Abstract
Background:
Theoretical models of behavior change argue that youth should decrease their time with cannabis-using friends and increase their time with non-using friends during treatment. Informed by behavior-change models of recovery and socialization and selection peer-influence models, the current study examined whether combining evidence-based psychosocial treatment with adjunctive pharmacotherapy helps youth decrease their affiliations with cannabis-using friends and increase their affiliations with non-using friends during cannabis misuse treatment.
Methods:
Youth ages 15-24 years (51% male), participated in a double-blind randomized clinical trial that tested the effects of motivational enhancement and cognitive behavioral therapy (MET-CBT) plus topiramate (N=39) or placebo (N=26) on cannabis craving and use. Ecological momentary assessment data, collected via smartphones throughout the six-week intervention, assessed youths’ time with cannabis-using and non-using friends, cannabis use, and craving in daily life. Multiple group multilevel structural equation modeling tested study hypotheses.
Results:
Across the topiramate (48% completion rate) and placebo (77% completion rate) conditions, greater time spent with cannabis-using friends promoted greater next day cannabis use and craving (socialization effect). In turn, cannabis craving, but not use, promoted continued selection of cannabis-using friends. This indirect effect was only supported in the placebo condition due to the selection piece of this cycle not being significant for youth who received topiramate. Neither cannabis craving nor use were associated with time with non-using friends the next day.
Conclusions:
MET-CBT and adjunctive topiramate pharmacotherapy interrupted youth selection processes. This finding suggests that changing peer affiliations could be one mechanism by which treatments can work.
Keywords: cannabis-using friends, peer influence, topiramate, MET-CBT, mediation
1. Introduction
Theoretical models of behavior change, such as the Social Identity Model of Recovery (Best et al., 2016) and Social Learning Theory (Maisto et al., 1999), posit that transitioning from substance-using to non-using peer affiliations is a critical mechanism of behavior change during treatment for substance misuse. Studies of this potential mechanism have largely focused on adults in Alcoholics Anonymous (AA) and treatments facilitating AA participation. On the whole, research shows that increasing time with non-drinking peers and decreasing time with drinking peers mediates treatments effects for adults (Kelly et al., 2012; Litt et al., 2016). Less work has examined this mechanism in youth (Kelly et al., 2018) and the evidence supporting peer affiliations as a mediator of AA among youth is less consistent (Chi et al., 2009; Kelly et al., 2014). Further, Hoeppner et al. (2014) found that youth were less likely than adults to increase their pro-abstinence social networks through AA attendance. No study has examined shifting peer affiliations in the context of cannabis treatment. In addition, research has focused on whether social networks mediate treatment outcomes and not how these changes in social relationships occur.
1.1. Mechanisms of Friendship Changes
Social developmental models identify selection of like friends and socialization within friend groups as mechanisms by which youth initiate and maintain substance use (Dishion et al., 1995; Hirschi, 1969; Sutherland et al., 1995). Youth who begin outpatient treatment are often entrenched in substance-using friend groups and have little motivation to change their social connections (Chung et al., 2015). For these youth, selection and socialization are likely reciprocal (Caouette and Feldstein Ewing, 2017). Social encouragement and pro-use attitudes of cannabis-using friends likely exacerbates youths’ cannabis cravings and use (socialization effect), which in turn, motivates youth to maintain these relationships with their cannabis-using friends (selection effect) (Brechwald and Prinstein, 2011).
Whereas much research has studied selection and socialization effects in relation to the initiation and escalation of adolescent substance use (McMillan et al., 2018; Osgood et al., 2015; Scalco et al., 2015), only two studies examined selection and socialization during treatment. Becker and Curry (2014) found evidence of selection and socialization for alcohol use but only selection for cannabis use among youth receiving motivational enhancement and cognitive behavioral therapy (MET-CBT). In a study assessing a brief parent-focused intervention, selection and socialization effects were found between baseline and 6-month follow up for both alcohol and cannabis; however, only selection was supported for alcohol and cannabis use between 6- and 12-month follow-ups (Becker et al., 2019). Although these studies speak to selection and socialization broadly, the months-long time gaps between assessment periods preclude an understanding of the day-to-day processes that may lead youth to continue to spend time with using friends and maintain high levels of cannabis use and craving during treatment. Assessing whether the presence of cannabis using friends relates to proximal changes in craving and use (socialization), and whether craving and use predict engagement with these friends (selection) may shed light on how treatments impact friendship processes (Hallgren et al., 2018).
1.2. MET-CBT Combined with Pharmacotherapy
Adjunctive pharmacotherapy is one proposed method to improve evidence-based psychosocial interventions for youth (Miranda and Treloar, 2016). Data for the current study stem from a randomized controlled trial (RCT) that compared the efficacy of topiramate and placebo in combination with MET-CBT on youth cannabis use (Authors, 2017). All youth received three sessions of MET-CBT, which focused on eliciting motivation to reduce cannabis use, building skills to manage cravings (e.g., recognize and avoid triggers), resisting peer influences (e.g., avoid using friends and high-risk situations) and increasing non-using supports and activities (e.g., spend more time with friends who don’t use).
Topiramate is a mixed GABA agonist and AMPA/kainite glutamate receptor antagonist that reduces alcohol, cannabis, and nicotine use in adults (Miranda et al., 2008; Oncken et al., 2014; Simeone et al., 2006). In the paper reporting the primary aims of this RCT, topiramate reduced how many grams of cannabis youth smoked on using days but did not affect how often they smoked (Authors, 2017). During weeks four to six of the RCT, when youth (n=40) reached the target dose for topiramate, Authors (2018) demonstrated that this effect may be due to topiramate blunting the stimulating effects of cannabis and reducing craving.
