Prevalence and Sociodemographic Correlates of Adolescent Use and Polyuse of Combustible, Vaporized, and Edible Cannabis Products

Key Points Question What are the prevalence, patterns, and correlates of use and polyuse of combustible, edible, and vaporized cannabis products among adolescents? Findings In a 2015 cross-sectional survey study of 3177 adolescents, ever use of edible (21.3%) or vaporized (10.5%) cannabis was appreciable, and most ever users of cannabis (61.7%) used at least 2 products. Current use was higher in boys than girls for vaporized (6.1% vs 3.0%) but not for combustible (13.8% vs 13.1%) or edible (8.4% vs 7.3%) cannabis. Meaning Health professionals should be aware that youth might use a diverse spectrum of combustible and alternative cannabis products.


Introduction
Adolescent cannabis use poses a substantial public health burden.[7] The manufacture of commercially available cannabinoid-infused edibles (eg, gummy bears treated with cannabinoid extracts) and drinks (eg, energy drinks with tetrahydrocannabinol [THC; the principal addictive compound in cannabis]) have diversified the types of ingestible cannabis products that might appeal to adolescents. 6Production and marketing of electronic vaporizer devices capable of heating cannabis products to temperatures high enough to aerosolize cannabis without heating beyond the point of combustion (ie, vaping) have also proliferated. 7In addition to devices for vaping dry cannabis plant material, the electronic cigarette (e-cigarette) industry has expanded to commercial cannabis products, including the sales of cannabinoid-infused e-cigarette liquids that are available in youth-oriented flavors (eg, bubble gum) and that use marketing strategies that might attract adolescents. 7itial reports 5,[8][9][10] find that appreciable proportions of adolescents use vaporized or edible cannabis products.A systematic study of the prevalence, patterns, and sociodemographic correlates of use and polyuse of alternative cannabis products (ie, use of Ն2 different cannabis products) among adolescents has not been published, to our knowledge.Data elucidating whether similar trends might emerge with adolescent use of cannabis products are needed to guide surveillance, policy, and prevention addressing youth cannabis use.
In this cross-sectional survey, we report the prevalence, patterns, and sociodemographic correlates of adolescent cannabis use and polyuse across combustible, edible, and vaporized administration methods among 10th-grade students in Los Angeles, California, in 2015.This location and time are of significance; the state of California legalized medical cannabis in 1996 and recreational cannabis sales in 2018.Because Los Angeles is among the most populous and sociodemographically diverse metropolitan regions in the United States, cannabis use patterns among youths in this region can be informative of a wide cross-section of US youths.

Participants and Procedures
Data were drawn from the Happiness and Health Study, an ongoing prospective cohort study of students in 10 high schools in the Los Angeles metropolitan area.Details on study design and methods are available elsewhere (D.B. and A.M.L., unpublished data, completed July 2018). 11

JAMA Network Open | Pediatrics
Prevalence and Sociodemographic Correlates of Adolescent Use and Polyuse of Cannabis Products were recruited based on their proximity to the research site and their representation of a diverse cross-section of urban and suburban communities, which collectively form a sociodemographically heterogeneous sample.Paper-and-pencil surveys were administered in students' classrooms on site.
Students not in class during data collections completed abbreviated surveys by telephone, internet, or mail.Survey items assessing cannabis use by method of administration were included in the (10th grade) survey from January 2 through October 6, 2015.Of 4100 eligible 9th-grade students in 2013, 3396 (82.8%) assented and consented to enroll.The institutional review board of the University of Southern California approved the study, which follows Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies 12 and the American Association for Public Opinion Research (AAPOR) reporting guideline for surveys. 13ong 3251 participants administered the spring 2015 survey, 3177 provided cannabis use data, constituting this report's analytic sample.The participant accrual diagram and comparisons of participant characteristics between cohort enrollees who were included vs excluded from this report due to attrition or missing responses can be found in the eFigure and eTable 1 in the Supplement.
Cohort enrollees who were included vs excluded from this report differed on race/ethnicity, were slightly younger, and had lower proportions of boys and combustible cannabis users in the fall of 2013 (9th grade) (eTable 1 in the Supplement).

