Conjoint Developmental Trajectories of Adolescent E-cigarette and Combustible Cigarette Use

Participants were adolescents in the ninth grade taking part in a longitudinal cohort study of the relationship among combustible cigarette smoking, e-cigarette use, and other tobacco use. Participants were enrolled in 1 of 4 public high schools in suburban Philadelphia, Pennsylvania. On the basis of published demographic data, the schools were selected such that our sample would be demographically representative of adolescents nationwide (sex, race, ethnicity, and annual household income). Demographic data in Table 1 provide support for a representative sample.

The cohort participants were drawn from 2198 students identified through class rosters at the start of ninth grade. Adolescents were ineligible to participate if they had a severe learning disability or did not speak fluent English. On the basis of the selection criteria, a total of 2017 of 2198 (92%) students were eligible to participate.

Parents were mailed a study information letter (active information) with a telephone number to call to obtain answers to any questions and to decline consent for their adolescent to participate (passive consent). Of the 2017 eligible adolescents, 17 (1%) had a parent who actively declined their adolescent’s participation. Adolescents with parental consent were approached to provide their written assent for study participation. Adolescents who were absent on the assent and baseline survey days (n = 124, 6%) and adolescents who did not provide assent (n = 41, 2%) because of lack of interest were not enrolled in the cohort. Thus, 1835 of the 2000 adolescents with consent (92%) provided their assent to participate and completed a 40-minute paper and pencil survey. This baseline survey, was completed on-site during compulsory classes in the fall of 2016.

Adolescents completed 6 paper and pencil follow-up surveys at 6-month intervals, with 92% completing a survey at wave 2 (n = 1687, spring 2017), 90% completing a survey at wave 3 (n = 1658, fall 2017), 89% completing a survey at wave 4 (n = 1643, spring 2018), 87% completing a survey at wave 5 (n = 1601, fall 2018), 84% completing a survey at wave 6 (n = 1538, spring 2019), and 83% completing a survey at wave 7 (n = 1530, fall 2019). The participants included in this study were adolescents who had complete data on the study variables at baseline (N = 1808). The Institutional Review Board of the University of Pennsylvania and the administration of each of the 4 high schools approved the study. Data analyses were conducted in January 2021.

The survey included an introduction explaining the types of products and devices labeled as e-cigarettes. Images of different e-cigarette devices were provided to facilitate clarity.^20,21  From baseline (wave 1) to wave 3, these images included e-cigarettes, e-hookah, vape pens, and mods. Images of USB-style pod vaporizers were added at wave 4. Excluding using an e-cigarette device for vaping marijuana, adolescents were asked, “Have you ever used an e-cigarette like the ones pictured above, even 1 or 2 times?” Adolescents who reported ever use of an e-cigarette were prompted to answer subsequent questions assessing lifetime frequency of e-cigarette use and time since last e-cigarette use. An ordered categorical variable defined progression in e-cigarette use as follows: 0, never used; 1, used, but not in the past 6 months; 2, used in the past 6 months; and 3, used in the past 30 days. Current use was defined as using an e-cigarette on at least 1 day in the past 30 days.^22,23  E-cigarette use was measured in all 7 waves.

Cigarette smoking was measured by asking adolescents, “Have you ever tried smoking a cigarette, even a few puffs?” Adolescents who reported ever smoking a cigarette were prompted to answer questions assessing the lifetime frequency of cigarette smoking and time since last cigarette use. An ordered categorical variable defined cigarette smoking progression as follows: 0, never smoked; 1, smoked, but not in the past 6 months; 2, smoked in the past 6 months; and 3, smoked a cigarette in the past 30 days. Current use was defined as smoking on at least 1 day in the past 30 days.²²  Cigarette smoking was measured in all 7 waves.

