Randomized clinical trial evaluating the preliminary effectiveness of an integrated anxiety disorder treatment in substance use disorder specialty clinics

Abstract

Objective

Anxiety and substance use disorders are highly comorbid and mutually maintain each other. Treatments for anxiety disorders that are well-integrated into substance use disorder treatment have the potential to improve both anxiety and substance use outcomes.

Methods

Ninety-seven individuals seeking treatment at a community-based, evidence-based Intensive Outpatient Program for substance use disorders who also had anxiety disorders were randomized to either (a) usual care (UC) at the Intensive Outpatient Program; or (b) UC + Coordinated Anxiety Learning and Management for Addiction Recovery Centers (CALM ARC), a 7-session, group-based, computer-assisted but therapist directed treatment for anxiety disorders adapted for individuals with anxiety disorder and substance use disorder comorbidity.

Results

CALM ARC + UC outperformed UC on measures of anxiety and substance use at post-treatment and at a 6-month follow-up.

Conclusions

Adding CALM ARC to UC for patients with comorbid anxiety disorders and SUD is superior to UC alone. Implications for future research and clinical practice are discussed.

The comorbidity of anxiety disorders and substance use disorders (SUD) is well-documented (Compton et al., 2007; Grant et al., 2004). Anxiety disorder comorbidity among those with SUDs is associated with greater clinical severity (Burns, Teeson & O’Neill, 2005; Glasner-Edwards et al., 2010) and poorer treatment outcomes (Smith & Book, 2010; Kushner et al., 2005) compared to those with SUDs only, making it a significant public health concern. Although effective treatments for anxiety disorders exist, with cognitive behavioral therapy (CBT) considered the treatment of choice (Arch & Craske, 2008), most individuals with anxiety disorders do not receive these treatments (Santucci, McHugh & Barlow, 2012). Thus, greater accessibility to CBT for anxiety disorders is needed. SUD specialty clinics represent settings with an elevated prevalence of anxiety disorders that typically go untreated (e.g., McGovern et al., 2006), and thus may be an important clinical setting to target.

Introducing CBT for anxiety disorders into SUD community practice settings has the potential to improve not only anxiety but also SUD outcomes. Despite the advances in SUD treatment, relapse rates and treatment dropout remain high (Dutra et al., 2008; Karwczyk et al., 2017). One possible way to improve SUD outcomes is to treat underlying negative affective states such as anxiety that may maintain and exacerbate problems with substance use (Stewart & Conrod, 2008). Models of care that integrate anxiety treatment into SUD treatment may be optimal to mitigate the mutual maintenance of anxiety and SUD symptoms. Indeed, studies examining the effects of anxiety disorder treatment on SUD outcomes are limited, with mixed findings (see Wolitzky-Taylor, Operskalski, Ries, Craske & Roy-Byrne, 2011), but the few studies that have integrated CBT for posttraumatic stress disorder (Najavits et al., 2006) and panic disorder (Kushner et al., 2006) into SUD treatment have had more favorable outcomes than non-integrated treatments. CBT for anxiety disorders that is well-integrated into SUD treatment may be more clinically relevant, consistent, and manageable for patients and their providers.

The Coordinated Anxiety Learning and Management (CALM) Study developed a computerized, therapist-directed CBT protocol for anxiety disorders that was delivered in primary care by nurses and social workers without previous CBT training (Craske et al., 2009). The CBT program focused on teaching overarching cognitive and behavioral skills and principles relevant across the anxiety disorders. CALM demonstrated moderate effect sizes compared to usual care in a large effectiveness trial (Roy-Byrne et al., 2010). Given its success, we aimed to develop and evaluate an adaptation of CALM for patients in SUD specialty clinics with comorbid anxiety disorders, which we called CALM for Addiction Recovery Centers (CALM ARC). The development of CALM ARC focused on creating a CBT program for anxiety disorders that would fit within the structure and culture of SUD specialty clinics in order to increase the likelihood of adoption and sustainability in clinical practice. As a result, we developed a brief, group-based CBT program for anxiety disorders that was delivered by SUD counselors using a CALM-like model: a computerized, therapist-directed, highly structured, and interactive program. Utilizing a hybrid efficacy/effectiveness randomized clinical trial design to balance internal and external validity, our aim was to evaluate the preliminary effectiveness of CALM ARC in reducing anxiety and substance use symptoms among patients with comorbid anxiety disorders and SUD enrolled in a community-based, evidence-based SUD Intensive Outpatient Program (IOP) compared to the evidence-based SUD IOP treatment alone. We hypothesized that those randomized to usual care (UC) + CALM ARC would show superior outcomes on measures of anxiety and substance use compared to those randomized to receive only UC.

