Multi-directional pathways of tobacco and marijuana use, including comorbid use, among a population-based cohort of adolescents in Texas: A six-wave cross-lagged model

Abstract

Purpose:

Past 30-day tobacco and marijuana use commonly occur among adolescents. It is unclear whether use of one product precedes the other, especially given the new climate surrounding marijuana legalization and the increasing popularity of e-cigarettes.

Methods:

Six-panel cross-lagged regression models, with six months between each panel/Wave (2014–17), were used to model stability paths, bi-directional paths, and comorbid paths (i.e., correlations) between past 30-day use of marijuana and tobacco products. Data were derived from three cohorts of adolescents (n = 3907; weighted N = 461,069) in 6th, 8th, and 10th grades at baseline.

Results:

Few bidirectional relationships between past 30-day tobacco and marijuana use were observed in early adolescence (6th grade). During the middle adolescence developmental period (8th grade), past 30-day marijuana use was prospectively associated with greater risk of past 30-day tobacco use. In late adolescence (10th grade), increased odds of past 30-day marijuana use given prior past 30-day e-cigarette use, and vice versa, were observed. For all cohorts, stability paths were common, especially for past 30-day marijuana use. Comorbid use was common in middle adolescence (8th grade) but small in magnitude.

Conclusions:

This is the first study to longitudinally situate comorbid, past 30-day use of tobacco and marijuana and simultaneously examine bi-directional past 30-day use of these products for adolescents. Marijuana use more often and more strongly predicted subsequent tobacco use than the reverse, especially during middle adolescence (13–15 years old). Marijuana use should be considered when creating interventions that address adolescent e-cigarette use in the U.S.

1. Introduction

For the first time, in 2011, prevalent past 30-day marijuana use surpassed past 30-day tobacco use among adolescents aged 12–17 years and in 2019 was 7.4%, compared to 2.3% past 30-day cigarette use (Substance Abuse and Mental Health Services Administration, 2020). In just two years, past 30-day use of e-cigarettes more than doubled among high school seniors and was 30.9 percent in 2019 (Johnston et al., 2020). Comorbid use of tobacco and marijuana in recent years has been well documented (Akbar, Tomko, Salazar, Squeglia, & McClure, 2019; McClure et al., 2018) and more commonly occurs among adolescents than adults (Substance Abuse and Mental Health Services Administration, 2020). Studies show that adolescent marijuana users have higher rates of cigarette use compared to non-marijuana users (Okoli, Richardson, Ratner, & Johnson, 2008) and that adolescent tobacco users have higher rates of marijuana use compared to nonsmokers (Richter et al., 2005).

Negative health effects of e-cigarettes and other combustible tobacco product use include impaired lung function, disrupted brain development, and respiratory infection (U.S. Department of Health and Human Services, 2016, 2014). For adolescents, marijuana use is causally associated with the development of schizophrenia, other psychoses, and marijuana use disorder (National Academies of Sciences et al., 2017). Comorbid use of these substances often leads to worse clinical outcomes, greater nicotine dependence, and potential to create new barriers to smoking cessation (Peters, Budney, & Carroll, 2012).

Previous research supports gateway (Beenstock & Rahav, 2002; Timberlake et al., 2007), reverse gateway (Peters et al., 2012; Richter et al., 2005), and substitution effects hypotheses (Peters et al., 2012) between marijuana and tobacco use among adolescents. However, these studies do not assess other alternative products, such as e-cigarettes, hookah, and cigars, limiting their ability to illuminate contemporary use patterns. In 2016, the Surgeon General called for more research on the causal relationships between e-cigarettes and combustible tobacco products and on the extent to which e-cigarette use is related to use of other substances—i.e. marijuana (U.S. Department of Health and Human Services, 2016).

