Cognitive Behavioral Interventions for Alcohol and Drug Use Disorders: Through the Stage Model and Back Again

Abstract

Cognitive behavioral therapy (CBT) approaches have among the highest level of empirical support for the treatment of drug and alcohol use disorders. As Psychology of Addictive Behaviors marks its 30^th anniversary, we review the evolution of CBT for the addictions through the lens of the Stage Model of Behavioral Therapies Development. The large evidence base from Stage II randomized clinical trials indicates a modest effect size with evidence of relatively durable effects, but limited diffusion in clinical practice, as is the case for most empirically validated approaches for mental health and addictive disorders. Technology may provide a means for CBT interventions to circumvent the ‘implementation cliff’ in Stages 3–5 by offering a flexible, low-cost, standardized means of disseminating CBT in a range of novel settings and populations. Moreover, returning to Stage 1 to reconnect clinical applications of CBT to recent developments in cognitive science and neuroscience holds great promise for accelerating understanding of mechanisms of action. It is critical that CBT not be considered as a static intervention, but rather one that constantly evolves and is refined through the Stage model until the field achieves a maximally powerful intervention that addresses core features of the addictions.

It should come as no surprise that cognitive behavioral approaches figured prominently in the first issue of Psychology of Addictive Behaviors in 1987. The first article in Volume 1, Issue 1 by Jerome Platt and David Metzger asked “…why is it that addicts seem to be unable to successfully cope with the problems they encounter in daily living” and described a “…set of cognitive skills in which addicts appeared to be deficient” (Platt & Metzger, 1987). The second article in that issue reported on a long term (6–10 year) follow-up of individuals after cognitive behavioral treatment for obesity and described factors associated with weight gain versus maintenance of weight loss over time (Jordan, Canavan, & Steer, 1987). Together, these articles anticipated important continuing themes in research on cognitive behavioral interventions: first, how best to convey cognitive and behavioral skills to help individuals successfully modify addictive behaviors, and second, how to reduce the risk of relapse and make such changes durable. Similar themes were also raised in the landmark 1985 publication of Relapse Prevention: Maintenance Strategies in the Treatment of Addictive Behavior (Marlatt & Gordon, 1985), the essential blueprint for cognitive-behavioral approaches for addictive behaviors.

Since then, cognitive behavioral approaches have been among the most-studied treatment approaches for addictive behaviors, with much of that literature published in Psychology of Addictive Behaviors. Currently, cognitive behavioral approaches have among the highest level of empirical support from well-controlled trials and are widely acknowledged as evidence-based approaches (U.S. Department of Health and Human Services, 2016) and included in a wide range of practice guidelines (American Society of Addiction Medicine, 2015; Center for Substance Abuse Treatment, 2004; National Institute on Drug Abuse, 2007; Veterans Administration, 2015).

As the history of cognitive behavioral approaches for addictive behaviors has been well covered elsewhere (Annis & Davis, 1989; Rotgers, Keller, & Morgenstern, 1996), this review will use the Stage Model of Behavioral Therapies Development (Onken, Carroll, Shoham, Cuthbert, & Riddle, 2013; Rounsaville, Carroll, & Onken, 2001) as an organizing principle to summarize the current evidence base for cognitive behavioral treatments for drug and alcohol use disorders (Stage 2 studies). We will then focus on Stage 3 and 4 issues, reviewing the status of dissemination of CBT in clinical practice, highlighting challenges to dissemination and the promise of technology-based approaches to address the ‘implementation cliff’ (Weisz, Ng, & Bearman, 2014). Finally, we speculate how CBT may evolve during the next 30 years, if informed by developments in technology, cognitive science and neuroscience. It should be noted that this review will for the most part concentrate on “classical CBT” rather than its many variants, including behavioral couples training, adaptations for specific comorbid conditions (such as mood management interventions), or combined approaches (such as the Community Reinforcement Approach).

Summary of the Evidence Base, Stage 2 Studies

Multiple meta analyses and reviews over the past 30 years have concluded that CBT is an effective treatment across a range of substance use disorders (Carroll & Onken, 2005; Dutra et al., 2008; Irvin, Bowers, Dunn, & Wong, 1999; Magill & Ray, 2009). The most recent and comprehensive meta-analysis included 53 controlled trials, published through 2006, of CBT for adults diagnosed with alcohol or drug use disorders, and reported a small but statistically significant treatment effect (g = 0.15) for CBT over control conditions across studies (Magill & Ray, 2009). Statistical transformations to a ‘success percentage’ indicated 58% of patients receiving CBT fared better than those in the comparison condition (Magill & Ray, 2009). Effect size varied according to the type of comparison condition, with a large effect size found for CBT in comparison to no treatment (random g = .80, p < .005), and small-sized effects in comparison to a passive (random g = .15, p < .005) or active treatment (random g = .13, p < .05). Also, larger effect sizes were found in studies of CBT combined with additional psychosocial treatment (random g = .31, p < .005) than for studies of CBT combined with pharmacotherapy (random g = .20, p < .05), or CBT alone (random g = .17, p < .05).

Additionally, there were indications of differences according to primary drug of abuse. Specifically, the largest treatment effects have been found in studies of marijuana use (moderate effect size, g = .51), whereas smaller effect sizes have been found across studies of alcohol, cocaine, stimulants, opiates, and polydrug use (ranging from g = .08 to g = .13). However, at the time of the meta-analysis, there were far fewer studies evaluating CBT for marijuana use compared to alcohol or other drugs, and many of these studies combined CBT with other psychosocial treatments to enhance effects (Magill & Ray, 2009). Clearly, there is room for improvement in enhancing the effectiveness of CBT for substance use disorders.

