Adolescent Cannabis Use and Conduct Problems: The Mediating Influence of Callous-Unemotional Traits

Abstract

Cannabis use has been linked to an increased risk of engaging in conduct problem behaviors. However, little existing research has considered intervening processes and shared risk factors that may contribute to this association. The current investigation examines whether callous-unemotional traits, which have shown associations with adolescent cannabis use and conduct problem development, may exhibit a mediating influence on this relationship. Using a longitudinal cohort of youth (n = 390) at increased risk for escalating in their use of cannabis, we found that baseline cannabis use (age~15) was associated with higher levels of trait-like conduct problems (ages~16 & 17), even after controlling for important autoregressive and cross-lagged effects, along with a number of other shared risk factors (e.g., co-occuring substance use, age, sex). Findings also revealed that callous-unemotional traits partially mediated this relationship, with the hypothesized model accounting for approximately one-third of the variance in the conduct problem outcome (R²=.34). These results indicate that callous-unemotional traits may play an important intermediary role in the association between cannabis use and the development of problem behaviors.

Introduction

The initiation of cannabis use most often occurs during adolescence, and recent findings show a decrease in the perceived risks of using cannabis among teenage youth (Johnston, O’Malley, Bachman, & Schulenberg, 2011). As a sensitive period for both normative and maladaptive facets of development, adolescence represents a time during which youth may be particularly vulnerable to the adverse effects of substance use (Rubino et al., 2009; Schneider & Koch, 2003; Spear, 2010; Steinberg, 2005). This, along with changes to the legal landscape in relation to the legalization of cannabis for medical and recreational purposes, have led to elevated concerns regarding the use of cannabis among adolescent youth (Pacula, Kilmer, Wagenaar, Chaloupka, & Caulkins, 2014; Volkow, Baler, Compton, & Weiss, 2014). The scientific community has echoed these concerns through an increased focus on cannabis. One particular area receiving heightened attention is the association between cannabis use and poor behavioral outcomes (Bennett, Holloway, & Farrington, 2008; Pardini, Bechtold, Loeber, & White, 2015; White, Bechtold, Loeber, & Pardini, 2015).

Individuals with cannabis use disorders often exhibit persistent impulsiveness (Hillege, Das, & de Ruiter, 2010; Kimonis, Tatar II, & Cauffman, 2012), a multifaceted personality construct that is commonly associated with addiction, poor decision-making, and maladaptive behavioral outcomes (Hodgins & Holub, 2015; Passanisi & Pace, 2017). In addition, early-onset regular cannabis use during adolescence has been associated with increased rates of delinquency, aggression, and violence (Brook, Zhang, & Brook, 2011; Green, Doherty, Stuart, & Ensminger, 2010; Pardini et al., 2015; Passarotti, Crane, Hedeker, & Mermelstein, 2015). While multiple shared risk factors (e.g., trauma, peer networks, susceptibility genes) have long been thought to contribute both to substance use and behavior problems (Gotham, Sher, & Wood, 2003; Malone, Taylor, Marmorstein, McGue, & Iacono, 2004), accumulating research suggests that cannabis may also exert a direct influence on the emergence and escalation of problem behaviors (Bechtold, Simpson, White, & Pardini, 2015; Brook, Lee, Finch, & Brook, 2014; Norström & Rossow, 2014; Schoeler et al., 2016). Specifically, cannabis use has been linked to greater engagement in hostile and high-risk impulsive behaviors across both, laboratory (Lane, Cherek, Tcheremissine, Lieving, & Pietras, 2005) and field (Ansell, Laws, Roche, & Sinha, 2015) settings. Evidence from long-term prospective studies has also shown persistent use during adolescence to be associated with increased risk for engaging in antisocial and violent behaviors into adulthood (Brook et al., 2011; Flory, Lynam, Milich, Leukefeld, & Clayton, 2004; Pardini et al., 2015; Tucker, Luu, & Derryberry, 2005). Yet despite a recent surge in research suggesting that persistent cannabis use during adolescence may contribute to an increased risk of engaging in conduct problem behaviors, our understanding of the factors that may drive this process remains remarkably limited.