The explosive popularization of e-cigarettes (Johnston et al., 2020; Audrain-McGovern, Stone, Barrington-Trimis, & Unger, 2018; Goodwin et al., 2018; Pacek et al., 2018) and emerging combustible tobacco products like hookah have slowed the two-decade-long declining tobacco use trajectory among adolescents (U.S. Department of Health and Human Services, 2016). Likewise, the liberalization of marijuana regulation has resulted in increased availability and new products to meet consumer demand (National Academies of Sciences et al., 2017). While there is longstanding evidence for unidirectional relationships between cigarettes and marijuana use (Agrawal & Budney, 2012; Audrain-McGovern et al., 2018; Peters et al., 2012) and the reverse (Beenstock & Rahav, 2002), there is still relatively little known about the transition from other combustible (e.g. hookah, cigars) or non-combustible (e.g. e-cigarettes) tobacco products to established use of marijuana or comorbid use during adolescence. To better characterize adolescents’ marijuana, tobacco, and comorbid substance use, the present study investigated bi-directional patterns in past 30-day marijuana and tobacco use and comorbid use, for all combustible (i.e., cigarettes, cigar products, hookah) and e-cigarette products among adolescents (age 11–17 years).

2. Methods

2.1. Study design, setting, and participants

Longitudinal data are derived from three population-based cohorts of adolescents (aged 11–18 + years) living in the five counties surrounding the four largest cities in Texas. The University of Texas Health Science Center at Houston, School of Public Health Institutional Review Board approved all protocols for the study. The Texas Adolescent Tobacco and Marketing Surveillance system (TATAMS) collected data from students in the 6th, 8th, and 10th grade at baseline, Fall 2014 (n = 3907; weighted N = 461,069). Among these, 48.9% were girls, 54.5% Hispanic, 21.4% non-Hispanic white, and 17.6% non-Hispanic black enrolled in 79 participating schools. TATAMS utilized a complex probability sampling design, described elsewhere (Perez et al., 2017) to recruit this population-based sample. Baseline data were collected in the classroom during the 2014–2015 academic year via tablets. Subsequent data have been collected outside the classroom approximately every 6 months using web-based surveys. This analysis uses data from Waves 1–6 (2014 to 2017). The retention rate ranged from 74% (Wave 1)-to 64% (Wave 6), comparable with other longitudinal studies of adolescent substance use with equivalent incentive structures. (Cantrell et al., 2018; Johnston et al., 2020).

2.2. Measures

Survey items were adapted from reliable, valid measures from national tobacco and marijuana use surveillance studies: Population Assessment of Tobacco and Health study (Hyland et al., 2017), National Survey on Drug Use and Health (Substance Abuse and Mental Health Services Administration, 2016) National Youth Tobacco Survey (Centers for Disease Control and Prevention, 2009), Monitoring the Future (Bachman, Johnston, O’Malley, & Schulenberg, 2005), and Youth Risk Behavioral Survey (Centers for Disease Control and Prevention, 2013).

2.3. Past 30-day tobacco and marijuana use

Past 30-day tobacco use was assessed at each Wave by product (conventional cigarettes, e-cigarettes, cigar products, and hookah) and by frequency (0–30 days). Use of conventional cigarettes, cigar products, and hookah were collapsed into a combustible tobacco product use category, following past research (Berg et al., 2017). Smokeless tobacco was excluded (prevalence < 2%) for all cohorts). As the majority of users were “light” users, categorized as using 1–5 days in the past 30 days (See Supplementary Tables 1–2), both past 30-day tobacco and marijuana use were dichotomized (0 days vs. any), consistent with past research (Okoli et al., 2008).

2.4. Cohort

Participants in 6th, 8th, and 10th grade cohorts were purposefully chosen to reflect three distinct developmental phases—early, middle, and late adolescence—corresponding to ages 11–13, 13–15, and 15–17 years, respectively (Perez et al., 2017). Cross-lagged analyses are presented separately for these cohorts to preserve the ability to interpret results from theoretical and applied perspectives on adolescent development (American Academy of Child and Adolescent Psychiatry, 2003).

2.5. Sociodemographic characteristics

Socioeconomic status (SES) was assessed by asking, “In terms of income, what best describes your family’s standard of living in the home where you live most of the time? Would you say your family is… (Very well off, living comfortably, just getting by, nearly poor, or poor)?” The lowest three item response categories were collapsed into a single category of “Low” versus “Living comfortably” versus “Very well off.” Race/ethnicity were also assessed. The non-Hispanic White/Other category for adolescents includes categories of White, Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, and Other.