The evidence supporting CBT has been generated from single-site studies, as well as from some of the landmark multisite studies of addiction treatment, including Project MATCH and Project COMBINE for alcohol (Anton et al., 2006; Project MATCH Research Group, 1997), the NIDA Cooperative Cocaine Treatment Study (Crits-Christoph et al., 1999), and the Marijuana Treatment Project (MTP Research Group, 2004). One of the distinguishing features of CBT has been its relative durability of effects, with significant treatment effects persisting through a follow-up period, in some cases with individuals showing greater improvement after treatment ends (i.e., ‘sleeper effect’) (e.g., Carroll et al., 2000; Carroll et al., 1994b; Rawson et al., 2002). It has also shown to be effective in combination with pharmacotherapies for substance use (e.g., Carroll et al., 2004; Schmitz, Stotts, Rhoades, & Grabowski, 2001), and has been a widely-used platform for pharmacotherapy trials, in other words, the ‘base’ treatment provided to all participants to enhance treatment retention and medication adherence, and to address other ancillary problems (Carroll, Rounsaville, & Kosten, Carroll, 1997; 2004).

CBT has been combined with other empirically supported treatments for alcohol and drug use disorders, such as Motivational Interviewing (MI) and Contingency Management (CM), as a strategy to bolster early treatment engagement and adherence. Several studies have investigated the combination of CBT and MI for various drugs of abuse, including amphetamines (Baker et al., 2005), cocaine (McKee et al., 2007; Rohsenow et al., 2004), methamphetamines (Bux & Irwin, 2006), and marijuana (Babor, 2004; Dennis et al., 2004). Although the findings have been mixed with respect to additive effects on drug use outcomes, there is some evidence to suggest that adding motivational enhancement to the early stages of CBT can be effective at increasing motivation and improving retention in treatment. Also, given that CM has strong immediate effects on substance use that tend to weaken after the contingencies are terminated (Prendergast, Podus, Finney, Greenwell, & Roll, 2006), while CBT tends to have more modest effects initially but is comparatively durable, there have been several investigations evaluating various combinations of CBT and CM. Results have largely indicated that CM is associated with better outcomes during the treatment period, but the combination of CM+CBT may produce greater rates of abstinence during the follow-up period (e.g., Budney, Moore, Rocha, & Higgins, 2006; Epstein, Hawkins, Covi, Umbricht, & Preston, 2003; Kadden, Litt, Kabela-Cormier, & Petry, 2007; Petitjean et al., 2014).

In contrast to the ample evidence regarding CBT’s efficacy, far less is known regarding the mechanisms of how it exerts its effects (Kazdin, 2007). As one of the primary elements of CBT is cognitive and behavioral skills training, most early studies of possible mechanisms of CBT focused on the improvement of these skills as a mediator of treatment effects. However, a seminal review by Morgenstern and Longabaugh (2000) concluded there was very little support for improvement in coping skills as a unique mechanism in CBT for alcohol use disorder. In the years since, some promising evidence has emerged supporting the acquisition and improvement in cognitive and behavioral control skills, as well as self-efficacy, as mediators (and potential mechanisms) of CBT’s effect on treatment outcomes. For example, improvement in the quality of individuals’ coping skills following computerized CBT was found to mediate treatment effects on abstinence from drugs, satisfying all criteria in the causal chain (Kiluk, Nich, Babuscio, & Carroll, 2010b). Also, increased self-efficacy has been found to mediate the relationship between drink refusal training (a specific ingredient of CBT) and drinking outcomes (Witkiewitz, Donovan, & Hartzler, 2012). Despite these findings, many trials have not found CBT to enhance coping or self-efficacy to a greater degree than comparison conditions, raining questions about the uniqueness of these mechanisms (Litt et al., 2008). Thus, consistent support for CBTs putative mechanisms of action remain elusive, as it does for many interventions (Emmelkamp et al., 2014).

Another key component of CBT, while not necessarily specific to it, is emphasis on extra-session practice assignments (hereafter referred to as homework) as a means of facilitating the generalization and maintenance of adaptive behavioral and cognitive skills. The general body of research on homework in psychotherapy has indicated a robust relationship between homework completion and treatment outcomes (Kazantzis, Whittington, & Dattilio, 2010). Although the amount of evidence supporting a homework-outcome relationship in CBT for substance use disorders is relatively sparse across various drugs of abuse, multiple reports have indicated greater homework completion has been associated with reduced drug use (Carroll, Ball, Martino, Nich, Babuscio, Gordon, et al., 2008; Carroll, Nich, & Ball, 2005; Gonzalez, Schmitz, & DeLaune, 2006). Moreover, there is evidence to suggest homework completion is not simply an indicator of symptom severity, or a proxy for motivation or treatment attendance (Decker et al., 2016). While more work in this area is needed, particularly with respect to predictors of homework completion, homework in CBT clearly plays in important role in successful treatment outcomes.

Traditional CBT, including relapse prevention, has been the prevailing empirically-based treatment approach for alcohol and drug use disorders for the past 30 years. However, several variants of CBT (i.e., ‘third wave’ behavioral therapies) have emerged that target the context and function of psychological events (e.g., thoughts, emotions, physical sensations) rather than primarily targeting the content, validity, intensity, or frequency of such events (Hayes, Villatte, Levin, & Hildebrandt, 2011). These ‘contextual CBTs’, such as Mindfulness Based Relapse Prevention (MBRP; Bowen et al., 2009), Dialectical Behavior Therapy (DBT; Linehan, 1993), and Acceptance and Commitment Therapy (ACT; Hayes, Luoma, Bond, Masuda, & Lillis, 2006), fit within a transdiagnostic approach to mental health and apply processes of acceptance, mindfulness, and values to treating substance use disorders. For instance, acceptance and emotion regulation skills are used to facilitate willingness to live with distressing events and reduce vulnerability to overwhelming emotions; mindfulness-based practices are used to increase awareness of triggers and consequences of substance use in a flexible and non-reactive manner; and living a valued and meaningful life is emphasized above a focus on reducing/eliminating addictive behaviors (Stotts & Northrup, 2015). While the evidence supporting the efficacy of these approaches toward treating alcohol and drug use is still emerging and not included in the meta-analyses described above (for reviews see: Dimeff & Linehan, 2008; Lee, An, Levin, & Twohig, 2015; Li, Howard, Garland, McGovern, & Lazar, 2017), they do appear to be promising enhancements to traditional CBT.