One potentially important mediating mechanism that has shown evidence of associations with both, cannabis use and conduct problem behaviors, is callous-unemotional traits. Callous-unemotional traits, which are analogous to the affective facet of adult psychopathy (e.g., a lack of guilt/remorse, insensitivity to others’ feelings, shallow affect), are posited to drive the downstream development of severe and protracted conduct problem behaviors (Byrd, Hawes, Loeber, & Pardini, 2016; Byrd, Loeber, & Pardini, 2012; Hawes, Byrd, Waller, Lynam, & Pardini, 2017) via a unique etiological pathway (Frick & Dickens, 2006; Frick & White, 2008). Notably, youths high in callous-unemotional traits exhibit deficits in emotion processing, and attenuated amygdala reactivity in response to emotion learning and recognition tasks (Blair, Budhani, Colledge, & Scott, 2005; Dolan & Fullam, 2006; Fairchild, Stobbe, van Goozen, Calder, & Goodyer, 2010; Finger et al., 2011), particularly when viewing fearful faces (Marsh & Cardinale, 2012; Marsh et al., 2008; Viding et al., 2012; White et al., 2012). This lessened ability to recognize fear and distress in others is considered central to the impaired development of empathic understanding among these youths, as well as their lack of concern over the effects of their behavior on others (Marsh et al., 2008). While callous-unemotional traits have been linked to the emergence and escalation of severe conduct problem behaviors, accumulating research also suggests that early-onset and persistent use of cannabis may influence the development of these traits.

Notably, research suggests that deficits in emotion processing, similar to those associated with the development of callous-unemotional traits, are often a consequence of cannabis use (Hindocha et al., 2014; Platt, Kamboj, Morgan, & Curran, 2010). Cannabinoid receptors (CB1) in the brain are densely distributed in cortico-limbic regions instrumental to emotion processing, such as the amygdala (Crane, Schuster, Fusar-Poli, & Gonzalez, 2013; Herkenham et al., 1990; Katona et al., 2001; Laviolette & Grace, 2006), and have been shown to attenuate amygdala reactivity when processing intensely fearful faces (Fusar-Poli et al., 2009). Several studies have found links between persistent use of cannabis and diminished amygdala responsiveness to fearful faces (Bhattacharyya et al., 2009; Bhattacharyya et al., 2010; Bossong et al., 2013; Gruber, Rogowska, & Yurgelun-Todd, 2009). Further, results from several large-scale longitudinal studies suggest that persistent substance use during adolescence may lead to long-term increases in callous-unemotional traits (Hawes et al., 2015; Pardini et al., 2015). In a recent study that followed 503 males for more than 20 years, Pardini et al. (2015) found that engaging in chronic cannabis use throughout adolescence led to increased risk for exhibiting callous-unemotional traits into participants’ mid-30s. Importantly, these findings were maintained even after controlling for co-occurring substance use, and a number of relevant pre-existing risk factors (e.g., early conduct problems, delinquency, baseline callousness, deviant peers).

To date, little existing research has considered which factors may drive the association between cannabis use and conduct problem behaviors. Identifying such influences is a critical step for developing appropriate targets for early prevention and treatment efforts. The current longitudinal investigation begins to address this limitation by examining callous-unemotional traits as a potential mediator of this association among a sample of youth at risk for escalation in their use of cannabis— a group for whom these relationships are particularly important. First, analyses were conducted to examine the influence of cannabis use assessed during middle adolescence (age ~ 15) on subsequent levels of conduct problem behaviors, measured prospectively across two annual follow-ups (ages ~ 16 & 17). Callous-unemotional traits were then examined as a potential mediator of this association. We hypothesized there would be 1) a significant direct effect of cannabis use on prospective conduct problems; and 2) callous-unemotional traits would partially mediate this relationship.

