The Association of Type of Cannabis Product Used and Frequency of Use with Problematic Cannabis Use in a Sample of Young Adult Cannabis Users

Abstract

Research exploring patterns and predictors of problematic cannabis use behaviors among young adults is limited. This knowledge is essential for intervening to prevent abuse and dependence outcomes.

Methods:

Young adult cannabis users (Mage=19.2[SD=0.8]) in Southern California (N=1007) were classified by patterns of problematic cannabis use from the Cannabis Abuse Screening Test, using Latent Class Analysis. Multinomial regression evaluated the association of frequency of use (no past 30 day use, infrequent [use on 1-2 of past 30 days], semi-frequent [use on 3-9 of the past 30 days], and frequent [use on 10 or more of the past 30 days]) for each cannabis product type (combustible, edible, vaporized, concentrate, blunt) with class membership.

Results:

Four distinct classes of cannabis-related problems were identified: “non-symptomatic” (no problems; 33.8%), “non-recreational use” (before noon and when alone; 34.5%), “moderate use problems” (before noon, alone, and memory problems; 8.0%), and “severe cannabis problems” (all 6 problems; 23.7%). Semi-frequent (AOR range: 1.85-4.63;ps<0.05) and frequent (AOR range: 9.18-24.2;ps<0.05) use of combustible and vaporized cannabis, frequent blunt use (AORs range: 4.03-10.3;ps<0.05), and semi-frequent, but not frequent edible use (AOR range: 2.57-2.73;ps<0.05) was associated with higher odds of classification in any problematic use class (vs. non-symptomatic).

Conclusions:

Differences in cannabis use problems across these classes and their predictors reveal the heterogeneity in cannabis-related problems experienced by young people. Combustible cannabis, vaporized cannabis, and blunt cannabis use may confer the most risk for cannabis abuse and dependency outcomes, with more frequent days of use contributing to increased patterns of risk.

1. Introduction

Amidst largescale regulatory shifts, cannabis remains the world’s most prevalent “illicit” drug, and the second most preferred substance of initiation in adolescence behind alcohol, with over a third of 12th grade students in America reporting past 30 day use in 2019 (Johnston, 2020). Changing regulations and growing acceptance of cannabis use have fueled concerns regarding the potential harms and consequences of increased availability and access, particularly among youth and young adult populations (Cerdá et al., 2019; Hall and Lynskey, 2020; Spindle et al., 2019). While the majority of youth use cannabis for recreational purposes, accumulating evidence suggests that daily and more heavy consumption patterns are becoming increasingly common (Cerdá et al., 2019; Johnston, 2019), with nearly 1 in 10 young adults (aged 19-28) having used cannabis on 20 or more occasions in the past 30 days in 2019 (Schulenberg et al., 2020). Given the unwavering popularity, and rapidly evolving cannabis product marketplace, monitoring and evaluating the magnitude of cannabis-related harms among young people is a public health priority.

To date, the majority of research concerning problematic cannabis use has been limited to adult samples, with a focus on prevalence of cannabis use disorder (CUD) diagnoses as defined by the International Classification of Disease or the Diagnostic and Statistical Manual of Mental Disorders (DSM) (Grant et al., 2006; Legleye et al., 2012). While this methodological approach has been useful in identifying clinical populations of cannabis users, the validity of the DSM in detecting less severe cannabis-related consequences has been criticized in the context of youth and young adults (Bashford et al., 2010; Caldeira et al., 2008; Davis et al., 2009). Moreover, the use of a single indicator (i.e. meeting CUD criteria or not meeting CUD criteria) alone ceases to acknowledge the extent by which individuals may vary in their expression and severity of cannabis-related harms. Thus, researchers suggest that a subclinical threshold of “problematic use” - defined as use that leads to negative health and/or social consequences (Beck and Legleye, 2008) - may be a better suited mechanism for detecting individuals experiencing immediate harms and at increased risk of experiencing future harms due to cannabis with or without meeting diagnostic requirements (Annaheim et al., 2008; Bashford et al., 2010; Beck and Legleye, 2008; Casajuana et al., 2016; Davis et al., 2009). Short screening measures have since been developed to assess problematic cannabis use in general population surveys (Annaheim et al., 2008).

