Demographic and substance use characteristics were similar across conditions as shown in . Thirty-six youth were enrolled in EXP and 33 in CONTROL. Fifty-two families had two-parent participation and 17 had mother-only participation. Overall, 65% of youth tested positive for THC at intake. On average, adolescents reported using marijuana 1.8 (SD=1.4) times per day on 13.3 (SD=10.3) days during the month prior to intake. Thirty-one (45%) adolescents met DSM-IV criteria for Marijuana Abuse only and 30 (43%) for Marijuana Dependence. Eight (12%) adolescents were diagnostic orphans. They did not endorse sufficient symptoms to meet criteria for Marijuana Abuse or Dependence, but met all other inclusion criteria including a THC positive urine drug test and/or reported use in the 30 days prior to intake. Those not meeting DSM criteria were less likely to have a THC positive intake urine drug test but the difference was not significant (No DSM: 50% THC positive vs. DSM Abuse or Dependence: 67%, X2(1)=.34, p=.92). Those not meeting DSM criteria reported significantly fewer days of use in the 30 days prior to intake (No DSM: Mean=4.4 days, SD=5.8 vs. DSM Abuse or Dependence: Mean=14.6 days, SD=10.1, F=10.2, p<.01). Fifteen (22%) also met criteria for Alcohol Abuse, and participants reported drinking an average of 3.4 (SD=4.0) drinks per day on 1.5 (SD=2.1) days in the month prior to intake. One met criteria for Opiate Abuse and one for Sedative Abuse. Forty (58%) adolescents were regular tobacco users.
Demographic and Substance Use Characteristics
As shown in rates of psychopathology were high in both conditions with rates of ODD or CD, ADHD, and Depression or GAD ranging from 39%–61% based on parent reports and from 14%–36% based on youth reports. The pattern of more psychopathology reported by parents than youth was also observed on the CBCL and YSR. Generally, EXP and CONTROL youth had similar mean CBCL and YSR internalizing, externalizing, total problems, and parenting scores at intake (all p’s >.05).
3.2. Participation and retention
Teens and mothers attended 11.3 (SD=4.0), and 10.7 (SD=3.9) of 14 sessions on average, respectively, with no between-condition differences. Fathers who attended one or more sessions (N=51) attended 9.1 (SD=4.5) sessions on average. Retention was good, as measured by attendance during the last treatment week: 77% for both conditions. Adolescents in both conditions provided a similar number of urine specimens (EXP: 23.4 (SD=8.1) vs. CONTROL: 22.6 (SD=7.4)). Participation rates for follow-ups also did not differ between EXP and CONTROL: 92% vs. 91%, 75% vs. 85%, 75% vs. 85%, and 78% vs. 79% for the discharge and 3-, 6-, and 9-month assessments, respectively. No adverse events were observed.
3.3. During-treatment abstinence
Primary marijuana use outcome measures indicated that EXP enhanced continuous abstinence outcomes. EXP youth had more mean weeks of documented continuous marijuana abstinence during treatment than CONTROL youth (EXP: 7.6 (SD=5.6) vs. CONTROL: 5.1 (SD=4.5), t=−2.1; p=.04, d=.48, medium effect). Those in the EXP condition were also more likely to achieve ≥8 weeks of continuous abstinence (53% vs. 30%, X2(1)=3.6, p=.06) and ≥10 weeks of continuous abstinence (50% vs. 18%, X2(1)=7.7, p=.006), while rates of briefer periods of abstinence were similar across the two treatment conditions (≥4 weeks: EXP 61%, CONTROL 55%; ≥6 weeks: EXP 56%, CONTROL 46%).
3.4. Post-treatment abstinence
Generalized Estimating Equation (GEE) analyses comparing the point prevalence of marijuana use at intake, discharge and at 3, 6, and 9 months indicated that marijuana abstinence based on urine toxicology testing followed a cubic pattern. Marijuana use decreased during treatment (linear time effect = −1.22, p<.01), but then increased during follow-up (quadratic time effect = 0.41, p<.01) and began to level off again (cubic time effect = −0.04, p<.01). There were no significant treatment or time × treatment interaction effects, although EXP showed higher rates of abstinence at each timepoint (see ). Power was .80 to detect effect sizes of d=.40 for treatment condition and .29 for the treatment × time interaction. The observed effect sizes were .21 for treatment and <.10 for treatment × time. Results were similar whether missing urine drug tests were coded positive or included as missing values in the GEE analysis.
Figure 2 GEE model of marijuana positive urine drug tests from intake to 9 months post treatment. Circles and squares represent observed percentages for the CONTROL and EXP conditions, respectively. The estimated cubic curves for each treatment condition are displayed. (more ...)
