Socioeconomic patterning of vaping by smoking status among UK adults and youth

Use of electronic cigarettes (e-cigarettes) rose precipitously in the UK from 2011 [1], before recently plateauing with approximately 6% of adults currently using them [2]. While health consequences of long-term e-cigarette use (or vaping) are largely unknown [1‐4], expert opinion regards vaping as substantially less harmful than smoking [1, 5]. Mounting evidence supports this view [2, 6‐8] and that vaping may aid smoking cessation [2, 9‐11]. Nevertheless, concerns remain [3, 12], particularly around risks of vaping introducing youth to nicotine and cigarette use [13‐16], though so far regular (at least weekly) vaping by youth in the UK has been rare [17, 18].

The implications of e-cigarettes for socioeconomic inequalities in smoking and hence health have received limited attention [2]. A recent review found socioeconomic inequalities in awareness and use (ever or current) of e-cigarettes, but findings were very mixed, with some studies, especially those rated as higher quality, suggesting greater awareness and use in higher income, better educated groups, but others finding the reverse, or no clear differences [19]. One possible reason for inconsistent findings, may be varying consideration of smoking status, with some studies focusing on the general population [20‐22], others stratifying by smoking status [20, 23], and others focusing only on current or former smokers [24‐26].

Stratifying by smoking status is critical to understanding potential impacts of e-cigarettes on health inequalities as conclusions will depend on how vaping interacts with smoking behaviour. Table 1 identifies six groups based on e-cigarette and cigarette use and which we use for analysis in this paper. Vaping among never-smokers represents a potential public health concern: ‘safer than smoking’ may not be the same thing as ‘safe’ [2, 3, 27, 28]. Vaping among youth who have never smoked may be especially concerning as nicotine addiction established at this life-stage could be long-lasting, potentially increasing the likelihood of cigarette use over the life-course and/or risk for psychiatric disorders, future substance use and poor later life cognition [2, 29‐31]. If vaping is more likely among disadvantaged never smokers it could indicate potential widening of health inequalities (unless vaping replaces and does not lead to take-up of smoking).

Perhaps most importantly, management of nicotine addiction with e-cigarettes only (i.e. vaping among ex-smokers) is likely to be substantially less harmful than with smoking [1, 2, 6, 7], and vaping by ex-smokers may often represent intentional replacement of cigarettes with e-cigarettes for health reasons [32]. There could also be other reasons for switching such as relative price, preference, or differences in regulation as to where the behaviour is allowed. Vaping ex-smokers could also include long-time ex-smokers who have returned to nicotine with e-cigarettes, perhaps instead of relapse to smoking.

Historically, more advantaged individuals have been more successful in smoking cessation than those who are more disadvantaged [33], but this inequality has reduced in recent years, and e-cigarettes have been suggested by others as one possible explanation for this trend [34]. Thus, while we might expect to see inequalities in whether smokers have quit, a key question is whether inequalities in ex-smoking status are smaller with vaping than without [19]. Such a pattern would indicate a potential for e-cigarettes to narrow inequalities in smoking (and therefore health). However, vaping instead of smoking could still carry some residual health risk [35] relative to using neither, so socioeconomic patterning of vaping among ex-smokers, may still represent some residual potential to widen health inequalities.

Vaping among current smokers probably often indicates some interest in quitting and/or concern over the health risks associated with smoking (perhaps less so among youth where it may just indicate a general propensity for experimental or risky behaviour). Although dual-use of e-cigarettes and cigarettes may be motivated by harm reduction, there is little evidence that dual-use is any healthier than smoking only (since dual-use still includes smoking, even if it reduces the number of cigarettes smoked) [7]. Dual-users could represent a prime-target group who may be amenable to interventions to aid smoking cessation (e.g. by switching solely to vaping, or using vaping as a step towards complete cessation of nicotine use). Understanding the social patterning of vaping among current smokers could help understand the potential impacts of dual-use targeted interventions on inequalities in smoking.

