Longitudinal Trajectories of E-Cigarette Use among Adolescents: A 5-year, Multiple Cohort Study of Vaping with and without Marijuana

Abstract

E-cigarette use harms adolescent health, yet it continues to escalate rapidly among teens nationwide. This longitudinal study sought to identify and differentiate between developmental trajectories of past 30-day e-cigarette use with and without marijuana (i.e., liquid THC) across adolescence (11–19 years old). Three population-based cohorts of adolescents (n=3,907; N=461,069) living in major metropolitan areas of Texas (Houston, Dallas-Ft. Worth, San Antonio, Austin) completed up to 9 Waves of an e-cigarette use survey, from 2014 to 2019. Growth curve models (GCMs) were used to identify average trajectories of past 30-day e-cigarette use, by cohort. Growth mixture models (GMMs) were used to investigate developmental patterns in these trajectories, by cohort. Sociodemographic differences in trajectories were also investigated. Stable trajectories of e-cigarette use with and without marijuana were identified, from 11 through 19 years of age. Trajectories varied by age of onset; frequency and escalation in use; and substance used. With one exception, all trajectories of e-cigarette use escalated with age. Moreover, age of onset and progression in use were positively related. The most problematic trajectories, corresponding to more frequent use, were observed among the younger cohorts compared to the oldest. Primary prevention is critical. Interventions to prevent the onset and progression in e-cigarette use among teens must begin early (e.g., in middle school) and be sustained throughout adolescence.

INTRODUCTION

In 2019 alone, over 5 million students nationwide reported using e-cigarettes at least once in the past 30-days, and nearly 1 million used them daily.^1,2 Sustained, substantial use among teens alarms physicians and public health professionals alike,^3,4 as it harms health in many ways.^4–6 A recent outbreak of acute, severe lung injury in 2019 (EVALI; E-cigarette & Vaping Associated Lung Injury) was especially worrisome, as >50% of cases occurred among those under the age of 25.^7–9 While most e-cigarette devices were designed to vaporize and deliver a nicotinized or denicotinized e-liquid, e-cigarettes can also be used to consume liquid THC (Δ-9-tetrahydrocannabinol), the main psychoactive ingredient in marijuana.¹⁰ Use of devices with and without THC were implicated in regards to EVALI.¹¹ E-liquids and e-cigarette aerosols contain carcinogens (e.g., formaldehyde) and heavy metals (e.g., lead) that damage the lungs and body.¹² Vaping nicotine increases teens’ risk for dependence¹³ and alters young people’s mood, memory, and learning.^14,15 E-cigarette use is also known to predict the onset of cigarette smoking among youth and young adults.^16,17

Unfortunately, the exponential increase in e-cigarette use among teens observed over the last decade has largely reversed the strong decline in overall tobacco use achieved the decade before.^1,18,19 Therefore, contemporary approaches to comprehensive tobacco control will need to successfully address and reduce e-cigarette use among young people. Constructing effective preventive interventions, however, will require a better understanding of the natural history of e-cigarette use behaviors. To date, most e-cigarette research has been cross-sectional in design^16,20–22 or, if longitudinal, has relied on a limited amount of data time points.^17,23 This, in turn, restricts our ability to fully understand the onset and progression in e-cigarette use behaviors in adolescence. As this developmental stage is often characterized by volatility in health behaviors,^24,25 there is a need to better elucidate stable, persistent patterns in e-cigarette use.^20,26–28 Such an understanding would help identify sustained and especially problematic e-cigarette use, and particular age(s) which it would be most relevant to target, with interventions.

Previous research specific to the natural history of cigarette smoking among young people shows it is possible to identify stable patterns or trajectories of behavior.^29–31 While some studies focus on understanding the average shape or form of a trajectory across adolescence,³² others identify clusters of trajectories to describe meaningful heterogeneity in developmental patterns of behavior.^29–31 Depending on the population under study, the latter approach, when applied to cigarette smoking, has identified sub-groups of young people who can be characterized as non-users; triers; erratic and/or occasional users; early escalators; late escalators; or quitters.^29–32 It is unclear whether similar patterns of e-cigarette use exist. The goal of this study, therefore, is to identify and differentiate between developmental trajectories of e-cigarette use, with and without marijuana (i.e., liquid THC). We present analyses of five years (2014–2019) of longitudinal, repeated measures data collected every 6 months specific to past 30-day e-cigarette use. These data span the entirety of adolescence (11–19 years old). This study builds upon emerging evidence-based perspectives on developmental trajectories of e-cigarette use with and without marijuana, given these goals, which are unique in the literature to date.^33–37

METHODS

Study design, setting, and participants

Data were collected from 3 population-based cohorts of adolescents living in major metropolitan areas of Texas (Houston, Dallas-Ft. Worth, San Antonio, Austin) who were in the 6^th (n=1,122, N=148,464), 8^th (n=1,322, N=160,083), and 10^th (n=1,463, N=152,522) grades at baseline. Each cohort is part of the Texas Adolescent Tobacco & Marketing Surveillance Study (TATAMS).³⁸ TATAMS used a complex, multi-stage, probability sampling design to recruit and enroll participants. All students in the 6^th, 8^th, and 10^th grades in Fall 2014 in 79 participating schools in these areas were eligible for the study.³⁸ More details about sampling methodology are found elsewhere.³⁸ On enrollment, students participated in a web-based survey in school about e-cigarette use, administered via tablet computers.³⁹ Approximately every 6 months thereafter, participants took the same web-based survey outside of school on their own tablet, computer, or smartphone. Longitudinal, repeated measures data from Wave 1 (Fall, 2014) through Wave 9 (Spring, 2019) are presented here. Retention across Waves ranged from 63–85%, on par with similar longitudinal studies of e-cigarette use among young people nationwide.⁴⁰ In our study, retention did not vary by cohort, at any Wave. Active parent consent and student assent were obtained for each participant (HSC-SPH-13-0377). After completing Wave 1, all participants were sent reminder e-mails and text messages about other surveys. To ensure participant retention, incentives were increased from $10 for Waves 1–2 to $25 for Waves 3–9.

