It has been shown that never-smoking adolescents who try electronic cigarettes (e-cigarettes) are at increased risk of subsequent conventional cigarette smoking. We evaluated associations between e-cigarette use and progression to established smoking among adolescents who had already tried cigarettes.
Among participants (age 12–17 years) in the nationally representative Population Assessment of Tobacco and Health survey who had smoked a cigarette (≥1 puff) but not yet smoked 100 cigarettes (N = 1295), we examined 3 outcomes at 1-year follow-up as a function of baseline e-cigarette use: (1) having smoked ≥100 cigarettes (established smoking), (2) smoking during the past 30 days, and (3) both having smoked ≥100 cigarettes and past 30-day smoking (current established smoking). Survey-weighted multivariable logistic regression models were fitted to obtain odds ratios (ORs) and 95% confidence intervals (CIs) adjusted for smoking risk factors.
Versus e-cigarette never use, having ever used e-cigarettes was positively associated with progression to established cigarette smoking (19.3% vs 9.7%), past 30-day smoking (38.8% vs 26.6%), and current established smoking (15.6% vs 7.1%). In adjusted models, e-cigarette ever use positively predicted current established smoking (OR: 1.80; 95% CI: 1.04–3.12) but did not reach statistical significance (α = .05) for established smoking (OR: 1.57; 95% CI: 0.99–2.49) and past 30-day smoking (OR: 1.32; 95% CI: 0.99–1.76).
Among adolescent cigarette experimenters, using e-cigarettes was positively and independently associated with progression to current established smoking, suggesting that e-cigarettes do not divert from, and may encourage, cigarette smoking in this population.
In previous studies of youth who have never smoked cigarettes, those who tried electronic cigarettes (e-cigarettes) were more likely to initiate conventional cigarette smoking compared with e-cigarette never users. In cross-sectional studies, e-cigarette use is associated with established youth smoking.
Among youth who already experimented with cigarettes but were not yet established smokers, having used e-cigarettes was prospectively associated with onset of current established cigarette smoking. For these youth, e-cigarettes appear to encourage progression to established smoking.
Electronic cigarettes (e-cigarettes) are increasingly popular among youth; from 2014 to 2016, more US middle and high school students used e-cigarettes than any other tobacco product, including conventional cigarettes.1 All currently available longitudinal studies have revealed that among never-smoking adolescents and young adults, e-cigarette use is associated with subsequent cigarette smoking.2,–6 This association was shown in studies taking place in California,3 Hawaii,2 and the Mid-Atlantic region,7 as well as in nationally representative US samples,6,8 Canada,4 and the United Kingdom.5 Seven of these studies were summarized in a recent meta-analysis, revealing more than a threefold increase in the risk of cigarette smoking initiation when comparing youth e-cigarette ever users to never users.9 Although this association between e-cigarettes and smoking initiation has been consistent across the literature and could be explained by using a proposed “catalyst” model,10 some have argued that the relationship partly reflects a shared propensity for experimentation with different nicotine-containing products.11
Many individuals at low risk of smoking initiation may be included in studies of baseline cigarette never-users. In contrast, youth who have already begun cigarette experimentation represent a population at high risk of progression to greater levels of cigarette use later in adolescence and into adulthood. Although smoking even 1 cigarette is concerning, becoming an established smoker in adolescence is of substantial clinical and public health concern and is strongly associated with continuing to smoke regularly.12 Therefore, in the present investigation we consider high-risk youth, as evident by having already tried smoking (ever smoked ≥1 puff) but not yet smoked 100 cigarettes, and evaluate whether e-cigarette use in this population predicts progression from experimentation to established cigarette smoking 1 year later.
In a previous cross-sectional analysis of the 2011 and 2012 National Youth Tobacco Surveys (NYTSs), among youth who had ever smoked a cigarette, ever use of e-cigarettes was associated with being an established smoker (lifetime smoked ≥100 cigarettes), including after adjusting for socio-demographic variables.13 However, the cross-sectional design of that analysis precluded causal conclusions because of uncertain temporal sequencing between e-cigarette use and established smoking.
In the current study, we used the nationally representative Population Assessment of Tobacco and Health (PATH) Study14 Waves 1 (2013–2014) and 2 (2014–2015) to examine these same relationships prospectively. We hypothesized that among PATH youth participants who had already tried cigarette smoking but not yet smoked a total of 100 cigarettes, use of e-cigarettes would be positively associated with becoming an established cigarette smoker within 1 year.