To date, a growing number of studies have been successful in determining subgroups of cannabis users using person-centered analytical approaches (e.g. latent class analysis, latent cluster analysis) (Grant et al., 2006; Krauss et al., 2017; Legleye et al., 2013; Manning et al., 2019; Pearson et al., 2017; Taylor et al., 2017). These findings provide evidence that heterogenous groups of cannabis users exist and are differentially characterized by demographic and personality factors. An important research question that has been left out of past work evaluating latent classes of problematic cannabis behaviors is the role of frequency and type of cannabis product used. Evidence suggests that different types of cannabis use products may confer differential levels of abuse and dependence risk (Arterberry et al., 2019; Craft et al., 2020; Freeman et al., 2018; Freeman and Winstock, 2015). For example, cannabis concentrate (eg, ‘dabs’ or use of extremely potent cannabis extracts such as wax, shatter, budder, or butane hash oil) users have reported increased CUD symptoms (Cinnamon Bidwell et al., 2018), greater physiological dependence (Meier, 2017), and more severe cannabis-induced withdrawal symptoms compared to flower cannabis users (Freeman and Winstock, 2015). However, other studies report no difference in cannabis-related harms between concentrate and flower cannabis users (Cinnamon Bidwell et al., 2018), leaving the relationship between type and quantity of use and cannabis-related harms unclear. Additional work is needed to evaluate the contribution of these factors in relation to problematic cannabis patterns among young adults.

Thus, the primary purpose of the present article is two-fold. First, we aimed to identify distinct classes of young adults according to patterns of self-reported problematic use behaviors. Second, we investigated whether type of cannabis product used and frequency of use were associated with patterns of cannabis-related harms in a sample of young adult cannabis users. If distinct profiles of problematic cannabis behaviors do exist, such evidence could be used to guide symptom-focused treatment efforts and prevention strategies geared towards deterrence from particular products or consumption practices that garner the most risk.

2. Methods

2.1. Source Population

Participants were enrolled in an ongoing prospective cohort study on mental health and health behaviors (Leventhal et al., 2015) originally recruited as students in 10 Southern California high schools in 2013. The study sample includes data from participants who completed the most recent survey wave conducted remotely via the Internet from October 2018 to November 2019. The analytic sample for this study was restricted to young adults who self-reported past 30 day use of at least 1 cannabis product and completed at least one item of the Cannabis Abuse Screening Test (CAST). The analytic sample (M[SD] age = 19.2[0.8] years; 36.8% female) was socio-demographically diverse (Table 1). Information on accrual and inclusion in this study’s analytic sample (N=1007) is depicted in supplemental Figure 1. All participants provided written informed consent. The University of Southern California Internal Review Board approved the study.

Table 1.

Characteristics of overall sample of lifetime cannabis users aged 18-21 (N=1007)

Variable	N (%)a
Sociodemographic Characteristics
Gender
Male	575 (57.1%)
Female	370 (36.7%)
Otherb	62 (6.2%)
Race/ethnicity
Hispanic/Latino	485 (48.2%)
Asian	145 (14.4%)
Black/African American	46 (4.6%)
White	166 (16.5%)
Otherc	145 (14.4%)
Missing	20 (2.0%)
Problematic Cannabis Use (CAST) d
Cannabis before noon	633 (62.9%)
Cannabis when alone	673 (66.8%)
Memory problems	586 (58.2%)
Told to reduce by friends or family members	316 (31.4%)
Unsuccessful attempt to reduce or quit	284 (28.3%)
Other cannabis-related problems	228 (22.7%)
Any cannabis problems	824 (82.4%)
No	178 (17.7%)
Yes	829 (82.3%)
Mean number of cannabis problems (SD)	2.7 (2.0)
Combustible cannabis use e
Non-current use	310 (30.8%)
Infrequent use	195 (19.4%)
Semi-frequent use	226 (22.4%)
Frequent use	274 (27.2%)
Edible cannabis use
Non-current use	640 (63.6%)
Infrequent use	196 (19.5%)
Semi-frequent	111 (11.0%)
Frequent use	60 (6.0%)
Vaporized cannabis use
Non-current use	416 (41.3%)
Infrequent use	197 (19.6%)
Semi-frequent use	212 (21.1%)
Frequent use	181 (18.0%)
Concentrate cannabis use
Non-current use	772 (76.7%)
Infrequent use	84 (8.3%)
Semi-frequent	88 (8.7%)
Frequent use	63 (6.3%)
Blunt use
Non-current use	478 (47.5%)
Infrequent use	206 (20.5%)
Semi-frequent use	184 (18.3%)
Frequent use	136 (13.5%)