Method

Design and Participants

Participants included 390 adolescents (46% girls) taking part in an ongoing longitudinal investigation examining the relationship between cannabis use, neurocognitive functioning, and socio-emotional development (DAxxxxxx; PI: Xxxxxx). This community sample of youth were between 14–16 years-old (x = 15.40, SD = .72) at the time of the baseline (‘T1’) assessment and were predominantly Hispanic/Latino (90%), reflecting the demographics of the Miami-Dade community from which they were recruited. Participants were recruited from the community via strategic placement of recruitment materials and study staff at locations identified as providing high visibility for youth in this age range (e.g., shopping malls, outside high schools, parks, movie theatres). Study incentives increased over time (i.e., Visit 1- $100, Visit 5- $140) and participants earned a bonus ($20) for completing all visits. A primary aim of the overall longitudinal investigation was to enroll a large number of youth who reported prior cannabis use, but who had a low prevalence of cannabis dependence at the time of their T1 evaluation. In turn, study inclusion criteria consisted of the ability to read and write in English, along with any self-reported use of cannabis, alcohol, or cigarettes, even if only minimal (e.g., a sip of alcohol, puff from a cigarette or joint). In addition, the study sample also included a small subset of youth (10%; n = 40) reporting no history of substance use at the time of the T1 assessment.

Study exclusionary criteria included self-reported developmental disorders, neurological conditions, birth complications, history of a significant psychiatric or mood disorder (i.e., received a diagnosis and were either prescribed medication for or underwent therapy in relation to that specific diagnosis), traumatic brain injury or loss of consciousness >10 minutes, history of significant alcohol or other substance use suggestive of an alcohol use disorder or a substance use disorder, use of other drugs (besides alcohol, cannabis, and nicotine) more than 10 times, use of any other drugs in the two weeks prior to assessment, and use of any other drug to an extent greater than their cannabis use. Of the 1,183 youth who were screened for this study, 782 were excluded based on these criteria, thus resulting in the final study sample at baseline (n = 401). All study procedures were approved by the Institutional Review Board at Florida International University and were in compliance with the Helsinki Declaration of 1975, as revised in 2013 (World Medical Association, 2015). Adolescent assent and parental consent were obtained prior to each assessment from participants and/or participant’s parent/guardian.

Measures

Substance Use.

The Drug Use History Questionnaire (DUHQ; Gonzalez et al., 2012; Sobell, Kwan, & Sobell, 1995) is a semi-structured interview that uses the time-line followback procedure to obtain self-reported frequency and quantity of substance use across 15 different classes of substances during a participant’s lifetime, the past 6 months, and the past 30 days (e.g., ‘When was the last time you used marijuana?’; ‘How many times have you used marijuana in the past 6 months’). This measure has demonstrated evidence supporting its validity in assessing drug use history for a range of substances across both, adolescent (Gonzalez et al., 2012) and adult samples (Gonzalez et al., 2012; Rippeth et al., 2004). Cannabis use frequency in the past 6 months (i.e., total number of days used in the past 6 months), assessed at the time of the T1 assessment, was the primary cannabis use measure in our analyses. This instrument was also used to measure and control for co-occurring use of alcohol and nicotine at T1. As with the cannabis use variable, use of alcohol and nicotine was assessed as frequency (i.e., total number of days) of use in the past 6 months.

Callous-Unemotional Traits.

The Inventory of Callous-Unemotional Traits (ICU; Frick, 2004), a 24-item self-report measure intended to assess callous-unemotional traits in children and adolescents, was completed by study participants at T1. The ICU was developed based on items from the Antisocial Process Screening Device (APSD; Frick & Hare, 2001) that most consistently loaded on the callous-unemotional factor in previous studies (e.g., “Concerned about the feelings of others,” “Feels bad or guilty when he/she does something wrong”). Prior work suggests the ICU consists of 3 underlying facets (Callousness, Uncaring, Unemotional), which are subsumed by a general callous-unemotional factor (Byrd, Kahn, & Pardini, 2012; Fanti, Frick, & Georgiou, 2009; Kimonis et al., 2008). In the current study, summed scores on these 3 facets are modeled as a latent callous-unemotional factor. The average observed total score on the ICU measure for participants in this study was 23.06 (SD = 7.30; range = 0 – 44; 25^th%_tile = 18, 50^th%_tile = 23, 75^th%_tile = 28)

Conduct Problems.