Nevertheless, while stratification by smoking status is important for interpretations of inequalities in vaping, stratification can also introduce a phenomenon known as collider bias [36, 37], which can arise when conditioning on a variable that is determined by both the exposure of interest and other variables that determine the outcome of interest. This is illustrated in Fig. 1, showing socioeconomic position (SEP) as a determinant of vaping, with potential (primary) confounders and smoking as a mediator, i.e. it is determined by SEP, and, in turn, may determine vaping. The dashed line indicates that the effect of SEP on vaping may differ by smoking status. Collider bias can arise if there are confounders of the effect of smoking on vaping (denoted intermediate confounders in the diagram), such as individual smoking histories, or parental smoking/vaping (for youth). Since smoking is determined by both SEP and the intermediate confounders, stratifying (or otherwise conditioning) on it induces a spurious association between SEP and the intermediate confounders and can bias estimates of association between SEP and vaping. If there were no causal relationship between SEP and the intermediate confounders then adjusting for these in stratified analyses would be sufficient. However, if any of the intermediate confounders are also determined by SEP as shown in Fig. 1 (they need not all be determined by SEP), then adjusting for them in stratified analyses will remove part of the effect of interest, while stratifying without adjustment will induce collider bias, and estimates will be biased either way [38, 39]. In these circumstances, marginal structural models can be used to estimate controlled direct effects (CDEs) [38, 39] within strata of smoking (assuming no unmeasured confounding).

Thus, our paper, based on a representative UK survey, uses marginal structural models to estimate socioeconomic inequalities in vaping and smoking cessation, specifically addressing the three research questions presented in Table 2.

Table 3 shows sociodemographic patterning of vaping among youth aged 10–15 by whether or not they have ever smoked. Overall, vaping was rare among never smokers but more common among ever smokers (1.4% compared to 28.4%). Youth vaping was more common among males than females, among older youth, those with single parents, and those whose parents used e-cigarettes, were in disadvantaged occupation categories, or had lower incomes. Aside from the patterning by gender and parental vaping, all these associations were diluted but held among never smokers. Among the small group of ever smokers (n = 247) only patterning by gender and age was still evident.

Table 4 shows the sociodemographic patterning of vaping among adults (aged 16+) by smoking status. Vaping by never smokers was very rare at 1.1%, more common among ex-smokers (6.9%) and markedly more common among current smokers (27.1%). Vaping was more common overall among males, at younger ages (especially ages 25–54), among White UK respondents, for those who were single, those who had children in the household, those in disadvantaged groups (though the most disadvantaged categories for education and occupation had relatively low rates of vaping -perhaps due to conflation with age) and those who had smoked in previous waves. However, most of these patterns varied by smoking status, for example, the association with education was reversed among current smokers.

Table 5 shows estimates of socioeconomic inequalities in vaping by smoking status among youth aged 10–15 from both unadjusted and marginal structural models. Unadjusted associations and CDE estimates both indicated vaping was more likely for disadvantaged youth. The associations and effects estimates tended to be weaker among ever smokers than among never smokers, but not significantly so.

Table 5 also shows socioeconomic inequalities in vaping by smoking status among adults. The unadjusted associations indicated that socioeconomic disadvantage was associated with a greater likelihood of vaping overall, but stratifying by smoking status this association disappeared among never smokers and was reversed among ex and current smokers. After adjusting for smoking and other confounders in the CDE estimates, socioeconomic disadvantage seemed even more strongly linked to vaping. However, this was clearly concentrated among ex-smoking adults, with little to no effects among never or current smoking adults, suggesting that the observed associations were affected by either confounding or collider bias.

Table 6 shows socioeconomic inequalities in ex-smoking among ever-smoking adults by vaping status. Socioeconomic disadvantage was associated with reduced odds of being an ex-smoker and this observed association differed little by vaping status. However, after adjustment for confounders, CDE estimates of SEP on ex-smoking were marginally weaker among those who vaped than those who did not (p = 0.081).

Findings from sensitivity analyses with truncated weights were largely consistent (and therefore not shown), but the CDE estimate for SEP on vaping among adult ex-smokers (from Table 5) was attenuated (from OR: 1.17; 95% CI: 1.09–1.26 to OR: 1.06; 95% CI: 1.02–1.10) and less clearly differentiated from the estimate of no effect among never-smokers (p = 0.173). The overall CDE estimate for SEP on vaping among all adult respondents was also attenuated (from OR: 1.12; 95% CI: 1.06–1.18 to OR: 1.02; 95% CI: 1.00–1.05).

