Individualized Assessment and Treatment Program (IATP) for Cannabis Use Disorder: Randomized Controlled Trial With and Without Contingency Management

Abstract

This study tested a new approach to the treatment of cannabis use disorder (CUD). CUD is difficult to treat, and achieving abstinence is particularly difficult. The Individualized Assessment and Treatment Program (IATP) was intended to address this problem by providing a highly individualized approach to the training of coping skills most relevant for each individual. To do this, an experience sampling procedure was employed prior to treatment to record patients’ marijuana use behavior and associated thoughts, feelings, coping behaviors and situations. This information was used by therapists to plan treatment that would address the specific strengths and weaknesses of each patient in drug-use situations. The present study tested IATP against a conventional combined motivational enhancement-cognitive-behavioral treatment (MET-CBT), with or without the addition of contingency management (CM) for abstinence. The patients were 198 men and women randomly assigned to one of four 9-session treatment conditions: MET-CBT; MET-CBT-CM; IATP; or IATP-CM. Patients were assessed out to 14 months. Planned contrasts indicated that the IATP conditions yielded greater levels of abstinence than the MET-CBT conditions. The addition of contingency management did not bolster the performance of IATP, but did do so for MET-CBT. As expected, IATP did lead to greater use of coping skills than the MET-CBT conditions. However, coping skills use was not a significant predictor of outcome when other variables were in the same analyses. Self-efficacy was a robust predictor, and mediator, of outcome. We suggest that the highly individualized IATP may act by enhancing self-efficacy.

Despite the gradual legalization of cannabis, marijuana continues to be the most commonly used illicit drug in the United States, with an estimated 22.2 million persons reporting having used it in any month (as of the 2015 National Survey on Drug Use and Health)(SAMHSA, 2015). Along with a recent sharp increase in daily use among adults (Mauro et al., 2018), the U.S. has seen an increase in the prevalence of cannabis use disorders (CUD), with an overall rate of about 2.9% of the population. It has been estimated that about 30% of those using cannabis develop a CUD (Hasin et al., 2015).

Despite its widespread use, treatment programs directed specifically to marijuana are relatively uncommon. When treatment is provided, abstinence is difficult to achieve, particularly in the long term. In the largest controlled trial of treatment for marijuana dependence to date, the multi-site Marijuana Treatment Project (MTP; The Marijuana Treatment Project Research Group, 2004), the highest abstinence rate achieved was 23% of participants at the 4-month follow-up in the Motivational Enhancement therapy + Cognitive-Behavioral treatment (MET+CBT) condition, declining to 15% at 9 months. Budney and colleagues (2006) combined contingency management with cognitive-behavioral skills training: 37% reported abstinence at a 12-month follow-up. In a comparable study by Kadden et al. (2007), a combination of MET+CBT+contingency management yielded 14-month abstinence rates of 35%.

In another study Budney et al. (2015) compared a computer-assisted treatment to treatment delivered by a therapist. The abstinence rate at the end of treatment was 46.7% in the Computer condition, and 44.8% in the Therapist condition. These rates dropped to 20% or below by the end of 9-month follow-up. In summary, despite rather intense treatments that include CBT and contingent reinforcement for abstinence, achieving and maintaining abstinence from marijuana has been difficult.

CBT presumes that those who relapse lack the skills to deal with the cues that trigger drug use. The aim of CBT is to provide the skills necessary to cope with stressors and high-risk situations in more adaptive ways (Marlatt & George, 1984; Marlatt & Gordon, 1980). To the extent that such skills are developed and practiced, success experiences should lead to increased self-efficacy for coping, that in turn will lead to greater use and persistence of coping behaviors, and greater abstinence over time (e.g., Larimer, Palmer, & Marlatt, 1999). This model of treatment has not always been supported, however.

Although improved coping appears to be related to better outcomes in some studies (e.g., Kiluk, Nich, Babuscio, & Carroll, 2010), it is not clear that coping skills per se are responsible for those outcomes. It is possible, for example, that both the use of coping skills and improved outcomes are the result of other processes. Nor is it clear that CB treatments result in increased use of coping skills as intended. To make a case for the role of coping in outcome, it would be necessary to show that a treatment intended to increase coping skills in fact did so, and that increased coping skills use was associated with abstinence or reduction of substance use (Longabaugh & Morgenstern, 1999). A few studies have shown this (e.g., Witkiewitz, Roos, Tofighi, & Van Horn, 2018, in alcohol treatment), but most have not. A study by Litt and Kadden (2015) indicated that coping skills did mediate outcome in treatment for cannabis dependence, but that the mediation occurred through effects on self-efficacy.

Carroll (1996) suggested that CBT or relapse prevention-based approaches are able to yield long-term improvements in substance use because they create the conditions by which patients may develop coping strategies even after treatment is finished. However, the coping skills that are developed long after treatment ends may be quite different from those trained during treatment. Few studies have measured use of coping skills at extended follow-up points, so the delayed adoption of new coping skills has not been substantiated. The true mechanisms of CBT are thus still unclear.

In contrast to the emphasis on skills in CBT, contingency management (CM) procedures treat abstinence behavior simply as an operant that is responsive to reinforcement. The addition of CM procedures to CBT for cannabis dependence treatment has proven effective at keeping patients in treatment, and more effective than CBT alone in promoting abstinence at treatment end (e.g., Budney, Higgins, Radonovich, & Novy, 2000). Some evidence suggests that the addition of CM to CBT may also increase abstinence in the long-term (Budney et al., 2006).