3. Data analysis

3.1. Sampling weights

All analyses were weighted to account for the complex survey design, clustering within schools, and for non-response so as to generalize back to the population from which the sample was drawn (Perez et al., 2017). 3907 students (weighted sample size of 461,069) provided data on socio-demographic characteristics and past 30-day use of tobacco and marijuana use.

3.2. Modelling longitudinal relationships and comorbid use using cross-lagged models

Six-panel cross-lagged path models, each approximately 6 months apart, were created to examine bi-directional relationships between past 30-day marijuana and tobacco use across Waves; and comorbid use, defined as past 30-day use of tobacco and marijuana at the same Wave. The following were simultaneously estimated: 1) stability paths, within each product between Waves, to model autoregressive direct effects; 2) bi-directional paths, between products and between Waves, to predict use of the other product at a subsequent Wave; and 3) correlation coefficients, between products within each Wave, to model comorbid use of marijuana and tobacco. Past 30-day marijuana- and tobacco-specific product use outcomes were modeled as binary variables (no use vs. any use), allowing for the calculation of odds ratios.

Nine sets of models—three per cohort—were created to estimate relationships between past 30-day marijuana use and (a) past 30-day tobacco use (any product); (b) past 30-day combustible tobacco use (combustible cigarettes, cigar products, hookah); and (c) past 30-day e-cigarette use. Analyses were conducted using Mplus (Los Angeles, CA, version 7.3) using maximum likelihood estimation methods with robust standard errors, assuming data were missing at random (<1% missing data) (Schafer & Graham, 2002) and the Weighted Least Squares with Mean and Variance Adjustment Estimation procedure (Asparouhov, 2005). Models present the adjusted (for sex, race/ethnicity, SES, and age), non-standardized βeta estimates and associated p-values across Waves 1–6. βeta estimates and standard errors were converted to odds ratios and 95% confidence intervals (95% CI), respectively.

4. Results

4.1. Descriptive analyses

At baseline in 2014, past 30-day combustible tobacco use was more prevalent with each sequential phase of adolescence, ranging from 2.18% (early adolescence) to 5.37% (middle adolescence) to 10.73% (late adolescence) (Table 1). Past 30-day e-cigarette and marijuana use trends were similar.

Table 1.

Baseline characteristics (Wave 1, 2014–15) among the entire adolescent cohort (n = 3907; N = 461,069), by cohort (6th, 8th, and 10th grades).

	6th grade		8th grade		10th grade		p-value
	(n = 1122, N = 148,465)		(n = 1322, N = 160,080)		(n = 1463, N = 152,524)
	% or mean	95% CI or SD	% or mean	95% CI or SD	% or mean	95% CI or SD
Sex (%, 95%CI)
Male	51.19	(39.10–63.28)	51.14	(42.63–59.65)	51.02	(45.03–57.00)	0.9996
Female	48.81	(36.72–60.90)	48.86	(40.35–57.27)	48.98	(43.00–54.96)
Age, mean (SD)	11.52	0.05	13.49	0.04	15.12	0.08	<0.0001*
SES, (%, 95%CI)
“Very well off”	22.58	(18.68–26.47)	21.43	(13.41–29.45)	15.1	(12.26–17.94)	0.0244*
“Living comfortably”	61.96	(58.76–65.17)	60.06	(54.53–65.59)	63.64	(59.64–67.63)
Low	14.74	(11.45–18.02)	18.48	(12.90–24.06)	21.03	(17.84–24.21)
Race/ethnicity, (%, 95%CI)
Non-Hispanic White/Other	27.09	(17.50–36.68)	27.18	(10.97–43.40)	29.47	(22.46–36.48)	0.9993
Non-Hispanic Black	16.88	(10.90–22.85)	17.86	(9.20–26.52)	17.95	(11.51–24.39)
Hispanic	56.03	(47.85–65.22)	54.96	(41.33–68.58)	52.58	(43.73–61.43)
Tobacco use
Ever use (yes)
Any product	8.74	(4.55–12.93)	24.01	(17.99–30.04)	40.71	(36.25–45.16)	<0.0001*
Combustible	5.34	(2.27–8.42)	15.07	(10.85–19.28)	26.38	(22.37–30.39)	<0.0001*
E-cigarette	7.72	(4.40–11.04)	22.65	(19.44–25.86)	39.99	(35.06–44.93)	<0.001*
Past 30-day use (yes)
Any product	3.29	(1.13–5.45)	9.33	(5.60–13.06)	18.26	(14.87–21.65)	<0.0001*
Combustible	2.18	(0.60–3.76)	5.37	(2.83–7.90)	10.73	(7.78–13.67)	<0.0001*
E-cigarette	1.98	(0.72–3.24)	6.69	(3.85–9.54)	13.48	(11.10–15.86)	<0.0001*
Marijuana use
Past 30-day use (yes)
Any product	2.65	(1.17–4.13)	10.32	(5.73–14.91)	15.86	(12.83–18.89)	<0.0001*