The Computerized Diagnostic Interview Schedule for Children, Version IV (CDISC-IV; Shaffer, Fisher, Lucas, Dulcan, & Schwab-Stone, 2000) was used to assess self-reported conduct disorder symptoms. The CDISC-IV is a structured interview that uses a series of standardized probes and follow-up questions to gather information about DSM-IV disorder symptoms (e.g., aggression towards other people or animals, destruction of property, deceitfulness or theft). It has demonstrated evidence of reliability and construct validity in previous investigations (for a review, see Malgady, Rogler, & Tryon, 1992). Youth self-reported conduct disorder symptoms were assessed during participant’s 1-year (T2) and 2-year (T3) annual follow-up assessments. In the current study, CDISC conduct disorder symptoms at T2 and T3 were used to model a time-invariant latent construct of conduct problems.

Baseline Conduct Problems.

The Youth Self Report (YSR; Achenbach, 1991) was used to assess conduct problems at T1 in order to control for autoregressive effects on the study outcome. YSR items are rated from 0 (“not true”) to 2 (“very true or often true”) and are summed together to form composite syndrome scale scores. The YSR Aggressive Behavior (e.g., physically attacks people, threatens people, destroys property) and Rule-Breaking (e.g., stealing, vandalism, lying or cheating) scales were used to model a latent construct of early conduct problems.

Data Analytic Plan

To evaluate callous-unemotional traits as a mediator of the association between cannabis use and conduct problems, the hypothesized structural model for the current study included 3 paths: 1) the effect of cannabis use on callous-unemotional traits [Path a]; 2) the effect of callous-unemotional traits on conduct problems [Path b]; and 3) the direct effect of cannabis use on conduct problems [Path c’]. Using a structural equation modeling (SEM) approach, the callous-unemotional traits mediator was specified by having the 3 subscales of the ICU measure load onto a latent callous-unemotional construct. In addition, a latent trait-like construct of conduct problems was modeled by constraining loadings and intercepts of CDISC-IV conduct problem symptom scores to equality across the T2 and T3 follow-ups.

Importantly, this approach is beneficial because latent variables are free of measurement error and result in increased power and less biased estimates (Busemeyer & Jones, 1983; Kenny & Judd, 1984; Little, Bovaird, & Widaman, 2006). This is particularly true in mediation analyses, where measurement error associated with a mediating variable can have a severe impact on parameter estimates (Muthén & Asparouhov, 2015). Further, an SEM framework provides the ability to analyze complex associations in a more parsimonious single model, compared to traditional regression techniques that involve testing multiple path analytic models. All study analyses controlled for relevant demographic information (i.e., age, ethnicity, and gender) and the influence of co-occurring use of alcohol and tobacco. Additionally, to provide a more stringent test of mediation, conduct problems assessed at T1 were included as a model covariate to account for potential autoregressive influences on the study outcome.

Analyses were conducted using Mplus 7.2 (Muthén & Muthén, 1998–2012). Significance of the mediation path (i.e., indirect effect) was assessed via the Mplus IND command. Standard errors and confidence intervals of model path coefficients were estimated using nonparametric bootstrap sampling procedures (10,000 resamples), which do not require normality of the sample distribution. This is of notable importance when evaluating mediation models, as indirect effects often have non-normal sampling distributions (Preacher & Hayes, 2008; Preacher & Kelley, 2011). We also report a standardized effect size of the indirect effect, kappa squared (κ²). This estimate, which is insensitive to sample size, represents the ratio of the maximum possible indirect effect for a given model. Maximum likelihood estimation was used for all study analyses. Due to the count nature of the observed conduct problem outcome, a negative binomial link function was specified to account for over-dispersion of the count variable in the calculation of standard errors (Raftery, 1995).

Missing Data

The proportion of participants with available data at the T1 evaluation was 400 (99%; 1 participant had missing data on all variables examined in this study) and retention rates across the study follow-up period remained high (94%-98%). Full-information maximum likelihood (FIML) estimates were used to handle missing data, as this procedure uses all available data points to construct parameter estimates under the assumption that the data are missing at random. The effective sample size for all study analyses was 390 (cannabis use data was unavailable for 10 participants; these participants were thus excluded from further analyses). No differences were found on any study variable when comparing participants with and without complete data. In addition, even when data are not missing at random, FIML estimation tends to produce less biased estimates than more traditional techniques for handling missing data (Enders & Bandalos, 2001).