3.5 Self-reported marijuana use outcomes
The mixed-model repeated measures analysis of secondary marijuana use outcome measure (self-reports of marijuana use) did not show significant treatment or time × treatment interaction effects (see ). For these analyses, power was .80 to detect effect sizes of d=.37 for treatment condition and .26 for the treatment × time interaction. The observed effect sizes were .24 for treatment and <.10 for treatment × time. The linear, quadratic, and cubic effects of time were all significant (p<.01). Marijuana use decreased through treatment (linear time effect = −131.16), began to increase after treatment (quadratic time effect = 39.68), but stabilized at a level lower than pre-treatment (cubic time effect = −3.69). Similar results and power estimates were obtained for self-reports of alcohol use, which included a significant quadratic time effect (.80, p<.01), but no treatment or time × treatment interactions. Percent of days used alcohol declined from intake to the 3 month follow up, but increased from 6 to 9 months.
Figure 3 Mixed model of reported percentage of days used marijuana in the 90 days prior to intake and between each subsequent assessment. Circles and squares represent observed percentages for the CONTROL and EXP conditions, respectively. The estimated cubic curves (more ...)
3.6 Parenting and Teen Psychopathology
Significant time effects emerged for all 3 APQ parenting scales, with no significant time × treatment condition interactions. All d’s were <.10 for the interactions, with power=.80 to detect interactions ranging from d=.12 to .29 across scales. Positive Involvement showed significant linear improvement over time (B=2.8, p<.05), with mothers reporting significantly more positive involvement with their teens (B=4.7, p<.01). Deficient Monitoring showed significant quadratic change (B=.53, p<.01), with scores improving from intake to 3 months, and worsening from 3 to 9 months post treatment. Negative Discipline showed significant cubic changes (B=−.41, p<.01), with scores improving from pre- to post-treatment, increasing slightly from 3 to 6 months post-treatment, but declining again from 6 to 9 months post treatment. In addition, for Negative Discipline, there was a main effect of treatment condition (B=1.2, p<.05, d=.25), with significantly worse scores for the CONTROL condition than the EXP condition.
Significant quadratic time effects emerged for both psychopathology scales (internalizing B=.75, p<.01; externalizing B=.67, p<.01), with scores improving (decreasing) from intake to 3 months post-treatment, and rising slightly from 3 to 9 months post treatment. There were no significant time × treatment condition interactions. Both d’s were <.10 for the interactions, with power=.80 to detect interactions of d=.10. However, for Externalizing, there was a main effect of treatment condition (B=2.4, p<.05, d=.30), with significantly higher scores for the CONTROL condition than the EXP condition. Both scales also showed significant informant effects, with teens reporting significantly fewer problems than mothers on internalizing (B=−7.0, p<.01) and externalizing (B=−4.5, p<.01), and fathers reporting significantly fewer problems for teens than mothers on internalizing (B=−2.8, p<.05).
Another potential indicator of parenting change was participation in the optional weekly urine drug tests in the 12 weeks following the end of counseling. EXP families were expected to show greater attendance because they had been taught contingency contracting based on test results. A large percentage of both treatment conditions attended one or more of these appointments (70% of CONTROL, 78% of EXP). EXP teens attended more times (EXP: 5.3 vs. CONTROL: 3.9), but the difference was not significant. The number of tests attended across conditions was significantly related to marijuana abstinence at discharge (e.g., negative drug test) at discharge (t(67)=3.2, p<.01) and at 3 months (t(67)=2.2, p<.05), but not at 6 or 9 months. These results suggest that parents of youth who responded well to treatment continued with testing, and continued testing may have helped maintain abstinence while it was available.
3.7 Predictors of Post Treatment Abstinence
We tested parenting, externalizing problems, and treatment condition as predictors of abstinence during the follow-up period (THC urine drug test results at 3, 6, and 9 months) in three exploratory structural models. Each model included: mean maternal and paternal CBCL externalizing ratings at intake and discharge, treatment condition, and a latent post treatment abstinence intercept construct with follow up urine test results (3,6,9 months) as indicators. Mean maternal and paternal Positive Involvement, Negative Discipline, and Poor Monitoring were each tested in a separate model. Models were tested in Mplus. The strongest results were found among parental negative discipline, adolescent externalizing problems, and post treatment substance use. As shows, negative discipline and adolescent externalizing were significantly correlated with each other at intake and discharge (Intake r=.40, Discharge r=.42) and showed significant stability from intake to discharge (Negative Discipline B=.72, Externalizing B=.76). Accounting for relations between intake and end of treatment allows us to interpret the end of treatment variables as reflecting during-treatment change. Consistent with the analyses presented above, treatment condition did not independently predict parenting or externalizing improvement, or marijuana use in the post-treatment period. However, changes in negative discipline (B=.49) were associated with marijuana abstinence in the follow-up period, with the model accounting for a total of 29% of variance in post treatment marijuana use. The identical pattern of results was found for Poor Monitoring, but Positive Involvement did not predict post treatment marijuana use. The final models fit well (RMSEAs=.00). These results suggest that inconsistent discipline and poor monitoring played an important role in marijuana use post treatment.
Figure 4 Structural model predicting a post treatment marijuana use construct, defined by marijuana (THC) positive urine drug test results at 3, 6, and 9 month follow ups. Factor loadings of the THC result variables were all constrained to 1. EXP=Experimental (more ...)