Socioeconomic status (SES) was assessed for adolescents on the TATAMS survey by asking, “In terms of income, what best describes your family’s standard of living in the home where you live most of the time? Would you say your family is…” (Very well off, living comfortably, just getting by, nearly poor, or poor). The item responses were categorized by collapsing the lowest three categories into a single category of “Low” versus “Living comfortably” versus “Very well off.”

Race/ethnicity: For TATAMS, Non-Hispanic White/Other category for adolescents includes race/ethnicity categories of White, Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, and Other.

Tobacco use variables include the reporting of ever or past 30-day use of (1) any tobacco product, including cigarette, cigar product –little filtered cigar, cigarillo, and large cigars–, hookah, and e-cigarettes; (2) combustible products, including cigarette, cigar product–little filtered cigar, cigarillo, or large cigar–, and hookah; and (3) e-cigarettes. All reporting is specific to cohort at baseline.

4.2. Stability paths for all cohorts

After controlling for age, race/ethnicity, SES, and sex, statistically significant autoregressive direct effects (stability in exclusive, past 30-day use, over time) were consistently observed across all Waves for marijuana use, for all cohorts. Although similar effects were observed for past 30-day combustible tobacco and e-cigarette use, statistically significant paths were less consistent and varied by tobacco product, Wave, and cohort (Figs. 1–3).

Fig. 1.

Cross-lagged models showing multi-directional pathways of use and comorbid use among 6th grade cohort in Texas across Waves 1–6 (2014–2017). βeta estimates and standard errors are shown in the following format: βeta estimate (Standard Error). Beta estimates can be exponentiated to obtain odds ratios. Statistically significant pathways are shown with solid arrows, and non-statistically significant pathways are shown with dashed arrows. All pathways are shown, and all models have been adjusted for race/ethnicity, gender, and socioeconomic status. Separate cross-lagged models are shown for past 30-day use of any tobacco and any past 30-day marijuana, past 30-day combustible tobacco and any past 30-day marijuana, and past 30-day e-cigarettes and any past 30-day marijuana. All tobacco and marijuana use outcomes have been dichotomized to show no days of use versus any days of use.

Fig. 3.

Cross-lagged models showing multi-directional pathways of use and comorbid use among the 10th grade cohort in Texas across Waves 1–6 (2014–2017). βeta estimates and standard errors are shown in the following format: βeta estimate (Standard Error). Beta estimates can be exponentiated to obtain odds ratios. Statistically significant pathways are shown with solid arrows, and non-statistically significant pathways are shown with dashed arrows. All pathways are shown, and all models have been adjusted for race/ethnicity, gender, and socioeconomic status. Separate cross-lagged models are shown for past 30-day use of any tobacco and any past 30-day marijuana, past 30-day combustible tobacco and any past 30-day marijuana, and past 30-day e-cigarettes and any past 30-day marijuana. All tobacco and marijuana use outcomes have been dichotomized to show no days of use versus any days of use.

4.3. Directionality of use and comorbid use, by cohort

4.3.1. 6th grade cohort (early adolescence)

After adjusting for age, race/ethnicity, SES, and sex, no statistically significant bi-directional paths were observed between past 30-day combustible tobacco product use and marijuana use, and only two were observed for e-cigarette and marijuana use, occurring between Waves 2 and 3 and Waves 5 and 6, in opposite directions (Fig. 1). Comorbid use of any tobacco, combustible tobacco, e-cigarette, and marijuana were fairly common but small in magnitude (increased odds < 2%) and generally occurred in later Waves, as participants neared middle adolescence.