Demographic variables included sex, race (White, Black, or other), ethnicity (Hispanic or non-Hispanic), and socioeconomic status. Receipt of free or reduced-cost lunch was used as an indicator of socioeconomic status because adolescent reports of economic status are unreliable.^24–27  This variable was assessed by asking adolescents if they receive free or reduced-price lunch (yes or no). In addition to the risk factors described below, these demographic variables were included in the statistical model to characterize adolescents who had a particular pattern of e-cigarette and combustible cigarette use.

Peer e-cigarette use was measured by asking adolescents whether their best friend and how many of their 4 other male best friends and 4 other female best friends used e-cigarettes.^28,29  Household e-cigarette use was measured with the question “Does anyone in your house use e-cigarettes?” (none versus at least 1).²⁸  Peer and household cigarette smoking were measured similarly to peer and household e-cigarette use.^30,31  Access to e-cigarettes was measured with the question “Do you think it would be difficult to get e-cigarettes if you wanted them?” Response options were the following: 0, yes, extremely difficult; 1, somewhat difficult; and 2, no, not difficult at all. Access to cigarettes was measured in the same manner.

E-cigarette risk perceptions (relative to combustible cigarettes) were measured with 2 items derived from previous research (ie, e-cigarettes might be less harmful for people to be around than cigarettes, e-cigarettes might be less addictive than cigarettes; 0 [strongly disagree] to 3 [strongly agree]).^32–34  Positive expectations of e-cigarette use and cigarette smoking were each measured with a 9-item Likert-style scale (0 [strongly disagree] to 3 [strongly agree]).^35–38  The items included “I think vaping e-cigarettes (smoking cigarettes) would…give me something to do when I’m bored, …help me deal with problems or stress, …feel more comfortable at parties.” The 20-item Centers for Epidemiology Studies of Depression scale²⁴  was used to assess depression symptoms over the past week (0 [rarely or none of the time] to 3 [most of the time]).^31,39  Sensation-seeking was measured with the 8-item Brief Sensation Seeking Scale (0 [strongly disagree] to 4 [strongly agree]).⁴⁰ 

Adolescents were asked if they had ever smoked a large cigar, little cigar, or cigarillo and, if so, whether they smoked in the past 6 months.⁴¹  Responses were coded as 1 for using any type of cigar and 0 for no use. Marijuana use and alcohol use were assessed by asking adolescents if they had ever used these substances.²²  Responses were coded as 1 for ever use and 0 for never use.

First, we sought to determine the number of latent classes of cigarette smoking and e-cigarette use with a sequential processes growth mixture model. A sequential processes growth mixture model is a latent variable mixture modeling method that identifies latent (unobserved) classes representing the joint development of 2 behaviors across time through repeated measures.⁴²  The modeling began with assessing the average growth trajectory of cigarette smoking and the average growth trajectory of e-cigarette use with latent growth curve models. Then latent variable mixture modeling was used to identify the optimal number of conjoint latent classes based on cigarette smoking and e-cigarette use patterns. Given 2 processes, the total number of latent classes reflects the product of the number of proposed latent classes for each behavior. We began with 2 classes for cigarette smoking and 2 latent classes for e-cigarette use, resulting in 4 possible latent classes. The Bayesian information criterion (BIC) was used to identify the optimal number of conjoint classes, with lower BIC values indicating a more optimal model.⁴³ 

Second, the latent classes were treated as a dependent variable in a multinomial logistic regression to assess the likelihood of belonging to each specific conjoint class compared with the comparison class for a unit increase in each predictor variable. Mplus 8.3 was used to identify the conjoint trajectories empirically, and SPSS (IBM SPSS Statistics, IBM Corporation) was used to characterize class membership through multinomial logistic regression analysis.

Four conjoint latent classes were identified. The log likelihood and BIC values for the 4 conjoint models were −8497.55 and 17 168.01, respectively. A 6 conjoint-class model was tested to determine if it was a better representation of the latent classes. The log likelihood (−9937.62) and BIC (20 123.26) suggested that 4 classes better represented the data. On the basis of the pattern of results (see Table 2, Fig 1), the 4 latent classes were labeled as later, rapid e-cigarette uptake (class 1: n = 230); no use of e-cigarettes or combustible cigarettes (class 2: n = 1141); earlier, steady e-cigarette uptake (class 3: n = 265); and dual use of e-cigarettes and combustible cigarettes (class 4: n = 204). Table 1 presents descriptive statistics for the total sample and the sample divided by class.