Building on evidence that MET-CBT may alter peer influences (Becker and Curry, 2014), topiramate + MET-CBT may help youth decrease their time with cannabis-using friends and increase their time with non-using friends. Ideally, medications compliment psychosocial treatments by making it easier for youth to implement skills (Potenza et al., 2011). Time with using friends is thought to promote youths’ cannabis craving and use (socialization) because the presence of using friends promotes high craving and is thought to make use more rewarding (Caouette and Ewing, 2017; Phillips et al., 2018). To address socialization effects, MET-CBT helps youth identify cons of using friendships and teaches skills to resist peer influence. By blunting the rewarding effects of cannabis use and craving in the presence of using friends (Authors, 2018), topiramate may help youth use skills to reduce craving or cannabis use after spending time with using friends. MET-CBT targets selection effects by minimizing the impact of lapses through seeking out non-using supports and encouraging engagement in non-using activities and spending time with non-using friends when experiencing cravings. Through blunting craving and the rewarding effects of cannabis use, topiramate may help youth use skills to spend less time with cannabis-using friends and increase their time with non-using friends.
1.3. Current Study
This study is the first to examine mechanisms that may account for youth decreasing time with cannabis-using friends and increasing time with non-using friends during treatment. Consistent with models of peer influence (e.g., Dishion et al., 1995), we hypothesized that:
Time spent with cannabis-using friends would predict greater cannabis craving and use the next day (socialization effects) (Hypotheses 1a, 1b).
Cannabis craving and use would, in turn, predict greater time with cannabis-using friends (Hypotheses 2a, 2b) and less time with non-using friends (Hypotheses 2c, 2d) the following day (selection effects).
The mediational pathways by which cannabis-using friends elicited greater cannabis craving and use the subsequent day, which in turn, predicted greater time with cannabis-using friends and less time with non-using friends, were hypothesized to be significant indirect effects (Hypotheses 3a-3d).
Because topiramate attenuates cravings elicited by cannabis-using friends, mediational pathways outlined in Hypothesis 3 may be significant only for the placebo group. Specifically, we hypothesized that:
-
4.
Socialization effects of time with cannabis-using friends predicting next-day cannabis craving and use would be significant only in the placebo group (Hypotheses 4a, 4b).
-
5.
Further, we hypothesized that selection effects of cannabis use and craving on use days predicting selection of cannabis-using friends would be significant only in the placebo group (Hypotheses 5a, 5b) and
-
6.
Selection of non-using friends the following day would be significant only in the topiramate group (Hypotheses 5c, 5d).
2. Materials and Methods
2.1. Participants
Eighty-five youth ages of 15 to 24 years were recruited from community settings to participate in this RCT (NCT01110434). Eligible youth used cannabis at least twice weekly in the past 30 days and experienced at least one cannabis use disorder symptom. Youth who were mandated to treatment, diagnosed with a psychological disorder other than cannabis, alcohol, nicotine, or disruptive behavior disorder, actively psychotic symptoms or suicidal, taking medications that contraindicated study medication, and pregnant were excluded. The final secondary-analysis sample consisted of the 65 youth who were randomized to either topiramate + MET-CBT (n=39) or placebo + MET-CBT (n=26) and contributed EMA data during the treatment period (Authors, 2017). The samples contains one less participant than Authors (2017) because one participant did not complete EMA.
2.2. Procedure
2.2.1. Overall Design.
Informed consent was obtained from youth ≥18 years and assent from minors. Parents of minors provided consent. Prior to randomization, eligible participants completed a comprehensive baseline assessment battery and a lead-in EMA period of approximately one week. EMA was collected during the 6-week treatment period, and MET-CBT was individually administered at weeks 1, 3, and 5. Youth were randomized on a 2:1 ratio (topiramate to placebo). Topiramate was titrated over four weeks and stabilized in weeks 5 and 6 (Albsoul-Younes et al., 2004). All procedures were approved by the University Institutional Review Board (Protocol number blinded for review).
2.2.2. EMA Procedures.
EMA was implemented using custom software on Samsung Omnia phones provided by the study (Samsung Electronics, Ridgefield Park, NJ). Participants were trained in the EMA protocol using an age-appropriate manual to self-initiate a begin-use report prior to engaging in cannabis use and a corresponding end-use report when that use episode ended. Non-use moments were sampled using audible-signaled prompts delivered during waking times; prompts were randomly scheduled in each 3-hour block. Participants could suspend prompts when unable to respond. Participants were compensated up to $10 per day.
2.2.3. Psychotherapy Platform.
Participants received three biweekly, 50-minute sessions of a manualized MET-CBT protocol. Masters- and doctoral-level interventionists, who received 20 hours of training and conducted at least two mock cases, delivered individual MET-CBT sessions. Participants and interventionists were blind to the participant’s medication condition.
2.3. Measures
2.3.1. Social Day and Weekend Status.
EMA reports were time and date stamped. Because participants’ social schedules did not always align with a 24-hour calendar day (i.e., awake past midnight), we created a “social day” index to calculate participants’ daily aggregates of focal predictors. Weekends (1) were defined as 6 p.m. on Friday through 6 p.m. on Sunday and weekdays (0) as all other times.
2.3.2. Social Contextual Variables.
Random-prompts and begin-use reports asked “Who are you with? (check all that apply),” with the following options: mother, father, brother(s), sister(s), own child/children, other relative(s), boy/girlfriend, friend(s), teacher(s), other(s), and no one. The categories of friend(s) and boy/girlfriend were combined to form a friends variable (0=friends not present, 1=friends present). Participants were also asked “Are you with people with whom you usually smoke pot?” with yes and no options. If participants indicated they were with friends and they were with people with whom they usually use cannabis, it was determined that the participant was currently in the presence of cannabis-using friend(s). Time with cannabis-using friends was calculated as the proportion of completed daily reports where youth endorsed being in the presence of using friends (coded as 1) relative to non-using friends and non-friends (e.g., family, alone; coded as 0). For example, if a participant completed nine random-prompts and begin use reports on a given day, three times they were with using friends (coded as 1) and six times they were not with friends (coded as 0), their time spent with using friends for that day would be 0.33. Time with non-using friends was calculated as the ratio of reports completed in the presence of non-using friends (coded 1) to using friends and non-friends (coded 0). If a youth reported being in the presence of non-using friends two out of nine completed reports that day, their time spent with non-using friends would be 0.22.
