Youth vaping and smoking and parental vaping: a cross-sectional survey

Background: Concerns remain about potential negative impacts of e-cigarettes including possibilities that: youth e-cigarette use (vaping) increases risk of youth smoking; and vaping by parents may have impacts on their children’s vaping and smoking behaviour. Methods: With cross-sectional data from 3291 youth aged 10-15 years from the Understanding Society Survey, we estimated effects of youth vaping on youth smoking (ever, current and initiation in the past year), and of parental vaping on youth smoking and vaping, and examined whether the latter differed by parental smoking status. Propensity weighting was used to adjust for measured confounders and estimate effects of vaping under alternative scenarios of no vaping vs universal adoption, and vs observed vaping levels. E-values were calculated to assess the strength of unmeasured confounding influences needed to negate our estimates. Results: Associations between youth vaping and youth smoking were attenuated considerably by adjustment for measured confounders. Estimated effects of youth vaping on youth smoking were stronger comparing no use to universal adoption (e.g. OR for smoking initiation: 32.5; 95% CI: 9.8-107.1) than to observed levels of youth vaping (OR: 4.4; 0.6-30.9). Relatively strong unmeasured confounding would be needed to explain these effects. Associations between parental vaping and youth vaping were explained by measured confounders. However, estimates for parental vaping on youth smoking indicated effects, especially for youth with ex-smoking parents (e.g. OR for smoking initiation: 11.3; 2.7-46.4) rather than youth with currently smoking parents (OR: 1.0; 0.2-6.4). Relatively weak unmeasured confounding could explain these parental vaping effects. Conclusions: While results for youth vaping and youth smoking associations indicated support for underlying propensities, estimated effects still required considerable unmeasured confounding to be explained fully. However, these estimates from cross-sectional data could also be explained by smoking leading to vaping. Stronger estimates for universal vaping adoption vs observed usage, indicated that if youth vaping does increase risk of youth smoking, this effect may be stronger in the general population of youth, than among those youth who typically vape. Associations of parental vaping with youth smoking and vaping were either explained by measured confounding or could be relatively easily explained by unmeasured confounding.

Nevertheless, potential public health benefits of e-cigarettes should be weighed against possible detriments [14,15]. Internationally, concerns have been raised, particularly relating to youth, that because the behaviours are similar and many (but not all) ecigarettes contain nicotine, vaping behaviour could help establish and/or maintain smoking behaviour [16][17][18][19][20]. As vaping prevalence has risen among adults, it is also important to understand the impacts this may have on young people who live with those adults. For example, if a smoking parent switches to vaping, what impact could this have on the risk of their children smoking and/or vaping? While vaping may be safer than smoking, nicotine exposure in adolescence specifically, may have concerning consequences including: increased risk for developing psychiatric disorders, effects on brain development and later-life cognition, and priming for future substance abuse [21].
The notion that vaping may increase risk for smoking can be contrasted with the notion of common liabilities [19,20]: that underlying propensities for both behaviours account for their close association among youth [2,9,17,[22][23][24][25][26]. While many studies [2,[22][23][24][25] have adjusted for measured differences in background, unmeasured common liabilities remain possible explanations [19,27], even for more recent longitudinal studies showing vaping preceding smoking [17,26,[28][29][30][31][32] and where respondents had stated no intention of smoking [33]. It is important to recognise that common liabilities and vaping increasing risk for smoking (or indeed, smoking increasing risk for vaping [32]) are not mutually exclusive explanations for associations between vaping and smoking among youth. Thus, it is more important to establish the relative contribution of each in explaining associations between youth smoking and vaping, than to try and establish any one of these as the 'true' explanation.
Parental smoking is among the most established risk factors for youth smoking [34], so it seems likely parental vaping could also be an important influence on youth behaviour, though it has been little studied as yet. Common liabilities are also viable explanations for associations between parental vaping and youth behaviour. A UK study has shown associations between parental vaping and youth initiation of smoking and vaping [32], though these were not the main focus of their study and were attenuated in adjusted models. If there are effects of parental vaping, these could feasibly differ depending on parental smoking status. If e-cigarettes are viewed as an aid to smoking cessation [3,14] then parental use could make smoking seem less normative and reduce risk of smoking initiation, especially if parents completely switch from smoking to vaping. On the other hand, dual-use by parents could result in the behaviours appearing to youth as linked and complementary and increase risk for both. Indeed, a study in Mexico showed youth susceptibility to vaping and smoking to be higher where family members were either cigarette or dual users, but not where family members only vaped [35].
With an interest in understanding these issues, we can be more specific in defining the effects of interest [36], recognising that in observational studies exposures are not experienced randomly and there may be important differences between those who are and are not exposed. For example, if we are interested in the effect of youth vaping on youth smoking, we may estimate a population average effect (often referred to as an average treatment effect or ATE, using language from experimental science), which represents a comparison of outcomes between a scenario where no youth vaped to one where all youth vaped. Alternatively, we might be interested in the effect of the observed level of youth vaping (referred to as the average treatment effect among the treated or ATT inclusion of all observed data from respondents with valid weights [38] (proportions missing for most variables were between 0 and 5.3%, though 18.6% for ethnicity; see Table 1). aAdolescents who had reported smoking in previous waves of the survey were excluded here (n = 216), so percentages indicate the proportion of those who had never smoked before who initiated smoking in this wave of the survey.