Methods
Researchers from the PATH Study selected participants using a 4-stage stratified probability design with oversampling for tobacco users, African Americans, and young adults (ages 18–24 years). The PATH youth sample consisted of adolescents (up to 2 per household) whose parents were selected for the PATH adult sample.14 Researchers from the PATH Youth Study enrolled 13 651 US adolescents ages 12 to 17 years at baseline (2013–2014), with 87.9% retention (unweighted) at Wave 2 (2014–2015).
In-home in-person computer-assisted interviews were conducted in administering the PATH questionnaire. In separate sections, participants were asked about their tobacco use (eg, ever use, number of lifetime uses, and number of days used in the past 30 days) for 8 types of tobacco and nicotine-containing products, including cigarettes and e-cigarettes. Tobacco use questions were repeated during the Wave 2 interview, including for individuals who reached age 18 before follow-up and were therefore administered the Wave 2 adult questionnaire.
In the present analysis, we included youth who had smoked ≥1 cigarette puff but had not yet smoked 100 cigarettes at baseline (smoking experimenters), with known smoking status at follow-up (N = 1295). We examined 3 outcomes at follow-up as a function of baseline e-cigarette use: (1) having smoked ≥100 cigarettes (established smoking), (2) smoking during the past 30 days, and (3) both having smoked ≥100 cigarettes and past 30-day smoking (current established smoking). We categorized e-cigarette use in 2 ways: (1) ever use or never use, and (2) never use, nonpast 30-day use (former use), or past 30-day use.
Logistic regression models (Stata 14; StataCorp, College Station, TX) were used to adjust for hypothesized confounding variables in 3 stages. First, 6 separate unadjusted models were fitted to cover each combination of independent variable (Wave 1 e-cigarette never or ever use and e-cigarette never or former or past 30-day use) and dependent variable (Wave 2 established smoking, current smoking, and current established smoking). In the second stage, we added sex, age (in years), and race and/or ethnicity (Hispanic and/or Latino, non-Hispanic white, non-Hispanic African American, other) as covariables in all models, matching the confounders used in a previous cross-sectional analysis of NYTS data.13 In the third stage, parent education (≥1 parent with a bachelor degree or greater), urban residence (based on sampling units), household tobacco use (lives with ≥1 tobacco user), alcohol ever use, tobacco advertisement receptivity15 (can recall brand of favorite advertisement), sensation-seeking score (scale from 3 to 15), cigarette warning label exposure (Likert-type scale), interview time of year (summer versus all other months), and ever use of any other tobacco product (ie, cigars, pipes, hookah, bidis, kreteks, snus, dissolvable tobacco, and conventional moist snuff or chewing smokeless tobacco) were also included in additional adjusted models. A sensation-seeking score was a composite of 3 Likert-type items (liking frightening things, willingness to break rules, and preference for exciting and unpredictable friends) and has been shown to correlate with youth tobacco use.16 Interview time of year was included because, for youth, the scholastic calendar may play a role both in opportunity and social pressure to experiment with tobacco products.
All models were weighted for sampling design and nonresponse by using balanced repeated replication to be representative of the Wave 1 target population.17 Multiple imputation was performed for missing observations (0.7% of data), with variance estimates adjusted accordingly.
An institutional review board at the University of California, San Francisco reviewed and designated the study protocol exempt for this analysis of deidentified survey data. The PATH Study protocol received a National Institutes of Health Certificate of Confidentiality and approval from the Westat Institutional Review Board. Parental consent was requested on behalf of participating youth. Youth who completed the questionnaire were given $25.
Results
Among baseline cigarette experimenters (mean age: 15.5 years; 48.3% girls), having ever used e-cigarettes was positively associated with progression to established cigarette smoking in Wave 2 (Table 1). Compared with e-cigarette never users, e-cigarette ever users were twice as likely to report Wave 2 established smoking (19.3% vs 9.7%; P < .001) and current established smoking (15.6% vs 7.1%; P < .001) and were more likely to report past 30-day smoking (38.8% vs 26.6%; P < .001).