2.3.3. Cannabis Craving.
Random-prompt, begin-use, and end-use reports asked “How strong is your urge to use pot right now?” Responses were recorded using a slider bar with endpoints labeled “no urge” and “strongest ever.” Responses were recorded as whole numbers from 0 to 10. An average of all craving reports for a given day estimated craving that day.
2.3.4. Cannabis Use.
End-use reports assessed daily cannabis use. Grams per episode were calculated by dividing the total number of grams by the number of people shared with. Grams per episode were summed within social day to capture total daily cannabis use. To address extreme outliers, values that exceeded three standard deviations from the mean (N=11, 0.52%) were recoded to three standard deviations from the mean (Tabachnick et al., 2007).
2.4. Analytic Approach
Hypotheses were tested with multiple group multilevel structural equation modeling (MSEM) using Mplus 8.3 (Muthen and Muthen, 1998-2018). MSEM accounted for study days being nested within participants, and allowed for the influence of time with cannabis-using friends on cannabis craving and use (a paths) and the influence of cannabis craving and use on time with cannabis-using friends (b1 path) and non-using friends (b2 path) to be tested simultaneously. The total observed variance was partitioned into day-level effects (level 1), which represent variability from a person’s average level of a behavior (e.g., how much craving a person is experiencing that day relative to their average level of craving across the trial), and person-level effects (level 2), which reflects individual differences (e.g., average level of craving across the trial). Models were estimated separately for cannabis craving and cannabis use to reduce complexity. Models were arranged with paths lagged across days such that time with cannabis-using friends (day X) predicted cannabis use and craving the next day (day X+1), which in turn, predicted time with cannabis-using friends and non-using friends the following day (day X+2). This time course was selected to ensure temporal precedence in the a, b1, and b2 paths. Further, evaluating day-to-day processes associated with behavioral change during treatment using intensive longitudinal data optimized the ability of any findings to help inform intervention development and clinical practice (Carpenter et al., 2020). Figure 1 provides our conceptual 1-1-1 multiple group MSEM.
Figure 1. Conceptual diagram of MSEM.
Note. CU=Cannabis-using, a= a path of mediational chain; within-person time with cannabis-using friends predicting cannabis craving/grams smoked the next day, b1= b1 path of mediational chain; within-person cannabis craving/grams predicting time with cannabis-using friends the following day, b2= b2 path of mediational chain; within-person cannabis craving/grams predicting time with non-using friends the following day, 1a=Hypothesis 1a; in the overall sample, time spent with cannabis-using friends would predict greater cannabis craving the next day, 1b=Hypothesis 1b; in the overall sample, time spent with cannabis-using friends would predict greater grams smoked the next day, 2a=Hypothesis 2a; in the overall sample, cannabis craving would predict greater time with cannabis-using friends the following day, 2b=Hypothesis 2b; in the overall sample, grams smoked would predict greater time with cannabis-using friends the following day, 2c=Hypothesis 2c; in the overall sample, cannabis craving would predict less time with non-using friends the following day, 2d=Hypothesis 2d; in the overall sample, grams smoked would predict less time with non-using friends the following day, 3a=Hypothesis 3a; the a*b1 craving mediational path would be statistically significant, 3b=Hypothesis 3b; the a*b1 grams smoked mediational path would be statistically significant, 3c=Hypothesis 3c; the a*b2 craving mediational path would be statistically significant, 3d=Hypothesis 3d; the a*b2 grams smoked mediational path would be statistically significant, 4a=Hypothesis 4a; path a for craving would be significantly weaker in the topiramate relative to the placebo condition, 4b=Hypothesis 4b; path a for grams smoked would be significantly weaker in the topiramate relative to the placebo condition; 5a= Hypothesis 5a; path b1 for craving would be significantly weaker in the topiramate relative to the placebo condition; 5b=Hypothesis 5b; path b1 for grams smoked would be significantly weaker in the topiramate relative to the placebo condition, 5c=Hypothesis 5c; path b2 for craving would be significantly stronger in the topiramate relative to the placebo condition, 5d=Hypothesis 5d; path b2 for grams smoked would be significantly weaker in the topiramate relative to the placebo condition. The level 1 covariate weekend status and within-person covariances between CU friends and non-CU friends are not depicted to reduce figure complexity.
Not all individuals randomized to topiramate (n=20) or placebo (n=6) completed the RCT (Authors, 2017). The most common side effect leading to dropout in the topiramate condition was memory difficulties (Gray et al., 2018). Models included all randomized individuals who provided EMA data (n=65) and were estimated with Robust Maximum Likelihood (MLR), which accounted for missing data. Time with non-using friends was non-normally distributed (skew=2.27, kurtosis=−5.01) and all other mediator and outcomes variables were normally distributed (skew<∣1∣, kurtosis<∣1.20∣).
Model building occurred in a sequential fashion. First, the mediational model constrained within-person parameters to be equal across the placebo and topiramate conditions. Second, the KNOWNCLASS command was used to relax this constraint. KNOWNCLASS specified the two groups of interest (0=MET-CBT + placebo and 1=MET-CBT + topiramate), allowing for comparison of the strength of the a, b1, and b2 paths between conditions. Since the KNOWNCLASS command specifies the use of mixture modeling with two known classes, traditional SEM model fit indices (e.g., CFI, RMSEA), were not available. Nested model tests were conducted using the likelihood ratio test to determine whether freeing parameters for a, b1, and b2 paths across conditions led to a significant increment in model fit. Of note, only within-person associations (see Figure 1) were freely estimated across medication conditions. Person-level effects could not vary across medication conditions since each individual was randomized to either topiramate or placebo.