Table 3 presents the multinomial logistic regression results with the odds ratio (OR) and 95% confidence interval (CI) of being in each class compared with dual use (class 4).

Being female was associated with more than a twofold increase in the odds of belonging to class 1 versus class 4 (OR 2.34; 95% CI 1.41–3.88). Peer e-cigarette use (OR 0.81; 95% CI 0.69–0.95), household smoking (OR 0.48; 95% CI 0.28–0.81), cigar use (OR 0.10; 95% CI 0.02–0.45), and marijuana use (OR 0.31; 95% CI 0.16–0.62) were associated with a 19%, 52%, 90%, and 69% decrease in the odds of being in class 1 compared with class 4, respectively. More positive e-cigarette expectations were associated with a 19% decrease in the odds of belonging to class 1 versus class 4 (OR 0.81; 95% CI 0.75–0.87).

Peer e-cigarette use (OR 0.76; 95% CI 0.67–0.86), peer smoking (OR 0.81; 95% CI 0.71–0.93), and household smoking (OR 0.42; 95% CI 0.27–0.67) were associated with a 24%, 19%, and 58% decrease in the odds of being in class 2 compared with class 4, respectively. Similarly, cigar use (OR 0.08; 95% CI 0.02–0.25), marijuana use (OR 0.24; 95% CI 0.14–0.43), and alcohol use (OR 0.54; 95% CI 0.33–0.90) were associated with a 92%, 76%, and 46% decrease in the odds of belonging to class 2 versus class 4, respectively. Higher depression (OR 0.97; 95% CI 0.95–0.99), greater sensation-seeking (OR 0.96; 95% CI 0.92–0.99), and more positive e-cigarette expectations (OR 0.79; 95% CI 0.73–0.84) were associated with a 3%, 4%, and 21% decrease in the odds of belonging to class 2 versus class 4, respectively.

Being female was associated with more than a twofold increase in the odds of belonging to class 3 versus class 4 (OR 2.46; 95% CI 1.55–3.90). Peer e-cigarette use was associated with a 10% increase in the odds of belonging to class 3 versus class 4 (OR 1.10; 95% CI 1.01–1.22), whereas peer smoking, cigar use, and marijuana use were associated with a 20% (OR 0.80; 95% CI 0.71–0.91), 61% (OR 0.39; 95% CI 0.21–0.75), and 42% (OR 0.58; 95% CI 0.34–0.99) decrease in the odds of belonging to class 3 versus class 4, respectively. More positive combustible cigarette expectations was associated with an 8% decrease in the odds of belonging to class 3 versus class 4 (OR 0.92; 95% CI 0.87–0.97). As cigarette access became easier, the odds of belonging to class 3 versus class 4 lessened (OR 0.64; 95% CI 0.45–0.91), but as e-cigarette access became easier, the odds of being in class 3 versus class 4 increased (OR 2.22; 95% CI 1.50–3.28).

This study provides new prospective evidence for distinct patterns and profiles of adolescents who progress to current e-cigarette use. Two e-cigarette–only trajectories were identified, differing in onset and rate of progression, with neither showing evidence for combustible cigarette smoking. A dual use trajectory identified adolescents who were initially cigarette smokers who progressed to current e-cigarette use.