The sampling frame for TATAMS included all private, public, and charter schools in these major metropolitan areas to provide a sample representative of student enrollment. Per data from the Texas Education Agency, the Texas Private School Accreditation Commission, and the National Center for Education Statistics, the sampling frame accounted for 97% of 6^th, 8^th, and 10^th grade students in these areas, and more than 40% of all 6^th, 8^th, and 10^th grade students in all of Texas in 2014–15.³⁸ At Wave 1, the sample was 48.9% female; 54.5% Hispanic, 21.4% non-Hispanic White, and 17.6% non-Hispanic Black; and the average age of each cohort was 11, 13, and 14 years old, respectively.

Measures

All questions on the survey were adapted from reliable, valid measures employed in national e-cigarette and marijuana use surveillance studies: the Population Assessment of Tobacco and Health study (PATH),⁴¹ the National Youth Tobacco Survey,⁴² and Monitoring the Future.⁴³ Primary outcome variables included past 30-day e-cigarette use with and without marijuana (i.e., liquid THC). Questions about e-cigarette use with marijuana were asked before questions without it. Other variables in analysis included age, sex, race/ethnicity, and SES.

Past 30-day E-cigarette Use without Marijuana.

At each Wave, students who reported ever e-cigarette use were asked “During the past 30-days, on how many days did you use an electronic cigarette, vape pen, e-hookah, MOD or tank system? Remember, marijuana DOES NOT count. Please enter the number of days (from 0 to 30 days).” Students that reported 0 days were considered non-users (referent, coded as 0) and those that reported 1 or more days were users (coded as 1). Past 30-day e-cigarette use without marijuana was assessed at all Waves.

Past 30-day E-cigarette Use with Marijuana.

At each Wave, students who reported ever marijuana use in an e-cigarette were asked “During the past 30-days, on how many days did you smoke marijuana (liquid THC) from an electronic cigarette, vape pen, or e-hookah? Please enter the number of days (from 0 to 30 days).” Students that reported 0 days were considered non-users (referent, coded as 0) and those that reported 1 or more days were users (coded as 1). Past 30-day e-cigarette use with marijuana (THC) was only assessed at Waves 3 through 9.

Statistical Analysis

First, we examined average trajectories of past 30-day e-cigarette use for each substance and cohort separately, then evaluated whether these trajectories overlapped at the same age(s) between cohorts. Growth curve models (GCMs) were applied to repeated measures data across all available Waves using generalized estimating equations (GEEs) with a log-link function. Log-linear models were employed, as they fit both e-cigarette use outcome variables best.^44–46 After average trajectories were identified, ageXcohort interaction effects were investigated using procedures described by Miyazaki & Raudenbush.⁴⁷ The quasi-likelihood-under-independence-model criterion (QIC) was used to test for differences in trajectories by cohort. If significant overlaps occur, data from multiple age cohorts (like these) could be combined into a single cohort for analysis using a cohort-sequential design.⁴⁷ We did not find this to be the case, here.

Next, we investigated whether the average trajectories for each cohort could be “split” into multiple trajectories to describe meaningful sub-groups of adolescents who followed a similar developmental pattern in past 30-day e-cigarette use with increasing age. To do so, growth mixture models (GMMs) were applied. Methods described by Nagin^48–50 and employed by others to identify developmental patterns in young people’s cigarette smoking behavior^29–31 were utilized. Models representing 2 to 7 sub-groups were considered; the model with the lowest BIC score and the highest posterior probability was chosen for each variable.^49,50 Multivariate logit models were used to test for differences between these patterns, by sex, race/ethnicity, and SES; the non-user trajectory group served as referent (Table 1, Table 2).

Table 1.

Differences in developmental patterns of e-cigarette use, by cohort and sociodemographic factors^b–f; the Texas Adolescent Tobacco and Marketing Surveillance System (TATAMS), Waves 1–9, 2014–2019