Formal mediation was assessed using the MODEL CONSTRAINT command, which calculated the indirect effects for the placebo and topiramate conditions separately (Preacher et al., 2010). All models initially included sex, weekend status, and medication compliance in both medication conditions (riboflavin concentration in urine) as control variables. To reduce model complexity, only weekend status was retained in the final models as it related to cannabis craving, grams, and time with cannabis-using and non-using friends.
3. Results
Participants had an average age of 19.7 (SD=2.17) and 49.2% were female. The sample was majority White (61%) and Black/African American (28%), and 21% identified as Hispanic. As shown in Table 1, treatment groups did not significantly differ at baseline on past 90 day cannabis use days, grams consumed per week, and cannabis use disorder symptoms. Youth in the topiramate condition were older than youth in the placebo condition (t=−2.82(63), p=.006). No other significant differences were observed across conditions at baseline. Of the 65 participants, 39 (60%) completed the study. Youth in the topiramate condition were more likely to drop out of the study (χ2(1)=5.17, p=.02) and thus attended fewer therapy sessions (t=2.88(63), p=.005).
Table 1.
Baseline and Treatment Information by Medication Condition
| Placebo | Topiramate | |
|---|---|---|
|
n (%) or M (SD) |
n (%) or M (SD) |
|
| N | 26.00(40%) | 39.00(60%) |
| Baseline Characteristics | ||
| Age | 18.81(2.08)a | 20.28(2.05)a |
| Male | 12.00(46.15%) | 19.00(48.72) |
| Minority | 13.00(50%) | 18.00(46.15%) |
| Grams Per Week | 4.40(4.14) | 2.87(2.75) |
| Total Use Days | 63.27(26.23) | 61.49(26.11) |
| Cannabis Use Disorder Symptoms | 4.69(2.29) | 4.15(2.08) |
| Treatment Characteristics | ||
| Completed Study | 20(76.92%)b | 19(48.72%)b |
| Sessions Attended | 2.65(0.75)c | 2.00(1.07)c |
| Week 3 Medication Compliance | 22(84.62%) | 26(92.86%) |
| Week 6 Medication Compliance | 19(95%) | 19(100%) |
| Random Prompt EMA Compliance | 2933(86.19%)d | 4048(90.74%)d |
Note. Tabled values for categorical variables are count (%), whereas values for continuous variables are mean (standard deviation). Differences in categorical variables were calculated using chi-squared tests, whereas differences in continuous variables were calculated using independent sample t-tests. Significant column differences are denoted with superscript letters.
3.1. EMA Descriptive Information
Table 2 provides descriptive information for cannabis craving, use, time with cannabis-using friends, and time with non-using friends over the course of the RCT. On average, youth smoked 1.99 grams (SD=1.29) of cannabis on use days during treatment. Youth reported an average craving rating of 4.32 (SD=2.46) on use days and 3.47 (SD=2.96) on non-use days. On use days, youth spent 49.6% (SD=0.33) with cannabis-using friends and 8.4% (SD=0.17) with non-using friends. On non-use days, youth spent 29.2% (SD=0.35) with cannabis-using friends and 13.3% (SD=0.25) with non-using friends.
Table 2.
Descriptive Statistics for Craving, Cannabis Use, and Cannabis-Using and Non-Using Friends
| Cannabis Craving | Grams of Cannabis Smoked | Time with Cannabis-Using Friends | Time with Non-Using Friends | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Overall | Topiramate | Placebo | Overall | Topiramate | Placebo | Overall | Topiramate | Placebo | Overall | Topiramate | Placebo | |
| M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | M(SD) | |
| Baseline | 3.98(2.46) | 3.62(2.45) | 4.53(2.39) | 1.40(1.57) | 1.22(1.40) | 1.65(1.77) | 0.43(0.35) | 0.43(0.35) | 0.43(0.36) | 0.08(0.19) | 0.09(0.19) | 0.07(0.19) |
| Week 1 | 3.83(2.56) | 3.61(2.50) | 4.13(2.62) | 1.22(1.46) | 1.09(1.28) | 1.41(1.67) | 0.40(0.36) | 0.40(0.36) | 0.40(0.36) | 0.10(0.20) | 0.07(0.17) | 0.14(0.24) |
| Week 2 | 3.92(2.76) | 3.74(2.81) | 4.18(2.67) | 0.96(1.39) | 0.81(1.18) | 1.18(1.63) | 0.37(0.34) | 0.36(0.34) | 0.38(0.34) | 0.12(0.22) | 0.11(0.22) | 0.12(0.22) |
| Week 3 | 3.88(2.8) | 3.62(2.81) | 4.2(2.75) | 0.86(1.30) | 0.79(1.08) | 0.96(1.53) | 0.38(0.35) | 0.38(0.34) | 0.38(0.37) | 0.10(0.20) | 0.10(0.20) | 0.11(0.21) |
| Week 4 | 4.00(2.83) | 3.56(2.71) | 4.45(2.89) | 0.71(1.14) | 0.65(0.97) | 0.77(1.30) | 0.42(0.37) | 0.42(0.35) | 0.43(0.39) | 0.12(0.22) | 0.10(0.21) | 0.13(0.23) |
| Week 5 | 3.89(2.84) | 3.75(2.92) | 4.03(2.76) | 0.69(1.12) | 0.45(0.80) | 0.94(1.33) | 0.37(0.36) | 0.35(0.35) | 0.39(0.37) | 0.10(0.21) | 0.11(0.21) | 0.10(0.21) |
| Week 6 | 3.68(2.99) | 3.58(3.01) | 3.77(2.97) | 0.66(1.23) | 0.52(0.99) | 0.80(1.41) | 0.40(0.36) | 0.45(0.37) | 0.35(0.35) | 0.13(0.23) | 0.14(0.25) | 0.12(0.21) |
Note. Baseline refers to the one week pre-randomization EMA period. Overall refers to the overall sample.