Methods

Experimental Design and Randomization

See Table 1 for the assessment schedule and randomization procedures and Figure 1 for the design and flow of participants.

Table 1.

Assessment Schedule and Measures

Assessment	Construct (Measure)	Purpose and Other Information
Period/Time
Initial Screening	Anxiety (Overall Anxiety Severity and Impairment Scale [OASIS] Campbell-Sills et al., 2009)	Potential participants (those enrolled at Matrix Institute IOP) scoring ≥ 8, indicating anxiety causing at least moderate distress or impairment, were invited for baseline eligibility assessment
Baseline eligibility	1. Anxiety (OASIS; Campbell-Sills et al., 2009)	Determine Inclusion Criteria: Participants must score ≥ 8 on OASIS at the baseline assessment; Cronbach’s α = .77 in this sample
	2. Psychiatric diagnoses (Mini Neuropsychiatric Interview [MINI] for DSM-5; Sheehan, 2016)	Determine Inclusion Criteria: Participants must meet DSM-5 diagnostic criteria for at least one of the following disorders: panic disorder, agoraphobia, social anxiety disorder, specific phobia, generalized anxiety disorder, obsessive compulsive disorder, post-traumatic stress disorder; and for substance use disorder (SUD), including alcohol or any drug and Determine Exclusion Criteria: unstable manic or psychotic symptoms, active suicidality Inter-rater reliability was assessed on 22.6% of the MINIs administered. κ = .80 across anxiety disorders and κ = .84 across SUDs.
	3. Cognitive impairment (Mini Mental Status Exam; Folstein Folstein & McHugh, 1975)	Determine Exclusion Criterion: Marked cognitive impairment (as indicated by scores < 24).
	4. Miscellaneous (demographic form, participant tracking and locator form)	The participant tracking and locating form is commonly used in SUD clinical trials to locate patients with SUD during the follow-up period (Hall et al., 2003), and was used to increase retention.
Pre-treatment assessment	Substance use (alcohol and drug use, specifying drug type) in past 30 days (Timeline Followback; TLFB, Sobell & Sobell, 1995)	Due to the heterogeneity of substances of dependence used in this real-world clinical sample, and the fact that over 60.3% of the sample met the SUD criteria for more than one substance of dependence, the TLFB was used because it allowed for assessment across a variety of substances of dependence. Interviewers were blind to condition. Inter-rater reliability was assessed on 22.9% of the TLFB interviews, with 100% agreement.
	Anxiety (Brief Symptom Inventory 12-item version [BSI-12]; Derogatis, 1993)	Cronbach’s α = .86 for the BSI in this sample
Week 1 (of CALM ARC or matched assessment in UC)	Treatment credibility and expectancy a. Modified version of Credibility and Expectancy Questionnaire (CEQ; Devilly & Borkovec, 2000). b. Ratings of expectation that treatment would improve anxiety and addiction, each on a scale from 0 (not at all) to 8 (certainly).
Post-treatment	1. Anxiety (BSI-12) 2. Substance use (TLFB) 3. Treatment Satisfaction (modified Treatment Satisfaction Questionnaire, TSQ; Cox, Fergus & Swinson, 1994)	On the TSQ, participants rated each CALM ARC component on likeability, usability, and how much they believed CALM ARC improved their anxiety and substance use. All ratings were on a 0 (not at all) to 7 (extremely) scale.
Follow-up	1. Anxiety (BSI-12) 2. Substance use (TLFB)

Figure 1.

Experimental Design and Flow of Participants through the Study (CONSORT Diagram).