Table 1a. Past 30-day E-cigarette Use without Marijuana (Waves 1–9, 2014–2019; n=3,878)
	Pattern 1 Non-users		Pattern 2 ‘Early’ escalators		Pattern 3 ‘Mid’ escalators		Pattern 4 ‘Late’ escalators
	%a	95% CI	%a	95% CI	%a	95% CI	%a	95% CI	p-value
6th grade cohort
n	1013	--	28	--	68	--
% of cohort	91%		3%		6%
Gender									0.533
Female	60.5	(57.4 – 63.5)	50.0	(30.6 – 69.4)	60.3	(47.7 – 72.0)
Male	39.5	(36.5 – 42.6)	50.0	(30.6 – 69.4)	39.7	(28.0 – 52.3)
Race/ethnicity									0.196
Non-Hispanic White	30.2	(27.4 – 33.1)	25.0	(10.7 – 44.9)	42.6	(30.7 – 55.2)
Non-Hispanic Black	13.6	(11.6 – 15.9)	10.7	(2.3 – 28.2)	4.4	(0.9 – 12.4)
Hispanic	40.9	(37.8 – 44.0)	42.9	(24.5 – 62.8)	38.2	(26.7 – 50.8)
Other	15.3	(13.1 – 17.7)	21.4	(8.3 – 41.0)	14.7	(7.3 – 25.4)
SES									<0.001
Low	11.8	(9.9 – 14.0)	46.4	(27.5 – 66.1)	11.8	(5.2 – 21.9)
Middle	63.1	(60.0 – 66.1)	42.9	(24.5 – 62.8)	55.9	(43.3 – 67.9)
High	25.1	(22.4 – 27.9)	10.7	(2.3 – 28.2)	32.4	(21.5 – 44.8)
8th grade cohort
n	1025		108	--	146	--	34	--
% of cohort	78%		8%		11%		3%
Gender									0.598
Female	54.0	(50.8 – 57.0)	52.8	(42.9 – 62.5)	48.6	(40.3 – 57.0)	58.8	(40.7 – 75.4)
Male	46.0	(43.0 – 49.2)	47.2	(37.5 – 57.1)	51.4	(43.0 – 59.7)	41.2	(24.6 – 59.3)
Race/ethnicity									<0.001
Non-Hispanic White	33.6	(30.7 – 36.5)	30.6	(22.1 – 40.2)	52.7	(44.3 – 61.1)	44.1	(27.2 – 62.1)
Non-Hispanic Black	15.9	(13.7 – 18.3)	13.9	(8.0 – 21.9)	4.1	(1.5 – 8.7)	0.0	(0.0 – 10.3)
Hispanic	36.1	(33.2 – 39.1)	41.7	(32.3 – 51.5)	28.8	(21.6 – 36.8)	32.4	(17.4 – 50.5)
Other	14.4	(12.3 – 16.7)	13.9	(8.0 – 21.9)	14.4	(9.1 – 21.1)	23.5	(10.7 – 41.2)
SES									0.747
Low	14.3	(12.2 – 16.6)	13.9	(8.0 – 21.9)	13.0	(8.0 – 19.6)	11.8	(3.3 – 27.5)
Middle	61.6	(58.5 – 64.6)	57.4	(47.5 – 66.9)	56.8	(48.4 – 65.0)	64.7	(46.5 – 80.3)
High	24.1	(21.6 – 26.9)	28.7	(20.4 – 38.2)	30.1	(22.8 – 38.3)	23.5	(10.7 – 41.2)
10th grade cohort
n	966		90	--	246	--	154	--
% of cohort	66%		6%		17%		11%
Gender									<0.001
Female	58.5	(55.3 – 61.6)	34.4	(24.7 – 45.2)	50.4	(44.0 – 56.8)	55.2	(47.0 – 63.2)
Male	41.5	(38.4 – 44.7)	65.6	(54.8 – 75.3)	49.6	(43.2 – 56.0)	44.8	(36.8 – 53.0)
Race/ethnicity
Non-Hispanic White	24.0	(21.4 – 26.8)	45.6	(35.0 – 56.4)	30.9	(25.2 – 37.1)	33.8	(26.4 – 41.8)	<0.001
Non-Hispanic Black	22.7	(20.1 – 25.4)	14.4	(7.9 – 23.4)	13.0	(9.1 – 17.9)	14.3	(9.2 – 20.8)
Hispanic	39.3	(36.2 – 42.5)	31.1	(21.8 – 41.7)	43.1	(36.8 – 49.5)	35.7	(28.2 – 43.8)
Other	14.0	(11.8 – 16.3)	8.9	(3.9 – 16.8)	13.0	(9.1 – 17.9)	16.2	(10.8 – 23.0)
SES									0.238
Low	17.3	(14.9 – 19.8)	24.4	(16.0 – 34.6)	21.1	(16.2 – 26.8)	24.7	(18.1 – 32.3)
Middle	64.1	(61.0 – 67.1)	60.0	(49.1 – 70.2)	61.0	(54.6 – 67.1)	60.4	(52.2 – 68.2)
High	18.6	(16.2 – 21.2)	15.6	(8.8 – 24.7)	17.9	(13.3 – 23.3)	14.9	(9.7 – 21.6)