4.3.2. 8th grade cohort (middle adolescence)

Controlling for age, race/ethnicity, SES, and sex, bi-directional relationships between past 30-day use of any tobacco product and past 30-day marijuana use were observed, and past 30-day marijuana use more consistently preceded an increase in past 30-day tobacco product use across Waves than the reverse (Fig. 2). The odds of any past 30-day tobacco use were higher for those who had used marijuana in the past 30-days at the previous Wave across all Waves, ranging from 16 times higher from Wave 1–2 and 2–4 times higher between other Waves (e.g., OR_{Wave 3 to 4} = 3.86, 95%CI 1.61–9.16). Conversely, increased odds of past 30-day marijuana use given any past 30-day tobacco use at a previous Wave were observed less frequently and were substantially smaller than those in the opposite direction. Specifically, odds of past 30-day marijuana use given past 30-day e-cigarette or combustible tobacco use were 7 to 15 times higher across Waves 3–6. Comorbid use of tobacco with marijuana was common across all Waves, especially for use of marijuana and e-cigarettes, but small in magnitude (increased odds ≤ 1%).

Fig. 2.

Cross-lagged models showing multi-directional pathways of use and comorbiod use among the 8th grade cohort in Texas across Waves 1–6 (2014–2017). βeta estimates and standard errors are shown in the following format: βeta estimate (Standard Error). Beta estimates can be exponentiated to obtain odds ratios. Statistically significant pathways are shown with solid arrows, and non-statistically significant pathways are shown with dashed arrows. All pathways are shown, and all models have been adjusted for race/ethnicity, gender, and socioeconomic status. Separate cross-lagged models are shown for past 30-day use of any tobacco and any past 30-day marijuana, past 30-day combustible tobacco and any past 30-day marijuana, and past 30-day e-cigarettes and any past 30-day marijuana. All tobacco and marijuana use outcomes have been dichotomized to show no days of use versus any days of use.

4.3.3. 10th grade cohort (late adolescence)

Overall, bi-directional associations for the cross-lagged paths linking past 30-day tobacco use and past 30-day marijuana use were smaller and occurred less frequently than exclusive tobacco or marijuana use (Fig. 3). Past 30-day combustible tobacco use at Waves 4 and 5 only resulted in increased odds of marijuana use at Waves 5 and 6, respectively, increasing the odds of marijuana use by about 2%. Past 30-day e-cigarette use at Waves 1, 3, and 5 preceded an increase in marijuana use at Waves 2, 4, and 6, respectively. The odds of past 30-day marijuana use given past 30-day e-cigarette use were 1.06 (95% CI 1.01, 1.12) from Wave 1 to 2 and were 1.33 (95% CI 1.15, 1.55) between Wave 3 and 4. The odds of past 30-day e-cigarette use given past 30-day marijuana use increased by 187% between Wave 2 and 3 by 43% between Wave 5 and 6. Comorbid use of combustible tobacco or e-cigarettes with marijuana was inconsistent and relatively small in magnitude.

5. Discussion

This is the first study to examine bi-directional relationships between past 30-day marijuana and tobacco product use over a two-and-a-half-year period (2014–17). A six-panel cross-lagged regression model was used to concurrently estimate stability in past 30-day marijuana and tobacco product use, separately; predictive paths between past 30-day marijuana and tobacco product use; and comorbid use of past 30-day marijuana and tobacco. In summary, we found few bi-directional relationships in early adolescence; many bidirectional relationships in middle adolescence, though past 30-day marijuana use more strongly and consistently predicted past 30-day tobacco use; and bi-directional relationships between past 30-day marijuana use and past 30-day e-cigarette use during late adolescence. For all phases of adolescence, stability in past 30-day marijuana use was observed; comorbid use of both product types occurred frequently, especially during middle adolescence, but was small in magnitude.