The short-term efficacy of contingency management procedures appears to be predicted by two processes resulting from reinforcement of abstinence behavior: increased retention in treatment, and enhanced periods of abstinence during treatment [see Petry (2000) and Petry and Simcic (2002) for reviews]. There is also evidence that continuous abstinence during treatment is one of the best predictors of longer term outcomes (Carroll, 1996; Higgins, Alessi, & Dantona, 2002; Higgins, Badger, & Budney, 2000; Sigmon, Steingard, Badger, Anthony, & Higgins, 2000). It is not clear, however, whether contingency management procedures promote long-term changes in outcomes, and if so, what changes occur in drug users to make this possible (e.g., changes in coping behaviors). In many studies (though not all, e.g., Budney et al., 2006; Petry, Alessi, Marx, Austin, & Tardif, 2005), CM procedures have tended to produce short-term gains that dissipated once the external reinforcement contingencies were withdrawn (cf. Litt, Kadden, & Petry, 2013).

Budney et al. (2006) suggested that early abstinence can be enabled by contingency management procedures, and maintained by skills training. Our own data indicate that the best predictors of long-term abstinence are increases in self-efficacy and use of coping skills, the latter measured by total scores on a 48-item Coping Strategies Scale (CSS; Litt, Kadden, Cooney, & Kabela, 2003). But an analysis of the CSS in over 900 marijuana users in treatment revealed that the coping strategies patients reported using to stay abstinent were not at all like the skills taught in conventional coping skills-based treatments (Litt, Kadden, & Tennen, 2012). Rather than reporting using skills such as avoidance of drug users, drug refusal, or waiting out their urge to smoke, many successful patients reported using cognitive strategies that (1) were consistent with a cognitive shift toward change (e.g., “made a commitment to myself to not smoke”), and (2) bolster self-efficacy (e.g., “remind myself about my accomplishments”). The implication is that the apparent ineffectiveness of many coping skills-based programs may be because they are not teaching skills that patients actually use in the long term.

The Individualized Assessment and Treatment Program (IATP) for cannabis use disorder was intended to provide patients with the coping skills they actually need. IATP employs experience sampling (ES) procedures to provide detailed assessments, multiple times per day, of patients’ pretreatment cannabis use and associated cognitions, affects, social situations, and coping behaviors. The resulting data are delivered to therapists prior to treatment and used to tailor skills training specific to each patient’s strengths and weaknesses. By personalizing, and presumably optimizing, coping skills training, it was hoped that this study would provide an opportunity to determine just how important coping skills are to outcome, particularly in comparison to standard CBT, with or without contingency management.

In the present study, men and women diagnosed with cannabis dependence were recruited from the community and randomly assigned to one of four 9-session treatment conditions in a 2×2 design that compared IATP versus standard MET-CBT, and CM versus No CM. The conditions were: MET-CBT; MET + CBT + Contingency Management (MET-CB-CM); IATP; or IATP-CM. Patients were followed at posttreatment (month 2), and at months 5, 8, 11, and 14. We expected that IATP-CM would yield the best long-term outcomes, and that MET-CB would show the poorest. We also expected that use of coping skills would be a significant predictor of outcomes over time.

Method

All procedures described here were approved by the Institutional Review Board of UConn Health.

Participants

The participants were 198 men and women recruited form the greater Hartford, CT, area from July 2013 through October 2016 using newspaper and radio advertisements. To be eligible, individuals had to be at least 18 years old, meet DSM-IV criteria for Cannabis Dependence (assessed by the SCID for DSM-IV, described below), and be willing to accept random assignment to any of the four treatment conditions. Exclusion criteria were acute medical/psychiatric problems that required inpatient treatment (e.g., acute psychosis, severe depression, suicide/homicide risk), reading ability below the fifth grade level, and lack of reliable transportation, or excessive commuting distance. Participants could meet criteria for dependence on other substances, but must have reported that marijuana was their primary substance of abuse.

Of the 336 persons who responded to advertisements and were screened, 316 were eligible according to the criteria described. Many (117) lost interest in treatment after screening but prior to randomization, and one more dropped out for other reasons before being randomized. The remaining 198 participants were randomly assigned to the four treatment conditions: MET-CBT (n=49); MET + CBT + Contingency Management (MET-CBT-CM) (n=51); IATP (n=48); or IATP-CM (n=50).

Based on effect sizes found in our earlier studies (Kadden et al., 2007; Litt et al., 2013), the present study was powered to detect differences in abstinence rates over time between the IATP conditions and MET-CB conditions. Sample data for a time-averaged binary data power analysis for a Generalized Estimating Equations (GEE) repeated logistic model were estimated. The power analysis indicated that combined group sample sizes of 160 (i.e., 80 each for the IATP and MET-CB conditions) would achieve 80% power to detect an odds ratio of 1.28 (a moderate effect size) for a Condition X Time interaction in a design with 5 repeated measurements having an AR(1) covariance structure, when the initial proportion from the IATP groups is 0.2, the correlation between observations on the same subject is 0.30, and the alpha level is 0.05. The sample size obtained was thus sufficient for our main hypotheses. A diagram showing the flow of participants through the study recruitment and treatment stages is shown in Figure 1.

Figure 1.

Flow of patients in the IATP project for treatment of cannabis dependence.