3.2. Mediation Results
Intraclass correlation coefficients calculated from intercept-only models supported examining the proposed mediational model at the within- and between-person level of analysis (see Table 1). Given space constraints and evidence that topiramate blunts craving and the reinforcing effects of substance use (Authors, 2018; Miranda et al., 2016), models are presented where the mediators of interest, cannabis craving and cannabis use, occurred on use days. Supplemental Materials 1 reports the results when analyzing all treatment days.
3.2.1. Craving.
In the overall sample (i.e. regardless of medication condition), spending more time with cannabis-using friends than on average on a given day predicted greater cannabis craving the next day (Hypothesis 1a), and experiencing greater craving than on average predicted more time with cannabis-using friends the following day (Hypothesis 2a) (see Table 2). Contrary to Hypothesis 2c, day-level craving was not associated with time with non-using friends. At the person-level, greater time spent with cannabis-using friends was associated with higher craving. Higher craving was, in turn, associated with less time with cannabis-using friends and more time with non-using friends. Mediational analyses supported the day-level indirect effect from cannabis-using friends predicting cannabis craving the next day, which in turn, predicted time with cannabis-using friends on the following day (Hypothesis 3a). No person-level indirect effects reached statistical significance.
A nested model test indicated that the day-level associations found in the overall sample significantly differed between the topiramate and placebo conditions (χ2=11.51(5), p=.04). Socialization effects observed in the overall sample, i.e., day-level time with cannabis-using friends predicting higher cannabis craving the next day, remained significant across both conditions (Hypothesis 4a). In line with hypothesis 5a, selection effects, i.e., day-level higher craving than on average on use days predicting a greater amount of time with cannabis-using friends the following day, differed across conditions such that selection effects were statistically significant in the placebo + MET-CBT condition but not in the topiramate + MET-CBT condition.. Contrary to Hypothesis 5c, craving was not associated with time spent with non-using friends in either condition. At the person-level, more time spent with cannabis-using friends predicted higher cannabis craving. Cannabis craving was also associated with less time spent with non-using friends. Indirect effects indicated that craving significantly mediated the association between spending time with cannabis-using friends the prior day and time with cannabis-using friends the following day for youth in the placebo + MET-CBT condition but not in the topiramate + MET-CBT condition. Hypothesis 5c was not supported since craving was not associated with non-using friends in either condition.
3.2.2. Grams Smoked.
Consistent with Hypothesis 1b, spending more time with cannabis-using friends than on average predicted greater grams smoked the next day. In contrast to Hypotheses 2b and 2d, grams smoked did not predict time with cannabis-using or non-using friends the following day. At the person-level, more time spent with cannabis-using friends was associated with greater grams smoked overall. Grams smoked was also associated with less time spent with non-using friends overall. Hypotheses 3b and 3d were not supported since indirect effects were not supported at the day-level. At the person-level, cannabis use significantly mediated the relationship between person-level time with cannabis-using friends and time with non-using friends. Youth who spent more time with cannabis-using friends smoked greater amounts of cannabis, which in turn, was associated with less time with non-using friends.
In contrast to Hypotheses 4b, 5b, and 5d, day-level coefficients for the a, b1, and b2 paths did not significantly differ across the placebo and medication conditions (χ2=7.11(5), p=.21). Supplemental Material 2 presents coefficients for these paths separately for the topiramate and placebo conditions.
4. Discussion
Peer influence predicts problematic cannabis use and is a main focus of youth cannabis misuse intervention efforts (Phillips et al., 2018; Winters et al., 2018). Yet, how changing peer affiliations relates to successful reductions in cannabis use during treatment is only beginning to be empirically explored (Becker and Curry, 2014; Becker et al., 2019). The present analysis tested fine-grained, temporally sequenced hypotheses about how MET-CBT with adjunctive pharmacotherapy may disrupt day-to-day associations between being with cannabis-using friends and cannabis-use behaviors.
4.1. Key Findings
One of the strongest and most consistent findings was that time spent with cannabis-using friends predicted greater cannabis craving and use the next day (socialization effects). These associations were present when examined in aggregate across the study period and when pinpointing day-to-day associations. We were able to also demonstrate that cannabis craving, in turn, predicted greater time with cannabis-using friends the following day (selection effects). Contrary to hypotheses, however, cannabis use itself did not predict greater time with cannabis-using friends the following day, and neither cannabis craving nor use predicted greater time with non-using friends. This was likely a function of the minimal time youth spent with non-using friends during treatment.
An important strength of this secondary analysis was the use of temporally sequenced data and an analytic approach that allowed for a stringent test of mediation. We found support for craving as an indirect pathway that sustained affiliations with cannabis-using friends. This indirect mechanism combined the socialization (i.e., more time with cannabis-using friends promoted greater cannabis craving the following day) and selection effects (i.e., greater craving promoted more time spent with cannabis-using friends the following day). Multiple group models indicated that while socialization effects were supported for youth in both the placebo and topiramate conditions, selection effects were only supported for youth in the placebo condition. There was no evidence of selection effect for youth in the topiramate condition. The present findings suggest that topiramate could operate through altering cravings elicited by cannabis-using friends to reduce friend selection effects.
Shifting from affiliations with cannabis-using friends to non-using friends did not appear to be a treatment mechanism. This is inconsistent with well-supported theoretical models of behavior change (Best et al., 2016; Maisto et al., 1999) and adds to prior work that has found increasing time with non-using friends to be less related to substance use outcomes for youth relative to adults (Hoeppner et al., 2014; Kelly et al., 2014). In conjunction with these prior studies, the results point to the need for future work to elucidate how youth can increase their engagement and affiliations with non-using friends in well monitored, structured, low-risk environments.