Progression From Cigarette Experimentation to Established Smoking, According to Baseline E-cigarette Use
. | n . | Weighted % With Outcome . | Unadjusted . | Adjusteda . | Adjustedb . | |||
---|---|---|---|---|---|---|---|---|
OR (95% CI) . | P . | OR (95% CI) . | P . | OR (95% CI) . | P . | |||
Outcome: smoked 100 cigarettes | ||||||||
Wave 1 predictors | ||||||||
E-cigarette never | 646 | 9.7 | Reference | — | Reference | — | Reference | — |
E-cigarette ever | 582 | 19.3 | 2.23 (1.55–3.21) | <.001 | 2.07 (1.41–3.04) | <.001 | 1.57 (0.99–2.49) | .05 |
E-cigarette never | 646 | 9.7 | Reference | — | Reference | — | Reference | — |
E-cigarette former | 406 | 18.6 | 2.13 (1.43–3.18) | <.001 | 2.04 (1.33–3.12) | .001 | 1.55 (0.94–2.56) | .09 |
E-cigarette in the past 30 d | 171 | 21.5 | 2.56 (1.58–4.14) | <.001 | 2.22 (1.31–3.74) | .003 | 1.69 (0.93–3.05) | .08 |
Outcome: smoked during the past 30 d | ||||||||
Wave 1 predictors | ||||||||
E-cigarette never | 699 | 26.6 | Reference | — | Reference | — | Reference | — |
E-cigarette ever | 596 | 38.8 | 1.75 (1.35–2.27) | <.001 | 1.65 (1.26–2.15) | <.001 | 1.32 (0.99–1.76) | .06 |
E-cigarette never | 699 | 26.6 | Reference | — | Reference | — | Reference | — |
E-cigarette former | 415 | 36.1 | 1.56 (1.15–2.12) | .004 | 1.48 (1.09–2.02) | .01 | 1.20 (0.86–1.68) | .29 |
E-cigarette in the past 30 d | 176 | 45.3 | 2.29 (1.64–3.19) | <.001 | 2.10 (1.47–2.99) | <.001 | 1.64 (1.12–2.41) | .01 |
Outcome: smoked 100 cigarettes and smoked during the past 30 d | ||||||||
Wave 1 predictors | ||||||||
E-cigarette never | 644 | 7.1 | Reference | — | Reference | — | Reference | — |
E-cigarette ever | 580 | 15.6 | 2.43 (1.55–3.80) | <.001 | 2.23 (1.39–3.59) | <.001 | 1.80 (1.04–3.12) | .03 |
E-cigarette never | 644 | 7.1 | Reference | — | Reference | — | Reference | — |
E-cigarette former | 406 | 15.5 | 2.41 (1.46–3.97) | <.001 | 2.29 (1.35–3.89) | .002 | 1.85 (1.02–3.36) | .04 |
E-cigarette in the past 30 d | 171 | 16.3 | 2.56 (1.52–4.32) | <.001 | 2.19 (1.24–3.88) | .007 | 1.76 (0.92–3.37) | .09 |
. | n . | Weighted % With Outcome . | Unadjusted . | Adjusteda . | Adjustedb . | |||
---|---|---|---|---|---|---|---|---|
OR (95% CI) . | P . | OR (95% CI) . | P . | OR (95% CI) . | P . | |||
Outcome: smoked 100 cigarettes | ||||||||
Wave 1 predictors | ||||||||
E-cigarette never | 646 | 9.7 | Reference | — | Reference | — | Reference | — |
E-cigarette ever | 582 | 19.3 | 2.23 (1.55–3.21) | <.001 | 2.07 (1.41–3.04) | <.001 | 1.57 (0.99–2.49) | .05 |
E-cigarette never | 646 | 9.7 | Reference | — | Reference | — | Reference | — |
E-cigarette former | 406 | 18.6 | 2.13 (1.43–3.18) | <.001 | 2.04 (1.33–3.12) | .001 | 1.55 (0.94–2.56) | .09 |
E-cigarette in the past 30 d | 171 | 21.5 | 2.56 (1.58–4.14) | <.001 | 2.22 (1.31–3.74) | .003 | 1.69 (0.93–3.05) | .08 |
Outcome: smoked during the past 30 d | ||||||||
Wave 1 predictors | ||||||||
E-cigarette never | 699 | 26.6 | Reference | — | Reference | — | Reference | — |
E-cigarette ever | 596 | 38.8 | 1.75 (1.35–2.27) | <.001 | 1.65 (1.26–2.15) | <.001 | 1.32 (0.99–1.76) | .06 |
E-cigarette never | 699 | 26.6 | Reference | — | Reference | — | Reference | — |
E-cigarette former | 415 | 36.1 | 1.56 (1.15–2.12) | .004 | 1.48 (1.09–2.02) | .01 | 1.20 (0.86–1.68) | .29 |
E-cigarette in the past 30 d | 176 | 45.3 | 2.29 (1.64–3.19) | <.001 | 2.10 (1.47–2.99) | <.001 | 1.64 (1.12–2.41) | .01 |
Outcome: smoked 100 cigarettes and smoked during the past 30 d | ||||||||
Wave 1 predictors | ||||||||
E-cigarette never | 644 | 7.1 | Reference | — | Reference | — | Reference | — |
E-cigarette ever | 580 | 15.6 | 2.43 (1.55–3.80) | <.001 | 2.23 (1.39–3.59) | <.001 | 1.80 (1.04–3.12) | .03 |
E-cigarette never | 644 | 7.1 | Reference | — | Reference | — | Reference | — |
E-cigarette former | 406 | 15.5 | 2.41 (1.46–3.97) | <.001 | 2.29 (1.35–3.89) | .002 | 1.85 (1.02–3.36) | .04 |
E-cigarette in the past 30 d | 171 | 16.3 | 2.56 (1.52–4.32) | <.001 | 2.19 (1.24–3.88) | .007 | 1.76 (0.92–3.37) | .09 |
ORs and CIs corresponding to model covariates are shown in Supplemental Table 2. —, not applicable.