Measures
Youth and parents self-reported vaping in response to the question: "Do you ever use electronic cigarettes (e-cigarettes)?" (Yes/No). The present tense "Do you" wording should primarily identify current vaping, though the wording is a little ambiguous and may feasibly have been interpreted by some respondents as "Have you ever used electronic cigarettes?" Our measures of vaping could therefore include both very infrequent and/or ever use in addition to current vaping. Although this was the first time respondents were asked about vaping, smoking was self-reported by youth and parents in this and in earlier waves of the survey. Youth smoking was coded in three binary outcomes: ever, current and initiation (with the latter defined as current smokers who started smoking in the year since the previous survey, i.e. with no indication of smoking from earlier waves). Parental smoking (never, ex, current) and vaping (yes/no) were coded according to the highest level of use from either parent.

Statistical analyses
We estimated the ATEs and ATTs using a propensity weighting procedure, which is designed to balance measured confounders across the main exposure groups, i.e. youth who did and did not vape, and youth with parents who did and did not vape [36,39]. This involves first running logistic regression models to predict each exposure, based on measured confounders (identified a priori). Gender, age, ethnicity, family structure, household SEP, UK country and interview date were treated as potential confounders throughout, as was parental smoking (except when stratifying on this variable). For estimating effects of youth vaping, parental vaping was included as an additional confounder.
The predicted probability of each individual's observed exposure status can then be used to calculate weights designed to help estimate ATEs and ATTs. Table 2 details these calculations. ATE weights re-weight both exposed and unexposed respondents to resemble the total sample (with regards to measured confounders), while ATT weights re-weight the unexposed respondents to resemble the exposed group. Prior to using these weights to estimate effects, validity of the weights was assessed by examining mean differences in confounders associated with the relevant exposure [39]. Weights were deemed valid if confounder differences, expressed in standard deviation units, were reduced close to zero (with differences < 0.2 considered close to 0). Models predicting exposure probability initially used main effects of confounders only, but where imbalance remained the model was revised by adding interactions terms and then re-assessed. Improvements in confounder balance from model revisions were balanced against sufficient overlap of propensity distributions between exposed and unexposed groups by confirming that mean ATE weights were close to 1 [39,40] (the same is not expected of mean ATT weights).
Deviations from this would suggest that some individuals were being assigned extreme weights, indicating risk of making inferences not strongly supported by the available data. [ Table 2 about here] Exposure effects were estimated in ATE-and ATT-weighted logistic regressions of each outcome on the exposure of interest. For comparison, we also present associations weighted for sample selection only (labelled "sample weighted associations"). Standard errors were adjusted for clustering of youth within households. Z-tests were used to compare differences in effect estimates between strata of parental smoking and between ATEs and ATTs [41].
Analyses of smoking initiation excluded 216 youth who had reported ever smoking in previous survey waves. These prior smokers were older, more likely to be vaping and to have single parents. Since this could introduce selection bias, these differences were reduced by additional weighting back to the total sample for analyses of initiation.
It is important to emphasise that the resultant effect estimates may not necessarily reflect the true effects of interest. For example, while our method aims to balance measured confounders between exposure groups, our effect estimates may still be biased by unmeasured confounders. For this reason, we calculate e-values for each point estimate and for the lower limit of the confidence interval [42]. E-values represent the minimum strength of association (OR in our analysis) that a set of unmeasured confounders would need to have with both the outcome and exposure of interest (independent of measured confounders), in order to explain away the association, or cause its lower confidence interval to include the null (if it already includes the null the e-value for the lower limit will be 1). We also include e-values for the sample-weighted associations, to indicate how much these were reduced by the adjustments made for measured confounders.