Stage 3 and Beyond: Dissemination and its Challenges

As is the case for other evidence-based approaches, it has been challenging to move CBT into widespread clinical practice (Emmelkamp et al., 2014; Harvey & Gumport, 2015; Institute of Medicine, 2001; Kazdin & Blase, 2011). General barriers to moving evidence-based behavioral treatments into clinical practice include lack of training and certification programs for clinicians, the cost of training and high rate of clinician turnover in many settings, lack of feasible means of evaluating and supporting fidelity in delivering evidence based therapies, different views on standards of evidence between researchers and practitioners, limited focus on sustainability, and several others (Addis & Krasnow, 2000; Crits-Christoph, Frank, Chambless, Brody, & Karp, 1995; Harvey & Gumport, 2015; Hoffman & McCarty, 2013; Institute of Medicine, 1998; McHugh & Barlow, 2010; McLellan, Carise, & Kleber, 2003; Olmstead, Abraham, Martino, & Roman, 2012; Weissman et al., 2006). Moreover, given the lack of a mandate for the substance use treatment system to track or report on meaningful treatment outcomes measures (Humphreys & McLellan, 2011) and weak, inconsistent efforts to require that providers demonstrate that they utilize evidence-based practice, there are few incentives for adopting evidence based treatments that might improve those outcomes (Carroll, 2014).

A particular barrier to effective dissemination of CBT is the lack of a system for training, supervision, and feedback to clinicians. Although ongoing monitoring and demonstration of clinician skill in delivering treatment and fidelity to manual guidelines is a methodological requirement for clinical trials evaluating behavioral therapies (Chambless & Hollon, 1998; Luborsky & DeRubeis, 1984; Rounsaville, Carroll, & Onken, 2001), systematic monitoring and feedback on clinicians’ implementation of evidence-based therapies is rare in clinical practice (Henggeler, Schoenwald, Liao, Letourneau, & Edwards, 2002; Hoffman & McCarty, 2013; Knudsen, Ducharme, & Roman, 2008; Martino et al., 2016; Roche, Todd, & O'Connor, 2007; Sholomskas et al., 2005). In recent years, multiple trials evaluating different methods of training clinicians to use evidence-based therapies have demonstrated that monitoring and supervision is significantly more effective than workshop-based training alone in a range of therapies (Beidnas & Kendall, 2010; Henggeler, Chapman, Rowland, Sheidow, & Cunningham, 2013; W.R. Miller, Yahne, Moyers, Martinez, & Pirritano, 2004; Schoenwald, Sheidow, & Chapman, 2009), including CBT (Rakovshik, McManus, Vazquez-Montes, Muse, & Ougrin, 2016; Rakovshik et al., 2013; Sholomskas et al., 2005). The lack of supervision and monitoring of clinician implementation of evidence-based treatments in clinical practice suggests that CBT and other EBPs, in practice, may bear little resemblance to the more closely monitored versions of those treatments as implemented in randomized clinical trials demonstrating their efficacy (Martino et al., 2016).

Data demonstrating very low fidelity in CBT among clinicians in community settings comes from a project exploring what constitutes ‘treatment as usual’ as part of two large multisite clinical trials evaluating Motivational Interviewing (MI) (Miller & Rollnick, 1991, 2002) and Motivational Enhancement Therapy (MET) for individuals seeking treatment for substance use disorders. In these two studies, clinicians working in community treatment settings who volunteered to participate in the protocols were randomized to either (1) a condition in which they were trained and supervised to deliver MET or MI with substance using individuals, or (2) continue to deliver the ‘standard treatment’ they usually would in that setting (treatment-as-usual, or TAU) (Ball et al., 2007; Carroll, Martino, et al., 2009). As a means of developing a fidelity rating system to evaluate implementation of MET/MI versus TAU in those trials, the 66 volunteer clinicians from the 11 participating sites were surveyed as to their usual theoretical orientation and techniques when working with clients at that site. Multiple orientations were endorsed, including 12-Step/disease concept, reality therapy, MI/MET, client centered, psychodynamic, and experiential; however, the most commonly endorsed orientation was relapse prevention/CBT (Ball et al., 2002). Nevertheless, review of taped TAU sessions by independent raters blind to treatment assignment indicated CBT techniques and strategies were among the most infrequently used in practice. Specifically, any clinician mention of cognitions or thoughts about substance use was identified in 14 of the 379 sessions rated and mention of skills training was detected only 13 times (Santa Ana et al., 2008). That these basic CBT components were detectable in less than 6% of all sessions rated suggests very limited success in disseminating CBT to the clinical community, at least in the settings included in those studies.

To summarize, while there are multiple barriers to dissemination of evidence based treatments as well as research/practice gaps in multiple fields outside the addictions (Institute of Medicine, 2001; Rogers, 1995), the relative complexity of CBT, the demand placed on clinicians and patients alike in terms of complexity of ideas and need for structure, as well as the need for ongoing supervision to support fidelity, are particular challenges for the dissemination of CBT and constrain the implementation of standardized, high quality CBT in clinical practice.

The Promise of Technology

Rapid developments in the sophistication, speed and reach of technology have opened up multiple new possibilities for disseminating evidence based therapies (Bickel, Christensen, & Marsch, 2011; Copeland, 2011; Emmelkamp et al., 2014; Gustafson et al., 2011; Kazdin & Blase, 2011; Marks, Cavanagh, & Gega, 2007; Marsch, Carroll, & Kiluk, 2014; Marsch & Gustafson, 2013), including cognitive behavioral therapies. Platforms for delivering addiction interventions via technology are diverse and multiplying rapidly. These include electronic screening and brief intervention (eSBIs)(Carey, Scott-Sheldon, Elliott, Bolles, & Carey, 2009; Copeland & Martin, 2004; Fachini, Aliane, Martinez, & Furtado, 2012; Gryczynski et al., 2015; Ondersma, Svikis, & Schuster, 2007); web-based multi-module programs and smartphone apps, with and without clinician involvement (Bickel, Marsch, Buchhalter, & Badger, 2008; Gustafson et al., 2014; Suffoletto et al., 2015) treatment delivered ‘live’ via Skype, telephone, or instant messaging (McKay, Lynch, Shepard, & Pettinati, 2005; McKay et al., 2004; McKay et al., 2011; McKay et al., 2010); monitoring via interactive voice response (IVR) and Ecological Momentary Assessment (EMA) and Ecological Momentary Treatment (EMT) platforms (Moore et al., 2013; Morgenstern, Kuerbis, & Muench, 2014) and several more (Muench, 2014). Overall, results from meta-analyses of such interventions are promising (Boumparis, Karyotaki, Schaub, Cuijpers, & Riper, in press; Carey et al., 2009; Riper et al., 2014; Rooke, Thorsteinsson, Karpin, Copeland, & Allsop, 2010; Tait, Spijkerman, & Riper, 2013), but methodological quality of studies within this young field is variable and often weak (Kiluk, Sugarman, et al., 2011). In the sections below we will cover only those which (1) are explicitly or predominantly cognitive-behavioral in focus (although several include components of MI and other interventions), (2) the primary targeted outcome is alcohol or drug use, and (3) the intervention is delivered online. We include an expanded description of a computer-based CBT program developed by our research group as a possible paradigm for how CBT evolve in the future.