Model covariates include the following: sex, age, and race and/or ethnicity.
Model covariates additionally include the following: parent education, urban residence, household tobacco use, alcohol ever use, tobacco advertisement receptivity, sensation-seeking score, cigarette warning label exposure, interview time of year, and ever use of any other tobacco product.
In models adjusted for sex, age, and race and/or ethnicity (Table 1), Wave 1 e-cigarette ever use (versus never use) was associated with approximately twice the odds of progression to Wave 2 established cigarette smoking (odds ratio [OR]: 2.23; 95% confidence interval [CI]: 1.55–3.21; P < .001), past 30-day smoking (OR: 1.75; 95% CI: 1.35–2.27; P < .001), and current established smoking (OR: 2.43; 95% CI: 1.55–3.80; P < .001). Associations were attenuated in fully adjusted models (Table 1), but e-cigarette ever use remained a positive and statistically significant predictor of current established smoking (OR: 1.80; 95% CI: 1.04–3.12; P = .035). Associations did not reach the threshold for statistical significance for established smoking (OR: 1.57; 95% CI: 0.99–2.49; P = .055) and past 30-day smoking (OR: 1.32; 95% CI: 0.99–1.76; P = .059).
When baseline e-cigarette former use (tried but not used in past 30 days) and past 30-day use were considered separately, there was a stepwise increase in the probability of progression to future established smoking from never to former to past 30-day e-cigarette use (Table 1). For example, the probability of Wave 2 past 30-day cigarette smoking rose from baseline e-cigarette never use (26.6%) to former use (36.1%) to past 30-day use (45.3%). Both e-cigarette former use and past 30-day use remained statistically significantly associated with all 3 Wave 2 cigarette outcomes in models adjusted for sex, age, and race and/or ethnicity (Table 1). In fully adjusted models, baseline e-cigarette former use remained a statistically significant predictor of progression to current established smoking (OR: 1.85; 95% CI: 1.02–3.36; P = .042), and baseline e-cigarette past 30-day use statistically significantly predicted progression to past 30-day smoking (OR: 1.64; 95% CI: 1.12–2.41; P = .010). Adjustment variables that were consistently associated with greater progression to established smoking included household tobacco use and tobacco advertisement receptivity (Supplemental Table 2).
Discussion
In this study, among youth who had experimented with cigarettes but had not progressed to established smoking, additional use of e-cigarettes was positively associated with future onset of current established smoking. Across 3 different definitions of established smoking and 2 different specifications of e-cigarette use, baseline e-cigarette users were at 1.5 to 2 times greater odds of progression to established smoking than e-cigarette never users, after adjustment for confounding variables. Fully adjusted associations with e-cigarette ever use were statistically significant for 1 definition of established smoking (current established smoking; P = .035) but fell just short of the a priori threshold for statistical significance for established smoking (P = .055) and past 30-day smoking (P = .059). The ORs in the present longitudinal analysis were in the same direction but smaller in magnitude than in the previous cross-sectional analysis of NYTS data13 in which new trials of e-cigarettes among previously established smokers could have been captured.