Participants were 58% male, with a mean age of 36 years (SD = 12.0), and were 51% White, 28% Black, 14% Hispanic, and 7% other. Their mean years of schooling was 13.7 (SD = 5.8), 60% were employed at least part time outside the home, and 20% were living with a spouse or partner. Participants reported, on average, using cannabis on 82 of the 90 days prior to intake, and consuming over 2 grams of marijuana per day. Table 1 shows the distribution of baseline characteristics by treatment condition, along with results of between-treatment tests (F-test or chi-square). No significant between-treatment differences were seen on any baseline variable.

Table 1.

Patient Characteristics by Treatment Condition.

	Treatment Condition
	MET-CBT (n=49)		MET-CB-CM (n=51)		IATP (n=48)		IATP-CM (n=50)		Total (N=198)		F or χ2
	M or %	SD	M or %	SD	M or %	SD	M or %	SD	M or %	SD
Age (years)	33.7	11.9	37.6	11.8	35.3	10.9	35.6	13.1	35.6	12.0	0.91
Sex (% Female)	41		43		48		38		42		1.06
Race/Ethnicity											6.17
% Non-Hispanic White	59		51		46		46		51
% African American	18		25		37		32		28
% Hispanic	14		14		13		16		14
% Other	9		10		4		6		7
Education (years)	13.3	3.1	13.8	2.0	13.2	1.9	14.3	10.9	13.7	5.8	0.42
Income ≥ $10,000/yr (%)	72		82		73		72		75
Married/Cohabiting (%)	18		24		23		16		20		1.22
Employed Full/Part Time (%)	55		59		64		64		60		1.13
Lifetime Alcohol/Drug Treatments	1.0	1.6	0.7	1.5	0.8	1.8	1.2	2.6	0.9	1.9	0.68
Days Used Cannabis in last 90	79.9	17.6	82.6	12.8	82.7	10.3	81.9	13.4	81.8	13.7	0.45
Cannabis Grams/Day	1.66	1.63	2.55	2.94	1.77	1.28	2.24	2.82	2.06	2.32	1.60

Measures and Instruments

The Structured Clinical Interview for DSM-IV Axis I Disorders, Patient edition, version 2.0 (SCID-I/P; First, Spitzer, Gibbon, & Williams, 1996), was used to determine whether subjects met inclusion/exclusion criteria for marijuana dependence, other drug dependence, and psychotic symptoms in the 90 days prior to intake. The diagnosis of marijuana dependence corresponds to DSM-V diagnoses of moderate-severe cannabis use disorder.

Marijuana, alcohol and other drug use for the 90 days prior to intake and at each follow-up were assessed using a version of the Time Line Follow-Back (TLFB; Sobell & Sobell, 1992) modified for collection of cannabis use data. The TLFB approach uses a calendar to prompt participants to recall the amount of substance use for every day of a given period. For this study marijuana use was estimated in grams per day. The TLFB has good test-retest reliability, and high validity for verifiable events in studying cannabis use (Litt et al., 2013; The Marijuana Treatment Project Research Group, 2004), with an overall agreement rate with biological verification measures of over 87% (Hjorthoj, Hjorthoj, & Nordentoft, 2012). Urine tests (OnTrak TesTcup®; Varian, Lake Forest, CA) were performed at intake, prior to each treatment session, and at the in-person follow-ups to verify non-use/use of cannabis [Δ−9-Tetrahydrocannabinol (THC) ≥ 50 ng/ml] and other drugs. If a discrepancy arose between self-reported use and the urine testing, the patient was further queried and the TLFB was corrected if necessary.

Three cannabis use dependent variables were calculated based on the daily self-reported TLFB data for each month. The probability of Abstinence at each month was based on a patient reporting no use of cannabis in that month. Proportion Days Abstinent (PDA) was calculated as the proportion of days each month that patients reported being abstinent. The Longest Duration of Abstinence (LDA) was calculated as the longest period of continuous abstinence, in days, recorded since the first day of treatment. The LDA variable was verified by negative urine samples taken at the nearest treatment session (during treatment) or in-person follow-up point; a positive urine finding would be recorded as cannabis use in a given month, ending the abstinence period. Due to the infrequency of urine testing in the posttreatment period the other cannabis use variables were not biochemically verified.

The Marijuana Problems Scale (MPS; Stephens, Wertz, & Roffman, 1993) is a self-report instrument that assesses marijuana-related problems in several domains (family, social, legal, employment, physical health, memory/cognitive), as well as common complaints of heavy users (e.g., procrastination, feeling bad about using). The internal reliability of the total score exceeded alpha=.89 at all assessments.

Treatment process variables were assessed at pretreatment, posttreatment, and at all follow-ups. The 48-item Coping Strategies Scale (CSS; Litt et al., 2003; Litt, Kadden, Stephens, & Marijuana Treatment Project Research Group, 2005) taps strategies used by patients in treatment to remain abstinent. Respondents rate, on a scale ranging from 1 (never) to 4 (frequently), the frequency of using specific strategies in the past 3 months. In this study the internal reliability of the total CSS exceeded alpha=.90 across all administrations.

Self-efficacy was assessed using a marijuana self-efficacy scale (MSE; Litt et al., 2005; Stephens, Wertz, & Roffman, 1995). The MSE scale assesses participants’ confidence (on a 7-point scale) in their ability to not use marijuana in different situations. In this study internal reliability ranged from alpha = .89 to .97 over all the assessments.