4.2. Limitations
First, while the within-treatment tests of mediation were temporally sequenced, we caution readers against inferring causal treatment mechanisms from this initial finding. MSEM examined mediation within treatment conditions and did not directly compare across treatment groups. Second, topiramate was poorly tolerated leading to greater attrition in this condition. Given the relatively small sample sizes in each condition and topiramate poor tolerability, the results from the current study should be interpreted with caution and future work should examine the mediational pathways identified in the current study with better tolerated medications shown to reduce cannabis craving and use. Third, EMA data prior to youth reaching target dose was included in the analyses to meet the sufficient ratio of free parameters to observations for our multiple group MSEM. Including days where youth were not at target dose of topiramate may have attenuated medication condition effects on cannabis use and craving.
4.3. Conclusions.
The high proportion of cannabis-related treatment admissions among youth in the United States (SAMSHA, 2017), coupled with modest and short-lived benefit of cannabis treatments (Silvers et al., 2019), calls for understanding mechanisms of successful treatment outcomes. The present study explored how friend affiliations during treatment promoted or inhibited factors supportive of continued cannabis misuse. Overall, being with cannabis-using friends promoted greater cannabis use and craving, which, in turn, promoted continued selection of cannabis-using friends. Youth treated with adjunctive topiramate pharmacotherapy did not show selection effects, however, suggesting the changing peer affiliations could be one mechanism by which treatments can work.
Supplementary Material
Table 3.
Day- and Person-Level Variability and ICCs for Mediator and Outcome Variables
| Parameter | Day-Level Variance (SD) | Person-Level Variance (SD) | ICC |
|---|---|---|---|
| Grams Smoked | 0.46(0.02) | 0.17(0.04) | 0.27 |
| Cannabis Craving | 2.48(0.08) | 5.04(1.00) | 0.67 |
| Cannabis-using Friends | 0.07(0.002) | 0.06(0.01) | 0.48 |
| Non-Cannabis-using Friends | 0.04(0.001) | 0.01(0.002) | 0.23 |
Note. Bayesian estimation was used to derive posterior standard deviations. All within- and between-person variances were significant at p<.05.
Table 4.
MSEM tests of mediational pathways involving craving on use days.
| Model | Overall Sample (N=65) | Hypothesis | Placebo Condition (N=26) | Topiramate Condition (N=39) | Hypothesis | |||
|---|---|---|---|---|---|---|---|---|
| B(SE) | 95% CI | B(SE) | 95% CI | B(SE) | 95% CI | |||
| Day-Level Associations | ||||||||
| CU FriendsX→CravingX+1 | 6.02(0.59)*** | (4.87, 7.17) | 1a | 6.92(0.78)*** | (5.40, 8.44) | 5.25(0.78)*** | (3.72, 6.78) | 4a |
| CravingX+1→CU FriendsX+2 | 0.02(0.01)* | (0.003, 0.03) | 2a | 0.03(0.01)** | (0.01, 0.04) | 0.01(0.01) | (−0.004, 0.03) | 5a |
| CravingX+1→Non-CU FriendsX+2 | −0.002(0.01) | (−0.01, 0.01) | 2c | −0.004(0.01) | (−0.02, 0.01) | −0.001(0.01) | (−0.01, 0.01) | 5c |
| CU FriendsX→CU FriendsX+2 | 0.004(0.03) | (−0.06, 0.07) | −0.03(0.04) | (−0.10, 0.05) | 0.04(0.05) | (−0.05, 0.12) | ||
| CU FriendsX→Non-CU FriendsX+2 | −0.02(0.02) | (−0.06, 0.01) | −0.02(0.03) | (−0.07, 0.04) | −0.03(0.03) | (−0.08, 0.02) | ||
| Person-Level Associations | ||||||||
| CU Friends→Craving | 2.40(1.16)* | (0.13, 4.67) | 2.40(1.16)* | (0.13, 4.67) | 2.40(1.16)* | (0.13, 4.67) | ||
| Craving→CU Friends | −0.02(0.01)* | (−0.03, −0.003) | −0.02(0.01) | (−0.03, −0.002) | −0.02(0.01) | (−0.03, −0.002) | ||
| Craving→Non-CU Friends | 0.01(0.01)** | (0.004, 0.02) | 0.01(0.01)** | (0.003, 0.02) | 0.01(0.01)** | (0.003, 0.02) | ||
| CU Friends→CU Friends | 1.00(0.04)*** | (0.92, 1.09) | 0.99(0.04)*** | (0.90, 1.07) | 0.99(0.04)*** | (0.90, 1.07) | ||
| CU Friends→Non-CU Friends | −0.07(0.05) | (−0.16, 0.03) | −0.07(0.05) | (−0.15, 0.02) | −0.07(0.05) | (−0.15, 0.02) | ||
| Day-Level Indirect Effect | ||||||||
| CU FriendsX→CravingX+1→CU FriendsX+2 | 0.11(0.05)** | (0.01, 0.20) | 3a | 0.18(0.07)** | (0.05, 0.30) | 0.06(0.04) | (−0.02, 0.14) | |
| CU FriendsX→CravingX+1→Non-CU FriendsX+2 | −0.02(0.03) | (−0.07, 0.04) | 3c | −0.03(0.05) | (−0.12, 0.07) | −0.004(0.03) | (−0.06, 0.05) | |
| Person-Level Indirect Effect | ||||||||
| CU Friends→Craving→CU Friends | −0.04(0.03) | (−0.10, 0.01) | −0.04(0.03) | (−0.10, 0.01) | −0.04(0.03) | (−0.10, 0.01) | ||
| CU Friends→Craving→Non-CU Friends | 0.03(0.02) | (−0.01, 0.08) | 0.03(0.02) | (−0.01, 0.07) | 0.03(0.02) | (−0.01, 0.07) | ||
Note. CU=cannabis-using, X=day x, X+1=day x+1, X=day x+2.
= p<.001
=p<.01
p<.05.
Table 5.