Measures Sociodemographic Characteristics
At enrollment, participants self-reported their sex, age, race/ethnicity (response options: white, black, Hispanic, Asian, Native American/Alaska Native, Native Hawaiian/Pacific Islander, multiethnic/ multiracial, or other), parent with the highest educational attainment (6 choices ranging from 8th grade or less to advanced college degree), and eligibility for the free or subsidized lunch program (ie, youth with family income Յ185% of federal poverty limit [yes or no]).Respondents who selected Native American/Alaska Native (32 [1.0%]), Native Hawaiian/Pacific Islander (136 [4.3%]), or other (48 [1.5%]) for the race/ethnicity item were collapsed into a category of other owing to smaller sample sizes for these groups.To capture educational attainment and income in a socioeconomic status (SES) composite variable, we classified youths with high SES as those who reported their parents attended college or higher educational attainment (1946 [61.3%]) and being ineligible for free or subsidized lunch (1424 [44.8%]), as in prior work. 14,15Remaining participants were classified with low SES.

Statistical Analysis Descriptive Results and Comparisons by Method of Administration
Data were analyzed from July 17, 2017, through July 12, 2018.For each cannabis product and pattern of use and polyuse, descriptive frequencies (number [percentage]) were reported in the overall sample and by sex, race/ethnicity, and SES.We estimated multivariable generalized linear mixed models (GLMMs), including age, sex, SES, race/ethnicity, and method of cannabis administration as simultaneous regressors with random intercepts to account for school-clustering effects.
Generalized linear mixed models allow the analysis of multiple observations nested within respondents, which permitted head-to-head comparisons in prevalence or frequency of use across the 3 products by inputting administration method (combustible vs edible vs vaporized) as a withinparticipant categorical regressor variable.To examine whether sociodemographic variation in use differed by cannabis administration method, within-by-between participant variable interaction terms were added (ie, administration method × sex, administration method × race/ethnicity, and administration method × SES) in subsequent models tested one at a time.Significant interaction effects were followed by stratum-specific univariable tests of associations of sociodemographic variables with cannabis use, completed separately for combustible, edible, and vaporized cannabis administration methods with school-level random effects.
Ever use and past 30-day use status (yes or no) were tested using binary logistic regression GLMMs with a binomial error distribution and log link.Frequency outcomes for past 30-day use were recoded into quantitative variables by inputting the mean whole number value within the range provided within each response category (range, 0-30), tested using linear regression GLMMs and including only the subsample of past 30-day users of cannabis products to characterize use frequency among users.

Polyuse
Polytomous (multinomial) logistic regression GLMMs 18 with a generalized logit link, which allows for a categorical outcome with 2 or more levels, were used to test associations of each demographic variable with a nominal 4-group variable reflecting the number of cannabis products used (0, 1, 2, or 3) with school-level random effects and 1 observation per respondent.Separate models were tested for designations of ever and past 30-day use.Polytomous logistic regression provided an omnibus P value indicative of sociodemographic differences across any of the 4 groups and 3 pairwise contrast estimates comparing odds of use of 1, 2, or 3 methods of administration vs a reference outcome group who used none of the cannabis products.
Odds ratios (ORs) or unstandardized linear regression parameters (Β values) with 95% CIs from GLMMs are reported.Analyses were conducted using SPSS software (version 23; IBM Corp).Raw P values from 2-tailed tests of GLMMs were considered statistically significant after correction for multiple testing using the Benjamini-Hochberg method to control for the false discovery rate to .05, which led to a corrected significance threshold of .028. 19

Missing Data
Sociodemographic data were unavailable for some participants due to item-level nonresponse or selecting "I don't know" in response to SES questions (eTable 2 in the Supplement provides numbers with available data for each sociodemographic group).Participants with missing sociodemographic data were excluded from regression models (n = 476).Sensitivity analysis using multiple imputation with the missing-at-random assumption for sociodemographic data did not appreciably change results (eTable 3 in the Supplement).

Participants
The   a Binary logistic regression models in overall sample with data available for all regressors (n = 2701).

Cannabis Use Prevalence and Frequency by Method of Administration
b Linear regression among past 30-day users (n = 393).
c Estimates from generalized linear mixed models of association of sociodemographic characteristics, administration method, and their interaction as simultaneous regressors, adjusted for school random effects and respondents' age.Main effects of sociodemographic variables are not presented.
d Indicates statistically significant with Benjamini-Hochberg correction using a P value threshold of .028as the criterion of statistical significance.
e Interaction terms were added to models one at a time; main effect estimates exclude interaction terms.
f Estimates from univariable generalized linear mixed models with school random effects.
g Stratum-specific associations of race/ethnicity with cannabis use for each administration method are reported in eTable 4 in the Supplement.