Although adolescents in the earlier, steady e-cigarette uptake class shared a pattern of earlier e-cigarette onset with adolescents in the dual use class, they had a greater number of e-cigarette–using peers, easier access to e-cigarettes, more difficulty accessing combustible cigarettes, fewer smoking peers, and less positive expectations of cigarette smoking, and were less likely to have ever used marijuana or smoked a cigar. The presence of e-cigarette risk factors and the relative absence of cigarette smoking risk factors likely accounts for the steady increase in current e-cigarette use from 18% to 35% across the following 36 months without concurrent cigarette smoking. The findings also highlight the need for evidence-based approaches before and during high school to reduce e-cigarette uptake and promote e-cigarette cessation, especially for girls who are more likely to make up the 2 e-cigarette classes. Approaches will require attention to the factors that foster e-cigarette use, rather than simply adapting cigarette smoking interventions.^44,45  Given that 20% of adolescents report current e-cigarette use, prevention interventions should be implemented across settings within the community, schools, and pediatric primary care.⁴⁶ 

Adolescents in the later, rapid e-cigarette uptake class had relatively low levels of risk factors, including fewer peers who used e-cigarettes, less positive expectations for use, and lower lifetime rates of other substance use. Although adolescents in this group had a later e-cigarette onset, the prevalence of past-30-day e-cigarette use doubled every 6 months, beginning at age 16, reaching 55% over the following 18 months. This period of time coincided with the popularity of pod-style e-cigarettes with efficient delivery of nicotine in higher doses.⁴⁷  The rapid rate of uptake in the context of a low-risk profile highlights the importance of product regulation and screening for e-cigarette use in the absence of risk factors. It is important to note that the risk factors were only measured at baseline, and we cannot rule out the possibility that the risk profile increased across the first 2 years of high school. Adolescents with rapid e-cigarette uptake may be difficult to identify early and may be at greater risk for nicotine dependence,⁴⁸  emphasizing the importance of e-cigarette prevention efforts mid through late adolescence.

Dual users have been conceptualized as a homogenous group with less attention to their pathway to dual use.⁴⁹  Adolescents who use e-cigarettes and then become regular cigarette smokers will likely have different risk factors than adolescents who smoke combustible cigarettes and become regular e-cigarette users. We found little support for the former pathway to dual use and evidence suggestive of the latter. At age 14, 19% of adolescents were current e-cigarette users, and 16% were current combustible cigarette smokers in the dual use group. Prevalence in e-cigarette use increased to 39% across the following 36 months, but there was modest growth in combustible cigarette smoking, from 16% to 19%. Adolescents who currently used both tobacco products initiated cigarette smoking at age 12 and e-cigarette use at age 13. Adolescents who smoke cigarettes may use e-cigarettes as a substitute for cigarettes in smoke-free situations, such as school,^50–52  which may promote smoking persistence.⁵³ 

As has been observed in previous research,^38,54,55  adolescents who dual use e-cigarettes and combustible cigarettes tend to have a greater number and severity of risk factors. Compared with adolescents who made up the e-cigarettes–only classes (ie, single use), adolescents in the dual use group tended to be male, use other combustible tobacco products (such as cigars), and have the highest level of lifetime marijuana use. Although these groups did not significantly differ in sensation-seeking or depression, the differences tended to reflect access, peer use, and expectations for e-cigarettes and cigarettes. Because this group has risk factors for e-cigarette use and combustible cigarette smoking at age 14 years, early prevention and intervention programs will need to address both to be successful.

In contrast, adolescents in the no use class differed most from the adolescents in the dual use class. These differences included fewer risk factors and, in most instances, the lowest level of any risk factor. These adolescents made up 62% of the sample.

As the first study to examine the unique patterns and predictors of e-cigarette and combustible cigarette uptake among adolescents, the study has strengths and limitations. Strengths include a diverse sample of adolescents measured during a vulnerable period for tobacco use, excellent participation and retention rate, modeling of e-cigarette use and cigarette smoking across 7 time points, and the inclusion of a rich set of risk factors relevant to both e-cigarette and combustible cigarette use. One potential limitation is that risk factors were only modeled at baseline, and these variables may have changed across time. Nevertheless, these findings inform opportunities for prevention and intervention based on distinct patterns of uptake and their unique risk factors.

BIC

Bayesian information criterion

CI

confidence intervale-cigarette, electronic cigarette

OR

odds ratio