MSEM results for mediational pathways for total grams of cannabis reported on use days.
| Model | Overall Sample (N=65) | Hypothesis | |
|---|---|---|---|
| B(SE) | 95% CI | ||
| Day-Level Associations | |||
| CU FriendsX→GramsX+1 | 2.64(0.16)*** | (2.32, 2.96) | 1b |
| GramsX+1→CU FriendsX+2 | 0.004(0.01) | (−0.01, 0.02) | 2b |
| GramsX+1→Non-CU FriendsX+2 | −0.003(0.01) | (−0.01, 0.01) | 2d |
| CU FriendsX→CU FriendsX+2 | 0.01(0.04) | (−0.06, 0.08) | |
| CU FriendsX→Non-CU FriendsX+2 | −0.02(0.02) | (−0.05, 0.02) | |
| Person-Level Associations | |||
| CU Friends→Grams | 1.25(0.27)*** | (0.71, 1.78) | |
| Grams→CU Friends | −0.01(0.01) | (−0.04, 0.01) | |
| Grams→Non-CU Friends | −0.02(0.01)* | (−0.04, −0.003) | |
| CU Friends→CU Friends | 1.00(0.04)*** | (0.92, 1.09) | |
| CU Friends→Non-CU Friends | −0.08(0.04) | (−0.16, 0.002) | |
| Day-Level Indirect Effect | |||
| CU FriendsX→GramsX+1→CU FriendsX+2 | 0.01(0.02) | (−0.03, 0.05) | 3b |
| CU FriendsX→GramsX+1→Non-CU FriendsX+2 | −0.01(0.01) | (−0.03, 0.02) | 3d |
| Person-Level Indirect Effect | |||
| CU Friends→Grams→CU Friends | −0.01(0.02) | (−0.05, 0.02) | |
| CU Friends→Grams→Non-CU Friends | −0.03(0.01)* | (−0.05, −0.002) | |
Note. CU=cannabis-using, X=day x, X+1=day x+1, X=day x+2.
= p<.001
=p<.01
p<.05.
Highlights.
Time spent with cannabis-using friends promoted greater cannabis use and craving
Youth treated with topiramate pharmacotherapy did not show selection effects
Across conditions, craving and use were unrelated to time with non-using friends
Changing peer affiliations could be one mechanism by which treatments can work.
Role of Funding Source
The National Institutes of Health supported this research (AA028414, AA024808, DA026778, AA026326).
Footnotes
Declaration of Competing Interest
The authors have no conflicts of interest to declare.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributor Information
Samuel N. Meisel, Center for Alcohol and Addiction Studies, Brown University, Providence, RI, 02912, USA.; E. P. Bradley Hospital, Riverside, RI 02915.
Hayley Treloar Padovano, Center for Alcohol and Addiction Studies, Brown University, Providence, RI, 02912, USA..
Robert Miranda, Jr., Center for Alcohol and Addiction Studies, Brown University, Providence, RI, 02912, USA.; E. P. Bradley Hospital, Riverside, RI 02915.
References
- Albsoul-Younes AM, Salem HA, Ajlouni SF, Al-Safi SA, 2004. Topiramate slow dose titration: improved efficacy and tolerability. Pediatr. Neurol 31(5), 349–352. [DOI] [PubMed] [Google Scholar]
- Authors, 2017. [Google Scholar]
- Authors, 2018. [Google Scholar]
- Becker SJ, Curry JF, 2014. Testing the effects of peer socialization versus selection on alcohol and marijuana use among treated adolescents. Subst. Use Misuse 49(3), 234–242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Becker SJ, Marceau K, Hernandez L, Spirito A, 2019. Is it Selection or Socialization? Disentangling Peer Influences on Heavy Drinking and Marijuana Use Among Adolescents Whose Parents Received Brief Interventions. Subst. Abuse 13, 1178221819852644. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Best D, Beckwith M, Haslam C, Alexander Haslam S, Jetten J, Mawson E, Lubman DI, 2016. Overcoming alcohol and other drug addiction as a process of social identity transition: The social identity model of recovery (SIMOR). Addict. Re. Theory 24(2), 111–123. [Google Scholar]
- Brechwald WA, Prinstein MJ, 2011. Beyond Homophily: A Decade of Advances in Understanding Peer Influence Processes. J. Res. Adolesc 21(1), 166–179. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Buckner JD, Zvolensky MJ, Crosby RD, Wonderlich SA, Ecker AH, Richter A, 2015. Antecedents and consequences of cannabis use among racially diverse cannabis users: an analysis from ecological momentary assessment. Drug Alcohol Depend. 147, 20 25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Caouette JD, Ewing SWF, 2017. Four mechanistic models of peer influence on adolescent cannabis use. Curr. Addict. Rep 4(2), 90–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Caouette JD, Feldstein Ewing SW, 2017. Four Mechanistic Models of Peer Influence on Adolescent Cannabis Use. Curr. Addict. Rep 4(2), 90–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carpenter RW, Squeglia LM, Emery NN, McClure EA, Gray KM, Miranda R Jr, Tomko RL, 2020. Making pharmacotherapy trials for substance use disorder more efficient: Leveraging real-world data capture to maximize power and expedite the medication development pipeline. Drug Alcohol Depend. 209, 107897–107897. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chi FW, Kaskutas LA, Sterling S, Campbell CI, Weisner C, 2009. Twelve-Step affiliation and 3-year substance use outcomes among adolescents: social support and religious service attendance as potential mediators. Addiction 104(6), 927–939. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chung T, Sealy L, Abraham M, Ruglovsky C, Schall J, Maisto SA, 2015. Personal network characteristics of youth in substance use treatment: motivation for and perceived difficulty of positive network change. Subst. Abus 36(3), 380–388. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Dishion TJ, Andrews DW, Crosby L, 1995. Antisocial boys and their friends in early adolescence: Relationship characteristics, quality, and interactional process. Child Dev. 66(1), 139–151. [DOI] [PubMed] [Google Scholar]