Note: ^aOf the 11 ineligible, 6 due to no anxiety disorder diagnosis, 3 due to uncontrolled psychotic features, 1 due to active suicidality, and 1 due to psychotic features and uncontrolled bipolar disorder. Randomization was conducted on a rolling basis during 6-week standardized periods of randomization into one cohort or another to more quickly enroll sufficient numbers of participants into a group (NIDA, 2003)

^bConducted one week prior to initiation of CALM ARC sessions or the matched weekly assessments in UC; attrition in both conditions between baseline eligibility/randomization and pre-treatment was due to treatment dropout from Matrix for a variety of reasons

^cImmediately following 7 weeks of CALM ARC or 7 weeks of matched weekly assessments in UC

^dConducted 6 months from the pre-treatment assessment

Participants

Inclusion and exclusion criteria are presented in Table 1. Demographic characteristics of the final sample (N = 75) who completed at least a pre-treatment assessment are presented in Table 3.

Table 3.

Patient Characteristics

Variable	Mean (SD) or %
Gender (% male)	57.3
Age	35.89 (11.77)
Race/Ethnicity
White	72.0
Hispanic/Latino	10.7
Asian-American	9.3
Multiracial	6.7
Pacific Islander	1.3
Principal Anxiety Disorder Diagnosis
Generalized anxiety disorder	50.7
Social anxiety disorder	29.3
Panic disorder	10.7
Obsessive compulsive disorder	4.0
Posttraumatic stress disorder	2.7
Specific phobia	1.3
Agoraphobia	1.3
Number of anxiety disorder diagnoses	2.83 (1.46)
Substance Use Disorder Diagnosis
Met criteria for drug use disorder and alcohol use disorder	57.8
Met criteria for alcohol use disorder only	33.3
Met criteria for drug use disorder only	8.0
Primary substance of dependence
Alcohol	56.8
Stimulants	17.6
Opiates	9.5
Cannabis	9.5
Tranquilizers	6.8

Measures

Baseline Eligibility Screening

Patients who enrolled at the community SUD clinic were invited to complete an anxiety screening measure (i.e., OASIS; Campbell-Sills et al., 2009) at their intake appointment or during their treatment. Those with scores ≥ 8 were invited to a baseline eligibility assessment, details of which are reported in Table 1.

Outcome Measures

Outcome measures are described in Table 1.

Measures of Therapist Fidelity and Patient Treatment Adherence

Therapist fidelity ratings (comprised of adherence and competence) were conducted on 22.3% of CALM ARC sessions by an independent licensed clinical psychologist with expertise in CBT for anxiety disorders. Adherence was determined by calculating the percentage of assigned components delivered in a given session. Competence was rated on a 0 (not at all competent) to 6 (highly competent) scale for each session. Participant adherence to CALM ARC was evaluated via weekly clinician ratings of participant homework compliance on a scale from 0 (no homework completed) to 7 (all homework completed) and by examining the number of CALM ARC sessions completed.

Procedures

Assessment Procedures

Interview-based assessments were conducted by research staff at the clinic. After randomization, participants in both conditions were given personal login access to the CALM ARC Web program, where questionnaires were completed.¹

Treatment Conditions

CALM ARC is a brief (orientation + six sessions), group-based, computerized but therapist-directed CBT program for anxiety disorders to be delivered by SUD treatment counselors to patients with comorbid anxiety disorders and SUD in SUD specialty clinics. Similar to the original CALM study (Roy-Byrne et al., 2010) but adapted for a group setting, CALM ARC was outlined in a Web-based program. This allowed clinicians with limited training in delivering CBT for anxiety disorders to deliver the treatment with a high level of adherence. Clinicians used the interactive Web-based program as a guide to deliver the treatment material. Therapist-directed sessions had corresponding home practice sections that allowed participants to enter responses to homework assignments, watch videos, and access material. Components of CALM ARC and integrated features are described in Table 2. Usual Care (UC) at The Matrix Institute IOP is CBT-based with elements of 12-step facilitation and traditional relapse prevention for SUD. For a description of, and evidence for the Matrix Model, see Rawson, Shoptaw, Obert, McCann, Hasson, 1995.

Table 2.