Procedures

Patients were assessed at several time points over 14 months. These inlcuded the initial screening evaluation, conducted by telephone, a pretreatment diagnostic evaluation, and follow-up assessments at posttreatment (month 2), and at months 5, 8, 11, and 14. The assessments at months 2, 8, and 14 were conducted in-person. Assessments at months 5 and 11 were conducted by telephone.

Initial Screening Evaluation, Diagnostic Evaluation, Informed Consent, and Solicitation of a Locator.

Prospective subjects were evaluated through an initial telephone screening procedure using the criteria described above. Callers were excluded (and referred elsewhere for appropriate treatment), or scheduled for an intake interview (Diagnostic Evaluation; DE) with a research assistant. At the DE prospective participants were administered the SCID-IV (see above), and urine specimens were collected and analyzed via instant screening for gross presence v. absence of controlled substances at intake. Those who met eligibility criteria and consented to participate were asked to give the name of a locator (usually a spouse or close friend) if available.

Assignment to treatment.

Those who agreed to participate were randomly assigned to treatment using an urn randomization (Stout, Wirtz, Carbonari, & Del Boca, 1994) computer program that balanced the four conditions on gender, age, and baseline drug use (marijuana PDA in the 90 days pre-intake). The Project Coordinator then informed participants of their treatment assignment and scheduled their first treatment appointment, approximately 2 weeks following their DE, and following their initial experience sampling period. Patients were not asked to stop using marijuana at this time (though some initial decline in use was expected).

Experience sampling for assessing in-vivo marijuana use variables.

An Experience Sampling (ES) procedure was used in all treatment conditions to assess use of coping skills while avoiding many of the problems inherent in retrospective recording. Although all patients engaged in ES data collection, the results from those records were only transmitted to the therapists when conducting the IATP treatment. Experience sampling was conducted using Interactive Voice Response (IVR) technology, in which assessments were made when the IVR system called participants on cellular telephones. Patients were free to use their own phones, and phones were distributed to those who did not have one. Patients were trained in the use of cell phones for answering the automated telephone questionnaires, and were asked to carry the phone at all times in Weeks 1 and 2 (after baseline interviews, but before treatment start) and again during Week 6.

During the ES self-monitoring periods of the study, subjects were engaged in signal-contingent recording only; that is, they only responded when the system called them. (Event-contingent recording, in which a respondent initiates a call in response to some event like drug use, tends not to be reliable, and defeats the purpose of experience sampling, which is to assess momentary cognitions, affects and behaviors without allowing the participant the time to calculate or reflect upon his/her response). The IVR system that administered the momentary assessment questions by voice, and recorded responses in the form of usable data, was developed by Telesage, Inc., Chapel Hill, NC.

During each ES period, the IVR system called subjects’ cell phones on a quasi-random basis 4–6 times per day, with 1–2 randomly scheduled prompts in each of four 210-minute time periods from 8:00AM to 10:00PM. This frequency of recording enabled us to capture many moments in a patient’s day without being disruptive. To promote protocol compliance during ES, subjects were paid $1 per call completed with a $4 bonus if all prompts were completed in a day, and a $25 bonus if at least 80% of total prompts were completed in the week. The possible total incentive was thus $60 per week during each ES period. At pretreatment the rate of responding ranged from 0% (in one person) to 100%; the average response rate was 76%, with no significant differences by treatment condition. The week 6 response rate averaged 67%.

Data collected during ES included items related to urges to use marijuana, situations, cognitions, affects, coping actions, and actual drug use (marijuana in grams) in the prior hour. Coping responses were assessed by asking the subject what he/she did to try to keep from using marijuana in the previous hour. Patients responded by voice. The free response was coded by research assistants as problem-focused v. emotion-focused, and cognitive v. behavioral, as well as by particular strategy (e.g., seeking support). Interrater reliability for coding responses was in the range of kappa=.70 to .80.

Treatments

The treatments were conducted on an outpatient basis in 60–70 min sessions, and were guided by detailed manuals that provided specific guidelines to therapists. Therapists were MA-level counselors with experience in CBT. The same therapists were used to conduct all treatments to minimize therapist effects. Therapists received extensive training in both MET-CB and IATP. Treatment was provided free of charge. Patients had 12 weeks to complete the 9-session programs.

MET-CBT.

MET-CB in this study was based on cognitive-behavioral principles and designed to remediate deficits in skills for coping with interpersonal and intrapersonal (e.g., anger, craving) antecedents to marijuana use. The treatment consisted of one session of motivational enhancement therapy (MET), followed by eight sessions of coping skills training. The motivational enhancement component was based on the MET program developed by Miller and colleagues (Miller & Rollnick, 2002). The MET session focused on a Personalized Feedback Report derived from the intake assessment, emphasizing identified problem areas, with the goal of moving patients towards accepting the need for change and identifying specific changes to make. Coping skills training started in the second session. This training, used in our previous treatment studies (Kadden et al., 2007; Litt et al., 2013), included brief didactic presentations, behavioral rehearsal, and homework exercises. Skills were trained according to a menu that included functional analysis of problems and identification of trigger situations, coping with cravings, marijuana refusal skills, and problem-solving, based on predetermined modules.

IATP.