Table 1b. Past 30-day E-cigarette Use without Marijuana (Waves 3–9, 2015–2019; n=2,733)
	Pattern 1 Non-users		Pattern 2 Escalators
	%a	95% CI	%a	95% CI	p-value
6th grade cohort
n	729	--	36	--
% of cohort	95%		5%
Gender					0.236
Female	58.0	(54.3 – 61.6)	69.4	(51.9 – 83.7)
Male	42.0	(38.4 – 45.7)	30.6	(16.3 – 48.1)
Race/ethnicity					0.168
Non-Hispanic White	33.9	(30.4 – 37.4)	30.6	(16.3 – 48.1)
Non-Hispanic Black	11.5	(9.3 – 14.1)	2.8	(0.1 – 14.5)
Hispanic	36.9	(33.4 – 40.5)	52.8	(35.5 – 69.6)
Other	17.7	(15.0 – 20.7)	13.9	(4.7 – 29.5)
SES					0.021
Low	13.9	(11.4 – 16.6)	30.6	(16.3 – 48.1)
Middle	64.3	(60.7 – 67.8)	50.0	(32.9 – 67.1)
High	21.8	(18.9 – 25.0)	19.4	(8.2 – 36.0)
8th grade cohort
n	835		78	--
% of cohort	91%		9%
Gender					0.657
Female	54.4	(50.9 – 57.8)	57.7	(46.0 – 68.8)
Male	45.6	(42.2 – 49.1)	42.3	(31.2 – 54.0)
Race/ethnicity					0.507
Non-Hispanic White	37.8	(34.5 – 41.2)	37.2	(26.5 – 48.9)
Non-Hispanic Black	13.2	(11.0 – 15.7)	9.0	(3.7 – 17.6)
Hispanic	32.6	(29.4 – 35.9)	39.7	(28.8 – 51.5)
Other	16.4	(14.0 – 19.1)	14.1	(7.3 – 23.8)
SES					0.597
Low	16.2	(13.8 – 18.9)	19.2	(11.2 – 29.7)
Middle	64.7	(61.4 – 68.0)	59.0	(47.3 – 70.0)
High	19.1	(16.5 – 21.9)	21.8	(13.2 – 32.6)
10th grade cohort
n	931		124	--
% of cohort	88%		12%
Gender					0.171
Female	58.5	(55.3 – 61.7)	51.6	(42.5 – 60.7)
Male	41.5	(38.3 – 44.7)	48.4	(39.3 – 57.5)
Race/ethnicity					0.455
Non-Hispanic White	26.2	(23.4 – 29.2)	29.8	(22.0 – 38.7)
Non-Hispanic Black	20.3	(17.8 – 23.0)	14.5	(8.8 – 22.0)
Hispanic	39.0	(35.8 – 42.2)	39.5	(30.9 – 48.7)
Other	14.5	(12.3 – 16.9)	16.1	(10.1 – 23.8)
SES					0.099
Low	23.5	(20.8 – 26.3)	24.2	(17.0 – 32.7)
Middle	65.9	(62.7 – 68.9)	58.9	(49.7 – 67.6)
High	10.7	(8.7 – 12.8)	16.9	(10.8 – 24.7)

^aColumn percentages represent prevalence within pattern and stratum (e.g., % female in Pattern 1).

^b3907 participants were enrolled in TATAMS at Wave 1. After excluding those not 11–19 years of age across Waves 1–9 (n=4), there were 1119 participants in 6^th Grade Cohort, 1322 participants in 8^th Grade Cohort and 1462 participants in 10^th Grade Cohort. Participants were further excluded from each cohort if they did not complete the survey at Wave 1, leaving 1109 participants in 6^th Grade Cohort, 1313 participants in 8^th Grade Cohort and 1456 participants in 10^th Grade Cohort (n=3878) for the final analysis sample.

^cTo assess Cohort, participants were asked ‘What Grade are you in?’. The response options at Wave 1 were ‘6^th’, ‘7^th’, ‘8^th’, ‘9^th’, ‘10^th’, ‘11^th’ and ‘12^th’. All 7th, 11th, and 12th grade students were removed and all 9th grade students were classified as 10th grade. 6^th, 8^th and 10^th Grade students were retained in the survey (n=3907) and followed longitudinally as ‘6^th grade cohort,’ ‘8^th grade cohort’ and ‘10^th grade cohort.’

^dTo assess Gender, participants were asked ‘What is your gender?’ and response options were ‘Male’ and ‘Female.’

^eTo assess Race/ethnicity, participants were asked ‘Are you Hispanic or Latino/a?’ and ‘What race or races do you consider yourself to be?’ respectively. The response options (Hispanic, White, Black or African American, Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander and Other) were used to derive the measure of race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Other).

^fTo assess SES, participants were asked ‘In terms of income, what best describes your family’s standard of living in the home where you live most of the time?’ and the response options provided were ‘Very well off,’ ‘Living comfortably,’ ‘Just getting by,’ ‘Nearly poor’ and ‘Poor.’ ‘Very well off’ was categorized as ‘High SES.’ ‘Living comfortably’ was categorized as ‘Middle SES’ and ‘Just getting by.’ ‘Nearly poor’ and ‘Poor’ were combined to derive ‘Low SES.’

^aColumn percentages represent prevalence within pattern and stratum (e.g., % female in Pattern 1).

^b3907 participants were enrolled in TATAMS at Wave 1. After excluding those not 11–19 years of age across Waves 1–9 (n=4), there were 1119 participants in 6^th Grade Cohort, 1322 participants in 8^th Grade Cohort and 1462 participants in 10^th Grade Cohort. Participants were further excluded from each cohort if they did not complete the survey at Wave 1, leaving 1109 participants in 6^th Grade Cohort, 1313 participants in 8^th Grade Cohort and 1456 participants in 10^th Grade Cohort (n=3878) for the final analysis sample.

^cTo assess Cohort, participants were asked ‘What Grade are you in?’. The response options at Wave 1 were ‘6^th’, ‘7^th’, ‘8^th’, ‘9^th’, ‘10^th’, ‘11^th’ and ‘12^th’. All 7th, 11th, and 12th grade students were removed and all 9th grade students were classified as 10th grade. 6^th, 8^th and 10^th Grade students were retained in the survey (n=3907) and followed longitudinally as ‘6^th grade cohort,’ ‘8^th grade cohort’ and ‘10^th grade cohort.’

^dTo assess Gender, participants were asked ‘What is your gender?’ and response options were ‘Male’ and ‘Female.’

^eTo assess Race/ethnicity, participants were asked ‘Are you Hispanic or Latino/a?’ and ‘What race or races do you consider yourself to be?’ respectively. The response options (Hispanic, White, Black or African American, Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander and Other) were used to derive the measure of race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Other).