Notes:

^aAvailable (nonmissing) data Ns range based on missing responses for each variable;

^bOther gender includes transgender male, transgender female, gender invariant/non-binary, other gender identity, prefer not to disclose and missing responses;

^cOther race/ethnicity includes multiracial/multiethnic, American Indian, Native Hawaiian or Pacific Islander, and other races;

^dProblematic cannabis use behaviors were assessed using the CAST: Cannabis Abuse Screening Test;

^eNon-current use defined as never users and prior users; Infrequent use defined as current use on 1 or 2 of the past 30 days; Semi-frequent use was defined as use on 3-9 of the past 30 days; Frequent use was defined as use on 10 or more of the past 30 days.

2.2. Measures

Problematic Cannabis Use:

For the current study, latent class indicators were created using the Cannabis Abuse Screening Test (CAST; (Legleye et al., 2012), a six-item questionnaire developed to screen for problematic cannabis use behaviors using the European Monitoring Centre for Drugs and Drug Addictions (EMCDDA) definition of problematic use meaning use that leads to negative health and social consequences for the individual user or community of users at large (Beck and Legleye, 2008). The CAST assesses the following aspects of cannabis consumption in the past 12 months, with response options of never/rarely/from time to time/fairly often/very often: non-recreational use (CAST 1 “Have you smoked cannabis before noon?”, CAST 2 “Have you smoked cannabis when you were alone?”), memory problems (CAST 3 “Have you had memory problems due to cannabis?”), social or relational problems (CAST 4 “Have friends or family members told you that you should reduce or stop your cannabis consumption?”), unsuccessful quit attempts (CAST 5 “Have you tried to reduce or stop your cannabis use without succeeding?”), and other problems linked to cannabis use (CAST 6 “Have you had problems because of your cannabis use (argument, fight, accident, bad result at school, etc?”). Following previously endorsed guidelines of detecting problematic cannabis use in young adult cannabis users (Cuenca-Royo et al., 2012), responses were dichotomized into binary indicators (0 [never] and 1 [rarely, from time to time, fairly often and very often]).

Cannabis Product Use and Frequency:

Cannabis use was assessed using well-validated items drawn from population-based surveys (Johnston, 2019). Participants who reported “yes” to any lifetime cannabis product use were asked a series of questions regarding frequency of use of five different types of cannabis products in the past 30-days, separately (categories: 0 days, 1-2 days, 3-5 days, 6-9 days, 10-19 days, 20-29 days, all 30 days). In primary analyses, responses were coded as multi-level ordinal response variables with the following specifications: non-current use = ever use, but no past 30-day use; infrequent use = use on 1 or 2 of the past 30 days; semi-frequent use = use on 3-9 of the past 30 days; frequent use = use on 10 or more of the past 30 days. Sensitivity analyses used cannabis product as a continuous predictor, using the median value for each category and rounding up to the nearest integer (0, 2, 4, 8, 15, 25, 30 days). Cannabis products included: 1) combustible cannabis [defined as smoked cannabis by joints, bowls, pipes, and bongs]; 2) blunts [defined as cannabis rolled in tobacco leaf or cigar casing]; 3) edibles [defined as cannabis or THC foods or drinks including pot brownies, edibles, cookies, cakes, butter, oil]; 4) vaporized cannabis [including liquid pot, cannabis oil, weed pen, PAX era]; and 5) concentrate cannabis [including dabs, wax, shatter, budder, butane hash oil, BHO].