Regardless of how Wave 1 e-cigarette and Wave 2 smoking variables were specified, positive associations persisted after statistical adjustment for sex, age, and race and/or ethnicity. Adding the full set of confounding variables to models, such as household tobacco use, warning label exposure, and baseline use of other tobacco products, reduced the strength of some of the observed associations to below the threshold for statistical significance. However, all associations remained positive in direction and similar in magnitude across different definitions of e-cigarette exposure and the smoking outcome.
Suggested in these results is that e-cigarette use is more likely to encourage youth smoking than to divert youth from smoking when considering individuals who have already experimented with cigarette use. Unlike adults, particularly cigarette smokers, who commonly report a desire to quit smoking as a main motivator for e-cigarette use,18 youth are more likely to cite curiosity as a reason to try e-cigarettes.19 E-cigarette use was not associated with cigarette quit attempts or with quit contemplation among US middle and high school students in any NYTS wave from 2011 to 2015.20
In existing studies of youth who had never smoked a cigarette at baseline, those who tried e-cigarettes were more likely to initiate cigarette smoking in the future.2,–6,9 In addition to smoking initiation among youth never-smokers, we demonstrate in the current study that e-cigarette use was also associated with progression to current established smoking among youth smoking experimenters.
In a study of California 10th grade students that included never smokers and current smokers at baseline, greater frequency of e-cigarette use at baseline was associated with subsequently greater levels of smoking frequency (days smoked in past month) and heaviness (cigarettes smoked per day) 6 months later.21 Similarly, in a school-based study of adolescent never and current smokers in Canada, baseline past 30-day e-cigarette use was associated with initiation of daily smoking 1 year later.4 In the results of a school-based study of baseline cigarette ever smokers in Hawaii, a statistically significant change at follow-up in smoking frequency (measured as numerical categories) between baseline e-cigarette ever and never users was not yielded.2 However, in a school-based study of adolescents in the United Kingdom, ever use (versus never use) of e-cigarettes was associated with “escalation” to smoking sometimes or usually among baseline nonsmokers who had used cigarettes in the past.5 Authors of that study reported an adjusted OR that was similar to the current study (OR: 1.89; 95% CI: 0.82–4.33) but not statistically significant (P = .13) in a smaller sample (n = 318).5
The smoking outcomes evaluated in the current study represent intensity levels of clear clinical and public health concern. Although smoking as infrequently as 1 day in the past month in adolescence is predictive of adult smoking,22 youth who reach higher levels of smoking are even more likely to continue to smoke.12 Additionally, although more recent (past 30-day) e-cigarette use was a stronger predictor of future established smoking than former e-cigarette use in unadjusted models, this pattern did not necessarily persist in fully adjusted models. We suggest that any level of e-cigarette use among adolescent cigarette experimenters may be a meaningful risk indicator of smoking progression.
Several study advantages strengthened the conclusions that can be drawn from this research. The large, prospective, and nationally representative nature of the PATH Study enhanced generalizability and certainty regarding the temporal sequence between exposure and outcome. The PATH questionnaire was rigorously pilot tested and administered under a consistent protocol.14 Furthermore, the magnitude of associations found in this study was largely consistent across different specifications of e-cigarette and cigarette use. Among other study aspects to consider, in-home administration of the PATH questionnaire could have led to differences in estimated tobacco use compared with school-based surveys. However, results of this analysis were qualitatively similar to previous work in which NYTS data was used.13 As with all observational studies, residual confounding from unmeasured variables cannot be ruled out, although associations remained positive and at the threshold for statistical significance after adjustment for an extensive set of variables known to predict youth cigarette smoking.23
In July 2017, the US Food and Drug Administration announced a plan for tobacco and nicotine regulation that delayed federal e-cigarette regulation from 2018 until 2022.24 However, local governments have taken regulatory action of e-cigarettes. For example, a 2017 San Francisco, California, ordinance prohibits the sale of flavored tobacco products, including e-cigarettes, with the intention of reducing the appeal of tobacco products to youth.25 It is indicated in our results that among youth cigarette experimenters, those who have also used e-cigarettes are more likely to progress to current established smoking than those who tried cigarettes alone. As long as e-cigarettes remain attractive to youth, concern persists that these products contribute to greater combustible cigarette smoking among adolescents.
Conclusions
Among youth cigarette experimenters, using e-cigarettes was positively and independently associated with future onset of current established smoking, suggesting that e-cigarettes do not divert from, and may encourage, cigarette smoking in this population. In weighing the overall public health impact of e-cigarette availability, regulation, and use, the potential to increase combustible cigarette smoking by youth deserves special consideration.