Results

Descriptive information, including bivariate correlations between study variables (or their observed indicators) can be found in Table 1. In addition to relationships between primary study variables, these descriptive results also highlight several other associations worth pointing out. Notably, there was evidence of an association between cannabis use and callous-unemotional traits, as small but significant effects were found between cannabis use and the Uncaring and Callousness subscales of the ICU measure. However, no such associations were found when examining the relationship between alcohol or tobacco use and callous-unemotional traits. Further, the magnitude of the effect between cannabis use and conduct problem behaviors was also stronger than the effects between alcohol or tobacco use and conduct problems. It is also worth noting that the Unemotional subscale of the ICU was not significantly associated with any variable other than the other ICU subscales. Table 1 provides additional descriptive information.

Table 1.

Descriptives and correlations for study variables.

	M (SD)	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.
1. DUHQ Cannabis Use	27.95 (46.01)	-
2. ICU Uncaring Subscale	7.69 (3.99)	.15**	-
3. ICU Callousness Subscale	7.97 (3.11)	.16**	.42***	-
4. ICU Unemotional Subscale	7.38 (2.82)	.02	.23***	.23***	-
5. CDISC-IV Conduct Problems (T2)	4.63 (3.77)	29***.	.29***	.33***	.05	-
6. CDISC-IV Conduct Problems (T3)	4.57 (3.84)	28***.	.31***	.24***	−.04	.78***	-
7. DUHQ Alcohol Use (T1)	4.92 (9.48)	22***.	.02	.03	−.03	.05	.12	-
8. DUHQ Nicotine Use (T1)	3.95 (17.65)	.20***	.02	.03	.03	.18***	.11	.11*	-
9. YSR Rule-Breaking (T1)	7.62 (4.92)	32***.	.35***	.33***	.06	.46***	.42***	.18***	.11*	-
10. YSR Aggressive Behavior (T1)	6.39 (5.20)	.16**	.31***	.28***	.03	.31***	.22**	.04	.04	.71***	-
11. Age(T1)	15.40 (.72)	.16**	−.01	.09	.05	.10	.11	.07	.10*	.15**	.08	-
12. Ethnicity (% Hispanic)	89%	.08	−.02	−.07	.01	.04	−.01	.10*	.06	−.01	−.06	−.05	-
13. Sex (% Male)	54.1%	−.12	−.17**	−.19***	−.02	−.32***	−.24**	.03	−.08	−.15**	−.02	−.10*	−.02	-

Notes.

^***= p < .001;

^**= p < .01;

^*= p < .05.

For correlations with participant’s sex, ‘Male’ is the reference category.

DUHQ = Drug Use History Questionnaire; ICU = Inventory of Callous-Unemotional Traits; CDISC-IV = Computerized Diagnostic Interview Schedule for Children (Version IV); YSR = Youth Self Report.

In evaluating study hypotheses, all models controlled for participant demographics (i.e., age, race, gender), co-occuring substance use (i.e., alcohol and nicotine use), and autoregressive effects of T1 conduct problems. First, examination of the direct effect of T1 cannabis use on the latent trait-like conduct problems outcome revealed a small-medium sized, positive association (β = .19, SE = .05, p < .001, 95% CI [.09, .27]). Subsequent to assessing this direct effect, analyses were conducted to evaluate the hypothesized mediation model. Figure 1 presents parameter estimates for all direct and indirect effects for this model. These results showed a significant, small-medium sized effect between cannabis use and callous-unemotional traits (Path a; β = .22, SE= .06, p < .001, 95% CI [.12, .33]). Higher levels of callous-unemotional traits, in turn, prospectively predicted more trait-like conduct problems (Path b; β = .28, SE = .11, p < .01, 95% CI [.10, .48]). As noted, there was a significant direct effect of cannabis use on the conduct problems outcome (i.e., Path c; β = .19, SE = .05), and after controlling for the mediating effect of callous-unemotional traits, this direct effect remained significant but this size of the effect was attenuated (Path c’; β = .14, SE= .05, p < .01, 95% CI [.04, 23]). The indirect path from cannabis use to conduct problems via callous-unemotional traits was also significant (Path a*b; β = .07, SE = .03, p = .02, 95% CI [.03, .14]), indicating there was a significant partial mediation effect. The magnitude of this indirect effect was in the medium range (κ² = .06). Overall, the hypothesized model accounted for nearly one third of the variance in the conduct problem outcome (R² = .34)¹. In addition, although there is little theoretical or empirical rationale to suggest an alternative temporal ordering of the study predictor and mediator, due to these variables being assessed contemporaneously, we examined the potential for a reverse causal influence (i.e., cannabis use as mediator). Results did not support this alternative model (indirect effect- β = .02, SE = .01, p = .19).