Covariates:

Gender (categories: male, female, transgender male, transgender female, gender invariant/non-binary, other gender identity, prefer not to disclose) and race/ethnicity (Asian, African American, Hispanic, White, Multiethnic/Multiracial, Native American, Pacific Islander, other race/ethnicity) were included as covariates in our analysis to examine their potential influences on latent class membership.

2.3. Statistical Analysis

To identify subgroups of individuals based on their problematic cannabis use profiles, we performed latent class analyses (LCA) using all six binary items of the CAST. LCA classifies respondents into mutually exclusive classes with distinct endorsement profiles (i.e. certain cannabis-related problems experienced). We conducted a series of models to determine the appropriate number of classes beginning with a one-class model. These models were sequentially fitted to the data using maximum likelihood estimation and multiple starting values (500 runs) to avoid local maxima. These analyses were conducted in Mplus Version 8.1 (Lubke and Muthén, 2005). Indices used to determine the optimal LCA solution included the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), the Lo-Mendell-Rubin (LMR) likelihood-ratio test, and the sample size-adjusted BIC (SSA-BIC). These criterion incrementally tested the improvement in model fit in comparison to the model with one less latent class (Lanza et al., 2007). Entropy (measure of class distinctiveness, with values approach 1 indicating clear delineation of the classes), class size and interpretability were also considered. Optimal model selection was then determined based upon recommended model-fit indices and highest entropy/quality of classification (Nylund et al., 2007). We stopped increasing the number of classes when there was no substantial decrease in information criteria and a non-significant LMR test. Univariate regression models comparing final class solutions were also conducted to evaluate substantial and meaningful associations between latent class profiles and covariates in our model.

After identifying the appropriate number of latent classes using the six indicator items on the CAST, we then examined how frequency of use of each cannabis product was associated with likelihood of belonging to a particular latent class, adjusting for gender and race/ethnicity. This was achieved using a three-step modelling process (Asparouhov and Muthén, 2014; Vermunt, 2010), which was chosen to independently evaluate the relationship between latent class membership assignment and our predictor variables. In summary, this process involved first creating a most likely class variable using the latent class posterior probability distributions. The most likely latent class variable was regressed on our predictor variables in a multinomial regression model using the R3STEP procedure in Mplus. This statistical approach allowed us to control for uncertainty in class assignment while maintaining the latent class structure and meaning found initially. This approach yielded adjusted odds ratios (AORs) and 95% confidence intervals (95%CIs) illustrating associations between covariates and latent classes. Significance was set to .05 (two-tailed). Missing data were managed with full information maximum likelihood (FIML) estimation. Sensitivity analyses were conducted using similar models, including frequency of use of each type of cannabis product as a continuous predictor.

3. Results

Overall, 2,548 participants completed the Fall 2019 survey, and 1,007 participants reported past 30 day cannabis use and completed the CAST, and were thus eligible for inclusion in the current study analysis (Supplemental Figure 1). Among our total sample of lifetime cannabis users, 27.2% (N=274) of young adults reported using combustible cannabis on 10 or more of the prior 30 days, 6.0% (N=60) participants reported using edible cannabis on 10 or more of the past 30 days, 18% (N=181) reported vaporized cannabis use, 6.3% (N=63) reported concentrate cannabis use, and 13.5% (N=136) reported blunt use on 10 or more of the past 30 days. Using cannabis alone (66.8%) and using cannabis before noon (62.9%) were the most commonly reported cannabis problems; unsuccessful attempt to reduce or quit cannabis (28.3%), and other cannabis-related problems (argument, fight, accident, poor results at school, etc.; 22.7%) were reported less often.

A comparison of model fit indicated a four-class model solution to best represent classes of problematic cannabis use in our sample. This was determined by a non-significant LMR test comparing the 5 versus 4-class models, and decreases in AIC and SSA-BIC values (Supplemental Table 1). 5-class solution showed increased SSA-BIC, compared to 4-class solution (5957.042 versus 5946.194), and LMR p-value was not significant (p= 586).