- CI
confidence interval
- e-cigarette
electronic cigarette
- NYTS
National Youth Tobacco Survey
- OR
odds ratio
- PATH
Population Assessment of Tobacco and Health
Dr Chaffee contributed to the design and conceptualization of the study and the analysis plan, conducted statistical analyses, and prepared the initial manuscript draft; Dr Watkins contributed to the design and conceptualization of the study and the analysis plan and conducted statistical analyses; Dr Glantz contributed to the design and conceptualization of the study and the analysis plan; and all authors revised and reviewed the manuscript, approved the final manuscript as submitted, and agree to be accountable for all aspects of the work.
FUNDING: Supported in part by the US National Cancer Institute (T32CA113710), the US Food and Drug Administration Center for Tobacco Products (P50CA180890), and the National Institute on Drug Abuse (R01DA043950). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the US Food and Drug Administration. The funding agencies played no role in the conduct of the research or the decision to submit the article for publication. Funded by the National Institutes of Health (NIH).
References
Competing Interests
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
Comments
The Chaffee et al response to the critique of their paper is inadequate
The response of Chaffee et al to the comments of Rodu & Plurphanswat is inadequate, disappointing and surprising.
It is inadequate because Rodu & Plurphanswat have highlighted a serious flaw in the treatment of confounding in the original paper. It is clear that the propensity to become a smoker will be better characterised by the number of cigarettes consumed in the range one puff to 99 cigarettes by an experimenter rather by assuming that all subjects in this category have the same propensity to smoke, regardless of consumption.
The authors’ response, alleging that a ‘statistical trick’ has been played, does not make any sense at all: Rodu & Plurphanswat have used a straightforward statistical procedure and a more rigorous approach than the authors. Rodu & Plurphanswat show that when the relationship is adjusted for the amount smoked by experimenters, the association between vaping and the onset of established smoking disappears. In other words, many of those vaping would have smoked whether or not they had vaped. This reanalysis is fatal to authors’ main findings and invalidates their controversial and politically-charged conclusion that “e-cigarettes do not divert from, and may encourage, cigarette smoking in this population”.
It is disappointing because the authors, apparently with the approval of the journal, have resorted to an ad hominem attack on Rodu and without showing any improper conduct or incomplete disclosure, or that their new analysis exhibits any sort of bias or weakness.
It is surprising because a neutral reader would expect a journal of the stature of Pediatrics to retract or insist on correction of such an obviously flawed paper when the flaws have been so clearly and dispassionately pointed out.
In Reply: Electronic cigarettes, smoking progression, and statistical adjustment
Using longitudinal data from the PATH study, we demonstrated a positive and independent association between adolescents' use of e-cigarettes and progression from conventional cigarette experimentation (smoked >1 puff but not 100 cigarettes) to current established smoking (smoked ≥100 cigarettes) one year later.(1) Drs. Rodu and Plurphanswat criticized our conclusion because we did not adjust for lifetime history of cigarette consumption at baseline as a confounding variable. They demonstrated that including this variable in statistical models attenuated the association between e-cigarette use and progression to established smoking.
We agree that, in lieu of experimental or quasi-experimental data, epidemiologic investigations must make methodological choices regarding how to account for confounding and variable specification. Because it is a predictor of established smoking, number of cigarettes smoked is a conceivable candidate as a confounding variable. However, within the context of our stated research question, statistical adjustment is not appropriate.
We consider the number of cigarette smoked to be a mediator rather than a confounder. If trial of e-cigarettes leads youth to smoke more conventional cigarettes on the way to becoming established smokers, the number of cigarettes smoked is on a pathway between e-cigarettes and established smoking. Including a mediator as an independent variable, as Drs. Rodu and Plurphanswat suggested, will introduce bias toward a null association.(2)
Furthermore, the proposed confounder (number of cigarettes smoked at baseline, up to 100 cigarettes) too closely resembles the outcome variable itself (smoking 100 or more cigarettes). Adjustment for baseline cigarette consumption would yield a regression coefficient with an unusual interpretation: the association with smoking 100 or more cigarettes, holding constant the number of cigarettes smoked before reaching 100 cigarettes. The originally defined the outcome variable was a widely accepted threshold of smoking experience, and adjusting away the cumulative number of cigarettes contributing to reaching that threshold substantively changes the nature of the research question. Essentially, such an approach places a component of the dependent variable on both sides of the regression equation: akin to adjusting for confounding by systolic blood pressure in a study of hypertension.