Alcohol use disorders

Multiple studies have evaluated technology-based CBT interventions for alcohol use disorders, most of which focus on problem drinkers/heavy drinkers, rather than those with alcohol use disorder. A six-week on-line cognitive-behavioral self-help intervention for adult problem drinkers showed promise in a randomized controlled trial (N=261) conducted in the Netherlands (Riper et al 2008). Participants who utilized the interactive self-help intervention (Drinking Less) reduced their drinking significantly more than participants who received an on-line psychoeducational brochure about alcohol use. Seventeen percent of those receiving the intervention reduced their drinking to levels considered low-risk in the Netherlands (no more than 20 g of alcohol per day) compared with 5.4 percent of those receiving the educational brochure.

Another study of adult problem drinkers in the Netherlands (Blankers, Koeter, & Schippers, 2011) randomized 205 problem drinkers to one of three interventions: wait list, Self-Help Online (SAO), a fully automated, internet based, self-guided treatment program based cognitive behavioral treatment and MI principles, and Therapy Alcohol Online (TAO), which combined Self Help Alcohol Online with up to seven synchronous text-based chat-therapy sessions with a trained therapist. Three-month follow-ups indicated significant differences in alcohol consumption and problems favoring both SAO and TAO relative to wait list control (27.0 versus 22.4 versus 35.5 mean drinks per week, respectively), but at 6 months, TAO had better outcomes than SAO (17.8 versus 26.2 drinks per week, p=.03), suggesting benefit of clinician involvement in terms of producing durable improvement in drinking.

Cunningham and colleagues compared an online SBI (Check Your Drinking) to Check Your Drinking plus an extended internet intervention called the Alcohol Help Center (AHC), which provides cognitive-behavioral and motivational components in a randomized clinical trial with 170 problem drinkers (Cunningham, 2012). Baseline drinking in this predominantly male sample was high (AUDIT mean score 22.1). Seventy-two percent of the sample accessed their assigned condition and 6-month follow-up self-reports were obtained from 90% of the randomized sample. Multivariate analysis of variance indicated greater reduction in drinking outcomes for those assigned to SBI+AHC relative to SBI alone (10.3 versus 11.5 mean highest drinks per occasion at 6 month follow-up, p=.02).

Hester and colleagues (Hester, Delaney, & Campbell, 2011) conducted a study in which 78 non-dependent problem drinkers were randomized to either Moderation Management alone (www.moderation.org) (either delivered in-person or web-based) or Moderation Management plus online training in moderation management using the “Moderate Drinking” app (www.moderatedrinking.com). While both groups significantly decreased the amount they drank, those assigned to the combination reported a higher percentage of days abstinent (43.9 versus 22.6%) and fewer alcohol-related problems than the group utilizing Moderation Management only.

In a subsequent study conducted entirely over the internet, Campbell, Hester and colleagues compared a cognitive-behavioral online program with components of MI called Overcoming Addictions in combination with an online and in-person mutual help group for problem drinkers called SMART Recovery (SR) (OA+SR) to SMART Recovery alone (Campbell, Hester, Lenberg, & Delaney, 2016; Hester, Lenberg, Campbell, & Delaney, 2013) in 189 heavy drinkers (AUDIT mean 24.7). Three and six-month follow-up indicated significant reductions in drinking outcomes, but no significant differences between conditions (PDA at 3 months follow up was 73.3 for OA+SR compared with 71.2 for SR only).

Gonzalez and Dulin reported on a 6-week pilot study of a MI/CBT smartphone app (Location-Based Monitoring and Intervention for Alcohol Use Disorder, LBMI-A) for 52 individuals with alcohol use disorder (Gonzalez & Dulin, 2015). Participants were randomized to either LBMI-A or the online Drinkers Check-up plus bibliotherapy (Hester, Squires, & Delaney, 2005). Both treatments were delivered as stand-alone interventions and uptake was very good. The two interventions were associated with significant reduction in drinking (PDA, percent heavy drinking days) over the 6 weeks, with comparable drinking outcomes overall, with some indicators of more rapid changes for LBMI.

Other substance use disorders

Turning to drug use disorders, Bickel, Marsch and colleagues developed the Therapeutic Education System (TES), a computerized version of the well-validated Community Reinforcement Approach (CRA)(Azrin, 1976; Budney & Higgins, 1998). TES combines over 60 online modules, many consistent with CBT, with voucher based contingency management where users receive vouchers upon submission of drug-free urine specimens. In their first study, TES was compared with therapist-delivered CRA (maximum value of vouchers was $1316 if all urines were negative for cocaine and opioid metabolites) with standard counseling for 135 opioid dependent adults maintained on buprenorphine. Those assigned to either form of CRA (therapist delivered or via TES) had significantly longer periods of continuous abstinence (8.0 weeks for therapist-delivered CRA and 7.8 for TES) compared with standard buprenorphine treatment (4.7 weeks) in this 23-week trial (Bickel et al., 2008). Subsequent trials of TES have indicated benefit of adding TES to standard methadone treatment (Marsch, Guarino, et al., 2014) as well as the benefits of the TES system itself in addition to contingency management in the context of buprenorphine maintenance treatment (Christensen et al., 2014). A recent multisite trial of TES plus prize based contingency management (Petry et al., 2005) (wherein participants earned chances to earn prizes for submitting drug free urine specimens or completing TES modules) was conducted in 10 outpatient settings with 507 drug-using individuals in a 12-week trial. Compared with standard outpatient treatment, TES was associated with significantly better retention and greater abstinence (Campbell et al., 2014). However, it was not possible to disentangle the effects of the TES program from the effects of the prize contingency management (participants earned an average of $277 worth of prizes over 12 weeks) and favorable effects of TES over standard treatment were not sustained through 3- and 6-month follow-ups.