5.1. Are e-cigarettes another gateway ‘drug’?

Although both e-cigarettes (U.S. Department of Health and Human Services, 2016) and marijuana (Berg et al., 2017; Johnston et al., 2020) have been promoted as safer alternatives to combustible tobacco use, they nonetheless can adversely affect public health, especially if they lead to increased likelihood of nonsmokers beginning to use combustible tobacco or initiate marijuana use at a young age (Dierker, Braymiller, Rose, Goodwin, & Selya, 2018). Our results showed strong, positive, bi-directional associations between e-cigarettes and marijuana and comorbid use of marijuana and e-cigarettes to be common, especially among 8th (middle adolescence) and 10th (late adolescence) graders. The magnitude of the bi-directional pathways observed among the middle adolescent cohort suggests that transitions from use of one product to the other product or both are likely. The fact that these bi-directional relationships to and from e-cigarette and marijuana use are reinforced in the final Waves for middle adolescent and across some Waves for late adolescent cohorts reveals the need for intensified efforts at preventing and reducing e-cigarette and marijuana use in tandem during these transitional times.

In prior work from our team, students in the eighth grade cohort were at highest risk, compared to 6th and 10th graders, for transitioning from non-susceptible to susceptible to e-cigarette use within twelve months because they presented the greatest number of risk factors (e.g. social and normative influences) (Carey et al., 2019). Others (Audrain-McGovern et al., 2018) have measured an increased in odds of initiation of marijuana use by 39.6% (95%CI 2.69–4.90) and current use of marijuana by 19.5% (95%CI 2.51–5.36) for ever versus never use of e-cigarettes after 24-months of follow-up (9th to 11th grade). Our results support these studies, are in line with the well-documented exponential increase in e-cigarette use that occurs at the onset of high school (Hair et al., 2019), and are consistent with Unger, Soto, and Leventhal (2016) who did not find e-cigarette use to be associated with either cigarette or marijuana smoking cessation (U.S. Department of Health and Human Services, 2016). Considering these results in the context of the national outbreak of e-cigarette product use-associated lung injury, often linked with reports of THC-containing e-cigarette products (Centers for Disease Control and Prevention, 2020), the impact of increased e-cigarette use among adolescents cannot be overstated.

5.2. No successful tobacco intervention without attention to marijuana

For all phases, but especially for middle adolescence (e.g., transition from middle to high school), our results showed strong, consistent relationships from past 30-day marijuana use to increased odds of combustible tobacco use and e-cigarette use. Marijuana use predicting subsequent combustible tobacco use was strongest in middle adolescence, and to our knowledge, this relationship has not been recently documented among this (or any other) adolescent age group. Marijuana use preceding an increase in combustible tobacco use is troubling, given the overwhelming evidence for associations between marijuana use and escalation of “tobacco involvement” (Agrawal & Budney, 2012).

Moreover, comorbid use of marijuana and combustible tobacco products was common for all cohorts, especially in middle adolescence. Comorbid use and bi-directional relationships between combustible tobacco and marijuana occurred in the late adolescent cohort, across Waves 4–6, as adolescents prepared for the transition out of high school. Given recent data showing that non-daily cigarette smoking is increasing among marijuana users (Pacek et al., 2018) and that marijuana use disorder is more common among non-daily and daily cigarette smokers compared with former or never smokers (Weinberger et al., 2018), these results are concerning for adolescents who exit high school and continue to use both products or progress to more regular or problem use of either.

Lastly, the strong, consistent stability in marijuana use we observed suggest that once onset of past 30-day marijuana use occurs, it continues steadily throughout adolescence. By comparison, past 30-day use of combustible tobacco and e-cigarettes may be more intermittent (Hair et al., 2019). As recreational marijuana acceptance is burgeoning (Johnston et al., 2020), acceptable social and legal environments for marijuana use are rapidly changing. Sustained intervention efforts throughout adolescence with particular emphasis on students’ transitional periods into and out of high school and on the social norms surrounding harm associated with marijuana use are therefore crucial to reduce rates of use and comorbid use of tobacco and marijuana among adolescents.

5.3. Middle adolescence as an important intervention period

We identified middle adolescence as the period with the most bi-directional associations between past 30-day combustible tobacco and marijuana use and past 30-day e-cigarettes and marijuana use. Onrust, Otten, Lammers, and Smit (2016) found that differences in psychological and cognitive needs and capabilities dictate that, in order for school-based programs to reduce and prevent substance use, they must align with developmental stage of the intended target group—e.g., childhood, early, middle, or late adolescence (Onrust et al., 2016). They found targeting social norms to benefit younger adolescents but found behavioral change achievable in middle adolescence only for those already demonstrating substance use problems who were willing to change (Onrust et al., 2016).