^fTo assess SES, participants were asked ‘In terms of income, what best describes your family’s standard of living in the home where you live most of the time?’ and the response options provided were ‘Very well off,’ ‘Living comfortably,’ ‘Just getting by,’ ‘Nearly poor’ and ‘Poor.’ ‘Very well off’ was categorized as ‘High SES.’ ‘Living comfortably’ was categorized as ‘Middle SES’ and ‘Just getting by.’ ‘Nearly poor’ and ‘Poor’ were combined to derive ‘Low SES.’

as ‘Middle SES’ and ‘Just getting by.’ ‘Nearly poor’ and ‘Poor’ were combined to derive ‘Low SES.’

Table 2.

Adjusted odds ratios testing for differences in developmental patterns of e-cigarette use, by cohort and sociodemographic factors^b–e; the Texas Adolescent Tobacco and Marketing Surveillance System (TATAMS), Waves 1–9, 2014–2019

Table 2a. Past 30-day E-cigarette Use without Marijuana (Waves 1–9, 2014–2019; n=3,878)
	Pattern 2 (‘Early’ escalators) vs. Pattern 1 (Non-users)		Pattern 3 (‘Mid’ escalators) vs. Pattern 1 (Non-users)		Pattern 4 (‘Late’ escalators) vs. Pattern 1 (Non-users)
	AORa	95% CI	AORa	95% CI	AORa	95% CI
6th grade cohort
Gender
Female	REF		REF		n/a	n/a
Male	1.567	(0.725 – 3.384)	0.942	(0.566 – 1.568)	n/a	n/a
Race/ethnicity
Non-Hispanic White	REF		REF		n/a	n/a
Non-Hispanic Black	0.738	(0.183 – 2.985)	0.235	(0.070 – 0.789)	n/a	n/a
Hispanic	1.106	(0.420 – 2.911)	0.664	(0.381 – 1.159)	n/a	n/a
Other	1.592	(0.516 – 4.908)	0.683	(0.324 – 1.440)	n/a	n/a
SES
Low	REF		REF		n/a	n/a
Middle	0.167	(0.074 – 0.378)	0.829	(0.375 – 1.830)	n/a	n/a
High	0.103	(0.028 – 0.371)	1.143	(0.490 – 2.664)	n/a	n/a
8th grade cohort
Gender
Female	REF		REF		REF
Male	1.060	(0.711 – 1.579)	1.215	(0.854 – 1.727)	0.815	(0.405 – 1.638)
Race/ethnicity
Non-Hispanic White	REF		REF		REF
Non-Hispanic Black	1.006	(0.528 – 1.917)	0.168	(0.071 – 0.395)	0.000	(0.000 – Inf)
Hispanic	1.344	(0.826 – 2.186)	0.518	(0.342 – 0.785)	0.662	(0.294 – 1.490)
Other	1.094	(0.574 – 2.084)	0.647	(0.383 – 1.093)	1.203	(0.495 – 2.924)
SES
Low	REF				REF
Middle	0.989	(0.545 – 1.794)	0.926	(0.540 – 1.587)	1.246	(0.419 – 3.705)
High	1.324	(0.681 – 2.574)	1.073	(0.591 – 1.948)	1.012	(0.291 – 3.521)
10th grade cohort
Gender
Female	REF		REF		REF
Male	2.666	(1.687 – 4.213)	1.382	(1.042 – 1.834)	1.126	(0.798 – 1.591)
Race/ethnicity
Non-Hispanic White	REF		REF		REF
Non-Hispanic Black	0.325	(0.168 – 0.628)	0.440	(0.279 – 0.695)	0.425	(0.249 – 0.727)
Hispanic	0.394	(0.235 – 0.661)	0.827	(0.588 – 1.164)	0.602	(0.396 – 0.915)
Other	0.308	(0.139 – 0.681)	0.701	(0.439 – 1.119)	0.781	(0.462 – 1.322)
Socioeconomic status
Low	REF		REF		REF
Middle	0.614	(0.360 – 1.048)	0.762	(0.531 – 1.094)	0.645	(0.425 – 0.979)
High	0.492	(0.239 – 1.012)	0.738	(0.466 – 1.172)	0.501	(0.284 – 0.884)

Table 2b. Past 30-day E-cigarette Use with Marijuana (Waves 3–9, 2015–2019; Waves 3–9, 2015–2019; n=2,733)
	Pattern 2 (Escalators) vs. Pattern 1 (Non-users)
	AORa	95% CI
6th grade cohort
Gender
Female	REF
Male	0.687	(0.327 – 1.443)
Race/ethnicity
Non-Hispanic White	REF
Non-Hispanic Black	0.238	(0.030 – 1.886)
Hispanic	1.462	(0.672 – 3.180)
Other	0.853	(0.288 – 2.530)
SES
Low	REF
Middle	0.362	(0.164– 0.798)
High	0.428	(0.159 – 1.157)
8th grade cohort
Gender
Female	REF
Male	0.866	(0.541 – 1.388)
Race/ethnicity
Non-Hispanic White	REF
Non-Hispanic Black	0.711	(0.301 – 1.679)
Hispanic	1.246	(0.718 – 2.162)
Other	0.863	(0.417 – 1.785)
SES
Low	REF
Middle	0.798	(0.428 – 1.486)
High	1.038	(0.486– 2.216)
10th grade cohort
Gender
Female	REF
Male	1.336	(0.915 – 1.951)
Race/ethnicity
Non-Hispanic White	REF
Non-Hispanic Black	0.676	(0.370 – 1.232)
Hispanic	0.964	(0.605 – 1.534)
Other	1.014	(0.563 – 1.826)
Socioeconomic status
Low	REF
Middle	0.854	(0.542 – 1.344)
High	1.531	(0.827 – 2.834)

^aModels are adjusted for all sociodemographic variables (gender, race/ethnicity, SES).