Focusing on individuals with cannabis use disorders, Budney and colleagues developed a computerized approach encompassing MI, CBT and contingency management (Budney et al., 2015). In a randomized controlled trial, 75 adults with marijuana use disorder were randomized to 2 sessions of brief treatment versus a 9-session clinician-delivered MET-CBT approach, or a 9-session version of TES emphasizing MET and CBT. Both MET/CBT conditions included a CM component in which participants could earn up to $435 in gift cards if all urines were negative for cannabis. Significantly more participants in clinician-delivered treatment (44.8%) and TES (46.7%) were abstinent at the end of treatment compared with the 2-session brief treatment (12.5%). Similarly, both the therapist- and computer delivered approaches were significantly more effective in reducing cannabis use compared with brief intervention alone during treatment, but effects weakened during follow-up and were no longer significant at the 6-month follow-up point. Again, it was not possible to disentangle the effects of MET/CM versus the contingency component in understanding what drove reductions in cannabis use. Interventions combining MI and CBT, without a contingency management component, have also shown promise among individuals with comorbid depression and alcohol or cannabis use disorders relative to brief intervention alone, including the SHADE program developed and tested in Australia (Kay-Lambkin, Baker, Lewin, & Carr, 2009; Kay-Lambkin, Simpson, Bowman, & Childs, 2014).

Several studies have developed web-based interventions for individuals with stimulant use disorders (amphetamines or cocaine). In a study conducted fully online in Australia, 160 individuals with self-reported amphetamine use problems were randomly assigned to a three-session computerized intervention based on MI with some components of CBT or a wait list control (Tait et al., 2014). Uptake of the computerized intervention was weak, with only 63% of those assigned to this condition accessing a module, and rates of 3-month follow-up were modest across conditions (57% of those in waitlist control and 48% of those assigned to computerized intervention). As standard outcomes (urinalyses or self-reported days of amphetamine use) were not reported, it was difficult to draw conclusions regarding the efficacy of the intervention in this sample.

Similar limitations occurred in a Swiss study of an 8-module internet-based program encompassing CBT and MI called Snow Control for individuals reporting cocaine use at least 3 times in the past 30 days (Schaub, Sullivan, Haug, & Stark, 2012). Participants were randomly assigned to the Snow Control program or an 8-session online psychoeducation control. Treatment engagement was very low, with only 18/96 (19%) allocated to the Snow Control program accessing a module and only 8 of the 100 allocated to control. Outcome data did not indicate significant differences in cocaine use outcomes by group.

Follow-up rates were also low (5.6% of the randomized sample were reached for 6-month follow-up assessment); making it difficult to make inferences regarding the efficacy of the program. The studies reviewed above highlight both the promise of technology-based interventions as well as their significant limitations, which include highly variable rates of retention and adherence and poor rates of follow-up, particularly for studies collected entirely on-line (Kiluk et al., 2010). Several studies have used use wait-list controls, hence limiting the inferences that can be drawn regarding the efficacy of the intervention evaluated. Issues of privacy and confidentiality are particularly important to consider when dealing with individuals who are users of illicit drugs, particularly in the era of electronic medical records (Ramsey et al., 2016). Finally, while validated technology based interventions are generally less expensive than traditional clinician-delivered interventions, the lack of a reimbursement structure for these interventions constrains their availability to date.

CBT4CBT and the Stage Model

Our research group at Yale has taken a programmatic approach to development of computerized CBT based on the Stage Model of Treatment Development (Onken, Carroll, Shoham, Cuthbert, & Riddle, 2013; Rounsaville et al., 2001). The Stage Model emphasizes a systematic, programmatic approach to treatment development and refinement so that the most potent version of an intervention is developed in a form compatible with clinical practice. In 2004, after recognizing that CBT was not being widely implemented with high levels of fidelity in clinical practice at Stage 3, we returned to Stage 1 with the goal of developing a maximally efficient, computer-delivered version of CBT that retained key features associated with clinician-delivered CBT, including durability of effects (Carroll et al., 1994a). Thus, we sought to develop a highly engaging, interactive approach that emphasized (1) conveying key cognitive and behavioral interventions as effectively as possible, and (2) practice of targeted skills via practice in the real world. As noted above, a significant limitation of many studies of online interventions is low levels of uptake and retention; thus, we sought to make the program as appealing as possible via a ‘media-rich’ format with minimal use of text (concepts are introduced by online narratives and animations), multiple interactive features including quizzes and interactive homework assignments, and extensive use of video vignettes (short, well-produced movies with likeable characters played by professional actors) demonstrating the targeted skills (Carroll, Ball, Martino, Nich, Babuscio, Nuro, et al., 2008).

CBT4CBT (computer based training in cognitive behavioral therapy) covers seven key cognitive behavioral skills, or ‘modules’, (functional analyses, coping with craving, refusing offers of drugs or alcohol, problem solving skills, recognizing and changing thoughts, decision making skills, and HIV/HCV risk reduction). Each of the seven modules follows a common format, roughly parallel to a traditional CBT session; check-in and review of homework, introduction of the skill to be taught in that module, a video vignette depicting an actor in a challenging situation, teaching of the skills, demonstration of the skill through another vignette with the same actor using the skill to avoid substance use, demonstration of practice exercise by the actor, a walk-through of the practice exercise, and a short quiz to test understanding of the skill.

