claim that vaping causes smoking. Instead, they likely reflect shared vulnerabilities among youth who engage in both behaviours. Studies that address these problems, using more robust designs to overcome inherent biases or which triangulate findings using different methods, are now needed to provide reliable insights into the important question of whether vaping causes youth to smoke.”

Prof Peter Hajek, Professor of Clinical Psychology, and Director of the Health and Lifestyle Research Unit, Queen Mary University of London (QMUL), said:

“The study found that 16-17-year-olds who vape are more likely to also smoke than those who do not vape. The press release suggests that this means that in these teenagers, vaping leads to smoking, but it is more likely that the finding just shows that people attracted to vapes are also attracted to cigarettes, in the same way that compared to teetotallers, drinkers of white wine are more likely to also try red wine.  The press release includes a question whether increased popularity of vaping threatens the decline in smoking in youth. There are data to answer this. In England, smoking among 15-year-olds has been tracked since 1982. In 1982, 25% were regular smokers. In 2018, 36 years later, the date of the last UK survey used in this study, 7% smoked. In 2022, after just 5 years, the figure was 3%.  The decline in smoking has if anything accelerated.”

‘Risk of adolescent cigarette use in three UK birth cohorts before and after e-cigarettes’ by Mongilio et al. will be published in Tobacco Control at 23:30 UK Time Tuesday 29 July, which is when the embargo will lift.

DOI: 10.1136/tc-2024-059212

Declared interests

Dr John Robins: No conflicts of interest to declare

Prof Peter Hajek: No COI

Prof Lion Shahab: None