^bTo assess Cohort, participants were asked ‘What Grade are you in?’. The response options at Wave 1 were ‘6^th’, ‘7^th’, ‘8^th’, ‘9^th’, ‘10^th’, ‘11^th’ and ‘12^th’. All 7th, 11th, and 12th grade students were removed and all 9th grade students were classified as 10th grade. 6^th, 8^th and 10^th Grade students were retained in the survey (n=3907) and followed longitudinally as ‘6^th grade cohort,’ ‘8^th grade cohort’ and ‘10^th grade cohort.’

^cTo assess Gender, participants were asked ‘What is your gender?’ and response options were ‘Male’ and ‘Female.’

^dTo assess Race/ethnicity, participants were asked ‘Are you Hispanic or Latino/a?’ and ‘What race or races do you consider yourself to be?’ respectively. The response options (Hispanic, White, Black or African American, Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander and Other) were used to derive the measure of race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Other).

^eTo assess SES, participants were asked ‘In terms of income, what best describes your family’s standard of living in the home where you live most of the time?’ and the response options provided were ‘Very well off,’ ‘Living comfortably,’ ‘Just getting by,’ ‘Nearly poor’ and ‘Poor.’ ‘Very well off’ was categorized as ‘High SES.’ ‘Living comfortably’ was categorized as ‘Middle SES’ and ‘Just getting by.’ ‘Nearly poor’ and ‘Poor’ were combined to derive ‘Low SES.’

^aModels are adjusted for all sociodemographic variables (gender, race/ethnicity, SES).

^bTo assess Cohort, participants were asked ‘What Grade are you in?’. The response options at Wave 1 were ‘6^th’, ‘7^th’, ‘8^th’, ‘9^th’, ‘10^th’, ‘11^th’ and ‘12^th’. All 7th, 11th, and 12th grade students were removed and all 9th grade students were classified as 10th grade. 6^th, 8^th and 10^th Grade students were retained in the survey (n=3907) and followed longitudinally as ‘6^th grade cohort,’ ‘8^th grade cohort’ and ‘10^th grade cohort.’

^cTo assess Gender, participants were asked ‘What is your gender?’ and response options were ‘Male’ and ‘Female.’

^dTo assess Race/ethnicity, participants were asked ‘Are you Hispanic or Latino/a?’ and ‘What race or races do you consider yourself to be?’ respectively. The response options (Hispanic, White, Black or African American, Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander and Other) were used to derive the measure of race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Other).

^eTo assess SES, participants were asked ‘In terms of income, what best describes your family’s standard of living in the home where you live most of the time?’ and the response options provided were ‘Very well off,’ ‘Living comfortably,’ ‘Just getting by,’ ‘Nearly poor’ and ‘Poor.’ ‘Very well off’ was categorized as ‘High SES.’ ‘Living comfortably’ was categorized as ‘Middle SES’ and ‘Just getting by.’ ‘Nearly poor’ and ‘Poor’ were combined to derive ‘Low SES.

Finally, we illustrated the average trajectories and developmental patterns in trajectories identified in analyses above, using non-parametric methods (Figure 1a, Figure 1b). As age is the primary independent variable in analysis, it is graphed on the x-axis; the number of days of e-cigarette use in the past 30 days (0–30), as the primary dependent variable, is on the y-axis. All methods were applied to all available data across all Waves for all students, providing robust estimates even in the presence of missing data, which was low.⁵¹ Final sample sizes for each cohort, along with a description of missing data are provided in the footnotes of the Tables.

Figure 1.

Stable trajectories of past 30-day e-cigarette use with and without marijuana, by cohort; the Texas Adolescent Tobacco and Marketing Surveillance System (TATAMS), Waves 1–9, 2014–2019

RESULTS

Past 30-day E-cigarette Use without Marijuana

The test of the ageXcohort interaction effect was statistically significant (p<0.01), providing evidence that the average trajectory or change in past 30-day e-cigarette use without marijuana with increasing age differed by cohort (Figure 1a, Panel 1a). While the trajectory for all cohorts increased with age, this increase occurred at younger ages for younger cohorts (6^th grade, 8^th grade), compared with the older cohort (10^th grade). At the same age (e.g., 16 years old), the frequency of past 30-day e-cigarette use without marijuana was, on average, highest for the 6^th grade cohort, followed by the 8^th grade and 10^th grade cohorts (Figure 1a, Panel 1a).

Next, within cohort, we identified sub-groups of adolescents who followed a similar developmental pattern of past 30-day e-cigarette use without marijuana as they aged. A 3-group solution fit the 6^th grade cohort best, and a 4-group solution fit the 8^th and 10^th grade cohorts best (Figure 1a, Panel 1b). With the exception of non-users (green pattern), all patterns of past 30-day e-cigarette use without marijuana escalated with increasing age for all cohorts. No declining patterns with increasing age were observed, for any cohort. Patterns included ‘early’ escalators (red pattern), ‘mid’ escalators (blue pattern), and ‘late’ escalators (purple pattern).