The “non-symptomatic” class (33.8%) was comprised of a response pattern characterized by low probabilities of endorsing all six problematic cannabis use behaviors measured on the CAST. The “non- recreational use” class (34.5%) consisted of high probabilities of endorsing using cannabis before noon (.831) and using cannabis alone (.914), but low probabilities of endorsing other problems. The “moderate use problems” class (8%) was characterized by high probabilities of using cannabis before noon (.826), using cannabis alone (.758), and experiencing memory problems due to cannabis (1.00). Lastly, the “severe cannabis problems” class (23.7%) was characterized by high probabilities of endorsing all six problematic cannabis-related behaviors. Latent classes were labelled based on the characterization of types of problematic cannabis behaviors as defined on the CAST (Legleye et al., 2011), and the frequency of problems endorsed.

Associations of cannabis product types and frequency of use with LCA classes

Current frequent use of combustible cannabis (vs. non-current use) was associated with greater odds of being classified in the non-recreational use class (AOR=13.9; 95%CI: 5.62, 34.5), moderate use class (AOR=9.18; 95%CI: 2.85, 29.5) and severe class (AOR=24.2; 95%CI: 9.11, 68.2) (Table 3). Elevated associations were also observed for semi-frequent (vs. non-current use) for each outcome class relative to non-symptomatic class, and for infrequent (vs. non-current) use for non-recreational vs. non-symptomatic classes (but not for other classes). Semi-frequent (vs. non-current) use of edible cannabis was associated with more than twice the odds of being in each symptomatic dependence class (vs. non-symptomatic class); elevated but non-significant ORs were observed for frequent use. Semi-frequent and frequent use of vaporized cannabis was associated with greater odds of being classified as non-recreational (Semi-frequent AOR: 1.85; 95%CI: 1.15, 2.99); Frequent: AOR=9.18; 95%CI: 3.91, 21.6), moderate (Semi-frequent AOR: 3.46; 95%CI: 1.64, 7.29); Frequent: AOR=14.7; 95%CI: 5.20, 41.5), and severe (Semi-frequent AOR: 2.52; 95%CI: 1.43, 4.42); Frequent: AOR=9.86; 95%CI: 3.97, 24.5), vs. non-symptomatic. Concentrate use was associated with elevated, but statistically non-significant ORs for each association, with the exception of infrequent use with severe symptoms (AOR=2.81; 95%CI: 1.16, 6.83). Blunt use was generally associated with increased odds of membership in all classes, vs. non-symptomatic class; frequent use was associated with 4.60 times the odds of being classified in the non-recreational use class (95%CI: 1.86, 11.4), 4.03 times the odds of being classified in the moderate class (95%CI: 1.31, 12.4), and 10.3 times the odds of being classified in the severe class (95%CI: 4.00, 26.6).

Table 3.

Multinomial regression predicting class membership relative to the Non-symptomatic class (N=1007)