In evaluating CBT4CBT, we adhered to the Stage Model by first testing it in small randomized pilot studies, emphasizing internal validity and mechanism at Stage 1, before conducting larger studies with fewer experimental controls (Rounsaville et al., 2001). As such, we attempted to avoid some of the methodological problems common to some studies of web-based interventions, such as weak controls (e.g., wait lists), lack of attention to treatment adherence and fidelity, reliance on unvalidated self-reports of change, low rates of follow-up and others (Kiluk, Sugarman, et al., 2011). The first pilot study randomized 77 substance users to either standard outpatient treatment at a community setting or standard treatment plus access to CBT4CBT for 8 weeks. This approach allowed monitoring of participants use of the program, verification of self-reports of drug use via biological markers, and a relatively strong comparison condition (treatment as usual, TAU). This trial demonstrated significant differences in drug free urines by condition (Carroll, Ball, Martino, Nich, Babuscio, Nuro, et al., 2008) as well as continuing improvement for those assigned to CBT4CBT through a 6-month follow-up (Carroll, Ball, et al., 2009). In addition, a role play assessment of acquisition of coping skills demonstrated that assignment to CBT4CBT was associated with a significant increase in the quality of coping skills from pre- to post-treatment relative to standard treatment, with continuing improvement for the CBT4CBT group through the 6-month follow-up (Kiluk, Nich, Babuscio, & Carroll, 2010a). Mediational analyses indicated that acquisition of coping skills accounted for the effect of CBT4CBT versus TAU treatment on substance use outcomes, pointing to the effect of CBT4CBT on skills training as a potential mechanism of action for CBT4CBT (Kiluk et al., 2010a). Thus, data from this initial pilot trial suggested that key features of traditional, clinician-delivered CBT appeared to be retained in its conversion to an easily disseminable web-based format, in that CBT4CBT appeared to improve the targeted skills and was associated with enduring effects. Secondary analyses of data from this trial also indicated (1) relative benefits of CBT4CBT over TAU in increasing self-reported use of coping skills (Sugarman, Nich, & Carroll, 2010), (2) that completing homework is strongly associated with better drug use outcomes in CBT4CBT (Decker et al., 2016), and (3) better response to CBT4CBT and acquisition of skills among individuals with higher estimated IQs at baseline (Kiluk, Nich, & Carroll, 2011). Other secondary analyses suggested comparable response to CBT4CBT by gender (DeVito, Babuscio, Nich, Ball, & Carroll, 2014) and ethnicity (Montgomery & Carroll, in press).

We then conducted a Stage 2 randomized trial, which also evaluated CBT4CBT as an add-on to standard treatment, but took place in a more homogeneous but challenging sample: 101 cocaine-dependent methadone maintained individuals (Carroll et al., 2014). In this study, CBT4CBT plus standard treatment was significantly more effective in reducing cocaine use than standard methadone treatment alone over the 8-week course of treatment and its effects were maintained through a 6-month follow-up (Carroll et al., 2014). The evaluation of acquisition of coping skills demonstrated slightly different findings from the first trial, as a larger percentage of participants had baseline scores indicating relatively high levels of skill, which might be expected in a group that had been maintained on methadone for some time and thus had considerable prior exposure to CBT-based group treatment. Thus, while a significant treatment by time effect in acquisition of coping skills was not found for the full sample, it was present among the subgroup with lower baseline levels of skills (Kiluk et al., under review).

The next step toward moving to dissemination involved evaluating whether CBT4CBT could be delivered safely as a stand-alone intervention rather than an adjunct to treatment. Therefore, our third trial involved an adaptation of CBT4CBT for individuals with primary alcohol use disorders in which 68 individuals were randomly assigned to one of three conditions: standard outpatient alcohol treatment in a community based setting (predominantly weekly group therapy, TAU), TAU plus CBT4CBT, or CBT4CBT delivered with only brief clinical monitoring (about 10 minutes each week). The clinical monitoring sessions were offered because this was moderately severe (baseline AUDIT mean was 19), treatment-seeking sample who met criteria for DSM-IV-R alcohol dependence) and the weekly brief check-ins provided an opportunity to closely monitor each participant’s clinical status and determine if this level of care was sufficient. Results indicated significantly higher rates of treatment completion in either condition offering CBT4CBT compared to TAU (CBT4CBT+TAU=65%; CBT4CBT alone, 63%, TAU 26%). Percentages of participants with no heavy drinking days in the last 4 weeks of treatment were 9% for TAU, 37% for TAU+CBT4CBT, and 33% for CBT4CBT as a virtual standalone. In addition to validating CBT4CBT for alcohol as an adjunct to treatment compared with standard care, this trial also indicated that CBT4CBT as a standalone was safe and highly acceptable to participants as an alternative to unstructured group therapy. Moreover, CBT4CBT as a standalone produced better retention and comparable outcomes at much lower cost, as multiple individuals assigned to TAU dropped out or were withdrawn from treatment and later sought expensive inpatient care (Kiluk et al., 2016). Similar findings are emerging from a recently completed study evaluating the drug version of CBT4CBT as a virtual stand-alone intervention in a diverse group of treatment-seeking substance users (Clinical Trials.gov NCT01442597). CBT4CBT has now moved into Stage 3 and 4, with a number of ongoing dissemination trials evaluating it as a standalone to enhance SBIRT (Screening, Brief Intervention, and Referral to Treatment) in primary care settings by providing the intervention on site rather than referring individuals to specialty care, as many do not follow through on referral. It is also being evaluated in a range of settings where access to evidence-based therapies has been limited, including rural drug courts, Native American and First National Health Centers, and office-based buprenorphine maintenance.

CBT in the Next Thirty Years

The studies reviewed above suggest that technology-based CBT interventions, provided that they are carefully constructed, developed to be as engaging as possible, and rigorously evaluated in methodologically sound clinical trials, have tremendous potential as a dissemination strategy to reach the majority of individuals with substance use problems who do not receive care due to issues of access, stigma, costs, concerns about confidentiality, and many more (Carroll & Rounsaville, 2010; Kazdin & Blase, 2011). While one important direction for technology-based interventions is as a dissemination tool to improve access to evidence based treatments, a second direction may lie in exploiting features of technology-based interventions to develop more potent, individualized, and scalable versions of CBT, better linked to and informed by cognitive science and neuroscience than traditional CBT interventions. In addition, technology offers strategies for enhancing our ability to study CBT and other interventions more systematically and more rigorously. In the sections below we elaborate on how these possibilities may accelerate development of cognitive behavioral interventions in the next 30 years.