For the 6^th grade cohort, 9% of participants fell into one of two use patterns: (a) for 4% (‘early’ escalators), past 30-day e-cigarette use began early then escalated rapidly at age 13; while (b) for 5% (‘mid’ escalators), past 30-day e-cigarette use both began and escalated at age 14. By 16 years of age, past 30-day e-cigarette use without marijuana was higher for the ‘early’ escalators ($\overline{\text{x}}>15$ days/month) compared to the ‘mid’ escalators ($\overline{\text{x}}<10$ days/month).

For the 8^th grade cohort, 20% of participants fell into one of three use patterns: (a) for 9% (‘early’ escalators), past 30-day e-cigarette use began early then escalated quickly at age 14; (b) for 9% (‘mid’ escalators), past 30-day e-cigarette use began and escalated at age 15; and (c) for 2% (‘late’ escalators), past 30-day e-cigarette use began and escalated just before age 17. By 18 years of age, past 30-day e-cigarette use without marijuana among the ‘mid’ escalators ($\overline{\text{x}}~20$ days/month) had surpassed the ‘early’ escalators ($\overline{\text{x}}~15$ days/month), while use for ‘early’ and ‘mid’ escalators exceeded that of the ‘late’ escalators ($\overline{\text{x}}~5$ days/month).

For the 10^th grade cohort, 34% of participants fell into one of three use patterns: (a) for 6% (‘early’ escalators), past 30-day e-cigarette use began early in adolescence, prior to the age of 14 and escalated thereafter; (b) for 18% (‘mid’ escalators), past 30-day e-cigarette use began before the age of 15 and escalated more rapidly after the age of 16; and (c) for 10% (‘late’ escalators), past 30-day e-cigarette use began and escalated at 17 years of age. By 19 years of age, past 30-day e-cigarette use without marijuana was highest among the ‘early’ escalators ($\overline{\text{x}}~22$ days/month), followed by ‘mid’ ($\overline{\text{x}}~15$ days/month) and ‘late’ ($\overline{\text{x}}~10$ days/month) escalator groups.

Finally, we investigated sociodemographic differences in patterns of past 30-day e-cigarette use without marijuana (Table 1a, Table 2a). For the 6^th grade cohort, most non-users were in the middle SES group; most ‘early’ escalators were in the low or middle SES groups; and most ‘mid’ escalators were in the middle or high SES groups (p<0.001). For the 8^th grade cohort, most non-users were non-Hispanic White or Hispanic, as were most ‘early,’ ‘mid,’ and ‘late’ escalators (p<0.001). For the 10^th grade cohort, most non-users were female, while most ‘early’ escalators were male (p<0.001). In addition, more nonusers were Hispanic, while most ‘early,’ ‘mid,’ and ‘late’ escalators were either non-Hispanic White or Hispanic (p<0.001). No other statistically significant differences were observed.

Past 30-day E-cigarette Use with Marijuana

The test of the ageXcohort interaction effect was statistically significant (p<0.01), providing evidence that the average trajectory or change in past 30-day e-cigarette use with marijuana with increasing age differed by cohort (Figure 1b, Panel 1a). While the average trajectory of past 30-day e-cigarette use with marijuana remained relatively flat for the 6^th grade cohort, it accelerated with age for the 8^th and 10^th grade cohorts. At the same age (e.g., 17 and 18 years old), the frequency of e-cigarette use with marijuana was higher in the 8^th grade cohort compared to the 10^th grade cohort (Figure 1b, Panel 1a).

Next, within cohort, we again identified sub-groups of adolescents who followed a similar developmental pattern of past 30-day e-cigarette use with marijuana as they aged. A 2-class solution fit all cohorts best: 7%, 8%, and 12% of the 6^th, 8^th, and 10^th grade cohorts fell into one e-cigarette use pattern (red; Figure 1b, Panel 1b), while the rest (green pattern) remain non-users. Past 30-day e-cigarette use with marijuana consistently escalated with increasing age for the 8^th and 10^th grade cohorts; but, for the 6^th grade cohort, it peaked at age 15 then declined somewhat. On average, the use of marijuana in these devices never exceeded 10 days/month.

We also investigated sociodemographic differences in patterns of past 30-day e-cigarette use with marijuana (Table 1b, Table 2b). For the 6^th grade cohort, most non-users were in the middle SES group (p<0.021). No other statistically significant differences were observed.

DISCUSSION

This is the first study to describe developmental trajectories of past 30-day e-cigarette use across adolescence, from 11 to 19 years of age. Importantly, it demonstrates that stable trajectories of e-cigarette use with and without marijuana (i.e., liquid THC) do exist, which has important implications for interventions. Our analysis described meaningful trajectories that occur ‘on average’ and include subgroup variation. Trajectories varied by cohort; age of onset; frequency and escalation in use; and substance used. With only one exception, all trajectories of e-cigarette use escalated with age. Moreover, age of onset and progression in use were positively related. That is, the younger one started, the more one used later, providing support for primary prevention as an important component of comprehensive tobacco control efforts that focus on e-cigarette use. Interventions to prevent the onset and progression in e-cigarette use among youth must begin early (i.e., in middle school) and be sustained throughout adolescence.