Regressors	Descriptive Statistics by LCA Classes				Associations of Cannabis Use with LCA Classes
	Non-symptomatic use	Non-recreational use	Moderate use problems	Severe cannabis problems	Non-recreational Vs. Non-symptomatic		Moderate Vs. Non-symptomatic		Severe Vs. Non-symptomatic
	N=340 N(col %)	N=347 N(col%)	N=81 N(col%)	N=239 N(col%)	AOR (95%CI)	p	AOR (95%CI)	p	AOR (95%CI)	p
Combustible
Non-current use	193 (56.8%)	77 (22.2%)	18 (22.2%)	22 (9.2%)	Ref		Ref		Ref
Infrequent use	78 (22.9%)	81 (23.3%)	11 (13.6%)	25 (10.5%)	1.87 (1.15, 3.05)	.01	0.88 (0.35, 2.22)	.79	1.43 (0.68, 2.99)	.35
Semi-frequent use	56 (16.5%)	82 (23.6%)	26 (32.1%)	62 (25.9%)	3.00 (1.70, 5.30)	<.0001	3.02 (1.29, 7.08)	.01	4.63 (2.23, 9.59)	<.0001
Frequent use	13 (3.8%)	105 (30.3%)	26 (32.1%)	130 (54.4%)	13.9 (5.62, 34.5)	<.0001	9.18 (2.85, 29.5)	<.0001	24.2 (9.11, 68.2)	<.0001
Edible
Non-current use	281 (82.6%)	213 (61.4%)	41 (50.6%)	105 (43.9%)	Ref		Ref		Ref
Infrequent use	45 (13.2%)	70 (20.2%)	22 (27.2%)	59 (24.7%)	1.20 (0.75, 1.92)	.65	1.73 (0.88, 3.41)	.11	1.39 (0.81, 2.39)	.23
Semi-frequent use	11 (3.2%)	46 (13.3%)	12 (14.8%)	42 (17.6%)	2.73 (1.26, 5.94)	.01	2.77 (1.02, 7.54)	.05	2.57 (1.12, 5.90)	.03
Frequent use	3 (0.9%)	18 (5.2%)	6 (7.4%)	33 (13.8%)	1.67 (0.40, 7.03)	.48	3.35 (0.65, 17.4)	.15	2.32 (0.54, 10.0)	.26
Vaporized
Non-current use	209 (61.5%)	129 (37.2%)	19 (23.5%)	59 (24.7%)	Ref		Ref		Ref
Infrequent use	79 (23.2%)	67 (19.3%)	18 (22.2%)	33 (13.8%)	1.18 (0.77, 1.81)	.44	2.21 (1.07, 4.60)	.03	1.26 (0.71, 2.24)	.43
Semi-frequent use	44 (12.9%)	75 (21.6%)	23 (28.4%)	70 (29.3%)	1.85 (1.15, 2.99)	.01	3.46 (1.64, 7.29)	.001	2.52 (1.43, 4.42)	.001
Frequent use	7 (2.1%)	76 (21.9%)	21 (25.9%)	77 (32.2%)	9.18 (3.91, 21.6)	<.0001	14.7 (5.20, 41.5)	<.0001	9.86 (3.97, 24.5)	<.0001
Concentrate
Non-current use	322 (94.7%)	268 (77.2%)	56 (69.1%)	126 (52.7%)	Ref		Ref		Ref
Infrequent use	9 (2.6%)	31 (8.9%)	12 (14.8%)	32 (13.4%)	1.84 (0.79, 4.3)	.16	2.89 (1.05, 7.93)	.04	2.81 (1.16, 6.83)	.02
Semi-frequent use	8 (2.4%)	28 (8.1%)	10 (12.3%)	42 (17.6%)	1.01 (0.40, 2.55)	.99	1.58 (0.52, 4.86)	.42	1.77 (0.70, 4.50)	.23
Frequent use	1 (0.3%)	20 (5.8%)	3 (3.7%)	39 (16.3%)	2.87 (0.34, 24.0)	.33	1.71 (0.15, 19.3)	.66	6.39 (0.77, 52.9)	.09
Blunt use
Non-current use	235 (69.1%)	157 (45.2%)	34 (42.0%)	52 (21.8%)	Ref		Ref		Ref
Infrequent use	58 (17.1%)	83 (23.9%)	17 (21.0%)	48 (20.1%)	1.93 (1.27, 2.94)	.002	1.50 (.75, 3.00)	.25	2.79 (1.65, 4.75)	<.0001
Semi-frequent use	39 (11.5%)	59 (17.0%)	18 (22.2%)	68 (28.5%)	1.25 (.75, 2.08)	.39	1.48 (.71, 3.09)	.30	3.14 (1.78, 5.54)	<.0001
Frequent use	7 (2.1%)	48 (13.8%)	12 (14.8%)	69 (28.9%)	4.60 (1.86, 11.4)	.001	4.03 (1.31, 12.4)	.02	10.32 (4.00, 26.6)	<.0001

Non-current use defined as never users and prior users; Infrequent use defined as current use on 1 or 2 of the past 30 days; Semi-frequent use was defined as use on 3-9 of the past 30 days; Frequent use was defined as use on 10 or more of the past 30 days

In sensitivity analyses, frequency of use of each type of cannabis product was associated with membership in each class relative to the non-symptomatic class, with ORs ranging from 1.07 to 1.19 for each additional day of use in the past 30 days (the association of concentrate cannabis with moderate vs. non-symptomatic dependence problems was not statistically significant but the effect was of similar magnitude; p=0.08).