Results
Sociodemographic patterning of youth vaping and youth smoking are shown in Table 1.
There were strong associations between youth vaping and youth smoking for all smoking measures. Youth whose parents vaped were more likely to vape and smoke themselves.

Youth vaping and youth smoking
Mean ATE weights were close to 1 (0.999) indicating stability. Figure 1 shows standardised mean differences in confounders associated with youth vaping before and after propensity weighting. Youth who vaped were more likely to be male, older, come from disadvantaged or single-parent households, and have vaping parents. Propensity weighting attenuated these differences to below the 0.2 standard deviation threshold, indicating successful balancing of confounder characteristics across exposure groups. Similar confounder balance was achieved among the sub-sample of youth who had not smoked before (results not shown).
[ Figure 1 about here]  ATT compares outcomes between a hypothetical scenario where no youth vape against one with actual observed youth vaping. The e-value for the OR indicates the minimum strength of association (OR) that an unmeasured confounder would need to have with both youth vaping and youth smoking to reduce this estimate to the null. The e-value for the lower limit indicates the minimum strength of association that an unmeasured confounder would need to have with both youth vaping and youth smoking for the lower limit of the confidence intervals around this estimate to cross the null (all confounders are unmeasured for the sample weighted associations).
[ Table 3 about here] Parental vaping and youth smoking and vaping There were only 12 cases of youth whose parents vaped but had never smoked. This was considered insufficient information to estimate effects of parental vaping among youth whose parents never smoked, so we present results for: all youth combined; youth with ex-smoking parents; and youth with currently smoking parents. Mean ATE weights for parental vaping were close to 1 (1.001) indicating stability. Figure 2 shows standardised mean differences in confounders associated with parental vaping for each of these groups.
Parental vaping was associated: with more parental current and less parental ex-smoking, and with socioeconomic disadvantage and ethnic majority status among all youth (Fig. 2a); with ethnic majority status and socioeconomic disadvantage among youth with ex-smoking parents (Fig. 2b); and with ethnic majority status and having couple parents among youth with currently smoking parents (Fig. 2c). Propensity weights successfully balanced measured confounders, with the exception of a small residual bias in interview date among all youth such that youth whose parents vaped tended to be interviewed later.
Similar balance was achieved for youth who had not smoked before (results not shown).  ATT compares outcomes between a hypothetical scenario where no parents vape against one with actual observed parental vaping. The e-value for the OR indicates the minimum strength of association (OR) that an unmeasured confounder would need to have with both parental vaping and youth outcomes to reduce this estimate to the null. The e-value for the lower limit indicates the minimum strength of association that an unmeasured confounder would need to have with both parental vaping and youth outcomes for the lower limit of the confidence intervals around this estimate to cross the null (all confounders are unmeasured for the sample weighted associations). Figure 1: Standardised Mean Differences in Confounders Associated with Youth Vaping The shaded band indicates the area we considered close to zero.
Among youth with ex-smoking parents (i.e. comparing youth with ex-smoking parents who vaped, against those with ex-smoking parents who did not vape), parental vaping was associated with youth current smoking, youth smoking initiation and youth vaping. For current smoking and initiation, ATE estimates were slightly stronger than the sample weighted associations but were attenuated by 41% for youth vaping. ATT estimates were attenuated by 33-41% relative to the sample-weighted associations, but still indicated a clear effect on smoking initiation. Again, the e-values indicated that relatively little unmeasured confounding would be required to explain these effects. Among youth with current smoking parents none of the estimates indicated much evidence for relationships with parental vaping.