JAMA Network Open | Pediatrics
Prevalence

Sociodemographic Correlates of Cannabis Use by Method of Administration
As shown in Table 2, sex and SES differences in prevalence varied by administration method (administration method × sex and administration method × SES interactions) for ever use and past 30-day use.As illustrated in Table 2   Prevalence is compared by sex (A) and socioeconomic status (SES) (B).Socioeconomic status was measured as high if at least 1 parent attended college and family income was greater than 185% of the federal poverty line (ineligible for free or subsidized lunch); low, if parental educational attainment was lower or family income was at or below 185% of the federal poverty line.P values were calculated using χ 2 test to compare girls vs boys or low vs high SES.
ethnicity × administration method) (stratum-specific associations between race/ethnicity and cannabis use for each administration method are reported in eTable 4 in the Supplement).
Sociodemographic differences in frequency of use among past 30-day users did not vary by administration method.

Discussion
This study of Los Angeles 10th-grade students in 2015 found that, although combustible cannabis remained the most popular cannabis administration method, use of cannabis via alternative administration methods was of appreciable prevalence and unequally distributed across sociodemographic strata.We found that polyuse via multiple administration methods was a  predominant pattern of cannabis use and report the first evidence, to our knowledge, of tripleproduct polyuse of combustible, edible, and vaporized cannabis among youths.These results add to a scant evidence base and begin to address critical early questions about the potential effect of alternative cannabis products on the pediatric public health burden.
A key question is whether a new pool of youth who have traditionally been at lower risk for combustible cannabis uptake might be drawn to alternative (noncombustible) cannabis products. 7ible and vaporized cannabis possess unique qualities absent in combustible cannabis (eg, availability of flavorings, no smell of smoke, attractive marketing and packaging, absence of harsh sensations to the airways and bitter taste of smoke), which might attract a wider segment of youths normally deterred from cannabis in combustible form. 20,21Although this study does not definitively address this question, observations from the present analysis fail to dispel such concerns.Among ever users of cannabis in this study, 7.8% had never smoked combustible cannabis, having used only edible or vaporized products.Whether this group of exclusive edible or vaporized cannabis product users would have avoided cannabis use if noncombustible products were unavailable is unknown.
Differences in cross-product frequency of use in this sample provide some of the first observational data indicative of the comparative abuse liability of combustible, edible, and vaporized cannabis in adolescents.Mean frequency of cannabis use was higher for combustible than edible cannabis.Evidence suggested that frequency of vaping was at the midpoint for the frequency observed for combustible and edible product use.[24] Recent surveys of adolescent Facebook users and northern California residents 8,9 suggest that noncombustible products appear to be commonly used in addition to combustible products, with polyuse of cannabis products being a predominant use pattern.The present study reinforces this conclusion and reveals a novel pattern of triple-product use of combustible, edible, and vaporized cannabis products, which represented 260 of 1077 cases (24.1%) of ever use and 99 of 474 (20.9%) of past 30-day use of cannabis in this sample.Relative to single-product use, polyuse of cannabis products might increase neuroexposure to a wider diversity or higher level of cannabinoids, which could accelerate the risks associated with cannabis use, including dependence and other neurocognitive consequences. 25Given such potential health hazards and the sizeable prevalence of dual-and triple-product cannabis use in this study, national surveillance of patterns of polyuse of cannabis products is warranted.With impending expansion of the cannabis product marketplace in states that recently passed legal recreational cannabis legislation, surveillance might be an urgent need.
The present data suggest that the emergence of alternative cannabis products might redefine populations at risk and the potential for sociodemographic disparities in cannabis use.Sex differences in use were more pronounced for vaporized (vs edible or combustible) cannabis, suggesting that this administration method might differentially appeal to male users and/or deter female users.Socioeconomic disparities in use were characteristically observed for combustible and edible cannabis, yet virtually absent for vaporized products, similar to recently reported evidence of associations of low SES with combustible but not vaporized tobacco product use in Connecticut youth. 26Whether the comparatively high initial cost of vaporized cannabis products is a financial barrier that prohibits access in teenagers with lower SES, certain features of vaporized products disproportionately appeal to higher-SES youths, or other factors explain this trend warrant future study.Considerably higher prevalence of polyuse of cannabis products was observed for lower (vs higher) SES, boys (vs girls), and certain racial/ethnic groups.Given these results, cannabis regulatory policies or prevention programs that target particular products might have an uneven effect on rates of cannabis use across different demographic subgroups.