4. Discussion

This application of LCA provided an empirical grouping of patterns of problematic cannabis use behaviors in a sample of young adult cannabis users in a market at a time of increasing product diversification and proliferation in routes of administration. To our knowledge, this is the first study to employ latent class analysis to define profiles of young adult cannabis users according to their expression of problematic cannabis use behaviors as measured by the CAST, and explore variations in types of cannabis products used, and frequency of use between those classes. LCA identified four distinct classes of problematic use that were divided between non-symptomatic participants, those who reported non-recreational use problems (using before noon, using alone), moderate use problems (using before noon, using alone, and experiencing memory problems due to cannabis) and severe cannabis problems (high probability of endorsing all six problematic cannabis items on the CAST). These distinct classes underscore the validity in using this methodological approach to capture the variability in problematic cannabis use behaviors expressed by young adults. Given the scarcity in literature involving hazardous cannabis use in this age group, this approach to identifying and understanding individuals in terms of specific clusters of disordered cannabis-related behaviors should be replicated in future research studies. Further, a brief screening measure such as the CAST could be a highly useful tool to adopt within primary health care settings to screen for individuals with heightened risk profiles.

More than 80% of all cannabis users in our study reported at least one cannabis use problem, with nearly a quarter of the sample being classified as having ‘severe’ cannabis use problems, with high probability of endorsing all 6 problems on the CAST. About a third of the sample was classified as having no problematic cannabis use, a group which was largely composed of individuals who had experimented with cannabis but had not reported use in the past 30 days. Problematic cannabis use has been linked to a number of physical and psychosocial hazards including lower educational attainment, motor vehicle crashes, cognitive decline, as well as increased risk for abuse and dependence (Hall and Lynskey, 2020; Hartman and Huestis, 2013; Lopez-Quintero et al., 2011; Squeglia et al., 2009; Taylor et al., 2017; Volkow et al., 2014). Our findings are similar to other studies that have sought to characterize problematic use of cannabis products. A recent study found four distinct classes of cannabis users characterized by varying frequencies of use and experiences of cannabis-related consequences in a sample of college students (Pearson et al., 2017); another recent study distinguished unique classes of adult cannabis users ranging from infrequent users with few problems to heavy users with more problems (Manning et al., 2019). These studies have identified important demographic (i.e. gender, race/ethnicity) and affective factors (i.e., emotion regulation, personality traits) associated with different classes of use (Krauss et al., 2017; Pearson et al., 2017; Piontek et al., 2011; Taylor et al., 2017), but have not explicitly examined the role of type of cannabis product used and frequency of use.

In our study, cannabis product type and frequency of use were consistently associated with classes of problematic cannabis use, supporting previous research indicating that cannabis abuse and dependence is associated with types of cannabis product used (Chan et al., 2017; Freeman et al., 2018; Meier, 2017). Combustible cannabis, vaporized cannabis, and blunt cannabis were the strongest predictors of membership in the highest risk class, which had been characterized by high rates of endorsing all six adverse cannabis-related items on the CAST, with more frequent use patterns (i.e., use on 10 or more of the past 30 days), conferring a greater odds of membership in this class. This finding is consistent with previous literature suggesting that higher potency cannabis products lead to higher tolerance and dependency compared to low potency cannabis products (Barrington-Trimis et al., 2020a; Loflin and Earleywine, 2014). Frequent use (10 or more of the past 30 days vs use non-current use) of these same cannabis products (combustible, vaporized, and blunt cannabis) were also associated with increased odds of being classified in the non-recreational or moderate use classes (vs. non-symptomatic). Notably, semi-frequent (but not frequent) use of edibles was associated with an approximate 2.7-fold increase in the odds of belonging to any of the three problematic use classes; the lack of association for frequent use may be due to sample size, as the proportion reporting such frequent use of edible cannabis was low, and confidence intervals were wide. Only infrequent use of concentrate cannabis was associated with membership in the moderate or severe (vs. non-symptomatic) class; higher patterns of use were not statistically associated with greater odds of membership in problematic use classes, but again, confidence intervals were wide due to small sample size, particularly for those using concentrate cannabis on 10 or more days but who were non-symptomatic (N=1). Overall, findings suggest that higher frequency of use is generally more strongly associated with more severe cannabis use problems, an association which appears to be strongest for use of combustible cannabis, which was also the most commonly use cannabis product.