Discussion
We found cross-sectional associations between vaping and smoking among youth, and some associations between parental vaping and youth smoking and youth vaping. Our analyses suggested that confounding was a major contributor to these associations, though associations between youth vaping and youth smoking remained after adjustment for observed confounders. Our effect estimates indicated risk of youth vaping increasing youth smoking, especially if youth adopted vaping more widely, but there was little evidence overall to suggest that current levels of parental vaping are increasing risk for youth smoking or vaping.
We used a rigorous propensity weighting procedure to estimate effects of youth vaping on youth smoking after balancing measured confounders (i.e. parental smoking and vaping, gender, age, ethnicity, family structure, household SEP, UK country and interview date) between youth who did and did not vape. Our estimates suggested that common liabilities related to these measured confounders explained considerable proportions of the close association between youth vaping and smoking. Nevertheless, unmeasured confounders (e.g. beliefs, values and personality) would need to have quite strong independent associations (ORs generally of magnitude 9 or more) with both smoking and vaping to explain the residual relationship. Unmeasured confounding of this magnitude is theoretically possible though unlikely in the form of a single strong confounding factor, but could be feasible as an aggregate effect from a set of weaker confounders [42].
We estimated the effect of youth vaping on youth smoking (as the most concerning direction of effect), but it is important to emphasise that our data were cross-sectional, and our effect estimates could be either partially or completely accounted for by effects of youth smoking on youth vaping (i.e. reverse causation, e.g. youth using e-cigarettes as a smoking cessation aid) [19], especially as others have shown longitudinal effects in both directions [32]. This would include our measure of smoking initiation as initiation within the past year could have led to vaping. However, reverse causation is less likely for the initiation estimates than for those relating to current smoking, as youth who have been smoking for longer would be excluded. Additionally, questions on vaping did not distinguish between different types of e-cigarette/vaping devices, different frequencies of, or motivations for vaping [10,43], and our estimates may have changed if these factors could have been taken into account. Also, our data were focused on youth in the age range of 10-15 years and associations and effects could differ among older youth, as they increasingly adopt more adult behaviours.
Our findings align with others showing strong associations between youth smoking and vaping [2, 9, 17, 19, 22-27, 31, 32, 44]. We were able to explain much of this association with a relatively limited set of measured confounders and the residual effect estimates could be at least partially explained by unmeasured confounding and/or reverse causation.
Taken together with other evidence such as observed increases in youth smoking after implementation of e-cigarette sale restrictions [45][46][47] and continued declines in youth pro-smoking attitudes while youth vaping has been rising [48], it seems that even if there is an effect whereby vaping increases risk for smoking, it may not be the primary or dominant explanation for these associations.
What is most novel in our findings is the suggestion of differential relationships between smoking and vaping in youth. We estimated the expected differences in youth smoking outcomes comparing no youth vaping to both all youth vaping (ATEs) and to actual observed levels of youth vaping (ATTs). The latter set of estimates suggested weaker effects that would be more easily explained by unmeasured confounding, especially for initiation of smoking. This suggests that any effect of vaping on smoking may be weaker for youth already pre-disposed to vaping by background factors. This is consistent with another study that found stronger effects of vaping among youth with no intention to smoke [33]. Thus, effects of vaping on smoking could become more salient and important if vaping were adopted much more widely and by a broader range of youth, as opposed to the current low prevalence. Governments may want to prioritise preventing wider adoption, e.g. with e-cigarette age-of-sale and advertising restrictions (actions that many public health actors agree on [49]), over changing or stopping current youth vaping behaviour.
Perhaps propensities for vaping and smoking are similar enough that vaping has little additional impact where propensity for smoking is already high. Indeed, one theoretical explanation for why young people may transition from vaping to smoking is that vaping provides experience and training in the social performance of a similar behaviour, which might otherwise be unfamiliar [20]. Such first-hand experience could come from infrequent or even singular experiences with vaping and could plausibly be more important for young people without a background predisposition to use, who may have less experience of seeing others smoke or vape. Mechanisms for why young people would begin vaping before smoking include e-cigarettes being viewed as less harmful, more acceptable, having attractive flavours, and being easier to conceal [20]; these mechanisms could also be more salient for those without a background propensity for use.
We found no support for our postulation that the effects of parental vaping would be weaker or reversed when parents had completely switched from cigarettes to e-cigarettes, compared to parents using both. Contrarily, parental vaping was most strongly associated with youth smoking and vaping among youth whose parents were ex-smokers, with little evidence of associations with parental vaping among those whose parents currently

Conclusions
Associations between youth smoking and vaping were attenuated substantially with adjustment for measured confounders, indicating support for common liabilities, but a relatively high degree of unmeasured confounding would be required to completely explain the association. Estimates of effects of youth vaping on youth smoking appeared weaker and among youth already predisposed to vaping. There was some evidence for effects of parental vaping on youth smoking and vaping, particularly among youth whose parents were ex-rather than current smokers, though unmeasured confounding could easily explain these associations. It may be important to prevent wide-scale adoption of vaping by youth, but evidence for negative impacts of parental vaping on youth was weak.

Ethics approval and consent to participate
The University of Essex Ethics Committee has approved all data collection on the Understanding Society main study and innovation panel waves. All participants gave verbal informed consent, and consent was also obtained from parents/guardians for respondents under 16 years of age.

Authors' Contributions
MG conceived and designed the study, carried out the analysis, and drafted the manuscript.
LG and HS both contributed to the interpretation of data and substantive revisions of the manuscript. All authors have read and approved the submitted version of the manuscript.  The shaded bands indicate the area we considered close to zero.