Figure 1.

Measurement model for cannabis use predicting conduct problems as mediated by callous-unemotional traits

Discussion

The current study investigates the prospective association between cannabis use and conduct problems during adolescence and examines whether callous-unemotional traits may exert a mediating influence on this relationship. Consistent with an accumulating body of longitudinal research (Bechtold et al., 2015; Brook et al., 2014; Norström & Rossow, 2014; Schoeler et al., 2016), we found that greater levels of cannabis use during adolescence were associated with increased risk for engaging in conduct problem behaviors. Importantly, findings revealed that this relationship continued to exist even when taking into account important shared risk factors and confounding influences. Extending on prior work in this area, and in line with study hypotheses, callous-unemotional traits were found to exhibit a significant mediating influence on the process by which cannabis use relates to the later expression of conduct problem behaviors. More specifically, these findings provide support for a pathway by which cannabis use during adolescence contributes to increased levels of callous-unemotional, which in turn lead to greater engagement in conduct problem behaviors.

As a critical period of neural maturation, adolescence marks a time during which youth are particularly vulnerable to the adverse effects of substance use (Clark, Thatcher, & Tapert, 2008; Squeglia, Jacobus, & Tapert, 2009). During this time, processes associated with affective problems, as well as maladaptive substance use and addiction outcomes, such as lack of effortful control and poor emotion regulatory abilities (Moffitt, Caspi, Harrington, & Milne, 2002; Pace et al., 2014; Rothbart, Ellis, Rueda, & Posner, 2003) undergo marked change (Oldehinkel, Hartman, Ferdinand, Verhulst, & Ormel, 2007; Zeman, Cassano, Perry-Parrish, & Stegall, 2006). Consequently, investigating the impact of cannabis use during adolescence is essential. Notably, deficits in emotion processing considered to result from chronic cannabis use overlap with those thought to precipitate callous-unemotional traits. Results from this study suggest that frequency of cannabis during adolescence may serve to increase callous-unemotional trait levels among some youth. In this regard, it is conceivable that chronic use of cannabis during this sensitive period could impair normative maturation in key brain regions responsible for emotional development (e.g., corticolimbic structures), leading to poorer fear recognition and reduced empathic concern, which are hallmark facets of callous-unemotional traits.

It is also worth pointing out, and somewhat surprising, that only cannabis use frequency and not alcohol or tobacco use was significantly associated with callous-unemotional traits (see Table 1). Similarly, associations between cannabis use and conduct problem behaviors were also stronger than those of alcohol and tobacco use. However, it bears repeating that although the effect of cannabis use frequency on callous-unemotional traits was significant in this study, the magnitude of this effect was relatively modest. This suggests that while adolescent cannabis use may exert some influence toward elevated callous-unemotional traits, the lion’s share of heterogeneity in this construct is accounted for elsewhere. Nonetheless, it is worth noting that the association between cannabis use and callous-unemotional traits remained significant even after accounting for several theoretically relevant confounds, including early conduct problems and concurrent use of alcohol and nicotine. This is consistent with recent research showing the association between cannabis use and callous-unemotional traits to persist even when controlling for a comprehensive set of confounding variables (Pardini et al., 2015).