E-cigarette use without marijuana was more prevalent than e-cigarette use with it, consistent with differences in national prevalence estimates, at present (i.e., 2019).^3,52 While marijuana use never exceeded 5–10 days per month, use of e-cigarettes without marijuana, at its peak, ranged from 10 to 20+ days per month. Our measure of the latter was not specific to quantity of nicotine used, which is difficult to disentangle in research. Most e-cigarette products are not clearly labeled in regards to their nicotine content, suggesting self-reports of use may be biased.^53,54 Future research that examines differences by device type or brand may be able to differentiate between trajectories with higher or lower/no nicotine, as certain brands (e.g., JUUL) and device types (e.g., vape pens, mods) are known to contain higher levels of nicotine than others.^55–57 Use of any e-cigarette product, with or without these substances, still confers risk.¹⁵ Our data suggest marijuana use in these devices may be highest for non-Hispanic Blacks, while use of nicotinized or denicotinized e-liquids is most common for non-Hispanic Whites. Notably, few differences by socioeconomic status were noted, and no differences by sex were observed.

While vaping nicotine and marijuana are distinct behaviors that have differential legal consequences, there are many similarities in risk factors for use, and their use patterns can be strongly linked.⁵⁸ Importantly, smoking combustible THC is still the most common mode of using marijuana among adolescents, followed by edibles, and vaping.⁵⁹ Conversely, vaping is now the preferred method for using nicotine among youth, having eclipsed cigarette smoking in 2014.⁶⁰ While a recent California study of trajectories of nicotine vaping and cannabis vaping across late adolescence into early adulthood showed similarities between trajectories,³⁵ our study from Texas showed differences. This contrast in findings may lie in differences between substances and/or contexts, particularly specific to legalization of these products.⁶¹ For example, THC vaping products may be more accessible (albeit still not legal to purchase) for adolescents in states like California where adults can purchase these products legally – and less accessible in states like Texas, where they are illegal to purchase for both adolescents and adults. Texas recently raised the legal age of tobacco use to 21,⁶² yet recreational marijuana (i.e., THC) use remains illegal in the state of Texas for everyone. Studies show marijuana (i.e., THC) use rates among youth vary between states that do and do not prohibit marijuana use among adults.^63–65

Cost may also be a factor in understanding differences in e-cigarette use trajectories by substance. Youth are more price sensitive than adults as they do not have as much disposable income as older age groups.⁶⁷ The cost of a 2-pack of JUUL (a popular brand of e-cigarette) cartridges is roughly $10,⁶² while a 1-gram cartridge of marijuana can cost over $50.⁶⁸ Ultimately, these differences may explain why youth in our sample report less frequent use of e-cigarettes with marijuana (max 5–10 days per month) as compared to nicotine (10 to 20+ days per month).

E-cigarettes were first introduced to the marketplace in the United States in 2007,⁶⁹ while monitoring and surveillance of its use among youth began in 2011.⁶ As prevalence escalated quickly for this vulnerable sub-group as a whole, few explicit comparisons have been made to determine whether introduction of this tobacco product may have differentially impacted younger or older youth. Unfortunately, our analyses suggest younger cohorts have and may continue to bear the brunt of this epidemic, as it continues to unfold over time. Onset of use occurred at earlier ages for younger cohorts (i.e., 6^th and 8^th grade cohorts) and progressed more quickly to more frequent use, compared to the oldest cohort (i.e., 10^th grade). This is worrisome, as earlier onset of substance use (e.g., alcohol use, tobacco use, marijuana use) often predicts worse health outcomes.^32,70,71 Future research is needed to determine what risk factors differentially predict these trajectories, across age groups. Early adolescence may simply be a developmental period characterized by more susceptibility to ENDS use, or other factors may be responsible for this finding, such as product characteristics (e.g., flavors, device type), industry advertising, or social media that may be especially appealing to young adolescents. Alternatively, young people’s perceptions of the harm or addictiveness of these products may differ by age and/or cohort or secular trends in ENDS use across this period (2014–19) may be responsible for findings.^72,73 Continued longitudinal research is also needed to track adolescents into young adulthood, to determine how e-cigarette use trajectories unfold over time. Trajectories, by definition, confer momentum. Young adulthood is the developmental period during which use of tobacco products stabilizes and becomes established, behaviorally speaking.^74–78 It is yet unclear whether these trajectories of e-cigarette use identified in adolescence will escalate, stabilize, and/or decline in young adulthood.

Limitations of this study include its reliance on self-report data from regional cohorts. At the time of study (2014–19), both marijuana (i.e., liquid THC) and nicotine (i.e., e-cigarette and other tobacco product use) were illicit substances for youth under the age of 18 in the state of Texas. Generalizability of findings may be limited to major metropolitan areas in Texas, though the prevalence of e-cigarette use behaviors reported here are comparable to national studies.⁷⁸ Findings specific to marijuana use, however, may not generalize to other places where medicinal and/or recreational use are no longer illegal among adults over the age of 21.

CONCLUSION

This longitudinal study provides 5 years of follow-up data (2014–19) across adolescence (11–19 years old). Comparisons are made not only between substances vaped, but also between cohorts, thereby differentiating between trajectories in early (i.e., 11–15 years old), middle (i.e., 13–17 years old), and late (i.e., 15–19 years old) adolescence, adding to emerging evidence-based perspectives on developmental trajectories of e-cigarette use with and without marijuana.^33–37 Stable trajectories of past 30-day e-cigarette use with and without marijuana do exist among young people, may be most problematic among younger cohorts of adolescents, and escalate with age. Interventions to re-direct the most problematic trajectories of e-cigarette use among adolescents are urgently needed and must start early in the life course.

HIGHLIGHTS.

• Identifies trajectories of e-cigarette use across adolescence (11–19 years old)

• Trajectories are modeled separately for use with and without marijuana (i.e., liquid THC)

• Trajectories vary by age of onset; frequency and escalation in use; substance use

• Primary prevention of e-cigarette use among younger adolescents is especially critical