Prior to the beginning of treatment, the 2 weeks of ES responses were collated by a Research Assistant and provided to the therapists as a functional analysis chart (see example; Figure 2) for use in the IATP conditions. Data regarding situations associated with drug use, affects, cognitions, and coping actions were used by the therapist and patient together to devise specific adaptive coping responses to specific high-risk situations. The same procedure was used at Week 6 during treatment, to modify and adapt treatment. In IATP, an analysis of situations and of the preceding thoughts, feelings, and behaviors, was used to identify the circumstances that pose the most risk, as evidenced by the ES records. Judgments regarding level of risk were based on the extent of the patient’s drug use in similar situations, proximity of the situation to actual use, strength of the accompanying emotions, and the content quality of the accompanying cognitions. Based on the combination of these factors, therapists initially focused on those situations that posed the highest risk for marijuana use. This entailed identifying ES records in which the most drug use or drug cravings occurred, and then looking for commonalities among those records. The therapist attended to the changing needs of the patients as indicated by the during-treatment ES records (i.e., Week 6 ES compared to pre-treatment), and adjusted the treatment plan with the patient as needed.

Figure 2.

Example of Functional Analysis Chart for IATP. Chart summarizes information regarding when drug use or drug urges occurred, plus antecedent thoughts, moods, and situations.

Contingency management (CM).

Two of the treatment conditions included CM for marijuana abstinence. A fishbowl-drawing procedure (Petry & Martin, 2002) was used to deliver reinforcement. Participants in all conditions provided a urine specimen at each treatment session. Patients in CM conditions with a cannabis-negative urine earned the opportunity to draw from the fishbowl. (In the No-CM Conditions, the therapist discussed the urine test results with the patient, and changes that might be made if the test was positive.) Because of the long period (20 or more days) over which marijuana may be detected in the urine of chronic users (Hawks & Chiang, 1986; Verebey, Gold, & Mule, 1986), drawings for prizes in the first two weeks were based on attendance alone; drawings for negative urines began in Week 3 (although urine samples were taken in Weeks 1 & 2 to establish the pattern of testing at each session). In week 3, patients started with 3 fishbowl draws. They earned an extra draw for each consecutive week their urine was cannabis-free. If a urine test was positive, they decreased to one draw for the next negative specimen, but could return to their prior best level of draws after 2 consecutive weeks of negative tests. The average total earned per subject was $63 (SD=$92) over 9 weeks of treatment, with a minimum of $0.00 and a maximum of $432.

Treatment fidelity.

Treatment fidelity was operationalized as therapists’ adherence to treatment protocols. Session outlines were developed for each session in both the MET-CB and IATP conditions. These outlines listed all necessary aspects of a session, and provided a step-by-step accounting of each element that should be completed, and in what order. Therapists were trained to follow the outline, and to check off each step in the session as it occurred. This method ensured that all necessary aspects of a session were completed, and minimized therapist drift. Session outlines were reviewed on a weekly basis by the Project Coordinator (PC). In addition, each treatment session was audio taped. The PC and one external rater listened to 25% of session tapes and evaluated each therapist’s adherence to session outlines (percentage of session outline material completed by the therapist). The rater (a trained research associate) also monitored sessions to assure that elements unique to one treatment were not employed in the other treatment. The rater and the PC had high interrater agreement for fidelity (ICC=.85). Therapist adherence rates all exceeded 90%.

Patient adherence to treatment.

Participants in all treatments attended a mean of 5.7 (SD=3.5; range 0–9) sessions, with no differences between conditions. In each treatment condition therapists recorded completion of assigned activities. Patients in all conditions completed an average of 62% (range 0%−90%) of assignments, with no differences between conditions.

Data Analysis

The primary substance use dependent variables were probability of total abstinence each month and Percent Days Abstinent (PDA) from marijuana each month. MPS score was the primary psychosocial outcome variable. The secondary outcomes included Longest Duration of Abstinence (LDA in days) during treatment, and the process variables CSS score and MSE score. All continuous variables were normally distributed. (Although PDA is a proportion, it was acceptably distributed overall, with a skewness of 0.25, and was treated as a continuous variable in analyses).

An intent-to-treat strategy was adopted for all analyses. All analyses were conducted using SAS software version 9.4 (SAS Institute, 2014). PDA and MPS score over months were analyzed using a linear mixed model approach adopting type 3 tests of fixed effects (SAS Proc MIXED). Outcomes were analyzed as a function of Treatment condition, repeated Time (in months), and Treatment X Time. In these analyses Treatment and Time were treated as fixed effects, and the intercept was treated as a random effect. A variance-components covariance structure was adopted (rather than an autoregressive structure) on the basis of accepted fit criteria (Judge, Griffiths, Hill, Lutkepohl, & Lee, 1985). The mixed modeling procedure employs maximum likelihood estimation to calculate parameter estimates, thus taking advantage of all data collected without imputing missing data.

A Type 3 generalized estimating equations (GEE; SAS Proc GENMOD) analysis was used to evaluate probability of patients reporting abstinence over all time points. GEE uses the “all available pairs” method, in which all non-missing pairs of data are used in estimating the working correlation parameters. Thus we lose only those individual observations that a given subject has missing, rather than losing the entire subject (Diggle, Liang, & Zeger, 1994). As in the analysis of PDA, the basic model analyzed outcome as a function of Treatment Condition, Time (in months), and Treatment X Time. Three planned contrasts were tested in these models: IATP v MET-CBT (Conditions 1&2 v. 3&4); CM v. No CM (Conditions 1&3 v. 2&4); and a linear contrast such that: MET-CBT < MET-CBT-CM < IATP < IATP-CM.