Components and Features of CALM ARC

Component	Session(s)	Ways in which the component was integrated into SUD treatment
Psychoeducation about SUD/anxiety comorbidity	Orientation	Mutual maintenance model of anxiety/SUD comorbidity (Stewart & Conrod, 2008) woven throughout all sessions
Psychoeducation about components of anxiety	1	Examples highlighting how components of anxiety interact in situations involving drugs and alcohol were presented
Cognitive restructuring skills	1	Examples and videos focused on patients with comorbid anxiety disorders and SUD
Development of a hierarchy of stimuli for in vivo, imaginal, and interoceptive exposure	2	Examples and videos focused on patients with comorbid anxiety disorders and SUD
Exposure to distressing and avoided stimuli, including situations, bodily sensations, images, and memories, conducted in-session and assigned as homework	Assigned for homework in 2–6; In-session in 3–5	1. The rationale for exposure worked within the context of addiction recovery [e.g., distinguishing between adaptive exposure to feared, but objectively safe situations v. problematic exposure to external cues (i.e., “people, places, and things”) that may lead to substance use in early recovery]. 2. Coping plans developed in the IOP were revisited in the context of managing anxiety as an “internal trigger” during or after exposure. 3. When possible, recovery activities were woven into the exposure exercises to mutually augment one another (e.g., assigning a patient with social anxiety disorder to speak at a 12-step meeting for home practice)
Relapse prevention and developing practice plans	6	Goal setting and practice plans were woven into general goal setting and relapse prevention in SUD recovery

Note: CALM ARC groups used closed (i.e., not rolling) admission because CBT for anxiety disorders requires incremental skill building and practice that does not lend itself well to an open enrollment as utilized in typical IOP for SUD. CALM ARC sessions were 2 hours long (with a break), with an average of ~6 members per group. Exposures were conducted in the group room individually, as a group, or broken into smaller groups, depending on the group members’ fear hierarchies. Imaginal exposures conducted in the group room were primarily done via writing. After writing, participants were instructed to read what they wrote multiple times to themselves in session, and then out loud at home for repeated exposure.

Controlling for Therapy Time

Participants randomized to CALM ARC attended CALM ARC group sessions in lieu of the family education group during the seven weeks of CALM ARC, and then resumed family education.

Therapist Training and Competency Testing

Therapist training included didactics, role plays, and watching videos developed for CALM ARC demonstrating therapy skills with mock patients. After training, therapists (n = 2) completed a knowledge quiz and were tested on their competence to deliver CBT skills in standardized role plays. Therapists were permitted to deliver CALM ARC when they achieved at least an 80% on the knowledge quiz and when their average role play scores were rated as at least “adequate” (i.e., ≥ 4 out of 7) on the Yale Adherence and Competence Scale (YACS; Corvino, 2000) for at least four of the six role plays conducted for each anxiety disorder. The two therapists’ average knowledge quiz score was 96.06%. Their average YACS ratings across role plays ranged from 3.92 (SD = 0.12) to 4.91 (0.12), indicating therapists learned to adequately deliver CBT for anxiety disorders prior to delivering CALM ARC. Therapists received weekly supervision during the trial. The training protocol and competency testing materials are available by request.

Statistical Analysis

One-way ANOVAs, t-tests, and χ² tests were used to examine between-group differences on demographic and baseline clinical characteristics, and measures collected at one time point. Our analyses of anxiety and substance use outcomes utilized a series of multilevel models (using SPSS Mixed) with time as a level 1 predictor and condition (CALM ARC + UC, UC) as a level 2 predictor of intercept and linear slope, which were both allowed to vary randomly across participants. These analyses involved two outcome observations at a time (e.g., pre and post; post and follow-up) to examine within- and between-group effects from pre- to post-treatment, and separately to examine maintenance of treatment effects by comparing post-treatment and follow-up observations. We utilized all data that was collected (regardless of whether participants “completed” their treatment protocol), and included any participant who completed at least one assessment (i.e., pre-treatment). No significant correlations were observed between completion of either post-treatment or follow-up assessments and key pre-treatment demographic or clinical severity measures, suggesting that the assumption of a missing at random mechanism is likely to be reasonable for these data. To avoid potential bias in the assessment of treatment effects due to missing data, we imputed values for the outcome variables across the entire span of the study using SPSS Multiple Imputation.² We report analyses of the imputed data, and footnote those instances where findings from multilevel models using the original dataset – which address missingness though Full Information Maximum Likelihood (FIML) estimation limited to only the small set of variables included in a given analysis -- differ. Assumptions of all strategies were met.