Reward-seeking behavior culminates in middle adolescence. Steinberg (2010) and Gunther Moor, van Leijenhorst, Rombouts, Crone, and Van der Molen (2010) found peer approval to be rewarding in itself to this age group (Gunther Moor et al., 2010; Steinberg, 2010). As such, and in conjunction with our findings, middle adolescence is an important, yet perhaps challenging target for behavioral change. Programs targeting prevention in early adolescence, peer education in middle adolescence, and development of one’s own identity in late adolescence (Onrust et al., 2016) are likely to prove more effective at reducing or preventing past 30-day tobacco or marijuana use.

5.4. Strengths and limitations

Using a novel application of autoregressive cross-lagged models to ensure temporality, multiple pathways of past 30-day marijuana, any tobacco, combustible tobacco, and e-cigarette products use were visualized. Additional strengths of this study include its demographically diverse sample measured during distinct developmentally important stages of adolescence; complex, longitudinal design; and use of a representative population-based sample. Further, six-month spacing between survey Waves allowed for a more complete picture of volatile past 30-day tobacco and marijuana product use.

Due to survey items and sample size, this study was unable to stratify the cross-lagged models by specific products (e.g., cigarettes only, or marijuana consumed in varied forms). To produce the most stable estimates for the cross-lagged models, we dichotomized the frequency measures of past 30-day use (0 vs. any days of use per month) and, as such, cannot provide more specific estimates of risk. Thirdly, adolescent tobacco and marijuana use were self-reported, though validated survey items (Bachman et al., 2005; Centers for Disease Control and Prevention, 2009; Hyland et al., 2017; Substance Abuse and Mental Health Services Administration, 2016) were used to minimize bias. We were unable to collect use behavior from some participants lost-to-follow-up, though complete data available for each participant at each Wave were used for this analysis. It is reasonable to assume that tobacco and marijuana use would be higher among those lost to follow-up (Howe, Cole, Lau, Napravnik, & Eron, 2016), so we may underestimate the strength of the stability paths and cross-lagged paths between tobacco and marijuana use, and comorbid use.

The sample was drawn from five major metropolitan areas of Texas, so generalizability may be limited, especially because recreational marijuana use in Texas was illegal at the time of this study, compared to other states. However, our estimates of past 30-day combustible tobacco, e-cigarette, and marijuana use at baseline were comparable with other national estimates such as National Survey on Drug Use and Health (Substance Abuse and Mental Health Services Administration, 2020). Our cohort is a probability sample (unlike other regional cohorts nationwide) and provides robust estimates for student populations in these metropolitan areas of Texas, which this sample represents. In addition, this analysis provides a clear picture of the relationships between tobacco and marijuana use among adolescents, in the absence of policies specific to marijuana use. As such, our results could be particularly impactful to consider as Texas considers changes to the marijuana landscape.

5.5. Conclusions and future research

We observed e-cigarette use to be prospectively associated with greater risk of future marijuana use, especially in late adolescence. In addition, marijuana use was prospectively associated with greater risk of future combustible tobacco and e-cigarette use, especially during middle adolescence. These results are foreboding when considered with a similar study among young adults that found combustible tobacco use and e-cigarette use to be prospectively associated with greater risk of future marijuana use, and vice versa (Rogers et al., 2020). These results should help structure interventions specific to different phases of adolescence to encourage prevention and curb escalation with maturation before young adulthood, when use of these substances could escalate, and the path to lifelong substance use is typically established (Ling & Glantz, 2002). Further research should monitor the impact of newly signed federal regulation that raised the age of sale of tobacco products from 18 to 21 years (U.S. Food and Drug Administration, 2020) on use of these products individually and comorbidly with marijuana.

Implications and contributions.

This study longitudinally assesses comorbid use between tobacco and marijuana use and simultaneously models use of these products for adolescents, by stage of adolescence. Findings underscore the need for interventions specific to stage of adolescence, targeting use and comorbid use in early adolescence to curb subsequent escalation in later adolescence.