The secondary outcome Longest Duration of Abstinence during treatment (LDA) was evaluated using one-way ANOVA. The treatment process variables Coping Scale Score and Marijuana Self-Efficacy Score were analyzed using SAS Proc MIXED as above.

Exploratory analyses were also conducted. A GEE analysis was conducted in which multiple predictors of abstinence outcome over time, measured at the five follow-up periods, were evaluated in the same model. These predictors included age, gender, and race (White v. Non-White), as well as Treatment Condition, Time (by follow-up period), treatment attendance, LDA, CSS total score at each time period, and MSE score at each time period. (Abstinence was evaluated at the follow-up periods, rather than monthly as in the main analyses, because the treatment process measures used in this model were only assessed at those time periods).

Results

Treatment Effects on Primary Outcomes

Figure 3 shows the proportion of patients continuously abstinent during each moth over the course of the study (panel A), the mean values for PDA (panel B), and the Marijuana Problem Scale scores (panel C) by treatment condition over time.

Figure 3.

Main outcome measures over time by treatment condition. MET+CBT+CM = motivational enhancement plus cognitive-behavioral treatment plus contingency management. IATP-CM=Individualized Assessment and Treatment Program plus contingency management.

Probability of abstinence.

The results of the GEE analysis on monthly abstinence over time are shown in Table 2. As indicated in the table, significant main effects were seen for treatment condition, and for time. As indicated by the Time effect, and reflected in the abstinence rates in Figure 3 (panel A), likelihood of abstinence tended to increase over time. The Treatment X Time interaction was not significant. In examining the planned contrasts, the comparison of IATP v. MET-CBT was significant. Analysis of mean estimates indicated that the IATP conditions were superior to the MET-CBT conditions in promoting abstinence. Likewise, the linear contrast was also significant. The means of the estimates indicated, however, that the most successful treatment was IATP alone, rather than IATP-CM. At months 13 and 14 abstinence rates among the IATP conditions exceed 43%. The contrast of CM v. No-CM was not significant; the addition of contingency management to treatment did not predict greater probability of abstinence.

Table 2.

Results of Mixed Model Analyses of PDA and MPS, and of GEE Analysis of Monthly Abstinence. Shown are Results for Effects and Planned Contrasts.

Dependent Variable	Effect Tested (F Values)			Treatment Contrasts (df=1, 2752)
	Treatment (df=3/2752)	Time (df=16/2752)	Treatment X Time (df=48/2752)	IATP v MET-CB		CM v No-CM		Lineara
				B	95% CI	B	95% CI	B	95% CI
PDA	3.23*	68.08	1.82***	0.16*	.016–.308	−0.09	−0.24 – 0.05	0.42*	.09 – .74
MPS	1.51	61.96***	0.94	−0.65	−4.12–2.83	2.17	−1.31–5.65	−3.46	−11.26–4.33
Dependent Variable	Effect Tested (χ2)			Treatmient Contrasts
	Treatment (df=3)	Time (df=13)	Treatment X Time (df=39)	IATP v MET-CB		CM v No-CM		Lineara
				B	95% CI	B	95% CI	B	95% CI
Monthly Abstinence	11.24*	32.30**	28.48	1.33*	0.31–2.34	−0.76	−1.77–0.26	3.41*	1.05–5.77

Note: B=unstandardized estimate; CI=Confidence Interval; PDA=Proportion Days Abstinent; MPS=Marijuana Problem Scale score; df=degrees of freedom

^aLinear contrast: MET-CBT < MET-CBT-CM < IATP < IATP-CM

^*p < .05;

^**p < .01;

^***p <.001

PDA.

Similar results were seen for PDA. Significant effects were seen for Treatment condition and for the Treatment X Time interaction (see Table 2). Examination of Figure 3 (panel B) indicates that PDA increased over time, and that the IATP condition yielded the most abstinent days. The significant contrast of IATP v. MET-CBT suggests that IATP yielded better outcomes than MET-CBT. The linear contrast was also significant, but again, the IATP condition without CM led to the best outcomes. The contrast of CM v. No-CM was not significant.

MPS.

Marijuana Problem Scale scores appeared to be insensitive to treatment differences. MPS scores decreased significantly from pre- to posttreatment, and remained low throughout the 14 months of follow-up.

Analyses of Secondary Outcomes, Process Measures

LDA.

One of the presumed mechanisms of change for contingency management procedures is increased abstinence during treatment (Petry, 2000; Petry & Simcic, 2002). In treatment studies the longest duration of abstinence (LDA) during treatment has been one of the best predictors of long-term outcomes (Carroll et al., 2006; Higgins et al., 2002). In the current study the mean LDA ranged from 14 days (SD=21.3) in the MET-CBT condition to 26 days (SD=27.2) in the MET-CBT-CM condition. Analysis of variance indicated a significant treatment effect on LDA (F (3, 1021) = 10.35; p < .001). Consistent with expectations, Treatment contrasts indicated that the CM conditions yielded significantly more continuous days of abstinence during treatment than did the No CM conditions (t (1021) = 4.46; p < .001).

Coping and self-efficacy.