Although the current findings add to the nascent literature on cannabis abuse and dependence, interpretation of the present analyses should be taken in light of a few potential study limitations. First, due to the highly data-driven nature of LCA, generalizability of the current findings are limited, and LCA should be replicated in another representative sample. In addition, all of our measures were self-report, which poses the inherent potential for biases to have influenced the data. For example, participants’ desire to be viewed positively may have resulted in artificially low prevalence of reporting problematic cannabis use behaviors. Additionally, our results may have been influenced by recall biases due to inconsistencies in time frames asked in key questionnaire survey items on cannabis use (e.g. lifetime behaviors, past 6 month behaviors, and past 30 day behaviors). There was substantial multi-collinearity between combustible cannabis and blunt cannabis; thus, analyses of blunt cannabis use were not able to be adjusted for combustible use. Lastly, because this study looked at 18-21 year-olds, these findings may not be generalizable to adolescent cannabis users. Nevertheless, these results provide a detailed description of the heterogeneity in cannabis use problems experienced by young adults and their differential associations with specific cannabis consumption practices and sociodemographic characteristics that have not yet been explored. Our findings illustrate the importance of future research investigating types of cannabis products, dosage, potency and problematic cannabis use to better discern the relationship these products and cannabis problems among young people.

5. Conclusions

Results from the current study indicated that young adult cannabis users demonstrate variations in cannabis-related harms, and that this variation may be attributable to more frequent use patterns and use of particular cannabis products over others. Combustible cannabis, vaporized cannabis, and blunt cannabis use may confer the most risk for cannabis abuse and dependency outcomes among young adults, with more frequent days of use contributing to increased patterns of risk. These findings suggest that young adults who report cannabis use may benefit from screening for problematic use, such as memory loss or social/relational problems that may signal risk of current or future abuse or dependency. Given recent evidence suggesting increases in daily cannabis use in young people, further research involving differential patterns of cannabis consequences and their associations with cannabis consumption practices are needed.

Table 2.

Conditional Item-Response Probabilities

	1 Non-symptomatic (33.8%; N = 340)	2 Non-recreational use (34.5%; N = 347)	3 Moderate use problems (8.0%; N = 81)	4 Severe cannabis problems (23.7%; N = 239)
1. Have you used cannabis before noon?	0.138	0.831	0.826	0.980
2. Have you used cannabis when you were alone?	0.186	0.914	0.758	0.987
3. Have you ever had memory problems when you used cannabis?	0.220	0.575	1.000	0.945
4. Have friends or members of your family ever told you that you should reduce your cannabis use?	0.023	0.189	0.000	1.000
5. Have you ever tried to reduce or stop your cannabis use without succeeding?	0.026	0.079	0.517	0.830
6. Have you ever had problems because of your use of cannabis (argument, fight, accident, bad result at school, etc.)?	0.032	0.037	0.406	0.685

Notes:

Item probabilities >.6 are bolded to indicate a high degree of class homogeneity

Highlights.

• Young adult cannabis users demonstrate distinct variations in their expression of cannabis problems.

• Cannabis product type and frequency of use were consistently associated with classes of problematic cannabis behaviors.

• Combustible cannabis, vaporized cannabis, and blunt cannabis were the strongest predictors of membership in higher risk classes.

• Future research involving types of cannabis products, dosage, potency, and problematic cannabis use patterns among young people are needed.