Results

Sample Characteristics, Treatment Credibility, and Treatment Expectancy

See Table 3 for diagnostic information. No between-group differences were observed on the number of days between Matrix IOP intake and pre-treatment assessment in the study, baseline symptom severity measures (Table 4), demographics, clinical characteristics, perceived treatment credibility, or treatment expectancies (ps ≥ .06).

Table 4.

Descriptive statistics and effect sizes by condition at each assessment period

Variable	Pre		Post		Follow-up
	CALM ARC+UC	UC	CALM ARC+UC	UC	CALM ARC+UC	UC	Post Between-groups Cohen’s d	FU Between-groups Cohen’s d
BSI-12	14.98 (7.75)	14.15 (7.74)	10.16 (6.35)	12.81 (7.00)	10.16 (8.56)	15.56 (11.21)	0.40	0.56
TLFB drinking days	5.88 (8.64)	3.23 (5.10)	3.26 (4.90)	6.00 (8.02)	4.81 (6.46)	6.72 (6.81)	0.43	0.29
TLFB drug use days	2.76 (5.75)	1.82 (3.93)	2.83 (4.74)	5.11 (8.57)	5.00 (7.34)	8.87 (9.78)	0.35	0.46
TSQ: liking			4.67 (1.27)
TSQ: usability			4.14 (1.18)
TSQ: helped anxiety			5.11 (1.27)
TSQ: helped addiction			4.81 (2.36)
TSQ: improved QOL			4.82 (1.33)

Note: Values are based on the multiply imputed sample. Parallel analyses were run without multiple imputation and findings were generally identical.

BSI scores < 6 indicate mild to no symptoms. 15.3% and 29.4% of UC participants and 34.3% and 53.8% of CALM ARC participants met this cutoff at post and follow-up, respectively.

Treatment Adherence: Therapist Fidelity and Patient Adherence

Therapists adhered to the CALM ARC protocol 99.3% of the time. The mean competence rating across all components and sessions rated on a scale from 0 (not at all competent) to 6 (highly competent) was 5.23 (SD = 0.35), indicating a high level of competence. Participants in CALM ARC completed an average of 4.92 (SD = 1.81) sessions. 69.2% of participants completed at least five sessions, but only 15.4% of participants completed all seven sessions. The average participant homework completion rating was 4.00 (SD = 2.08), corresponding to “about half” of the assigned homework. Thus, the average participant in CALM ARC could be characterized as moderately engaged.

Clinical Outcomes

Table 4 presents descriptive data and between-group effect sizes (Cohen’s d) across assessment periods for outcome data.

Anxiety symptom outcomes (BSI-12)

BSI scores significantly declined over time (pre- to post-treatment) in UC + CALM ARC [b = −4.82, t (211.42) = −4.49, p < .001] but not in UC (p = .27). There was a significant time (pre, post) x condition effect, b = −3.61, t (323.66) = −2.32, p < .05, with UC + CALM ARC outperforming UC on pre- to post-treatment BSI improvement. There were no significant effects of time from post-treatment to follow-up within either condition, and no time (post-treatment, follow-up) x condition effects (all ps > .19), indicating that effects were maintained from post-treatment to follow-up.

Substance use outcomes (Timeline Followback)

When examining pre- to post-treatment change in drinking days, within-group tests revealed a significant effect of time in CALM ARC + UC, b = −2.99, t (677.44) = −2.19, p < .05. In contrast, there was no statistically significant effect of time in UC (p = .07).³ There was a significant time (pre, post) x condition effect, b = −5.74, t (907.32) = −2.83, p < .01, with CALM ARC + UC participants showing greater pre- to post-treatment reductions in drinking days compared to UC. Next, we conducted a parallel set of analyses examining drug use days. Within-group tests showed no significant effect of time within UC + CALM ARC (p = .99), but a significant increase in days of drug use from pre- to post-treatment in UC, b = 3.18, t (1527.01) = 2.53, p < .05. There was a time (pre, post) x condition effect that approached significance, b = −3.19, t (1641.12) = −1.90, p = .057,⁴ with UC + CALM ARC outperforming UC. There were no within- or between-group effects from post-treatment to follow-up on drinking days or drug use days (ps > .13), with one exception: there was a significant effect of time on drug use days within UC, b = 3.90, t (161.41) = 2.34, p < .05, indicating greater drug use in UC from post-treatment to follow-up.