Table 3 shows the results of linear mixed model analyses on the CSS scores and the MSE scores. As indicated, there was a main effect for Time on CSS scores, such that coping scores increased from pre- to posttreatment, with significant contrasts, such that IATP conditions yielded higher coping scale scores than did the MET-CBT conditions (see Figure 4, panel A). The contrast of CM v. No-CM conditions was not significant. Analysis of MSE scores indicated a main effect for Treatment, and for Time, reflecting a significant increase in self-efficacy from pre- to posttreatment. The Treatment X Time interaction was not significant, but evaluation of means indicated that the IATP conditions elicited higher self-efficacy ratings than did the MET-CBT conditions. This was borne out by the planned contrast analyses.

Table 3.

Results of Mixed Model Analyses of CSS and Marijuana Self-Efficacy Over Time. Shown are Results for Effects and Planned Contrasts.

Dependent Variable	Effect Tested (F Values)			Treatment Contrasts (df=1,2752)
	Treatment (df=3/194)	Time (df=5/777)	Treatment X Time (df=15/777)	IATP v MET-CB		CM v No-CM		Lineara
				B	95% CI	B	95% CI	B	95% CI
CSS	2.44†	21.38***	0.60	0.17*	0.01 – .33	−0.14	−0.30 – 0.02	0.48**	0.13 – 0.84
MSE	5.51**	43.58***	0.87	2.83**	3.88–21.82	−7.72†	−16.69 −1.25	33.42**	13.25 – 53.60

Note: CSS=Coping Strategies scale; MSE=Marijuana Self-Efficacy; B=unstandardized estimate; CI=Confidence Interval; PDA=Proportion Days Abstinent; MPS=Marijuana Problem Score; df=degrees of freedom

^aLinear contrast: MET-CB < MET-CB-CM < IATP < IATP-CM

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001

Figure 4.

Process measures. Coping Strategies Scale scores and Marijuana Self-Efficacy scale scores over time by Treatment Condition. MET+CBT+CM = motivational enhancement plus cognitive-behavioral treatment plus contingency management. IATP-CM=Individualized Assessment and Treatment Program plus contingency management.

Multiple Variable Prediction of Abstinence Over Time

The results of the multiple-predictor analysis of abstinence over time are shown in Table 4. As seen in the table, both gender and race predicted outcome. Women were more likely to report abstinence over time than men, and white persons were more likely to be abstinent than were non-white people. We also see that when treatment process and demographic variables are entered into the analysis, the effect of Treatment differences is no longer significant. The Time effect is significant, reflecting the fact that abstinence increases over time. LDA emerges as a significant predictor in this model, as does marijuana self-efficacy over time. (The effect of MSE over time indicated that abstinence at any given time period was predicted by MSE measured during the same time period.) Although CSS score, a key variable in IATP, predicted abstinence when examined alone (B=0.82; SE=.23; OR=2.27; p < .001), it drops in significance when other variables are in the model.

Table 4.

Multiple Variable Prediction of Abstinence over Time (Follow-Up Periods). Results from GEE Model.

Predictor	B	SE	z	OR	Lower 95% CI	Upper 95% CI
Age	0.01	0.01	1.26	1.01	0.99	1.03
Gender (1 Male; 2 Female)	0.62	0.24	2.53*	1.86	1.39	2.33
Race (Non-White v White)	0.61	0.24	2.50*	1.84	1.37	2.31
Treatment
MET-CBT	−0.42	0.35	−1.21	0.66	−0.03	1.34
MET-CBT-ContM	−0.35	0.33	−1.05	0.70	0.06	1.35
IATP	0.18	0.3	0.60	1.20	0.61	1.79
IATP-ContM	--	--	--
Time	0.12	0.03	4.45***	1.13	1.07	1.19
Attendance	−0.02	0.05	−0.30	0.98	0.88	1.08
LDA	0.04	0.01	7.26***	1.04	1.02	1.06
CSS Score	0.20	0.16	1.20	1.22	0.91	1.54
MSE Score	0.04	0.01	8.27***	1.04	1.02	1.06

Note: LDA=Longest Duration of Abstinence during treatment; CSS=Coping Strategies Scale over time; MSE=Marijuana Self-Efficacy over time; -- = No estimate: reference group; B= logistic coefficient; SE=Standard Error; Z=Z-score for logistic coefficient; OR=Odds ratio; CI=Confidence Interval

^*p < .05;

^***p < .001

Mediation of PDA Outcome

Given the importance of self-efficacy in predicting outcome in this study, a multilevel test of mediation was conducted to determine if during-treatment change in self-efficacy mediated the effects of IATP on PDA outcome over time. The analysis was conducted using a procedure based on product of coefficients described by Krull and Mackinnon (2001), in which the IATP v MET-CBT treatment contrast was used to predict PDA over time, and the change in MSE from pre- to posttreatment was the mediator. The results indicated that, relative to MET-CBT, IATP conditions significantly predicted change in MSE, and MSE was a significant predictor of PDA over time. The mediation analysis indicated that change in self-efficacy was a significant mediator of the effects of IATP on PDA outcome over time (z=2.14; p =.016).

Discussion

The IATP treatment was intended to address a possible reason for the frequent failure to find strong evidence for the use of coping skills in achieving good outcomes in treatment for CUD. Although the coping skills model of treatment is compelling, its success is not well supported. As indicated above, to the extent that patients report use of coping skills to manage drug-use situations, the skills endorsed bear little resemblance to those taught in a standard MET-CBT program (Litt et al., 2012). Thus it appeared that those who fared well in the long-term were those who “stumbled onto” some coping strategies that worked well for them. If a treatment could help all patients find those effective individual strategies in the first place, the chances of success should be improved. Based on its conceptualization and its results in this study, IATP might be considered as a means of identifying and developing such strategies.