Modularization and evaluation of effective components

In the years ahead, we may move towards technology driven, modularized versions of CBT in an effort to optimize treatment and away from ‘one size fits all’ approaches toward precision behavioral intervention. For example, in developing CBT4CBT, the focus was on developing a small set of core modules, each of which would convey a fundamental cognitive and behavioral principle as clearly and effectively as possible. While each module used examples and vignettes relating specifically to drug or alcohol use, it was made explicit that each skill was generalizable to a wide range of behaviors or problems other than substance use. Thus, as shown in Figure 1, the ‘concept of functional analysis’ demonstrates understanding patterns of substance use by analyzing thoughts, feelings, and behaviors before and after an episode of substance use, but also makes clear that the same basic behavioral analysis can be used to better understand, and then change, any problematic behavior. The module ‘coping with craving’ teaches skills associated with recognizing craving for drugs and alcohol and multiple coping strategies (urge surfing, monitoring, relaxation), but also emphasizes that these same strategies can be used to help tolerate any strong feeling without impulsive responding (i.e., affect tolerance). The ‘refusal skills’ module focuses on refusing offers of drugs or alcohol, but makes explicit links with assertive responding, persisting with goals under stress, and effective interpersonal functioning. The module on ‘problem solving skills’ provides multiple examples of applying basic problem solving strategies to issues commonly reported by individuals with substance use disorders, but the various practice exercises demonstrate how these basic steps are applicable to almost any scenario or problem. The module on recognizing and changing thoughts uses examples related to cognitions commonly associated with drug use, but relates this to negative thoughts as well and their relationship to strong affect and behavior. The module on decision making skills, or ‘seemingly irrelevant decisions’, uses the classic example of a man unexpectedly confronting a high risk situation (Marlatt & Gordon, 1985), but also includes exercises on the need to anticipate consequences of any decision as a means of temporizing behavior and reducing impulsive responding. Thus, we have moved toward conceptualizing CBT4CBT as a set of cognitive control strategies, each focused on a different aspect of impaired control over behavior. In so doing, we are also moving toward a more transdiagnostic model of CBT (Sauer-Zavala et al., 2017), which recognizes that individuals with substance use disorders typically have a range of psychiatric and psychosocial problems (Hasin, Stinson, Ogburn, & Grant, 2007; McLellan, Cacciola, Alterman, Rikoon, & Carise, 2006; Rounsaville et al., 1991). As they are often more concerned with these other problems than their substance use, this broader conception of CBT may broaden its appeal and applicability.

Most work to date has involved evaluating CBT interventions as a package, as in traditional CBT. Thus, it has been challenging up till now to evaluate the individual contribution of each of these components, or modules. For example, when CBT is delivered by clinicians, multiple concepts may be discussed in each session and the frequency or intensity of any specific concept or intervention is often relatively low (Carroll, Nich, Sifry, et al., 2000), making it difficult to isolate or evaluate individual components. Furthermore, isolating specific effects from important ‘relational effects’, such as the therapeutic alliance, is fundamentally impossible when interventions are delivered in the context of an interpersonal relationship (Miller & Moyers, 2015), thereby complicating the evaluation of mechanism. With standardization and modularization of CBT components via technology, however, it is much more feasible and straightforward to evaluate the contribution (or lack thereof) of any single component of CBT to the efficacy of the package. In other words, the effect of any single module could be evaluated in terms of whether it successfully conveys the targeted skill and whether it contributes to outcome; or, does the “coping with craving” module actually have an effect on craving or affect tolerance, and does including or excluding it in the package matter?

Another potential, but largely unexplored advantage of technology-based, modularized CBT approaches is that they are compatible with evaluation and refinement via the MOST (Multiphase Optimization Strategy) approach of Collins and colleagues (Collins, Murphy, & Stretcher, 2007; Collins et al., 2007; Collins et al., 2014). Briefly, MOST approaches utilize factorial (and fractional factorial) designs to efficiently evaluate individual components of an intervention and their contribution to producing outcome. MOST designs have been successfully implemented in smoking research to refine multicomponent interventions for smoking (Piper et al., 2016; Schlam et al., 2016). Given that attrition remains high in most clinician- and technology-delivered interventions and few individuals actually complete a full course of the intervention, MOST strategies for refining interventions and delivering highly parsimonious and effective components to the largest possible samples have clear promise, as the majority of substance users remain in treatment for only a matter of weeks and thus it is imperative we deliver our most potent and effective interventions from the outset.

Another promising pathway afforded by conceiving of CBT as a set of components is using single, highly targeted interventions to focus on single core features of addictive behaviors. Examples of this approach are proliferating and include the use of strategies designed to modify attentional biases in individuals with addictive behaviors (Leeman, Robinson, Waters, & Sofuoglu, 2014; Schoenmakers et al., 2010; Wiers, Eberl, Rinck, Becker, & Lindenmeyer, 2011; Wiers et al., 2015). Bickel has evaluated intensive working memory training as a means of reducing delay discounting in substance users and fostering greater future orientation (Bickel, Yi, Landes, Hill, & Baxter, 2011). Another example is an ongoing trial building an online intervention for craving based on Kober’s Regulation of Craving Task (Kober, Kross, Mischel, Hart, & Ochsner, 2010), wherein smokers are trained over several weeks in either mindfulness-based (experience craving) or cognitive-based (think about consequences) strategies in response to visual smoking cues (ongoing, clinicaltrials.gov NCT 02153749).

While the development of effective, individual components demonstrated to address a single core feature of addiction (e.g., attentional bias, craving, delay discounting) may have limited impact if delivered alone, it has the potential to lead to strategies that allow us to more efficiently tailor treatments for complex and heterogeneous disorders like the addictions. Thus, individual “A’ who is impulsive and discounts future rewards might receive a version of CBT that contains components addressing problem solving, working memory, decision making skills, and cognitive control training, while individual ‘B’, who is depressed and concerned with craving for cocaine, might receive a set of components focusing primarily on attentional bias training, affect tolerance, challenging cognitions, and even modules from computerized CBT for depression.