For these reasons, the approach proposed by Drs. Rodu and Plurphanswat represents unnecessary and inappropriate statistical adjustment that will bias the analysis toward the null. We stand by our results as reported.
Revision notice:
The initial comment dated 3/30/2018 was revised on 4/24/2018 to focus specifically on the concerns raised in the comment by Drs. Brad Rodu and Nantaporn Plurphanswat.
References
1. Chaffee BW, Watkins SL, Glantz SA. Electronic Cigarette Use and Progression From Experimentation to Established Smoking. Pediatrics. 2018;141(4).
2. Babyak MA. Understanding confounding and mediation. Evid Based Ment Health. 2009;12(3):68-71.
RE: Electronic Cigarette Use and Progression From Experimentation to Established Smoking
Using data from the first two waves of the Population Assessment of Tobacco and Health survey, Chaffee et al. (1) conducted a study finding that “Among adolescent cigarette experimenters [never smoked 100 cigarettes, in Wave 1], using e-cigarettes was positively and independently associated with progression to current established smoking [in Wave 2]…” Outcomes were having smoked 100+ cigarettes, smoking in the past 30 days, and the combination (established smoking), according to never, past 30-day and experimental e-cigarette users at Wave 1.
Logistic regression models were adjusted for standard variables, including sex, age, race/ethnicity, and for other factors related to smoking (household tobacco use, alcohol use, tobacco ads receptivity, sensation-seeking, cigarette warning label exposure and ever use of other tobacco products). But the authors omitted lifetime cigarette consumption (LCC) at Wave 1, which is well known to be a profoundly important variable relevant to adolescent progression from experimental to established smoking.(2)
To demonstrate the importance of this variable, we reproduced Chaffee’s unadjusted odds ratios (ORs), and then we added the cigarette consumption variable (one or more puffs but never a whole cigarette, one cigarette, 2-10, 11-20, 21-50 and 51-99 cigarettes).
The results in the table demonstrate that all of Chaffee’s elevated ORs in Wave 2 for e-cigarette users are substantially reduced and not statistically significant after adjusting for cigarette consumption. Our analysis demonstrates that Chaffee’s model is inadequate, and the conclusion that e-cigarette use is “independently associated” with progression to smoking is incorrect. Substantial revision or retraction of this study is warranted.
Brad Rodu
Nantaporn Plurphanswat
References
1. Chaffee BW, Watkins SL, Glantz SA. Electronic cigarette use and progression from experimentation to established smoking. Pediatrics 2018; 141: https://doi.org/10.1542/peds.2017-3594 </p>
2. Choi WS, Pierce JP, Gilpin EA, Farkas AJ, Berry CC. Which adolescent experimenters progress to established smoking in the United States. Am J Prev Med 1997; 13: 385-391
Odds Ratios for Smoking Outcomes in the Wave 2 PATH Survey, According to E-Cigarette Use at Wave 1
BR-NP, Rodu-Plurphanswat reproduction of Chaffee et al. (1) odds ratios
LCC, Lifetime Cigarette Consumption
*Trier = Former smoker in Chaffee et al. (1)
National Academies of Sciences, Engineering and Medicines’ (NASEM) comprehensive review on Electronic cigarette and health
A new congressionally mandated National Academies of Sciences, Engineering and Medicines’ (NASEM) comprehensive review of more than 800 peer reviewed scientific studies on health effects of Electronic cigarette (ECs) on adolescents concluded that “There is substantial evidence that e-cigarette use by youth and young adults increases their risk of ever using conventional cigarettes.”(1) Therefore, the NASEM recommendation should appropriately reflect in the future pediatric e-cigarette clinical guidelines, in order to protect our adolescents from e-cigarette use.
Meanwhile, a recent study shows that few adolescents try to quit smoking using any sort of cessation aid or professional intervention. Importantly, this study also highlight that the efficacy of e-cigarettes in improving smoking cessation in youth is equivocal (2).
In December 2016, the US Surgeon General released a comprehensive action plan, based on best evidence information, to address how to protect our adolescents from e-cigarette use. This e-cigarette prevention plan includes implementing comprehensive educational campaigns tailored for young people (3).