This robust analysis of data from a large and representative UK survey provides evidence of complex variations in vaping inequalities by age and smoking status. Our findings concur with reports of low vaping prevalence among youth, especially youth who have never smoked [1, 2, 17, 18, 32, 44]. Vaping was more likely for youth in disadvantaged than more advantaged groups, especially among youth who had never smoked, though concerns should be tempered by the low overall prevalence [45], and because some of this unequal take-up of vaping could be replacing unequal take-up of smoking, which would be more health damaging. Regarding adults, like others we found that more advantaged ever-smokers seemed more likely to use e-cigarettes, and to have quit smoking [1, 2, 24, 33], but the association between socioeconomic advantage and vaping among current smokers disappeared with adjustment for confounding and collider biases. Our adjusted analyses indicated that socioeconomic disadvantage increased the likelihood of vaping among ex-smokers, while there was little to no effect of SEP on vaping among never or current smokers.

With respect to the association between socioeconomic advantage and ex-smoking among ever-smoking adults, confounding adjustment showed slightly weaker effects of SEP among ever smokers who vaped than those who did not. Although only marginally different, this is encouraging and worthy of further research. Studies already suggest that vaping can aid smoking cessation [2, 9‐11], but if e-cigarettes are especially appealing or helpful to disadvantaged smokers, then this could have a welcome long-term impact of reducing health inequalities. There have been reductions in inequalities in successful smoking cessation in recent years [34], and some have attributed this trend to e-cigarettes. Our findings were somewhat consistent with this notion, but did not examine trends in cessation specifically, and further study of how inequalities in smoking cessation are impacted over time by e-cigarettes is needed.

Given the cross-sectional nature of the data, the adult findings (regarding both inequalities in vaping and cessation) could have been brought about by disadvantaged ex-smokers being more likely than advantaged ex-smokers to take up vaping (i.e. relapsing to nicotine but not cigarette use), as well as by disadvantaged smokers switching to vaping. We have not yet been able to investigate longitudinal dynamics between e-cigarette and cigarette use [9, 15, 46]. This will be possible with future waves of the study and clarifying inequalities in transitions between particular smoking and vaping states will be an important issue for such longitudinal research.

Other limitations included the question on vaping not distinguishing between different types of e-cigarette devices, or different frequencies of vaping [9, 47]. Associations might differ in either direction if we were able to further stratify by such variables, and frequency and device types could also affect how vaping should be interpreted in terms of public health impact (e.g. infrequent vaping may be less effective for cessation [4]). Nevertheless, theory regarding vaping leading to smoking does not necessarily rely on frequent use, mere experience of the social performance of nicotine use may be sufficient to facilitate transition to smoking [48]. Moreover, without data on intentions or attempts to quit, our comparisons of ex-smoking against current smoking potentially conflate socioeconomic differences in intentions to quit with socioeconomic differences in successful quitting. Youth was defined as ages 10–15 due to the structure of the survey data, but patterns could be distinct among young adults aged 16+ (who were grouped with adults here), though the age-range covered here is similar to that in other studies of UK youth [17, 18]. Further, while we have adjusted for many relevant confounders (including recent smoking history) while preserving the effects of SEP on these confounders, the effect estimates presented here assume no unmeasured confounding [46]. We did not have information on prior vaping, which could have particularly biased analyses of smoking cessation, as respondents recorded here as ex-smokers who do not vape could have already used vaping as an intermediate step to help them quit nicotine completely [49].

Nevertheless, understanding impacts of vaping on socioeconomic inequalities in smoking is especially important considering that few population-level tobacco control interventions (except taxation) have had much success in this regard [50‐52]. The role of e-cigarettes in smoking cessation continue to be a focus of policy and regulatory debate. Interventions to reduce smoking should aim to maximise potential benefits of e-cigarettes and consider possible impacts on inequalities. Taxation that keeps cigarette prices high relative to e-cigarettes may be an avenue worthy of study, and could help encourage disadvantaged smokers to switch to e-cigarettes or use vaping as a step towards no nicotine use. Measures may also be needed to protect vulnerable groups and minimise potential harms, e.g. by restricting availability of e-cigarettes to youth or at least closely monitoring inequalities in uptake among youth [2].

Assuming smoking is more harmful than vaping and that vaping is more harmful than no use of nicotine, the socioeconomic inequalities that we found in vaping among never smoking youth and ex-smoking adults could potentially lead to some future widening of socioeconomic inequalities in health. Conversely, we found weaker inequalities in smoking cessation among smokers who vaped, and this could have an opposing effect leading to narrowing of health inequalities. In other research, the potential impacts of e-cigarettes on smoking prevalence and health have been estimated using various predictive models with different assumptions [3, 53‐55], and similar efforts to estimate their likely impact on inequalities in smoking and health would now seem advisable to understand the potential net impacts of these opposing trends.