- Gray JC, Treloar Padovano H, Wemm SE, Miranda R Jr., 2018. Predictors of topiramate tolerability in heavy cannabis-using adolescents and young adults: A secondary analysis of a randomized, double-blind, placebo-controlled trial. J Clin Psychopharmacol 38(2), 134–137. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hallgren KA, Wilson AD, Witkiewitz K, 2018. Advancing analytic approaches to address key questions in mechanisms of behavior change research. J. Stud. Alcohol Drugs 79(2), 182–189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hirschi T, 1969. A control theory of delinquency. Criminology theory: Selected classic readings 1969, 289–305. [Google Scholar]
- Hoeppner BB, Hoeppner SS, Kelly JF, 2014. Do young people benefit from AA as much, and in the same ways, as adult aged 30+? A moderated multiple mediation analysis. Drug Alcohol Depend. 143, 181–188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kelly JF, Bergman BG, Fallah-Sohy N, 2018. Mechanisms of behavior change in 12-step approaches to recovery in young adults. Curr. Addict. Rep 5(2), 134–145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kelly JF, Hoeppner B, Stout RL, Pagano M, 2012. Determining the relative importance of the mechanisms of behavior change within Alcoholics Anonymous: a multiple mediator analysis. Addiction 107(2), 289–299. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kelly JF, Stout RL, Greene MC, Slaymaker V, 2014. Young adults, social networks, and addiction recovery: post treatment changes in social ties and their role as a mediator of 12-step participation. PLoS One 9(6), e100121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Litt MD, Kadden RM, Tennen H, Kabela-Cormier E, 2016. Network Support II: Randomized controlled trial of Network Support treatment and cognitive behavioral therapy for alcohol use disorder. Drug Alcohol Depend. 165, 203–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Maisto SA, Carey KB, Bradizza CM, 1999. Social Learning Theory. Psychological Theories of Drinking and Alcoholism 2, 106–163. [Google Scholar]
- McMillan C, Felmlee D, Osgood DW, 2018. Peer influence, friend selection, and gender: How network processes shape adolescent smoking, drinking, and delinquency. Soc Netw 55, 86–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Miranda R Jr., MacKillop J, Monti PM, Rohsenow DJ, Tidey J, Gwaltney C, Swift R, Ray L, McGeary J, 2008. Effects of topiramate on urge to drink and the subjective effects of alcohol: a preliminary laboratory study. Alcohol Clin. Exp. Res 32(3), 489 497. [DOI] [PubMed] [Google Scholar]
- Miranda R Jr., MacKillop J, Treloar H, Blanchard A, Tidey JW, Swift RM, Chun T, Rohsenow DJ, Monti PM, 2016. Biobehavioral mechanisms of topiramate's effects on alcohol use: an investigation pairing laboratory and ecological momentary assessments. Addict. Biol 21(1), 171–182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Miranda R Jr., Treloar H, 2016. Emerging Pharmacologic Treatments for Adolescent Substance Use: Challenges and New Directions. Curr. Addict. Rep 3(2), 145–156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Muthen LK, Muthen BO, 1998-2018. Mplus user's guide, in: Muthen M (Ed.). Los Angeles, CA. [Google Scholar]
- Oncken C, Arias AJ, Feinn R, Litt M, Covault J, Sofuoglu M, Kranzler HR, 2014. Topiramate for smoking cessation: a randomized, placebo-controlled pilot study. Nicotine Tob. Res 16(3), 288–296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Osgood DW, Feinberg ME, Ragan DT, 2015. Social networks and the diffusion of adolescent problem behavior: Reliable estimates of selection and influence from sixth through ninth grades. Prev. Sci 16(6), 832–843. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Phillips KT, Phillips MM, Lalonde TL, Prince MA, 2018. Does social context matter? An ecological momentary assessment study of marijuana use among college students. Addict.Behav 83, 154–159. [DOI] [PubMed] [Google Scholar]
- Potenza MN, Sofuoglu M, Carroll KM, Rounsaville BJ, 2011. Neuroscience of behavioral and pharmacological treatments for addictions. Neuron 69(4), 695–712. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Preacher KJ, Zyphur MJ, Zhang Z, 2010. A general multilevel SEM framework for assessing multilevel mediation. Psychol. Methods 15(3), 209. [DOI] [PubMed] [Google Scholar]
- SAMSHA, 2017. Key substance use and mental health indicators in the United States: Results from the 2016 National Survey on Drug Use and Health in: Statistics CFBH, and Quality SAAMH, Administration S (Eds.). Rockville, MD [Google Scholar]
- Scalco MD, Trucco EM, Coffman DL, Colder CR, 2015. Selection and Socialization Effects in Early Adolescent Alcohol Use: A Propensity Score Analysis. J. Abnorm. Child Psychol 43(6), 1131–1143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Silvers JA, Squeglia LM, Rømer Thomsen K, Hudson KA, Feldstein Ewing SW, 2019. Hunting for what works: Adolescents in addiction treatment. Alcohol Clin. Exp. Res 43(4), 578–592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Simeone TA, Wilcox KS, White HS, 2006. Subunit selectivity of topiramate modulation of heteromeric GABAA receptors. Neuropharmacology 50(7), 845–857. [DOI] [PubMed] [Google Scholar]
- Sutherland EH, Cressey DR, Luckenbill D, 1995. The theory of differential association. Deviance: A symbolic interactionist approach, 64–68. [Google Scholar]
- Tabachnick BG, Fidell LS, Ullman JB, 2007. Using multivariate statistics. Pearson Boston, MA. [Google Scholar]
- Winters KC, Botzet AM, Stinchfield R, Gonzales-Castaneda R, Finch AJ, Piehler TF, Ausherbauer K, Chalmers K, Hemze A, 2018. Adolescent substance abuse treatment: A review of evidence-based research, Adolescent Substance Abuse. Springer, pp. 141 171. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.