Discussion

Consistent with the hypothesis, adding CALM ARC to a SUD Intensive Outpatient Program was superior to SUD treatment alone in reducing anxiety symptoms and substance use. These findings provide further evidence that SUD treatment alone is insufficient for comorbid anxiety disorders and SUD (see Wolitzky-Taylor et al., 2011). Importantly, anxiety was effectively treated during SUD treatment. This finding challenges the conventional wisdom that the SUD should be treated first before anxiety can be effectively treated (Nace, 1988), a practice that typically results in untreated mental health problems (Havassy, Alvidrez & Mericle, 2009).

In line with tension-reduction (e.g., Conger, 1956) and mutual maintenance (Stewart & Conrod, 2008) models of comorbid anxiety disorders and SUD, treating anxiety symptoms that were likely to be maintaining SUD improved substance use outcomes as well. These findings add to the small body of literature supporting the use of integrated models of treatment for anxiety disorder and SUD comorbidity (e.g., Najavits et al., 2006; Kushner et al., 2006). CALM ARC’s effects are particularly notable given its brevity, moderate level of patient engagement, and its comparison to an evidence-based treatment for SUD (Rawson et al., 1995). Psychiatric diagnostic assessments in SUD clinics should be conducted in order to identify patients for whom typical SUD treatment may not exert its optimal effect without the addition of CBT to treat underlying anxiety.

In contrast to the high level of fidelity to the CALM ARC protocol by SUD providers, patients were only moderately adherent. Effects may have been even larger with greater treatment engagement. Low treatment engagement and dropout are common clinical challenges in SUD populations, particularly among those with comorbid psychiatric disorders (Krawczyk et al., 2017), and are critical to address given the link between treatment engagement and clinical outcomes (Simpson, Joe, Rowen-Szal & Greener, 1995). Future research should evaluate adjunctive efforts to enhance participation in CALM ARC (e.g., Miller & Rollnick, 2002). Notably, treatment adherence was not a problem specific to the CALM ARC condition. Attrition was apparent in both conditions, even between the baseline eligibility assessment and the beginning of the treatment phase (see Figure 1). Thus, outpatient SUD treatment-seekers represent a population in need of interventions to increase treatment adherence and retention.

The primary limitation of the study was the relatively small sample size, which may have precluded us from detecting small effects. Future studies with larger samples are needed to evaluate whether findings replicate, and to examine anxiety disorder- and substance-specific outcomes. Also, although the hybrid effectiveness/efficacy design is a strength of the investigation by addressing both internal and external validity, this approach is a balancing act that naturally leaves some questions unanswered. For example, future research should examine whether CALM ARC exerts the same effects without weekly expert supervision or in less “research friendly” clinics. Uncovering mediators and moderators of CALM ARC should also be the focus of future research that can provide information to refine and augment outcomes.

These preliminary findings suggest that CALM ARC is a feasible, acceptable, and effective adjunctive intervention for comorbid anxiety disorders in SUD treatment seekers. Its transdiagnostic approach makes it translatable to real-world settings, in which treating specific anxiety disorder and substance type combinations may be impractical. This study adds to the growing body of literature that aims to overcome the disconnect between typical clinical practice and the emerging theoretical (Stewart & Conrod, 2008) treatment (Kushner et al., 2006) and policy (Barry & Huskamp, 2011) models that argue for the advantages of an integrated model of mental health and addiction care.

Public Health Significance Statements.

1. Patients with comorbid anxiety and substance use disorders who received an adjunctive cognitive behavioral intervention for anxiety during substance use disorder treatment showed greater improvement in anxiety and substance use outcomes than those who with this comorbidity who received only substance use disorder treatment.

2. This intervention was feasible for delivery in real-world substance use disorder clinics. Its delivery platform, brevity, and adaptation to meet the needs of this population increases the likelihood of adoption in community practices.

3. Treating substance use disorder patients for comorbid anxiety disorders using an integrated approach has a significant impact on clinical outcomes.