Linkages with neuroscience and cognitive science

Development of maximally potent, optimized treatment components tailored for specific individuals would require a different approach to assessment, one that is more closely linked to core features of addiction. The Addictions Neuroclinical Assessment (ANA), a new neuroscience-based framework for assessment, identified three core functional domains relevant to defining meaningful subtypes of individuals with addictive disorders (including behavioral addictions) and functional outcomes: executive functioning, negative emotionality, and incentive salience (Kwako, Momenan, Litten, Koob, & Goldman, 2016). Although aspects of ANA will require data-driven refinement over many years, it is an exciting starting point and highly useful heuristic framework for developing interventions better linked to and informed by neuroscience and cognitive science (DeVito, Carroll, & Sofuoglu, 2016).

Figure 1 demonstrates how this framework might be applied to CBT, using CBT4CBT as an example. The CBT4CBT modules targeting executive functioning include functional analyses, problem solving, and decision making skills, all of which are focused on temporizing behavior and goal-setting. Negative emotionality is addressed by the modules teaching coping with craving/affect tolerance and addressing and changing negative thoughts. Incentive salience is addressed by the coping with craving and refusal skills modules. While it is encouraging to consider how these traditional CBT components map on to these more novel conceptualizations of the core features of addiction, it is imperative that we now move towards evaluating how well CBT interventions actually affect change in these dimensions. The ANA suggests relevant assessments of each of these domains (Kwako et al., 2016), many of which have been included in the batteries used in the various CBT4CBT studies (NB: NIDA is currently developing a similar phenotyping battery for drug use disorders).

Figure 1.

CBT concepts and relationship to core features of addictions

We have begun linking these aspects of the core features (negative emotionality, incentive salience, and executive function) to response to CBT. For example, in terms of executive functioning, we have included relevant neuropsychological tasks from CANTAB (Cambridge Automated Neuropsychological Test Battery) (Robbins et al., 1994) in our studies. In our most recent study of the cognitive enhancer galantamine and CBT4CBT in cocaine-dependent methadone-maintained individuals (Carroll, DeVito, Shi, Nich, & Sofuoglu, in press), we found that while these indicators of executive function demonstrated little change during treatment, they were nevertheless consistently associated with treatment outcome. Baseline measures were significantly positively correlated with percentage of urine specimens submitted that were negative for all drugs, including sustained attention (CANTAB RVP A’), visual memory (CANTAB PRM% correct), Digits Backward, and a composite of all cognitive measures. This is consistent with numerous findings linking impaired cognitive function and poorer outcome in CBT (Aharonovich, Nunes, & Hasin, 2003; Bates, Buckman, & Nguyen, 2013; Litt, Kadden, Cooney, & Kabela, 2003; Sofuoglu, DeVito, Waters, & Carroll, 2013). Given that (1) cognitive impairment may not be directly improved by CBT, and (2) CBT’s relatively high level of cognitive demand may be particularly challenging for patients with difficulties with attending to interventions, remembering them, and implementing them effectively, we are exploring whether cognitive training, delivered prior to the initiation of CBT, might improvement response, by introducing these relatively complex cognitive and behavioral coping skills delivering outcome after interventions that target memory and attention (ongoing, clinicaltrials.gov NCT02174614).

In terms of with negative emotionality, we found that individuals higher in alexithymia, as measured by the Toronto Alexythymia Scale (Taylor, Parker, & Babgby, 1990) had better cocaine use outcomes when assigned to CBT4CBT versus TAU alone (Morie, Nich, Hunkele, Potenza, & Carroll, 2015), possibly by teaching CBT skills without the emotional demand associated with interpersonal interactions associated with group or individual therapy. In terms of incentive salience, we included an emotional Stroop task (Hester, Dixon, & Garavan, 2006; Waters et al., 2005) in the 2014 study (Carroll et al., 2014). Evaluation of pre-to post changes in reaction time indicated individuals assigned to CBT4CBT reduced reaction time in response to cocaine words significantly more than individuals assigned to TAU, suggesting CBT4CBT may be associated with reducing attentional bias for cocaine cues in this cognitive control task (DeVito, under review). This work is in fairly early stages, but Figure 2 suggests how this general approach might be explored more systematically in the future, for example, through a series of small trials investigating the effect of each module on the behavioral target, similar to the strategy implemented by Baskin-Somers and colleagues (Baskin-Sommers, Curtin, & Newman, 2015). Eventually, ANA-linked frameworks might be used for pre-intervention assessment of individuals who would then be exposed to specific interventions designed to modify their own pattern of specific deficits and strengths. Again, while this strategy would be complex and cumbersome if implemented with clinicians, it would be relatively straightforward if both assessment and treatment delivery were largely computerized.

Figure 2.

Strategies for linking CBT and cognitive neuroscience

Mechanisms and neuroscience

Finally, we anticipate the most promising strategy for improving and refining CBT in the next 30 years will be by consistent focus on mechanism. As outlined above, we must consistently seek to demonstrate that our interventions change the basic behavioral and cognitive targets of addiction and understanding why CBT has its enduring effects. Unlike many other therapies, CBT seeks to convey strategies for achieving and maintaining cognitive and behavioral control, and new technologies may help us do so more effectively. Neuroimaging, EEG, and advances in cognitive assessment should be harnessed in order to demonstrate how CBT changes cognition, behavior and the brain (Cabrera et al., 2016; Verdejo-Garcia, 2016; Weingarten & Strauman, 2015). As noted by Garland and Howard (2009), “Neuroplasticity represents a plausible biological mechanism through which psychological interventions may exert some of their therapeutic effects”. Powerful, rapidly developing tools like neuroimaging should help us explore how effective interventions like CBT may affect cognitive functions and identify markers of who responds to CBT (Chung et al., 2016; Konarski et al., 2009; Morgenstern, Naqvi, Debellis, & Breiter, 2013; Potenza, Sofuoglu, Carroll, & Rounsaville, 2011). It will take some time before this approach bears fruit, as we are far from being able to reliably disentangle changes associated with delivery of common or specific treatment elements, natural recovery that occurs during prolonged abstinence, premorbid cognitive functioning and drug-related changes in cognitive function (Chung et al., 2016) or understanding the neuroscience of successful abstinence (Garavan, Brennan, Hester, & Whelan, 2013). Nevertheless, the future of our field may lie in demonstrating that behavioral therapies like CBT are indeed biological therapies (Kandel, 1998).