References:
(1). National Academies of Sciences and Engineering and Medicine. Committee on the Review of the Health Effects of Electronic Nicotine Delivery Systems. Public health consequences ofe-cigarettes. (accessed Feb 24, 2018).The National Academies Press, Washington, DC; 2018. https://www.nap.edu/catalog/24952/public-health-consequences-of-e-cigare...
(2). Wolfenden L, Stockings E, Yoong SL. Regulating e-cigarettes in Australia: implications for tobacco use among young people. Med J Aust 2017; 208 (1):1. doi: 10.5694/mja17.00787
(3). Surgeon General’s report (2016) US Department of Health and Human Services. E-cigarette use among youth and young adults: A report of the Surgeon General-http://e-cigarettes.surgeongeneral.gov/documents/2016_SGR_Full_Report_no....
RE: Electronic cigarette and adolescents’ health
Recent nationally representative Population based study found that teens who use e-cigarettes are more likely to continue using e-cigarettes. Moreover, this study further suggests that e-cigarettes use is also likely to encourage, combustible cigarettes smoking in this population (1).
Aaccording to US Center for Disease Control (CDC) every day, more than 3200 adolescents under the age of 18 experiments with their first cigarette and an additional 2100 go on to become daily smokers. In fact, over 95% of adults who are or were daily smokers tried their first cigarette before age 21 (2). Therefore, empowering adolescents to avoid unhealthy behavior is paramount, as we believe it is an important social and public health responsibility. This would help reduce the burdens of disease during adulthood and associated economic costs in the future.
Pediatricians can ensure that adolescents are protected from commercial interests and are able to make informed decisions based on strong evidence and scientific fact. Globally there is an increasing trend of harmful e-cigarette use by adolescents (3, 4).
Meanwhile, the Commissioner of the US Food and Drug Administration (FDA) Dr. Scott Gottlieb is attempting to differentiate e-cigarettes from traditional tobacco by pointing to the lack of evidence for the harm or risks of e-cigarettes. It openly promotes e-cigarettes as a valid smoking cessation and harm reduction tool, even though there is no sufficient evidence to support this claim (5).
Adolescents are very puzzled over the contradictory information provided by social media, the e-cigarette industry, the widely divided stance in the academic community and the FDA. The online propaganda for e-cigarettes is so powerful that adolescents simply cannot resist. There is no doubt that comprehensive global prevention programs targeting adolescents are urgently needed. In fact, we urge that American Academy of Pediatrics (AAP) take a leadership role to counteract the e-cigarette and the tobacco industry sponsored popular online media myth spreading machine.
As we know that doctors are one of the most trusted sources of health information in the new-media environment, pediatricians need to explore the Internet and social media platforms, as a viable medium for communicating with their adolescent patients. Through such influence it potentially allows the usage of e-cigarettes amongst adolescents to reduce. We thank AAP for continued advocacy in support of the adolescent health and well-being.
REFERENCES:
(1). Chaffee BW, Watkins SL, Glantz SA. Electronic Cigarette Use and Progression From Experimentation to Established Smoking- http://pediatrics.aappublications.org/content/early/2018/03/01/peds.2017...
(2). CDC-(Center for Disease Control and Prevention) 2016. Youth and Tobacco Use. Retrieved from: http://www.cdc.gov/tobacco/data_statistics/fact_sheets/youth_data/tobacc...
(3). Barrington-Trimis JL, Urman R, Berhane K, Unger JB, Cruz TB, Pentz MA, Samet JM, Leventhal AM, McConnell R.E-cigaratte and future cigarette use. Pediatrics. 2016 Jul;138(1). pii: e20160379. doi: 10.1542/peds.2016-0379. Epub 2016 Jun 13.
(4). Park JY, Seo DC, Lin HC. E-cigarette use among US youths is associated with intention to smoke but not with intention to quit smoking.Am J Public Health. 2016 Apr;106(4):672-8. doi: 10.2105/AJPH.2015.302994. Epub 2016 Jan 21.
(5). Gottlieb S, Zeller M. A nicotine-focused framework for public health. N Engl J Med 2017 (Epub ahead of print: 16 Aug 2017).doi:10.1056/NEJMp1707409
(1). Aki Nilanga Bandara
Founding Chair-Global coalition to empower adolescent and youth on harmful therapeutic interventions to prevent combustible tobacco use.
Undergraduate student, University of British Columbia, British Columbia, Canada, 2357 Main Mall, Vancouver, BC V6T 1Z4 Canada.