Limitations
This study had several limitations.First, because this report was cross-sectional, whether youth polyusers first initiated cannabis use with noncombustible products and later transitioned to combustible cannabis or vice versa is unclear, which can be addressed in future longitudinal work.
Second, study survey items did not differentiate cannabis product potency, strain (sativa vs indica vs hybrid), or certain types of cannabis formulation used, including the use of waxes, topicals, tinctures, concentrates, and cannabis-tobacco mixtures.These distinctions might be important because use of high-potency cannabis concentrates, especially via vaporized administration methods, appears to be popular among adult cannabis users and is associated with living in states with recreational cannabis policies, younger age, and lower perceived risk of cannabis use. 27Third, specific features of the sampling design (eg, a convenience sample in 2015 of 10th-grade students from a single region with a high prevalence of racial and ethnic minorities, attrition before the 10th grade survey) may have introduced selection bias, and results may not generalize to the current year or across historical periods, locations, race/ethnicity, or age ranges.Fourth, this study used a measure of SES based on parents' educational attainment and the students' eligibility for free or subsidized lunch, which may not generalize to other SES indicators.Fifth, although we assumed that data are missing at random conditional on the values of variables used for multiple imputation, 28

Figure 1 .
Figure 1.Prevalence of Combustible, Edible, and Vaporized Cannabis Product Use

Figure 2 .
Figure 2. Prevalence of Ever Use Patterns of Single and Multiple Products in Cannabis Users

JAMA Network Open | Pediatrics Prevalence
and Sociodemographic Correlates of Adolescent Use and Polyuse of Cannabis Products

Table 1 .
Depending on the time interval, odds of use were 3.21 to 4.15 times greater for combustible vs vaporized cannabis, 1.74 to 1.90 times as large for combustible vs edible cannabis, and 1.69 to 2.38 times greater for edible vs vaporized cannabis (Table2).Among participants who reported using cannabis in the past 30 days, mean frequency of use of combustible cannabis was higher by 2.65 Prevalence and Frequency Distribution of Cannabis Use by Administration Method in Overall Sample Most cannabis users (665 [61.7%]) used multiple administration methods.As depicted inTable1, prevalence of ever use was highest for combustible cannabis (993 [31.3%]), followed by edible (676 [21.3%]) and then vaporized (333 [10.5%]) cannabis.A similar pattern was found for past 30-day use JAMA Network Open | Pediatrics Prevalence and Sociodemographic Correlates of Adolescent Use and Polyuse of Cannabis Products JAMA Network Open.2018;1(5):e182765.doi:10.1001/jamanetworkopen.2018.2765September28, 2018 4/12 Downloaded From: https://jamanetwork.com/ on 09/27/2023 (426 [13.4%] for combustible, 249 [7.8%] for edible, and 156 [4.9%] for vaporized cannabis).aThe denominator of each column is the number of past 30-day cannabis users in each administration method.

Table 2 .
Differences in Cannabis Use and Sociodemographic Correlates of Cannabis Use by Administration Method and Sociodemographic Correlates of Adolescent Use and Polyuse of Cannabis Products 95% CI, 1.40-3.91days;P= .001)thanthemean frequency of use for edible cannabis and 1.75 days higher (95% CI, 0.59-2.90days;P= .003)thanfrequency of use for vaporized cannabis.The difference in mean frequency of past-month use for edible vs vaporized cannabis (−0.91 days; 95% CI, −2.33 to 0.52 days; P = .21)wasnotsignificant (Table2).

Table 3 .
Association of Sociodemographic Characteristics With Use and Polyuse of Cannabis via Multiple Administration Methods b Indicates statistically significant with Benjamini-Hochberg correction using a P value threshold of .028as the criterion of statistical significance.
we cannot rule out potential bias due to missingness caused by unmeasured aspects of respondents' sociodemographic characteristics.Sixth, relative to youth in this report, those excluded owing to cohort attrition before the spring 2015 survey or other sources of missing data (219 of 3396 [6.4%] cohort enrollees) were slightly overrepresented by boys, older youths, certain racial/ethnic groups, and combustible cannabis users based on data collected in 9th grade.Consequently, the figures reported herein might underestimate cannabis use.Prevalence and Sociodemographic Correlates of Adolescent Use and Polyuse of Cannabis Products ConclusionsIn this 2015 cross-sectional survey of Los Angeles 10th-grade students, use of cannabis via alternative administration methods was of appreciable prevalence, predominantly observed in conjunction with polyuse of other cannabis products, and unequally distributed across sociodemographic strata.JAMA Network Open | PediatricsJAMA Network Open.2018;1(5):e182765.doi:10.1001/jamanetworkopen.2018.2765September 28, 2018 9/12