Callous-unemotional traits exhibited a moderate sized mediation effect (β = .07, SE = .03) on the pathway from cannabis use to conduct problems in this study, even after controlling for the autoregressive influence of early conduct problems. Although the overall model explained a substantial portion of variance in the study outcome, it did not fully explain the association between cannabis use and conduct problems. Including callous-unemotional traits in the model led to an attenuated direct effect; however, the link between cannabis use and conduct problems remained significant. This is not surprising, as robust and multifaceted associations between substance use and maladaptive behaviors have been extensively documented (Elliott, Huizinga, & Menard, 2012; Loeber, Farrington, Stouthamer-Loeber, & Van Kammen, 1998). We also note that the Unemotional subscale of the ICU was not significantly associated with any other study variables other than the Uncaring and Callousness subscales of the ICU. This finding is consistent with results from several other studies that have suggested the Unemotional subscale of the ICU may not tap into the nomological network of callous-unemotional traits as precisely as intended (Hawes et al., 2014; Pechorro, Hawes, Gonçalves, & Ray, 2016; Waller et al., 2015). Finally, it is important to point out that even though this study investigates a novel mediation pathway controlling for important shared risk factors, this is without question an oversimplification of a very complex relationship. Indeed, these processes are almost certainly interconnected by a number of reciprocal and bidirectional associations, as well as other third variable influences (e.g., stressful life events, peer networks, resource availability, gene by environment interactions) that impact how each unfolds across development.

Strengths and Limitations

This study had several important strengths including the use of sample consisting of a large number of youth who reported prior cannabis use and who were then assessed prospectively during adolescence. Furthermore, the implementation of a latent variable modeling framework reduced the influence of measurement error allowing for a more parsimonious examination of these associations than would be permitted by traditional regression techniques. We also controlled for several theoretically important confounds, including early conduct problems, to reduce potential bias resulting from autoregressive influences on the study outcome. However, study findings should also be interpreted within the context of several limitations. This study primarily consisted of Hispanic youth and findings may not generalize well to non-Hispanic populations or different developmental periods. Youth enrolled in this study exhibited diverse levels of cannabis use at the time of their baseline evaluation, but did not meet criteria for dependence, which may also affect generalizability. A small subset of youth screened for inclusion into this study (<5%) reported having been diagnosed with major depression or an anxiety disorder and reported undergoing treatment (i.e., medication or therapy) for that diagnosis, and as a result were excluded. Although this base rate is relatively small, it is possible that this exclusion criteria may affect associations between some study variables. In addition, cannabis use was assessed based on participant’s self-reported frequency of use and therefore, other factors that may be important for understanding developmental patterns of cannabis use were not taken into account (e.g., mode of administration, potency, quality, quantity, or type of cannabis). Replication across a larger temporal window, using multi-informant assessments is an important avenue of future research. While temporal precedence was established for all study variables in relation to the conduct problem outcome, the study predictor and mediator were assessed concurrently. Although there seems to be little theoretical or empirical evidence to support an alternative temporal ordering (i.e., callous-unemotional traits exerting a direct influence on cannabis use), we nonetheless examined this model, finding no evidence of a significant mediation effect.

Clinical Implications and Future Directions

Acquiring a better understanding of the complex developmental pathway from adolescent cannabis use to conduct problem behaviors is essential for informing treatment and prevention efforts. Investigating this association during key points of development and is particularly important for enhancing ecological validity of study findings, a critical component for improving targeted interventions. The results presented here suggest a relatively persistent association between cannabis use and conduct problems over time which exists beyond the influence of multiple commonly shared risk factors. Efforts to reduce cannabis use, particularly among high risk youth, may facilitate the attenuation of problem behaviors across longer term development. Further, the reduction of callous-unemotional traits may be an important goal for interventions aimed at treating adolescents with comorbid cannabis use and problem behaviors. Notably, recent evidence demonstrates substantial heterogeneity in the developmental course of callous-unemotional traits, suggesting their amenability to change (Byrd et al., 2016; Fontaine, McCrory, Boivin, Moffitt, & Viding, 2011). Thus, treatment programs may be tailored to target multiple domains of risk that show promise for reducing callous-unemotional traits in youth such as emotion recognition training, social skills and problem solving, and parent training (for a review see Wilkinson, Waller, & Viding, 2015). Future research that replicates these findings is essential, as are studies that investigate other mechanisms that may contribute to the association between cannabis use and poor behavioral outcomes.