Another possible important component of treatment was also tested: contingency management (CM). Due in large part to the pioneering work of Nancy Petry and colleagues (e.g., Petry, 2000; Petry, 2012; Petry & Martin, 2002), contingency management has increasingly been incorporated into treatment for substance use. The efficacy of CM procedures has been well documented, particularly for producing significant reductions in substance use in the short term (see Davis et al., 2016; Sayegh, Huey, Zara, & Jhaveri, 2017 for reviews). It was hoped that the addition of CM to IATP would optimize treatment. The CM component would promote early abstinence, allowing the patient to better focus on treatment, and to build on that early success. In the present study CM did prompt greater levels of abstinence in the months immediately following treatment. In the long-term, however, the CM conditions were not as successful as IATP without CM.

The results of this study indicate that IATP was a successful treatment strategy. Patients in the IATP condition reported over 60% days abstinent at 14 months post-intake. And 44% of patients in IATP reported complete abstinence at 14 months. These are excellent rates of abstinence, particularly over such a long-term.

It is notable, also, that the improvements in abstinence, at least in the IATP condition, appeared to increase over time. This effect was not due to the loss of non-performers at the longer-term follow-ups: those lost did not differ from those retained on any demographic or baseline substance use variable. Though this general increase in abstinence over time was seen to some extent in all conditions, it was most pronounced in the IATP condition. This supports the idea that for many patients treatment is only the beginning of a change in substance use, and that patients may build on the changes learned in treatment and carry those out afterwards.

The reasons for the success of the IATP condition in this study, however, are not quite clear. It was expected that IATP would encourage patients to use a variety of coping skills, both those discussed in treatment, but also those that they discover on their own. Thus their net CSS score should have increased. This did occur, from pre- to posttreatment, but CSS scores did not continue to increase over time as PDA increased. Moreover, the same pattern of CSS scores were seen in the MET-CBT conditions. Finally, when evaluated in the context of other measures, CSS score failed to emerge as a significant predictor of outcome.

One variable that predicted outcome over time, and which may have reflected influence of IATP, was marijuana self-efficacy. MSE increased from pre- to posttreatment and, among the IATP patients, tended to stay high (over 90%) out to 14 months. As indicated above, the effects of IATP were mediated in part by pre-post increase in self-efficacy.

It is not clear why CSS scores were not better predictors of outcome in this study. It is possible that both treatments operated through coping, but that coping itself operates through self-efficacy as we saw in an earlier study (Litt & Kadden, 2015). Once again, self-efficacy emerged as one of the most robust predictors of substance use treatment outcome (Kadden & Litt, 2011). It is possible that the high level of individualization in the IATP approach engendered greater confidence over time, which translated into better outcomes a year later.

An additional possibility is that coping actions per se become less salient, and thus less reportable, over time, as abstinence becomes easier and more automatic. That is, perhaps people who have been long abstinent do not need to exert so much conscious effort to stay abstinent. In that case, however, we would expect a decline in scores over time, but this decline was not found here; scores stayed at their posttreatment highs throughout follow-up.

Another finding was that the addition of CM to IATP did help patients in the months just after treatment, but did not enhance treatment effects in the long-term. Indeed, the best results were seen for those patients in the IATP No-CM condition. The results are reminiscent of those seen for the addition of CM to Network Support treatment for alcohol use disorder patients (Litt, Kadden, Kabela-Cormier, & Petry, 2009). In that study the best results were seen for the Network Support No-CM condition, and it appeared that the addition of CM (for homework completion in that study) was counterproductive. On the other hand, the addition of CM to MET-CBT was largely successful here. It is possible that the IATP treatment may lead to a somewhat closer patient-provider relationship due to the detailed examination of patient behavior involved. In this case it may be that the addition of CM could be perceived as disrupting that relationship. This is purely speculative, however. In future studies we will evaluate what effects, if any, the IATP treatment has on the working alliance between patients and therapists.

One limitation of the present study was that we relied upon self-report for our measures of cannabis use at extended follow-ups. As we indicate, however, the TLFB has demonstrated extremely high validity for cannabis use. And our rate of discrepancies between reported abstinence at follow-ups and urine test results is estimated as less than 5%. We thus feel confident about the accuracy of our data. An additional issue is that, in the CM conditions, it was possible for patients starting week 3 of treatment to report abstinence, but still test positive on urinalysis, thus delaying their receipt of prizes and potentially putting them at a disadvantage. This occurred in 24 cases. The performance of these persons (on monthly PDA) was slightly better than, though not statistically different from, those who were not “unfairly” delayed in receiving prizes.

The present study did show that a highly personalized treatment, based on actual patient behavior measured in-vivo and with (we trust) a reduced degree of recall error and retrospective bias, is effective in treating persons with CUD. We believe that this approach will allow therapists a better understanding of the determinants of their patients’ behavior, and their ability to cope. The procedure as used here, however, is highly dependent on research assistants to aggregate ES records for therapist use. As the technology for the kind of recording conducted here improves, and coding of responses becomes more automated, it will be possible for approaches like IATP to be more widely adopted and tested with new substance using populations.